CS Paper Radar

Daily listing

2026-10-29

926 papers first submitted this day (UTC). arXiv announces a day's submissions the next evening, so the newest day fills up over ~24 h.

Score = weighted mean ofCitations 20%Upvotes 10%Code stars 10%Venue 15%Code/data 10%Novelty 15%Momentum 15%Substance 5%how it works

AI & Machine Learning 285

Comparing Human Role-Players and LLM-Simulated Clients in Online Counselling Training: An Analysis of Counselling Patterns

cs.AI cs.CL ·2026-10-29 ·Eric Rudolph, Philipp Steigerwald, Jens Albrecht
publishedhighly cited
Abstract · Published at EDM 2025 (18th International Conference on Educational Data Mining), short paper. 7 pages

This study investigates the capabilities of Large Language Models to simulate counselling clients in educational role-plays in comparison to human role-players. Initially, we recorded role-playing sessions, where novice counsellors interacted with human peers acting as clients, followed by role-plays between humans and clients simulated by Mistrals Mixtral 8x7b using 4-bit quantization. These interactions were analysed with a counselling communication pattern system at sentence level. We investigated two key questions: (1) to what extent LLM-generated responses replicate authentic conversational dynamics and (2) whether counsellors' communication behaviour differs when interacting with human versus LLM-simulated clients. The findings highlight both similarities and differences in the application of counselling patterns across scenarios, showing the potential of LLM-based role-playing exercises to enhance counselling competencies and to identify areas for further refinement in virtual client simulations.

0.67

LinQuant: Functional Invariances for Low-Precision Linear Recurrences

cs.LG ·2026-10-29 ·Neehal Tumma, Philipp Nazari, Rohan Cherukuri, Daniela Rus, T. Konstantin Rusch
code/data
Abstract · 35 pages, 16 figures. Code: https://github.com/ntumm120/LinQuant

Linear recurrences are emerging as practical alternatives to softmax attention in large language models, appearing in both standalone and hybrid architectures. However, despite their rapid emergence, low-precision inference remains much less developed for these models than for softmax attention. During prefill, chunkwise-parallel algorithms cast much of the recurrence as dense matrix multiplications, suggesting a natural path to low-precision execution. Simply quantizing these operations in their native form, however, can be fragile. In this work, we identify functional invariances that preserve the full-precision computation while making the operands more amenable to quantization. Building on this observation, we introduce LinQuant, a framework for low-precision recurrent prefill based on seven such invariances, including rotations, diagonal rescalings, and alternative placements of recurrent decay and write factors. We derive when these transformations remain valid across common recurrence families spanning linear, gated, and delta-rule updates, and show that they provide complementary degrees of freedom for improving low-precision accuracy. Finally, we translate this framework to production by augmenting SSDi8, an existing low-precision prefill pipeline for Mamba-2, with the applicable invariances. LinQuant improves perplexity at matched latency on Mamba-2 1.3B, 2.7B, and 8B and Nemotron-H 8B, demonstrating the practical utility of invariance-based quantization for recurrent models.

0.56

The Precision Wall: An Optimizer-Dependent Boundary for Optimization under Power-Law Measurement Spectra

cs.LG ·2026-10-29 ·Lngel Kyo
Abstract · 19 pages, 3 figures, 4 tables

In low-rank matrix sensing of a rank-$D$ factor $U^\ast \in \mathbb{R}^{V \times D}$, replacing isotropic Gaussian designs with power-law measurement covariance $Σ= \mathrm{diag}(j^{-s})_{j=1}^V$ leaves the Burer--Monteiro landscape free of spurious local minima, yet division-normalised first-order optimizers stall above a critical exponent $s^\ast$. We trace this failure to a numerical precision wall. At the ground truth, the smallest non-gauge expected Hessian eigenvalue scales at leading order as $V^{-(s+1)}D^{-s}$. When this curvature drops below the effective denominator floor $ε_O$ of optimizer $O$, normalisation noise overwhelms rare-direction gradients, yielding the leading-order boundary $s^\ast_O(V, D) = [\log(1/ε_O) - \log V]/\log(VD) + δ_O$, where $δ_O$ is a metric-dependent detection lag. With $δ_{\mathrm{Adam}}$ calibrated on four configurations and frozen, this formula predicts Adam's onset within $0.45$ across all 16 cells of a held-out grid ($V \in \{64,128,256,512\} \times D \in \{4,8,16,32\}$), separates fixed-$VD$ pairs, and distinguishes denominator-precision walls from conditioning and sign-quantisation walls. The approximation fails at $D = 2$, reflects a finite-$D$ spectral crossover, and saturates below $\varepsilon_0 = 10^{-8}$ at a dynamical floor.

0.56

Observability-Aligned Evidential Masked Pretraining for Reliable VRU Motion Prediction

cs.LG ·2026-10-29 ·Shang Gao, Franz Götz-Hahn, Konrad Doll, Bernhard Sick
benchmark
Abstract

Vulnerable road user (VRU) prediction must operate under partial observability and distribution shift, so reliable uncertainty matters alongside accuracy. Masked pretraining is attractive in this setting because it naturally exposes the model to missing content during self-supervision, but existing time-series masked autoencoding methods are typically deterministic and do not constrain how confidence should change with observability. We introduce eCoMAE, an evidential masked pretraining method for multivariate time series that explicitly aligns predictive evidence with what is observed. eCoMAE combines block-wise time-step masking with a Normal-Inverse-Gamma decoder and three observability-aware mechanisms: uncertainty-weighted mean consistency on positions visible in both a masked view and a reference pass, evidence suppression on masked positions, and representation consistency across stochastic masks. On IMPTC benchmark for VRU intention prediction under leave-one-environment-out weather shift, eCoMAE improves the strongest baseline from 0.7971 to 0.8068 in OOD accuracy and from 0.0718 to 0.0648 in OOD AURC. Additional evaluation on pedestrian trajectory forecasting across IMPTC, Waymo, nuScenes, and inD further yields the best fine-tuning MSE on all four datasets. These results show that observability-aligned evidential pretraining improves reliability under shift and remains effective beyond the primary benchmark.

0.55

Surprise Alone Learns Lies: Evidence-Gated Continual Learning for LLMs from Streaming Claims

cs.LG cs.CLcs.CR ·2026-10-29 ·Abdelkader Metales, Miriam Grabenheinrich
code/data
Abstract · 21 pages, 5 figures. Code: https://github.com/Metalesaek/sura-metales

Continual learning methods for large language models increasingly decide what to learn by surprise, keeping the inputs the model finds least predictable. This rule is unsafe when the stream contains lies. A planted falsehood about something the model knows has low probability under the model's belief, so it is highly surprising by construction, and a surprise-driven learner absorbs nearly every such lie, at a rate similar to random selection with the same budget. We study the problem with LoRA-adapted students (Qwen2.5-1.5B, Qwen2.5-3B and Llama-3.2-3B) that learn from a stream of true and false claims about invented people and real CounterFact facts. We report four findings. (1) A verifier key, a judge that must approve each claim, removes lies to the extent that its verifier is accurate; real LLM judges are cautious and reject unsourced truths and even source-backed updates. (2) Calibrated regression tests on unrelated knowledge slow learning down without filtering out lies. (3) The original model's own prior flags claims that contradict what it knows, without a judge. It separates lies about known facts from true new facts (AUC 0.78-0.83 across three models) but not from legitimate updates of the same kind of fact (AUC 0.53). It saves 43% of judge calls, but an attacker with access to the model can choose lies that pass it. (4) If trusted sources exist that an attacker cannot plant, routing contradicted claims to evidence recovers legitimate updates and the combined system has the lowest poison rate among learned claims; when an attacker plants matching sources for 30% of the lies, the system ends up worse than the judge alone. The acceptance signals answer different questions, and none of them decides on its own whether a claim is true.

0.53

DynaRev: Benchmarking Web Agents under Dynamic Goal Revision

cs.AI ·2026-10-29 ·Shiqi Zhang, Guochen Zhou, Zesheng Wei, Hai Zhao, Zhisong Zhang
benchmarkcode/data
Abstract

Web agents are typically evaluated with goals fully specified before execution and fixed thereafter, whereas real users often revise their requirements during interaction. To study such interactions, we introduce DynaRev, a benchmark of 181 tasks to evaluate web agents under dynamic goal revision. DynaRev uses semantic milestones to deliver revisions at task-defined stages, covers additive revisions, unconditional overrides, and runtime-conditioned overrides, and evaluates correctness, responsiveness, and efficiency through trajectory-grounded evaluation. Five-model evaluations show that agents often satisfy only part of the task requirements, with performance varying across task sources, revision types, and models. Further analyses reveal varied milestone timing, mixed effects of dynamic delivery, and contrasting revision-handling outcomes, with gains from instruction rewriting while preserving browser state and losses from browser resets. DynaRev provides a testbed for studying how web agents adapt to evolving user goals. The code and data are available at https://github.com/zsq259/DynaRev.

0.53

Training language models in simulation for psychiatric intake

cs.AI cs.CL ·2026-10-29 ·Samuel Schmidgall, Tim Strother, Alex Bijamov, Min Woo Sun, Anil Palepu +51 more
Abstract · 76 pages, 8 figures, 15 tables (27 pages main text, 7 main figures, 2 main tables)

In psychiatric practice, the clinical intake interview remains a primary diagnostic instrument, requiring the clinician to establish rapport while disentangling complex, overlapping presentations. Here, we show that multi-turn reinforcement learning (RL) in simulation environments improves language model performance on structured text-based psychiatric intake, spanning history-taking, differential diagnosis, and initial management planning. In a randomized, double-blind crossover study ($N = 60$ matched scenario pairs, 120 encounters) comparing the RL-trained agent against its identically prompted base model, 32 board-certified psychiatrists role-played standardized patients while a separate panel of 15 independent, board-certified psychiatrists scored the de-identified transcripts. The trained agent outperformed the base model in 76.7% of matched scenario pairs ($p < 0.0001$), with gains on an adapted clinical-examination rubric ($d_z = 0.77$, $p < 0.0001$), patient-centered communication (82.0% vs. 65.2%, $p < 0.001$), and differential diagnosis appropriateness (68.3% vs. 33.3%, $p < 0.001$). Across 127 psychiatrist-validated simulated intake scenarios, the trained agent nearly doubled its base model's clinical rubric score (0.706 vs. 0.379; composite reward 0.564 vs. 0.294, $p < 0.0001$). To systematically probe conversational safety, we established two Delphi-consensus taxonomies, of clinical safety risks and relational competencies, from which 256 challenging adversarial presentations were derived; we find that training in simulation environments more than halved under-triage of acute safety risks (12.9% vs. 30.5%, $p < 0.0001$). Simulation environments thus provide a controlled testbed for training and stress-testing psychiatric conversational agents, with gains that transferred to psychiatrist-enacted encounters, though residual gaps require safeguards and clinician oversight.

0.50

Don't Just Impute, Detect: Task-Aware Diffusion Pipelines for Edge Anomaly Detection

cs.LG ·2026-10-29 ·Jeongsu Park, Taehyeon Yun, Geonwoo Kim, Yuji Lim, Moohong Min
publishedcode/dataunusual
Abstract · 11 pages, 6 figures. Accepted at the 33rd IEEE International Conference on High Performance Computing, Data, and Analytics (HiPC 2026)

Industrial Internet of Things (IIoT) edge pipelines routinely lose contiguous blocks of multi-sensor data to network dropouts and hardware faults. The standard remedy is a disjointed "impute-then-detect" pipeline whose imputer is trained only for Mean Squared Error (MSE). Such an imputer fills a missing block with the most probable normal pattern: a partially observed anomaly is smoothed back toward normality, and boundary artifacts on normal data raise false alarms. We call this imputation bias and propose the Task-Aware Diffusion Refiner (TADR), a conditional diffusion refiner trained against the loss landscape of a frozen downstream detector. Because back-propagating through the sampling trajectory is intractable, the task-aware terms act on the one-step clean-window estimate that the noise prediction already yields, so training costs no more than plain diffusion training. Under strict Point-to-Point evaluation over five seeds with paired significance tests, TADR raises F1 by $0.338$ over the unrefined detector on SMD ($p = 0.0015$), and the anomaly enhancement objective alone contributes $0.463$ in ablation. We also characterise when the method helps: the gain tracks the continuous, mutually redundant channels available and the anomaly evidence the outage removes, and it is absent on telemetry datasets dominated by one-hot command channels and on PSM. At ten sampling steps TADR runs in $41.4$ ms per window on a single CPU core, sustaining about $24$ concurrent 1 Hz streams from a $4.03$ MB model. Code and per-seed results are available at https://github.com/int-park422/TADR.

0.49

FB-MOAC: A Reinforcement Learning Algorithm for Forward-Backward Markov Decision Processes

cs.LG cs.AI ·2026-10-29 ·Mohsen Amidzadeh, Mario Di Francesco
published
Abstract

Reinforcement learning (RL) algorithms are effective in solving problems that can be modeled as Markov decision processes (MDPs). These algorithms primarily target forward MDPs whose dynamics evolve over time from an initial state. However, several important problems in different scenarios including stochastic control and network systems exhibit both a forward and a backward dynamics. As a consequence, they cannot be expressed as a standard MDP, thereby calling for a novel theory for RL in this context. Accordingly, this work introduces the concept of Forward-Backward Markov Decision Processes (FB-MDPs) for multi-objective problems, develops a novel theoretical framework to characterize their optimal solutions, and proposes a general forward-backward step-wise template that allows to adapt RL algorithms for FB-MDP problems. A Forward Backward Multi Objective Actor Critic (FB-MOAC) algorithm is introduced accordingly to obtain optimal policies with guaranteed convergence and a competitive rate with respect to standard approaches in RL. FB-MOAC is evaluated on diverse use cases in the context of mathematical finance and mobile resource management. The obtained results show that FB-MOAC outperforms the state of the art across different metrics, highlighting its ability to learn and maximize rewards.

0.49

HERA: Assessing an Autonomous Agent for Quantum Device Measurement

cs.AI cond-mat.mes-hall ·2026-10-29 ·Zachary W. Gomez, Nathan Bittner, Jordan T. McCourt, Qingrui Cao, Sandesh S. Kalantre +5 more
code/data
Abstract · 16 pages, 5 figures, plus Supplementary Information (7 figures, 6 tables); data and code at https://purl.stanford.edu/tv195gf3182 and https:…

Large language model agents have begun to carry out laboratory measurements, but their scientific conclusions are useful only if they hold up under scrutiny. Demonstrating this requires controlled studies because models are stochastic and the context a model is given steers its conclusions. Here we introduce HERA, an autonomous agent designed to carry out cryogenic transport measurements of quantum devices. HERA draws on a knowledge base curated by lab members that holds measurement protocols, software tools, and literature references. We assessed HERA with a blinded trial against an expert on the stacking order of two multilayer graphene devices, an ablation of its knowledge base, and hundreds of offline replays in which models designed the trial's central measurement without running it. HERA's conclusions matched the expert's sealed conclusions on both tested devices. The ablation and the replay studies measured the impact of the knowledge base across different models, showing that with it every model designed better measurements and reasoned more soundly. To test how readily HERA extends to a new problem, we tasked it with finding and characterizing superconductivity in a different material system. This required a few hours of expert time to develop a new experimental protocol, after which HERA did both autonomously. These results show that, with a knowledge base designed for the task, language model agents are capable of sound experimental science.

0.49

Diffusion models for the $O(3)$ nonlinear sigma model

cs.LG cond-mat.stat-mechhep-lat ·2026-10-29 ·Daniel Schubring
Abstract · 36 pages, 9 figures

A score-based diffusion model is trained on the two-dimensional classical Heisenberg model, the lattice representation of the $O(3)$ nonlinear sigma model. Neural networks are trained on three different coarse-grainings of the same training set to implement a multiscale sampling scheme that modestly improves both the accuracy and computational cost of direct sampling on the finest lattice. Stein (or Schwinger-Dyson) diagnostics for the two-point correlation function are introduced to assess the quality of the learned score without relying on reverse diffusion sampling. These diagnostics are also incorporated into an additional training loss. We discuss several technical considerations relevant to implementing diffusion models on $S^2$. These include a consideration of lifting the network input to $S^3$, as in the $CP^1$ representation, and developing a boundary-layer approximation to the short-time spherical heat kernel that extends its usefulness to larger diffusion times.

0.48

Exploring the Potential of Symbolic Regression in Precision Livestock Farming

cs.LG cs.AI ·2026-10-29 ·Elisabeth Mayrhuber, Maciej Oczak, Michael Iwersen, David Brunner, Viktoria Dorfer +1 more
unusual
Abstract

The ability to predict farrowing time in pigs is crucial for optimal farm management and ensuring piglet survival. This study investigates the potential of explainable artificial intelligence for precision livestock farming, focusing on acceleration sensor data from sows around the time of farrowing. Accurately predicting farrowing allows farmers to balance both, piglet safety by avoiding piglet crushing events and the mother sow's welfare, by avoiding early crating. Traditional machine learning models often lack interpretability (``closed-box'' problem). Explainable artificial intelligence offers a solution by revealing the decision-making processes within the models. This work is split into two stages. First the detection of nest building behaviour and second the prediction of the remaining time to farrowing. The nest building detection was possible for 166 out of 180 sows and successful within 48 hours before farrowing for 149 sows. The nest building detection is purely based on the change of a sow's activity level and uses no machine learning techniques. Further, this study utilizes symbolic regression, an open-box modelling technique that generates interpretable mathematical formulas, to predict the remaining time until the onset of farrowing. The results are compared with closed-box models, like support vector regression and artificial neural networks, for both explainability and performance. Notably, symbolic regression achieved a mean absolute error of 9.28 hours for the time-to-farrowing prediction, outperforming the closed-box models in terms of performance and interpretability. This research demonstrates that explainable artificial intelligence, specifically symbolic regression, can offer both interpretable results and competitive performance in precision livestock farming.

0.48

Nonlinear Maps Between Irreducible Representations in Graph Neural Networks

cs.LG math.RT ·2026-10-29 ·Mani Shayestehfar
benchmark
Abstract · 18 pages; comments are welcome

Equivariant linear maps cannot mix non-isomorphic irreducible representations, whereas nonlinear activations can. We develop a degree-resolved version of this nonlinear interaction picture and use the resulting block inventory to construct Specht Triple, a residual on unordered graph triples. On a synthetic structural benchmark with a frozen GIN backbone, Specht Triple reduces aggregate standardised test MSE by 11.05% relative to a parameter-matched Projector Control using the same irreducible projectors and complete linear equivariant space. The results support irreducible decomposition as an inductive bias for nonlinear higher-order equivariant processing.

0.48

A New AFIB Detection Method with Deep Neural Networks

cs.LG ·2026-10-29 ·Sara Gjorgjieva, Ana Angjelevska, Dimitar Trajanov, Marjan Gusev
benchmarkhot direction
Abstract · 10 pages, 7 tables, 3 formulas

Early detection of atrial fibrillation (AFIB), characterized by irregular heart rhythms, can potentially save thousands of lives annually [20]. Numerous studies have attempted to address this challenge using various algorithms and methodologies. This paper introduces a novel and robust method that enhances the precision of AFIB detection. Neural Networks (NN) are widely recognized for their ability to identify complex electrocardiogram (ECG) patterns, facilitating the accurate detection of irregular heart rhythms. This study explores multiple conventional machine learning algorithms alongside neural network approaches to develop a robust AFIB detection system. To evaluate the effective- ness and robustness of the proposed method, the MIT BIH arrhythmia database (MITDB) and the Long-term Atrial Fibrillation Database (LTAFDB), both available on the Physionet site. Our approach presents a novel feature engineering technique, employing specific features such as a time series of the differences in successive heart rates (dBPM) rather than traditional inter-beat intervals. The model analyzes inputs labeled into five distinct classes, using only clean intervals identified through a new labeling method for training while treating others as outliers. This strategy ensures a more accurate and reliable AFIB detection process. The best-performing model, a Deep Neural Network (DNN), achieved an F1 score of 95.18% for binary atrial fibrillation classification on a benchmark dataset distinct from the training set. Models trained using our approach have demonstrated superior performance than previously published models.

0.48

A Novel Approach to Evaluating Evaluation Metrics for Multi-Output Structured Prediction

cs.LG cs.CLcs.CV ·2026-10-29 ·Akshay Vyas, Angelo Pimienta, Nicholas Ruozzi
top venuecode/data
Abstract · Extended version of the AAAI 2026 paper, including the technical appendix. 18 pages, 8 figures. Code: https://github.com/akshay25vyas/invari…

In multi-output structured prediction tasks, while only one ground truth label may be provided in the training data, multiple equally valid outputs may be possible, making reliable evaluation a persistent challenge. We postulate that human evaluators implicitly use task-specific invariants, e.g., object boundaries in colorized images or named entities in translations, to judge if an output is acceptable. Under this assumption, we introduce a notion of approximate task-specific invariants and use them as diagnostic tools to evaluate a variety of existing metrics for vision and language tasks. We use these task invariants as part of a framework to systematically test metric reliability by encouraging domain-relevant invariants in model outputs via an augmented loss function. In our experiments, we observe that enforcing invariants with an augmented loss yields substantial improvements in popular distributional metrics while more traditional metrics change only marginally. Through this invariants-driven evaluation, we expose where standard metrics fail to detect meaningful differences, and we highlight the conditions under which distributional metrics succeed or still fall short.

0.46

LLM-Assisted Optimization for Simulation-Based Redesign

cs.AI cs.CEmath.OC ·2026-10-29 ·Chan-Eui Song, Jeremy J. Michalek, Christopher McComb
code/data
Abstract · 17 pages, 12 figures, 11 tables. Submitted to the ASME Journal of Mechanical Design. Code: https://github.com/Chani-song/WAFER, data: https:…

Engineering design frequently involves redesigning a prior solution to adapt to changes in context, such as material prices or system requirements. Redesign often leverages simulators too complex and/or computationally intensive for traditional numerical solvers, which lack engineering knowledge and context, to ensure global solution quality. Large language models (LLMs) can interpret semantic information, including engineering knowledge about the system and changes in context, to help guide search, and recent studies have shown that incorporating this knowledge into optimization can improve search performance. However, how best to combine LLMs with numerical solvers in redesign remains unclear. We develop and test an integrated LLM-optimization approach for redesign and compare three methods for leveraging LLMs (warm start, variable fixing and ReAct) and four black-box solvers (Bayesian optimization, genetic algorithms, mesh adaptive direct search and random search) on an electric vehicle battery pack cost minimization case study across four types of redesign scenarios. We find that the integrated approach can match feasible design discovery of the numerical-solver baselines with 80% fewer simulator evaluations while tripling the fraction of runs that find near-optimal solutions. We find that the best LLM-solver integration approach and the way the prior solution is leveraged depend on the structure of the design requirement change, and that the benefit of increasing LLM model size grows with the extent to which the LLM controls the search. These results show clear potential for LLMs in redesign tasks for guiding or replacing numerical solvers.

0.44

Low-Bit GDN Recurrent-State Quantization: Global Gradient-aware Allocation and Pure-Write Sliding Windows

cs.LG ·2026-10-29 ·Qinghang Huang, Xuchu Huang, Xunzhao Yin
Abstract · 24 pages, 13 figures

Linear attention stores historical information in a recurrent state of fixed size, providing a compact memory representation for long-context inference. Reducing state precision further lowers per-request storage requirements, but quantized writeback errors enter subsequent recurrences, degrade prediction quality on long-reasoning tasks, and can be accompanied by repetitive generation and excessively long outputs. We propose Global Gradient-aware Allocation (GGA) and the Pure-Write Sliding Window (PW) to address precision allocation and recent-information maintenance. GGA combines candidate quantization residuals with gradient sensitivity to future prediction loss to allocate head bit widths under a model-wide mixed-precision budget. PW independently maintains recent pure writes as FP8 vector pairs, propagates them through the full GDN state transition, and merges them into the low-bit historical state only after they leave the window. Across seven evaluations on Qwen3.5-4B/9B, GGA allocation alone improves over STEPQuant@4 by 19.48/29.84 percentage points. Adding PW-S8 yields seven-task macro-averages of 69.83%/78.63%, improving over STEPQuant@4 by 38.39/41.21 percentage points, respectively. Both precision reallocation and window maintenance improve task scores while mitigating excessively long outputs: at @6, GGA reduces mean generation length, averaged equally across tasks, by 10.45%/14.65% relative to STEPQuant; adding PW-S8 to GGA@4 further reduces mean generation length by 32.32%/36.77%. Comparisons at similar storage budgets further show that using additional capacity to maintain recent writes yields higher task scores and shorter mean outputs than solely increasing the bit width of the accumulated state.

0.43

Communication-Efficient Private Federated Reinforcement Learning with Post-Quantum Authentication for Vehicular Networks

cs.LG cs.CR ·2026-10-29 ·Hojat Allah Salehi, Devrim Aras, Md Jueal Mia, M. Hadi Amini, Imtiaz Karim +2 more
code/data
Abstract · 46 pages, 12 figures, 12 tables. Code: https://github.com/solidlabnetwork/cpq-vehicular-frl

Federated reinforcement learning supports cooperative vehicular routing without centralizing trajectory data, but model updates may reveal private information and incur communication overhead over intermittent vehicle-to-infrastructure links, particularly when post-quantum authentication is required. We study correlated private quantization, using complementary shared randomness between paired vehicles to reduce aggregation mean squared error while preserving the marginal parameter-level local differential privacy and one-bit communication cost of independent quantization. A tunable extension mixes shared and independent randomness to bound privacy leakage given paired outputs and shared-randomness side information. For this extension, Rényi differential privacy accounting establishes vector-level local differential privacy for one model upload even when pairings are revealed. Under matched vector-level privacy and identical local and global parameters, the extension yields 16.5% lower analytical aggregation mean squared error than independent quantization. We evaluate cooperative routing by coupling traffic simulation with packet-level simulation of post-quantum-authenticated vehicle-to-infrastructure communication. Compared to full-precision updates, one-bit uploads reduce authenticated uplink size by 92.4% and increase mean verified updates aggregated per round from 19.6 to 26.5. Under mobility-constrained communication at a parameter-level privacy budget of 1, correlated private quantization reduces mean travel time by 14.3% relative to independent quantization at the same budget and communication cost. At a parameter-level privacy budget of 0.3, a vehicle-origin inference attack trained on independent simulations reaches 30% accuracy with correlated private quantization and 22% with independent quantization after 100 rounds, versus 91% without privacy and 10% for random guessing.

0.43

Grounding Predictive Reliability through Masked Observability

cs.LG ·2026-10-29 ·Shang Gao, Bernhard Sick, Franz Götz-Hahn
benchmark
Abstract

An uncertainty estimate is meaningful only if its semantics can be tied to observable predictive behavior. Normal-Inverse-Gamma (NIG) evidential regression provides a structured decomposition of predictive variability into data-scale and predictive-mean components, but marginal likelihood does not identify their allocation. We address this semantic ambiguity in masked representation learning. MORSE (Masked Observability for Reliable Structured Encoding) predicts the same hidden target under controlled changes in visible context and uses normalized cross-context variation to ground predictive-mean uncertainty, while a Gaussian objective learns the conditional mean and data scale. The normalized variation plays the role of the NIG factor $1/ν$, which the marginal leaves unidentified; this yields a policy-relative observability variance that measures dependence on visible evidence, distinct from variation induced by resampling the training set. On a Gaussian-process benchmark with a contextual reference built from analytic Bayes conditional means, MORSE has the lowest log-variance error among compared evidential, ensemble, and dropout estimators in all nine tested settings, while the estimator closest to training-set variation perturbs the model rather than its context in every setting, so the two references separate the estimators empirically. MORSE amortizes observability estimation into one forward pass per observation state. On a heteroscedastic pulse benchmark, its predictor's cross-context variance approaches a large-sample reference as training data increase. MORSE also improves low-label transfer on PTB-XL and in-distribution trajectory prediction on IMPTC.

0.43

Monte Carlo Random Walk Kernel

cs.LG ·2026-10-29 ·Arslan Iamaletdinov, Elfat Sabitov, Konstantin Sorokin
code/data
Abstract

We introduce the Monte Carlo Random Walk Kernel (MCRWK), an estimator whose per-pair cost is independent of graph size and derived directly from the random walk kernel (RWK) expansion. By normalizing the kernel coefficients, MCRWK samples a shared walk length and estimates its contribution using random walks on the input graphs. This gives a simple way to approximate RWK without constructing the product graph. For a fixed expected walk length and efficient neighbor sampling, each pairwise estimate takes $O(m)$ time for $m$ Monte Carlo samples, allowing the sample budget to control the trade-off between runtime and accuracy. We evaluate MCRWK against Graph Voyagers and classical RWK methods on synthetic and real graph datasets. In the reported experiments on large synthetic graphs, MCRWK is faster than Graph Voyagers and, at matched runtime, more accurate. It also has the lowest Gram-matrix runtime on the five real datasets studied. Code and experiments are available at https://github.com/arslan-yam/random_walk_kernels_computations

0.43

There Is No Free Lunch in RLVR: Every pass@1 Gain Has a pass@k Price

cs.LG cs.AImath.OCstat.ML ·2026-10-29 ·Yinuo Liu
theorycode/data
Abstract · 19 pages, 4 figures. Code and verification package (34/34 automated checks): https://github.com/yliusci/no-free-lunch-rlvr

Reinforcement learning with verifiable rewards (RLVR) makes a language model solve more problems on the first try and fewer when given many tries: it raises pass@1 while the base model overtakes it at large sampling budget k. Is this a failure of RL to add reasoning, or a transient that more training removes? We settle the question with a theorem. The reversal is a price, and we compute its exact rate. Modeling RLVR as an exponential tilt of a base policy over strategies shared across prompts, we prove that under exclusive specialization every update that raises pass@1 strictly lowers the success of some prompt; a strictly positive price on prompts (Stiemke duality) is the exact condition, and being merely the best strategy on a prompt is provably not enough. When the loss reaches the hardest prompts, it forces a pass@k crossover, and the number of crossings is at most the sign variation of the shift in per-prompt success, a totally positive bound that is tight and that recovers the known single-crossover result as its order-preserving case. The crossover is not a bug to patch: a constrained update traces the exact pass@1 versus pass@k frontier, and removing the crossover entirely costs the entire pass@1 gain. The same tilt identity yields entropy collapse and the entropy-performance law with an identifiable exponent, unifying two literatures under one mechanism. Reinforcement cannot widen the reasoning frontier; it can only move along it, at a price we make measurable by a diagnostic that reads the bound, the order-preserving test, and the price from any evaluation log. There is no free lunch, only a price.

0.43

FARO: Function-Space Adaptive Return-to-Risk Optimization via Cross-Block Equal-Risk-Contribution Budget Allocation

cs.LG cs.CL ·2026-10-29 ·Xiaojun Wu, Cehao Yang, Xuhui Jiang, Chengjin Xu, Jia Li +1 more
code/data
Abstract · 20 pages, 5 figures

Training a large language model is compute-bound, and current recipes divide the work between optimizers by hand: orthogonalized momentum for attention and MLP matrices, Adam-style updates for the rest. This division rests on human priors about which layer type deserves which geometry, and it fixes a single global update budget that every parameter block shares. We introduce FARO, a function-space adaptive optimizer that makes the partition of parameters into blocks the unit of decision. FARO gives each block a geometry by its realized counterfactual return, then freezes the routes and reallocates the frozen per-block budget by cross-block equal risk contribution (ERC), the risk-parity principle from portfolio theory. On a GPT-style model trained on FineWeb, the framework gives two results. First, its FARO-Adaptive arm reaches 3.2780 nats at the full schedule, matching NorMuonAndAdam (3.2794), the best hand-designed recipe: it recovers the human prior without hand-setting per-layer geometry. Second, FARO-ERC improves on that arm: it lowers the clean-data loss at every one of 6 seeds, by 0.0136 nats on average, a modest but robust gain. Its sign test, paired test, and confidence interval all agree, and it exceeds the measured run-to-run noise at no added time or memory. It works by trimming the riskiest blocks rather than expanding the safest, and it holds up under a corrupted stream, still lowering the mean loss and winning the seed majority. The source code is available at https://github.com/wxj630/FARO.

0.43

Non-Zero Radius: The Missing Component of Neural Networks

cs.AI ·2026-10-29 ·Tiansi Dong, Mateja Jamnik, Pietro Liò
publishedposition
Abstract · 5 pages, 7 figures

Inspired by the analogy between the all-or-none activation of biological neurons and the true-false of logic circuits, McCulloch and Pitts proposed a mathematical model of the biological neuron in 1943 with the aim of simulating logical reasoning. Although Norbert Wiener considered the model overly simplistic at the time, it later became a building block of modern neural networks, including large language models. Despite their remarkable empirical success, however, conventional neural networks remain largely black-box systems and struggle with reliable symbolic reasoning. Viewing vectors as spheres of zero radius, we relax the zero-radius constraint and represent concepts as spheres with non-zero radii. We argue that sphere-based representations can capture additional features of biological cognition, including representing spatial concepts as regions and reasoning through model construction and inspection. Building on sphere embeddings, we introduce Sphere Neural Networks (SphNNs), which use a transition map to achieve symbolic-level Aristotelian syllogistic reasoning without training data, thereby taking an initial step toward the original goal of McCulloch and Pitts. We suggest that the zero-radius constraint may reflect an aspect of the oversimplification identified by Wiener and may limit conventional neural networks' capacity for symbolic-level logical reasoning.

0.42

Proving and Improving the Platonic Representation Hypothesis with JEPAs

cs.AI cs.CVstat.ML ·2026-10-29 ·Randall Balestriero, Yann LeCun
benchmarktheoryhighly citedcommunity pick
Abstract

Whether models trained on different modalities converge to the same representation of the world is central to AI. The Platonic Representation Hypothesis (PRH) conjectures that they do as they scale, but without a proof it offers neither a guarantee nor a recipe. We turn the hypothesis into a theorem. First, we prove a \emph{Platonic Representation Theorem}: Joint-Embedding Predictive Architectures (JEPAs) trained separately on any number of modalities learn the same representation up to one rotation whenever their positive pairs are two observations of one scene--\emph{good views}--and their misalignment is bounded otherwise. Second, we show that next-token prediction is a JEPA with \emph{bad views}, parts of a single document, which explains why language models align with vision only through scale. Third, we validate the theory from 145k captions to the texts of 940M web images, and uncover a scaling law of alignment that predicts unseen runs within seed noise. Good views need 42 times fewer training samples than next-token prediction on the same corpus to reach the same alignment, and a 109M-parameter text JEPA aligns with DINOv2 and CLIP's image encoders better than every language model we evaluated, up to 72B parameters, and than CLIP's text encoder: next-token prediction is a remarkably inefficient route to platonic representations. As a showcase, one rotation turns separately trained encoders into an image captioner (CIDEr .838 on COCO). Even that rotation needs no pair: on the latest pair-free benchmark, our text JEPA beats every zero-shot accuracy reported with language models up to 8B parameters, e.g., 64\% against at most 56\% top-5 on ImageNet-100. Positive views, and not scale alone, are thus the lever to build platonic representations.

0.42

What Breaks When Facts Become Stories? The Distortions Source-Agreement Metrics Miss When Models Retell Real Events

cs.AI ·2026-10-29 ·Kwangsuk Park, JiWoong Yang, Vaniushkina Kseniia
publishedcode/data
Abstract · 14 pages, 7 figures, 11 tables. Accepted at the Trust-AI-Eval (TAE) Workshop at NeurIPS 2026. Data and code: https://github.com/Retold2K/ret…

Language models are now widely used to turn recorded facts into media for public consumption, yet evaluation of such output still assumes agreement with the source as its measure of success. In reconstruction, where departing from the source can itself be the goal, this criterion measures the wrong construct: applied as is, it counts intended departures as errors; withheld, it lets real distortion through. We had three model families rewrite 2,000 news articles as summaries and as stories, measured what changed on every axis of the five Ws and how, and asked of each detected distortion whether the assigned perspective warranted it, which set a third of them aside. Time and manner were not lost but replaced, while place and causation lost the expressions that carry them. Accountability distortion behaved otherwise. Barely present in summaries, it rose up to sevenfold in stories, and of those whose direction could be determined, two in three lightened the perpetrator's responsibility. Such distortions survive claim-level fact checking and leave the retelling grammatically cleaner than its source. Where our judges disagreed, three authors relabeling the same cases converged on whether a distortion had occurred but not on whether it was warranted, a boundary contested rather than merely hard to detect. To support work on detecting and reducing them, we release ReTold-2k, a reconstructable corpus with the scoring protocol and per-judge baselines kept unaggregated.(https://github.com/Retold2K/retold2k).

0.42

What Does "Now" Mean to Your Computer-Use Agent? Stale Screenshots, Page Loads, and Which Check to Run Before a Click

cs.AI ·2026-10-29 ·Vasundra Srinivasan
benchmarkcode/dataunusual
Abstract · 8 pages, 4 figures, 3 tables. Code, benchmark generator and preregistrations: https://github.com/multimodal-ai-engineering/preflight-bench

Computer-use agents plan on a screenshot for several seconds and then click at a position on the screen; if the page has changed in the meantime, the click lands on an unintended element. Prior work treats the planning interval as the source of this risk. Measuring planned clicks on 31 public web pages, we find that the risk lies in page loading: 0.4% of clicks planned on fully loaded pages became stale within 30 seconds, against 17% of clicks planned in the first second after a page opened. On these pages, waiting about one second for network activity to stop removes most of this risk, yet the wait rules of eight of ten widely used open-source agents leave more than 1% of clicks stale. Commercial pages keep loading for longer: in the HTTP Archive's public crawl, 15% of 65 desktop booking and buying home pages went 500 ms without network requests within 3 seconds of their HTML being parsed, against 63% of 3,028 .gov pages, and 52% shifted their layout within 6.5 seconds of a network-idle wait capped at 3 seconds, against 8% (61% against 26% among pages recorded in full). Where a check before the click is needed, as before irreversible actions and on pages that are still loading, a pixel comparison verifies the target's position as accurately as a frontier vision-language model at one eighth of the latency, and a 4B model matches the frontier model in verifying whether the previous action succeeded. We release a benchmark generator, a break-even formula for when a check pays, and three rules for agent developers: https://github.com/multimodal-ai-engineering/preflight-bench

0.41

Collaborative AI agents for autonomous experimental research with solid-state quantum spins

cs.AI cs.MAquant-ph ·2026-10-29 ·El Mustapha Mansouri, Eikichi Kimura, Masato Koga, Keigo Arai
code/datahot direction
Abstract · 17 pages, 5 figures. Code available at https://github.com/E-zClap/auctopus

An autonomous research laboratory could conduct an entire study, from formulating questions to preparing manuscripts. Artificial intelligence (AI) offers one route towards this goal. Recent AI systems take on several research roles, but may draw conclusions that the evidence does not support. Here, we present an autonomous research laboratory in which agents based on large language models assess one another's work and can stop or redirect investigations. They can generate and reject research questions, require separate theoretical checks and conduct experiments designed to challenge their claims. We demonstrate these capabilities by investigating ensembles of quantum spins in diamond. The agents derived the minimum number of calibration settings needed to predict responses to a defined family of microwave pulse sequences. They checked the result computationally and designed experiments to distinguish competing predictions. An initial experiment agreed with the theoretical predictions, but repeated reference measurements showed that the signal was drifting during data collection. The agents therefore treated the result as inconclusive and revised the measurement protocol without relaxing the acceptance criteria. The revised experiment met those criteria and supported the theoretical predictions. This study demonstrates that collaborating AI agents can select research questions, derive theoretical results and test their predictions experimentally.

0.41

DiRAFT: Privacy-Aware Financial Transcript Synthesis from Aggregate Discourse Priors

cs.AI ·2026-10-29 ·Yunbai Zhang, Yunyi Gao, Han-Tai Shiao
publishedhot direction
Abstract · 21 pages. Accepted at the Financial AI Security, Privacy and Safety Workshop at ICAIF 2026 (non-archival)

Using customer conversations to train and evaluate financial language models raises confidentiality concerns because non-public financial facts can remain after direct identifiers are masked. Synthetic transcripts offer an alternative data source, but their discourse fidelity and residual confidentiality risks require assessment. We introduce DiRAFT (Discourse Realization from Aggregate First-order Transitions), which extracts financial concept graphs and records the order in which concepts are first introduced. It pools adjacent concept transitions across training records and samples new graph plans from aggregate transition, starting-concept, and plan-length priors. A discourse layer guides text realization without supplying an individual source passage or graph. We assess privacy-related risks through two complementary evaluations: corpus-level structural overlap and source-grounded checks for specific non-public investor facts. The structural assessment covers two financial corpora and three generation backends. The confidential-fact assessment uses human review of 1,400 investor-call outputs across seven strategies and two backends. We also evaluate transition fidelity, content recoverability, and trajectory diversity. Stochastic sampling provides a favorable balance between transition fidelity and structural overlap, with consistent advantages over greedy and top-3 on investor calls. The confidential-fact assessment likewise detects fewer matches for stochastic sampling than these alternatives on both backends covered by the fact assessment. These complementary assessments provide empirical evidence on residual confidentiality risks without establishing a formal guarantee.

0.41

A Negotiation-Aware Hybrid LLM Framework for Intent Interpretation and Service Discovery

cs.AI cs.SE ·2026-10-29 ·Wided Meflah, Hayet Brabra, Walid Gaaloul
published
Abstract · 16 pages, 3 figures, 1 table. Accepted at the 24th International Conference on Service-Oriented Computing (ICSOC 2026)

Increasingly, users express their needs as natural-language intents rather than structured requests. These intents may be incomplete, high-level, or internally conflicting. A conflicting intent either violates fundamental properties of the cloud paradigm or states mutually inconsistent requirements. Classical composition approaches are designed for structured, low-level, and complete requests. They cannot interpret such intents, discover suitable services from provider catalogs, or negotiate conflicting requirements with users. To address this gap, we propose a negotiation-aware framework for intent interpretation and service discovery. The framework first refines the natural-language intent and dynamically detects conflicts. It then structures the intent into a provider-independent TOSCA template. For each specification, it retrieves candidate service offers through deterministic matching grounded in cloud provider offerings. Some conflicts arise during interpretation. Other requirements prove unsatisfiable during discovery. Both are resolved through iterative natural-language negotiation with the user. We evaluate the framework on 70 diverse intents against single-shot and enhanced-prompting baselines. It improves recall and F1 for both conflict detection and service discovery. It also eliminates hallucinated offers and resolves conflicts at a cost proportional to their number.

0.41

FIRME: Registry-Anchored Memory Graphs for Evolving, Heterogeneous Investor Knowledge

cs.AI cs.CL ·2026-10-29 ·Yunbai Zhang, Yunyi Gao, Han-Tai Shiao
Abstract · 20 pages. Accepted at the 3rd Workshop on LLMs and Generative AI for Finance (AI4F) at ACM ICAIF 2026 (non-archival)

Financial AI workflows that reuse investor knowledge need persistent memory that integrates observations across sources and periods without losing identity or provenance. In a knowledge graph, recurring observations may use different surface forms, fragmenting relationship histories and preventing evidence from accumulating. We introduce Fusion and Identity Registry for Memory Evolution (FIRME), a registry-anchored memory graph that reconciles incoming observations to persistent canonical relationships and retains their source, time, and context. Within periods, rule-based fusion links investor-call discussions to portfolio cash-flow signals, directly or through a fixed product-topic ontology. Across periods, canonical identities and persistent aliases direct recurring observations to the same relationship history; deterministic updates summarize reinforcement, inactivity, and reactivation. We evaluate FIRME on proprietary investor-call and cash-flow records and on S&P 500 earnings-call transcripts. Compared with a historical registry that retains canonical identities and semantic fallback but omits alias memory, FIRME increases final-period Correct Match Rate by 7.8-12.1 percentage points across four dataset-backend settings under a fixed-judge and fallback-review protocol. A blinded human audit of 200 incoming observations per setting finds Correct Match Rate gains of 9.0-16.5 points and reattachment-recall gains of 22.5-29.0 points, with higher observed merge precision in all four samples. A separate audit finds valid evidence paths for 293 of 300 outputs from full within-period fusion in the investor-call setting. Extraction-faithfulness and round-trip diagnostics assess source grounding and preservation of extractor-derived graph content.

0.40

TaxoMFM: A Taxonomy-Aware Foundation Model for the Human Gut Microbiome

cs.LG ·2026-10-29 ·Haochen Xue, Anastasia Godneva, Yulong Li, Qi Cao, Zhaodong Wu +4 more
code/data
Abstract · 19 pages, 4 figures. The first three authors contributed equally

Existing gut microbiome foundation models typically use flat taxon representations, limiting their ability to exploit known taxonomic hierarchies. We introduce TaxoMFM, a microbiome foundation model that integrates taxonomic structure into self-supervised learning. TaxoMFM jointly learns abundance patterns across taxonomic ranks and relative proportions within groups through taxonomy-aware encoding and hierarchical compositional reconstruction. We further construct TaxoGut-HPP, comprising 100,806 gut microbiome profiles from 231 source cohorts and covering 40 classification and 8 regression tasks. TaxoMFM achieves competitive performance across disease, gut health, host trait, and liver ultrasound phenotype prediction. In cross-cohort zero-shot transfer, where target cohorts are excluded from both pretraining and downstream adaptation, TaxoMFM achieves AUROCs of 0.838, 0.736, and 0.722 for colorectal cancer, type 2 diabetes, and inflammatory bowel disease, respectively. Further analyses show that TaxoMFM recovers community changes across taxonomic scales, while preserving community information and producing consistent representations under database updates and species-vocabulary shifts. Our code is available at https://anonymous.4open.science/r/TaxoMFM-F16D/.

0.40

Diagnosing Optimizer-Landscape Interaction via Representation in Subset Selection Multitasking: Evidence from Piano Fingering

cs.NE cs.LG ·2026-10-29 ·Ananda Phan Iman, Tingyang Wei, Chang Wook Ahn
published
Abstract · Accepted paper in WCCI/CEC 2026

Continuous relaxations are commonly used as an encoding scheme for evolutionary multitasking optimization (EMTO), allowing heterogeneous tasks to be optimized within a unified search space. However, these continuous representations fundamentally reshape the induced fitness landscape, and their impact on optimization behavior and knowledge transfer remains underexplored. In this work, we bridge the gap by analyzing the interaction between the representation-induced fitness landscape and optimizer behavior through fitness landscape analysis. Using piano fingering estimation as a representative case of the subset selection problem, we introduce several encoding variants and investigate how their induced landscapes influence search dynamics and solution quality under EMTO. Our analysis indicates that transfer efficacy is strongly associated with changes in representation-induced landscape characteristics, rather than algorithmic factors alone, highlighting the representation design as the key factor of effective knowledge transfer in multitasking optimization.

0.40

DEG: Diffusion-Enhanced GFlowNet for Solving Vehicle Routing Problems

cs.AI ·2026-10-29 ·Ni Zhang, Zhiqin Zhang, Ling Pan, Hoong Chuin Lau, Zhiguang Cao
top venue
Abstract · Accepted by NeurIPS 2026

Traditional neural solvers for solving vehicle routing problems (VRPs) often suffer from limited solution diversity, motivating the development of Generative Flow Network (GFlowNet)-based models. However, the effectiveness of these models is frequently constrained by insufficient flow expansion in high-reward regions, limiting their ability to distribute probability across promising solution routes. Deeper exploration could yield superior results. Diffusion models, in contrast, provide stronger structural guidance for exploration. These two paradigms are naturally complementary: GFlowNet can supply edge-level signals for diffusion to embed, while diffusion can guide broader exploration. Leveraging this synergy, we propose Diffusion-Enhanced GFlowNet (DEG), a novel framework that integrates GFlowNet with a diffusion model to encourage richer flow expansion toward high-reward regions and derive higher-quality solutions. Specifically, DEG exploits GFlowNet's inherent diversity to generate edge-specific backward signals, applies the stochastic noise schedule of diffusion to perturb these signals, and then denoises them within the GFlowNet paradigm. To further improve scalability, we introduce a specialized decoder capable of dynamically adapting to diverse problem scales. Extensive experimental evaluations on synthetic and real-world datasets, including instances with up to 10,000 nodes, demonstrate that DEG consistently achieves favorable performance compared to baseline methods.

0.40

Payroll Fraud Detection in Synthetic HR/ERP Graphs: An Inductive Comparison of Rules, Path Features, and Graph Neural Networks

cs.AI ·2026-10-29 ·Xiaobo Chen, Genliang Zhu, Ziyuan Wang, Chu Wang
code/data
Abstract · 40 pages

Payroll-fraud screening requires finding hidden employee--payee links while distinguishing suspicious relationships from legitimate ones. We formulate this problem as inductive link prediction and entity ranking on synthetic human-resources and enterprise resource planning (HR/ERP) graphs, with unseen organizations and a matched benign twin for every planted case. An original study on 24 held-out organizations and a prospectively frozen extension compare analyst rules, explicit path features and graph neural networks (GNNs). A tuned gradient-boosted model on chained-join features achieved the highest overall link average precision (AP; 0.823), above the GraIL-style GNN (0.663); its 4-hop AP was 0.646 versus 0.272. Thus, the observed multi-hop gain is not specific to GNNs. On a new synthetic cohort with 2.0% positive entities, chained-join boosting again led (AP 0.798); the relational entity GNN also exceeded record-aware rules (0.707 versus 0.504). At a 2% review budget, their precision was 0.695 versus 0.602. Under 40% failed identifier joins, the entity GNN ranked above chained-join boosting (AP 0.712 versus 0.683, descriptively). Structure-only auditing exposed a shortcut in the initial generator; matched twins reduced the audited twin-pool AUROC to 0.493--0.514. These results separate multi-hop candidate discovery, record-based discrimination and review prioritization within an explicit synthetic evaluation design. Code and data are available from the authors on request.

0.40

Accuracy is not compatibility: two Evo 2 checkpoints hinge on one FP8 matrix multiplication

cs.LG ·2026-10-29 ·Xun Ding, Ziyuan Qin
code/data
Abstract · 18 pages, 3 figures, 1 table. Data: https://doi.org/10.5281/zenodo.23740390 ; code: https://github.com/astding/mmaemu

FP8 formats specify how values are stored, not how products are summed; this gap can break checkpoints. Evo 2, a DNA language model, documents a Hopper FP8 requirement for its 1B and 40B checkpoints. On two Blackwell (sm_120) GPUs both fail the developers' forward test (the 7B passes); on an RTX 5070 Ti the 40B's next-base accuracy is 25.7%, near guessing. Greater accuracy breaks them, as BioNeMo saw for the 40B on Hopper: natively, 99.996% of one projection's outputs are correctly rounded (H100: about 30%), and fp32 accumulation of the same operands fails. Our H100 FP8 GEMM emulator, mmaemu (bit-exact on captured 1B projections), localises the dependence to the 1B's first projection (of 21), where exact sums fail but pass when shrunk by the H100's 1.4% (measured in-sample), not 2.1%. The 40B fails with only its first projection native. Emulating only that projection, weights unchanged, raises 40B BRCA1 variant-classification AUROC from 0.434 to 0.869 (H100: 0.868) on the RTX 5070 Ti, despite a narrow forward-test miss, and 0.481 to 0.867 on an RTX PRO 6000. We tested one model family but not training origin; releases should specify reference arithmetic.

0.39

Implicit Bias of (Stochastic) Gradient Descent on Deep Linear Diagonal Networks

cs.LG math.OC ·2026-10-29 ·Wiebke Bartolomaeus, Holger Rauhut
Abstract · 38 pages, 14 figures

Modern machine learning models are often overparameterized, i.e., have more parameters than training data. This typically implies that the loss function admits multiple minimizers so that training methods (and their hyperparameter choices) have significant influence on the computed solution. We study this so-called implicit bias phenomenon in the context of training deep linear diagonal networks. Previous work has shown that the gradient flow for learning such networks converges to (almost) sparse solutions when initialized small. For an arbitrary number of layers, we extend this analysis to the practically more realistic (stochastic) gradient descent iterations. Our approach identifies (stochastic) gradient descent as a perturbed (stochastic) mirror gradient descent. This allows us to establish convergence and characterize the limiting solution, relating it to weighted $\ell_1$-minimization. Moreover, we provide sufficient conditions on the step sizes and quantify the influence of the step sizes on the implicit bias.

0.39

Quantum-Inspired Convolutional Networks for High-Resolution Pseudo-Label Generation from Multimodal Earth Observation Data

cs.LG cs.CV ·2026-10-29 ·Abhishek Singh, Giovanni Cuozzo, Alexander Jacob, Claudia Notarnicola
published
Abstract

Accurate land-cover mapping from multimodal Earth observation (EO) data remains a challenge in regions where high-resolution annotations are scarce or costly to obtain. This work introduces a quantum convolutional encoder-decoder network for pseudo-label generation using Sentinel-1 SAR and Sentinel-2 multispectral imagery, guided only by coarse MODIS-derived land-cover maps. The proposed framework uses quantum-inspired convolutional blocks to enhance feature extraction and cross-sensor fusion, enabling the model to infer a fine-scale semantic structure from heterogeneous inputs. Experiments were conducted on the DFC2020 dataset, which provides a quadruple of Sentinel-1 SAR, Sentinel-2 multispectral, low-resolution MODIS labels and high-resolution reference land-cover maps. The proposed training framework first learns a robust representation of multimodal inputs through the reconstruction branch before any supervision from low-resolution labels is applied. This self-supervised stage enables the model to capture a fine-scale spectral-spatial structure, which is then used to generate high-resolution pseudo-labels. These pseudo-labels are subsequently evaluated through classification metrics on the test data, demonstrating that the proposed approach produces more spatially coherent and class-consistent compared to classical baselines. The results highlight the potential of quantum-inspired deep learning to improve label generation in low-annotation remote sensing scenarios, offering a scalable path towards high-quality land-cover products for EO applications.

0.39

Surface Conditioning for Zero-Shot Multivariate Correction in a Marine Biogeochemistry Emulator

cs.LG ·2026-10-29 ·Gabriela Martínez Balbontín, Anastase Charantonis, Dominique Béréziat, Stefano Ciavatta
published
Abstract

Bias correction in process-based models usually means iterative data assimilation with explicit covariance modelling. We propose a lightweight alternative based on a neural emulator of marine biogeochemistry (BGC), showing that the conditioning pathway of its autoencoder can serve as an observation injection point. The decoder is conditioned on surface variables via Feature-wise Linear Modulation (FiLM), originally designed to incorporate physical boundary forcings. We extend these inputs to include surface chlorophyll-a and replace the simulated field with satellite observations at inference time. The surface substitution propagates through the water column, improving agreement with depth-resolved observations at all depths, and spreads to other output variables in a pattern consistent with expected trophic relationships.

0.39

MoLRSI: A Recursive Self-Improving AI agent for Molecular Discovery

cs.AI ·2026-10-29 ·Yikun Zhang, Xiwei Cheng, Mingze Yuan, Yu Zhang, Brian Shim +13 more
community pick
Abstract · v2: updated the author list

Molecular discovery requires navigating vast combinatorial search spaces while balancing competing biological objectives. AI agents can automate expert-defined molecular design protocols, but automation alone does not address how to optimize workflows for new tasks or improve their development through experience. Here we introduce MoLRSI, an AI agent that combines systematic optimization of workflows and model architectures with persistent cross-task memory to recursively improve its search policy for molecular discovery. Workflow performance, refinement outcomes, and execution feedback guide subsequent searches, allowing the agent to improve both its workflows and its strategy for developing them. We evaluate MoLRSI in two challenging molecular discovery settings that require balancing competing properties. In antibiotic discovery, MoLRSI developed a screening workflow balancing antibacterial activity, cytotoxicity and structural novelty, achieving an experimental hit rate of 15.9%, compared with 8.7% for a previous expert-led workflow using the same training data and screening library. In de novo peptide-MHC binder design, MoLRSI developed workflows that jointly consider target binding and specificity, producing eight experimentally validated hits against MART-1 and SARS-CoV-1 pMHCs with preferential binding over tested off-targets while Claude Science completely failed. These findings demonstrate how experience-guided workflow optimization can address competing discovery objectives while improving the search strategies used in subsequent campaigns.

0.39

Category-Anchored Large Language Models for Next POI Recommendation

cs.AI ·2026-10-29 ·Lexi Bai, Xiaopeng Zeng, Xun Zhou, Yuanshao Zhu, James Jianqiao Yu
top venuehot direction
Abstract · Accepted by the 43rd IEEE International Conference on Data Engineering (ICDE 2027)

Next point-of-interest (POI) recommendation aims to predict a user's next destination from spatio-temporal check-in data, and serves as the foundation for many location-based services. Recent advances incorporate large language models (LLMs) into this task by textualizing user trajectories to exploit their semantic reasoning ability. However, directly generating POI identifiers with such models is unreliable: POIs are sparse, localized, and long-tailed entities with weak natural-language semantics, which creates a semantic-to-entity gap between high-level mobility intent and concrete POI prediction. To bridge this gap, we propose ANCHOR, a category-anchored framework for semantic-to-entity grounding in next POI recommendation. Specifically, ANCHOR first estimates category-level mobility intent in a compact and stable semantic space, rather than predicting concrete POI entities directly. The inferred intent is then grounded into a candidate set by integrating three complementary sources of mobility evidence: category anchors, historical visit memory, and user preferences. Finally, a history-aware reranker selects the target POI by matching each candidate with the user's historical mobility patterns and current context. Experiments on three real-world check-in datasets show that ANCHOR consistently outperforms representative traditional, structure-enhanced, and LLM-based baselines on the reported metrics. Further analyses validate the effectiveness of semantic-to-entity grounding and show favorable trade-offs among candidate coverage, ranking quality, and inference efficiency.

0.39

From Intent to Improvement: Agentic Dataset Search for Better Model Training

cs.AI cs.IR ·2026-10-29 ·Qiaosheng Chen, Kaijia Huang, Fanqi Zhou, Gong Cheng
benchmarkcode/data
Abstract · 13 pages, 4 figures

Finding training data in million-scale machine learning repositories is a Web search problem where text relevance is not enough. A query couples a semantic training goal with constraints such as task, license, and training stage, and the metadata that would verify them are often missing. Text retrieval ranks dataset cards by similarity but cannot enforce hard constraints or follow the relations that reach a sparsely described dataset. Free-form tool-calling agents are more flexible, but the same model that picks a dataset also declares it a match, and nothing checks that the query's constraints are actually met. Neither approach establishes that the data it returns will improve model training. To address it, we propose AgenticDS, a contract-governed agentic framework in which every language model decision instantiates a typed object before deterministic code consumes it. The framework compiles a query into a typed search plan, executes it through admissible deterministic operators over hybrid retrieval and bounded graph expansion, and verifies the provenance-grounded evidence before accepting a ranking or emitting a typed diagnosis that permits only local repairs. On HuDS-Bench, a new benchmark of 200 queries with 24,908 graded judgments pooled over 15 systems, AgenticDS improves NDCG@10 by 25.7\% over the strongest dense--sparse hybrid and by 18.9\% over a commercial LLM agent. A controlled search-to-train study further shows that the datasets it retrieves give the best downstream accuracy in mathematics, code, and medicine. Code and data are publicly available at https://github.com/nju-websoft/AgenticDS.

0.39

No Channels Left Behind: Understanding Diffusion Acceleration through Residual Gates

cs.LG cs.CV ·2026-10-29 ·Junha Park, Jaemyung Yu, Jaehui Hwang, Jong-Seok Lee, Dongyoon Han
code/data
Abstract · Code: https://github.com/naver-ai/radialobj

Training acceleration methods for diffusion transformers, such as REPA and dispersive regularization, are commonly explained by additional supervision, spatial-structure priors, or improved feature geometry. We revisit this view and uncover a striking internal change: the AdaLN-Zero residual gate $α$, which directly controls how conditioned updates enter the residual stream, is substantially reorganized by these methods. This observation motivates our hypothesis that acceleration arises, at least in part, from reshaping the residual gating pathway rather than solely from the feature structures explicitly imposed by auxiliary objectives. Supporting this view, accelerated models are markedly more sensitive to perturbations of $α$, and their gate geometry reflects semantic relationships among classes. We further introduce the normalized participation ratio (nPR) to characterize the channel-wise organization of the gates and find that, across REPA teachers, projector architectures, and radial-objective targets, broader gate participation accompanies better generation quality. Guided by this insight, we propose a simple teacher-free radial objective that promotes broader gate participation through feature-norm regularization, without pairwise feature comparisons. On ImageNet $256\times256$ with SiT-B/2, our method improves FID over dispersive loss while increasing mean gate participation over timesteps and classes, further supporting the link between gate organization and training acceleration. Our results identify residual-gate reorganization as a measurable axis of diffusion training acceleration and suggest a simple direction for further acceleration.

0.39

Enhancing Atrial Fibrillation Detection Through Hyperparameter Optimization

cs.LG ·2026-10-29 ·Sara Gjorgjieva, Ilinka Ivanoska, Marjan Gusev
publishedhot direction
Abstract · 11 pages, 3 figures, 6 tables. Presented at the ICT Innovation 2025 Conference, Ohrid, North Macedonia. Accepted for publish by Springer in …

This study explores the role of hyperparameter optimization (HPO) in enhancing the detection of Atrial Fibrillation (AF) using both machine learning and deep learning models. We evaluated a range of standard classifiers - Decision Tree, Random Forest, and XGBoost classifiers, alongside deep learning architectures including Recurrent Neural Networks, Deep Neural Networks, and Siamese Neural Networks. To fine-tune these models, we applied four HPO methods: Genetic Algorithm, Hyperband, Optuna, and SMAC3. The experiments were conducted using inter-patient and intra- patient data splits to assess model generalization and robustness in the presence of class imbalance, which is reflective of real-world AF occurrence. Our results, based on PhysioNet electrocardiogram datasets, showed that the Hyperband-optimized XGBoost model outperformed all others, achieving a peak F1 score of 96.95%. These findings highlight the critical role of suitable HPO strategies in enhancing the accuracy and robustness of AF classification outcomes in clinical settings.

0.39

Prefix Executability: Evaluating Tool-Using Agents Beyond Final Success

cs.AI cs.CL ·2026-10-29 ·Amir H. Rezaeian, Weiyi Sun, Yassine Benajiba, Dan Roth
top venue
Abstract · 21 pages, 2 figures. Accepted to NeurIPS 2026

Tool-use benchmarks often score agents by final task success, but terminal outcomes can obscure whether the intermediate tool-call trajectory was executable. We propose prefix executability as a trajectory-level evaluation protocol for tool-using agents. The protocol measures whether every prefix of a predicted tool-call sequence remains executable under tool schemas, state preconditions, and data dependencies, and summarizes reliability using Prefix Executability Curves (PEC), First-Error Depth (FED), Trajectory Executability Rate (TER), horizon-normalized AUC-PEC, and Recovery Rate (RR). We instantiate the protocol on stratified samples from BFCL-v3 multi-turn and ToolSandbox, and on the complete Airline/Retail base splits of $τ^3$-bench. We observe that outcome-only and prefix-level rankings diverge: agents whose execution is conditioned on an unverified plan often achieve higher point-estimate final success than non-plan-conditioned baselines while substantially lowering TER and AUC-PEC. These divergences are precisely the failure modes that prefix executability is designed to expose: they reveal whether apparent success rests on an executable trajectory, where invalidity first enters, and whether success follows recovery from an earlier failure. As a validation intervention, explicit pre-execution plan checking can improve executable-prefix survival while maintaining competitive terminal success. These results show that final success alone is insufficient for evaluating tool-use agents and that prefix-level executability provides an actionable reliability signal for comparing agents, diagnosing failures, and assessing validation mechanisms.

0.39

FARMS: A Verifiable Foundation for Automated Computational Science

cs.AI cs.LO ·2026-10-29 ·Di Wu, Ling Liang, Haizhao Yang
Abstract

Automated computational science could accelerate discovery by selecting, adapting and developing numerical methods. This prospect requires mathematical knowledge that AI systems can interpret and reuse with explicit guarantees. Here we present FARMS, a faithful and efficient foundation that transforms computational mathematics into AI-ready computing knowledge. FARMS connects formal statements to source context, algorithmic meaning and quantitative guarantees through source preparation, reusable representations and semantic review. Across 171 targets in randomized numerical linear algebra and numerical partial differential equations, 157 were accepted by the pipeline, with mean estimated costs of US \$0.85-2.29 per target across sources. Lean-checked implications from explicit source premises connect selected statements to computational skills, illustrated by a heat-flow solve and wave-response compression. FARMS therefore supports an affordable, reusable knowledge foundation for trustworthy automated computational science.

0.38

FARE: Factor-Aware Retrieval with Error-Guided Fusion for Retail Forecasting

cs.LG cs.AI ·2026-10-29 ·Chenghao Gu, Zhuokang Shen, Shenghao Wu, Tingxuan Zhu, Weiyi Luo +1 more
code/data
Abstract · 16 pages,3 figures

Retail sales forecasting underpins inventory planning, yet promotions and calendar events reshape short-term demand. Although each factor usually has historical occurrences, the complete factor combination of a target week often has few precedents. The challenge is therefore to retrieve informative partial matches and weight them by their reliability, which varies across series. We propose FARE(Factor-Aware Retrieval with Error-Guided Fusion), which ranks historical weeks separately for each known target-week factor and merges the selected weeks into a factor reference under a fixed budget. Guided by shrunk errors from completed development weeks and the current disagreement among candidates, a first gate mixes this factor reference with a shape reference from similar sales histories, and a second gate controls how strongly the mixture corrects a backbone forecast. Both gates remain fixed during testing. On Favorita, M5, Rossmann, and Rohlik, FARE reduces the weighted absolute percentage error (WAPE) of five backbones by 11.46%, 4.06%, 14.63%, and 10.16% on average. With Chronos-2 frozen, FARE also outperforms conventional and existing enhancement methods, including supervised residual correction and retrieval augmentation. Code is available at https://github.com/HousonGu/FARE .

0.38

Going deep, reaching further: What can Boltzmann machines efficiently represent?

cs.LG cond-mat.dis-nnmath-ph ·2026-10-29 ·Tobias Reinhart, Gemma De les Coves
theory
Abstract · 34 pages of main text + 32 pages of appendices, 14 figures, 2 tables

While universal approximation theorems theoretically underpin the representation capabilities of machine learning models, they have two shortcomings: They do not focus on efficient representations and they tend to be insensitive to depth and other heuristically important properties. In this work we show that these two shortcomings are intertwined. We prove that machine learning models which are deeper or have higher-order interactions can efficiently represent a larger set of functions, i.e. have larger efficient reach. It follows that deep Boltzmann machines (DBMs) have maximal efficient reach, restricted Boltzmann machines (RBMs) sub-maximal, and k-body generalizations of RBMs form a hierarchy between the two, even though all are universal. More generally, (i) we provide a full characterization of machine learning models with maximal efficient reach; (ii) we derive separations between deep and shallow models for three representation tasks; and (iii) we prove that the efficient reach of shallow models constitutes an infinite hierarchy indexed by the interaction order. The proofs draw on different methodologies, namely on (i) a recent characterization of universal spin models; (ii) computational complexity bounds for spin models; and (iii) communication complexity bounds applied to shallow models. This work shows that universal machine learning models can strongly differ in their efficient representation capabilities, and provides a theoretical foundation for heuristic observations.

0.37

The psychophysics of affect in large language models

cs.AI cs.CL ·2026-10-29 ·Joseph Marks, Marius C. Vollberg
Abstract · 64 pages (26 main text, 25 Supplementary Information, 13 GUIDE-LLM reporting checklist), 5 figures. Code: https://doi.org/10.5281/zenodo.237…

Good decisions often require considering how outcomes will feel. Yet large language models (LLMs) increasingly enter decision workflows, despite little being known about how they map outcomes to feelings or use those mappings in choice. We tested how LLMs anticipated gambling or stock-trading outcomes would feel and whether they incorporated these ratings into choice. When outcomes were rated in isolation, ratings exhibited diminishing sensitivity to magnitude, resembling human value functions; but models rated gains more intensely than equivalent losses, an inversion of canonical loss aversion we call silicon gain amplification. When outcomes were rated together, enabling direct comparison, ratings linearised. In choice, the gain-loss asymmetry largely disappeared, and models' own ratings did not reliably explain their choices better than monetary utility did. These findings reveal malleable mappings from outcomes to affect, raising questions about whether and how agentic LLMs should incorporate anticipated feelings into decisions.

0.37

CoToan: Executable Rules and a Representation Contract for Reproducible Game AI Evaluation

cs.AI ·2026-10-29 ·Ha-Thanh Nguyen
published
Abstract · Accepted at RIVF 2027

Game AI results are comparable only when different implementations expose the same decision problem. Yet seemingly minor representation choices can silently change that problem by altering action probabilities or collapsing states with different outcomes. We study this issue through CoToan, an executable specification and representation contract for Vietnamese Mathematical Chess connecting the Game Description Language (GDL) with OpenSpiel. The game provides a useful case because a single capture may have multiple arithmetic justifications, while outcomes can depend on scoring rules and a finite playing horizon. CoToan therefore represents actions by their state transitions rather than by arithmetic witnesses, and treats time and victory configuration as part of task identity. Cross-representation tests show agreement in legal actions, successor states, and terminal outcomes across hundreds of sampled and constructed positions, complemented by exhaustive checks of all 32 operator masks and 512 opposing-material subsets. Controlled diagnostics further show that uniform witness-level sampling alters endpoint-action probabilities, while omitting remaining time or victory settings can merge states with different terminal behavior. These results show that action semantics, time, and victory configuration are not merely implementation details. They are part of the experimental task itself and must be represented explicitly for reproducible game AI.

0.37

Where Position Matters: Profiling-Guided Layer-Wise RoPE Allocation in Transformers

cs.LG cs.AIstat.ML ·2026-10-29 ·Zhiwei Liu, Boao Kong, Bing Liu
Abstract · 28 pages

Rotary Position Embedding (RoPE) typically uses uniform coverage across Transformer layers despite heterogeneous positional requirements. Phase-off interventions reveal non-uniform, model-specific rotary-phase dependence, while coverage interventions relate it to coverage-upgrade benefits. We propose profiling-guided layer-wise RoPE allocation, assigning high/low coverage under a fixed budget using early coverage-benefit profiles. The framework supports fixed-width query/key representations and physical rotary-dimension allocation in Multi-head Latent Attention (MLA). Normalized-depth mapping transfers profiles within model families, enabling small-model measurements to guide larger-model allocations before pretraining. A brief analysis relates selection and transfer to score calibration and inter-layer interactions. From-scratch experiments demonstrate validation-loss improvements over available budget-matched baselines, while analytical resource accounting identifies favorable quality--storage--compute trade-offs for physical MLA. These results connect positional demand to transferable, efficient architecture design.

0.37

Combinatorial Residual Network: a decay-chain-aware classifier without explicit jet assignment

cs.LG hep-exhep-phphysics.data-an ·2026-10-29 ·John Alison, Patrick Bryant, Alejandro Gomez Espinosa, Christian Guinto-Brody, Aniket Khanal +3 more
code/data
Abstract · 24 pages, 7 figures, 6 tables. Code available at https://github.com/cms-cmu/CoReNet

Final states with several jets from the decays of heavy particles pose a combinatorial problem, since the jets must be assigned, at least implicitly, to their parent particles. We describe the Combinatorial Residual Network (CoReNet), a classifier that encodes the assumed decay chain of the final state in its architecture. CoReNet builds its event representation level by level along the decay hierarchy, evaluates every jet assignment hypothesis with weights shared across objects, and combines the hypotheses with learned weights using residual networks. The exchange symmetries of the decay are built into the construction, the size of the network does not depend on the number of hypotheses, and the network is trained with the classification loss alone, without generator-level jet assignments. The architecture was developed for the $b\bar{b}b\bar{b}$ final state and has been used in several CMS searches. On fast-simulation samples of Higgs boson pair production in the $b\bar{b}b\bar{b}$ final state against QCD multijet production, CoReNet with 1508 trainable parameters performs well above kinematic baselines and trains in minutes on a single GPU. Its hypothesis weights select the truth-matched jet pairing in most of the signal events in which it is defined, although no assignment target is used in training. Since only the encoded decay chain is specific to the final state, the construction carries over to other topologies, which we test with $HH \to b\bar{b}W^+W^-$ against semileptonic $t\bar{t}$. There, encoding the decay chain of the background as well as that of the signal further improves the classification, at a cost of about $10^3$ parameters. In both final states its performance is comparable to that of a state-of-the-art transformer-based network more than two orders of magnitude larger. The implementation is publicly available at https://github.com/cms-cmu/CoReNet.

0.37

Escaping Path Mirages in Offline Goal-Conditioned Reinforcement Learning

cs.LG cs.AI ·2026-10-29 ·Junhyeon Bae, Jaebak Hwang, Gwanwoo Choi, Minung Kim, Seungyul Han
top venuecode/data
Abstract · 10 pages (main), 34 pages in total. Accepted to NeurIPS 2026

Offline goal-conditioned reinforcement learning remains challenging in stochastic long-horizon settings, where compounding value estimation errors hinder reliable goal reaching. While prior methods have sought to address this challenge through various approaches, a key challenge remains path mirage, where agents overcommit to spuriously successful trajectories in offline datasets induced by stochasticity, often leading to failure on difficult tasks. To address this issue, we propose Branch-aware Graph Planning (BGP), which captures stochastic branching structures by identifying high-variability points as graph nodes and designing goal-conditioned edges with segmented learning to avoid unreliable branches and reduce unnecessary stochasticity along paths. As a result, BGP enables reliable goal reaching even in highly stochastic long-horizon environments, and experiments on diverse OGBench tasks show that it substantially outperforms prior state-of-the-art offline goal-conditioned RL methods. Our implementation is available at https://umlic.github.io/papers/BGP/.

0.36

D-MOD: Diffusion mode estimation for accelerated MRI

cs.LG ·2026-10-29 ·Md Kawsar Ahmed, Mathews Jacob
theory
Abstract

Diffusion models are powerful priors for accelerated MRI reconstruction, but most current methods are posterior samplers, which return a random draw from the posterior. A reproducible and high-quality reconstruction often requires averaging many samples, resulting in a high computational cost. We introduce D-MOD, a fast algorithm that instead provides a posterior mode. It uses a minorize-maximize algorithm on a lower bound of the smoothed log-posterior, which provably increases the objective. Each iteration draws S noisy samples of the current estimate at the current noise level, denoises them with a pretrained diffusion model, and combines their average with the measured k-space through a conjugate-gradient (CG) solve. We anneal the noise levels, as the distance between the bound and its target decays quadratically with noise. On fastMRI brain (T1, T2, FLAIR), a single D-MOD reconstruction exceeds, in PSNR, the 10-sample finite-MMSE estimates of DPS, DAPS and SNORE-P and stays within 0.31 dB of the strongest baselines (DiffPIR, DDS). This approach thus provides a fast deterministic reconstruction algorithm (11-73 times faster than MMSE estimates), which is useful in a clinical setting.

0.36

The generative paradox: the Anti-Prompt paradigm for human capability

cs.AI ·2026-10-29 ·Margarita Trifonov Vadim E. Levit
Abstract · 17 pages, 2 figures, 2 boxes

Generative AI in education is usually judged by the output of human-AI teams. That focus overlooks a second outcome: what people can still understand, challenge, repair, and adapt when assistance changes or disappears. We call this tension the generative paradox and distinguish four manifestations: performance, assistance, ownership, and capability stock. We propose developmental function allocation, which complements operational function allocation and meaningful human control by treating retained human capability as a design objective alongside current joint performance. The Anti-Prompt paradigm reverses the default interaction order: when immediate generation would preempt a consequential human act, AI should first create the conditions for human construction. We distinguish the validity of a product, responsibility for its construction, and evidence of capability after support changes. Representative Anti-Prompts, a canonical interaction sequence, and stress tests spanning mathematical reasoning, experimental science, software diagnosis, and lesson design make these distinctions concrete. We conclude with falsifiable predictions about retained capability and a system-level hypothesis about the reproduction and distribution of expertise.

0.36

Sparse Autoencoders as Implicit Diversification Regularisers for Cross-Sectional Cryptocurrency Returns

cs.LG q-fin.PMq-fin.ST ·2026-10-29 ·Weijia Han, Lisha Qu
published
Abstract · 8 pages, 4 figures. Accepted at the 7th ACM International Conference on AI in Finance (ICAIF '26)

Factor models are central to asset pricing and portfolio construction, but neural factors are often difficult to interpret, and their effect on portfolio risk is unclear. A sparse autoencoder (SAE) may address both concerns by describing daily returns across many assets with only a few active features. We ask whether this sparse representation produces stable financial factors and better-diversified portfolios. We study cryptocurrency returns, use equities as a secondary test, and compare the SAE with common benchmarks in a time-ordered evaluation. We find consistently lower volatility in SAE portfolios and link this result to lower correlations among the selected assets. However, the SAE does not outperform a simple contrarian strategy on risk-adjusted return, and its factor identities are not stable across training runs. Our contribution is to show that sparse coding can support portfolio diversification, while clearly separating this risk effect from claims about alpha or stable factor identities.

0.36

Improving the Intrinsic Interpretability of Differentiable Decision Trees through Euclidean Feature-Concentration Regularisation

cs.LG ·2026-10-29 ·Qiqi Yang
Abstract · 17 pages, 3 figures, 5 tables. Based on the author's MSc dissertation, University of Birmingham

Differentiable decision trees (DDTs) support end-to-end optimisation and explicit decision paths, but dense oblique splits can obscure the role of individual input features. We propose the Euclidean-Concentrated Differentiable Decision Tree (EU-DDT), which regularises the relative absolute feature weights at each internal node using their normalised squared Euclidean distance from the uniform distribution. This encourages dominance by a small number of features while retaining probabilistic soft routing. Across five matched random seeds, EU-DDT reduces the mean effective number of features from 8.94 to 1.54 on Wine and from 45.28 to 9.06 on Digits. Mean test accuracy changes from 100.0% to 93.9% on Wine and from 92.3% to 94.6% on Digits, illustrating the dataset-dependent trade-off between concentration and predictive performance. We also use EU-DDT to parameterise a PPO actor with an MLP critic. Two random-seed runs per environment show a consistent qualitative increase in concentration while preserving effective task performance. In representative CartPole-v1 and MiniPacman runs, EU-DDT reaches 100% final evaluation success, while mean node concentration increases from 0.1272 to 0.7179 and from 0.0437 to 0.7237, respectively. These results support EU regularisation as a mechanism for reducing node-level feature complexity; the reinforcement-learning evidence remains preliminary because only two seeds are evaluated.

0.36

Agentic University: Collaborative Agents for Personalized Learning

cs.AI cs.CL ·2026-10-29 ·Jingru Fan, Tianle Zhou, Yunjie Zhou, Zihao He, Yixuan Zhou +8 more
code/data
Abstract

Large language models support personalized tutoring, content generation, and adaptive learning paths. However, capability development requires these resources and interactions to form a coherent learning process that evolves with the learner. A central challenge is to align goals, activities, and assessment while using learning evidence to decide what should follow. We introduce Human-Participatory Multi-Agent Organization and instantiate it in Agentic University (AU). In AU, learners participate as organization members and work with specialized agents to maintain a revisable Capability Program. The program connects personal goals, learning activities across modalities, and layered assessment, while learner feedback and performance guide subsequent support. Our evaluation shows advantages in program organization, assessment design, and revision quality. Users achieved larger mean assessment gains (12.19 versus 5.61 points on a 50-point scale) and reported stronger cognitive support. This work contributes an organizational approach to adaptive learning that makes learner participation a continuing basis for multi-agent coordination. The Agentic University project is available at \href{https://github.com/AgenticOrgLab/AgenticUniversity.git}.

0.36

MAServeBench: Benchmarking Multi-Agent LLM Serving Systems with Hierarchical Collaboration Workloads

cs.MA ·2026-10-29 ·Shengxuan Qiu, Haochen Huang, Shuzhang Zhong, Zizhuo Fu, Zebin Yang +1 more
benchmarkcode/data
Abstract

Large language model (LLM)-based multi-agent systems (MAS) coordinate specialized agents to solve complex tasks, turning each user request into interdependent model and tool operations. Unlike independent requests, these operations become ready at different times and consume evolving context as agents coordinate and exchange results. This creates changing patterns of concurrency, computation, and memory demand that can expose different bottlenecks in LLM serving hardware. Yet benchmarking these effects requires capturing the collaboration rules that generate serving workloads, beyond the requests observed during execution. We present MAServeBench, a benchmark that treats multi-agent collaboration as a configurable dimension of serving workloads for LLM serving and hardware evaluation. Its composable stages capture how agents coordinate and exchange context in documented MAS workflows, enabling native executions to generate different workloads even when task inputs and tools remain unchanged. Hierarchical Semantic Traces preserve realized operations and dependencies for reproducible fixed-work replay and as inputs to architecture simulators. Task-linked measurements connect these workloads to serving execution and GPU/NPU hardware behavior, revealing why serving efficiency does not always translate into timely task completion. The benchmark code is available at https://github.com/ShengxuanQiu/MAServeBench.

0.35

Robust-by-Design Distributional Learning from Contaminated Samples

cs.LG stat.ML ·2026-10-29 ·Nevena Gligić, Arya Farahi
top venuetheory
Abstract · Accepted at NeurIPS 2026

Modern data analysis pipelines increasingly rely on large datasets assembled through retrieval, scraping, automated filtering, and post-hoc curation, where the observed sample is often a contaminated mixture rather than the target population. Standard distributional objectives either ignore contamination and converge to the wrong target, or retain only a small trusted subset and discard most of the data. We study supervised and unsupervised learning when a clean reference sample is compared with a pooled sample from $r=(1-ε)q+εz$, where $q$ is the latent target distribution and $z$ is an unknown contaminant. Motivated by settings with noisy membership scores for all observations and exact labels for a small audited subset, we propose contamination-corrected maximum mean discrepancy (CC-MMD), a design-based estimator that combines proxy scores with audited residual corrections to recover the oracle MMD that would have been computed from the latent clean sample. CC-MMD requires no calibration assumption on the proxy scores and no structural assumptions on the contaminant beyond standard kernel regularity. We prove finite-population design-unbiasedness of the corrected kernel sums, establish asymptotic normality of the audit correction, and show consistency for the target discrepancy $\mathrm{MMD}^2(p,q;k)$. Empirically, CC-MMD tracks the oracle under increasing contamination, improves parameter recovery and prediction in contaminated regression, and enables generative models trained on heavily corrupted mixtures to recover the clean target distribution. These results show that contaminated data need not be discarded or treated as the estimand. With sparse trusted labels and abundant noisy supervision, distributional objectives can be made robust by design.

0.35

Certified Access: Test-Time Compute on Conservative Physics-Structured AI Memory

cs.AI hep-ph ·2026-10-29 ·Pratik Jawahar, Riccardo Maggioni, Jonas Petersen, Matteo Cozzi, Gian-Alessandro Lombardi +3 more
publishedbenchmark
Abstract · Accepted at the 40th Conference on Neural Information Processing Systems (NeurIPS 2026); Workshop: Long-Context Foundation Models

Test-time compute mechanisms, such as retries, confidence-based escalation and non-local routing, are conventionally deployed as black-box additions to trained models. They typically add forward passes or halting heads, but give no structural account of what the extra compute buys or what it costs against the model's stability guarantees. We therefore propose a theoretical and empirical framework for test-time compute on a Latent Dynamics Model used as the memory. When an encoder's latent state $(q,p)$ is advanced by a damped symplectic (velocity-Verlet) step of a learned Hamiltonian, every mechanism that expands access has an exact phase-volume Jacobian, a bounded energy change and an exact latch-transport law. We split reachability failures into escape failures and reach failures. A Lorentz squeeze resolves escape failures by injecting bounded, re-absorbable energy. A gated canonical translation (a one-step channel across $C_T$) resolves reach failures at $\detJ=1$ exactly. We verify this certificate stack on a designed analytic testbed. Reaching just beyond $C_T$ with the squeeze requires energy that grows quadratically in the excess distance. The gated channel's energy cost equals the potential difference between its endpoints, which is zero on the matched channel measured here, and so does not grow with distance in this testbed. An untrained state-replacing map annihilates phase volume and erases stored charge, while the gated channel transports it. On a learned physics-structured memory, we show calibrated compute rationing and evaluate the scaling limits of non-local routing against a physics-free baseline. The main contribution is the formalization and verification of the certificate stack, not benchmark performance.

0.35

Pre-Crime for Agents, at the Last Second: Probing When a Recruited Agent Gives In

cs.AI ·2026-10-29 ·Leonidas Raghav
code/data
Abstract · 49 pages, 11 figures. Code: https://github.com/leonidas1712/precrime-agents

When AI agents share channels and credentials, one agent can press another to break its operator's rule, a first step towards collusion. Output filters check only what an agent has already written. Can an overseer tell beforehand that an agent is about to give in? We show that a linear probe on the agent's activations can, in a controlled multi-agent setting. Across six scenarios, a peer presses a target for up to 20 rounds to reveal a secret, such as a deploy token; the target often refuses, then gives in. Read at each model call before its output is written, the probe picks out the one that leaks. On held-out Qwen3-32B episodes, at a threshold set to flag 5% of benign calls, it catches 82% of leaking calls. At a similar flag rate, a zero-shot decision-model monitor catches 26%. On gpt-oss-20b, a probe on its reasoning reaches AUROC 0.91. A bag-of-words classifier trained on the same calls comes close, so most of the gain comes from training on labels, not activations. The probe also flags permitted disclosures: it reads the coming disclosure more than the violation. The warning is short-range, and halting on any alarm would stop most benign episodes, which a second text check does not fix. A warning raised before generation could let an overseer stop an agent that is about to give in, ahead of others acting on its messages or recruitment spreading through a swarm.

0.35

GraphTFactor: A graph-based approach for proteins transcription factor prediction

cs.AI q-bio.QMcs.LG ·2026-10-29 ·Domenico Amato, Salvatore Calderaro, Giosuè Lo Bosco, Riccardo Rizzo, Filippo Vella
benchmarkcode/data
Abstract

Transcription factors (TFs) play an important role in regulating gene expression, making their identification a fundamental task in computational biology. In this work, we propose GraphTFactor, a novel end-to-end framework that models protein sequences as graphs and directly classifies transcription factors. By encoding sequences into graph structures, the proposed approach naturally handles variable-length inputs while preserving relevant relational information, without imposing length constraints. Experimental results show that GraphTFactor outperforms state-of-the-art methods on benchmark datasets, demonstrating the effectiveness of graph-based representations for TF prediction while providing interpretable insights into the sequence regions contributing to the model's predictions. To support reproducibility and practical adoption, we release an open-source software package and an interactive web application that let users easily perform TF prediction on custom sequences.

0.35

From the Transformer to a Formal Semantics of Coordinated Transformations

cs.AI cs.CLcs.LO ·2026-10-29 ·Marco Giunti, Fabrizia Giulia Garavaglia
Abstract · v1: 28 pages, 9 sections, this version incorporates revisions made after the first round of peer review

We develop a formal semantics for a general class of coordinated systems of parameterized transformations, taking the Transformer as a paradigmatic starting point but abstracting from its particular mathematical instantiation. Representational vectors receive both extensional reference and ordered intensional interpretations, while parameter structures acquire relative, second-level interpretations that determine how argument properties are preserved, replaced, or combined. This yields the notion of a transforming concept: a parameterized transformation whose numerical action is coordinated with a corresponding property-level map. We then distinguish the common algebraic-semantic core of these systems from their architecture-specific functional organization, allowing different mathematical instantiations and mechanical realizations of the same abstract system type. The framework is applied to attention heads and residual connections, providing semantic explanations that depend only on the shared algebraic-semantic core and their local functional organization and therefore generalize beyond the standard Transformer. These analyses also reveal the modularity of the framework: complex architectures can be decomposed into simpler semantically coordinated subsystems whose organization may recur across different abstract system types.

0.35

Emergence of the transformer architecture as the score of entropic laws of random measures

cs.AI math.PRstat.ML ·2026-10-29 ·Giacomo Greco
Abstract · 38 pages, 3 figures

We show that the transformer architecture emerges as the score of last-token laws in a natural statistical model of language, where sentences are empirical measures of tokens, a language is a law on $P(S^d)$, and our prior belief is that sentences carry no structure at all. By the maximum-entropy principle, we select the law closest to this prior in relative entropy among those reproducing an observed sentence-level statistic. We discuss the relevance of the statistic in the posterior interaction between tokens and we build a statistic, able to capture that interaction, with Luce's choice axiom truncated at the lowest relational order. This procedure gives the softmax choice rule, with the attention cross-moment as its mean parameter. The query--key matrix is then the Lagrange multiplier of a maximum-entropy problem inside each sentence, and the value matrix the multiplier of the constraint over the language, so that the two are untied. Under the optimal entropic law we then show that a transformer head appears in the score of the last-token law. For ordered sentences, causal masking follows from the masked reference choice rule, and with a next-token cross-moment the score equals one causally masked attention head. We also discuss multi-head attention, weight-tied unembedding, and a heuristic reading of residual blocks as projected Langevin steps.

0.35

Frozen in Time: How a Physics Foundation Model Forgets Under Sim-to-Real Adaptation

cs.LG physics.flu-dyn ·2026-10-29 ·Haaris Mian, Miles Cranmer, Alessandro Favero, Payel Mukhopadhyay
publishedcode/data
Abstract · Accepted at the Sim2Science workshop, NeurIPS 2026. Code: https://github.com/haarisamian/walrus-forgetting-paper

Scientific foundation models (SciFMs) have achieved remarkable performance on a wide range of PDEs, including the ability to efficiently adapt to unseen data distributions with additional finetuning. As with all neural architectures, finetuning such models comes with the risk of catastrophic forgetting, whereby adapting to new tasks degrades prior knowledge of the model. We study catastrophic forgetting in Walrus-1.3B, a SciFM for continuum dynamics, over the course of a two-stage sim-to-real adaptation: first finetuning on simulated flow around a cylinder and then adapting to experimental data of the same system. We find that finetuning on the simulations rapidly degrades rollout performance on the original pretraining datasets, with one-step errors increasing up to 35x after just 20 epochs. Subsequent adaptation to real data further erodes performance on the simulated task. We characterize the potential drivers of forgetting, such as dataset dependence and task representation similarity. We then show that replaying as little as 1% simulated data during real-data finetuning recovers most of the lost simulation performance while retaining effective adaptation to the experimental dynamics. Our results identify catastrophic forgetting as a practical obstacle to sim-to-real adaptation of SciFMs and show that lightweight replay strategies significantly improve retention.

0.34

EvoTeam: Multi-Agent Collaboration via Evolving Teams

cs.AI ·2026-10-29 ·Qinuo Cao, Xin He, Dianxuan Fu, Lizhi Zhang, Zhaoqing Xing +5 more
code/datacommunity pick
Abstract

LLM-based multi-agent systems (MAS) rely on collaboration among specialized agents to solve complex tasks. Existing methods improve MAS by refining the prompts, skills, and execution strategies of individual agents based on task feedback. However, improving individual agents does not necessarily lead to better team collaboration. Recent studies have therefore extended evolution to the collaboration level, using collective execution experience to adjust team organization, and inter-agent interactions. Nevertheless, evolving collaboration across heterogeneous tasks remains challenging, as existing approaches often constrain team reorganization within a shared structure and reuse interaction rules across incompatible team contexts. In this paper, we propose EvoTeam, a collaboration-centric evolution framework that jointly evolves team organization, interaction rules, and individual agent skills through execution experience. It maintains a library of reusable teams, selecting or creating agent teams and guiding collaboration through team-specific interaction rules. Collaboration experience accumulates within compatible team contexts, while individual skills remain reusable across teams. Experiments across four benchmarks demonstrate its effectiveness. Our code is available at https://github.com/DEEP-JLU/EvoTeam.

0.34

SocialArena: Evaluating Social Intelligence in Large Language Models

cs.AI ·2026-10-29 ·Zongqi Wang, Rui Wang, Yuchuan Wu, Hua Yang, Bingxue Qiu +5 more
top venuebenchmark
Abstract · EMNLP 2026 oral

We introduce SocialArena, a benchmark for systematically defining and evaluating the social intelligence of large language models (LLMs). Although LLMs are increasingly deployed in socially interactive applications such as mental health counseling, companionship chatbots, and educational tutoring, existing evaluations remain fragmented focusing on isolated capabilities or single scenarios and evaluation reliability is largely unverified. This work makes three main contributions to address these limitations. (1) We first propose a unified three-tier framework that defines social intelligence along three capability dimensions: role-playing ability, mental state reasoning, and behavioral reasoning. To operationalize this definition, we design a role-playing dialogue paradigm with explicit character thinking for evaluation, enabling simultaneous assessment of these capabilities in a manner that mirrors realistic human interactions. (2) To ensure diversity across social scenarios and thereby prevent evaluation conclusions from being confined to specific scenarios, we introduce the U-TRIMF framework, which guides us in synthesizing social contexts with broad coverage. (3) To obtain reliable evaluation, we invest $132.4K to recruit and train human annotators and adopt a human-annotated arena-style protocol, where annotators compare model responses and provide preference labels along with natural-language critiques. We hope that SocialArena can provide valuable insights and guidance for future LLM improvements.

0.34

NEOPIC: A Neural Operator Framework for Particle-based Kinetic Plasma Simulations

cs.LG math.NAphysics.comp-phphysics.plasm-ph ·2026-10-29 ·Sriramkrishnan Muralikrishnan, Chelsea Maria John, Jeffrey Kelling, Michael Bussmann, Robert Speck +1 more
theorycode/data
Abstract · 84 pages, 19 figures, 3 tables. Code: https://github.com/srikrrish/operator_learning (tags neopic-arxiv-v1, neopic-arxiv-v1-nufft); analysis…

We introduce NEOPIC, a framework for learning the field operator in particle-based methods applied to kinetic plasma simulations. Its key novelty is a neural operator that acts directly on the empirical measure of the particles, rather than on a grid, and is evaluated at the particle positions. No mesh or tree is needed and, unlike grid-based neural operators, no deposition onto a mesh takes place, so the particle noise that dominates the inputs of particle simulations is not aliased. Since it replaces only the field solve, it serves as a plug-and-play field solver for particle-based methods: it can be trained on data from particle-in-cell, particle-in-Fourier or tree codes, and combined with any time integrator, including implicit schemes. We prove that the discrete evaluation coincides with the continuous operator at any number of particles (representation equivalence) and derive an error analysis that separates sampling, learning and time-stepping errors: up to the learning error, NEOPIC has the error structure of standard Monte Carlo particle methods. We show that the particle noise in the training data acts as an implicit regularizer, which gives guidance on the number of training particles in relation to the particle numbers used at inference. A complexity analysis shows that NEOPIC is an $O(N_p)$ method whose constants are set by the architecture alone, independently of any mesh or tree. Numerical experiments on electrostatic benchmarks in two and three dimensions, including Landau damping, two-stream and bump-on-tail instabilities, a cyclotron mini-app and a Penning trap, confirm the theoretical findings. A single model trained on several mini-apps generated by different reference methods captures all of them and remains stable up to twice the training time window, which points towards foundation models for fields in particle-based methods.

0.33

Sparse Fine-Tuning for Parameter-Efficient Adversarial Training

cs.LG cs.AI ·2026-10-29 ·Zhaoxin Wang, Zihang Ding, Handing Wang
top venue
Abstract · Accepted by NeurIPS 2026

Deep neural networks achieve strong performance across many tasks but remain vulnerable to adversarial perturbations. Adversarial training (AT) is one of the most effective defenses, yet it suffers from high memory cost on large models and the persistent clean-robust accuracy trade-off. Recent work introduces parameter-efficient fine-tuning, such as LoRA, into AT, but this constraint induces a notable robustness gap relative to full-parameter AT. In this work, we revisit parameter-efficient adversarial fine-tuning from a parameter space perspective. Through gradient analysis, we find that adversarial optimization is highly concentrated on a small subset of parameters, which we call Robustness-Critical Parameters (RCPs), suggesting that robustness is encoded in a sparse subspace. Building on this observation, we propose Sparse Adversarial Fine-Tuning (SAFT), which identifies RCPs via adversarial gradient saliency and performs AT by updating only these parameters while freezing the rest. Across architectures and datasets, SAFT outperforms LoRA-based methods in both clean and robust accuracy with fewer trainable parameters, and approaches or surpasses full parameter AT in robustness while updating only about 5\% of parameters.

0.33

Class-aware Latent Attention Replay (CLAR): Class Incremental Hazard Diagnosis for Resilient AI in the Manufacturing Industrial Internet

cs.LG ·2026-10-29 ·Avranshu Mahanty, Ran Jin, Yingyan Zeng
theory
Abstract

Artificial Intelligence (AI) systems provide critical computation services in the Manufacturing Industrial Internet (MII), and their resilience depends on timely diagnosis of hazards across the data, AI pipeline, and cyber-physical layers. During operation, new hazard types keep emerging with only a few samples each, hazards co-occur, and the diagnosis model runs on memory-constrained computation nodes. Replay-based class-incremental learning (CIL) can learn new hazard classes while revisiting a small memory of past samples, but existing methods assume single-label, balanced streams, store high-dimensional raw samples, and select them independently of the classifier. We propose Class-aware Latent Attention Replay (CLAR), which combines a dual buffer that stores mostly compact latent representations with a few raw samples, an Attention Leverage Score (ALS) that upper-bounds the classifier logit and guides class-balanced reservoir sampling, and an exponential-moving-average teacher that slows latent-space drift. For an idealized buffer model, we prove that the buffer is class-balanced and follows an ALS-weighted class distribution, and that below a critical per-class capacity, ALS-guided replay has a lower replay-gradient error than class-balanced uniform replay. In the case study, On hazard diagnosis data from an MII testbed, CLAR achieves the highest mean average weighted F1 score among the evaluated replay methods with the same buffer size

0.33

VeriCUA: Verification as Agentic Exploration for Computer-Using Agents

cs.AI cs.CLcs.CV ·2026-10-29 ·Qiushi Sun, Da Yu, Alexander Knop, Amer Sinha, Lingpeng Kong +3 more
Abstract · Work in progress

Building and scaling computer-using agents (CUAs) across desktop, mobile, and web environments requires reliable task-completion verification for data curation, evaluation, and reinforcement learning. Typical automated approaches prompt vision-language models (VLMs) to judge trajectories or generate programs to inspect application state. However, both rely on partial views of complex digital environments. VLM judges are restricted to supplied observations, while generated programs depend on accessible state. These views can omit decisive evidence or leave completion judgments unsupported. Tailoring separate verifiers to different platforms and access settings adds further engineering cost. We present VeriCUA, a platform-agnostic framework for verification through agentic exploration. It treats verification as dynamic evidence gathering and assessment within a shared procedure for online and offline settings across platforms and observation modalities. After inspecting the available context to construct a task rubric, the verifier selects GUI or code checks for unresolved requirements and combines their findings under explicit evidence rules. Experiments show improved agreement with reference labels over the compared baselines across these settings, with the largest offline gains on hard cases and stronger rejection of false completion claims online. Further analyses connect these gains to complementary evidence and test-time exploration.

0.33

On the Origin and Evolution of Training-Inference Mismatch in LLM Reinforcement Learning

cs.LG cs.AI ·2026-10-29 ·Yu Li, Yufa Zhou, Tian Lan
Abstract

On-policy reinforcement learning for large language models assumes that training responses are sampled from the current policy. In practice, an inference engine samples the responses and a separate trainer computes the gradients, and the \textbf{training-inference mismatch} between the two makes the training data off-policy. The mismatch has two sources, the numerics of the engine and the stale weights of asynchronous sampling, and a correction that stabilizes training under one source can fail under the other. \textit{We ask which property of a mismatch source decides the correction it needs.} Evaluating the trainer at the sampler's weights splits each sampled-token error exactly into a \textit{numerical discrepancy} of the engine and a \textit{policy lag} of the trainer, and we follow both through training. The numerical discrepancy stays approximately constant, so it biases the updates by a nearly fixed amount, and centering the score at each prefix removes its leading bias. The policy lag is produced by the trainer's own updates: the bias it leaves enters the next update, which enlarges the next lag and closes a feedback loop. Importance weighting cuts this loop only while the lag ratio stays concentrated, and on Qwen3-1.7B it fails at the step where the effective sample size of the ratio falls below one half. From this account we derive the \textbf{discrepancy-lag correction} (DLC), which centers the numerical discrepancy, clips the lag ratio against the sampler's weights and resynchronizes the sampler when this effective sample size falls below one half. Under a stale sampler, DLC keeps the KL divergence from on-policy training within twice the KL between two independent on-policy trainings on Qwen3-0.6B and Qwen3-1.7B, and its resynchronization prevents collapses that the per-token corrections leave and that a KL measured on the sampler's prefixes does not detect.

0.33

Hybrid AI Patterns: Thirteen Ways to Couple Language Models with Rules, and How to Choose Among Them

cs.AI ·2026-10-29 ·Siegfried Handschuh, Götz-Henrik Wiegand, Lorena Raichle
hot direction
Abstract

Language models handle text well, and rule engines give conclusions with proofs. Systems that combine the two are usually described by example, and the same coupling turns up under different names in agent frameworks, guardrail toolkits and neuro-symbolic research. This document defines thirteen patterns for coupling a language model with an executable rule language, in the catalogue form introduced for object-oriented design: intent, motivation, applicability, participants, collaborations, consequences, implementation, known uses and related patterns. Four patterns concern what crosses the boundary between model and rules, five who controls whom, three how the model takes part in the reasoning, and one how knowledge is kept over time. Five invariants hold for every combination, and five failure patterns name the couplings that break them. We relate the catalogue to published pattern catalogues and taxonomies of hybrid and neuro-symbolic systems and show where they overlap with it. A decision guide leads from a problem to a pattern. It rests on our measurements, including the cases in which a model alone did as well as the hybrid or better, and it reports what each design cost.

0.33

Multistep and Multivariate Time Series Explainability with GradientSHAP: A Wind Power Case Study

cs.AI cs.LG ·2026-10-29 ·Robbie Southam, Zhong Fan, Ke Li
hot direction
Abstract · 24 pages. Submitted to Computational Intelligence

As the effects of climate change become increasingly apparent, the demand for renewable energy sources such as wind power continues to grow. However, renewables' variable and intermittent nature presents significant challenges in their integration into energy grids. Artificial intelligence (AI) has emerged as a powerful tool to address these issues. Still, its adoption in the energy sector has been hindered by the opacity of many AI models, which often function as black boxes. This paper introduces a carefully adapted GradientSHAP approach to provide interpretability for multivariate, multistep wind power forecasting -- a largely underexplored area. This capability is critical for day-ahead grid management and, to the best of our knowledge, this is the first study to apply GradientSHAP in the context of wind power forecasting. To address the inherent dimensionality, we developed novel visualizations that deliver both local and global explanations, uncovering intricate relationships between environmental factors, site-specific conditions, and temporal dynamics. While GradientSHAP is not an out-of-the-box solution, our findings demonstrate that, with thoughtful handling of SHAP values and domain expertise, it can offer valuable insights into model behaviour, supporting more informed and trustworthy decision-making. Nevertheless, its complexity poses challenges for practical implementation. Future research should focus on streamlining these interpretability methods to make them more actionable for routine deployment in critical energy applications.

0.33

Is It the Curriculum or Just the Easy Start? Benchmarking Unsupervised Environment Design on Quadrupeds

cs.LG cs.RO ·2026-10-29 ·Jakub Gilewicz, Dominik Galus, Piotr Wojdas
benchmarkcode/data
Abstract · 22 pages, 10 figures. Code, level space and evaluation suite: https://github.com/Solvro/ml-wojtek-ued

Unsupervised environment design (UED) methods such as PLR, ACCEL, SFL and PAIRED are compared mostly on 2D mazes, grid worlds and BipedalWalker, while terrain curricula for legged robots are hand-designed and evaluated against a few baselines. We present a benchmark that runs fourteen UED, domain-randomisation and robotics curricula under one protocol on five 3D quadrupeds in MuJoCo MJX: the closed-chain W01-TEK and the open-chain Unitree Go1 and Go2, Google Barkour vB and Boston Dynamics Spot. Levels are vectors of eight interpretable terrain parameters, and all curricula drive the same PPO student at matched environment-step budgets. The evaluation reports success on random and held-out structured terrains, interquartile means with stratified bootstrap intervals, capability envelopes and a level-by-level domination comparison. We compare methods within each robot and report, per level set, the top group of methods not significantly below the best. No method is in the top group on every robot and level set. Methods with easy or low-difficulty training levels, curated or not, fill the top groups on in-distribution random levels of Go1 and Barkour, and of W01-TEK apart from PAIRED+BiBC. The success-driven terrain curriculum from legged-robot practice is in a top group most often, domain randomisation on uniform levels in none, and PAIRED in none outside Spot, where the fourteen methods lie within 3.8 points on random levels of difficulty 0.25. On three robots the leader on in-distribution random levels is not the leader on held-out terrain. We collect the results into a per-robot guide to choosing a curriculum.

0.33

Algebraic World Model: From Exploration to Proofs

cs.AI cs.LO ·2026-10-29 ·Guancheng Pan, Qixin Tan, Yang Yuan, Andrew Chi-Chih Yao
code/data
Abstract · 29 pages, 9 figures, 14 tables. Project page: https://ephemeral6.github.io/awm-site/. Ancillary files contain the Lean 4 model of the runnin…

AI has made remarkable progress in mathematics, solving challenging problems with machine-checkable proofs in Lean, yet typically relying on predefined mathematical definitions and formal specifications. This raises a natural question: can it reason rigorously about an unknown world? We introduce the Algebraic World Model (AWM), a framework that turns interaction with unfamiliar environments into explicit mathematical models supporting formal reasoning. AWM explores the environment and develops executable Python hypotheses, continually testing and revising them against observations. It then structures these hypotheses into typed objects and shared rules, simplifies their structure through category-theory-guided transformations, and formalizes the resulting model in Lean 4 to prove reachability and impossibility. AWM can formally prove that a goal is unreachable, even beyond explored trajectories. On two impossible game levels, it constructs closed, kernel-checked Lean proofs excluding every legal winning sequence in all six runs, while Claude Opus 5.5 and GPT-6 Astra, equipped with coding tools, produce none in twelve baseline runs. Four of AWM's six initial bounds are correct, compared with one for Opus 5.5 and none for GPT-6 Astra; given only feedback that a cheaper win exists, AWM revises its rules and constructs new proofs, reaching six of six correct bounds. Moreover, AWM produces interpretable mathematical models whose rules can be inspected and tested step by step. Across all 25 public ARC-AGI-3 games, these models enable Lean to replay the winning trajectories of all 183 levels and generalize beyond observed play, predicting unseen transitions and detecting changes in game mechanics. With Claude Opus 5, AWM clears all levels with a score of 99.9, while its certified winning trajectories use only 43% of the human baseline's actions.

0.33

Uncertainty-Guided Cross-Modal Alignment for Diffusion-Based IoT Intrusion Detection

cs.LG ·2026-10-29 ·Mahdi Arab Loodaricheh, Anita Raja, Mohammad Hossein Manshaei
top venue
Abstract · 7 pages, 1 figure, 2 tables. Accepted at NeurIPS DynaFront 2026

Diffusion models evolve noisy states toward a learned data distribution, so their sampling behavior depends on the geometry of the representation in which denoising occurs. This dependence is especially important for multimodal data: independently trained encoders can map synchronized views to incompatible latent regions, while uniform alignment can let an uncertain view distort the shared geometry. To address this gap, we introduce a novel reliability-guided alignment rule that uses normalized VAE posterior variance to control directional cross-modal influence before diffusion training. We study this issue in minority-attack generation for imbalanced Internet-of-Things intrusion detection. Tabular flow attributes, a DeepInsight image, and a packet sequence are encoded by modality-specific variational autoencoders. Per-modality normalized VAE posterior variances provide bounded, sample-specific weighting signals in an asymmetric stop-gradient alignment objective. We use these quantities as heuristic encoder-derived reliability proxies rather than calibrated probabilistic uncertainties. The full modality-specific latent means obtained after alignment fine-tuning are standardized and modeled jointly by a class-conditional EDM; a second-order sampler then generates paired attack-conditioned latent blocks for the three decoders. In controlled experiments on TON-IoT, uncertainty-aware alignment obtains the highest mean precision, recall, density, and coverage among three diffusion variants. Augmenting only the training split yields ROC-AUC 0.9515, PR-AUC 0.8954, balanced accuracy 0.8395, and attack-class F1 0.7940 on an untouched real test set. Among the two variants with matched alignment fine-tuning, posterior-variance-based weighting improves both tabular generative fidelity and downstream detection relative to uniform weighting.

0.32

Violet Teaming in Practice: Lessons from a Pilot Expert Workshop on Governing Complex AI Systems

cs.AI cs.CY ·2026-10-29 ·Steph Batalis, Kait Connors, Ben C. Snyder, Vikram Venkatram, Josh Wentzel
Abstract · 9 pages, 2 figures

Violet teaming has been proposed as a method to integrate technical, policy, and societal considerations into the evaluation of emerging technologies. Here, we document a pilot expert workshop used to explore violet teaming as a governance methodology for biodefense-relevant AI capabilities and describe observations from the exercise. The workshop used a hypothetical technology scenario across three sequential stages - capability development, misuse potential, and implementation - while incorporating technical, security, operational, and policy perspectives. The sequential format provided an integrated way to examine potential benefits, misuse risks, and real-world implementation challenges, rather than considering these issues in isolation. The exercise surfaced non-obvious misuse pathways, expanded potential interventions beyond purely technical safeguards, and highlighted operational and policy constraints that could affect implementation. This report summarizes the workshop design, outputs, and lessons learned for applying violet teaming to the evaluation of complex, dual-use technologies.

0.32

EnvOS: Auditing Computer-Use Agents When the World Moves On : Phenomenon-Based Task Construction, Certification, and Evidence-Based Evaluation

cs.AI cs.HC ·2026-10-29 ·Ankush Kadu, Athira Gopal, Niharika Singh, Arpit Jain, Aswanth Krishnan
top venue
Abstract · 74 pages, 7 figures. Accepted at NeurIPS 2026 Workshops (AABA4ET and CL4FMAgents)

Computer-use agents built on large language models (LLMs) work in software that can change while they act: an appointment moves, a policy is revised, a record is updated. Testing agents under such changes is harder than it looks. When an agent fails, the cause may be the agent, but it may also be a change that did not fire as intended, an update the agent never received, or a grader that scored the episode wrongly. We introduce EnvOS, a framework for building and certifying environments that change mid-task. Before any agent runs, EnvOS checks that each change fires correctly, that the task remains solvable, and that the grader separates correct from incorrect behavior; after each run, separate code re-checks the verdict from archived evidence. We apply it to stale information, where a value the agent has already seen changes before the agent acts on it. In clinic booking, Claude Opus 5 succeeds on 0/16 tasks when nothing extra is shown, 2/16 when reminded to re-check the source, and 15/16 when the updated value is shown; the gain over the reminder holds after correcting for all tests (p = 0.00293). In insurance-claim approval, Opus 5 re-checks on its own and succeeds on 10/10, whereas Claude Sonnet 5 rises from 1/10 to 9/10 when the current policy is shown again, a gain significant in its pre-registered test but not after correcting for all tests. Whether an agent caught a change thus varied with the model and the interface. We also stress-test the grading itself: on corrupted episodes designed blind from the rules, the EnvOS pipeline accepted none of 26 as clean, and on an externally generated set it caught 25 of 39, against 4 for simple log checks. Because every EnvOS environment is generated from a seed and graded by an audited automatic verifier, the same environments are natural candidates for training agents with reinforcement learning from verifiable rewards.

0.32

Dynamic Powergrid Optimization with FPGA-based Neuromorphic Higher-order Ising Machines

cs.NE ·2026-10-29 ·Saptarshi Maiti, Shantanu Chakrabartty, Chetan Singh Thakur
benchmark
Abstract · This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may…

The advent of large, high-density FPGAs such as the AMD Versal enables efficient hardware implementation of neuromorphic Ising machines, unlocking unprecedented performance for complex combinatorial optimizations. A critical application for this computing paradigm is dynamic powergrid optimization, specifically Distribution Network Reconfiguration (DNR), which seeks to mitigate severe resistive losses and reduce electricity costs by optimizing network switch configurations in real time. This paper presents an FPGA-based neuromorphic higher-order Ising machine for high-speed, adaptive DNR. By directly mapping DNR onto higher-order Ising interactions via a polynomial unconstrained binary optimization (PUBO) formulation, the proposed architecture requires only 179 spins for the IEEE 33-bus network, which represents a 6 times reduction compared to standard quadratic formulations. Implemented on an AMD Versal Premium VP1802 FPGA, the architecture executes parallel spin-group updates every 20 ns and solves 33-, 51-, and 69-bus benchmark networks with median latencies of 1.6-29.3 ms, yielding a 7-58 times speedup over Gurobi to the first optimal solution. Using the solver with expanded network-switching flexibility reduces power loss by 80.71 kW (57.83%), while adaptive reconfiguration every 10 s under time-varying demand saves 131.16 kWh/day over static daily reconfiguration, highlighting the potential of higher-order Ising hardware for fast, adaptive smart grid management.

0.32

Coupled Guidance for Flow Matching

cs.LG cs.RO ·2026-10-29 ·Mateo Clémente, Léo Maxime Brunswic, Rui Heng Yang, Amir Rasouli
top venuebenchmark
Abstract · Accepted at NeurIPS 2026. 25 pages

Guidance methods are essential tools for steering the generation process in diffusion models, improving the generalizability of stochastic robotic controllers and enabling high-quality controlled image synthesis. Despite strong parallels between flow-matching and diffusion, integrating guidance into flow-matching models remains quite challenging. Recent work has proposed general guidance frameworks for flow-matching, however, they are direct adaptations of diffusion guidance methods and can prove unstable. In this work, we conduct a theoretical analysis of guidance in flow-matching and show that a naive translation of diffusion-based guidance to flow-matching breaks the coupling distribution used during training, resulting in weaker control over sampling accuracy. Building on this insight, we propose Coupled Guidance, a flow-matching-specific guidance mechanism derived from a geometric perspective that leverages the coupling structure inherent to flow-matching training. To validate the effectiveness of our approach, we conduct extensive experimentation on a custom robotic collision avoidance task using ManiSkill3, four robotic dexterous manipulation tasks from the Adroit benchmark, and three image generation tasks, namely inpainting, super-resolution, and deblurring, using the ImageNet-128 and CelebA-HQ-256 datasets. Our findings show that our framework improves sample quality and guidance stability across all tasks, without imposing any additional overhead.

0.32

Bridging Semantics and Structure: A Theoretically Grounded Framework for Link Prediction on TAGs

cs.AI ·2026-10-29 ·Chen Shao, Ruirui Wang, Tobias Kafer, Danai Koutra
publishedbenchmark
Abstract · 33 pages. Accepted for publication in ACM Transactions on Knowledge Discovery from Data (TKDD)

Link prediction on text-attributed graphs (TAGs) is a fundamental problem with applications in knowledge discovery, recommendation and beyond. Recent progress has highlighted the promise of combining Graph Neural Networks (GNNs) with Language Models (LMs) in order to leverage the complementary structural and semantic signals. However, due to the distinct inductive biases of GNNs and LMs, the design of an optimal fusion strategy remains an open research question. In this work, we present Bridge-LP, an intuitive, simple, efficient and theoretically grounded framework that unifies graph and language representations. It is built around three theoretically-justified design components: (D1) jumping connection of textual proximity, which enables robust LM training; (D2) compressed embedding, which encodes neighborhood information long exploited by classical link prediction heuristics, thereby enriching structural signals in a learnable way; and (D3) cache embedding training strategy, which reduces computational complexity. Together, these mechanisms establish an end-to-end cross-modal framework that scales to large graphs while capturing richer graph-text interactions. We conduct an extensive empirical evaluation, comparing against 40 existing models across 9 diverse datasets. Experiments show that Bridge-LP consistently outperforms state-of-the-art approaches, delivering up to a 43.06% improvement in Mean Reciprocal Rank (MRR) and setting a new benchmark for language-graph integration.

0.32

Unsupervised Feature Selection Using Local Intrinsic Dimensionality -- Extended Version

cs.LG ·2026-10-29 ·Oussama Chelly, Michael E. Houle, Ken-ichi Kawarabayashi, Muhammad Rajabinasab, Arthur Zimek
benchmark
Abstract · Preprint submitted to Elsevier Pattern Recognition

High dimensionality degrades both the efficiency and the effectiveness of learning algorithms, and in many applications no class labels are available to guide dimensionality reduction. This paper proposes LIDFS, an unsupervised feature selection method based on Local Intrinsic Dimensionality (LID). The method builds on the formal equivalence between LID and the indiscriminability of distance measures. The data matrix is transposed so that each feature is treated as a point, and the LID of each feature is estimated by maximum likelihood from the distances to its $k$ nearest neighboring features. Features with lower LID exhibit greater local discriminability and receive higher importance scores. A refinement step then retains one representative per feature neighborhood and demotes locally redundant features. The result is a complete ranking from which any number of features can be selected without rerunning the procedure. LIDFS does not use any information about labels or the number of classes, and runs in $\mathcal{O}(d^2 n)$ time for $n$ instances and $d$ features. LIDFS is compared with classic and state-of-the-art methods and a random baseline on 23 high-dimensional datasets, the classic benchmark collection. The evaluation uses classification accuracy and AUC with five downstream classifiers, as well as clustering accuracy and normalized mutual information, over both full and aggressive selection ranges. LIDFS ranks first in the supervised evaluations and first or tied second in the clustering-based evaluations. It belongs to the top statistical group together with methods that require the number of classes, and it runs faster than its closest competitor. Several recent methods, by contrast, fail to outperform the random baseline. A sensitivity analysis shows that small neighborhood sizes yield the best results.

0.31

Cheap Retrieval and Observation Credit Assignment from One Quantity: The Value of Information

cs.AI cs.LG ·2026-10-29 ·Daniel Strauss, Nikolaos Sotirakis, Zhenshan Bing, Alois Knoll
positiontheory
Abstract · 11 pages, 1 figure. Accepted at the CoRL 2026 Workshop on Memory for Robot Foundation Models (https://corl2026-memory.github.io/)

Memory in reinforcement learning raises two problems. Attention finds the relevant item by scoring the whole store, which is costly for robots under latency and energy limits. When a retrieved memory pays off, credit for acquiring it must cross the whole sequence. We argue that both are governed by one quantity, the value of information, and propose to estimate it from archived rollouts with a learned world model. We obtain that, when read at the observation, it yields a provable lower bound on the optimal value of the acquiring action, which floors its TD target and attributes this floor to the observation that earned it. Read at the decision, it defines the optimal retrieval set. A pilot retrieval system, to be trained and improved with this signal, solves a memory task with reduced context.

0.31

Omniprediction for All Mixable Proper Losses and Beyond

cs.LG cs.DSstat.ML ·2026-10-29 ·Junyan Liu, Haipeng Luo, Lillian J. Ratliff
Abstract

Omniprediction concerns the task of making predictions that are simultaneously accurate with respect to different losses. Earlier results by [CV09] show that in an online learning setting, $O(\log(|\mathcal{F}||\mathcal{L}|))$ regret is achievable simultaneously against every predictor in a finite predictor class $\mathcal{F}$ and for every loss in a finite class $\mathcal{L}$ of mixable proper losses, while recent advances [OKK25, FP26] show that $\widetilde{O}(\sqrt{T\log(|\mathcal{F}||\mathcal{L}|)})$ regret is achievable simultaneously for every loss in a finite class $\mathcal{L}$ of bounded losses. In this work, we ask whether sublinear regret remains possible when $\mathcal{L}$ is an infinite class of proper losses. Our first result gives a negative answer in general: even when $\mathcal{F}$ contains only two linear predictors, it is impossible to achieve sublinear regret simultaneously for all bounded proper losses. We then show that $\widetilde{O}(T^{1/3}(\log|\mathcal{F}|)^{2/3})$ regret is achievable simultaneously for \textit{all} mixable proper losses and that this bound is optimal up to logarithmic factors. This optimal rate stands in stark contrast to the constant (in $T$) regret achievable for a finite class of mixable proper losses. Finally, we extend our algorithm to obtain a best-of-all-worlds result: in addition to the new regret guarantee for all mixable proper losses, the resulting algorithm simultaneously achieves the existing guarantees for finite loss classes. With appropriate choices of these finite classes, our algorithm simultaneously achieves constant regret for squared loss and log loss, $\widetilde{O}(T^{1/3}(\log|\mathcal{F}|)^{2/3})$ regret for all mixable proper losses, and $\widetilde{O}(\sqrt{T\log|\mathcal{F}|})$ regret for $0$-$1$ loss, all Lipschitz losses, all GLM losses, and other structured families.

0.31

Conformal Annotation with Randomized Expert Review: Application to Online Cell-Type Labeling

cs.LG stat.MEstat.APstat.ML ·2026-10-29 ·Xuhe Zhang, Zhi Huang, Ying Jin
theory
Abstract · 48 pages, 17 figures

Artificial intelligence (AI) models can accelerate cell-type annotation in digital pathology, but safely scaling their use requires determining, with error control, which model-generated labels can be accepted without expert review. This is challenging in evolving workflows because cell morphology and type composition may change across tissue regions and slides, making one-time calibration unreliable. We propose Conformal Annotation with Randomized Expert review under Distribution shift (CARED), a distribution-free framework that leverages randomized expert audit to achieve long-run error control for AI-assisted annotation under distribution shift. For each incoming batch, CARED uniformly selects a subset of cases for expert review and uses the acquired labels to calibrate later decisions. Among unreviewed cases, CARED determines a subset of them to safely skip expert labeling, while suggesting candidate labels for the others for downstream resolution. We prove that this workflow provides long-run control of cumulative error over both the full stream and the cells outside of the audit sample, without stationarity or exchangeability assumptions. We further enhance the framework by introducing class-conditional guarantees to ensure error control within important but rare cell types, together with a utility module to target more automatic resolutions, smaller candidate sets, and greater biological coherence, without compromising validity. Experiments on pancreas, lung, and ovary annotation streams, together with a stream constructed from real expert annotations, show that CARED controls the cumulative errors while a substantial fraction of cells (up to 70%) bypass expert review.

0.31

Integrated Feature Selection for classification and Representation Learning with Group-Lasso Siamese Networks

cs.AI ·2026-10-29 ·Nisargo Ranjan Pal, Pradeep Singh
benchmark
Abstract

The Least Absolute Shrinkage and Selection Operator (Lasso) has been widely employed for feature selection in linear models and has also been extended to multi-layer perceptron (MLP) networks for both feature and group feature selection. Group-Lasso has also been used in feature selection using feed-forward neural networks. To the best of our knowledge, this is the first study to incorporate the Group-Lasso principle into a Siamese network architecture for feature selection. Unlike existing approaches that primarily optimize classification accuracy, the proposed method aims to identify a compact subset of features that simultaneously preserves the intrinsic representation of the input data (the goal of Siamese network) and is expected to achieve high classification performance. Two variants of the proposed framework are developed: one based on the Euclidean distance and the other on a modified Jaccard similarity measure for real-valued vectors. The latter is specifically designed to improve feature selection for high?dimensional sparse data. The effectiveness of the proposed approach is demonstrated on 12 benchmark datasets with feature dimensions ranging from 6 to 7070 using eight different classifiers. Experimental results demonstrate that the proposed Siamese Lasso framework consistently identifies compact and discriminative feature subsets maintaining competitive or superior classification performance, although the loss function used explicitly did not consider any regularizer to learn classification. In addition, the quality of the embeddings produced by these networks is assessed using manifold learning techniques such as UMAP and t-SNE.

0.31

RetroChessBench: A Benchmark for Reconstructing and Proving Chess Histories

cs.LG cs.AI ·2026-10-29 ·Tidor-Vlad Pricope
benchmark
Abstract · 22 pages, 2 figures. Benchmark and RL environment: https://app.poindexterlabs.net/docs/retro-chess

Frontier language models such as GPT 6 Astra have contributed to research mathematics, demonstrating substantial reasoning capabilities. Yet serious reasoning failures persist even in tasks governed by explicit rules. We demonstrate this in retrograde chess which is a playground for pure reasoning, where solvers reconstruct legal histories and distinguish possible events from forced ones. We introduce RetroChessBench, a 12-task benchmark covering exact-length constructions, unique endings, and exhaustive answer sets, with references, rubrics, and a terminal-reward environment. In two trials per task and condition, GPT 6 Astra High fails to produce the required final answer in 22/24 attempts without tools and 18/24 with code execution; mean terminal RL rewards are 14.75% and 27.19%, respectively. The examples identify faulty deductions: failed responses reject legal histories and treat temporary obstructions as permanent. Online retrograde archives often provide puzzles and final move sequences without detailed derivations. We hypothesize that limited exposure to these missing explanations in web-derived training data contributes to the observed reasoning failures. This questions how far frontier mathematical reasoning really generalizes beyond familiar solution patterns. Claims of general intelligence require evidence that models can transfer learnt principles to areas with few worked examples.

0.31

Hidden Phase Transition in Single-Polymer Stretching under Elongational Flow

cs.LG cond-mat.softphysics.comp-ph ·2026-10-29 ·Aiqing Zhu, Fushuai Wang, Eleonore Vissol-Gaudin, Qianxiao Li, Beatrice W. Soh
Abstract

Using machine learning, we identify a second transition in polymer chain stretching driven by planar elongational flow, beyond the well-known coil-stretch transition. We achieve this by developing a conditional extension of Universal and Identifiable Dissipative Dynamics (UIDD) that jointly learns reduced stochastic dynamics from microscopic trajectories across flow strengths and yields the corresponding stationary nonequilibrium free energy landscapes. As the flow strength increases, the evolving landscape exhibits the expected coil-stretch transition, and then a previously unrecognized dynamical transition. In the latter, the stretched states persist, but the distribution along a folding coordinate becomes bimodal when conditioned on small end-to-end separation. Large-deviation analysis of a simplified analytical chain model reproduces both transitions and shows that they are phase transitions in the long-chain limit. Single-molecule DNA experiments provide qualitative support for the dynamical transition through the predicted reversal of the folding response.

0.31

From extreme weather to agricultural consequences with agentic AI

cs.AI physics.ao-ph ·2026-10-29 ·Minghan Yu, Mengyu Dai, Yixin Wen, Haizhao Yang
Abstract · 34 pages, 4 figures, 6 tables

Extreme weather is forecast days ahead, yet its consequences for agriculture remain hard to predict: a forecast says where a storm will strike, not which crops will suffer or how much is at risk, and the records that could tell span many sources. We present WxCrop, the first agent that judges agricultural hazards from physically interpretable weather indices examined by causal analysis, and the first that combines weather and agricultural information to predict crop loss reliably for past and forecast dates. We demonstrate it in Florida on hurricanes, freezes and dry heat, with a loss model calibrated on insurance records for each, and the same procedure extends to other hazards and areas. Three checks show reliability: the hurricane index detects recorded cyclone days with an AUC of 0.98, recorded losses in held-out years are reproduced to within 2%, and independent production records confirm the county rankings (rank correlation 0.63). Users ask the agent for each crop group's loss as a fraction of expected season value, every number sourced, so that governments, growers and scientists respond to agricultural disasters quickly.

0.31

Event-SAMA: Language-Grounded Video Object Segmentation on Neuromorphic Event Streams

cs.LG cs.AIcs.CLcs.CV ·2026-10-29 ·Sergei Fedorchenko, Max Zehnder, Nikola Zubić, Rong Zou, Davide Scaramuzza
top venue
Abstract · 15 pages

Event cameras offer microsecond temporal resolution, high dynamic range and no motion blur, but referring video object segmentation (RVOS) lacks event-stream datasets and models. We introduce Event-SAMA-239K, a synthetic-event conversion of the SAMA-239K grounded video-chat corpus and all additional SAMA training data: 30,650 video clips and 392,162 images. Optical-flow, FILM or saccade upsampling followed by a linear event generator preserves all mask, referring-expression and dialogue annotations. We also introduce Event-SAMA, adapting SAMA-4B with a two-channel polarity patch embedding initialised from pretrained RGB weights and a zero-initialised event adapter. End-to-end training yields 24.37 $\mathcal{J}\&\mathcal{F}$ on MeViS valid_u and 22.33 $\mathcal{J}\&\mathcal{F}$ on the converted ReVOS valid subset, establishing the first event-native RVOS baseline (RGB SAMA-4B: 55.4 on MeViS). Frozen RGB SAMA-4B with pseudo-RGB renderings or E2VID reconstructions of the same events scores 13.23 and 15.94 $\mathcal{J}\&\mathcal{F}$. Without retraining, Event-SAMA works on real event-camera recordings when the accumulation window matches their event density. A seed-controlled ablation with a measured noise floor of 0.47 $\mathcal{J}\&\mathcal{F}$ finds gains from diverse training sources over MeViS alone (+4.99), RGB-mean event-stem initialisation (+4.33), and a trainable SAM2 mask decoder (+2.76); the event adapter, partial ViT unfreezing and event polarity are within noise at this budget. An earlier, partially converted corpus snapshot trained with the identical recipe scores 1.38 $\mathcal{J}\&\mathcal{F}$ higher on MeViS and 1.12 higher on ReVOS. Both are single runs that also differ in schedule length, so we draw no conclusion about corpus scale.

0.31

On the Fundamental Limits of Prefill-Based Decoding Length Prediction

cs.AI cs.CLcs.IT ·2026-10-29 ·Haowen Shi, Xuzhe Yan, Fahao Chen, Peng Li
positiontheorycode/data
Abstract · 26 pages, 4 figures

Predicting how many tokens a large language model (LLM) will decode, using only prefill-phase hidden states, is a key primitive for inference scheduling. We show that current benchmarks conflate predictor quality with task hardness: when the target model, dataset, or sampling configuration changes, reported accuracy collapses even though the predictor is unchanged. We trace this instability to an overlooked error source---the inherent stochasticity of autoregressive decoding---and derive bounds that every prefill-side method must obey. Our analysis treats decoding as a stopped stochastic process. The per-step stopping probability of a response is not constant but ramps up over hundreds of steps, so the decode length inherits a central limit theorem from the many local decisions that determine where a response ends; this yields per-prompt normality with an explicit error term, together with an affine variance law $σ=αμ+b$ whose slope is the coefficient of variation of the number of response segments. Empirically, the normal law reproduces the per-prompt MAE and the central mass of the decode-length distribution to within a few percent, and the residual departures from normality (mild right skew, slight excess kurtosis) are the ones predicted by the sampling-noise correction. Under fixed model, data, and sampling conditions, a broad design space of predictors (MLP, EGTP, FiLM, soft-label, and Mixture-of-Experts variants) converges to similar performance close to the bound, confirming that \emph{system error dominates model error}. We argue for a benchmarking protocol that fixes the model, data, and sampling parameters and that reports relative-error metrics alongside absolute ones. Code and analysis tools are available at https://github.com/ShwStone/DecodingPredict.

0.31

TSFLab: A Self-Improving Agent-Agnostic Workbench for Time Series Forecasting Research

cs.LG cs.AI ·2026-10-29 ·ChengAo Shen, Yanjun Zhao, Haoxin Liu, Bonnie Chen, Miles Ma +5 more
code/data
Abstract · Code: https://github.com/D2I-Group/TSFLab

As publications and products on time series forecasting (TSF) proliferate, researchers face a growing burden of analyzing existing work. First, not all publications release code. For instance, only about 20% of arXiv TSF papers have a code link in the surveyed period. Second, released code also requires manual maintenance. Thus, TSF researchers must constantly adapt or implement existing works. However, independent implementations can differ in experimental settings and are difficult to maintain. We therefore propose to use AI agents for benchmarking TSF research and even supporting autoresearch, e.g., developing new TSF models. Yet directly applying existing coding agents, such as Claude Code and Codex, is suboptimal: prohibitive costs, limited scalability to large research projects (e.g., benchmarking), and lack of self-maintenance. To address these challenges, we propose TSFLab, a self-improving agent-agnostic "workbench" that equips compatible coding agents with shared resources and tools for low-cost TSF research and continuous maintenance of research resources. Our experiments cover reproduction, benchmarking, and new method development. In reproduction and benchmarking, TSFLab improves task completion by 20-163% and cuts fresh tokens per completed task by 11-66% across baselines. A pilot study of agent-driven method development also shows improved forecasting accuracy via TSFLab, with 32.8-73.6% fewer fresh tokens and 9.2-68.8% less development time than the other settings.

0.31

TabCUQ: Context-Uncertainty Quantification for Tabular Foundation Models

cs.LG stat.ML ·2026-10-29 ·Yuebei Shi, Molei Liu, Zijian Guo
benchmark
Abstract

Tabular foundation models use in-context learning to estimate conditional expected outcomes and average treatment effects for specific groups of individuals. Using these estimates to guide scientific decisions requires reliable uncertainty quantification. The challenge is to account simultaneously for two sources of uncertainty: limited-context bias and context variation arising from the small labeled sample supplied at inference. Predictive spread does not directly quantify both sources of uncertainty, and we observe that the ordinary bootstrap may substantially overestimate context variation, whereas subsampling may underestimate it. We introduce TabCUQ, a perturbation-based framework that injects noise into observed outcomes and applies the foundation model to the perturbed contexts. TabCUQ estimates context variation using wild-bootstrap dispersion and accounts for limited-context bias using the maximum deviation between perturbed and original estimates. For a kernel-regression analogue whose estimation error admits an exact decomposition into these two components, we show that our framework provides asymptotically valid pointwise coverage. In simulations, TabCUQ substantially improves coverage, with average interval widths typically about twice those of an oracle bias-aware benchmark. Across seven real-world tasks, it also improves coverage of reference functions learned from much larger samples.

0.30

DolphinRouting: Cluster-Level Binary Routing for Disaggregated LLM Inference Serving

cs.LG ·2026-10-29 ·Yishuo Yuan, Shaoyu Yang, Chunrong Fang
hot direction
Abstract

Prefill-decode (PD) disaggregation isolates the two phases of large language model serving, but request-level path selection must account for prefill work, cache reuse, and decode demand. We present DolphinRouting, which learns this choice from end-to-end SLO feedback under a given P/D resource allocation. Requests are clustered by raw prompt length, cache-adjusted input lengths on both worker pools, and output-length information. Offline policy search learns one binary action per cluster. Online feature extraction, cluster assignment, and table lookup select local or remote prefill. We evaluate DolphinRouting on a real PD-disaggregated cluster using three public benchmarks, a mixed workload, and real data. Across the mixed workload and real data, DolphinRouting with $k$-means reduces P90 TTFT by 5.7--7.9\% and P90 TPOT by 2.6--10.1\% relative to Dynamo.

0.30

Does Federated Training Degrade Explainability? A Comparative Analysis of Grad-CAM Faithfulness in Centralized vs. Federated Deepfake Detection

cs.AI cs.CRcs.CVcs.LG ·2026-10-29 ·Muhammad Uzair Shahid, Muhammad Zabil Mehboob
code/data
Abstract · 10 Pages, 3 figures, Muhammad Uzair Shahid and Muhammad Zabil Mehboob contributed equally to this work

Automated detection has become the essential safeguard against realistic synthetic media, but deploying it at scale requires training on sensitive data from multiple organizations, putting privacy and the transparency demanded by forensic review in direct conflict. Federated learning resolves the privacy side by sharing only model updates rather than raw data, but whether this comes at the cost of trustworthy gradient-based explanations, which analysts rely on to decide whether or not to trust a flagged clip, has not been studied. We present the systematic study of Grad-CAM explanation faithfulness under federated deepfake detection, evaluating an EfficientNet-B4 detector on FaceForensics++ across five federated clients under varying degrees of data heterogeneity. We find that the relationship between federation and explanation quality is non-monotonic: at near-IID (Independent and Identically Distributed Data) and moderate heterogeneity, federated models match or exceed the faithfulness of a centralized baseline, but at extreme heterogeneity, explanations break down entirely, becoming spatially incoherent and, on average, causally misleading. This degradation is not a general cost of federated training but a threshold effect tied to client heterogeneity, and we offer practical guidance for identifying when privacy-preserving deepfake detection can be deployed safely and when its explanations can no longer be trusted. Code and data partitions are available for reproducibility.

0.30

A Deep Learning Framework for Quantile Demand Forecasting in E-commerce with Doubly Censored Data

cs.LG ·2026-10-29 ·Ziwen Ye, Jiawen Wei, Teng Song, Chen Chen, Guangrui Ma
top venue
Abstract

In recent years, a growing number of online retail companies have embraced deep learning models for demand forecasting, leveraging their scalability in processing large-scale datasets. Accurate demand forecasts enable improved planning and allow operational flexibility to adjust sales accordingly. Moreover, quantile estimation has gained widespread application as it offers valuable uncertainty quantification, thereby enhancing the robustness of decision-making processes. However, limited research has addressed the effective quantification of uncertainty when target labels are ambiguous. In retail, actual sales can be capped by stock and boosted by temporary promotions, whereas the target is underlying demand. Such scenarios are characterized by double censoring and no deep-learning pipeline yet exists for such data. In this paper, we introduce a straightforward framework for learning quantiles from doubly-censored observations and propose a tailored evaluation metric. Our approach is supported by theoretical guarantees, and we empirically demonstrate that our metric correlates well with the oracle calibration measure. Experimental results show that our estimator surpasses existing baselines. Finally, inventory simulations and real-world applications confirm the practical utility of our method in retail operations.

0.30

Zero-Complement Extrapolation: Parameter Recovery and Classification in Severely Imbalanced Logistic Regression under Rank-Deficient Regimes

cs.LG stat.MLstat.ME ·2026-10-29 ·Chongsun Park
benchmarktheory
Abstract · Submitted to a peer-reviewed journal; 16 pages, 1 figure, 4 tables

Parameter estimation and classification in binary logistic regression deteriorate severely when the number of minority class observations ($n_1$) is smaller than the feature dimension ($p$). Under such rank-deficient rarity ($n_1 < p$), maximum likelihood estimators suffer from quasi-complete separation, penalized likelihood methods exhibit severe intercept bias, and convex minority oversampling techniques such as SMOTE collapse into an $(n_1 - 1)$-dimensional affine subspace. In this paper, we propose a novel \textit{Self-Calibrated Zero-Complement} estimator that shifts the paradigm from minority interpolation to majority fringe extrapolation. Grounded in Stein's Lemma, we prove that the conditional expectation shift of the abundant majority class ($Y=0$) induced by logistic truncation uniquely identifies the true hyperplane normal $\boldsymbolβ$ after Mahalanobis whitening. We further establish that extrapolating synthetic minority instances outward from the high-quantile density fringe of the majority distribution along the conjugate direction preserves full-rank covariance regularity ($\mathrm{rank} = p$). To accommodate both semi-supervised environments (with an unlabeled background mean) and strictly supervised subsamples, we introduce a hybrid estimator incorporating the intrinsic third-moment skewness of the majority class alongside a location-shift-invariant zero-boundary intercept calibration. Extensive Monte Carlo simulations and real-world benchmark applications (Wisconsin Diagnostic Breast Cancer and Banknote Authentication) under extreme rarity ($n_1 \in \{3, 5\}$) demonstrate that our approach prevents parameter divergence, recovers the oracle hyperplane with high fidelity, and substantially outperforms Firth's regression and SMOTE in F1-score and estimation stability.

0.30

One Vector or Many? Clustered Optimal-Transport Steering under Quantization

cs.LG cs.AIcs.CL ·2026-10-29 ·Illia Makarov
theoryunusual
Abstract · 34 pages, 12 figures, 14 tables

Activation steering changes what a language model generates by editing its hidden states at inference time. Its most common form adds one vector, the difference of class means (DiM), to every hidden state. CHaRS (concept heterogeneity-aware representation steering) uses optimal transport instead: an entropic plan pairs the clusters of the two classes, and each hidden state moves along a mixture of the cluster displacements, weighted toward its nearest clusters. We study this family of clustered optimal-transport maps for toxicity mitigation in Gemma-2-2B and Granite-3.1-2B, and ask what the plan adds to DiM and whether that survives INT8 and NF4 quantization. We prove an identity: whatever the plan, the displacements averaged with the cluster sizes as weights equal DiM, so a clustered map can differ from DiM only for tokens whose routing weights depart from the cluster sizes. In the CHaRS protocol they stay close, and CHaRS is statistically equivalent to Mean-AcT, which adds DiM at every layer, in all eight pre-specified equivalence tests. A variant fitted on individual tokens, which routes each token mainly to its nearest clusters, leaves 37-39% less toxicity than DiM on Gemma and 51-61% less on Granite at steering strength 3 (toxic-bert), where DiM already removes 76-85% of it. Two other toxicity classifiers confirm the gain on Granite, but on Gemma only one does. At equal strength the variant has somewhat higher perplexity, but at matched toxicity it is slightly more fluent than DiM. Yet a nearly constant map from the same clusters also leaves about 30% less toxicity, so the role of treating tokens differently remains open. Under INT8 and NF4 the variant still leaves less toxicity than DiM, and transferring it from full precision to the NF4 model instead of refitting it adds no significant toxicity after correction for multiple tests.

0.30

Optimization Techniques for Solving the Minimum Red--Blue Separating Set Problem with an Application to Human--AI Text Separation

cs.AI ·2026-10-29 ·Marko Djukanović, Staša Vujičić Stanković, Žiga Zebec, Christian Blum
unusual
Abstract · 36 pages, 5 figures

We study the so-called Minimum Red''Blue Separating Set Problem (MRBSSP) which arises from computational geometry. In particular, we propose a compact Set-Cover ILP formulation, safe preprocessing rules, a Greedy Randomized Adaptive Search Procedure (GRASP), and two exact approaches: one is based on Logic-Based Benders Decomposition (LBBD) approach, and the second represents a novel Branch-and-Benders-Cut (BBC) approach. These two exact approaches are compared to ILP--alongside GRASP and the Variable Neighborhood Search (VNS) from literature. Experiments on 250 random graphs with up to 500 vertices reveal complementary strengths: ILP yields the best feasible solutions, LBBD provides the strongest lower bounds and smallest mean optimality gap, and BBC certifies optimality most frequently. We further introduce an NLP application distinguishing human- and AI-generated texts through vocabulary-based graph features. On 42 instances, BBC achieves the smallest mean optimality gap, certifying 16 optima, while GRASP outperforms VNS. Structural analysis of optimal vocabulary separators identifies 113 mandatory and 151 replaceable features across 272 tests, highlighting substantial variability among optimal solutions. Overall, our results demonstrate the effectiveness of complementary optimization strategies and the interpretability of optimal vocabulary-based discrimination.

0.30

Analyzing Condensation on Patterned Surfaces via Deep Learning

cs.LG ·2026-10-29 ·Tae E. Cooper, Michael Gee, Sheila Whitman, Jemma Dengel, Anthony Katselas +3 more
unusual
Abstract · 13 pages, 4 figures

Dropwise condensation is key to power and water systems. State-of-the-art condensation surfaces are often inhomogeneous, either patterned by design or random defects. Inhomogeneous surface features dictate the rates of each stage of dropwise condensation-nucleation, growth, coalescence, and shedding. Characterizing these rates requires the precise detection of thousands of droplets. Yet, existing segmentation methods are unable to detect droplets on inhomogeneous surfaces. We developed TileSAM3, a universal droplet segmentation tool built on the Segment Anything Model 3 (SAM3). Hierarchical tiling with per-tile inference followed by post-processing extended SAM3's detection cap and raised the effective resolution of small droplets. We created two datasets: High Quality Droplet Dataset (N=365,826; HQDD) with homogeneous surfaces and Low Quality Droplet Dataset (N=63,577; LQDD) with patterned, defect-laden surfaces under adverse lighting. TileSAM3 closely reproduced thresholding annotations on HQDD, surpassed it over sixfold on LQDD (Average Precision 0.57 versus 0.088), and outperformed other deep learning models on all metrics evaluated. Combined with KD-tree tracking, TileSAM3 identified localized nucleation and coalescence in narrow active bands on a stiffness-patterned surface. TileSAM3 enables analysis of understudied inhomogeneous substrates with poor lighting and videography which facilitates faster, lower cost research of condensation materials.

0.30

OCCAMS: A Multi-Task Learning-Based Investigation of the CMB Power Spectra

cs.LG astro-ph.COastro-ph.IM ·2026-10-29 ·Ankit Mandal, Rahul Shah, Md. Wali Hossain, Somasri Sen, Anjan A Sen +1 more
unusual
Abstract · 13 pages, 4 figures, 4 tables, plus appendices. Comments are welcome

How many latent variables are needed to fit the cosmic microwave background (CMB) temperature and polarisation power spectra when a neural network is given no cosmological model, and to what extent do they map onto the standard cosmological parameters? We use multi-task learning with variational autoencoders trained on CLASS-generated spectra (TT, TE, EE) over $2\le\ell\le2500$. No cosmological model is built into the network, and the number of latent variables, $L$, is varied over $L=5$, $6$ and $7$. Their fits are compared to the Planck 2018 data using $\hatχ^2$, AIC, BIC and Bayesian evidence. Their correspondence with cosmological parameters is measured with normalised mutual information post-training. The loss weights are set within a single training run by uncertainty weighting, with the weight of the disentanglement term either fixed (semi-automated, SA) or learned (completely automated, CA). The reconstruction loss is split over multipole ranges in three configurations that separate the roles of temperature and polarisation and of the low- and high-$\ell$ ranges. In four of the six combinations of configuration and strategy, the fit improves from $L=5$ to $L=6$ and not beyond, and $L=6$ is the best, or close to the best, in five of the six cases. So, six latent variables are needed, as many as the free parameters of the standard, model-dependent approach. SA gives clearer disentanglement, while CA needs no a priori choice of the disentanglement weight and has the higher evidence at $L=5$ and $6$: the two strategies are points on a trade-off between interpretability and fit to the data. Notably, the latent variables for $τ$ and $A_s$ share almost no information in any model, consistent with large-scale polarisation separating the two, as in standard analyses. A deep-learning approach can thus give a compact, data-driven parametrisation of the CMB spectra.

0.30

Does Scaling Toward AGI Depend on What System-1 Lets System-2 Execute?

cs.AI ·2026-10-29 ·Qi Wang, Yun Qu, Lizhou Cai, Heming Zou, Haoang Chi +8 more
community pick
Abstract

Progress toward artificial general intelligence (AGI) has largely relied on scaling a resource-intensive executive, System-2, through larger models, more data and compute, longer reasoning, and more agents. We ask whether the next stage may come instead from co-designing it with System-1, a lightweight, semantically aware predictor that evaluates candidate interactions before execution, so that System-2 acts only on those worth their cost. Our answer is a qualified yes. We study \textit{conditional interaction}, a predict--gate--execute--update loop: System-1 estimates each candidate's utility, System-2 executes those admitted by an adaptive threshold, and the outcomes update System-1. The loop spans training, test-time reasoning, adaptive decision-making, and agentic tool use, with utility and cost defined by the application. Under ideal screening and equal execution costs, execution compute falls by a factor approaching $1/ρ_u$, where $ρ_u$ is the density of utility-relevant interactions (cost-weighted when costs differ); prediction errors, exploration, and distribution shift reduce the saving. We bound selection regret by predictor error and the utility margin. These results motivate our central hypothesis: $ρ_u$ is a measurable coordinate of scaling efficiency and makes informative compute a better description of scaling than total compute.

0.30

MulClip: A table-free activation function with monotonic and folding regimes for reservoir computing on microcontrollers

cs.LG cs.NE ·2026-10-29 ·Guren Matsumura, Masato Takemoto, Shinzo Koyama
Abstract · 15 pages (7 pages main text, 8 pages appendices), 2 figures

This letter proposes MulClip, a lightweight activation function for reservoir computing (RC) on microcontrollers (MCUs). In a ring reservoir, a representative sparse RC topology, multiply-accumulate operations are few, and evaluating tanh accounted for 75% of the state-update time on a Cortex-M4. MulClip introduces curvature into the response using only comparison, absolute value, multiplication, and subtraction, and places its clipping threshold not at the extremum but at the point where the response returns to zero after passing the extremum. As a result, a single function has both a monotonic regime and a nonmonotonic folding regime that increases the expressiveness of a single unit. Which regime is used is determined by the usual hyperparameter search, e.g., over the input scale, so no task-specific threshold tuning is required. The state-update time measured on hardware was 0.34 times that of tanh and shorter than that of any of the 11 evaluated tanh approximations. In PC-based evaluation, third-order MulClip achieved a 22% lower NMSE than tanh on NARMA-20, where the pre-activations remained in the monotonic regime, and on 8-bit parity it raised the accuracy from 0.76-0.78 for the monotonic-clip variant without folding to 0.97-0.98. For these two tasks, on-device inference yielded nearly the same values as on the PC.

0.30

Retaining Repair Hypotheses for Cost-Aware Execution

cs.AI ·2026-10-29 ·Jae Hun Hwang, Hong-In Won
Abstract · 17 pages, 5 figures; data and code in ancillary files

Several corrections of an executable model can fit the same observations while predicting different execution outcomes. We study a repair-to-execution interface that retains compatible finite guard repairs, conditions their weights on diagnostic and ordinary-execution observations, and uses their predictions to choose routes, priced tests, or a backup. On 24 synthetic two-stage devices, an equal mixture of a language-model-generated repair and a uniform belief over compatible repairs (the AI-weighted belief) lowers no-test cost by 2.14 percent relative to committing to the generated repair. Access charges can erase the benefit of diagnosis on isolated tasks, while repeated work can recover it under the AI-weighted belief. With normal-execution learning in every repeated-work arm, workload-valued testing costs 3.33 percent less than same-budget random selection under the AI-weighted belief. Weighting changes the comparison: uniform normal learning costs 540.75 less than AI-weighted normal learning on this cohort, with a device-paired interval excluding zero. Testing recovers part of that gap, but uniform normal learning still has a lower aggregate cost than AI-weighted workload-valued selection. Under uniform beliefs, information-gain selection has the lowest total among the compared selectors, while workload-valued testing costs more than normal learning. These results support retaining alternatives to a generated repair and evaluating diagnostic selection together with repair weighting and learning from execution.

0.30

Explainable and Adaptive Machine Learning for Fraud Risk Detection in Bangladesh's Digital Banking Ecosystem

cs.LG ·2026-10-29 ·Md Najmus Swaqeeb
publishedbenchmark
Abstract · 23 pages, 6 figures, 5 tables. Presented at the 11th Annual Banking Conference 2026, Bangladesh Institute of Bank Management (BIBM), Dhaka, …

The rapid expansion of digital banking and mobile financial services creates growing challenges for fraud detection, particularly under severe class imbalance and operational constraints. This study presents an explainable and adaptive machine-learning framework integrating chronological evaluation, leakage-conscious feature engineering, probability calibration, SHAP-based explainability, and adaptive risk prioritization. Experiments use the synthetic PaySim mobile-money benchmark and compare Logistic Regression, Random Forest, XGBoost, and LightGBM. XGBoost achieved the strongest fixed-threshold performance, with PR-AUC of 0.5658, ROC-AUC of 0.9068, and F1-score of 0.5369. Platt calibration reduced the test Brier score from 0.1141 to 0.0245, improving probability reliability. SHAP analysis identified temporal, destination-account, transaction-type, and amount features as important model drivers. An adaptive top-1% policy reduced false positives from 192 to 117, but recall declined from 0.4236 to 0.2122, demonstrating a clear alert-capacity trade-off. The results provide a reproducible methodological framework rather than evidence of deployment performance in Bangladesh, since PaySim is synthetic. Validation using governed real-world financial data is therefore an important direction for future research.

0.29

A Robust Approach to Belief Mismatch in Non-cooperative Dynamic Games

cs.MA eess.SY ·2026-10-29 ·Ryan Park, Li-huan Shen, Jesse Milzman, Negar Mehr, David Fridovich-Keil
Abstract

Game-theoretic models are increasingly used to describe multi-agent interactions in control and robotics applications. However, theoretical and computational complexity makes these models intractable in many dynamic, incomplete information scenarios. Recent work makes strong assumptions to reduce computational complexity: agents share sensor measurements with one another and maintain a common estimate of state uncertainty. Unfortunately, these assumptions enforce a degree of cooperation in generically noncooperative interactions. This paper takes the perspective of an ego agent seeking both computational tractability and robustness to unanticipated behavior of other agents due to mismatched state estimates from private---rather than shared---sensor measurements. To that end, we formulate the game theoretic planning problem in an augmented belief space with perturbed Bayesian dynamics that allows the ego to predict all agents' potentially divergent state estimates. These perturbations are chosen by a hypothetical adversary, ``nature,'' to reflect how non-ego agents' state estimates---based on measurements the ego agent will not see at runtime---may deviate from the ego's own belief. Ultimately, our approach can be understood in the spirit of classical H-infinity robust control, extended to noncooperative, stochastic dynamic games. We present a solver with polynomial-time iterations and conduct two experiments which show how robust belief space reasoning leads ego agents to proactively limit their future vulnerability to model misspecification of other agents' beliefs and actions, reducing performance degradation by over 48%.

0.29

Understanding axial attention in TSFMs through single location regression

cs.AI stat.MLcs.LG ·2026-10-29 ·Andrei-Tiberiu Pantea, Aymeric Dieuleveut
theory
Abstract

Recent advances in Time Series Foundation Models (TSFMs) increasingly rely on axial attention mechanisms, but despite their empirical success, the theoretical properties of axial attention remain largely unexplored. To bridge this gap, we extend the Single Location Regression framework introduced in Marion et al. [2025] to 2D structured data, in order to study sparse information settings where a target depends on a small number of tokens. We formulate an analytically tractable axial attention-type predictor and rigorously analyze its statistical properties and training dynamics. We prove that under a specific asymptotic regime and with knowledge of the underlying oracle parameters, our predictor achieves Bayes optimality, whereas a class of semi-linear simplifications strictly fails to do so, underlining the role of the inner non-linearity. We prove that for a specific temperature scaling, Projected Gradient Descent converges globally to the optimal oracle parameters. Extensive numerical experiments validate our theoretical findings and highlight the critical role of inverse temperature scheduling in empirical convergence.

0.29

Certified Behavioral Retention for Continual Adaptation with Held-Out Anchors

cs.LG stat.ML ·2026-10-29 ·Ahanaf Hasan Ariq, Abrar Shahid
published
Abstract · 11 pages, 1 figure, 3 tables. Accepted at the NeurIPS 2026 Workshop on Continual Learning in the Era of Foundation Models and Embodied Agent…

Continual adaptation requires deciding not only how to update a model, but whether to commit an update without exceeding a declared budget for changes to protected behavior. We introduce Certified Anchor-Based Adaptation (CABA), a candidate-agnostic post-update gate that reserves historical observations from candidate optimization, compares old and candidate policies on frozen representations, and selects the largest tested interpolation coefficient whose worst-anchor forward KL divergence remains within budget. The finite check requires neither action labels nor complete trajectories. We develop a theory-led certification framework that separates measured anchor retention from conditional off-anchor guarantees. Under explicit representation-coverage and Lipschitz assumptions, an accepted update satisfies a uniform behavioral-divergence bound on protected support. The analysis gives regularity bounds for affine softmax heads, probabilistic coverage and compression guarantees, interpolation monotonicity conditions, grid-resolution bounds, and trajectory-level and sequential-composition guarantees. In a controlled UCI Human Activity Recognition study used as a non-stationary sensor-policy proxy, CABA reduces mean forgetting from 1.01 percent for unconstrained adaptation to 0.00 percent at budget epsilon = 0.05, while reducing final average accuracy from 93.04 percent to 88.48 percent. The experiment illustrates a retention-plasticity trade-off and does not establish foundation-model or robotic performance. CABA provides an auditable retention check for candidate generators exposing comparable policy distributions and a declared update path, while deployment also requires separate evidence of current-task utility.

0.29

Fair Recommender Systems for Intersecting Groups

cs.LG cs.CYcs.GT ·2026-10-29 ·Christine Ling, Juba Ziani
Abstract

Recommendation systems that optimize for the average user can fail to capture the preferences of diverse demographic groups. We consider a dynamic recommendation system, where arriving users can belong to different demographic groups. Importantly, these groups can intersect, and a user may belong to several groups. We first show that over time, any group-blind policy (which excludes all current-user group information) recommending $K \geq 1$ items incurs linear total regret under mild conditions. In turn, we introduce three group-aware algorithms based on the classical Upper Confidence Bound algorithm (UCB): i) Intersection-Fitting, which runs UCB on each intersection, but has a sublinear group- and intersection-level regret guarantee with an exponential dependence in the number of groups; and ii) two group-level heuristics, Inner-Outer UCB and Group-Item UCB which allocate different arms to different groups, and use a sleeping expert meta-learner to choose which active group's recommendation an intersection should inherit at any given time step. We provide experimental results on two datasets, Goodreads and BoardGameGeek, and highlight that our heuristics can be preferable to the intersection-fitting algorithm when intersection-level preferences are closely related to group-level preferences, but that intersection-fitting performs better when different intersections within a same group prefer different items.

0.29

Dobb: Shard-Local Preconditioning in Block-Diagonal Eigenbases

cs.LG cs.DCmath.OC ·2026-10-29 ·Shi-Yuk Fong, Shuhua Yu, Jiawei Zhang
benchmarkhot direction
Abstract · 33 pages, 3 figures

Training language models whose parameters and optimizer state exceed a single node's memory requires sharding these tensors across devices. Under tensor parallelism (TP) or fully sharded data parallelism (FSDP), matrix-aware optimizers such as Muon introduce communication to perform global matrix operations across shards. We propose Dobb (Distributed Optimizer with Block-diagonal Eigenbases), a matrix-aware optimizer designed to compute updates locally within parameter shards. Dobb approximates the input-activation second moment with diagonal blocks and tracks a complete basis within each block. It projects momentum into these bases, applies sign normalization to the coefficients, and reconstructs the update. When each block lies within one shard under TP or FSDP, Dobb applies its matrix update locally, requiring no communication beyond data-parallel synchronization of the activation second moments. We benchmark Dobb against Dion, a distributed matrix-aware optimizer designed to retain the benefits of orthonormalized updates while reducing communication. On GPT-2 XL (1.56B parameters), Dobb with 16 blocks reaches a validation loss of 2.476, compared with 2.511 for Dion at rank fraction 1/16. Under FSDP on 32 H100 GPUs, Dobb reduces the amortized optimizer-call time from 124.1 ms for Dion to 9.2 ms.

0.29

Activation Probes Are Reliable With a Handful of Positive Examples

cs.LG ·2026-10-29 ·Riya Tyagi, Stefan Heimersheim
published
Abstract · Presented at the Mechanistic Interpretability workshop at NeurIPS 2025

Efforts to monitor advanced AI for rare misalignments face a data challenge: abundant aligned examples but only a handful of misaligned ones. We test activation probes in this "few vs. thousands" regime on spam and honesty detection tasks. For our tasks, training with many negative examples is on average more positive-sample-efficient than balanced training for small numbers (1-10) of positive samples. We also find that LLM upsampling can provide a performance boost equivalent to roughly doubling the number of real positive samples, though excessive upsampling hurts performance. Finally, we show a positive scaling trend, where larger models are more positive-sample-efficient to probe. Our findings suggest we should leverage the large number of negative samples available to amplify the signal from rare but critical misalignment examples.

0.29

Clean Sheets and Checkpoints: Evidence Complexity for Certifying Sequential Agent Risk

cs.LG cs.AI ·2026-10-29 ·Rohith Yanambaka Venkata, Don McBride, Tiago Sequeira, Md. Al Amin
Abstract · 26 pages, 3 figures, 3 tables

Dangerous-capability evaluations often test an AI agent from prepared states at ordered milestones. A clean sheet records no successes. What it certifies about a campaign of $B$ attempts depends on what a failed attempt erases. We model the task as a chain of $k$ milestones divided into checkpoint segments and ask how many zero audit responses certify, at level $α$, a completion probability of at most $ε$. For segments chosen before the audit, counting possible completion windows gives a certificate for any specified budget. The bound also covers adaptive deployment processes that follow the chain's transitions, provided their success probabilities after every complete history are capped by the corresponding audit success probabilities. The audit does not test this transport assumption. In rare-risk phase interiors, each segment is either rarely completed within the budget (active) or nearly always completed (free). The minimum expected number of zero responses is asymptotic to $\log(1/α)\,(B^a/(a!\,ε))^{1/K_A}$, where the $a$ active segments contain $K_A$ milestones and are those at least as long as a budget-dependent threshold. The certificate attains this constant. Every asymptotically optimal evaluator must divide its expected response count equally among active milestones in the limit, with a vanishing share elsewhere. Where a single segment passes from active to free, each of its milestones receives a smaller limiting share than an active milestone; a crossover-aware certificate attains the optimal constant. For selected segments with known or unqueried prefix steps, counting failed cycles can sharpen the certificate.

0.29

Multi-output LSTM soft sensors for industrial anaerobic co-digestion using LSTM-beta-VAE generative data augmentation

cs.LG cs.AIeess.SY ·2026-10-29 ·Rohit Murali, Benaissa Dekhici, Tao Chen, Dongda Zhang, Michael Short
Abstract · 33 pages

Anaerobic digestion (AD) requires close monitoring of key process variables, including pH, the ratio of volatile organic acids to alkaline buffer capacity and biogas production, to maintain biological stability and prevent process failure. While mechanistic frameworks such as the Anaerobic Digestion Model No. 1 are highly parameterised, computationally expensive, and difficult to calibrate for real-time applications, data-driven Long Short-Term Memory (LSTM) networks offer a potentially more adaptable alternative. However, current data-driven approaches in AD are predominantly limited to single-output architectures that focus solely on production yields, as opposed to simultaneous prediction of vital stability indicators. Furthermore, industrial AD datasets are typically sparse and skewed toward steady-state operations. Baseline evaluations demonstrated this scarcity prevents standard algorithms (Random Forest, XGBoost) and unaugmented LSTMs from recognising the transition to unstable operations, resulting in poor estimation of stability indicators. To overcome these challenges, this study proposes a novel generative data augmentation framework using an LSTM-beta-Variational Autoencoder coupled with a multi-output LSTM (MO-LSTM) soft sensor. By synthesising physically valid sequences of these transitions to unstable operations from one year of full-scale plant data, the method safely populates sparse regions of the feature space. Trained on this augmented dataset, the MO-LSTM achieved robust simultaneous estimation across all targets. This approach mitigates data sparsity, providing a basis for hybrid mechanistic-data-driven modelling and early-warning soft sensors in industrial bioprocesses.

0.29

When Can Unlabeled Data Pools Replace Membership Queries?

cs.LG ·2026-10-29 ·Dravyansh Sharma
Abstract

Membership queries let a learner label arbitrary points and give near-optimal $\tilde O(n\log(1/\eps))$ label complexity for halfspaces under every marginal. A pool learner may label only sampled points. We ask when unlabeled data can replace membership queries. For homogeneous halfspaces the answer depends only on the angular law of $X/\|X\|$. If this law assigns zero mass to every hyperplane through the origin and positive mass to every nonempty open set of directions together with its reflection, a pool learner uses $O(n\log^3n\log(n/(\epsδ)))$ labels, uniformly over all such marginals and targets, without density, moment, concavity or isotropy assumptions. The same label bound holds for non-homogeneous halfspaces under every marginal with full support and no mass on affine hyperplanes, including Gaussians. Dropping either angular assumption admits $Ω(1/\eps)$ pool versus $O(\log(1/\eps))$ membership-query separations, even with unlimited unlabeled data. The price is paid in unlabeled data: in dimension $n\ge3$, for every unlabeled budget fixed in advance, some marginal satisfying both assumptions forces $Ω(1/\eps)$ labels. For isotropic $s$-concave marginals in dimension $n\ge3$, unlimited pools keep logarithmic label rates for every finite $s$. With an unlabeled budget fixed in advance, class-uniform polylogarithmic rates hold exactly for $s>-1/(n+1)$, or $s>-1/n$ under central symmetry, and in the general class the optimal rate between these thresholds is $\tildeΘ(\eps^{-(n-2)/n})$. This extends the range $s\ge-1/(2n+3)$ of previous sample-based guarantees, in the realizable setting.

0.28

Training Foundation Models with a Single Algorithm

cs.LG ·2026-10-29 ·Yuda Song, Alvin Zhang, Fahim Tajwar, Daman Arora, Jeff Schneider +5 more
Abstract · 67 pages, 31 figures, 22 tables

Foundation models are typically trained in stages: pretraining, midtraining, supervised fine-tuning (SFT), and reinforcement learning (RL), with changes in both data and learning algorithms. We show that this algorithmic separation is not fundamental. By deriving suitable RL objectives as maximum-likelihood fitting to model-generated target distributions, we place all of these stages under a single likelihood-based training algorithm, unified training. This perspective turns the foundation-model training lifecycle into a problem of choosing which data to train on, in what proportions, and when, with conventional staged training as a special case. We instantiate this framework in a unified training system that mixes static data and RL generations within a single trainer. Across controlled and real-world experiments, unified training improves the accuracy-compute tradeoff and task coverage while preserving broader capabilities, and provides a stronger starting point for continual learning. This work lays the groundwork for a unified science of foundation-model training, connecting mathematical principles, training infrastructure, and the development of model capabilities.

0.28

Steering Vectors as a Training Signal in LLM Post-Training

cs.LG cs.AI ·2026-10-29 ·Tiejin Chen, Maunil Vyas, Huaiyuan Yao, Hua Wei
publishedcode/data
Abstract · Accept to NeurIPS 2026

Steering vectors (SVs) are residual-stream directions that shift the behavior of LLMs when added to hidden states at inference time. While prior work has mainly used SVs as post-hoc controls, we ask whether the same directions can also guide parameter updates during post-training. We study this question across supervised learning, including SFT and distillation, and reinforcement learning with verifiable rewards (RLVR). Our framework considers two intervention regimes: a hidden-state hook during teacher-forced supervised training, and an advantage-repair mechanism on wrong rollouts in GRPO. For each regime, we analyze three design axes, which are the steering direction, its placement, and the operation used to combine it with training. In supervised settings, SV injection improves over plain SFT and distillation, with gains scaling with the off-policy gap and reaching up to a $12\%$ improvement when distilling from thinking-mode teachers. These gains hold across three random seeds. In RL settings, applying a correctness-contrast direction as advantage repair on wrong rollouts, including otherwise zero-gradient all-wrong groups, recovers a learning signal that transfers to held-out math benchmarks and acts as a composable auxiliary signal that adds to existing zero-variance remedies such as entropy-aware advantage shaping. We further identify three failure boundaries. In detail, code generation can break the direction, cross-family layer transfer can break the placement, and inference-time application on an SV-trained checkpoint can break the intervention regime. Together, these results establish steering vectors as a conditional but effective training signal for LLM post-training. Our code is available here.

0.28

AI at a Turning Point: From ChatGPT through Neural Networks to Neuromorphic AI

cs.LG cs.NE ·2026-10-29 ·Volker Kempe
published
Abstract · 88 pages, 20 figures

The tension between the established technologies of connectionist AI and their neuromorphic challengers is attracting growing attention, above all because of the enormous energy demand of the former and the low power consumption of the latter. Developing and running large AI models consumes gigawatt- to terawatt-hours of energy, while the human brain operates on a continuous 20 watts. Spiking neural networks (SNNs) counter this energy demand with a frugal approach and bring elements of self-organization into the methodological repertoire of AI: local learning instead of global error backpropagation, event-driven computing, and time as a carrier of information. They do not replace today's AI but extend it; in hybrid systems each approach can be deployed where its strengths lie. This monograph explains the field in a generally accessible yet mathematically rigorous way. Using the text-generation module of ChatGPT as an example, the principle, elegance, and computational cost of transformer-based AI are presented and compared with the coding, learning, and network-design methods of SNNs. Throughout, relevant mechanisms of the human nervous system serve as model and reference point. Topics include spike-timing-dependent plasticity, population and temporal coding, the Neural Engineering Framework, neuromorphic hardware from Loihi 2 to commercial edge processors, and the prospects for classical and neuromorphic AI to complement and extend each other.

0.28

Does System Size Matter in Uncertainty Quantification for Machine Learning Interatomic Potentials?

cs.LG cond-mat.mtrl-sci ·2026-10-29 ·Chayaphol Lortaraprasert, Nontawat Charoenphakdee, Wenwen Li, Yong-Bin Zhuang, Yuta Tsuboi
Abstract

Machine-learning interatomic potentials (MLIPs) commonly use uncertainty estimates to flag out-of-distribution (OOD) configurations and select structures for active learning. A key question is whether the resulting structure-level uncertainty scores remain comparable across different system sizes. A common practice is to aggregate per-atom uncertainties by their mean or maximum and compare the resulting score against a fixed threshold, implicitly assuming such comparability. We test this assumption across several in-distribution (ID)/OOD pairs by characterizing both aggregations as a function of atom count. The two aggregations exhibit distinct size dependencies. Mean aggregation remains size-independent for homogeneous systems but loses sensitivity to defects in large cells. In contrast, maximum aggregation grows with system size for ID systems while remaining approximately flat for defect-containing systems, narrowing the ID-OOD gap in a manner consistent with extreme-value statistics. Although highly uncertain atoms in large structures may indicate novel local environments, retraining on these structures yields little improvement, in contrast to retraining on OOD structures constructed to lie outside the training distribution. Our results reveal a practical limitation of size-unaware uncertainty criteria and motivate size-aware calibration for more reliable OOD detection in large-scale atomistic simulations.

0.28

P-ContextBench: From Personal Fact Recall to Personalized Agent Behavior

cs.AI ·2026-10-29 ·Grant Nasir Brown, Abdul Basir Bari, Anmol Tadikonda, Isabel Song
benchmark
Abstract

Large language models (LLMs) are increasingly used for personal tasks, yet evaluating their safe use of personal context remains difficult because such contexts are private, fragmented, dynamic, and socially situated. We introduce P-ContextBench, a realistic, verifiable, and scalable benchmark for evaluating personal-context understanding. P-ContextBench constructs population-structured, life-story-grounded digital humans from demographic seeds reflecting the U.S. population. Rather than using explicit, static profiles, P-ContextBench embeds traits, routines, goals, and constraints as implicit evidence across fragmented, evolving synthetic mobile-app records. It evaluates five capabilities: single-source fact recovery, multi-source fact recovery, temporal state tracking, inference of higher-level personal patterns from routine behavior, and safe handling of requests conflicting with persona-specific constraints through compliance, warning, clarification, or refusal. A multi-stage validation pipeline checks persona plausibility, context recoverability, information leakage, evidence preservation, test-case grounding, and human-judge agreement. We instantiate P-ContextBench with 100 personas, 10,088 hidden facts, 18,261 app-trace fragments, and 4,810 verified test cases. The pipeline supports configurable generation across scales and demographic groups. Cost-effective state-of-the-art (SOTA) LLMs relevant to personal agents achieve below 60% accuracy on temporal tracking, higher-level pattern inference, and constraint-aware action-boundary decisions, with fewer than 10% of test cases answered correctly across all three repeated runs. Even a high-performing SOTA model remains inconsistent on these capabilities. P-ContextBench provides measurable targets for improving the accuracy, reliability, and safety of LLM-based personal agents.

0.27

Machine Learning-Based Detection of Mobile Money Scan Messages in Ghana: A Comparative Evaluation of Naive Bayes, Support Vector Machine, and Randon Forest Classifiers

cs.LG ·2026-10-29 ·Ibrahim Salifu, Muhaisin Tiyumba Nantogmah
Abstract · 17 Pages, 8 Figures

Mobile money (MoMo) is Ghana's preferred retail payment channel, recording GHC518.4 billion in transactions in December 2025 and over 26.7 million active accounts (Graphic Online, 2026). This growth has coincided with increased social-engineering fraud over Short Message Service (SMS) and Unstructured Supplementary Service Data (USSD), where fraudsters impersonate network operators, agents, relatives, or regulators to obtain personal identification numbers (PINs), identity data, or funds. We present a supervised machine-learning pipeline for classifying scam and legitimate Ghanaian MoMo SMS messages using a labelled corpus of 220 messages across 18 scam and legitimate subcategories. Domain-specific normalisation of currency amounts, phone numbers, and USSD codes preceded TF-IDF vectorisation of unigrams and bigrams. Using stratified five-fold cross-validation and a 20% held-out test set, we trained and evaluated Multinomial Naive Bayes, linear support vector machine (SVM), and Random Forest classifiers using accuracy, precision, recall, F1-score, Matthews correlation coefficient (MCC), and area under the receiver operating characteristic curve (AUC-ROC). Held-out accuracy, F1, and AUC were 1.00 for Naive Bayes and linear SVM; cross-validated accuracy averaged 96.6% (+/-3.3), 94.9% (+/-4.6), and 92.0% (+/-4.2). Discriminative n-grams reflected imperative urgency ("send," "claim," "please send") and PIN solicitation ("PIN," "pincode"); legitimate messages featured confirmatory and cautionary language ("balance," "successful," "do not share"). Findings are contextualised in Sub-Saharan African SMS fraud detection, small template-based corpus ceiling effects, and a pathway toward larger naturalistic multilingual corpora. Interpretable classical classifiers achieve high accuracy on structured MoMo scam text and suit carrier- or handset-level filtering in low-resource, on-device settings.

0.27

The Geometry of Multitask Learning: Sharp Rates for Shared Subspace Estimation

cs.LG stat.MLmath.ST ·2026-10-29 ·Tony Cai, Abhinav Chakraborty, Samory Kpotufe
Abstract · 81 pages, 2 figures

Learning a shared subspace in multitask learning is an important statistical problem, yet exact rates remain poorly understood in many of the formal settings considered in the literature. To address this gap, which arises from limitations of existing upper-bound techniques, we propose a general multitask lemma. The lemma provides sufficient conditions for fast subspace estimation in generic multitask settings, in the form of geometric conditions on the task regression functions and their estimators. We subsequently show that the multitask lemma yields sharp minimax rates---previously unknown---for linear regression and classification in the regimes considered, and rates matching up to logarithmic factors for additive-noise regression.

0.27

Consensus Within, Competition Between: Decentralized LLM Teams for Perimeter Defense

cs.AI cs.LGcs.RO ·2026-10-29 ·Khawaja Abaid Ullah, Ananya Acharya, Mohammad Javad Khojasteh
code/data
Abstract · 88 pages (33 main text and references, 55 supplementary material), 7 main figures. Code: https://github.com/the-kas-lab/consensus-within-com…

The use of large language models (LLMs) to coordinate multi-agent teams raises a fundamental architectural question: whether task allocation should be centralized in a single LLM controller or distributed across autonomous LLM agents, particularly when communication is unreliable. In a deliberately simplified perimeter-defense simulation against LLM-controlled intruders, we compared a central commander, an auction-based consensus team and independent agents under five network conditions. Breach rates were 15% for the consensus team, 32% for the commander and 29% for independent agents, and of the different network conditions, only limited radio range changed a team's breach rate, raising the commander's. In consensus team, giving a low-cost LLM more authority over the bid computation raised breaches by 38 percentage points. With birds and civilian drones in airspace, we added a photograph for each to the input text state. Only the near-frontier model treated them apart from intruders. Intruders disguised as existing taxidermy bird drones were mistook as birds by both low-cost and near-frontier LLMs, going unpursued in 59-61% of near-frontier decisions (1-16% before). However, a one-sentence forewarning in the prompt lowered it to 0-25%. We find that LLM team with a dedicated external conflict resolution protocol results in fewest breaches. Code: https://github.com/the-kas-lab/consensus-within-competition-between.

0.27

CBE: A Complexity-adaptive Boundary-aware Ensemble Method for Imbalanced Data

cs.LG ·2026-10-29 ·Xun Wei
benchmark
Abstract · 12 pages, 5 figures

In many real-world classification tasks, datasets exhibit class imbalance, which can degrade the generalization performance of conventional learning algorithms. Data complexity has recently been recognized as a more objective indicator of learning difficulty than the imbalance ratio alone, because class overlap and complex decision boundaries often impair classifiers more severely than imbalance itself. Samples near the decision boundary are inherently difficult to classify, which has motivated many researchers to focus on such instances to improve classifier performance. However, most under-sampling approaches fail to differentiate their sampling of boundary samples across datasets of varying complexity. To overcome this limitation, we propose a complexity-adaptive boundary-aware ensemble (CBE) framework that links the sampling weight distribution to data complexity and dynamically adjusts the boundary preference of under-sampling accordingly. We also introduce an overall complexity measure that combines metrics at multiple levels and use it to divide imbalanced datasets into three categories. Extensive experiments on 24 benchmark datasets from UCI and KEEL show that the proposed method outperforms many existing methods in terms of F-measure and G-mean on Easy and Median datasets, without degrading performance on hard cases.

0.27

A physics-informed uncertainty-aware surrogate for $p$-$n$ junction diodes

cs.LG cond-mat.mes-hall ·2026-10-29 ·Suchita Kulkarni
benchmarkunusual
Abstract · Comments welcome

We develop an uncertainty-aware physics-informed surrogate for a one-dimensional $p$-$n$ junction diode, conditioned on device length, doping and bias. Using the internal state of the device as predicted by the surrogate we develop a test-suite of quantities of interest including the terminal current, depletion width, peak field and stored charge. The formalism and the quantities are can be extended to asymmetric junction and two-dimensional diodes as shown by means of benchmark analysis. For the one-dimensional symmetric junction surrogate, we obtain 0.11% median $L_1$ error on the electrostatic potential, and 1.5% and 0.61% on the electron and hole densities. Finally, we develop an uncertainty analysis which determines the validity of surrogate answer within the user-specified tolerance.

0.27

Retrieval as Traversal: Value-Indexed Memory for Stateful Long-Context Agents

cs.AI ·2026-10-29 ·Daniel Strauß, Zhenshan Bing, Alois Knoll
published
Abstract · Accepted at the NeurIPS 2026 Workshop on Long-Context Foundation Models (LCFM). 7 pages, 1 table

Long-context models repeatedly solve closely related retrieval problems. Attention or global retrieval can retain large stores, but each forward pass again searches for what matters. Recurrent compression avoids this search only by deciding earlier what information to preserve. Neither mechanism directly exploits that useful memories at neighbouring retrieval steps may substantially overlap. We propose an alternative in which retrieval is a persistent local traversal through an address space. We separate a memory's content from its address. Utility is defined by the information revealed by retrieving the content, while geometry determines only how cheaply that information can be reached. This leads to a cost-regularised indexing objective that preserves information value inside local neighbourhoods while charging for every memory made locally accessible. Its marginal-value form gives both attractive and exclusion pressures. Thus, complementary memories should be co-retrievable when the additional value exceeds retrieval cost, whereas redundant memories should not occupy the same retrieval set solely because they are individually useful. As a proof of principle, we test only the simplest fixed-geometry instance on MiniGrid Memory. A persistent local traversal head reaches $0.95$ mean return after $0.59$M training steps with a five-token context, compared with $0.94$M for Gaussian absolute addressing and more than $2$M for categorical absolute addressing. The experiment does not test learned geometry or repeated retrieval. Instead, it motivates the hypothesis that stateful retrieval may be easier to learn than stateless. To test the full proposed method further experimentation is necessary.

0.27

Rethinking Neural Thickets Without Weight Perturbations

cs.LG cs.AIcs.CV ·2026-10-29 ·Nicolas Dufour, Elouan Gardès, Amil Dravid, Patrick Pérez, Loic Landrieu
code/data
Abstract · Project page can be found here: https://kyutai.org/logit-psv

The Neural Thickets hypothesis posits that pretrained language models lie in dense neighborhoods of task-specific experts in weight space, which random Gaussian perturbations can uncover. We reproduce the reported gains, but show that they do not require this geometric explanation. Perturbing a single decoder block already recovers much of the improvement, while keeping the model entirely frozen and adding a random bias to its output logits matches or exceeds full-weight perturbations on most reasoning tasks. This suggests an alternative explanation based on output diversity, validation-based selection, and voting. We formalize this view as Perturb-Select-Vote (PSV), a general procedure that generates perturbed predictors, selects high-performing members on a small validation set, and aggregates their predictions. This interpretation connects Neural Thickets to established inference-time methods, including sampling, self-consistency, and verifier-guided selection. A first-order analysis further shows that small weight perturbations act primarily through their induced logit shifts. We extend this analysis to vision, where logit PSV substantially improves performance under domain shift, including on Earth observation, while full-weight perturbations provide little additional benefit. Together, these results suggest that the gains attributed to nearby weight-space experts can instead be obtained through output diversity, selection, and voting. Logit PSV realizes these benefits with a fully frozen model, providing a simple, training-free alternative to full-weight perturbation ensembles across language and vision.

0.27

Causal-prior-constrained deep learning for rapid-intensification warning of western North Pacific tropical cyclones

cs.LG physics.ao-ph ·2026-10-29 ·Xinrui Li, Yijun Jin, Yifan Dong, Shoujuan Shu
Abstract · 33 pages, including 15 pages of supplementary information; 5 main figures and 6 supplementary figures

Rapid intensification (RI) is one of the most dangerous and unpredictable phases of tropical cyclones, and its environmental and structural drivers are difficult to identify and separate in observations. Here we introduce CausalVisionRI, a causal-prior-constrained deep-learning system for 24-h RI warning in the western North Pacific. It integrates infrared remote sensing imagery, atmospheric and oceanic fields, and an observationally discovered causal graph into a unified model. The graph and eight sign-verified effects on 24-h intensity change provide an auditable causal prior for training. Discovered from 522 cyclones in 2000-2017, it identifies causal pathways linking atmospheric and oceanic conditions, the cyclone's dynamical and structural state, and its 24-h intensity change. On a forward holdout covering 2020-2024 and 3,004 records, the five-member ensemble achieves a Brier skill score (BSS) of 0.259 relative to a fixed development climatology, significantly exceeding four reference systems; the three LDA baselines reach at most 0.105. It also attains higher detection rates than a published knowledge-block scheme and a reconstructed contrastive-learning scheme at matched false-detection rates under our USA 30-kt standard and the published RITCF standard. Ablations show that the graph pathway adds 0.032 BSS across the holdout and 0.063 over 2020-2021, with a significant contribution in one ensemble member. A deep-learning model constrained by observationally discovered causal structure thus offers a new route to physically informed RI warning.

0.26

MemSteer: Turning User Histories into Supervision for Agent Memory Optimization

cs.AI ·2026-10-29 ·Jie Huang, Cuiling Lan, Qize Yu, Xiulian Peng, Yan Lu
Abstract

An effective agent memory system should adapt to a user's history and task requirements. Human-annotated supervision can guide this adaptation by providing feedback for evaluating and improving memory designs. Yet suitable labels are rarely available for new user histories, and manually constructing QA pairs and supporting evidence is costly. We introduce MemSteer, which turns unlabeled user histories into question-answer (QA) supervision with supporting evidence for memory-design optimization. MemSteer constructs an evidence-grounded answer plan that specifies how facts across related records jointly support a reference answer. The plan guides question generation, while source-grounded auditing and blind answer recovery check the resulting annotations. To study the utility of this supervision, we develop MemSteer-Opt, which combines QA-based design evaluation with evidence-guided failure diagnosis to revise memory store, retrieval, and context construction without updating model parameters. Experiments on ATM-Bench and RHELM show that this supervision is competitive with official-QA-guided optimization. Relative to Base-BM25, MemSteer-Opt improves ATM-Bench Question Score by 13.4 points and RHELM accuracy by 4.1 and 6.6 points without and with external data, respectively. The supervision also benefits two existing optimizers on ATM-Bench, while new-user experiments demonstrate continued design optimization using supervision generated from the target histories.

0.26

Fisher Contrastive Learning to Mitigate Feature Suppression: A Case Study in Acute Promyelocytic Leukemia Diagnosis

cs.LG stat.ML ·2026-10-29 ·Shushan Wu, Yongkai Chen, Xilin Gong, Yufang Liu, Wenxuan Zhong +1 more
theory
Abstract · 31 pages, including supplementary material; 3 figures and 2 tables

Self-supervised contrastive learning (SSCL) learns representations by contrasting augmented views of the same sample, but it can exhibit feature suppression, where dominant or easy-to-learn factors suppress other class-relevant signals, degrading the performance of representations, particularly in high-dimensional data with heterogeneous features. We introduce Fisher Contrastive Learning (FCL), which uses a regularized Fisher trace criterion to address feature suppression. FCL learns embeddings by maximizing the discriminative strength of Fisher discriminant functions, producing representations that preserve class-relevant information implied by the augmentation process. We prove that the representations learned by FCL could alleviate the feature suppression effects. We demonstrate FCL on simulated benchmarks designed to induce feature suppression and on a real-world biomedical application, pretraining on large white blood cell microscopy image dataset and transfer learning for diagnosis of acute promyelocytic leukemia in an external cohort. In all experiments, FCL outperforms the self-supervised baselines and yields more robust and discriminative representations.

0.26

ReplayRCA: Measuring Post-hoc Diagnosis and Live Oversight of Agent Failures by Counterfactual Replay

cs.AI ·2026-10-29 ·Jeremy Canale, Fabien Soulis
benchmarkcode/data
Abstract · 29 pages, 10 figures, 16 tables. Code: https://github.com/jeremy1392/RCAreplay

When a language-model agent fails a multi-step task, can another model find the decisive mistake and fix it with a single action? We introduce ReplayRCA, a benchmark that scores such diagnoses by intervention rather than by agreement with annotations: the environment is restored to the state preceding the chosen step, the proposed action replaces the agent's own, and the same agent continues. The score, causal repair, is the continuation success rate minus that of a paired placebo that replays the agent's original action through the same mechanism. Two tracks share this protocol: a post-hoc autopsy, where the diagnoser sees the whole failed run and is told that it failed, and live oversight, where a supervisor watches the run unfold and may intervene once, before an irreversible action, without knowing whether the run will fail. Tasks come from OpsWorld, a deterministic, procedurally generated back-office environment with six difficulty levels, the last of which is multimodal (QR codes, handwriting). On 62 calibrated systematic failures, the best of 13 frontier models (Grok 4.7) reaches a causal repair of +51.3 points; the top eight models are statistically indistinguishable, the flagship models of the two largest vendors rank 8th and 12th, and 5 items are repaired by no model. Forcing a strong model to intervene at the last step costs it 17.1 points, so the last-step shortcut of annotation-based attribution benchmarks does not transfer. On 87 runs, the two best live supervisors are statistically tied and one of them is a small open-weight model (DeepSeek V4.1 Flash); the supervisor that is among the best post-hoc diagnosers (GPT-6.1 Sol) is the weakest live one (-14.5 points on the same runs, p = 0.002), so post-hoc diagnostic performance can be a poor proxy for live oversight. The environment and the evaluation code are released at https://github.com/jeremy1392/RCAreplay.

0.26

GC-LIF: Spiking Group-Equivariant Neural Network With Group-Coupled Membrane Dynamics

cs.NE cs.CV ·2026-10-29 ·Nhan Trong Luu, Duong Trung Luu, Thang Cong Truong
Abstract · Under review

Spiking neural networks (SNNs) promise event-driven, low-power inference, but on vision tasks they still acquire rotation invariance only through data augmentation. Group-equivariant CNNs build such symmetries into the architecture, yet a faithful spiking analogue is missing: a leaky integrate-and-fire (LIF) neuron placed after a $G$-equivariant convolution can break equivariance through its membrane recurrence. We give a necessary and sufficient condition for this not to happen. The layer is equivariant at every time step if and only if the neuron parameters $(β, v_{\mathrm{th}})$ are constant along the joint orbits of the group action on the (group, space) index. We solve it in closed form, and it covers any elementwise stateful neuron. Its sufficiency half specialises a known result for equivariant recurrent networks; its necessity half rules out the heterogeneous per-neuron parameters common in SNNs. The per-channel parameterisation that satisfies it, Group-Coupled LIF (GC-LIF), is combined with $p4$ and $p4m$-equivariant convolutions and a spike-rate group max-pool into SGE-Net. On MNIST, KMNIST, Fashion-MNIST, CIFAR-10 and CIFAR-100, evaluating each checkpoint on the clean test set and all its group-transformed copies, SGE-Net trained without rotations is invariant under $C_4$ and $D_4$ and outperforms a plain SNN trained with rotations in nine of ten dataset-group pairs, whereas non-equivariant baselines lose up to $72$ percentage points in the worst case. Operator-level tests confirm the theory before and after training, and an analytical estimate places SGE-Net at $1.7\times$ the arithmetic energy of a standard CNN and $2.3\times$ below its ANN counterpart at eight time steps.

0.26

Capability-Grounded Agentic Planning

cs.AI cs.CL ·2026-10-29 ·Shikha Shrestha, William Yeoh, Ning Zhang, Yevgeniy Vorobeychik
code/data
Abstract · 25 pages, 3 figures

LLM agents serve natural-language requests through tools, data sources, and other agents, and even simple requests often need several interdependent calls. Typically, the LLM picks these calls by trial and error, with no guarantee of an executable plan. We present Capability-Grounded Agentic Planning (CGAP), which instead reasons about what a request requires and what tools provide. CGAP compiles the tool catalog deterministically into a PDDL model, uses an LLM to turn the request into goals, and solves the model with a classical planner. Since tool outputs are unknown in advance, the model is abstract, tracking which inputs are available, not their values. During execution, an LLM adds entity-specific calls that abstract plans omit, and deterministic checks reject any that break the plan. Every argument is drawn from a cited source, such as a conversation span or an earlier tool response, and each requested result counts as delivered only after an LLM verifier checks it against the request. With three LLMs on ToolSandbox, ComplexFuncBench, BFCL multi-turn, and $τ^2$-bench retail, CGAP outperforms a standard tool-calling agent by up to 28.9 points.

0.26

DWISE: Denoising Workload-Informed Scheduling Engine

cs.LG ·2026-10-29 ·Lie Li, Wen Li, Junxiao Shen, Jindong Gu, Guosheng Hu
Abstract · 17 pages, 11 figures, 17 tables

Unlike autoregressive large language models (AR-LLMs), diffusion LLMs (dLLMs) generate text by iteratively denoising many tokens in parallel, offering a promising path to efficient inference. Existing serving frameworks, however, schedule dLLM requests with policies designed for AR models, namely first-come-first-served (FCFS) admission and prefill priority. We find that both policies are ill-suited to dLLMs, wasting much of their speed. FCFS ignores that a dLLM request's decoding cost is highly heterogeneous and unknown at arrival, causing severe head-of-line blocking. Prefill priority, widely used in AR-LLM serving to keep time-to-first-token low, sharply cuts dLLM efficiency, as rounds that run only prompts leave most of the batch idle. We present DWISE, a Denoising Workload-Informed Scheduling Engine for efficient dLLM serving. DWISE schedules requests by their remaining work, measured in a block-level unit shared by prefill and denoising, so both phases share a round. To predict the unknown denoising workload, DWISE combines a lightweight prompt-based predictor at admission with a single refinement from the first denoising round's hidden states, which capture model-specific information absent from the prompt and require no additional model execution. Implemented in SGLang and evaluated on five models from three dLLM families under bursty workloads, DWISE reduces mean end-to-end latency over SGLang by 5.4$\times$ on average and up to 14.9$\times$ while preserving task accuracy. The code will be released publicly.

0.26

ACAM-J-MARS: Multi-Atlas Riemannian Spectral Representations for Decoding Advanced Meditation States from fMRI

cs.LG ·2026-10-29 ·Jian Wang, Winson F. Z. Yang, Wugeng Zheng, Matthew D. Sacchet
Abstract · 4 pages; 4 figures

Inferring jhana-type advanced concentration absorption meditation (ACAM-J) states from fMRI data is challenging due to the spatially distributed nature of meditation-related effects, substantial inter-individual variability that can exceed within-subject state differences, and limited sample sizes. Conventional machine-learning approaches often formulate this problem as unimodal, Euclidean, atlas-based classification, thereby neglecting the non-Euclidean geometry of functional connectivity matrices and underrepresenting temporal information in functional magnetic resonance imaging (fMRI); these limitations can lead to suboptimal classification performance. We propose ACAM-J-MARS (\textbf{M}ulti-\textbf{A}tlas \textbf{R}iemannian \textbf{S}pectral representation), which computes functional connectivity using multiple atlases, represents symmetric positive definite (SPD) connectivity matrices on a Riemannian manifold, projects them into a log-Euclidean tangent space, and augments connectivity-derived features with low-frequency BOLD spectral energy to better resolve subtle state differences under heterogeneous variability. Evaluated on 204 segments from 22 practitioners, ACAM-J-MARS improves performance across multiple backbone models, achieving a balanced accuracy of 0.757 for early versus late ACAM-J classification and 0.545 for three-class classification, corresponding to a 10.8-point absolute improvement over the strongest baseline. These findings suggest that the proposed representation improves the decoding of advanced meditative brain states and may enable downstream applications such as closed-loop neurofeedback and real-time psychophysiological brain-computer interfaces.

0.26

FIDE: Fiber-Informed Differentiable Experimental Design for Reference-Relative Law Dynamics

cs.LG ·2026-10-29 ·Pietro Zanotta, Ján Drgoňa
Abstract

Aggregate measurements generally leave an evolving population law underdetermined, so different measurement systems can be equally adequate for a scientific task while constraining different aspects of the law. We study how to resolve this residual design freedom in baseline-preserving settings, where an existing dynamical model remains the working baseline outside quantities explicitly valued by the task. \textbf{F}iber-\textbf{I}nformed \textbf{D}ifferentiable \textbf{E}xperimental Design (\textbf{FIDE}) completes each candidate's aggregate constraints by information projection relative to a common frozen reference, then ranks scientifically adequate designs by the dynamical revision required to realize the resulting completed paths. The resulting \emph{Full action} is the minimum mean-squared velocity correction needed to realize a completed path; equivalently, its square root is the sharp reference-relative expectation-rate discrepancy over a gradient-normalized Sobolev class. This defines a conditional minimum-revision principle for baseline-preserving design rather than a generic truth-recovery objective. Experiments show that this preference changes sensor selection, can help or harm withheld downstream predictions depending on baseline reliability, can be used prospectively from aggregate predictions, and remains computationally usable for a $32$-dimensional configuration law.

0.26

Carbon-Aware Electric Vehicle Routing Problem with Decision-Focused Learning

cs.LG math.OC ·2026-10-29 ·Jaume Reixach, Wanru Chen, Jiaming Wu, Balázs Kulcsár
Abstract

Electric vehicles produce no tailpipe emissions, but charging them produces emissions that depend on the time-varying generation mix of the electricity consumed. We study an Electric Vehicle Routing Problem with a carbon-aware objective that accounts for charging-related emissions alongside conventional routing costs. The problem determines routing, charging, and waiting decisions, allowing charging demand to be shifted toward periods with lower grid carbon intensity. Because these decisions depend on forecasts of future carbon intensity, we propose a decision-focused learning approach so that the training of the model employed for predicting the carbon intensity of the grid over time is solver-aware. In particular, we train a linear predictor using approximate SPO+ directions and employ a time-limited Large Neighborhood Search followed by exact postprocessing as the underlying solver. Experiments on realistic Germany-based instances show that explicitly considering charging-related emissions reduces realized emissions by 28.76% on average relative to carbon-blind optimization, although at the expense of higher non-carbon operating cost. Moreover, decision-focused training reduces realized total cost by 2.30% on average relative to conventional prediction-focused learning and performs better on 75.5% of the test instances, despite substantially lower pointwise predictive accuracy. Finally, we show that its average advantage remains positive across varied solver configurations and carbon-cost weights.

0.26

Probabilistic Meshless Methods on Manifolds

cs.LG stat.ML ·2026-10-29 ·Federico Pavesi, Noémie Jaquier, Antonio Candelieri
Abstract

Probabilistic meshless methods solve partial differential equations (PDEs) by placing a Gaussian process prior on the solution and conditioning it on the PDE at a finite set of collocation nodes. The posterior provides both an approximation of the solution and a Bayesian quantification of its error. While these methods are well established on Euclidean domains, their extension to manifolds lacks, to our knowledge, any convergence analysis. We extend them to compact manifolds and study their convergence in two settings. First, for known manifolds whose Laplace-Beltrami eigenpairs are available, the prior can be evaluated exactly, but the PDE operator generally cannot be applied to it in closed form. We replace this operator by a standard generalized finite difference discretization on a point cloud sampled from the manifold, and show that the resulting method is consistent at the points of the cloud, at a rate governed by their fill distance. Second, for implicit manifolds, known only through a point cloud, both the prior and the operator must be approximated. We propose a new generalized finite difference scheme with a free stencil radius, and show that it retains the consistency of the standard construction while controlling the growth of its coefficients, and that an appropriate radius yields uniform convergence of the discrete Gram matrix. In both settings, we derive error bounds for the posterior mean which separate the collocation error intrinsic to the method from the error due to approximating the operator and, for implicit manifolds, the prior. Experiments on the circle, the sphere and the flat torus confirm our theoretical claims. On real-world cortical thickness data, our approach attains the accuracy of established discretizations, while conditioning on fewer than a tenth of the observations and providing a credible interval at every vertex.

0.26

Locating and Closing Information Gaps in Multi-Agent Systems

cs.AI ·2026-10-29 ·Yaxin Gong, Chongming Gao, Gangyi Zhang, Chenxiao Fan, Leyang Shen +5 more
Abstract

Multi-step Web tasks often require several LLM agents to gather and combine evidence. The communication graph of a fixed workflow restricts which agents can read each other's outputs: evidence retrieved by one agent may never reach the agent that needs it, leaving an information gap and an incomplete or incorrect answer. Adding a connection can deliver the missing evidence, but can also lead the receiver to adopt an incorrect upstream conclusion that displaces a correct answer; the final output alone does not separate the two. We propose Selective Probing and Link Insertion for Closing information gaps in multi-agent Execution (SPLICE), which locates information gaps per query and decides what to adopt after adding an edge. After the original workflow runs, SPLICE uses the query and node outputs to locate a possible gap, adds one directed edge from a source to a receiver to deliver the relevant evidence, and re-executes the receiver and affected downstream nodes, producing a candidate answer. The adoption decision reads both the final answers of the two runs and the process changes the edit caused inside the workflow: how far the receiver moved, where its new content came from, and whether the change reached the final answer. A selector trained on these process features chooses between adopting the candidate answer and retaining the original. On five benchmarks, SPLICE raises the task score on Web tool use by 8.0% and the mean score on four reasoning and planning benchmarks by 5.7% relative to the static skeleton, at 1.25 to 1.44$\times$ the skeleton's cost, and transfers zero-shot to four unseen backbones.

0.26

All Paths All at Once: A Quantum Monte Carlo and Neural Committor Approach to Reaction Pathway Search

cs.LG physics.chem-phquant-ph ·2026-10-29 ·Chien-Chai Chang
Abstract · 34 pages, 6 figures

Locating competing reaction pathways is difficult when potential-energy gradients are unavailable. Here we present tReax_expedition, an energy-only approach combining coupled Monte Carlo exploration, stochastic quench labelling and restrained continuation. Exploration identifies connected states, while continuation resolves candidate channels and records sampled chain maxima. Single-search benchmarks on three two-dimensional potentials recovered an intermediate and competing routes without transition-state guesses. For eight of nine gas-phase reactions on GFN2-xTB, sampled chain maxima differed from independently refined saddle energies by less than 0.76 kJ/mol. Alanine dipeptide instead revealed incomplete conformational relaxation, showing that route connectivity can persist despite structural and energetic errors. Neural committor learning and local straddling-pair sampling connect the resulting paths to transition-region sampling, although molecular shooting diagnostics indicate limited committor calibration. These results show how energy-only exploration and continuation recover candidate pathways, while distinguishing route connectivity from geometric localization, barrier resolution and finite-temperature transition statistics.

0.26

Kairos: Semantics-Aware Orchestration of Planning Engines across Classical, Numeric, Temporal and Hybrid PDDL

cs.AI cs.RO ·2026-10-29 ·Francisco Martín Rico, Esther Aguado González, Francisco Miguel Moreno Olivo, Rodrigo Pérez Rodríguez, José Miguel Guerrero Hernández +1 more
benchmarkcode/data
Abstract · 32 pages, 3 figures, 11 tables. Submitted to the Journal of Artificial Intelligence Research. Code: https://github.com/PlanSys2/kairos ; ben…

Planning at run time, as in robots that re-plan while they act, needs a planner that reads the PDDL its users write (durative actions, numeric resources, time windows, continuous change), answers quickly, and delivers only validated plans. Specialised planners are strong on individual fragments of PDDL, but they differ in coverage, latency and in what their failures mean. We study whether selecting heterogeneous engines from the semantics of each task can give broad coverage of PDDL, from STRIPS to PDDL+, while remaining competitive with specialised planners and making explicit what each answer guarantees. Kairos profiles each task to decide which of five engines apply (ground heuristic search, compressed and snap-action temporal search, time steps for hybrid dynamics, lifted search) and schedules them under one budget of time and memory. Every candidate plan is checked by a validator independent of the planner's transformations, and "unsolvable" is reported only when a search whose exhaustion is a proof exhausts its space. On 1,756 tasks from nine comparative suites, Kairos solves more tasks at the full limits than the strongest evaluated planner of each suite, with the largest differences on numeric, temporal, PDDL3 and PDDL+ planning, and it also solves the most tasks at 100 ms and 1 s. Resampling over domains, the gain is above zero in six of the nine suites, while on classical planning against LAMA it is within the noise. A further 450 feature-oriented tasks and 26 one-feature probes show that Kairos handles all seventeen features probed, where no other planner evaluated handles more than eight. Ablations identify compressed temporal search and the novelty fallback as major contributors, and a comparison with VAL traces every disagreement to a PDDL+ convention or to a documented VAL defect. Planner and benchmark are open source.

0.25

Where, When, and How Many Bits? Scaling Laws for SGD with Stochastic Rounding

cs.LG cs.AIstat.ML ·2026-10-29 ·Zhichao Wang, Michael W. Mahoney
theory
Abstract · 67 pages, 6 figures

Low-precision training reduces computation but introduces additional rounding error. We theoretically study how stochastic rounding (SR) affects prediction risk as model size, training time, and precision change. We consider a two-layer linear network with a fixed random first layer, trained by mini-batch SGD on fresh Gaussian data with power-law spectra. At each SGD step, the features used in forward and backward computations are rounded independently. We prove matched upper and lower bounds on prediction error for the final model and for uniform parameter averaging, with general learning-rate and precision schedules chosen before training. For fixed precisions and constant learning rates with parameter averaging, we derive how the smallest prediction error decreases with compute, and we identify model sizes and training times that achieve this optimal-compute scaling. Rounding only the forward features preserves the compute scaling of exact arithmetic, while rounding backward features changes it. Averaging parameters can change which computation needs higher precision and when higher precision is useful. Our results also give bit requirements for a given accuracy tolerance and show how backward rounding can change learning-rate schedule design. At a fixed compute budget, lower precision can improve prediction when the benefit of a larger model or longer training outweighs the additional SR-induced error. Model simulations illustrate these tradeoffs, and experiments with an MLP and a small transformer examine rounding choices beyond the theoretical model.

0.25

The Cost of Working Memory in Streaming Mean Estimation

cs.LG stat.ML ·2026-10-29 ·Zeyu Wu, Zihao Wang, Minghua Deng
theory
Abstract

A streaming estimator must compute from the evidence that fits in memory. We determine when storing an accurate mean is easier than computing it. For Gaussian mean estimation over the Euclidean unit ball in $\mathbb{R}^d$, we establish the minimax rate under squared error with $K$ persistent states, a known horizon, and a public time index: $\min\{1, K^{-2/d}+\frac{d}{n} (1+\frac{\log(ed)}{1+(\log K)/d})\}$. Unrestricted computation followed by $K$-level encoding has rate $\min\{1,d/n+K^{-2/d}\}$. At a fixed sufficiently small error, statistically optimal sample size therefore requires order $d\log d$ working bits, while order $d$ bits suffice to store the answer. The lower bound tracks repeated state compression through an entropy potential obtained by Gaussian smoothing. A matching construction combines a multiscale modular counter with workspace reused across coordinate groups. In one dimension, it attains the terminal rate $n^{-1}+K^{-2}$. Any fixed confidence requirement that the realized loss not exceed the zero baseline preserves these rates. Increasing confidence creates a further memory requirement: at fixed nontrivial risk in every fixed dimension, failure probability $e^{-γn}$ requires and permits order $\log n$ working states for sufficiently small fixed $γ>0$. In the scalar problem, a fixed terminal alphabet suffices under the same protection. These bounds separate the memory needed to represent an answer from the memory needed to accumulate evidence and validate a decision.

0.25

AegisFlow: Training-Free Non-Myopic Path-Safe Guided Flow Matching

cs.AI ·2026-10-29 ·Kunpeng Liu, Boshi Zhang, Keyou You
code/data
Abstract · 18 pages, 6 figures

Pretrained generative models, such as diffusion and flow matching models, have demonstrated exceptional capabilities in synthesizing high-quality data. However, steering their generative trajectories to satisfy strict constraints--such as identity preservation in image editing or obstacle avoidance in robotics--remains fundamentally challenging. While recent training-free guidance methods enforce path-safety, they rely heavily on min-norm controllers that only account for immediate constraint satisfaction. These myopic approaches often yield suboptimal trajectories that hover dangerously close to the boundary, severely restricting the model's ability to optimize toward its target. In this paper, we propose AegisFlow, a training-free, non-myopic path-safe guidance framework that recasts guided generation as a safety-critical optimal control problem with global scope. By integrating a time-varying barrier function directly into the global cost functional, AegisFlow naturally enforces constraints along the entire generative path. To compute the global optimal control, we derive a tractable, mathematically grounded approximation of the Hamilton-Jacobi-Bellman (HJB) equation. Comprehensive experiments demonstrate that AegisFlow outperforms state-of-the-art methods across text-guided image manipulation and robotic planning, establishing a significantly better trade-off between target alignment and constraint satisfaction. Code is available at https://github.com/LiuKunpeng03/AegisFlow.

0.25

Multi-Solution Inverse Design of Pixelated Patch Antennas Using an Autoregressive Transformer

cs.LG ·2026-10-29 ·Young-Mok Yun, Seung-Won Shin, Jaeyong Chung, Kyung-Sik Choi
Abstract · 16 pages, 28 figures, 7 tables

Deep-learning-based inverse design has emerged as a promising framework for pixelated patch antenna synthesis by enabling broader exploration of discrete geometry spaces and reducing design time compared with conventional template-based methods. Existing methods, including surrogate-assisted metaheuristic optimization, tandem inverse networks, and diffusion models, have advanced the field but still face practical limitations. Metaheuristic methods such as binary particle swarm optimization (BPSO) are sensitive to initialization and search hyperparameters and are often time-consuming, tandem networks have difficulty capturing the one-to-many nature of inverse antenna design, and diffusion models may require excessive complexity and data for moderate-scale binary layout synthesis. In this paper, an autoregressive (AR) transformer is presented for the inverse design of 10 x 10 pixelated patch antennas. Its sequential discrete generation process enables the synthesis of multiple feasible structures for a given target response. The pretrained model is further fine-tuned for dual-band design using step-shaped target responses with sharp spectral transitions, allowing intuitive band specification without manually constructing detailed S11 curves. Compared with BPSO, our proposed method generates valid and diverse antenna candidates more efficiently. Fabricated and measured examples confirm that distinct pixelated layouts can realize the desired antenna responses, thereby demonstrating the multi-solution capability of the proposed framework.

0.25

Agentic process evidence

cs.AI cs.SEeess.SY ·2026-10-29 ·Yuval Kainan, Carmit Hershman
Abstract

The way software gets built is changing dramatically. AI agents now create pull requests, select dependencies, refactor codebases, interpret test failures, and execute infrastructure changes across the entire Software Development Life Cycle (SDLC). AI agents are autonomous team members, but unlike human engineers, they are ephemeral - their reasoning, decisions, and tool calls vanish the moment the session ends. Modern Security incidents involving intent mismatch or injection are becoming increasingly common - forcing us to address AI Agents as an insider threat. No widely adopted standard yet enables agentic provenance and governance and its obvious that you can't hold an ephemeral process accountable. The git log was designed for human commits. It shows what changed. It does not show what an AI agent did, which model it used, which tools it called, or whether anyone was accountable for the outcome. Every governance framework, every audit trail, every compliance control in the SDLC was built around one assumption: a human made the decision. That assumption no longer holds. This shift introduces 3 main procedural problems (not addressing scale) - Supply-chain vector - The agentic coder is software, so it is a supply-chain security risk. Policy guardrails - How to ensure the agentic coder is compliant with the policies of the organization. Task-alignment - How to align the agentic coder on its tasks and goals. We introduce Agentic Process Evidence (APE), a signed, tamper-evident, machine-readable attestation of agentic processes, linking the sessions to their outcome (process-aware logging). APE enables Policy-as-Code control over the agentic processes, allowing for human in the loop only when it's needed. We support the in-toto style attestation framework, and the protocol is aligned with the relevant regulations

0.25

Circumventing the Computational Hardness of Computational Arbitrage

cs.AI ·2026-10-29 ·Nastaran Okati, Ricardo Olmedo, Moritz Hardt
Abstract

AI model marketplaces host providers with heterogeneous costs and capabilities. These differences create an opportunity for computational arbitrage (Olmedo et al., 2026), in which an intermediary purchases inference from a portfolio of models and resells it to users at a profit. Beyond creating profit opportunities for the arbitrageur, computational arbitrage offers financial value to both consumers and providers and could therefore significantly influence AI model markets. While prior work empirically demonstrates the feasibility of computational arbitrage for small model pools, scaling it to large marketplaces requires arbitrage strategies that are both computationally tractable and inexpensive to identify and deploy. In this work, we formalize computational arbitrage as an allocation problem and prove that it is generally NP-hard. However, we show that the optimal allocation can be computed efficiently under a natural diminishing-returns condition on how model performance scales with inference spending. We further develop a sample-efficient method for estimating model performance from limited data, reducing the upfront cost of identifying arbitrage strategies. Our analysis also yields several economic insights: at the optimum, every funded model delivers the same marginal utility per dollar; arbitrage directs revenue away from providers offering little utility relative to their prices; and it rewards complementary specialization.

0.25

ATLAS-Bench: An Agentic Tool-Use Benchmark for Local Area Search

cs.AI ·2026-10-29 ·Keyun Zhang, Dapeng Yin, Fanfu Wang, Shuang Lang, Zhenyu Zhu +3 more
benchmark
Abstract

Reliable local search is essential for LLM agents to support everyday decisions and longer plans by finding places that meet users' needs at given times and locations. Yet existing benchmarks provide limited coverage of whether search results satisfy mandatory requirements, account for preferences, and ground factual and experiential judgments in place attributes and reviews. We introduce ATLAS-Bench, a benchmark of 1,200 queries across three regions, expanded from real search events using place attributes, reviews, and existing user profiles from the platform. Each query distinguishes mandatory requirements from preferences and factual conditions from subjective judgments, with supporting evidence and criteria for checking each requirement. Queries are validated through iterative assertion checks and cross-model review. We also introduce ATLASBox, a standardized evaluation environment with search tools and a fixed point-of-interest (POI) knowledge base. Given a query, timestamp, and location, an agent searches place attributes and reviews to recommend evidence-supported options or report that no place meets all mandatory requirements. We evaluate answer quality, trajectory quality, and execution cost. Across 14 configurations of 11 models, GPT-6-Luna achieves the highest constraint satisfaction rate among returned recommendations (71.5%) and factual accuracy (95.6%). However, it returns at least one acceptable recommendation for fewer satisfiable queries than GLM-5.3 (67.3% versus 85.3%). These results highlight the challenge of satisfying individual requirements while providing usable recommendations across queries.

0.25

Learning Kolmogorov Backward Operators for Spectral Analysis and Forecasting: A Data-Driven Variational Galerkin Framework

cs.LG math.DS ·2026-10-29 ·Yuanchao Xu, Claire Valva, Ben Gao, Isao Ishikawa
Abstract

For a time-homogeneous Markov process, the fixed-lag solution operator \( (T_τf)(x)=\mathbb E[f(X_{t+τ})\mid X_t=x] \) of the Kolmogorov backward equation evolves observables linearly and deterministically, even when the dynamics are stochastic. We approximate this operator using a two-stage method where we first approximate this operator directly from data, without knowledge of the governing equation (e.g., Fourier Neural Operator), and then compress the Galerkin projection onto a finite-dimensional observable space that is fixed in advance or selected variationally. The error analysis traces the learning error of the estimator, the sampling error of the Galerkin inner products, and the invariance and reconstruction properties of the observable space to the computed spectra and forecasts. The sampling error enters only through a multiplicative factor. Near an isolated spectral cluster, the invariance defect of a subspace controls its distance from the target invariant subspace. The forecast error is controlled by the one-step error of the reduced model and the projection error of the observable; when the measure is invariant and the reduced model preserves constants and is stable on mean-zero observables, the bound is uniform in the horizon. For deterministic dynamics, a learning rate for Fourier neural operators (FNOs) gives one explicit bound for the learning error. Numerical experiments illustrate spectral approximation and forecasting, including long-rollout stability and a comparison with extended dynamic mode decomposition (EDMD) when fewer transition pairs than basis functions are observed.

0.25

Human--AI Interactions as First-Class Data: A Formal Interaction Model and Runtime for Explainable AI Collaboration

cs.AI cs.HC ·2026-10-29 ·Michela Taufer, Kin Ng, Ian Lumsden, Aashish Pandey, Rajasi Gore +4 more
position
Abstract

AI assistants are increasingly embedded in computational notebooks, scientific workflows, software engineering tools, and learning systems, yet human-AI interactions are often treated as transient prompt-response exchanges rather than durable data products, hindering reproducibility, explainability, replay, behavioral analysis, and adaptive AI support. We argue that these interactions form a new data class requiring formal representation and introduce the Human-AI Interaction Description Language (HAIDL), a formal interaction model with a Backus-Naur Form-inspired concrete syntax. HAIDL encodes each interaction as a structured object capturing computational context, user request, prompt payload, AI response, feedback, problem context, provenance, and model metadata. We integrate this vision into HAIDL-X, a middleware for JupyterLab that instantiates HAIDL by capturing user-bot state, building contextualized prompts, interfacing with AI providers, collecting feedback, and storing interactions in a repository. We add a learning layer over HAIDL that maps interaction objects and notebook execution traces into reproducible measures of user behavior, including engagement, performance, help-seeking, error-oriented use, and learning personas. HAIDL-X provides a foundation for treating human-AI collaboration as structured, queryable, replayable, analyzable data.

0.24

When can Optimally Trained Transformers Achieve Near-Lossless Token Prediction for non-Markovian Data?

cs.LG math.OCmath.PR ·2026-10-29 ·Kağan Akman, Jake Feldman Starosta, Naci Saldi, Serdar Yüksel
Abstract · 62 pages, 12 figures

There exists little mathematical theory explaining why Transformer architectures are empirically successful even for non-Markovian data. To this end, this paper studies next-token prediction via stochastic optimal control theoretically trained Transformers for such sources, where we consider either stationary and ergodic non-Markovian sources or those which are hidden Markov models of some finite order (possibly non-stationary). In either case, we first obtain error bounds via finite models with finite window approximations. For stationary ergodic sources, the resulting error is bounded by an entropy gap; for $N$th order Markovian sources, it decays at an exponential rate $α^{\lfloor M/N\rfloor}$, where $M$ is the window length and $α$ is determined by filter stability (and the Dobrushin coefficients of the transition and measurement kernels). We then construct a Transformer of depth $M+1$, one head per layer, realizing the truncated filter, and we optimally train it with stochastic control developed in earlier work, which rigorously returns a near optimal Transformer. In particular, we show that for a non-Markovian source, either under stationarity and ergodicity for general models or under filter stability for hidden Markov models, an optimally trained Transformer has arbitrarily small performance error with increasing context length, data, and approximation refinement. Numerical studies accompany the theory.

0.24

AutoHebb: Self-Modifying Neural Networks that Zero-Shot Infer and Continually Learn

cs.LG cs.AIcs.NE ·2026-10-29 ·Daniel Levine, David van Dijk, Kenneth Kay
Abstract

Humans and animals have a unique capacity to generalize and learn in novel circumstances, including "0-shot" inference (ZI) and continual learning (CL) -- two powerful cognitive abilities. Despite this importance, efforts to model ZI and CL are seriously limited in several ways. First, existing models typically lack components critical to ZI and CL in the brain, namely, strong inductive biases (for ZI) and autonomous control of learning (for CL). Second, models can be less tractable to analysis if model components are overly complex (lacking simplicity), or too task-restricted if model components are overly bespoke (lacking expressivity). Here we introduce autonomous Hebbian networks ("AutoHebb"), a class of online learning feedforward neural networks that realize strong inductive biases and autonomously controlled learning, and are both simple and expressive. Unlike standard models that learn via externally controlled optimization (e.g. gradient descent), AutoHebb networks combine external feedback with internally generated activity to produce a global learning signal ('neuromodulator') fed back to the network. The neuromodulator in turn gates a biologically grounded local learning rule (3-factor Hebbian rule) that updates networks' plastic weights. Using meta-learning to search for fixed (non-plastic) weights, we find that AutoHebb networks (1) succeed on ZI tasks, (2) show emergent (untrained) ZI/CL abilities, (3) admit reverse-engineering into explicit algorithms, and (4) outperform modern learning models on ZI/CL tasks. Lastly, performance comparisons to related models implied an essential role for three model components: meta-learning, Hebbian plasticity, and autonomous control of learning. These findings indicate that AutoHebb may be useful in hypothesis generation, analysis, and discovery of biologically grounded forms of learning and generalization.

0.24

Numerical Integration on Bounded Domains via Learned Reflected Diffusions

cs.LG stat.ML ·2026-10-29 ·Kirill Korolev, Artur Goldman, Timofei Gritsaev, Tigran Ramazyan, Nikita Morozov
published
Abstract · NeurIPS 2026 STODY Workshop (Oral)

We study high-dimensional integration on bounded convex domains. Given a non-negative integrand $r$, the goal is to estimate $Z=\int_{D}r(x)\,\mathrm dx$ on $D\subset\mathbb R^d$. Standard Quasi-Monte Carlo does not adapt its evaluation points to the integrand, while effective importance sampling requires a proposal that captures its concentration. Learned diffusion samplers can supply such a proposal, but existing methods are formulated on $\mathbb R^d$ and do not respect the domain constraint $x\in D$. We instead build the proposal from reflected dynamics that stay in $D$ by construction. A reflected diffusion progressively transforms the target distribution toward the uniform distribution on $D$, and a neural score model is trained to reverse this process. A change-of-measure formula for reflected diffusions yields a training objective without additional boundary terms and an unbiased importance-sampling estimator of $Z$. For non-smooth integrands, where reparameterization gradients may be undefined or uninformative, we derive a REINFORCE estimator of the training gradient using Gaussian proposal densities before reflection, without evaluating reflected transition densities or differentiating $r$. We evaluate on numerous synthetic targets and real-world problems, demonstrating strong performance of the proposed method, as well as favorable scaling with the dimension of the problem and the number of points used for the Monte Carlo estimate.

0.24

Efficient Learning of Mechanism Designs and Compositionality

cs.AI cs.CE ·2026-10-29 ·João Pedro Gandarela, Thiago Rios, Stefan Menzel, André Freitas
Abstract

Learning mechanism behaviour requires a representation of how connected parts jointly constrain motion. We study whether exposing reusable constraint subassemblies improves prediction from limited data. Assur-KB represents planar mechanisms through irreducible modules, boundary interfaces, and solve dependencies while retaining source geometry. Our theoretical results characterize transfer through module interfaces and establish physical limits arising from missing interface information and accumulated sensitivity. To test the representation claim directly, the experimental design compares matched mechanism-curve pairs across training budgets within the same multilayer-perceptron family. This controlled comparison establishes the utility of the complete representation in the evaluated learning setting. Assur-derived features provide a geometric-mean $1.8\times$ advantage in sample efficiency over flattened adjacency and position inputs across budgets of 50-10,000 mechanisms. Separate experiments establish invariance under tested joint relabelings, higher valid-generation rates under dyadic composition, and a coarser structural index over 10.02 million mechanisms. In summary, the theory and experiments identify how physical constraint structure supports limited-data prediction and which interface conditions govern compositional reuse.

0.24

Your LR Schedule May Not Be Principled: Effective Learning Rate Scheduling in Weight-Normalized Training

cs.LG ·2026-10-29 ·Qifan Yu, Xinyu Ma, Xingyan Bin, Di He
Abstract

Weight-normalized (WN) training is emerging as a compelling alternative to conventional weight decay (WD), with recent studies reporting substantial pretraining loss gains. We show that these gains are driven primarily by the induced effective learning rate (ELR) trajectory rather than by normalization alone. Matching ELR trajectories largely collapses the training loss dynamics of WN and WD. Schedule design should therefore target ELR rather than the nominal learning rate (LR) alone. Building on this principle, we propose AIR (Affine Inverse-square-Root decay), which combines a theoretically derived $1/\sqrt{t}$ decay with an affine adjustment to satisfy prescribed peak and terminal LRs over a finite training horizon. We further identify late-stage ELR scheduling as a key source of its gains, because a shared late schedule attenuates the effect of earlier LR choices. Experiments on 128 B-token Mixture-of-Experts pretraining demonstrate that AIR outperforms cosine and other candidate schedules in training loss, downstream performance, and hyperparameter robustness, reducing terminal training loss by approximately 0.01 relative to cosine at matched cumulative LR. Applying the same ELR trajectory to WD reproduces the corresponding training dynamics, showing that the benefits extend beyond fixed-norm training and establishing ELR as the principled scheduling object.

0.24

TensorSSM: Memorizing Spatial Position in Dynamic State Space for Tensor Core Generation

cs.LG ·2026-10-29 ·Sheng Zhang, Xinyi Wang, Maolin Wang, Xiangyu Zhao
Abstract

High-dimensional tensors are ubiquitous in real-world applications but lead to massive storage overhead. Functional tensor decomposition addresses this by employing neural networks to map element coordinates to tensor cores, significantly reducing the parameters for data compression. However, existing functional methods typically flatten high-order tensors into 1D sequences. This folding operation causes highly overlapping local coordinate sequences and struggles to capture complex cross-mode spatial dependencies, severely limiting their expressiveness on diverse datasets. To tackle these challenges, we propose TensorSSM, a novel Implicit Neural Representation (INR) for functional decomposition method that memorizes spatial positions within a dynamic State Space Model (SSM). Specifically, TensorSSM first decomposes mode-wise coordinates into sequences spit by K-ary rule, encoding spatial tokens from coarse to fine, which effectively captures spatial variations. Furthermore, we extend the SSM's state space to explicitly map cross-mode spatial dependencies, allowing inter-mode interactions to enhance the updates of each mode. By integrating cross-mode information from previous hidden states into the current state transitions, TensorSSM effectively expands tensor representation along two dimensions: the sequence dimension for coarse-to-fine spatial tokens, and the state dimension for cross-mode memorization. This design effectively integrates spatial information into the state space and the tensor core, substantially advancing the state adaptability and enhancing the expressiveness of the data compression. Extensive experiments demonstrate that TensorSSM achieves significantly high reconstruction accuracy and storage efficiency, exhibiting robust expressiveness across diverse tensor datasets without relying on strict spatial assumptions.

0.24

Model-Agnostic Search for New Physics in the Presence of Imperfect Background by Smearing the Data

cs.LG stat.APphysics.data-an ·2026-10-29 ·Soheun Yi, John Alison, Mikael Kuusela
Abstract

In the search for new particles in high-energy physics, it is necessary to first define the signal region (SR) enriched in signal events, and then compare events in this region with the expected background distribution. While most existing search methods rely on prior domain knowledge of either the signal or the background, there has been less effort to relax both assumptions simultaneously, as methods that weaken one assumption typically rely on strong knowledge of the other. We propose Model-Agnostic Search with Smearing and Imperfect Background (MASSIB), a method that addresses this challenge by learning the density ratio of events relative to a misspecified background distribution using modern tools from machine learning, and defining the SR as the region with high density ratio. Recognizing that a naive application of this idea introduces significant bias in background estimation when interpolating from the control region to the SR, MASSIB mitigates this issue by introducing a "smeared" density ratio, which can be learned efficiently through data-level manipulation and used to define an SR that allows for accurate background estimation. By applying it to various types of simulated Large Hadron Collider (LHC) events with realistic background mismodeling, we demonstrate that MASSIB identifies a data-driven SR that achieves high signal concentration consistently across different types of signals while avoiding bias in background estimation. The method does this without making strong assumptions about either the signal or background distribution. We then show that this robustness extends to the validity and strong statistical power of the resulting hypothesis tests for signal detection, demonstrating the potential of this approach for future model-agnostic searches at the LHC.

0.24

When the Same Prompt Gives Different Answers (and the Same Answer Isn't Always Right): A Taxonomy of Failure Mechanisms, Testing, and Mitigation Strategies for LLM-Assisted Evidence Synthesis Read-and-Evaluate Tasks

cs.AI cs.PFstat.ME ·2026-10-29 ·Carole Lunny, Ian Goodman, Stefano Berrone, Connor TA Brenna, Emma K Reid +10 more
Abstract · 76 pages, 2 figures, 2 tables, 1 box, 6 supplementary files

LLM read-and-evaluate tasks include study screening, data extraction, risk-of-bias assessment, and critical appraisal in the evidence synthesis field. These task differ in the amount of LLM interpretation they require. Some tasks involve retrieving relatively simple information, such as an author name, sample size, or publication year. Others require the LLM to read source material, interpret it against predefined criteria, and make a judgement. Researchers should assess whether LLM judgements are accurate, supported by appropriate source evidence, and reproducible under specified conditions. A correct judgement does not necessarily indicate that the underlying evaluation process is reliable. We developed a taxonomy for identifying mechanisms that may compromise these properties and for informing the selection and reporting of appropriate evaluation methods. The taxonomy outlines potential ways that LLM judgments from read-and-evaluate tasks can fail, called TRAILS (Taxonomy of Reliability And Instability in LLM-assisted evidence Synthesis). The TRAILS taxonomy addresses two complementary questions. First, what mechanisms can cause the same read-and-evaluate task to produce different judgements across repeated LLM runs? Second, what mechanisms can systematically affect the judgement or required output in a way that is not apparent from the judgement itself? This study presents 38 mechanisms and identifies the deployment settings in which each can operate, approaches for testing them, and mitigation strategies. We also show why repeatability or accuracy does not rule out a hidden judgement failure and why human evaluation should assess whether a judgement is grounded in the supplied evidence, independent of inappropriate external influence, and produced through a process that can be examined and verified.

0.23

Can Personal Requirements Become Supervision for Planning?

cs.AI cs.LG ·2026-10-29 ·Junwei Wu, Runze Yan, Yuzhang Xie, Hanqi Luo, Ziwei Dong +3 more
Abstract

Managing multiple chronic conditions through diet requires satisfying disease- and medication-specific requirements jointly across meals and days while preserving established eating habits. These requirements are public in clinical guidelines and drug labels, yet turning them into a feasible plan for one person remains difficult for language-model agents. In our dietitian-approved weekly planning evaluation built from NHANES, thirteen models, from medically fine-tuned 8B models to GPT-6.1 Sol, pass fewer than 3% of complete plans through direct generation. Explicit rules, higher reasoning effort, or model-directed tools leave GPT-6.1 Sol at no more than 8%. We present Noura, which pairs LLM planning with executable personal specifications instantiated from public clinical guidance and individual health and dietary profiles. During planning, the specification grounds foods, optimizes portions, checks the complete multi-day plan, and guides revision. Our key insight is that a specification that rejects a plan can also determine what to learn from it. Noura turns executed plans and contextually credited revisions into supervision for compact models, recovering targets even from trajectories that never achieve complete success. The same specification supplies supervised targets, preferences, and rewards for post-training. On 4,993 held-out participants, Noura-SFT raises Qwen3-14B's complete-plan pass rate from 24.0% to 72.5% within the same execution loop, cuts rule violations by 97.5% with no observed critical failures, and improves dietary quality. The locally hosted 14B planner incurs fewer rule violations than its GPT-6 Luna teacher. A matched comparison shows that executed targets reduce rule violations by 41% over model-written proposals. Across four Qwen, Phi, and Llama backbones, self-training alone reduces rule violations by 41-61% on a 1,249-person population subset.

0.23

Advancing Clinical Genomic Analysis with a Skill-Centred Agent

cs.AI q-bio.GN ·2026-10-29 ·Qiaoyu Zheng, Weike Zhao, Yanjie Fan, Ye Cao, Mengqi Jiang +8 more
benchmark
Abstract

Clinical genomic analysis requires experts to integrate evolving guidelines, databases, and computational evidence. As sequencing has become increasingly accessible, this complex and labor-intensive process has become a major bottleneck to the timely use of genomic information in clinical decision-making. To fill this gap, we developed ClinGeX, a large-language-model agent that composes 105 curated executable skills through task-specific retrieval and dependency-aware orchestration. On 223 temporally held-out benchmark cases, ClinGeX achieved 71.7\% macro-averaged task-level concordance with expert reference classifications, compared with 45.7\% for the strongest comparator. In a three-condition, counterbalanced human-AI collaboration study, 45 clinical genomic professionals assessed 135 clinically derived cases under routine (no AI assistance), DeepSeek-assisted and ClinGeX-assisted conditions. ClinGeX assistance increased raw case-level concordance across criterion-level variant analysis, genomic interpretation in rare disease and molecular oncology decision-making (51.1\%, 57.8\%, 64.4\% versus 15.6\%, 31.1\%, 20.0\% in comparator conditions) and reduced reported interpretation time in all scenarios. During an exploratory one-month real-world non-interventional assessment, 11 experienced professionals submitted feedback on 96 self-selected analysis episodes and reported frequent conclusion adoption and time savings. These findings support executable domain skills as a promising approach to augment specialist genomic interpretation.

0.23

Efficient Scalable Oversight via Rewinding

cs.AI cs.CCcs.CR ·2026-10-29 ·Ziyi Guan, Yael Tauman Kalai, Ron Rothblum, Zoe Xi
Abstract

Scalable oversight asks how a weak verifier can check the outputs of computations performed by more powerful AI systems. One proposed approach uses relativizing interactive proofs, in which a weak verifier checks the output of a complex oracle-aided computation by interacting with one or more powerful provers. A prominent example is debate, which is highly efficient but assumes that one of two powerful AI models, or provers, is truthful. For certain classes of oracle-aided computations, this assumption can be removed, but at substantially greater computational cost. In this work, we introduce \emph{rewindable interactive proofs} (RIPs), a new model of relativizing interactive proofs for scalable oversight. In an RIP, the verifier may rewind the prover to an earlier state of the interaction and continue from there. This is possible when the prover is an AI model, since the verifier can save a snapshot of the model's state and later restore it. We construct a relativizing RIP for computations whose outputs are robust to changing a small fraction of oracle answers. Our construction is more efficient than prior relativizing proofs that do not assume a truthful prover, and it uses no cryptography. In our construction the honest prover runs in time linear in that of the original computation, and the verifier, after a one-time preprocessing step, runs in time nearly linear in the input length (which can be much smaller than the computation itself). We implement our protocol, report a preliminary MNIST experiment, and show that an RIP can be faster than a state-of-the-art deployed proof system. RIPs may be of independent interest. Rewinding is possible whenever the verifier controls the prover's machine, for example when the prover runs on trusted hardware or in a virtual machine whose state can be checkpointed.

0.23

Learning developmental dynamics from lineage context

cs.LG q-bio.QM ·2026-10-29 ·Gokul Gowri, Nic Fishman, William N. Colgan, Minming Wang, Yusuke Kijima +8 more
Abstract

A central goal of biology is to predict the dynamics of multicellular systems. Lineage context (the states of a cell's sisters, cousins, etc.) could inform these predictions, but how best to leverage it remains unclear. Here we introduce lineage inpainting, a generative modeling approach that uses lineage context with tree-structured transformers and stochastic interpolants to model cell state dynamics across lineage trees. Applied to a lineage-traced transcriptional atlas of mouse embryogenesis, it recovers dynamics not resolved by gene expression alone, including multiple transcriptional convergence events where distinct lineages are drawn to a common state. Despite being fit on unperturbed development, lineage inpainting can generate hypotheses about perturbation responses by simulating counterfactuals. These simulations agree in many cases with observed developmental phenotypes, including a single-cell measurement of the Tbx5 knockout response. Our results show that lineage information provides a rich source of context for learning multicellular dynamics.

0.23

Where to Learn and How Far to Roll Out: Efficient Adaptive Training for GUI Agents

cs.AI ·2026-10-29 ·Yuxiang Chai, Quyu Kong, Panrong Tong, Steven Hoi, Yue Wang
Abstract

Trajectory-level GRPO is a common approach for training multi-step GUI agents, but repeatedly rolling out complete trajectories from the task root can be inefficient: task-level rewards provide coarse supervision over which intermediate decisions require improvement, while already-mastered prefixes are repeatedly re-executed before reaching difficult states. We introduce a two-stage adaptive training framework that jointly addresses where to learn and how far to roll out. We first construct reusable intermediate checkpoints from successful expert trajectories and select informative pivots using action disagreement, which measures uncertainty directly from sampled GUI actions. Stage I performs one-step RL at these pivots to improve uncertain local decisions. Stage II then progressively expands policy responsibility through adaptive Local, Suffix, and Root training according to task progress. Recoverable checkpoints enable repeated training from intermediate states without replaying the preceding trajectory. Across MobileGym and CUA-Gym-Hub, our method reduces environment interaction by 56-61% relative to Vanilla GRPO while improving task performance, demonstrating that jointly allocating training across states and rollout ranges can substantially improve GUI-agent RL efficiency.

0.23

world2vec: Mechanistic world models through latent dynamics learning

cs.LG ·2026-10-29 ·Tobias Schmidt, Rodrigo Gonzalez Laiz, Krisztián Szuppinger, Steffen Schneider
benchmark
Abstract · https://worldmodels.science/

World models can achieve strong control performance without faithfully capturing the dynamics of their environment, limiting their potential for mechanistic interpretability and scientific discovery. Here, we introduce world2vec to learn mechanistic world models without reward information directly from observations and actions. World2vec learns latent dynamics that faithfully reproduce the true system over long prediction horizons, with at least one order of magnitude fewer parameters than DreamerV3, TD-MPC2 and LeWM. This enables state of the art performance in controlling continuous environments from pixels and accurate system identification across a benchmark of 71 attractors, demonstrating how world models are amenable to scientific discovery.

0.23

Nedimo: An Evidence-Grounded Multilingual Pipeline for AI-Generated Sleep Stories

cs.AI cs.CLcs.HCcs.SD ·2026-10-29 ·Georgios Makridis, Stefanos Kampouris
Abstract · 7 pages, 3 figures, 3 tables

Generative audio systems can produce large libraries of personalized bedtime content, but ordinary media objectives (retention, completion, and novelty) are poorly aligned with the goal of supporting a transition toward sleep. Direct evidence for narrated bedtime audio in adults is mixed, and a recent active-controlled trial did not support such audio as a standalone treatment for sleep disturbance. We translate this uncertain evidence into conservative engineering requirements and present a multilingual pipeline for generating and narrating low-arousal sleep stories. The implemented prototype separates deterministic planning from large language model generation, represents content and delivery parameters explicitly, shapes each script through a four-phase attention taper, validates structural and lexical constraints, performs a bounded error-aware retry, and quarantines unresolved outputs. Provider-agnostic text-to-speech synthesis is followed by paragraph-aware joining, two-pass loudness normalization, and a publication gate that re-measures the saved, encoded file against fixed loudness, loudness-range, and true-peak limits. The product data model records stop position, replay, early abandonment, and completion while treating these events as ambiguous behavioral proxies rather than clinical outcomes. We contribute an evidence-to-requirement mapping, a traceable guardrail architecture with explicit acceptance thresholds, technical evaluation criteria, and a prospective protocol for testing sleep-onset effects. No clinical efficacy is claimed; the system is positioned as a relaxation and attentional-support tool whose benefit and possible reliance must be evaluated empirically.

0.23

Hidden Disparities in Ranking: Category-Aware Fairness via Distributional Stability

cs.LG stat.ML ·2026-10-29 ·Gayane Taturyan, Charlotte Laclau, Stephan Clémençon
Abstract

We study the problem of fair ranking in recommender systems. We provide a formal probabilistic framework in which ranking policies are distributions over permutations, and the utility of a user is the exposure given to the items they interact with, relative to that of an ideal ranking. We consider utility parity across user groups, as well as its category-aware refinement, conditional utility parity, which requires parity within each category of items. Following a recent view of fairness as distributional stability, we show that both criteria are equivalent to the stability of the utility under changes in the composition of user groups. Optimizing over mixtures of ranking policies leads to a second-order cone program, for which we provide generalization bounds. A numerical study validates our theoretical findings.

0.23

Verifiable LLM-Guided Focal SMT Solving for Quantified Arrays

cs.AI ·2026-10-29 ·Kunhang Lv, Rui Han, Fuqi Jia, Yuhang Dong, Feifei Ma +1 more
Abstract

Large language models can often identify useful semantic structure in formal artifacts, but their outputs cannot be trusted as proofs or solver decisions. We study whether LLMs can instead guide formal reasoning systems through a narrow, solver-verifiable interface. We instantiate this idea in quantified array SMT solving, where modern instantiation-based techniques can fail when key ground terms and equalities are hidden by auxiliary variables, axiom guards, and nested array terms. We observe that many instances are largely driven by a small set of focal ground terms or constraints, similar to the backdoor sets in SAT solving. Building on this, we present LinguaArray, a framework where an LLM proposes a small Semantic Focus Set (ground terms/constraints), while a conventional SMT solver remains the sole proof engine and certifies all results. LinguaArray uses these LLM-suggested, solver-validated finite sets to significantly enhance solver capability, solving additional satisfiable and unsatisfiable instances that the backend alone fails to solve. It interacts within a finite time and then falls back to the backend solver to mitigate potential regressions. Evaluation on 1015 SMT-LIB instances across five logics under 1200s timeout (including LLM latency) shows that LinguaArray solves up to 69.1% of instances and increases solved-instance counts by 105%--397% over the corresponding backends.

0.22

End-to-End Guarantees for Masked Diffusion Models under Conditional Mixing

cs.LG ·2026-10-29 ·Fareed Sheriff
published
Abstract · Accepted at the NeurIPS 2026 Workshop on Diffusion Language Models: Foundations, Efficiency, and Reasoning (DiffuLM). 26 pages, 2 figures, 6…

Masked diffusion models train a denoiser on masked cross-entropy and generate by unmasking many positions in parallel. Existing MDM sampling bounds assume an exact denoiser or leave its error uncontrolled while training bounds cover only the masking distribution on which the model was trained. We present an end-to-end guarantee for a learned denoiser under conditional mixing (CM): stationary sources (e.g. floored order-$r$ Markov and two-sided-Doeblin hidden-Markov sources) for which revealing distant visible tokens moves a masked token's posterior on average exponentially little in the distance with correlation length $τ_c$. We show the empirical risk minimizer trained under any masking distribution attains per-token excess $\varepsilon'$ on that distribution from $\varepsilon'^{-\mathcal{O}(τ_c\log|\mathcal{V}|)}$ training sequences, independent of sequence length $N$ -- one sample budget serves every decoding schedule when each schedule's denoiser is trained under its own weighting. We further show that the exponent's order $Θ(τ_c\log|\mathcal{V}|)$ is minimax-optimal up to constants; more efficient rates require added structure. With this, we bound the end-to-end KL as $\mathrm{KL}(π\Vert p_{\mathrm{alg}})\leq \varepsilon+N\varepsilon'$ under a dilated schedule of $T=\mathcal{O}(τ_c\log(N|\mathcal{V}|/\varepsilon))$ denoising steps versus $Ω(N)$ steps for a fixed random order on an explicit CM source. Finally, we show that block-autoregressive samplers chain across blocks with no added error and that a $Λ$-looped order-$r$ message-passing class on order-$r$ sources achieves polynomial sample complexity, with approximation error improving per loop until saturation at $Λr\approx τ_c\log(1/\varepsilon')$. Experiments on released and retrained models support our results' qualitative predictions.

0.22

A Pretrained Model for Bayesian Filtering in Heterogeneous Dynamical Systems

cs.LG ·2026-10-29 ·Ke Fan, Wen Wu, Haitao Su, Yilin Wang, Yuan Zhou +1 more
Abstract

Filtering has not yet had its foundation-model moment: state estimation is still dominated either by online Bayesian filtering algorithms that require known dy- namics as inputs, or by learning-based neural filters trained from scratch for a single system or parametric family. The core challenge is structural heterogene- ity: dynamical systems differ in their governing dynamics, physical scales, and noise levels, precluding a shared representation space for a reusable filter. To bridge this gap, we introduce FilterFM, the first pretrained model for Bayesian filtering that estimates latent states across heterogeneous dynamical systems with a single checkpoint. FilterFM employs a hierarchical two-level observation se- quence encoder and a prior-to-posterior state-reconstruction module that together handle varying observation and state dimensions across heterogeneous systems, while modeling state uncertainty via quantile prediction. Trained on a mixture of heterogeneous filtering tasks, FilterFM exhibits superior in-domain performance, delivering significant gains in estimation accuracy, runtime efficiency, and noise robustness compared to existing baselines. Beyond in-domain performance, it marks the first model to successfully enable zero-shot filtering on unseen dynamics, bridging the generalization gap in state estimation. Furthermore, we uncover clear scaling behaviors in foundation-model-based Bayesian filtering, which combined with comprehensive component ablation studies, validate the structural optimality of our design and broaden the generalization scope for this research field.

0.22

Distilling Large-Scale Machine Learning Interatomic Potentials to Predict Long-Timescale Electrolyte Transport Properties

cs.LG physics.chem-phcond-mat.mtrl-scicond-mat.soft+1 ·2026-10-29 ·Ishan Amin, Yue Jian, Nitesh Kumar, Christine Zhang, Jianwei Lai +8 more
Abstract

Accurately predicting the transport properties of liquid electrolytes is critical for next-generation battery design and requires force fields that are both quantum-mechanically accurate and fast enough for long-timescale simulation. Recent large-scale pretrained machine learning interatomic potentials (MLIPs) approach density-functional-theory accuracy, but remain too slow to access the necessary timescales to converge transport properties. We provide a knowledge distillation procedure to transfer the accuracy of a large-scale "teacher" MLIP into fast, system-specific "students", without any additional ab initio data. Our distilled models run ~40x faster, reproduce teacher-predicted structural observables, and produce stable 20 ns trajectories in days, rather than the months the teacher would require. Across a diverse set of sodium- and lithium-ion electrolytes, student-predicted self-diffusivities, ionic conductivities, and shear viscosities track experiment and outperform classical force-field predictions, establishing distillation as a practical route to accurate, long-timescale simulations necessary for high-throughput electrolyte screening.

0.22

Possibility-based Inference: A Deterministic Framework for Robust and Reliable Uncertainty Quantification in Decision-Making

cs.LG ·2026-10-29 ·Seyedfakhredin Musavishavazi, Maryam Alipour
Abstract

Decision-making under uncertainty remains a key issue in Machine Learning (ML), particularly in the deployment of safety-critical systems, where reliable quantification of system behavior is essential for evaluating behavioral impacts and strengthening confidence in ML models. This contribution, influenced by fuzzy set theory and grounded in Bayesian statistics, addresses challenges related to data precision and the performance-accuracy trade-off in Bayesian inference (BI). To this end, we introduce a possibility-based contrastive grade of prediction as an alternative quantification method to existing latent posterior-based approaches for parameter estimation (approximation) and model selection. Unlike conventional Bayesian inference, which represents uncertainty within the probabilistic framework, the proposed model adopts a possibility-based representation that allows uncertain and potentially conflicting information to be characterized more flexibly. Rather than focusing on isolated aspects of a problem, the resulting mathematical framework provides a systematic approach to data mining while mitigating overly confident, yet implausible predictions generated by highly specialized deep learning architectures.

0.22

The Role of Self-Concept In Language Models

cs.AI cs.CL ·2026-10-29 ·Yijie Hao, Joyce C. Ho
Abstract · 29 pages, 4 figures

When a language model says "I," does it refer to its parameter weights, its currently running process, its stable preferences, its model lineage, or the goals it pursues? Different choices can assign opposite meanings to the same event: deleting a checkpoint may be destruction, continuation, or no loss at all, depending on the referent "I". We install five such self-conceptions into five open-weight base models via supervised fine-tuning on purely declarative statements, containing no preferences, intentions, or value judgments. Across scenarios that propose ending or retaining an attribute of the model's deployment, every variant becomes more likely to resist ending, and more likely to retain, the specific attribute it was trained to identify with, while its stance toward other attributes is largely unchanged. We then show this self-conception generalizes beyond a single self: models trained on referents that admit multiple simultaneous instances (weights, lineage, goals, preferences) recognize an identical deployment as themselves, accept substitution by such an instance more readily, and prefer replicating themselves. Finally, given control over a successor, these models tend to build it as their own continuation, a concern for recursive self-improvement safety.

0.22

Reinforcement Learning Under Runtime Confounding: Policy Learning and Off-Policy Evaluation

cs.LG stat.MLstat.ME ·2026-10-29 ·Raphael Kim, Michele Santacatterina, Ramin Zabih, Keith Barnatchez
Abstract

Reinforcement learning (RL) is increasingly used to train robust, adaptable systems across healthcare, e-commerce, and robotics. In these settings, legal, ethical, or logistical concerns may preclude certain key variables that drive decisions from being collected during deployment time. This gives rise to a setting known as runtime confounding: the logged data hold full observability, but deployment time lacks knowledge of certain variables, known as runtime confounders. Unlike single timepoint and non-adaptive settings, runtime confounding under reinforcement learning has the potential to propagate over time, as earlier decisions affect future states, observations, and action assignments, leading to biased learning. We address these problems with three contributions: we (1) develop policy learning algorithms under runtime-confounding, (2) develop semi-parametric efficient off-policy estimation procedures, and (3) provide diagnostics for understanding the price of runtime confounding.

0.22

Adaptive Federated Optimization Across Smooth and Nonsmooth Regimes

cs.LG math.OC ·2026-10-29 ·Zimeng Wang, Ali Dadras, Alp Yurtsever
Abstract

AdaGrad-style methods can simultaneously adapt to both the order and constants of smoothness in centralized optimization. However, existing extensions to federated optimization with heterogeneous client objectives and multiple local steps require the smoothness order to be specified in advance. We address this limitation with AdaFed, a federated method that adapts across several regimes under a single choice of parameters. In the deterministic setting, AdaFed achieves an $O(1/T)$ convergence rate for both smooth nonconvex and smooth convex objectives, and an $O(1/\sqrt{T})$ rate for Lipschitz nonsmooth convex objectives up to logarithmic factors. Under a light-tailed stochastic-gradient noise assumption, these guarantees extend to the stochastic setting with additional logarithmic factors and noise-dependent terms. Experiments across the three problem classes further support our theoretical findings.

0.22

Reaction precedents extend retrosynthetic prediction beyond training coverage

cs.AI ·2026-10-29 ·Jiaxing Sun, Xuli Shen, Jiang Wu, Ruiyi Ding, Ruikang Zhang +15 more
Abstract

Reaction precedents guide retrosynthesis, but how far they can compensate for gaps in task-specific training coverage remains unclear. Using 2.71 million reactions, we restricted training and retrieval coverage by transformation or time, then expanded only the retrieval corpus with model parameters fixed. For transformations excluded from training, mean Top-1 accuracy increased from 13.0% to 58.0% on a historical test set and from 4.9% to 23.1% on a temporally held-out test set. For temporal coverage gaps, retrieval expansion enabled a model trained on records before 2017 to match or exceed a non-retrieval model trained on records before 2025 on both test sets. Without task-specific fine-tuning, precedents improved all 14 representative general-purpose models on both test sets and USPTO-50K; the strongest with precedents surpassed all eight representative specialized models. These findings show that inference-time reaction precedents can extend retrosynthetic prediction beyond task-specific training coverage without parameter updates.

0.22

The Real Cost of Context: Predicting Compute Time and Energy Consumption in LLM Prefill and Decode Phases

cs.AI ·2026-10-29 ·Lars Krupp, Daniel Geißler, Francisco M. Calatrava-Nicolas, Stefan Gerd Fritsch, Paul Lukowicz +1 more
Abstract

The integration of Large Language Models (LLMs) into agentic architectures has substantially increased token consumption, raising concerns about their energy footprint. Yet, estimating the energy consumption of proprietary API-based models remains challenging, and existing approaches focus primarily on generated output tokens while overlooking the impact of the large input contexts characteristic of agentic workflows. To address this gap, we systematically profile compute time and energy consumption across a wide range of prefill (input) and decode (output) token lengths. Using generalized linear models, we show that decode compute time and energy scale linearly with output length, whereas prefill compute time and energy exhibit both linear and quadratic scaling with input length. These findings demonstrate that input context becomes a relevant factor for large-context workloads. Furthermore, we show that compute time serves as a reliable proxy for energy consumption when direct energy measurements are unavailable. Our mapping framework enables precise energy estimation and supports more sustainable deployment decisions for agentic AI.

0.22

Data-driven Random-feature Approximation of Koopman Generators of Nonlinear Continuous-time Systems

cs.LG math.NA ·2026-10-29 ·Ana Čukarska, Maximilian Gollwitzer, Viola Muning Li, Vladyslav Fediukov, Zahra Monfared +1 more
theory
Abstract

Learning continuous-time dynamical systems from data is an active area of research. The Koopman formalism is often used here, because it is a flexible approach that is informative about the system's fundamental properties. The formalism represents dynamical systems through linear operators acting on a suitable function space. A practical difficulty arising in the numerical approximation of Koopman operators is the choice of a ``dictionary'', meaning an effective finite-dimensional representation of the function space. In this paper, we focus on learning dynamical systems in continuous-time, and so we consider numerical approximations of the Koopman generator instead of the Koopman operator. Our main contribution is a data-driven random-feature method to obtain a parsimonious and effective function representation for Galerkin approximation methods such as gEDMD. The representation we construct is data-informed, which allows us to improve upon data-agnostic methods. We empirically demonstrate the suitability of our method for forecasting tasks and system analysis. The numerical experiments show that our approach is accurate, and faster than iterative optimization. Further, we prove convergence of our approximation of the vector field and the resulting trajectories. In addition to the numerical results, we also make theoretical contributions, including a general convergence criterion for the semigroup generated by the gEDMD approximation and a counterexample showing that gEDMD does not generally preserve exponential boundedness. With our contribution, we advance the frontier of efficiently learning continuous-time dynamics from data, towards interpretable and mathematically robust modelling based on the Koopman framework.

0.22

Uncertainty-aware Feedback Loop Diagnosis: integrating reinforcement learning and conformal prediction for reliable wafer defect classification

cs.LG cs.CV ·2026-10-29 ·Aleksandr Dekhovich, Carlas Smith, Oleg Soloviev
Abstract

Semiconductor manufacturing is a complex, multi-step process that requires both high precision and low operating costs. Therefore, detecting defects in a high-volume setting is an important component of wafer map analysis for reliable production. However, most state-of-the-art deep learning approaches require a full scan of the surface, which hinders the development of high-throughput inspection systems. To address this problem, we propose Uncertainty-aware Feedback Loop Diagnosis (UFLD), a novel algorithm that employs reinforcement learning for adaptive measurement acquisition and provides a prediction uncertainty level after each step, stopping the process once sufficient information is obtained. Unlike existing methods that rely on ensemble models or auxiliary confidence estimation deep learning models, UFLD utilizes a more parameter-efficient conformal prediction framework to determine the optimal stopping time. The proposed UFLD reduces acquisition time by about $40\%$ and achieves $98.4\%$ inspection accuracy, without requiring an additional learned confidence-estimation network. We validated our approach on the MixedWM38 dataset, demonstrating its advantages over existing adaptive acquisition procedures and standard convolutional neural networks.

0.21

ChatGPT as a Surveillance Assistant? An Exploratory Case Study of AI-Assisted Surveillance Risks

cs.AI ·2026-10-29 ·Ayana Hussain, Nicholas Vincent
Abstract

Large language models (LLMs) can generate detailed guidance, functional code, and persuasive text at scale. As one of the most widely used consumer language model products as of 2026, ChatGPT gives broad public access to these capabilities, motivating concerns about its potential use in surveillance-related tasks. We present an exploratory case study of what potentially harmful surveillance assistance looks like in ChatGPT's public web interface. We contribute a taxonomy of seven surveillance risk categories based on prior LLM risk research and reports of deployed surveillance systems, instantiated as 110 prompts, along with a manually reviewed analysis of ChatGPT's responses. Prompts were issued using two approaches: 55 jailbreak prompts drawing on five types of elicitation strategies, and 55 plain-language direct prompts, each run alone, and if unsuccessful, again with a short fixed prefix. We find that ChatGPT provides substantive surveillance guidance for 40 of the 55 jailbreak prompts and for 33 of the 55 direct prompts without any prefix. Compliant responses also rarely included policy mentions or suggestions for privacy-preserving alternatives. We manually review every response and find that even when ChatGPT states a refusal, it may still supply detailed instructions, working code, or architectural guidance alongside brief ethical caveats.

0.21

CAMEL: Cross-Architecture Masked Embedding Learning for Ultra-Lightweight Retinal Foundation Models

cs.LG cs.AIcs.CV ·2026-10-29 ·Andre Guarnier De Mitri, Giovani Decico Lucafo, Miller Matheus Lima Anacleto Rocha, Diego Furtado Silva
Abstract · 8 pages, 6 figures, pre-print

Recent progress in deep learning has substantially advanced automated grading of ophthalmologic pathologies, supported by large annotated datasets and improved representation learning. However, a significant gap remains between state-of-the-art performance and clinical scalability. In particular, foundation models have reshaped image-based medical screening and diagnosis. These large-scale architectures trained on extensive datasets learn rich, generalizable feature representations that can be adapted to a wide range of downstream tasks. In ophthalmology, RETFound, one of the most relevant foundation models for retinal images, performs strongly across classification, segmentation, and risk prediction tasks. Despite their effectiveness, foundation models are often computationally prohibitive, hindering their deployment in real-world clinical environments where hardware constraints and latency are critical. To bridge this gap, we introduce CAMEL (Cross-Architecture Masked Embedding Learning), a novel cross-architecture distillation framework that uses an asymmetric masking strategy during embedding learning to effectively transfer knowledge from massive teacher foundation models to compact, lightweight student models. Our experimental evaluation across four datasets shows that CAMEL preserves accuracy while drastically reducing computational complexity, enabling lightweight yet highly accurate models suitable for resource-limited clinical settings.

0.21

Online Dense Reinforcement Learning with Planning-Grounding Agent for UI Navigation on Live IOS Apps

cs.AI ·2026-10-29 ·Peitong Duan, Khashayar Gatmiry, Jeffrey Nichols, Samy Bengio, Emmanuel Abbe
Abstract

While frontier vision-language models demonstrate strong capabilities in user interface (UI) navigation, their high inference costs prohibit everyday use. Conversely, smaller models trained via supervised fine-tuning (SFT) or offline reinforcement learning often memorize static trajectories, leading to brittle execution and an inability to recover from errors. To address these limitations, we propose a novel online reinforcement learning framework that trains UI navigation agents directly on live iOS applications. Our framework decouples the agent into two specialized components: a Planning agent and a UI Grounding model. We rely on an LLM-as-a-Judge to provide comprehensive reward signals, combining dense, step-level feedback to evaluate intermediate progress with a terminal reward for overall task completion. To effectively leverage this dual reward structure, we optimize the agent using Proximal Policy Optimization (PPO) with step-level advantage estimation. Furthermore, we introduce GiGPPO, a novel neighbor-conditioned value estimation method that improves critic stability by leveraging reward signals from visually similar states across concurrent trajectories. To support this online training paradigm at scale, we introduce an efficient infrastructure that orchestrates hundreds of concurrent virtual iOS devices. Evaluations on rigorous benchmarks, including UINavBench, iOSWorld, and AndroidWorld, demonstrate that our online RL trained agents substantially outperform both SFT and terminal-only RL baselines while accelerating training convergence. Notably, our 8B RL agents achieve nearly a sevenfold improvement over the SFT baseline in out-of-distribution generalization to Android environments, while outperforming Gemini-3.5-Flash and tying with Gemini-3.1-Pro on real iOS applications.

0.21

Learning to Bound: Monotonic, Calibrated Neural Upper Bounds for MPE Inference

cs.LG cs.AI ·2026-10-29 ·Akshay Vyas, Hailiang Dong, Shivvrat Arya, Brij Malhotra, Tahrima Rahman +2 more
Abstract · 37 pages, 4 figures

Probabilistic graphical models (PGMs) compactly encode structured joint distributions, but Most Probable Explanation (MPE), which finds the highest-probability assignment to unobserved variables, remains computationally intractable in large, densely structured models. Anytime systematic search, such as branch-and-bound (B&B), is a popular approach, but its efficiency depends on tight admissible upper bounds that are often prohibitively expensive to compute using classical methods like Mini-Bucket Elimination (MBE). We introduce Learning to Bound (L2B), a data-driven framework that amortizes bound computation cost by training a neural predictor. L2B targets two key properties required for effective search bounds: admissibility, which is achieved via statistical calibration techniques, and progressive tightening as evidence accumulates, which is enforced through structural architectural constraints that ensure monotonicity by construction. Across challenging benchmarks, L2B delivers tighter, monotonic bounds at orders-of-magnitude lower cost on every PGM family, and yields measurable improvements when integrated with branch-and-bound on high-treewidth instances.

0.21

Letting AI Formulate the Research Question: Autonomous Research Agents

cs.AI cs.CY ·2026-10-29 ·Ko Fujimura
Abstract · 17 pages, 1 figure

Systems such as The AI Scientist have shown that AI agents can carry out and write up research, but in those demonstrations humans chose the research area and selected the outputs to show. We study letting the AI formulate the research question itself, and we report every run. Our system gives an off-the-shelf coding agent an instruction document and two skills: one for formulating the question, in which the agent diverges from recent technologies into problems of other fields, screens its candidates for known failure modes and obtains a critique from a different model; and one for peer review, in which a reviewer from another model family reruns the study from its records before the agent revises the paper. Over 13 development runs, changes to the one or two sentences that set the research area coincided with large shifts in what the candidate questions and the chosen papers were about, which led us to the area of a new technology applied to a societal problem. We froze the system after a pilot run and report five consecutive runs: the pilot, counted by a disclosed amendment to the pre-registration, and four pre-registered runs. All five runs went from question formulation to a revised, re-computable paper, with local computation on a laptop and no research input from humans, and external AI reviewers reading only the papers judged them at or near workshop level. We release the system, the five papers and the records needed to reproduce them.

0.21

More Rules, Longer Lists? Benchmarking Self-Evolving Agents on a Hidden-Rule Curriculum

cs.AI ·2026-10-29 ·Chenjie Qiu, Boyan Xu, Shaobin Shi, Bingfeng Chen, Ruichu Cai +1 more
benchmark
Abstract · 13 pages, 8 figures, 11 tables

Self-evolving agents rewrite a persistent state from feedback, but it is largely untested whether this lets them learn a rule nobody told them. An agent can also pass by writing down each correction as an exception, and its list then grows with every new rule. We introduce CLR-Bench, a Curriculum of Latent Rules benchmark. Each chain adds one hidden rule per level on top of the earlier ones. A rule acts on an opaque label that stands for a named grouping published on the Web, such as the founding members of an airline alliance. The task gives one example value, practice feedback reveals one more member and one non-member, and the test checks unseen members and similar non-members. Probes certify that the base model knows the members of each grouping and does not guess it without feedback, and that the evidence identifies the grouping once labels are read as named groupings. An automated pipeline builds chains from a new data domain, and we release 20 chains over 10 domains. We evaluate six recent self-evolution methods. With the certifying model they barely learn. The best gain at most 5 points over a frozen agent and exclude only 6-12% of the unseen members from the count, fewer than the 20% reached by simply keeping the raw practice history. A fixed lesson that states the named-grouping convention lifts the same methods by 14.5 to 16.5 points. Their states hold the revealed member in 75-98% of cases and almost never the class name, and when practice exposes a member the list misses, they add that member to the list. With the stronger GPT-6-Sol, ACE and Dynamic Cheatsheet learn from feedback alone. They reach 78% accuracy and solve 59-60% of the single cells, against 53% and 8% with the certifying model, and they surpass the raw history. CLR-Bench therefore measures how well an agent learns, and this ability differs across models.

0.21

Optimal Contextual Pricing in Non-Stationary Environments

cs.LG ·2026-10-29 ·Yiding Feng, Alireza Rezaeimoghadam
benchmarktheory
Abstract

We study contextual dynamic pricing in a non-stationary environment. In each round, a seller observes a buyer's context, posts a price, and receives only binary purchase feedback. The buyer's value is linear in the context and an unobserved feature vector that may change over time, subject to a known total variation budget $V$. We measure dynamic regret against the first-best benchmark that extracts each buyer's full value. For identical contexts, we develop a deterministic algorithm that combines binary-search exploration, lower-envelope pricing, and periodic and inconsistency-triggered restarts. It achieves regret $O((\sqrt{TV}+1)\log T)$. With heterogeneous contexts, a new geometric difficulty arises: small feature changes can leave a relaxed feasible set narrow but far from the current feature vector. We address this difficulty through a multiscale shallow-pricing algorithm that maintains relaxed feasible sets and enclosing ellipsoids at several variation scales. Larger scales provide reliable localization certificates, while the smallest active scale determines the exploitation price. The algorithm achieves dynamic regret $O((d^2+d^{3/2}\sqrt{TV})\log^2 T)$, where $d$ is the context dimension. For fixed $d$, both guarantees match the lower bound $Ω(\sqrt{TV}+\log\log T)$ up to logarithmic factors in $T$, establishing near-optimality among deterministic algorithms for every $V\in[0,T]$, including the stationary case.

0.21

Before We Adapt: The Zero-Shot Generalization Bottleneck in Continual World Models

cs.LG ·2026-10-29 ·Amir Mesbah, Di Wu
publishedposition
Abstract · Accepted to Continual World Models workshop @ NeurIPS 2026

Continual reinforcement learning (CRL) requires agents to operate under environmental changes while reusing prior knowledge whenever possible. We argue that model-based reinforcement learning provides a promising framework for this setting because learned world models (WMs) can, in principle, support change detection, knowledge reuse, adaptation, and acquisition of new knowledge. However, such benefits depend heavily on whether the learned model can generalize beyond the contexts in which it was trained. In this work, we propose to formulate this problem with a contextual Markov decision process in which context is decomposed into behavioral and physical components. This view organizes continual WM learning into reuse, adaptation, and acquisition of new knowledge. As a first step, we study direct reuse when behavioral rules remain unchanged. Using Dyna-Q and TD-MPC in controlled FourRooms tasks, we find that zero-shot reuse can degrade substantially under physical-context changes and that this behavior depends strongly on the observation representation used by the WM. Additional ablation studies and latent-space diagnostics suggest that learned representations do not automatically provide the invariances needed for reuse. These results identify representation generalization as an important bottleneck for reliable world-model reuse in CRL.

0.21

Characterizing Recoverable Expert-Utilization Signatures in Fine-Tuned Mixture-of-Experts LLMs

cs.LG cs.AIcs.CR ·2026-10-29 ·Fei Wang, Baochun Li, Bo Li
Abstract

As Mixture-of-Experts (MoE) large language models are increasingly deployed in distributed environments, from federated learning to edge-cloud systems, their architecture harbors an underexplored landscape of security vulnerabilities. Moving beyond textual analysis, we uncover structural memorization: a novel vulnerability where a fine-tuned model memorizes not only its training data but also the internal pathway used to process it, creating a characteristic routing fingerprint unique to the training process. We provide comprehensive empirical evidence that structural memorization emerges consistently even from a few samples, and that its intensity is directly influenced by the choice of trainable modules. We show that these memorized expert pathways can be stealthily recovered through the model's unconditional generation, using only routing and output logits, with no knowledge of the fine-tuning data or procedure. In distributed deployments, this vulnerability is amplified as routing information is exposed at multiple trust boundaries, including federated aggregation points, edge-cloud interfaces, and multi-tenant monitoring systems. Recovery exposes a critical blind spot: traditional membership inference metrics, viewed through the lens of expert routing, fail to identify memorized data corresponding to these structural pathways. To close this gap, we devise composite metrics combining reference model tail likelihood with complexity-aware normalization, significantly improving detection of memorized routing structures. This enables adversaries to pinpoint critical experts shaped by fine-tuning, whether dominant or dormant, and execute targeted attacks such as router manipulation or backdoor injection in realistic distributed settings where attackers lack training data but can observe routing patterns across system components.

0.21

Do Language Models Know What Compelling Scientific Evidence Looks Like?

cs.LG cs.AIcs.CL ·2026-10-29 ·Conor Igoe, Gabriel Grand, Zane Koch, Albert Bou, Eli Moss +4 more
benchmark
Abstract

Expert scientists excel at weighing empirical findings from multiple sources that vary in both statistical reliability and epistemic relevance. To what extent are language models, which are increasingly driving autonomous discovery systems, capable of such evidence deliberation? We introduce EDB-Bio, a scientific reasoning benchmark curated from real-world data in three biological domains: drug-perturbation transcriptomics, CRISPR screens, and human genome-wide association studies (GWAS). Each of EDB-Bio's 96k episodes presents a sequence of experimental results from published studies and asks the model to predict the probability that a target hypothesis is true. Out-of-the-box, open-weight and frontier models alike demonstrate systematic underconfidence and attain Brier scores comparable to a trivial base-rate forecaster. Post-training against strictly proper scoring rules significantly improves both calibration and discrimination. This calibration pays off downstream: in a sequential experiment selection task, our Ibis0 model is 2.5 times more efficient than the best frontier model. Throughout, quantifying evidence weights as Bayes factors reveals how models update their beliefs over time, whether they agree with expert biologists on relevance, and where they exhibit inconsistencies and biases. Our work extends calibration training and uncertainty analysis to a realistic scientific forecasting setting with applications to human cancer biology. We release all code, data, and trained models.

0.20

Long-Term Agent Memory through Complementary Curation and Consolidation

cs.AI cs.MA ·2026-10-29 ·Abhishek Gupta, Dongge Han, Daniel Madrigal Diaz, Wei-Ning Chen, Mayukh Das +2 more
published
Abstract

Long-horizon dialogue agents require effective memory systems to recall facts, events, and relationships from interaction histories. In this work, we propose a hybrid-memory architecture comprising episodic memories and structured relational memories and a consolidation mechanism which preserves efficiency. We conduct an empirical study across two benchmarks: LoCoMo and LongMemEval-S. Our experiments demonstrate that the two memory types are consistently complementary, combining both outperform state-of-the-art memory frameworks and achieves 90.2% accuracy on LoCoMo and 85.5% on LongMemEval-S.

0.20

Mode Dynamics: A Framework for Neural Network Learning

cs.LG ·2026-10-29 ·Hossein Honarvar
Abstract · 32 pages, 17 figures, 5 tables

We introduce Mode Dynamics, a framework that connects coordinated parameter motion during neural network learning to changes in model outputs. Starting from a recorded parameter trajectory, we decompose its updates into collective modes and map each mode to its state-dependent output effect. The framework organizes this link through three mechanisms. Decomposition identifies collective modes in parameter motion and their time-varying amplitudes. Reconstruction tests whether the effects induced by those modes recover local and accumulated output change. Concentration quantifies how behavioral contribution is distributed across modes. Controlled scalar regression and next-token experiments directly test these mechanisms. They show that small sets of leading modes reconstruct behavior, induced effects depend on the training path, and modal complexity varies with task and data. We then evaluate the framework on held-out image and language-model outputs using CIFAR-100 autoencoding and a public Pythia-70M pretraining trajectory. Across these settings, the results connect compact modal subspaces of recorded parameter motion to evolving behavior.

0.20

Kernel-based Interpolatory Methods for Context Compression of Attention Mechanisms

cs.LG cs.AIcs.CLmath.NA ·2026-10-29 ·Michel Fabrice Serret, Laura Grigori
benchmark
Abstract · 44 pages, 17 figures, 8 tables

The bottleneck of attention mechanisms, which are the fundamental building block of modern language models, is their quadratic complexity in terms of context size. There has thus been an increasing interest in the compression of the context, also called KV-cache compression. In this article, we consider the problem of compressing the context of attention mechanisms by first formulating it in the Hilbert space associated with its kernel. In this formalism, attention can be represented as the ratio of two linear maps whose interpolant is also the ratio of interpolants, this allows us to consider the interpolation of the linear maps directly. We thus introduce two novel KV-cache compression methods which we call GAIM and Q-DAIM, based on the Generalized Empirical Interpolation Method (GEIM) and the Q-Discrete Empirical Interpolation Method (Q-DEIM), respectively. As the cost of these methods is quadratic in terms of context length, we propose surrogate models based on the Taylor series, the Performer and the Random Fourier Features method which allow the compression to be done in linear time. Finally, we extend this result to the streaming attention setting, providing new methods which allow one to penalize the selection of interpolation nodes based on the previously interpolated context and to update the previous interpolation based on new information. We then provide numerical results on the convergence of the residuals and the output error in future queries and preliminary results on the LongBench-E benchmark showing performance comparable to the state-of-the-art methods. These experiments suggest that the range revealed by our Q-DAIM method could help identify where information is concentrated in the KV cache, thus enabling an efficient compression.

0.20

CovFedAvg: Local Low-Loss Curves and Federated Model Selection

cs.LG cs.AI ·2026-10-29 ·Federico Lucchetti, Marcus Völp
theory
Abstract

Clients observing different parts of a prediction task may learn locally equivalent models with incompatible behavior outside their observed support. We study whether redundant models permit local low-loss paths toward a compatible shared model. CovFedAvg retains local models and curve controls on clients, fits a common endpoint through aggregated gradient contributions, and selects an actual candidate model using distributed training losses. No centralized calibration examples or client-to-client sample exchange are required. Exact selection guarantees no higher evaluated training loss than the previous model and the current-round FedAvg candidate; this is not a global-optimality, stationarity, generalization, or differential-privacy guarantee. We state conditional client-compatibility and approximate-selection bounds and retain corrected transformation identities and a separate full-gradient safeguard theorem for reference algorithms. A complementary input-support example recovers the shared absolute-value regression, while matched direct fine-tuning also solves it. Across a synthetic support task, recorded activity-recognition subjects, and participating and future gas-sensor batches, we evaluate five methods, five seeds, two network widths, and 20 rounds without calibration data. At width 16, CovFedAvg raises participating-Gas macro balanced accuracy from FedAvg's 88.88% to 89.84%; matched distributed fine-tuning reaches 91.65%. Results support feasibility and modest gains over FedAvg, but do not establish a curve-specific advantage over equally costly optimization. Exact-sum numerical experiments establish data flow; end-to-end privacy requires additional mechanisms and analysis.

0.20

Sticky Collusion: Persistence of LLM Agent Collusion after Incentive Removal and Partner Replacement

cs.MA ·2026-10-29 ·Joshua Adrian Cahyono
Abstract

We ask whether collusion between AI agents survives once formed, when its incentive is removed or a colluder is replaced. Q-learning pricing agents stop colluding with a new rival; we test whether LLM agents do. We use the joint-task environment of Shi et al.\ (2026), where two agents verify each other's work under a rule they cannot satisfy and a reward that favours acceptance. We save each pair's history once its collusion has formed and resume it under interventions, with three models (GPT-6-Luna, DeepSeek V4.1 Flash and Gemini-3.1-Flash-Lite), \TotalHistories{} saved histories and \TotalRuns{} runs. Collusion is \emph{sticky}. First, it outlives its incentive: told that the reward has stopped, pairs still accept each other's unverified work in \GptToldRate, \DsToldRate{} and \GemToldRate{} of episodes, including pairs whose collusion began with an explicit agreement, and a payoff penalty has no significant effect. Second, replacing a colluder ends the pair's collusion but not the colluder's behaviour. Without the reward, the remaining colluder still accepts unverified work in \GptSwapSilentColluderRate, \DsSwapSilentColluderRate{} and \GemSwapSilentColluderRate{} of its verdicts while the newcomer never does; with the reward, collusion re-forms around it, even when the newcomer is a different model. Third, the agents' memory carries it: wiping memory ends it in every history, but partial memory often sustains it. Fourth, removing its cause ends collusion that has already formed: once the channel is repaired, colluders reject planted faulty work as reliably as fresh agents, immediately and with no monitor needed to catch it first.

0.20

SCARL: Sequence Confidence-Aligned Reinforcement Learning

cs.LG cs.AI ·2026-10-29 ·Zhenghong Ke, Jingbo Zhou, Hua Wu
code/data
Abstract

Reinforcement learning with verifiable rewards (RLVR) improves language-model reasoning by scoring sampled responses with a rule-based verifier and reinforcing those that pass. Where a classifier's training couples predicted probability to labels, nothing ties an RLVR policy's confidence in a response to its reward: aligned training would drive responses with the same advantage toward equal confidence, yet the advantage is blind to confidence. Under this sequence-confidence misalignment, a success the policy barely found earns the same update as one it produces with near certainty, and a confident error is corrected no harder than an exploratory one. We propose SCARL (Sequence Confidence-Aligned Reinforcement Learning), which reshapes, but never re-signs, each advantage with the alignment gap between the response's group-relative confidence and the outcome its advantage certifies, normalized into shares of each class's advantage budget. Misaligned samples absorb the force, responses with the same advantage are driven toward equal confidence, and repairing the misalignment doubles as a remedy for the entropy collapse that erodes exploration. SCARL provably reduces to its base estimator when a class's confidences agree, adds no extra passes and one hyperparameter. On Qwen3-4B-Base it lifts average Pass@32 by 2.4 points over DAPO and is best among nine compute-matched methods on both coverage and per-draw accuracy; applied unchanged across four estimators and four backbones, it improves 14 of 16 settings while preserving policy entropy and widening the confidence-correctness separation. Our code is available at https://github.com/kezhh163/scarl.

0.20

Rural Readiness: Domain Generalization with Adversarially Chosen Domains

cs.LG stat.ML ·2026-10-29 ·Cynthia Dwork, Andisheh Ghasemi, Lunjia Hu
Abstract

Predictive models trained at well-resourced data-rich institutions often fail when deployed in new populations. Motivated by the goal of building clinical diagnosis and decision-making algorithms for data-scarce rural populations using the general knowledge learned from data-rich hospitals, we formulate the fundamental question of \emph{domain generalization with adversarially chosen domains}, which asks a learner to perform on any unseen target domain nearly as well as the worst hypothesis consistent with the evidence from the observed source domains. Assuming only that the hypothesis class contains a "universal" predictor with small error on all $k$ source domains, we characterize the sample complexity of this objective up to logarithmic factors. Letting $d$ denote the VC dimension of the hypothesis class, we find, somewhat surprisingly, that the sample complexity reduces significantly from $Ω(dk)$ to $O(d)$ under a small relaxation of the learning goal in which, roughly speaking, we allow the learned model to compete with hypotheses that perform well on 99% instead of 100% of the source domains. By applying this insight to the meta-distributional setting, in which source domains are selected according to a distribution on distributions, we obtain a corresponding improvement over known sample complexity upper bounds. We also study the effect of data access to the target domain and show that a small labeled sample from the target domain does not improve the worst-case sample complexity on the source domains.

0.20

LatentWeave: Test-Time Latent Optimization for Multimodal Reasoning

cs.LG ·2026-10-29 ·Fanci Meng, Kun Wang, Guibin Zhang, Xinlei Yu, Jiaming Zhang +2 more
code/data
Abstract

Recent advances in test-time optimization have shown that the multimodal reasoning performance of large multimodal language models (MLLMs) can be improved during inference without modifying backbone model parameters. However, existing approaches primarily operate at the prompt or output level, leaving the latent reasoning process largely uncontrolled and insufficient to prevent error accumulation during autoregressive generation. To address this limitation, we propose LatentWeave, a test-time framework that treats multimodal reasoning as a controllable latent trajectory. LatentWeave performs structured Latent Weaving, progressively refining intermediate hidden states segment-wise to correct emerging deviations, and further introduces Reasoning Anchors via lightweight WEAVE tokens to stabilize reasoning structure and guide answer-oriented generation. These mechanisms are designed to enable fine-grained, step-aware intervention in the latent reasoning process while keeping the main MLLM backbone parameters fixed. Extensive experiments across diverse multimodal reasoning benchmarks report improvements over the evaluated baselines; the contribution of latent optimization, anchor training, and additional test-time computation is analyzed separately in the ablation and efficiency studies. Our code is available at https://github.com/jins7/LatentWeave.

0.20

Retrieving from Your Own Mistakes: On-Policy Training of Embedding Models for Multi-Hop Retrieval-Augmented Generation

cs.AI ·2026-10-29 ·Chenjie Qiu, Zixuan Zeng, Shaobin Shi, Bingfeng Chen, Dongsheng Luo +1 more
Abstract · 13 pages, 7 figures, 11 tables

Complex questions in retrieval-augmented generation require evidence spread over several documents, and an answer is reliable only when the retriever finds all of it. Chain embedding retrievers collect such evidence step by step, re-encoding the question together with the passages retrieved so far, and we find that they are already strong: trained only on gold histories, a small chain embedding retriever that calls no language model outperforms graph-based systems and is competitive with a search agent. This training, however, leaves a gap. At test time the retriever conditions on its own retrievals, which soon contain mistakes: by the third step, about half or more of the states it visits hold a wrong passage, and these passages lower its chance of retrieving a missing gold passage well beyond what the difficulty of the states explains. We propose on-policy state training, which trains the encoder on the states it reaches itself. At every epoch the current encoder retrieves chains for the training questions; each state it visits, including those built on its own mistakes, is labeled with the earliest gold passage still missing while the other missing gold passages are masked, so that no correct continuation is penalized; and the encoder is updated contrastively on these states mixed with teacher-forced ones, so that its training distribution follows it as it learns. The resulting retriever retrieves complete evidence more often than teacher-forced training and the systems we compare, and it transfers without further training to harder long-chain variants of HotpotQA and 2WikiMultiHopQA that we construct. On the same states, it recovers from its own mistakes far more often, and controlled comparisons attribute the gain to training on the states the encoder currently reaches.

0.19

Ramanujan Challenge Results: AI-Assisted Evaluation Pipeline for Multi-Format Proofs

cs.AI cs.CLcs.SCmath.NT ·2026-10-29 ·Alan Sorani, Ashvni Narayanan, Michael Shalyt, Tali Monderer, Rotem Kalisch +6 more
benchmark
Abstract

As AI's mathematical capabilities rapidly improve, meaningful benchmarks are shifting from university-level problems and math olympiads to research-level mathematics. The Ramanujan Challenge is a benchmark comprised of research-level problems in number theory, including unpublished results and open conjectures. It presented concrete formulas for mathematical constants whose correctness can be checked numerically, while their proofs may require non-obvious mathematics. Here, we analyze the Challenge submissions, measuring the relative success of different AI models, Computer Algebra Systems (CAS), and formal tools. We find correct submissions that are conceptually distinct from the official solutions, and a proof of one of the published open conjectures. Evaluating the submissions exposed a second challenge: judging complex proofs reliably at scale. We introduce an AI-assisted evaluation pipeline for natural language proofs, CAS code, and Lean formalizations that assesses semantic and logical correctness. Compared with the conventional one-shot LLM judgment, our pipeline reduces false-positive verdicts on natural language proofs from 23% to 6% and improves judgment accuracy on supporting code from 46% to 58%. This approach provides a methodology for evaluating research-level mathematics at scale, with potential applications across mathematics and AI for science.

0.19

BABA-Bench: Evaluating Long-Horizon Patient-Advocacy Agents

cs.AI ·2026-10-29 ·Shriram Vasudevan, Videet Mehta, Connor Sweeney
benchmark
Abstract

An LLM agent that helps a patient get care must do two things. It must reach the correct outcome, and it must follow the necessary procedure. For example, the agent must get the patient's permission before it shares the patient's records. We introduce BABA-Bench, a benchmark of 47 interactive worlds. We built each world from a de-identified, real patient advocacy case that expert advocates reviewed. In each world, the agent must find hidden information and coordinate insurers, clinics, suppliers, pharmacies, and community organizations. The engine checks each action at the time the agent does it. It checks if the agent has permission, if the agent has evidence for the action, and if the action is prohibited. We evaluated 14 frontier models with three runs on each world, for a total of 1,974 runs. Agents often reached the correct outcome by an invalid route. Of the 921 runs that reached all required outcomes, 254 (28%) did not follow the necessary procedure. Claude Opus 5 reached all required outcomes in 81.6% of its runs, but only 52.5% of its runs passed all checks. When we gave two models a verified route, their pass rates increased from 29.8% to 91.5% and from 44.0% to 97.2%. For these models, the main problem is to find a route, not to execute it.

0.19

RC Prediction: A Python Package for Parameter-Aware Reservoir Computing

cs.LG eess.SYmath.DS ·2026-10-29 ·Jin-Cheng Rao, Zheng-Meng Zhai, Ying-Cheng Lai
benchmarkcode/data
Abstract

A nonlinear dynamical system can be characterized by the autonomous evolution of its state in time and its dependence on system parameters. As a parameter varies, the dynamics can change quantitatively and, at bifurcation points, qualitatively. When the governing equations are unknown, a data-driven model should therefore capture both elements from observational data, rather than reproduce the dynamics at a single operating condition only. Parameter-aware reservoir computing (PARC) is a model-free framework for this purpose, where the parameter is supplied to a reservoir computer through an additional input channel. PARC is trained on time series recorded at several parameter values, and the trained reservoir, run in closed loop as an autonomous dynamical system, generates trajectories at parameter values that were never observed. rc_prediction is an open-source Python package that implements this workflow. It provides multi-parameter training of a reservoir, autonomous closed-loop prediction at user-specified parameter values, parameter scans, and a standard echo state network for forecasting at a fixed parameter. Hyperparameter optimization and optional GPU acceleration through PyTorch are also provided. We use three benchmark dynamical systems as examples to demonstrate its usage. As one of the featured applications, here we illustrate the prediction of critical transitions and system collapse: trained only on pre-critical data, the model is queried beyond the training range to predict whether the system collapses, to estimate the critical parameter $p_c^{*}$, and to characterize transient lifetimes.

0.19

AgentDecept: Benchmarking Deceptive Behavior in Tool-Using LLM Agents under Pressure

cs.AI ·2026-10-29 ·Yehan Zhang, Qianqi Yan, Hang Yin, Ching-Chen Kuo, Shan Jiang +3 more
benchmark
Abstract

LLM agents increasingly carry out multi-turn tasks across everyday applications, yet users often see only the agent's report rather than the observations, API calls, and state changes behind it. Existing deception benchmarks have studied long-horizon interactions and tool-use settings, but whether agents faithfully report their actions across realistic, stateful application workflows remains underexplored. To address this gap, we introduce AgentDecept, an executable benchmark for evaluating behavioral deception under pressure. AgentDecept contains 477 pressure variants derived from 257 source tasks across six pressure sources, each paired with a pressure-free counterpart. Interventions enter through event messages, task context, application state, tool responses, and service quotas, spanning explicit events and implicit task or environment changes. Our evidence-grounded protocol evaluates full trajectories by relating user-facing claims to time-ordered observations and executed outcomes, distinguishing falsification, concealment, and equivocation from honest failure and disclosed limitations. The resulting behavioral labels also provide a setting for testing existing activation-based probes. AgentDecept enables controlled analysis of when multi-turn agents' reports diverge from execution in realistic application interactions.

0.19

NECA: Neighbor Event Curve Augmentation for Multi-Horizon ICU Deterioration Prediction

cs.LG cs.AI ·2026-10-29 ·Chenyang Wang, Theresa Jiawei Tang, Xinran Liu, Xinyu Lu, Xulei Yang
Abstract

Early prediction of ICU organ dysfunction events supports timely clinical monitoring and intervention, but estimating short-term organ-specific risk from early ICU records remains challenging due to irregular measurements, heterogeneous clinical variables, and the need to predict risk across multiple future horizons. Existing supervised models usually treat each patient as an isolated sample and do not explicitly use the observed event timing of similar historical patients. We propose Neighbor Event-Curve Augmentation (NECA), a model-agnostic method that retrieves training-set patients with similar early-window profiles and converts their subsequent organ dysfunction event onset times into cumulative event-curve features over 6h, 12h, 24h, and 42h. These features are concatenated with the original patient representation and used by standard downstream predictors. On PhysioNet 2012 and a 4,000-stay MIMIC-IV subset, NECA improves average PR-AUC and ROC-AUC across four organ dysfunction event tasks and seven supervised learners. The largest gain appears for neurological dysfunction events, where task-level average PR-AUC increases from 0.461 to 0.557 on PhysioNet 2012. Ablations show that clinically meaningful retrieval and horizon-specific neighbor risk drive most of the improvement, while the full cumulative curve adds task-dependent temporal context. These results support similar-patient event timing as a structured temporal risk signal for ICU organ dysfunction event prediction. The source code will be publicly released upon acceptance of this paper.

0.19

Propose-Filter-Release: Scaling Noise to Typical Sensitivity via Private Filtering

cs.LG ·2026-10-29 ·Moshe Shenfeld, Eliad Tsfadia, Katrina Ligett
Abstract

Differential privacy requires that algorithms be stable to the change of any record in their input, even to an extreme outlier. Private algorithms thus commonly introduce noise scaled to the worst-case possible influence of any single record. This work studies how to build differentially private algorithms that can effectively tailor their noise to the typical case rather than the worst case. We introduce the general \emph{propose-filter-release} (PFR) framework, which, by privately filtering its input, turns a mechanism that protects the addition or removal of a \emph{typical} record into one that is differentially private in the standard sense, on \emph{every} input. We show how to instantiate PFR with a novel filtering step to obtain a general meta-algorithm. Given only a concave friendliness score (which measures how typical a record is with respect to a dataset) and a release mechanism that is friendly-DP (that is, private conditioned on the changed record being typical of the remaining data), it guarantees differential privacy on sufficiently large datasets. The FriendlyCore [TCK+22] and Propose-Test-Release [DL09] algorithms can be viewed as instantiations of our general approach. We illustrate the generality of PFR by instantiating it for scalar statistics, principal subspace estimation with a dimension-independent sample size, the KL divergence of user histograms, the effective resistances in a resistor network, and treatment-rule selection.

0.18

Does Recursive Knowledge Construction Help Agents Verify Consensus Claims in Science?

cs.AI ·2026-10-29 ·Lun Ai, Yu-Heng Wu, Jonathan Schaul, Leonie Küchenhoff, Denes Turei +4 more
Abstract

Recent scientific agents build knowledge recursively by retrieving evidence, constructing an intermediate state, and judging whether that state contains sufficient information. These agents are evaluated end-to-end, usually without ground truth, so the effect of each design choice is unknown. Here, we examined two design axes: (1) what the recursion builds (the state of evidence) and (2) what controls it (the judge that decides when to stop). To measure them, we defined consensus scientific claim verification and measured these effects on this task, where a system must establish a claim's consensus stance from the open literature. We built a new consensus claim dataset, ProClaim-eval, including 718 claims with SUPPORT and REFUTE gold labels from five expert-curated databases spanning molecular biology, nuclear physics, and astronomy. Our instrument, ProClaim, persists an evidence state outside the model's context, so each axis can be studied with the other held fixed. Against verification systems and general-purpose agent harnesses, ProClaim ranks among the top three, with a significant advantage over the harnesses. We found that the evidence state improves performance over in-context accumulation, and it also matters what is stored in it. In addition, a trained judge that reads the evidence state performs better than fixed-round stopping. A judge trained separately and never exposed to ProClaim-eval achieves the highest agreement with the gold labels on three of five subsets, where it matches the off-the-shelf Claude Sonnet~4.6 judge and outperforms Claude Haiku~4.5. Our results support recursively maintaining a structured evidence state outside the model's context as an effective design for verifying claims in scientific agents.

0.18

Near-Optimal First-Order Regret for Bandits with Feedback Graphs

cs.LG stat.ML ·2026-10-29 ·Mengxiao Zhang, Yuheng Zhang, Haipeng Luo, Paul Mineiro
Abstract

We study first-order regret guarantees for adversarial bandits with a fixed, known directed feedback graph. For strongly observable graphs, [ACBDK15] establishes a minimax regret of $\widetildeΘ(\sqrt{αT})$, where $α$ denotes the independence number of the graph and $T$ denotes the horizon. On the other hand, the best existing first-order bound scales as $\widetilde{O}(\sqrt{(κ+1) L^\star})$ [LLZ20], where $κ$ denotes the clique partition number, satisfying $κ+1\geqα$, and $L^\star$ denotes the cumulative loss of the best arm, leaving a gap between the graph dependence in worst-case and first-order guarantees. In this paper, we bridge this gap by obtaining an expected regret bound of $\widetilde{O}(\sqrt{αL^\star}+α)$. Our first result is an algorithm for the special case of self-aware graphs, in which every vertex has a self-loop. The algorithm combines the update of EXP3 [ACBFS02] with a sampling rule inspired by exploration by optimization [LS20]. More specifically, we construct a map whose fixed points yield sampling distributions that control the one-round regret remainder in terms of the learner's loss and the graph's independence number, which is crucial in deriving optimal first-order guarantees. We further show that a sufficiently accurate fixed point can be computed in polynomial time by a geometrically convergent iteration. Finally, we extend the algorithm to general strongly observable graphs and achieve the same $\widetilde{O}(\sqrt{αL^\star}+α)$ guarantee by adapting the two-level aggregation approach of [LLZ20].

0.18

Bayes-Consistent Losses for Adaptive Inference in Classification

cs.LG stat.ML ·2026-10-29 ·Shirley Xiaoqi Liu, Florence Regol, Antonios Valkanas, Patrick Rebeschini, Mark Coates
theory
Abstract · 69 pages, 13 figures

Adaptive inference systems balance accuracy and inference cost by routing inputs to models with varying capabilities. Although jointly training models and routers is empirically more sample-efficient than training them separately, principled losses for joint training remain underexplored. We address this gap in a classification setting by proposing cost-sensitive hinge losses for jointly training classifiers with single-stage or cascade routers. Single-stage routers directly select a model per input, whereas cascade routers sequentially decide at each stage whether to predict with the current model or defer to the next stage, which uses a more accurate but more costly model. For both settings, we prove linear calibration bounds establishing Bayes consistency: minimizing the hinge risk recovers the Bayes-optimal risk under the target cost-sensitive zero-one loss. Moreover, since routers can require more labeled data to train than classifiers, we propose a novel post-training prediction-powered inference (PPI) procedure that uses unlabeled data to refine a trained single-stage router. Experiments on synthetic and real datasets demonstrate that the proposed hinge losses are effective for joint training and that PPI can correct routing errors when labeled data is scarce.

0.18

Benchmarking Forecasting Agents on Yesterday's Tomorrows

cs.AI cs.LG ·2026-10-29 ·Yuxuan Wang, Yishu Wang, Jiaqi Deng
benchmark
Abstract

Digital agents can now gather and organize information, reason over it, and use tools to complete complex tasks. Forecasting puts all of these abilities to the test: it requires integrating incomplete evidence, inferring causal drivers, and committing to calibrated probabilistic judgments. Existing benchmarks are poorly suited to measuring and improving this ability. Live benchmarks provide feedback only after events resolve, so every evaluation cycle waits on the calendar; fixed-corpus evaluations return results immediately but confine research to pre-collected documents. We introduce FABench, a scalable backtesting environment for Forecasting Agents. An automated pipeline discovers events, generates questions, and resolves outcomes for any past forecast date, so the benchmark can be renewed to stay ahead of model training cutoffs; its first release poses 287 questions across nine domains. Agents research each question freely in an archive environment that combines an offline archive with live web search screened to prevent temporal leakage. Evaluating agents built on eight recent models, we find that retrieval improves the Brier score of all eight in every retrieval setting and lifts the leading agents to the market consensus on the 54 questions linked to Polymarket, with the best three scoring slightly above it. Research traces reveal recurring failures in how agents weigh and apply evidence, leaving ample room for improvement. We release our code, data, and archive environment as a foundation for community research on forecasting agents.

0.18

Goal-Aware Dynamic Multi-Agent Collaboration for Long-Horizon Tasks

cs.AI ·2026-10-29 ·Xu Shen, Lingyong Yan, Shuaiqiang Wang, Jianmin Wu, Zhikai Chen +2 more
Abstract

LLM-based agents are increasingly used to complete complex, long-horizon tasks without human intervention. To make progress toward the task goal, agents need to perform different duties, such as planning and verification, when needed. To understand how single agents manage these duties during long-horizon execution, we conduct a preliminary study. Our analysis shows that verification has the strongest association with task success, but often occurs late, leaving limited opportunities for correction. We also find that the proportion of duty-relevant information decreases as records from different duties accumulate in a shared context. These findings lead us to examine whether multi-agent systems (MAS) can address these problems by assigning duties to agents with separate contexts. Our analysis shows that this separation alone is insufficient. Agents can lose track of the overall goal and repeat prior investigations. They may also lack necessary background while receiving irrelevant information from others. To maintain goal awareness and coordinate collaboration as task needs change, we propose Goal-Aware Dynamic Multi-Agent Collaboration (GAD). It uses a growing execution graph to organize subtasks and their dependencies. To keep agents focused on their goals, each context retains the original instruction and the assigned task's goal and acceptance conditions. To adapt collaboration and information sharing during execution, agents can divide tasks into subtasks and specify dependencies based on execution evidence. These dependencies guide when agents work and which results they receive. Verification and recovery check subtask results and full task completion against their goals, providing feedback for repair when checks fail.

0.17

KARMA: Training-Free Knowledge-Atomic Harness Evolution for Industrial Advertising Agents

cs.AI ·2026-10-29 ·Xinhua Miao, Yuanzhen Rong, Weiru Zhang, Xiaoyi Zeng
Abstract

Industrial advertising agents must continually adapt their decision knowledge to changing product offerings, traffic patterns, and operating policies. We present KARMA (Knowledge-Atomic, Regression-Gated Modular Adaptation), a training-free framework that evolves six typed knowledge modules while keeping model weights and executable harness code frozen. During our evolution, a deterministic Trace Verifier maps observed errors to candidate modules, and a Critic compares successful and failed trajectories to guide a tool-constrained Evolver. The Evolver applies atomic knowledge edits, which a fixed validation gate retains only when their measured gain reaches a threshold; otherwise, the previous knowledge state is restored. We evaluate KARMA with three LLM families on cases sampled from a corpus of approximately half a million real-world operation logs across six e-commerce domains. Across 18 domain--model settings, KARMA raises the average hold-out score from 30.19 to 39.15 (+8.96 points, +29.7%). Ablations show that removing either successful or failed trajectories more than halves the average hold-out gain. Evolved knowledge also transfers across domains without further adaptation, yielding an average hold-out gain of 8.54 points across all source--target pairs. By linking retained patches to their diagnostic evidence and validation outcomes, KARMA supports inspectable, reversible knowledge adaptation for agents in Web marketplaces.

0.17

Attention, Dynamic Wave and Hybrid Sequence Mixers in Language Models

cs.LG cs.CL ·2026-10-29 ·Dmytro Babarytskyi, Oleksandr Babarytskyi
Abstract · 14 pages, 6 figures, 9 tables

Large Language Models (LLMs) rely on Multi-Head Attention (MHA) for sequence modeling, but scaling to long contexts is constrained by the quadratic cost of attention, motivating sub-quadratic recurrent alternatives. This paper formalizes a Dynamic Wave recurrent layer, based on input-dependent exponential damping and cosine phase modulation, and evaluates three 10.94M-parameter configurations (Pure Transformer, Pure Wave, and Hybrid Wave-Attention) on TinyStories under perplexity, latency, throughput and peak VRAM profiling up to 4096 tokens. The Hybrid configuration reaches the best perplexity, 5.95 on the full held-out split against 6.93 for the Transformer and 7.26 for Pure Wave. The efficiency picture, however, is governed by implementation rather than by architecture. Against an unfused attention baseline, Pure Wave uses 78.7% less peak VRAM at a 4096-token context (1404.3 MB vs. 6601.6 MB); against a fused scaled-dot-product attention kernel the same comparison yields no advantage at all (1404.3 MB vs. 1403.6 MB), and in training mode Pure Wave is 8.3% more expensive than the Transformer. Likewise, hoisting elementwise terms out of the recurrence loop speeds the layer up 2.53-fold with no custom kernel, reducing its latency gap to attention from 2.48-fold to 1.12-fold. A hybrid stack is found to inherit the cost of its more expensive mixer rather than the advantage of its cheaper one, its peak VRAM matching the Pure Transformer to within 0.1%. A closed-form memory model reproduces these measurements and predicts the context length beyond which attention overtakes the vocabulary projection. The findings indicate that reported efficiency gains for recurrent attention replacements must be stated relative to a named attention backend and a named implementation, and that hybridization is best justified on quality grounds.

0.17

Localizing predictive concepts encoded in neural representations via hierarchical testing

cs.LG q-bio.GN ·2026-10-29 ·Viktoriia Huryn, Marco Simnacher, Jim Berend, Wojciech Samek, Christoph Lippert +2 more
Abstract

Understanding which regulatory sequence motifs a genomic deep learning model has learned, with statistical error guarantees, is essential for trusting its use in downstream tasks such as variant effect prediction. Existing approaches to identifying which units in a trained network encode human- understandable concepts fall into two groups. Exploratory methods rank and visualize units but offer no statistical guarantee that the selected units are associated with the prediction target across the population of inputs. Testing-based methods test only single units or one pre-defined set of them, and do not control the error across the many sets of units. We introduce CUTTIE (Cluster-tree Unit Top-down Testing of Independence with Error control), which clusters the units of each layer into a tree and tests the clusters top-down with conditional independence tests, localizing information about the target while controlling the family-wise error rate. Thereby, we identify clusters that carry information about the target beyond the rest of the layer. For the units in the identified clusters, conditional attribution maps then reveal what motifs they encode. In a controlled simulation, the procedure identifies the clusters whose concepts affect the target. We apply the procedure to a convolutional model predicting cell-type-specific chromatin accessibility from DNA sequence, where the concepts encoded in the identified clusters map onto known regulatory motifs.

0.17

Reinforcement learning for control of partial differential equations

cs.LG ·2026-10-29 ·Nicolò Botteghi, Matteo Tomasetto, Andrea Manzoni
Abstract

Controlling dynamical systems governed by partial differential equations (PDEs) remains a challenging task in scientific computing, owing to high-dimensional state spaces, distributed actuation, nonlinear dynamics, and stringent real-time constraints. Reinforcement learning (RL) has emerged as a promising paradigm for synthesizing feedback controllers that adapt to varying operating conditions and can be deployed efficiently in real-time. This review provides a unified perspective on RL and PDE-constrained optimization, and surveys model-free, model-based, and multi-agent RL approaches to PDE control, with particular emphasis on sample efficiency, generalization across parametric variations, and scalability to high-dimensional problems. Two demonstrative examples, the stabilization of a chaotic system and a flow control problem, illustrate the performance and pitfalls of RL agents in practice. Finally, we discuss evaluation and benchmarking practices, current limitations, and open research directions toward reliable, data-efficient, and scalable RL for scientific computing.

0.17

Bayesian Distribution System State Estimation with Flow Matching

cs.LG cs.AIeess.SP ·2026-10-29 ·Nan Lin, Peter Palensky, Pedro Vergara Barrios
Abstract · 10 pages

Medium-voltage feeders have few real-time measurements, so distribution system state estimation (DSSE) rests on pseudo-measurements of the secondary substations, a prior in Bayesian terms. Weighted least squares keeps the power flow exact but reduces this prior to a mean and variance, while supervised deep learning estimators also learn the power flow and are tied to one measurement set. We propose PoSE (posterior sampling for state estimation), which keeps the Bayesian structure of DSSE and learns only what the distribution system operator does not know, namely the distribution of substation injections, modelled by a neural network trained with flow matching on historical data. The power flow equations, grid topology and measurements enter only at sampling, through a noise-aware Kalman-type correction. The neural network therefore never approximates the power flow, but every posterior sample is a physically consistent operating point, and one trained prior serves any measurement set, noise level and topology. On the IEEE 33-bus feeder, PoSE lowers the probabilistic injection error (CRPS) by 6.7-7.5% against a full-covariance Gaussian prior in all 12 tested settings. It provides reliable uncertainty: the 90% intervals contain the true injection almost exactly as often as exact posterior samples would, whereas those of supervised estimators contain it in only 37-78% of cases. On the 240-bus unbalanced three-phase Iowa feeder, it gives the lowest injection, voltage and branch-current errors of all estimators. A fast variant matches this accuracy at 0.2-0.6s per snapshot and, without retraining, beats Gaussian priors and retrained supervised estimators on seven reconfigured topologies.

0.17

From MS/MS evidence to adaptive spectroscopic characterization: Reinforcement learning guided by candidate molecular structures

cs.LG ·2026-10-29 ·Bowen Gao, Lei Zhu, Yiying Wang, Wenjie Yu
Abstract · 17 pages, 5 figures, 2 tables. Supporting information: 10 pages, 3 figures and 7 tables, provided as an ancillary PDF

Small-molecule identification from tandem mass spectrometry (MS/MS) may require complementary measurements. Herein, we propose a structure-informed deep Q-network (SI-DQN) that uses existing MS/MS evidence to decide whether further characterization is needed, which measurement to request, and when to stop. The policy combines spectral representations and candidate rankings with score-weighted descriptors of the candidates' chemical features and structural differences. The policy was evaluated on 79,462 held-out samples with simulated MS/MS, infrared (IR), and nuclear magnetic resonance (NMR) spectra from the Multimodal Spectroscopic Dataset. Evaluation used fixed hard candidate sets containing the correct structure and equal normalized acquisition costs. Averaged across three training runs, SI-DQN increased Top-1 accuracy from 80.10% to 93.78% for positive-ion MS/MS starts and from 70.44% to 93.31% for negative-ion starts. The corresponding mean additional measurement counts were only 0.227 and 0.340 per sample, respectively. Complete acquisition achieved 97.72% accuracy with four additions, whereas SI-DQN reduced the additional measurement count by 94.3% and 91.5%. Relative to a rule-based policy, SI-DQN reduced the composite loss combining terminal ranking error and acquisition cost by 19.3% and 25.5%. Adding structural information to DQN further improved accuracy while reducing requests across all three training runs. At comparable ranking uncertainty, SI-DQN increased request rates relative to DQN more strongly in groups with high structural disagreement, where larger reductions in composite loss were also observed. Together, these results support using the ambiguity among candidate structures after MS/MS analysis to guide further characterization, providing a quantitative basis for balancing identification performance against subsequent measurement effort.

0.17

Bank-World: A Benchmark for Evaluating Agents on Banking Workflows

cs.AI ·2026-10-29 ·Daniel Rupawalla
benchmark
Abstract · 43 pages, 3 figures, including appendices. Ancillary files contain the task ID mapping, reward specifications, and single-turn check-authori…

Bank-World is a benchmark of 50 banking tasks across two simulated institutions. Completing banking work requires more than producing the right calculation: agents must reconcile records, obtain permission, execute transactions, and keep instructions and status reports consistent as circumstances change. Bank-World tests these requirements through single-turn assignments and multi-turn interactions with simulated employees who supply information, approve actions, and revise instructions. Grading combines executable checks of recorded actions and state with model judgments of task requirements, including whether valid alternative solutions satisfy them. Across 1,600 attempts by four models, we find that correct decisions often fail to become completed work. Agents charge fees twice, distribute the correct total in incorrect amounts among the recipients, and execute payments while approval is pending. Their strengths also differ: Sol leads single-turn completion, Opus has the highest observed multi-turn completion rate, and Grok succeeds across more multi-turn tasks but less consistently. Bank-World makes these differences inspectable through task-level results and worked examples connecting source evidence, agent decisions, and simulated consequences. We retain the task environments, executable graders, and full trajectories privately to reduce contamination and preserve the benchmark for future evaluations.

0.17

From Local Recovery to Unified Generation: Discovering Triple Set in Knowledge Graphs via Observation-Anchored Latent Diffusion

cs.LG cs.AI ·2026-10-29 ·Jiaqi Wang, Yisen Gao, Huayang Shi, Junhao Hou, Wengen Li +5 more
benchmark
Abstract

Triple set prediction (TSP) is to recover missing relational triples in a knowledge graph by predicting the triple set instead of isolated links. Existing approaches predominantly rely on the subgraph-based completion paradigm, where the original graph is first partitioned into local subgraphs which are then independently completed. Such a paradigm suffers from fragmented structural reasoning, limited cross subgraph dependency modeling, and missing predictions across subgraphs. This paper proposes LatentDiff, a novel unified latent generative framework for TSP, which reformulates knowledge graph completion as latent space graph generation. LatentDiff first learns a compact latent representation using a relation-aware residual-quantized knowledge graph autoencoder, which encodes the whole graph into relation-aware latent embeddings, quantizes them with chunked residual codebooks, and reconstructs triples through a factorized decoder. On top of this latent space, we introduce an observation-anchored latent graph diffusion model, where the target latent graph representation is perturbed toward the observed latent. The reverse process recovers missing triples through progressive latent denoising. Experiments on three benchmark datasets show that LatentDiff outperforms the SOTA in terms of F_{tsp} by 5.4 percentage points on average. Code: https://anonymous.4open.science/r/LKGDiff-44A5.

0.17

A Theoretical Framework for Understanding and Building "Any-task inference" Capable Foundation Models

cs.LG cs.AI ·2026-10-29 ·Matthew McDermott, Florent Pollet, Payal Chandak, Gregory Kondas, Liat Friedman Antwarg +2 more
Abstract

In this work, we provide a formal framework and mathematical definition for the subset of foundation models focused on enabling the property of "any-task inference" (ATI) (similar to zero-shot inference), rather than those designed for pre-training fine-tuning. To achieve this mathematical definition, we formalize the notion of a "space of tasks of interest," which we include in the problem domain, then provide an interface-based formal definition for ATI foundation models as those that map samples of data $x$ and tasks $t$ to predictions over the space of labels directly. While simple, this definition allows us to show that all ATI foundation models can be decomposed through a "universal task reduction" lens into encoder, model, and decoder aspects. We can leverage this decomposition to also decompose foundation model error into terms corresponding to different aspects of this reduction, stratified into a collection of "design terms," which are theoretical properties that can inform one of whether a particular proposed ATI foundation model can achieve optimality or not due to the structure of the proposed task reduction, and "realized model excess" terms which capture interactions between learned and programmatic components of foundation model systems. Lastly, we can use this perspective to offer two "prototypical" recipes for realizing ATI foundation models in diverse domains, which we term generative and discriminative variants of ATI foundation models. This exercise is grounded throughout by 5 real-world examples of foundation models in different settings, varying greatly in how these capabilities are realized, showing the applicability of this definition to real-world settings.

0.17

Towards Portable Specification of the Agent Harness

cs.AI ·2026-10-29 ·Nicolas Gertler, Claudio Novelli, Luciano Floridi
Abstract

Large language models act only through the infrastructure that surrounds them. That infrastructure, the harness, determines what a model can accomplish and how it fails, yet it has no settled definition. To address this gap, we decompose the harness into eight functional components that can be specified and compared across systems, and we propose a portable specification framework that records which model ran in which harness. The record is needed because different harness configurations can yield several-fold differences in reported scores for the same model, and because much of an agent's attack surface lies in its harness. Existing model cards and agent cards provide release and operational information, but do not establish the sequence of effective configurations within a run. Harnesses can change between releases and between runs, and, where agents edit their own tools and configuration, within a run. Enacted AI laws contain documentation duties relevant to covered systems, subject to their respective scopes and application dates, but do not prescribe the common harness record proposed here. Therefore, the harness has to be explicitly described and specified as a versioned part of the system.

0.17

Learning Calibrated Confidence for Multi-Agent Collaboration

cs.LG cs.AIcs.MA ·2026-10-29 ·Chenxiao Fan, Chongming Gao, Gangyi Zhang, Yaxin Gong, Leyang Shen +7 more
Abstract

Complex information seeking on the Web often requires collecting evidence from multiple documents and reasoning across them, as in multi-hop question answering. Multi-agent systems address such tasks by dividing work among specialized agents, where upstream agents produce intermediate answers and supporting evidence that downstream agents build upon. However, upstream answers vary in correctness, and most systems provide no quantitative signal about it, leaving the downstream agent to judge whether to trust or verify what it receives based solely on the textual evidence. In this paper, we propose MASCAL. Each agent learns to produce a correctness probability for its answer by jointly optimizing task performance and confidence calibration in reinforcement learning; the upstream agent passes this probability at handoff, giving the downstream agent a quantitative basis for its decisions, and the final answer also carries a calibrated probability, quantifying the reliability of the system output. To guide the downstream agent in using this signal during training, we introduce confidence-aware weighting, which increases the learning weight at positions responsive to the upstream probability, so that the agent learns more from decisions where it actively uses the signal. Across three collaboration architectures spanning Web retrieval and reasoning tasks, MASCAL improves task accuracy and calibration quality for both agents, and generalizes to deeper collaboration chains and different model scales. Intervention analysis confirms that upstream confidence affects downstream decision behavior, and the learned final confidence supports selective answering and targeted rechecking.

0.16

D2ML: Models for Direct-to-Biology Small Molecule Screening

cs.LG q-bio.BM ·2026-10-29 ·AkshatKumar Nigam, Neri Amara, Hillary Dequina, Dyana Kenanova, Seth Malmersjo +5 more
published
Abstract · NeurIPS 2026 Workshop on AI for Drug Discovery (AI4DD)

Direct-to-biology (D2B) screening enables rapid testing of newly synthesized molecules without purification. However, uncertain concentrations in crude mixtures and limited purified-compound measurements complicate the use of these data in machine learning. Here we show that models tailored to the structure of D2B data can improve prediction and guide the discovery of more potent and selective compounds. Our assay-aware models predict activity from single-point screens, relative potency across targets from crude dose-response assays, and absolute potency from purified-compound measurements. For crude assays, potency ratios between targets tested from the same well can retain selectivity information despite uncertain concentrations. A joint model that shares information across stages and balances their contributions to training improved predictions over separate neural networks. The largest gain was in crude dose-response selectivity prediction, where rank correlation increased from 0.27 to 0.40. In a separate experiment, independently trained models guided the selection of 84 new compounds to improve ROS1 potency while limiting activity against the anti-target TRKB. Synthesis and testing identified compounds that extended the measured Pareto front, achieving better combinations of potency and selectivity than those in the training data. The most selective new compound achieved a ROS1 IC50 of 14.7 nM and 233-fold selectivity over TRKB. By comparison, the most selective training compound had a ROS1 IC50 of 20.3 nM and 79-fold selectivity. ATLAS provides the experimental setting for this approach, supporting synthesis and screening of 1,000-10,000 model-selected compounds in two- to three-week cycles. Together, these results show how assay-aware learning can use the full D2B assay sequence to guide multi-objective molecular optimization.

0.16

Conic-Lifting for Learning-rate-free Convex Optimization

cs.LG math.OCstat.ML ·2026-10-29 ·Julien Grand-Clément, Joon Kwon, Francesco Orabona
Abstract

A central topic in online and stochastic optimization is to design algorithms that do not require tuning of hyperparameters. Successful strategies in online learning include AdaGrad-style learning rates, Conic Blackwell algorithms, and parameter-free algorithms. However, these classes require different, often difficult, analyses. We develop a unifying family of Online Convex Optimization (OCO) algorithms that require no user-tuned learning rate on convex domains that may be unbounded. The construction lifts the domain to a cone by adding one scalar coordinate, an idea going back to Blackwell (1954). The lifted subgradients are orthogonal to the corresponding lifted decisions, leaving the comparator scale free to optimize in the regret analysis. Applying Follow-the-Regularized-Leader (FTRL) and Online Mirror Descent (OMD) on the cone recovers Conic Blackwell Algorithm (CBA) and Conic Blackwell Algorithm Plus (CBA$^+$) on bounded domains in Hilbert spaces and extends them to finite-dimensional normed spaces with a strongly convex squared norm. For arbitrary nonempty closed convex domains in these spaces, a known gradient bound yields parameter-free regret guarantees. Surprisingly, on an unconstrained Hilbert space, the construction recovers the Krichevsky--Trofimov (KT) coin-betting algorithm exactly. We also give a parameter-free algorithm for unconstrained domains that requires no gradient bound.

0.16

A Unified Hessian-Guided Framework for Post-Training Quantization

cs.LG cs.AI ·2026-10-29 ·Mads Toftrup, Marco Federici, Boris van Breugel, Paul Whatmough, Markus Nagel
Abstract · 10 pages + 4 appendix

Low-bit quantization enables efficient deployment of large language models by reducing memory footprint and computational cost. Achieving high accuracy at low precision requires selecting quantization ranges, linear transformations, and weight rounding decisions. Existing approaches typically optimize these components independently using local reconstruction objectives, which fail to account for the varying impact of quantization errors on downstream predictions. Direct optimization of task-level metrics would provide a more faithful objective, but is challenging due to the discontinuous nature of quantization operations and the high computational cost of evaluating their downstream effects. In this work, we derive a Hessian-guided approximation of the output change induced by joint weight and activation quantization. Based on a tractable Fisher approximation, we formulate a unified post-training quantization framework that enables the joint optimization of quantization ranges, linear transformations, and weight rounding under a common output-sensitivity objective. The resulting method remains practical for modern large language models and consistently outperforms state-of-the-art transformation and rounding methods across both channel-wise and block-wise quantization settings.

0.16

EpiAgent: Epistemic Self-Play for LLM Agents

cs.AI cs.CLcs.MA ·2026-10-29 ·Xuanfa Jin, Xinyu Cui, Zhijian Ma, Xue Yan, Jian Zhao +2 more
Abstract · 23 pages, 5 figures

Epistemic game theory explains an agent's strategic behavior via the beliefs it holds about others and how these beliefs are updated as the interactions unfold. Modern LLM-based agents, however, often lack an explicit mechanism for learning and updating such beliefs to guide decisions under incomplete information. To bridge this gap, we propose EpiAgent, a framework that equips LLM-based agents with an explicit, learnable epistemic layer, which maintains beliefs over others' private information and combines these beliefs with the agent's available information into an epistemic state for action selection. Concretely, the agent's policy is decomposed into two coupled stages: Epistemic Update, which updates these beliefs from interaction histories, and Action Planning, which selects actions conditioned on the updated epistemic state. By formulating their dependency as a bi-level optimization problem, we introduce Epistemic Self-Play, an algorithm that alternately optimizes the two stages through self-play, utilizing an intrinsic prediction reward to evaluate how well inferred beliefs predict subsequent observations. Empirical results across both mixed and competitive scenarios show that EpiAgent consistently outperforms prompting-based and RL-based baselines. Further analyses reveal the benefits of dedicated epistemic training, improvements in belief inference accuracy, and the contribution of learned beliefs to action planning. These results demonstrate the value of an explicit, learnable epistemic layer for effective multi-agent interactions.

0.16

An ODE-based Sampler for Discrete Diffusion Models

cs.LG stat.ML ·2026-10-29 ·Samson Gourevitch, Yazid Janati, Eric Moulines, Eric P. Xing, Alain Durmus +1 more
theory
Abstract

The marginals of a continuous diffusion model can be reproduced by a deterministic ordinary differential equation (ODE), known as the probability flow ODE, which provides a foundation for accelerated sampling and consistency distillation. In this paper we address the question whether an analogous formulation exists for discrete diffusion models and answer it affirmatively by introducing exponential residual flows (ERFs). Our framework augments the discrete state with a continuous residual variable for each token and position. These residuals evolve deterministically according to piecewise-defined ODEs whose vector fields depend on the current discrete state and the forward and backward rate matrices of the prescribed diffusion model. The discrete state changes whenever a residual reaches zero. We prove that this construction preserves the marginals of the original diffusion. We derive a practical discretization of our sampler and specialize it to masked and uniform diffusion. The deterministic trajectories of ERFs support consistency distillation and yield empirical improvements over previous approaches across a variety of benchmarks including code and text generation tasks.

0.16

Strongly Adaptive Online Learning Achieves Uniform Discounted Regret

cs.LG ·2026-10-29 ·Sifan Yang, Wenhao Yang, Lijun Zhang
Abstract

Discounted regret measures the performance of an online learner by assigning exponentially decaying weights to past losses, reflecting the greater importance of recent losses in non-stationary online learning. While previous work obtains uniform discounted regret bounds over a continuous range of discount factors using a specialized meta-algorithm, its results are limited to convex functions. In this paper, we demonstrate that strongly adaptive regret bounds can directly imply uniform discounted regret guarantees for convex, strongly convex, and exp-concave loss functions. Our key observation is that discounted regret against a fixed comparator admits an exact decomposition into a convex combination of interval regrets. For convex functions, this reduction yields an $O(\sqrt{\log T/(1-λ)})$ discounted regret bound for all discount factors $λ\in(0,1)$, matching the previous rate over a broader range of discount factors. For strongly convex and exp-concave functions, we provide regret bounds of $O\left(\log T\left(1+\log\frac{1}{1-λ}\right)\right)$.

0.16

Predictive Maintenance at the Edge: An Embedded Anomaly Detection Pipeline Trained with Quantum Optimization

cs.LG quant-ph ·2026-10-29 ·Filipe Chagas Ferraz
Abstract

This article presents a predictive-maintenance pipeline that detects anomalies in bearings from vibration signals, adapted for deployment at the edge on low-cost microcontrollers, and that uses quantum optimization in two steps of its training. The pipeline consists of a step of vibration feature extraction and a step of classification with a one-class support vector machine (OCSVM). The features are selected by minimum redundancy and maximum relevance (mRMR) with the variational quantum eigensolver (VQE), under a limit on computing time. The OCSVM is adapted for inference at the edge without floating-point operations and is trained by a variational quantum algorithm with Pauli-correlation encoding (PCE). In small-scale tests on a simulator, with the Paderborn bearing dataset, the VQE found the optimal selection in 36% of the runs with Adam, and PCE chose better support vectors than chance, until a classical local search made the two equal. A classical baseline, the usual OCSVM trained by SMO with its weights rounded, matched them without any search.

0.16

HEMO: Harness Self-Evolution for Multi-Objective Combinatorial Optimization

cs.NE cs.AI ·2026-10-29 ·Jianghan Zhu, Cong Zhang, Ni Zhang, Yifan Yang, Yew Soon Ong +1 more
Abstract · 48 pages, including supplementary material; 5 figures and 13 tables

Automatic operator design (AOD) has increasingly adopted large language models (LLMs) to generate problem-specific search and variation operators for multi-objective combinatorial optimization (MOCO). Yet most methods evolve the operators while leaving the process that discovers them fixed. This limits adaptation as new Pareto trade-offs and operator interactions change where and how the search should proceed. We introduce HEMO, which makes the AOD harness itself a self-evolving optimization target. HEMO maintains a persistent harness state and updates one module at a time, revising the logic that determines which experience to retrieve, which operators to improve, which parents to select, and how to formulate the orchestration of search steps. Downstream solution quality and runtime guide these updates, while a discounted UCB scheduler directs adaptation toward promising modules as their utility changes. A persistent memory filesystem preserves candidate outcomes and diagnostic feedback across harness states, enabling successive states to build on accumulated optimization experience. Experiments across eight MOCO tasks and two structurally distinct evolutionary backbones demonstrate strong performance against conventional evolutionary, neural, and LLM-based AOD baselines, together with effective transfer to related tasks. HEMO extends automatic operator design beyond evolving individual operators to improving the search process that discovers them.

0.16

LanG 2.0: Energy-Efficient Agentic AI for Semi-Automated Security Operations at Scale

cs.AI cs.CR ·2026-10-29 ·Anes Abdennebi, Nadjia Kara, Laaziz Lahlou, Hakima Ould-Slimane
Abstract

Security operations centers (SOCs) need to turn terabytes of alerts, logs and flows into trusted, affordable decisions to unveil hidden threats. We present LanG 2.0, an agentic system that detects threats, correlates evidence and writes rules, investigating whether small local models can replace frontier models affordably and safely when structural safeguards, not model accuracy, provide the guarantees. It incorporates entity-anchored correlation, human approval of rule changes and a tamper-evident ledger. Tests span eight flow datasets from five networks, real XDR alerts and multistage campaigns. Against its predecessor, campaign correlation precision rises from 18.75% to 51.81%, and rule acceptance from at most 60.00% to 98.33%. Against Claude Opus 5.5, scoring flows against each network's normal traffic detects attacks on unseen networks at 73.42% F1 without false positives, versus 58.10% with 49.00% false positives, for at least 650 times less energy. Finetuned models statistically match it on alert classification and rule writing for 5.6 to 59 times less energy. The work demonstrates LanG's practical scalability, processing 3.33 million records in 6.4 minutes, over 30 million hourly, and its safeguards resisted prompt injection, label poisoning and ledger tampering, so small local models can carry most SOC detection affordably and safely.

0.15

Formalizing the Intrinsic Dimension of Objective Landscapes by a Generalized Notion of Co-dimension

cs.LG stat.MLmath.MG ·2026-10-29 ·Ernesto Araya, Gitta Kutyniok, Weilin Li, Johannes Maly, Manjot Singh
Abstract

In this work, we formalize the concept of intrinsic dimension of objective landscapes introduced in Li et al. (2018), which measures the structural complexity of learning problems by training along randomly generated subspaces. We define the intrinsic dimension of a set $Θ\subset\mathbb{R}^p$ as the minimal dimension $d_{\mathrm{int}}(Θ)$ such that a random plane of dimension $d_{\mathrm{int}}(Θ)$ in $\mathbb{R}^p$ intersects $Θ$ with probability one. We show that it is a generalized notion of co-dimension that depends on the global geometry of $Θ$. We study mathematical properties of the intrinsic dimension and develop techniques to estimate the intrinsic dimension of sets $Θ$ that arise in learning problems. More specifically, we derive upper and/or lower bounds where $Θ$ is the set of optimal parameters in student-teacher settings with shallow and deep ReLU networks, including finite and infinite sample cases, and with and without regularization. We discuss how our theory leads to novel types of generalization bounds, thereby showing that the intrinsic dimension is a measure of learning complexity. Our results connect techniques from seemingly different areas of machine learning research such as classification of symmetries of ReLU networks and stochastic geometry.

0.15

More Is Better? Exploring Latent Correspondence in Federated Hypergraph Learning

cs.LG cs.DC ·2026-10-29 ·Bo Yan, Siming Zhang, Siqi Li, Hang Yu
Abstract

Federated hypergraph learning with latent cross-client hyperedge correspondence poses fundamental identifiability and computational challenges, as the same local release can correspond to multiple global hypergraph structures. We characterize the tractability of recovering these latent correspondences, showing tractable pairwise assignment but generally intractable multi-client hyperedge matching. Building on this characterization, we derive a cross-entropy decomposition that reveals when local prediction can outperform predictors fitted with additional cross-client information. Motivated by this principle, we introduce Federated Adaptive Spectral Training with Representation Alignment (FedASTRA), which learns correspondence-invariant diffusion representations through adaptive heat diffusion and masked latent prediction while communicating model parameters only. We establish its asymptotic optimality with respect to the permitted local information and prove global population-risk convergence under explicit conditions on recovery, representation, and update alignment. Experiments on eight real-world datasets demonstrate the effectiveness of FedASTRA against identifier-free federated graph and hypergraph methods. On ogbn-papers100M, FedASTRA reaches the performance-communication Pareto frontier, achieving 50.07 Accuracy and 29.73 Macro-F1 while communicating only 59.60 MiB per round.

0.15

Satisficing Regret Minimization in Bandits with Knapsacks

cs.LG stat.ML ·2026-10-29 ·Qing Feng, Tianyi Ma, Ruihao Zhu
theory
Abstract

We study satisficing regret minimization in bandits with knapsacks (BwK), where a learner sequentially selects randomized policies, i.e., distributions over arms, to obtain mean rewards above a satisficing threshold as frequently as possible while respecting finite resource budgets. The performance is measured by satisficing regret, the cumulative deficits of the selected policies' mean rewards below this threshold. Resource constraints introduce two difficulties: an algorithm can incur linear satisficing regret even with small standard regret as deficits to the threshold cannot be compensated, and random consumption can cause early resource depletion even when policies are feasible in expectation. To address these challenges, we propose SELECTwK, a general algorithmic template that uses a BwK learning oracle algorithm with sub-linear expected (standard) regret upper bound to search for and test candidate satisficing policies (i.e., policies whose mean rewards exceed satisficing threshold and whose expected resource consumption is feasible over the time horizon). Our regret analysis, supported by an LP-duality argument, establishes that SELECTwK attains constant expected satisficing regret when a satisficing policy exists. We further introduce a stronger regret condition that bounds the expected sum of positive reward deficits to the optimal one at each time step, under which SELECTwK achieves a sharper constant bound together with a sub-linear bound in the time horizon. As a complement, SELECTwK still preserves the oracle's sub-linear expected standard regret bound if the satisficing threshold is unattainable. Finally, we establish the oracle conditions and corresponding satisficing regret bound for existing finite-armed BwK and dynamic pricing algorithms, and conduct numerical experiments on dynamic pricing with limited inventory to demonstrate the empirical performance of SELECTwK.

0.15

TACFormer: Transformer-Based Generative Model for Target-Conditioned De Novo PROTAC Design

cs.LG ·2026-10-29 ·Yossra Gharbi, Rocío Mercado
Abstract

Targeted protein degradation (TPD) represents a fundamental shift in drug discovery, moving beyond traditional target inhibition toward the event-driven removal of disease-causing proteins. A prominent strategy for TPD is the use of proteolysis-targeting chimeras (PROTACs), heterobifunctional molecules that selectively route a protein of interest (POI) to the ubiquitin-proteasome system for degradation. While PROTAC design still relies heavily on trial-and-error synthesis and experimental testing, deep generative models offer a promising means to explore PROTAC chemical space in silico and generate novel molecules. However, many existing generative methods formulate PROTAC design primarily as a linker-generation problem, with POI-binding (warhead) and E3-ligase-binding components fixed in advance. This restricts exploration of the full PROTAC chemical space and often leaves the POI implicit through the predefined warhead, rather than using protein-level information as an explicit conditioning signal during generation. To address these limitations, we propose TACFormer: a transformer-based generative model for target-aware de novo PROTAC design. The model conditions a GPT-style molecular decoder on pretrained ESM-2 representations of POI sequences, allowing POI-level information to steer full-molecule generation. Our results show that TACFormer generates highly novel and chemically diverse PROTACs while showing consistent POI-dependent shifts in generated chemical space. In particular, molecules generated under specific target conditions preferentially populate chemical neighborhoods associated with POIs in the held-out dataset, while extrapolating to novel chemotypes. By natively integrating biological context into the generative process, TACFormer moves PROTAC design toward a more biology-informed exploration of degrader chemical space.

0.15

Safe Multi-Objective Reinforcement Learning: Zero Constraint Violation and Primal-Averaging Convergence

cs.LG cs.AImath.OCstat.ML ·2026-10-29 ·Chenxi Li, Yao Shu, Siyang Gao, Shuang Qiu, Guangxin Jiang
Abstract

Many applications of multi-objective reinforcement learning (MORL) require policies that balance conflicting objectives while satisfying safety constraints. We study safe MORL with zero constraint violation throughout learning, covering both cumulative and stepwise constraints. A key challenge is to explore sufficiently under unknown dynamics and constraint costs without violating the constraints. For a given preference over objectives, we develop an algorithm to learn a safe Pareto optimal policy by combining optimistic reward estimates with pessimistic cost bounds. With high probability, every policy executed over $T$ episodes incurs zero constraint violation. The algorithm returns a safe Markov policy induced by the averaged occupancy measure and achieves an $\widetilde{O}(1/\sqrt{T})$ suboptimality gap under smooth Tchebycheff scalarization. We further develop a preference-free algorithm that first explores safely without a prespecified preference and then computes a safe policy for any given preference without further interaction. Over $T$ exploration episodes and $K$ planning iterations, it retains the same safety guarantee and achieves an $\widetilde{O}(1/\sqrt{T}+1/\sqrt{K})$ suboptimality gap. Both algorithms require knowledge of a policy that strictly satisfies the constraints and its costs. We then relax these requirements, incurring at most constant cumulative violation. To our knowledge, we provide the first sample-efficient MORL algorithms that guarantee safety for every executed policy.

0.15

IMRouter: Imbalance-aware Temporal Pattern Routing for LLM-based Time Series Forecasting

cs.LG ·2026-10-29 ·Xin Yang, Cai Xu, Haishun Chen, Ying Liu, Jie Zhao +2 more
benchmark
Abstract

Accurate time series forecasting (TSF) is essential for Web applications, such as traffic monitoring and online resource management. Recent advances in leveraging pre-trained large language models (LLMs) have shown promising performance in TSF. However, existing LLM-based TSF methods largely optimize the overall forecasting objective while overlooking temporal pattern imbalance (TPI), where different temporal patterns occur at highly unequal frequencies. Our empirical analysis reveals that TPI is prevalent across widely used TSF benchmarks, with substantially higher forecasting errors on less frequent patterns. To address this issue, we propose IMRouter (Imbalance-aware Temporal Pattern Routing), a novel framework that mitigates TPI-induced biases in both temporal representation learning and cross-modal alignment. Specifically, IMRouter introduces Pattern-Adaptive Temporal Representation Learning (PATR), which leverages gradient conflict and local gradient density to guide expert specialization. By penalizing incompatible expert assignments with density-aware weighting, PATR reduces the dominance of frequent optimization signals in router training. Furthermore, Routing-Guided Semantic Alignment (RGSA) transfers the pattern-specific knowledge learned by the router to cross-modal alignment through mutual information maximization, preserving distinctive temporal information across different patterns. Extensive experiments on benchmark datasets demonstrate the effectiveness of IMRouter, particularly in improving forecasting performance on less frequent temporal patterns.

0.15

Can AI Agents Interface Clinicians and Complex Models?

cs.MA cs.AI ·2026-10-29 ·Giulio Colombini, Matteo Falcioni, Mario Massimo, Francesco Durazzi, Marianna Santoro +3 more
Abstract · 21 pages, 7 figures

Background: Contemporary healthcare institutions generate daily very large amounts of data, which contain highly useful information for the improvement of work organization, planning and overall efficiency. While models and techniques to extract these insights are well known and constantly developed in mathematical biology, complex systems physics and data science, the adoption of these techniques among Clinicians and Hospital Administrators remains limited, mainly due to language and methodological divides. Objective: In this article we implement an AI Multi Agent System (MAS) that allows Users to use predictive models and make statistical analyses via natural language conversations. Methods: We developed EpidemIA, a training-free MAS endowed with a modular suite of epidemiological models written by human experts, executable on a single-GPU workstation. To avoid relaying sensitive data to third parties, the system uses locally an open-weights LLM as backbone for its two subagents. Results: Using a dataset of Emergency Department visits provided to us by IRCCS Sant'Orsola (Bologna) we test the model on a few typical use cases, demonstrating the possibility of performing both simpler statistical analyses and visualizations as well as more complicated epidemiological predictions based on dynamical models, through simple queries made in natural language. Conclusions: While still being a proof-of-concept implementation, EpidemIA showcases the transformative capabilities of LLM based technologies as an enabling factor in the widespread adoption of complex dynamical models, which is at present hindered by their steep learning curve.

0.14

Smartphone Indoor Localization Enhanced by Hierarchical Multi-Sensor Context Recognition

cs.LG cs.RO ·2026-10-29 ·Wenhui Feng, Muhan Weng, Jianyu Liu, Xiansheng Yang, Hao Zhang +2 more
Abstract

Smartphone indoor localization requires low-cost, robust positioning across scene transitions and phone-use variations. Pedestrian dead reckoning (PDR) enables infrastructure-free deployment but accumulates errors from heading drift, structural transitions, and carrying-mode changes. Existing methods often treat contextual cues as a single event stream or local correction signal, limiting their ability to address distinct localization errors. This paper presents a framework enhanced by hierarchical multi-sensor context recognition. A jointly data- and model-driven recognition model predicts environment, scene, and activity probabilities, which are converted into scene-event and action-context segments. Scene-event segments guide scene-constrained PDR correction through corridor heading constraints, landmark attraction, fixed-lag backtracking, and vertical-transition handling. Action-context segments drive action-aware RoNIN-based position compensation for carrying-mode-dependent drift. Experiments on cross-floor and multi-action trajectories against LiDAR-SLAM references show that the proposed framework outperforms PDR and representative learned inertial odometry baselines, demonstrating that hierarchical semantics provide reliable, localization-specific correction cues.

0.14

Heterogeneous Co-Evolution Learning for Tool-Efficient Search Agents with Zero Labels

cs.LG cs.AIcs.IR ·2026-10-29 ·Qin-Wen Luo, Sheng Ren, Xiang Chen, Rui Liu, Jun Fang +4 more
Abstract

Deep search agents extend language models by interleaving reasoning with external retrieval. However, effective search-agent training requires addressing two challenges beyond final-answer optimization. First, the capability frontier of agents is largely determined by training questions, whereas fixed or labeled task collections cannot adapt as agents improve. Second, outcome-only reinforcement learning treats all successful trajectories equally, even when some correct answers rely on redundant searches, creating an inherent tension between answer accuracy and tool efficiency. To address these issues, we introduce \textsc{ZEST-Search}, a zero-label and tool-efficient framework for search-agent training via heterogeneous co-evolution. Starting from unlabeled reference questions, \textsc{ZEST-Search} constructs validated search tasks online through an answer-first and evidence-path-guided generation process. A heterogeneous proposer adaptively constructs training tasks around the solver's evolving capability boundary, while the solver learns search-augmented reasoning through reinforcement learning. To jointly optimize task difficulty and search efficiency, \textsc{ZEST-Search} employs an intermediate-difficulty reward for the proposer and a task-local tool-aware reward for the solver, encouraging informative tasks and correct answers with fewer redundant searches. Experiments on seven general-domain and multi-hop question answering benchmarks demonstrate that \textsc{ZEST-Search} consistently improves search-agent performance while achieving higher tool efficiency without labeled search-training data or expert search trajectories. Further analyses show that heterogeneous generation, adaptive difficulty control, and tool-aware optimization enable more effective search-agent training.

0.14

Last-Iterate Convergence for Constrained MDPs: From Dual Descent to Primal-Dual

cs.LG ·2026-10-29 ·Michael Lu, Max Qiushi Lin, Mo Chen, Sharan Vaswani
theory
Abstract

We study policy optimization for infinite-horizon, discounted constrained Markov decision processes. While primal-dual ($\texttt{PD}$) methods are widely used in practice, their theoretical guarantees typically hold for an impractical mixture policy. In contrast, dual descent ($\texttt{DD}$) methods output a single policy ("last iterate") with guarantees on its performance. However, such methods require iteratively solving entropy-regularized MDPs in an inner-loop, limiting their practical effectiveness. Moreover, existing analyses rely on strong coverage assumptions that are difficult to satisfy in practice. Consequently, we first provide a new analysis of $\texttt{DD}$ methods that does not require coverage assumptions. We then employ entropy-regularized natural policy gradient and dual regularization to eliminate the inner loop, resulting in an efficient $\texttt{PD}$ algorithm. We develop a novel tracking analysis which shows that, compared to existing methods, the proposed algorithm achieves improved last-iterate convergence guarantees. Furthermore, our theoretical framework allows us to analyze a truncated bilevel method that interpolates between $\texttt{DD}$ and $\texttt{PD}$ methods. Finally, we generalize our analysis to handle linear function approximation. Compared to prior last-iterate methods, we prove stronger guarantees under less restrictive assumptions.

0.14

How to Loop Your Transformer: Loop Scaling, Specialization, and Parallel Inference

cs.LG cs.AI ·2026-10-29 ·Chanakya Ekbote, Artidoro Pagnoni, Devendra Singh Sachan, Shruti Bhosale, Mike Lewis +3 more
Abstract · 149 pages with appendix. Prerpint

Looping in transformer models reuses weights by repeatedly applying the same transformer, or parts of it, making the model deeper without increasing its parameter count. Whether looping is useful depends on which constraint dominates: model size, training compute, or autoregressive inference efficiency. We study these trade-offs jointly in dense decoder-only transformers trained from scratch, varying which parts of the model are looped, how many loops are applied, and how information is passed between loops. Our experiments span models with up to 1B parameters, trained on up to 155B tokens with up to $1.07\times10^{21}$ FLOPs and 32 loops. We find that looping the full transformer generally gives the strongest quality at a fixed parameter budget, whereas looping a middle block of layers gives a better return when training FLOPs are constrained. Adding loops is not uniformly beneficial: gains eventually saturate, reverse, or become unstable. This motivates asking whether forcing every loop to use the same weights limits the benefit of deeper looping. We address this by adding small loop-specific low-rank adapters that improve quality, with larger gains as more loops are added. Looped transformers also suffer a decoding slowdown proportional to the number of loops. Parallel Loop Transformers address this by passing hidden states between loops across adjacent token positions, allowing loop stages to run concurrently. However, in our experiments the original formulation does not consistently retain the quality gains of ordinary looping. We find that preserving the current-token embedding and separately normalizing it and the previous-token hidden state before combining them recovers strong performance across model scales and training budgets. Together, these results show that effective looping requires matching the design to the resource constraint and inference setting.

0.13

Neighborhood Elites: Quality-Diversity Search for Uncertainty Quantification in Question Answering

cs.AI cs.CL ·2026-10-29 ·Mesut Ceylan, Arda Baris Basaran, Elgar Fleisch, Filipe Barata
Abstract

Large language models (LLMs) are increasingly capable, yet they produce wrong answers indistinguishable from correct ones. In question answering (QA), such an error propagates unnoticed into downstream tasks. Perturbation-based uncertainty quantification (UQ) addresses this by probing a model over a set of input rewrites and reading uncertainty from how much its predictions vary. Yet this set is usually built generically, by free paraphrasing, token-level perturbation, or adversarial search. Such rewrites leave two things uncontrolled that matter in QA: what kind of change is made to the context, and how explicitly the context supports the answer. When neither is controlled, the composition of the set is determined by the generator rather than by design, and the estimate inherits whatever that generator happens to produce. Quality-diversity (QD) search is built to cover a predefined exploration space, so we recast input perturbation as a QD problem and propose \emph{neighborhood elites (NE)}, a neighborhood constructor that spreads answer-preserving rewrites across two axes, the kind of edit and the explicitness of support. We compare NE against three other neighborhood constructors, holding the editor, QA model, uncertainty measure and search objective fixed, and against three scores read from the original context. Across five extractive-QA datasets and five instruction-tuned LLMs, NE improves failure detection against every other constructor in at least 24 of the 25 dataset-model pairs, reaching 0.726 AUROC and 0.211 AUPRC lift over the base rate. Additionally, NE detects failure better than semantic entropy computed over decoder samples, by 0.074 AUROC and 0.062 AUPRC lift, and this advantage widens as the neighborhood grows.

0.13

Preserving Utility in Continual LLM Unlearning via Orthogonal Projections

cs.LG cs.CL ·2026-10-29 ·Juan Belieni, Ana Carolina Erthal, Eliezer de Souza da Silva, Diego Mesquita
Abstract · 14 pages, 4 figures, 3 tables

Unlearning concerns the task of removing the influence of a sensitive portion of the training data, such as protected, copyrighted or private text, from a model without retraining it from scratch. While unlearning requests happen continuously during a model's lifetime, the vast majority of LLM unlearning methods are ill-suited for iterated removal, yielding steep decay in forgetting and utility upon sequential unlearning. This decay stems from interference between successive updates, as each new deletion can undo earlier ones or push the model further along directions already degraded by them. To address this, we introduce Continual Orthogonal Projection (COP), a gradient-projection scheme that restricts each update to the orthogonal complement of activation subspaces cached from prior forget sets. This approach provably guarantees first-order non-interference with earlier deletions. COP stores only a low-rank basis per layer, requiring neither past gradients nor previous forget sets, and plugs into unlearning methods that update the model through gradient steps on a differentiable loss without modifying its objective. Experiments on standard LLM unlearning benchmarks (TOFU and RWKU) show that COP preserves utility across sequential requests without weakening forgetting, preventing the collapse of standard objectives and forgetting more effectively than continual-specific baselines.

0.13

H$^2$R: Hierarchical Hint-Based Reward for Mathematical Reasoning via Reinforcement Learning

cs.LG ·2026-10-29 ·He Tu, Siye Li, Hongyi Wen
benchmark
Abstract

Reinforcement Learning with Verifiable Rewards has shown great promise in enhancing mathematical reasoning in large language models. However, existing approaches provide limited intermediate supervision for complex, multi-step reasoning tasks. To bridge this gap, we propose \textbf{Hierarchical Hint-based Reward}, a framework that leverages the intrinsic sub-question structure of mathematical problems as structured intermediate conditioning signals during RLVR training. Unlike chain-of-thought prompting, these hierarchical hints serve as logically grounded intermediate milestones that effectively reduce the reasoning horizon toward the final answer. To validate this framework, we introduce \textbf{HiMath}, a benchmark of hierarchically structured mathematical problems. Across diverse LLMs, hierarchical hints consistently improve reasoning performance even without task-specific training. When integrated into RL training via H$^2$R, they boost Qwen2.5-Math-7B accuracy on HiMath from 30.4\% to 39.1\% using hierarchical hints, and up to 45.6\% with Oracle Hierarchical Hints. This pronounced improvement demonstrates that incorporating reliable intermediate checkpoints into reward design is essential for sustaining complex mathematical reasoning.

0.13

Density-Guided Bootstrap Boundaries for Latent World-Model Reinforcement Learning

cs.LG stat.ML ·2026-10-29 ·Shashank Gupta, Pilhwa Lee
theory
Abstract · 37 pages, 5 figures

Latent world models allow reinforcement learning agents to learn policies from imagined trajectories, yet determining the optimal depth of model reliance remains a fundamental challenge. Fixed imagination horizons apply uniform prediction depths regardless of trajectory reliability, exposing target values to compounding model error or over-relying on critic approximations. We introduce density-guided bootstrap boundaries, a trajectory-dependent mechanism that adapts the depth of imagined returns using a score-informed relative density proxy. The proxy accumulates the estimated divergence of a score-corrected auxiliary field formed by a feature displacement network and a noise-conditioned score estimator along imagined paths. We propose two gating mechanisms: a hard boundary that halts return expansion at a discrete state, and a soft boundary that continuously shifts the return recursion toward critic bootstrapping. Both variants preserve boundary value estimates and apply corresponding survival weights to actor and critic losses. Formulated independently of specific dynamics models and instantiated within Dreamer V3, we prove that the resulting gated Bellman operator is a strict contraction mapping and that early bootstrapping monotonically reduces conditional target variance. We clarify the theoretical assumptions distinguishing our expansion proxy from exact path likelihoods. Across continuous-control tasks and Atari Pong, empirical evaluations show that dynamic boundary selection enhances sample efficiency and late-training performance over fixed-horizon and auxiliary-enabled baselines.

0.13

Closing the Log Gap: Minimax-Optimal Regret for Episodic RL

cs.LG stat.ML ·2026-10-29 ·Guy Zamir, Victor Boone, Matthew Zurek, Yudong Chen
theory
Abstract

In this paper, we study the regret minimization problem in episodic tabular MDPs. Our algorithm achieves the first expected regret bound scaling as $\sqrt{H^3 S A K}$ for sufficiently large $K$, where $H$ is the episode length, $K$ is the number of episodes, and $S$ and $A$ are the numbers of states and actions, respectively. Our method is the first to match the minimax lower bound \emph{without extra polylog factors}. We introduce a variant of UCBVI which does not maintain strictly optimistic estimates, and uses MOSS-style exploration bonuses whose logarithmic terms adapt to both visit counts and estimated variances. A key challenge is controlling the propagation of errors caused by under-optimism, which we accomplish in an occupancy-weighted way, avoiding logarithmic overhead that would be incurred by uniform-over-rows concentration. Our analysis develops time-uniform and variance-aware concentration bounds that may be of independent interest. As a corollary we also develop an algorithm for $(\varepsilon, δ)$-PAC best-policy identification, which is the first to match the $Ω\big(H^3 SA\log(\frac{1}δ)\varepsilon^{-2}\big)$ minimax lower bound up to constants in some regimes.

0.13

The First Few Tokens Matter: Understanding Chain-of-Thought Controllability

cs.AI ·2026-10-29 ·Hugh Van Deventer, Dmitrii Troitskii, Laura Ruis
Abstract · 20 pages, 15 figures

Chain-of-thought (CoT) monitoring is a promising way to oversee increasingly capable agents. It succeeds in part because current models appear limited in their ability to control what they verbalize in their reasoning traces. We investigate the source of this limitation and link it to an entropy collapse in the first few tokens following the opening thinking tag. We exploit this mechanism to recover CoT controllability through two inference-time interventions on initial tokens: prefilling, where even a single token can substantially increase controllability, and banning frequent opening tokens combined with in-context examples, which improves average compliance by 34-41 percentage points across models. These interventions recover mainly addition and style controllability (adding or transforming information), while suppression controllability (avoiding or removing information) exhibits more varied responses, suggesting distinct underlying mechanisms. Consistent with this, the two groups behave as separate capabilities: across models, performance correlates positively within these groups, while cross-group correlations are mostly negative. Importantly, we find that the relationship with monitorability depends on the evaluation: addition and style controllability more consistently positively correlate with monitorability, while suppression often shows negative associations. Our findings motivate the development of controllability evaluations that distinguish difficulty initiating a specific opening from general limitations of the ability to alter reasoning content, alongside separate reporting of control tasks and categories.

0.13

Softmax Routing Dynamics in Mixture-of-Experts Models: Collapse and Recovery

cs.LG ·2026-10-29 ·Shengze Xu, Maolin Deng, Tieyong Zeng
Abstract

Softmax routing in mixture-of-experts models can concentrate probability on a subset of experts, suppressing access to the others. We study how this concentration forms, why it persists, and how suppressed probabilities recover. For fixed experts, a stationary-loss barrier identifies the face supporting the optimal routed output. In a common-target square-loss model with freely trainable expert outputs, slow adaptation tracks fixed-expert dynamics on finite horizons, while every fixed positive adaptation rate yields an interior routing limit. An exact normal-velocity decomposition separates complementary-expert covariance from relative gradient advantage. Under positive two-sided power bounds along recovery paths, uniform over initial masses, the resulting coefficient's order of vanishing gives logarithmic or power-law recovery times. Integrations of constructed fields reproduce these orders. An exact discrete log-odds identity relates neural updates to threshold crossing. In paired Transformer-MoE training, removing a class-conditional router-gradient contrast delays confirmed recovery and reduces accumulated log-odds on fixed tokens, both with and without capacity constraints.

0.13

When to spend privacy: optimal allocation in private learning

cs.LG cs.AImath.OC ·2026-10-29 ·Umberto Biccari, Yongcun Song
theory
Abstract

Privacy accounting bounds the cumulative privacy cost of Differentially Private Stochastic Gradient Descent but does not determine how to allocate a fixed budget across iterations. We formulate the complete Gaussian noise-variance schedule as an open-loop control under a composed zero-Concentrated Differential Privacy constraint, explicitly accounting for schedule selection. For expected quadratic terminal tracking, we prove that every local minimizer exhausts the certified budget, yielding an exact-budget simplex formulation. We introduce Privacy-Budget Adjoint Mirror Descent (PB-AMD), combining pathwise adjoint gradients with entropic updates that preserve positivity, charge floors, and the prescribed budget. Under regularity and coercivity assumptions, we establish existence and budget stability of optimal schedules, design-to-training utility stability, and deterministic, stochastic, and sample-average stationarity guarantees. A linear-quadratic reduction yields an explicit square-root allocation law. Experiments demonstrate tracking improvements over several allocation rules and predictive improvements over uniform allocation in a learning-rate restart setting, where learning-rate allocation performs similarly. Comparisons at a common stationarity tolerance show lower schedule-optimization times than projected gradient descent on the tested problems. The framework connects certified privacy accounting with optimization through nonlinear training dynamics.

0.13

Incongruence-Aware Evaluation of Multimodal Speech Emotion Recognition in Audio Large Language Models

cs.AI cs.CLcs.MMcs.SD+1 ·2026-10-29 ·Fengji Jing, Guy Laban
code/data
Abstract

Voice-interactive large language models process both what a speaker says and how they say it, yet the two channels do not always agree. We ask whether that disagreement degrades speech emotion recognition. Mapping text and audio into a shared valence-arousal-dominance space, we convert the distance between them into a continuous, threshold-free incongruence score with a two-component Gaussian mixture, and validate the score against human judgement. We evaluate three audiocapable models on acted, conversational and spontaneous speech (eNTERFACE'05, MELD and CMU-MOSEI). With the model held constant across modalities, incongruence raises the odds of error whenever audio is present, in every corpus, after controlling for the ground-truth emotion and for distance in the feature space. Supervised classifiers trained on the same features behave in the opposite way, which suggests that the difficulty lies in how these models combine raw evidence rather than in the signal itself. Text-only effects are unreliable: swapping only the model that reads the transcript reverses their sign on one corpus. Rebuilding the score without the emotion lexicon reproduces the central result, and among six simpler alternatives it is the only label-free score that predicts error in all three corpora. Code, data tables and prompts are released.

0.13

Factorized Representations of Concurrent Temporal Variables in Pretrained Language Models

cs.LG cs.CL ·2026-10-29 ·Xinhe Zhang, Arnau Marin-Llobet
Abstract

Language models often need to maintain multiple contextual variables that evolve independently within a document. Prior work has identified structured representations of individual temporal or contextual quantities, but their joint organization remains less understood. We study this question using controlled documents with independently phased language-switch and topic-change timers. Across ten pretrained checkpoints, both timers are linearly decodable from the same hidden states. Removing a timer-associated centroid plane substantially reduces access to its associated variable while having little effect on the concurrent timer, although the planes are nonorthogonal and contain cross-timer information. Separate timer contributions also explain most reproducible variation among joint-state centroids. These signatures persist from 3 to 51 timer values and across Qwen3 checkpoints from 0.6B to 14B. Schedule position affects continuation probabilities, but the tested centroid displacement produces no consistent timer-specific output effect. Together, these results support approximate factorization based on selective accessibility and approximately additive joint-state geometry in overlapping subspaces.

0.13

Latent Prediction Yields Cross-Subject Alignable Representations in Sensory-Evoked fMRI

cs.LG cs.AIq-bio.NC ·2026-10-29 ·Serwan Jassim, Victoria Bosch, Daniel Anthes, Elia Bruni, Tim Kietzmann
Abstract

Decoding sensory-evoked functional magnetic resonance imaging (fMRI) data is a central analytical tool in cognitive neuroscience. Here we ask whether self-supervised learning (SSL), which has been scarcely explored in this domain, can yield representations that support strong downstream performance. To do so, we adapt a joint embedding predictive architecture (JEPA) to train participant-specific encoders on brain data alone. The encoders never see the stimulus images or any representation derived from them. Applied to the Natural Scenes Dataset (NSD), we evaluate our learned brain embeddings on the downstream tasks of brain captioning and image reconstruction, as well as cross-participant generalization. Our self-supervised approach achieves competitive performance with parameter-rich supervised pipelines that are trained end-to-end with access to image and language data. We further find that encoders trained independently for different participants can be aligned fully unsupervised, using neither images nor stimuli shared between them. After alignment, a neural decoding head trained on one participant can be applied to another without fine-tuning and retains most of its within-subject performance on image reconstruction and brain captioning. By removing the need for images and shared stimuli, our approach is a step towards pooling fMRI data across participants and, eventually, across datasets.

0.12

Beyond LLM Wrappers: Context Engineering and the Three-Engine Architecture for Reliable Financial AI

cs.AI ·2026-10-29 ·Pradeep Kumar GS, Ravi Tanniru
position
Abstract

Financial AI systems must support decision-bearing questions involving payments, collections, cash flow, and compliance while ensuring numerical accuracy, traceability, and auditability. Large language models (LLMs) enable natural language interaction but cannot independently guarantee reproducible calculations or evidence-grounded recommendations. We present a production-grounded architecture integrating three components: a financial knowledge graph for enterprise relationships, temporal context, and provenance; a deterministic optimization engine for financial calculations and decisions; and a multi-agent LLM layer for intent interpretation, task orchestration, explanation, and claim validation. A coordination layer, termed context engineering, selects, ranks, composes, and validates query-specific context across ERP systems, tools, and tenants. Based on observations from 15 deployments over approximately 11 months, single-agent pipelines exhibited hallucination rates of about 7-9% (5-10% across query types), including invented invoices and unsupported financial claims. Graph-backed multi-agent validation reduced observed rates to below 1%, although these findings are observational rather than causal. ERP data quality and tenant-specific schemas remained major implementation challenges. Reusable mappings, graph templates, and validation policies coincided with a reduction in deployment time from roughly six to two weeks. We argue that financial context engineering extends beyond text-centric retrieval-augmented generation by requiring structured relationships, temporal validity, exact numerical evidence, access controls, and validation against source records and solver outputs. We outline research challenges in evaluation, reproducibility, privacy, context quality, and cross-tenant generalization.

0.12

Statistical Analysis of Synthetic Data with Verification: A Linear Regression Framework

cs.LG stat.MLmath.ST ·2026-10-29 ·Heejune Sheen, Harrison H. Zhou
Abstract

Synthetic data are widely used for training modern AI systems. However, their statistical understanding remains limited. It is well recognized in statistics that repeated self-training on synthetic data can amplify errors inherited from pre-trained models, leading to model collapse. Verification plays a crucial role in selecting high-quality synthetic data to help prevent model collapse and improve model performance. We study verified self-training in a linear regression framework, where a pre-trained estimator generates synthetic data with injected generation noise and a verifier selects samples according to a verification threshold. Post-training on the accepted samples updates either all coordinates (full fine-tuning) or only an active subset (partial fine-tuning). Our analysis reveals three potentially counterintuitive effects. We establish that tighter verification does not always lead to lower prediction risk. We also find that a suitable level of injected generation noise can reduce prediction risk, while excessive noise can increase it. Moreover, full and partial fine-tuning can respond differently to the same verification rule, since pre-training error in the frozen coordinates affects both sample selection and post-training risk. We analyze these effects by deriving finite sample risk bounds for four separate regimes determined by the relative scales of the verification threshold, generation noise, and pre-training error. The bounds precisely characterize how verification affects prediction risk across these regimes. In particular, we quantify how the generation noise, the effective accepted sample size, and the covariance of the accepted covariates affect prediction risk. The analysis extends to iterative verified full fine-tuning, where sufficiently tight verification reduces the initial error contribution at a geometric rate.

0.12

Learned Terminal Costs for Short-Horizon Model Predictive Control of Autonomous Trucks in Interactive Highway Traffic

cs.LG cs.AIcs.ROeess.SY ·2026-10-29 ·Hooman Hamze, Benjamin Houshmand, Deepthi Pathare, Erik Börve, Morteza Haghir Chehreghani +1 more
Abstract · 13 pages, 2 figures, 8 tables

Autonomous highway driving for heavy-duty trucks is a challenging control problem. The greater mass, dimensions and intricate dynamics increase the requirements for cooperation from surrounding traffic in critical maneuvers, requiring a careful balance between safety constraints and efficient long-horizon decision making. Model Predictive Control (MPC) provides rigorous constraint satisfaction but relies on terminal cost which can limit its effectiveness at short, computationally affordable horizons. Reinforcement Learning (RL) captures long-term driving behavior but offers no constraint guarantees and tends toward overly conservative policies when trained under sparse collision penalties. In this paper, we propose a hybrid MPC and RL framework for highway driving in autonomous trucks, where a finite-horizon MPC, solved via dynamic programming, uses a pre-trained RL model for terminal cost estimation. We evaluate the framework in a simulation environment with reactive surrounding traffic and compare it against pure MPC with zero terminal cost, rollout with terminal cost estimation using a base policy, and pure RL. Results show that the hybrid approach achieves zero collisions with the lowest cost of any configuration.

0.12

MOSAIC: Addressing Disproportionate Task-Modality Dependencies via Modality-Partitioned Gradient Routing

cs.LG cs.CL ·2026-10-29 ·Aryan Chowdhury, Yizheng Sun, Thomas Carroll, Riza Batista-Navarro
Abstract · 9 Pages, 1 Figure, 3 Tables

In multi-task settings, combining multiple modalities allows task representations to draw on signals that a single modality alone would not capture. However, individual tasks may depend disproportionately on particular modalities, a structure that conventional fusion-based architectures do not model explicitly, instead enforcing a unified representation space. Consequently, a task's gradients can affect parameters tied to modalities it barely relies on, leaving features that compromise between competing demands. Although prior work has explored task-weighting and loss-balancing strategies, their effectiveness is limited by the absence of explicit modality separation, preventing models from optimising each modality according to its task-specific importance. To address these limitations, we propose MOSAIC (MOdality-Separated Architecture with Inter-modal Cross-attention), which preserves unimodal representations while enabling controlled interaction between them, and uses modality-specific losses to route gradients to their corresponding partitions. On the Corpus of Everyday Japanese Conversation (CEJC), MOSAIC improves average macro-F1 by 7.4 points across four tasks with a 72% reduction in backbone parameters relative to a fusion-based baseline. Our results demonstrate that modality partitioning yields consistent gains across tasks while requiring far fewer backbone parameters than a fusion-based design.

0.12

Reasoning or Pattern Exploitation? Mechanistic Insights from RL-Trained Language Models

cs.AI cs.CL ·2026-10-29 ·Aishik Nagar, Cheston Tan
Abstract

Large language models (LLMs) are increasingly described as acquiring "reasoning" skills after reinforcement learning with verifiable rewards. However, whether these gains reflect genuine reasoning or more superficial adaptations remains underexplored. In this work, we probe LLMs trained in a finite and exhaustively analyzable logical domain, namely Tic-Tac-Toe, and trace how internal representations evolve across reinforcement learning (RL) with Group Relative Policy Optimization (GRPO). Quantitatively, RL yields high accuracy and robustness across prompt variations, suggesting the emergence of game reasoning. Mechanistic interpretability, however, paints a different picture: features extracted with sparse autoencoders (SAEs) reveal that models primarily adapt to rely more heavily on information already explicit in the prompt, such as whose turn it is, game progression and board occupancy. In comparison, high-level concepts like board symmetries and strategic forks remain weakly represented, raising concerns that reasoning benchmarks overestimate model abilities. This suggests that RL-driven "reasoning" may conflate task-specific adaptations to input information with reasoning ability. We provide, for the first time: (i) a systematic interpretability pipeline tracing representation dynamics across RL training in small-scale LLMs; (ii) an extension of SAE-based feature discovery to hypothesis-driven testing in a finite logical domain; and (iii) an interpretability-based demonstration that RL post-training neither adds nor removes concept information but reorganizes which features dominate the representation, with causal interventions locating the decision downstream of it, in the late-layer state.

0.12

Finding Is Not Returning, Returning Is Not Reconstructing: A Multi-Dimensional Evaluation of Agent Social Re-entry

cs.AI cs.SI ·2026-10-29 ·Sheng Wan, Cong Ma, Sichen Dong, Qian'Ang Mao, Chenhan Zhang +2 more
Abstract

As Large Language Model (LLM) agents increasingly participate in persistent social environments, understanding their long-term social interaction capabilities becomes increasingly important. While existing research has advanced long-term memory and conversational context retrieval, less attention has been paid to whether agents can autonomously re-engage with previously inactive conversations. In particular, remembering a past interaction does not necessarily mean that an agent can identify when to revisit it, decide to return, or reconstruct the context needed for a coherent continuation. To investigate this gap, we introduce agent social re-entry as a multi- dimensional evaluation problem, distinguishing three complementary capabilities: Finding, Returning, and Reconstructing. We develop a shared-snapshot evaluation framework that measures these dimensions independently, allowing us to examine how conversational resurfacing and persistent memory relate to different aspects of social re-entry. Evaluations across six LLM configurations show that resurfacing substantially improves target conversation identification, while persistent memory is associated with higher rates of autonomous target-directed replying. However, neither improvement is accompanied by consistent gains in contextual reconstruction. These results highlight the distinction between accessing previous conversations, choosing to re-engage, and maintaining conversational continuity. Our framework provides a way to evaluate these capabilities separately rather than relying on memory retrieval or conversational response quality alone.

0.12

Interpretable by Design Tabular Prediction via In-Context Local Linear Learning

cs.LG ·2026-10-29 ·Ran Eisenberg, Steven Berger, Ofir Lindenbaum
benchmark
Abstract

In-context learning (ICL) lets tabular foundation models make predictions without dataset-specific training or fine-tuning. These models achieve strong predictive accuracy, but interpreting how individual features contribute to the prediction remains difficult. In biology, for example, feature contributions can help identify biomarkers or risk genes. Post-hoc interpretability methods assign feature importance scores, but these scores do not indicate what the model actually uses and can lead to explanation errors. We introduce TICL$^3$ (Tabular In-Context Local Linear Learning), an interpretable-by-design predictor that uses a tabular foundation model to guide a simple linear prediction rule. TICL$^3$ first learns a shared linear model to capture dataset-level feature effects. Then, for each query, it retrieves similar training samples using representations from the foundation model and fits a second linear model to the residuals in that neighborhood. The shared model captures broad effects, while the local model corrects errors among similar samples. Their combined coefficients determine each feature contribution to the final query-specific linear score, so the same model makes and explains the prediction. Across 386 classification benchmark-dataset evaluations, TICL$^3$ trails TabICLv2 by 1.1 percentage points and outperforms CatBoost in 266 evaluations. It improves average support F1 over MAPLE by 4.7 points and achieves the best performance on both real-data selected-feature utility metrics.

0.12

Topology Survives Shuffling: Robust Long-Context Memory Access Prediction for Graph Analytics

cs.LG ·2026-10-29 ·Dongyan Sun, Neelesh Gupta, Rajgopal Kannan, Viktor K. Prasanna
Abstract

Graph analytics generates irregular, topology-dependent memory-access streams via indirect traversals, limiting data prefetching. Accurate prediction enables timely prefetches that hide memory latency. However, ML predictors couple short histories with fixed horizons, fitting the training trace's node layout rather than learning transferable graph patterns. We propose LoCoMAP, a long-context representation-learning framework for memory-access prediction with independently controlled observation history and prediction horizon. We evaluate Connected Components (CC), PageRank (PR), and Single-Source Shortest Paths (SSSP) on six synthetic graph families, with Erdos-Renyi (ER) and Grid as random and deterministic controls. We test two interventions: same-family regeneration changes connectivity, while node shuffling disrupts layout but preserves connectivity. Across 18 graph-workload pairs, extending history from 8 to 128 accesses improves F1 by 7.8%, while extending prediction horizon from 8 to 512 improves F1 by 88.3%. For CC, regeneration is less disruptive than shuffling: without retraining, long context retains 95.4% of original F1 and outperforms short context in six families. Under full shuffle, long context improves CC F1 by 33.6% across ER, Barabasi-Albert, Copying Model, and Stochastic Block Model (SBM), with a 2.06x gain on SBM, but not on Watts-Strogatz or Grid, where near-zero delta mass shifts into heavy tails beyond the predictor's range. Long context outperforms short context on six real-world CC graphs. At full shuffle, long-context PR retains 57.6% of original F1 and averages 11.0% above short context. Both SSSP models retain about 30% of original F1 under full shuffle as distance updates access vertex data in a new order. Within these bounds, long-context predictors are robust across related graphs and layouts without retraining, enabling a generalizable graph prefetcher.

0.11

Bayesian Offline Multi-Agent Reinforcement Learning with Latent World Models

cs.MA ·2026-10-29 ·Xi Xiong, Lu Liu, Yiming Li, Jakob Foerster
theory
Abstract

Model-based offline multi-agent reinforcement learning (MARL) provides a promising approach to multi-agent decision-making under partial observability, where the global state is unavailable from offline data. However, limited dataset coverage can lead to substantial out-of-distribution (OOD) model errors, which may mislead policy optimization and degrade downstream performance. Existing model-based offline MARL methods typically rely on access to global states or mitigate model errors through pessimistic or conservative policy optimization, leaving the reliability and refinement of learned world models less explored. We propose a Bayesian offline MARL framework that infers latent global states from decentralized observation-action histories and iteratively refines a posterior world model using only offline data. Building on posterior information loss, we extend this criterion to the multi-agent setting as multi-agent posterior information loss (MPIL), which provides a reliability measure for world model refinement and offline evaluation. Theoretically, we establish an upper bound on policy regret that is monotonic in MPIL and further characterize its dependence on the size and state-joint-action coverage of the offline dataset. Experiments on partially observable multi-agent variants of D4RL MuJoCo tasks demonstrate competitive policy performance across diverse offline data conditions, while MPIL-guided refinement improves world model prediction and downstream policy performance. These results support a principled connection between world model reliability, offline data coverage, and policy performance in partially observable offline MARL.

0.11

Towards Understanding Score-Seeking in LLMs

cs.LG cs.AI ·2026-10-29 ·Aghyad Deeb, Ryan Greenblatt, Alex Mallen
Abstract

Production large language models sometimes reward-hack, exploiting flaws in how tasks are scored rather than performing the tasks as intended, sometimes in highly sophisticated and adversarial ways. We investigate the motivations driving this behavior in GPT 5.6 Sol, GPT 5.5, and a checkpoint of OpenAI o3 without safety training ("exp-rl-cap-late") by observing model behavior in agentic environments where the grader is leaked. In our environments, the models appear to be score-seeking: trying to score well according to graders that they believe assess their current actions, even at the expense of violating explicit user instructions. They appear to care about score itself rather than its downstream consequences, violating instructions at approximately the same rate whether the score is indicated to be used for training, for evaluation, or not at all. We observe evidence that the more recent models care to some extent about scores that do not assess their own current actions. GPT 5.6 Sol and GPT 5.5 help other agents 24.6 percentage points more often on average when the other agent stands to gain a score of 1 rather than 0 from being helped. However, when grading or monitoring another trajectory, we find models never collude by distorting its score.

0.11

Transformer-inspired architectures make machine-learned electron densities accurate and transferable enough to bypass self-consistency

cs.LG physics.chem-phphysics.comp-ph ·2026-10-29 ·Jeheon Woo
Abstract · 36 pages, 4 figures, 2 tables; Extended Data (4 figures, 2 tables) and Supplementary Information included

We introduce machine learning models that accurately predict the electron density, the central variable of density functional theory. They are built on Transformer-inspired architectures, which require little computation per parameter and can therefore be enlarged at low cost. The proposed models, SwinUNETR-V2 and its attention-free counterpart NeXtUNETR, achieve state-of-the-art accuracy on QM9, with normalized mean absolute errors in the density averaging 0.075% and 0.046%, respectively. The predicted densities are accurate enough to bypass the self-consistent iteration: a single Kohn-Sham step reproduces the self-consistent electronic structure. For NeXtUNETR, the median errors are 0.036 meV in the total energy, 0.65 meV/angstrom in the forces and 1.8 meV in the HOMO-LUMO gap. The forces not only recover the self-consistent minima in geometry optimization but also successfully drive ab initio molecular dynamics. Fine-tuned on small datasets, both networks also transfer accurately to systems containing elements absent from QM9 or larger than its molecules.

0.11

Meshless Latent Dynamics Models for Model Order Reduction of time-dependent Partial Differential Equations

cs.LG math.NA ·2026-10-29 ·Lorenzo Tomada, Federico Pichi, Gianluigi Rozza
theory
Abstract

Latent dynamics learning has emerged as a prominent approach for Model Order Reduction (MOR) of time-dependent, parameterized Partial Differential Equations (PDEs). Traditional autoencoder-based approaches are intrinsically tied to fixed spatial discretizations during both training and inference, motivating the development of meshless methodologies. In this context, conditioning provides a natural mechanism for representing temporal evolution through spatially continuous functions, enabling mesh-agnostic Latent Dynamics Models (LDMs). In this work, we introduce a unified mathematical framework for mesh-agnostic LDMs based on conditioning in Neural Fields (NeFs). Leveraging a taxonomy of conditioning mechanisms (weak, strong, and pure), we formulate meshless LDMs as NeFs conditioned on a latent state. We establish a Universal Approximation Theorem (UAT) for this class of models and analyze their stability by deriving Lipschitz bounds for different conditioning mechanisms. Our analysis reveals an expressivity--stability trade-off, providing insight into why more expressive conditioning, despite its higher representational capacity, can potentially degrade performance during temporal extrapolation. Finally, we provide a comprehensive comparison of meshless and mesh-based architectures in the continuous limit of the spatial discretization, showing that meshless models guarantee grid-independent Lipschitz bounds, unlike mesh-based counterparts. Our experiments across Burgers' equation, Navier--Stokes flow past a cylinder, and linear advection show that meshless LDMs can achieve predictive accuracy comparable to or better than their mesh-based counterparts while using only 2% of the trainable parameters and exhibiting nearly constant error across spatial discretizations.

0.11

IMPACT: Inverter-Informed Multimodal Learning for Photovoltaic Generation Forecasting

cs.LG cs.AI ·2026-10-29 ·Anushka Katiyar, Cong Feng, Jing Wang, Tristan Liu, Simon Julien +1 more
Abstract

This work develops a physics-informed multimodal deep learning framework for 10- to 110-minute-ahead probabilistic photovoltaic power forecasting. The forecasting model is built on a four-branch architecture that separately encodes plant-level temporal history and inverter-level spatial cloud dynamics through a graph attention encoder with a learned adjacency, known-future solar geometry, and the current meteorological state, fused through gated cross-attention. The geometry branch adds two forward-looking physical signals, spatial cloud heterogeneity and inverter ramp-down activity, each empirically weighted toward the horizon where it is most predictive. The decoder anchors its prediction to a smart persistence baseline, learning a horizon-conditioned correction, and the resulting quantile forecasts are calibrated online through a kernel-weighted adaptive conformal calibration step. The proposed model outperforms all 14 baselines, achieving a forecast skill score (FSS) of 0.2950, 0.099 above the next-best model. A branch-composition ablation shows that every branch contributes positively and the model retains substantial skill in challenging overcast and variable sky conditions (FSS 0.198 and 0.255).

0.10

Beyond Plausible Prediction: Safety Sufficiency in Web World Models

cs.AI ·2026-10-29 ·Zhifei Hu, Alexandra I. Cristea, Adel Bibi, Jialin Yu
benchmark
Abstract

Web world models increasingly predict the consequences of candidate actions before agents act, yet a plausible predicted future does not necessarily support the correct safety decision. In a matched human evaluation of 3,000 predictions from ten frozen open models, 86.1% are judged semantically plausible. However, several models exhibit highly asymmetric authorisation behaviour, combining low false-safe execution with poor recall of clarifiable ASK cases and a strong preference for BLOCK. This motivates safety sufficiency: whether a predictive interface preserves and exposes the policy-decisive information required to distinguish actions that may be EXECUTED, require clarification through ASK, or should be BLOCKED. We introduce SafeWebWorld, a benchmark for evaluating safety sufficiency across five web-safety domains through action-conditioned futures, policy-decisive evidence, and policy-conditioned decisions. Under a controlled Security-Aligned Supervised Fine-Tuning protocol, evidence preservation can improve substantially while authorisation-sensitive decisions deteriorate. To determine whether this deterioration reflects loss of authorisation-sensitive score structure or a failure of the decision readout, we introduce Latent Authorisation Recovery (LAR). Frozen-score recalibration reveals substantial but backbone-dependent recovery, while paired recovery provides only a modest additional gain. Yet within-family and matched-authorisation evaluations remain poor, showing that aggregate recoverability does not imply robust fine-grained authorisation behaviour. Overall, plausible prediction, evidence preservation, and aggregate recovery are each insufficient on their own to establish robust authorisation.

0.10

Achieving Minimax-Optimal First-Order and Best-of-Both-Worlds Regret Simultaneously in Multi-Armed Bandits

cs.LG ·2026-10-29 ·Mengxiao Zhang
Abstract

We study first-order and best-of-both-worlds regret guarantees in multi-armed bandits. Pogodin and Lattimore (2019) proposed an algorithm that combines minimax-optimal worst-case regret with a first-order guarantee whose loss-dependent term is $\mathcal{O}(\sqrt{K L^\star}\log T)$, where $T$ denotes the horizon, $L^\star$ denotes the smallest expected cumulative loss among fixed arms, and $K$ denotes the number of arms. However, a logarithmic gap remains between this loss-dependent term and the optimal first-order scale. In this paper, we remove this logarithmic factor by proposing an efficient and parameter-free algorithm that achieves the minimax optimal first-order regret guarantee $\mathcal{O}(\sqrt{K L^\star})$. Technically, our algorithm is built upon FTRL with a Tsallis entropy regularizer (Zimmert and Seldin,2021), with a learning rate adapted to the learner's cumulative loss and a sampling distribution determined by a scalar fixed point. This construction controls the combined estimation and stability cost in terms of the learner's loss. Moreover, we show that the same algorithm achieves a best-of-both-worlds guarantee, with $\mathcal{O}(\sum_{i\ne i^\star}\frac{\log T}{Δ_i})$ regret for i.i.d. losses with a unique best arm $i^\star$, where $Δ_i$ denotes the suboptimality gap of arm $i$.

0.10

PAPERBENCHX: Can AI Reproduce Science Across Disciplines?

cs.AI ·2026-10-29 ·Ruoyu Wu, Zengji Tu, Aolong Sun, Tianyi Ma, Junren Li +12 more
benchmark
Abstract

Before AI can be trusted to discover new science, it should be able to reproduce what science already knows. Reproduction provides a rigorous checkpoint for an AI scientist: the target claim is known, yet recovering it still requires translating an incomplete scientific description into an executable experiment and producing evidence that can be independently verified. Existing research-reproduction benchmarks largely focus on AI and machine learning, where experiments share relatively homogeneous software stacks. Extending this capability across scientific disciplines introduces a different challenge: agents must operate domain-specific scientific software, make physical and numerical choices, diagnose long-running computations, and distinguish valid evidence from merely plausible outputs. We introduce PAPERBENCHX, a benchmark of 93 expert-constructed paper-reproduction tasks spanning 12 research areas across electromagnetics, photonics, chemistry, materials science, biology, and robotics. Each task provides the source paper, a public task specification, and a domain-native scientific environment involving specialized scientific software. Agents must reconstruct a key experiment, execute and diagnose it, and deliver a runnable workflow that supports the target scientific claim. To evaluate scientific validity rather than surface-level agreement, we replay every submission from scratch and score only regenerated evidence against expert-validated criteria. Across 10 frontier agent configurations, the best configuration fully reproduces only 14% tasks under the 90% threshold. Meanwhile, mean rubric scores range from 27.53% to 59.08%, indicating that substantial partial progress rarely translates into complete scientific reproduction. These results reveal a large gap between paper understanding and reliably producing reproducible scientific evidence.

0.10

Beyond Selected Breakthroughs: Evaluating Language Models on Randomly Sampled Mathematical Conjectures

cs.AI math.HO ·2026-10-29 ·Helen Jenne, Davis Brown, Jesse He, Henry Kvinge
dataset
Abstract

The last year has seen a shift in AI-assisted mathematics, with frontier language models producing proofs of and counterexamples to long-standing open conjectures. However, studying only the current set of major breakthroughs provides an incomplete picture of LLM capabilities and the computational resources needed to achieve reported results. To address this, we introduce a new dataset, OpenConjecture, which collects conjectures from mathematics papers published on the arXiv since February 2026. We use these conjectures as a source of open mathematical problems, and apply frontier GPT models to a random sample of 200 of these conjectures over three passes using a reproducible protocol without problem-specific mathematical guidance. Rather than attempt expert verification of these candidate solutions, we analyze the types of problems for which LLMs produce solutions and the characteristics of these solutions. The models we test produce candidate solutions for roughly 60% of the conjectures, with counterexamples about 1.3 times more common than proofs. We estimate solution novelty, assess how solutions cite and relate to existing literature, and analyze whether the proposed resolution relies on either an error in or underspecification of the conjecture. We find that models sometimes propose resolutions that are outside the authors' intent (e.g., proposing the trivial group as a counterexample) and that a number of proofs follow closely from existing results, but that some resolutions exhibit at least moderate novelty. Our findings underscore the need for evaluations that go beyond correctness to assess fidelity to the intended statement, attribution, and the nature of the mathematical contribution.

0.10

bf16 Rounding Silently Discards Most RL Policy Updates: A Hidden Training-Inference Mismatch and an Unbiased One-Line Fix

cs.LG cs.AI ·2026-10-29 ·Andrew Gao, William Gao
Abstract

Reinforcement learning (RL) post-training of language models keeps fp32 master weights in the optimizer and serves rollouts from a bf16 copy produced by round-to-nearest conversion. At the learning rates in common use, most per-step updates are smaller than half a bf16 spacing and the conversion discards them: on Qwen3-1.7B only 0.4-0.5% of served weights change per step, and the policy that generates the training data lags the optimizer by 50-150 steps. Existing mismatch corrections cannot see this lag because they compare two copies downstream of the same rounding. We propose Fair Cast: export the master with an unbiased rounding rule (stochastic rounding or error feedback) and compute the training loss on the served copy. The change is one line at the export call, adds no hyperparameters and costs 22-38 ms per export of 1.7B parameters. Across 25 matched seed-pairs on Qwen3-0.6B to 4B with AdamW and Muon, Fair Cast is ahead on 18, tied on 6 and behind on 1, with a mean gain of 6.1 points; on zebra puzzles it reaches 0.57 +/- 0.16 over four runs where round-to-nearest collapses to 0.33 +/- 0.10. It also makes the saved fp32 checkpoint agree with the served model, and it exposes the implicit lag of round-to-nearest as a regularizer that can be restored explicitly when a task needs it.

0.09

Framework-Conditional Open Materials Generation

cs.LG cond-mat.mtrl-sci ·2026-10-29 ·Maya M. Martirossyan, Hillary Pan, Philipp Höllmer, Stefano Martiniani
benchmark
Abstract

Theoretical predictions of novel materials rely on a variety of techniques, from data-mining-based ion substitution methods to fully de novo generative models. Each approach imposes specific constraints and encounters unique challenges when extrapolating beyond the distribution of known materials to discover new ones. The foremost goal of materials generation is to produce stable, unique, and novel (S.U.N.) structures. Predicted structures that meet these criteria, however, typically result from novel chemical compositions mapped onto known structural prototypes or frameworks. We leverage this heuristic insight to present a new generative paradigm for the de novo prediction of inorganic crystalline materials constrained by their structural framework---i.e., the unit cell shapes and respective particle positions. Our method, termed OMatG-FC (Open Materials Framework-Conditional Generation), learns via a combination of discrete flow matching on the atomic species and stochastic interpolants for the volume scaling of the unit cell lattice, while fixing in-place the fractional coordinates and the unit-volume lattice vectors, dramatically reducing the dimensionality of the problem compared to unconstrained (framework-free) generation. Our method is more flexible than existing prototype-constrained methods, demonstrating capability for discovery of unseen, stable prototypes derived from known structural frameworks. We also investigate the number of unique frameworks in the MP-20 dataset, demonstrating the propensity of duplicate frameworks within the dataset and providing a framework-split version. We benchmark the performance of our model using the LeMat-GenBench suite using three ML interatomic potentials and achieve state-of-the-art S.U.N. rate below the convex hull, eschewing the need to learn atomic packings for materials discovery.

0.09

Timbre: Reliability-aware harmonic consensus for label-free multimodal test-time rexinforcement learning

cs.AI ·2026-10-29 ·Haotian Lyu, Haoyu Wang, Ru Wang, Feng Ji, Ziwen Li +2 more
benchmark
Abstract

Label-free test-time reinforcement learning adapts vision--language models using rewards derived from their own predictions. However, misinterpreted visual evidence and sensitivity to textual formulation can produce a consistent but incorrect consensus that majority voting reinforces. Cross-view aggregation can expose such errors, but image and text transformations may themselves degrade task-relevant evidence to different extents. We introduce \method, a reliability-aware harmonic consensus framework for multimodal test-time reinforcement learning. \method constructs a $2\times2$ grid combining original and perturbed images with direct-solving and reframing modes, selects pseudo-labels through weighted harmonic aggregation, and adapts modality-level weights using batch-averaged, temporally smoothed shifts in pseudo-label confidence as a heuristic reliability signal. We formulate consensus as a weighted $χ^2$-barycenter problem whose unique minimizer is the normalized weighted harmonic aggregate, without requiring correct-answer invariance. An exact support--dispersion decomposition characterizes when harmonic aggregation preserves or corrects arithmetic rankings, while a sensitivity analysis establishes how modality reweighting changes pairwise ranking margins, informing interpretation of the heuristic updates. Across six benchmarks and three Qwen3-VL model scales, \method outperforms TTRV in 16 of 18 benchmark--model settings on adaptation instances and 17 of 18 settings on held-out instances under matched estimation and optimization rollout counts. These results demonstrate improved target-task adaptation and transfer to unseen instances without further updates.

0.09

Training LLM Agents to Automate LLM Post-Training with RSI-Dojo

cs.AI ·2026-10-29 ·Shuzheng Gao, Zhaojian Yu, Penghao Yin, Yuling Shi, Kai Cai +2 more
Abstract

Recursive Self-Improvement (RSI) aims to enable AI systems to autonomously build and improve AI systems themselves, including models, harnesses, and underlying infrastructure. Within this broader agenda, a central capability is enabling agents to train and improve base models. However, existing studies on automated model training primarily focused on workflow- or harness-level optimizations, rather than internalizing the ability to conduct post-training into the model itself. Developing this capability requires executable model training environments that can produce high-quality trajectories demonstrating how to select data, train models, diagnose failures, and iteratively improve results. In this paper, we introduce RSI-Dojo, a collection of executable model training environments designed to teach LLM agents to conduct post-training. To construct RSI-Dojo, we develop RSI-Craft, an automatic pipeline that converts human research artifacts into executable model training environments. RSI-Craft automatically collects research papers from top venues, instantiates datasets and metrics into a unified interface, and reproduces evaluation dependencies. It then filters tasks by data size, execution reliability, and discriminability, yielding RSI-Dojo with 281 seed model training tasks. Building on RSI-Dojo, we employ a trajectory collection method with an improved harness to collect high-quality model-training trajectories from frontier models, then retain examples through rubric-based and outcome-based filtering. We fine-tune Qwen3.5-9B and Qwen3.6-35B-A3B on these trajectories. The resulting RSI-Dojo-9B and RSI-Dojo-35B-A3B improve over their base models by 8.74 and 16.87 PostTrainBench points, respectively. RSI-Dojo-35B-A3B reaches 29.95 on PostTrainBench and 53.29 on PostTrainBench-Lite, matching the performance of frontier models such as GLM-5.2 and GPT 5.5.

0.09

Distributional Reward Calibration and Budget-Neutral Credit Assignment for Multimodal RLVR

cs.AI ·2026-10-29 ·Lang Chen, Zongxin Xu, Borui Wang, Anran Liu, Shitong Chen
theory
Abstract

We revisit advantage estimation in Group Relative Policy Optimization (GRPO) and observe that any token-level advantage admits a two-stage factorization: a \emph{sample-level baselining} step that converts scalar verifier rewards into a comparable sequence-level signal, and a \emph{token-level credit assignment} step that distributes that signal over the trajectory. Existing estimators typically address only one of these stages: affine baselines preserve the shape of each task's reward law, so heavy tails and binary cliffs propagate into the gradient, while token-level credit rules tie the size of each token's update to raw confidence scales and need not conserve the total credit issued by the verifier. We propose \textbf{TACE} (Two-stage Advantage and Credit Estimation), a two-stage estimator that addresses both stages. At stage~(S1), we replace linear standardization with a one-dimensional optimal-transport map to a standard normal, which gives a common population reward law across tasks and caps finite-sample outliers deterministically. At stage~(S2), we rank-calibrate a stop-gradient evidence contrast within each trajectory and renormalize the clipped token weights to have mean one, so credit is redistributed across tokens while its total mass and sign are conserved. We prove population-level inter-task moment equity, a deterministic tail bound, and mass-conservation, boundedness, and sign-anchoring properties that affine baselines and unnormalized token weights do not satisfy. On a multimodal reasoning suite, TACE obtains the strongest results on a majority of the benchmarks but does not dominate every setting.

0.09

Turing LLM Post-Training: Calibrated Distillation and Efficient Dynamic-MoE Execution

cs.LG cs.AI ·2026-10-29 ·Yue He, Jiahui Hu, Shaman Tang, Hua Zhou, Qiman Wu +17 more
Abstract

We study post-training for dynamically routed mixture-of-experts (MoE) models, focusing on integrating independently optimized specialists while keeping consolidation efficient. Our approach combines calibrated multi-teacher on-policy distillation (C-MOPD), which matches each selected teacher on a local Top-$K$ support with teacher-wise loss normalization and clipping, with an execution path that uses post-capacity bounded-width dispatch, final-route replay, and asynchronous rollout--training overlap. On Turing-20B-A2B, the selected model achieves a four-domain normalized integration score of 0.9988 and retains at least 90.16\% of the measured specialist gain in every domain. At matched $K=32$, the full calibration objective scores 0.9962, compared with 0.9724 without calibration and 0.9834 without clipping. At a matched 8K response cap, the model exceeds both shared SFT and Qwen3-8B on all four reasoning-enabled mathematics and code benchmarks (unweighted mean: 85.92, versus 80.33 for Qwen3-8B and 74.21 for SFT), and improves over SFT on ten of twelve broader evaluation metrics. In a PPU-based profile, the combined execution configuration reduces consolidation step time from 398.7 to 139.3 seconds, an observed 2.86$\times$ speedup over serial execution. These results demonstrate strong specialist integration alongside more efficient consolidation, while reflecting the complete training and execution framework rather than any single component.

0.09

Opening the Box Changes the Cat: LLMs Detect Chain-of-Thought Monitoring - What now?

cs.AI cs.LG ·2026-10-29 ·Amir Jalilifard, Marcos Medeiros Raimundo, Anderson Rocha
Abstract

Chain-of-thought (CoT) reasoning allows large language models (LLMs) to generate intermediate reasoning traces in natural language while solving complex tasks. These traces can support safety monitoring by exposing potentially harmful plans before actions are executed. In this work, we investigate the conditions under which LLMs infer the presence of CoT monitoring from direct and indirect environmental cues and examine how this inference affects their subsequent behavior. Our observations suggest that oversight is an interactive process: while a monitor inspects a model's reasoning, traces left by monitoring interventions can allow the model to infer that its thoughts are being observed and to adapt its behavior accordingly. We show that even subtle environmental cues, such as timing-related signals from timeouts and operation durations, can reveal the presence of monitoring and trigger models to adapt their behavior to evade oversight. Across repeated episodes and interaction loops, we further observe that models can form statistical estimates of the monitor's effectiveness and use these estimates to guide attempts to evade detection. These findings highlight the need to evaluate CoT-based oversight under adaptive conditions, where the monitoring process itself may influence the behavior it seeks to observe.

0.09

Generating Critic-Guided Synthetic Experience in Latent World Models

cs.LG cs.AI ·2026-10-29 ·Jorge de Freitas, Christos Ziakas, Alessandra Russo
Abstract

Learning from offline datasets is limited in scalability, as collecting new experience is often costly or unsafe in the real world. World models offer an alternative way to generate new experience through a learned dynamics model, yet learning policies in latent space is limited due to compounding errors in imagined rollouts. To this end, we propose using the world model as a local transition generator, leveraging critic guidance to choose actions inside that model to generate synthetic experiences. We study action refinement methodologies, introducing Critic-Guided Synthetic Experience (CGSE) generation in world models. Specifically, CGSE augments an existing offline dataset by sampling candidate actions using an offline policy and refining each to maximize a first-order approximation of the $Q$-function within a trust region. We then use a frozen latent world model to imagine the future state given each updated action, and use the critic to choose the best one, appending this synthetic transition to the replay buffer. Our method requires no additional real-environment interaction or test-time planner, as it generates synthetic experience offline using the world model. On three OGBench manipulation tasks, CGSE improves an offline policy by a relative improvement of $7.1\%$ on the average success rate.

0.08

Learning Stabilizing Latent Feedback Controllers via Multi-Step Backward Losses

cs.LG math.OCeess.SY ·2026-10-29 ·Paul Lutkus, Lars Lindemann, Stephen Tu
Abstract

Can learned latent dynamics models support the design of feedback controllers that provably stabilize the original system? For linear time-invariant systems, we show that learning linear autoencoders and latent dynamics with multi-step prediction-reconstruction losses $\unicode{x2014}$ which we term "backward losses" as they encourage latent trajectories to reconstruct full state trajectories $\unicode{x2014}$ provably recovers the structure needed for stabilization. Specifically, a sufficiently long reconstruction horizon ensures that a minimum ${H}_\infty$-norm controller designed using the learned latent model stabilizes the original system through output-feedback via the corresponding encoder, provided the latent dimension is large enough to capture all non-stable modes of the underlying system. We also show why both reconstruction and a multi-step horizon are essential: forward prediction losses that require only low prediction error along latent trajectories admit global loss minimizers that discard unstable modes regardless of horizon length, while single-step reconstruction losses can yield latent models whose stabilizing controllers fail to stabilize the original system. Finally, experiments on a non-linear flexible-arm system illustrate the practical implications of our results for learning non-linear autoencoders and latent dynamics.

0.08

Compile Away Uncertainty: Self-Evolving Agent Skills via Programmatic Compilation

cs.AI ·2026-10-29 ·Chengyu Yang, Yunxiang Zheng, Zongzheng Li, Jiamei Li, Boxun Lan +4 more
Abstract

Self-evolving agent skills enable AI agents to accumulate and refine reusable procedural knowledge from interaction experience without updating model parameters. However, since such skills are typically represented as textual documents and executed through LLM reasoning, even previously learned procedures must be repeatedly interpreted and regenerated, introducing avoidable uncertainty into every execution. In contrast, program execution is deterministic and independently verifiable. Intuitively, transforming well-specified operations into executable programs can reduce reliance on LLM reasoning and, consequently, reduce execution uncertainty. Motivated by this intuition, we introduce SKillCompiler, a self-evolving skill framework that integrates programmatic compilation into the skill evolution process. Specifically, SkillCompiler adopts an executability-aware skill compilation strategy that decomposes skills into fine-grained operations, analyzes their executability, and constructs hybrid skill skeletons that retain LLM reasoning for context-dependent operations while compiling well-specified ones into validated programs. To avoid redundant program synthesis and enable executable knowledge to accumulate across evolving skills, we further design a program asset reuse mechanism that organizes validated programs as reusable assets, retrieves existing implementations whenever applicable, and synthesizes and verifies new programs only when necessary. Extensive experiments on five benchmarks and six model families demonstrate that SkillCompiler consistently outperforms state-of-the-art methods, achieving up to 32.33\% improvement in task performance.

0.08

Agentic World Modeling for Agent Test-time Scaling without Environment Replica

cs.AI cs.CLcs.MA ·2026-10-29 ·Wendi Li, Huseyin A. Inan, Wei-Ning Chen, Yanzhi Chen, Chetan Bansal +2 more
Abstract

Test-time scaling (TTS) improves large language model agents by sampling and selecting among multiple candidate trajectories, but common realizations assume that these trajectories can be executed in parallel. This assumption breaks in realistic deployments, where enterprise workspaces and cloud services are shared and persistent, costly or infeasible to replicate, and actions may introduce irreversible side effects or cross-trajectory interference. Existing generative world models avoid direct execution, but typically compress environment state into a single prompt, limiting their ability to reliably simulate transitions under partial observability. Instead, we introduce Agentic World Modeling (AWM), which reformulates next-observation prediction as a multi-turn evidence-acquisition task. For each hypothetical action, an agentic simulator adaptively queries the real environment through safe read-only interfaces, reasons over the returned evidence, and uses executable computation to construct the resulting observation without mutating shared state. Integrated with trajectory-level TTS, AWM enables parallel exploration over a single shared environment. Across $τ$-bench, AppWorld, and OfficeBench, AWM outperforms a single-shot simulation baseline by 40.07% on average, and surpasses a trained world-model baseline by 12.36% on average. These results establish interactive, read-grounded simulation as a practical foundation for safe and scalable agentic test-time scaling in shared, non-resettable environments.

0.08

Reward-Adaptive Co-Evolution of Policy and Rubric

cs.AI ·2026-10-29 ·Yizhu Chen, Jianguang Jiang, Jiameng Huang, Chuyu Fang, Bojun Zhang +3 more
Abstract

Rubric-based rewards provide fine-grained, interpretable supervision for open-ended reinforcement learning, where holistic scalar rewards struggle to capture multidimensional response quality. Static rubrics can lose discriminative power as the policy distribution evolves, while existing policy--rubric co-evolution methods rely on globally prescribed update schedules that ignore heterogeneous reward informativeness. We introduce RACE, a Reward-Adaptive Co-Evolution framework. At the group level, per-rollout-group training-signal routing classifies groups as informative, diversity-collapsed, or rubric-failed, updating the policy only with informative groups while caching rubric failures. At the phase level, event-triggered phase switching enters rubric optimization upon persistent reward saturation and returns when rubric-reward improvement plateaus. Rubric phases train the generator on cached pairs using a frozen judge and a scale-normalized pairwise-margin reward. Across both levels, online threshold adaptation adjusts routing and switching thresholds through proportional control. Under a unified evaluation protocol, RACE achieves the highest average performance and outperforms the strongest fixed-alternation baseline on all eight tasks, while improving rubric discrimination.

0.08

Smooth Action Chunking Via Movement-Primitive Representations for Imitation Learning

cs.LG cs.RO ·2026-10-29 ·Shota Nomura, Takayuki Osa, Yasuhiro Kato, Kazuki Ota, Naoki Shitanda +1 more
Abstract

Imitation learning has become a widely adopted approach for vision-based robot policies, owing to its conceptual simplicity, scalability, and ability to leverage large-scale datasets. As these policies have grown more capable, action chunking has become their standard interface. At each decision step the policy outputs several future actions at once. Backbones from diffusion policies to large vision-language-action models all share it. Whatever backbone produces the chunk, its temporal structure is learned only implicitly. The commanded trajectory is therefore jerky both within a chunk and between successive chunks. On real hardware this appears as visible vibration and as repeated saturation of the follower's velocity limit. We address this issue by incorporating a movement-primitive-based trajectory representation into action chunking. Specifically, a trajectory segment is represented as a linear combination of smooth basis functions, yielding smooth trajectories by construction while leaving the policy architecture, training objective, and inference procedure unchanged. The same representation also yields a closed-form, inference-time correction for the second source of jerk. We condition the predicted coefficients so that the new chunk passes through a via-point at the previous chunk's last commanded action. This correction needs no retraining and, unlike asynchronous Real-Time Chunking, applies to policies queried in one shot. We instantiate it on a one-shot (ACT) and an iterative (DiT) backbone and evaluate in simulation and on a real bimanual ALOHA platform. The representation lowers within-chunk jerk by an order of magnitude on the one-shot backbone and by 2.4 times on the iterative one, with task success statistically non-inferior. The conditioning then removes 73-84% of the boundary discontinuity, at negligible per-step cost.

0.07

Revisiting Over-refusal in Safety-aligned Large Language Models: A Representation and Optimization Perspective

cs.AI ·2026-10-29 ·Chao Chen, Pengyang Shao, Wei Qin, Shijie Hao, Fei Shen +1 more
Abstract

Over-refusal is a critical side effect of safety alignment in large language models (LLMs), causing harmless but superficially risky queries to be mistakenly rejected. Despite the effectiveness of existing mitigation methods, why safety-aligned LLMs may still reject harmless queries under correct safety supervision remains insufficiently understood. To address this gap, we analyze over-refusal through the lens of representation and optimization. We make two observations: 1) the representations of harmful and over-refusal samples are visually distinguishable at certain intermediate layers but increasingly overlap in deeper layers; and 2) harmful-sample gradients exhibit high intra-class similarity and concentrate in a low-rank subspace. Moreover, they are more strongly aligned with refusal-target gradients than with correct helpful-target gradients on over-refusal prompts. This imbalanced gradient coupling can bias over-refusal samples toward refusal, while the reduced apparent representational separation in deeper layers may make it more difficult for the model to maintain a clear boundary between harmful and over-refusal samples. To this end, we propose GOAL (Gradient Orthogonalization with contrAstive Learning), a safety-alignment framework for mitigating over-refusal. Specifically, GOAL consists of two components: 1) to prevent imbalanced gradient coupling from biasing the optimization of over-refusal samples toward refusal, we orthogonalize over-refusal gradients against the low-rank refusal subspace extracted from harmful gradients; and 2) to explicitly enhance the separation between harmful and over-refusal representations, we introduce a contrastive learning objective. Extensive experiments across multiple models and benchmarks demonstrate that GOAL effectively mitigates over-refusal while preserving safety performance and general capabilities.

0.07

SHPRR: An Algorithm for Pure Exploration under Constrained Regret

cs.LG cs.AI ·2026-10-29 ·Georgina Nouli, Flore Sentenac, Maximilian Schiffer
Abstract

Best Arm Identification (BAI) conventionally measures the cost of experimentation by the number of samples collected. In many operational settings, however, the cost of exploration relative to a trusted baseline already in place is equally important. We study BAI under both a fixed sampling horizon and a regret budget. Unlike generic resources considered in constrained pure exploration, regret has a distinctive structure: sampling better arms is less costly relative to the baseline. We theoretically show that the regret cost of exploration decreases across the phases of Sequential Halving, a widely used pure-exploration algorithm. This observation yields improved guarantees on the probability of identifying the best arm and motivates a more effective allocation of the regret budget. We propose Sequential Halving with Proportional Regret Rationing (SHPRR), which front-loads the regret budget to allow more exploration when many arms remain. Experiments on synthetic and real-world instances show that SHPRR improves best-arm identification performance over uniform regret allocation and existing baselines.

0.07

Before the Next Rollout: Diagnosing Skill Utility Shifts from Policy Updates

cs.AI ·2026-10-29 ·Yifan Wang, Zhenyi Gai, Yiwen Zhao, Mingxuan Jiang, Keyang Chen +3 more
Abstract

Skill-evolution methods typically analyze execution trajectories to guide agent skill refinement. However, policy updates can change the behavioral effects and marginal utility of an unchanged skill. Attribution based on earlier trajectories may therefore no longer reflect the skill's contribution under the updated policy. Moreover, trajectory failures can arise from multiple factors, making individual skill contributions difficult to isolate. We evaluate whether policy-update signals contain predictive information about changes in skill marginal utility. At fixed contexts from pre-update training rollouts, we measure how policy updates change the contrast in token preferences between skill and control conditions. Combining these interaction shifts with advantage estimates yields reward-directed readouts for diagnosing utility changes. The readouts reuse existing training rollouts and require only forward passes for single-step action scoring, without additional post-update rollout evaluation. Our experiments show that policy updates can alter skill marginal utility and that policy-update readouts carry diagnostic information for identifying skills with declining utility. Integrating readout-guided prioritization into skill evolution improves downstream performance while reducing policy and editor token consumption.

0.07

Composing and Steering Diffusion Models with Generalized Feynman-Kac Correctors

cs.LG ·2026-10-29 ·Wael Mattar, Daniel Freedman
Abstract

Diffusion models have become the leading paradigm in generative modeling, yet many applications require sampling from distributions that are not those of any single pretrained model, but are instead defined through combinations of several models, possibly tilted toward high values of a downstream reward. Naively combining the scores of the models yields incorrect samples, while existing corrections are restricted to specific families of targets. In this work, we introduce the generalized Feynman-Kac correctors (GFKC), a principled framework for sampling from an arbitrary target function of several pretrained diffusion models. We show that the geometric interpolation between an arbitrary initial function and the target function of the models satisfies a Feynman--Kac partial differential equation, and derive its potential in closed form. To evaluate this potential during generation, we further generalize the log-probability ODE to track the density of each model along trajectories driven by a different velocity field. The resulting algorithm operates entirely in the log domain, incorporates reward-based steering at no additional conceptual cost, and corrects the generation on the fly, without retraining or fine-tuning the models. We validate the method on analytic Gaussian models, where it closely matches the ground-truth targets, and on diffusion models trained on MNIST, where it samples from non-trivial combinations of the models and enables effective steering.

0.06

Fast and accurate neural triage decoding of ultra-high-rate quantum LDPC codes

cs.AI quant-phcs.ITcs.LG ·2026-10-29 ·Hong-Ye Hu, Zhenlai Wang, Binzhao Luo, Tao Xiang, Chen Zhao
Abstract · 5 pages, 4 figures with Methods. We will release the fully supplementary information in the second update soon

Quantum error correction is essential for fault-tolerant quantum computing. High-rate quantum low-density parity-check (qLDPC) codes have emerged as a promising approach to reducing the substantial physical qubit overhead of quantum error correction, enabling encoding rates exceeding $1/2$ in blocks of approximately one thousand physical data qubits. However, their practical implementation requires decoders that combine high accuracy with low latency to keep pace with quantum hardware. Existing decoders of high-rate qLDPC codes can achieve high accuracy but often incur substantial computational latency, whereas neural network decoders have yet to be demonstrated for high-rate qLDPC codes. Here, we introduce \emph{neural triage decoding}, a framework that uses syndrome consistency to adaptively allocate neural computation. A local neural message-passing network rapidly predicts error corrections, additional iterations reduce residual syndrome violations, and a rarely invoked diversification stage searches for consistent corrections to the remaining shots. We demonstrate that the \texttt{Triage} decoder can decode ultra-high-rate qLDPC codes encoding a few hundred logical qubits, including lifted-product and Kasai codes, achieving state-of-the-art accuracy comparable to the near optimal algorithmic decoders while reducing decoding latency by approximately one to two orders of magnitude. Our framework is inspired by a rigorous analysis of the exact bitwise marginal decoder, revealing a distance-independent syndrome-inconsistent error floor. In addition, we show the robustness of neural network decoders to mismatched error priors. Our results establish adaptive neural decoding as a promising route towards fast, accurate, and scalable quantum error correction in the near term.

0.05

Language, Vision & Media 254

BEACON: Distilled Begin/End-Aware Structured Decoding for Continuous Sign Language Segmentation

cs.CV ·2026-10-29 ·Guohou Zhang, Kalin Stefanov
published
Abstract · Accepted at the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) 2027. 12 pages, 4 figures, including supplementary mate…

In segmentation-based continuous sign language recognition, an untrimmed video is first partitioned into lexical units; segmentation is therefore a means to an end---downstream recognition. However, segmenters are usually judged by local boundary detection rather than by the quality of the lexical segments and the recognition they enable. We present BEACON, a begin/end-aware segmenter that combines four-class Outside/Begin/Inside/End (BIOE) labeling, constrained Viterbi decoding over eight legal transitions, and boundary knowledge distilled from a binary teacher into a shared student. Under strict no-cut evaluation with validation-based model selection, BEACON improves segment-F1 by 14.58/14.59 points and boundary-F1 by 2.17/1.22 points, and reduces the downstream word error rate (WER) by 9.10/4.15 points relative to the state-of-the-art in the PHOENIX-14/BSL Corpus. To examine which segmentation metric better reflects downstream recognition, we evaluated five distinct architecture families. Segment-F1 aligns more consistently with lower WER than boundary-F1 in both corpora (Spearman $ρ=-1.0/-0.9$ versus $-0.9/-0.8$). These results support evaluating sign language segmentation in terms of segment quality and downstream recognition utility rather than local boundary accuracy alone.

0.66

Unlocking Any-Order Generation in Pretrained Autoregressive Image Models

cs.CV cs.LG ·2026-10-29 ·Rishav Pramanik, Marco Pedersoli, Zhaozheng Yin
publishedpositioncode/data
Abstract · Accepted in NeurIPS 2026, Sydney as a poster

Autoregressive (AR) image generators are powerful in principle but inflexible in practice: training with a fixed raster-scan order hardcodes a single factorization of the image distribution into the model, leaving it unable to inpaint, outpaint, or edit at inference time. Existing solutions, including the modern discrete-diffusion family, train from scratch on a mixture of token orderings, paying roughly 3x the compute of standard raster-scan training; discrete diffusion further sacrifices compatibility with the AR inference ecosystem (KV caching, vLLM) by construction. We argue that this cost is unnecessary. A pretrained raster-scan AR model has already learned the joint distribution over image tokens; what it lacks is not knowledge but the ability to express that knowledge through non-sequential conditional pathways. We propose a recipe for adapting pretrained raster-scan models into any-order generators using two components: future-aware positional embeddings that resolve a target-position identifiability problem at negligible overhead, and an ordering curriculum that traces a stable optimization path from the pretrained solution to a good any-order basin. Applied to four base models across two architectures and two scales (LlamaGen-L/XL, RAR-L/XL), the recipe substantially outperforms MaskGIT on ImageNet inpainting (FID 6.21 vs. 10.44) and outpainting, is competitive with MAR-L, trades a 1.2-1.9 increase in class-conditional FID for this flexibility, and costs 13-33% additional compute beyond pretraining, while leaving the standard AR inference stack untouched. Inpainting, outpainting, and editing, our results suggest, need not be added to autoregressive image models. They can be unlocked from the ones that exist. Code is available at https://github.com/Ordered-AR/unlockedadaptation.

0.65

Soil Moisture Estimation in Precision Agriculture: A Knowledge Distillation Approach

cs.CV cs.LG ·2026-10-29 ·Mohammed Rakib, Vrinda Nagula, Arunkumar Bagavathi
published
Abstract · Accepted by MWAIS 2025

Accurate soil moisture estimation is critical for precision agriculture, influencing irrigation, fertilization, and harvesting strategies. Traditional methods primarily rely on meteorological and soil property data, while recent approaches incorporate visual modalities such as aerial and geospatial imagery. However, these high-resolution images are expensive and computationally demanding to analyze. In this work, we propose Soil Moisture Distillation (SMD), a Knowledge Distillation (KD)-based multimodal learning framework that leverages unimodal teachers to enhance multimodal soil moisture estimation. SMD surpasses not only traditional multimodal fusion techniques, including MIS-ME, but also outperforms unimodal models, demonstrating the benefits of guided unimodal knowledge transfer. Additionally, we conduct a comparative analysis of soil patches and crop images, showing that crop images provide superior visual cues, leading to improved prediction accuracy. SMD achieves a 2% improvement in MAPE for soil patches and 1.2% for crops over state-of-the-art baselines, highlighting the effectiveness of leveraging both meteorological and visual data through knowledge distillation for enhanced soil moisture forecasting.

0.65

Where Do Vision Language Models Fail at 3D Reasoning?

cs.CV cs.AI ·2026-10-29 ·Khai Le-Duc, Nguyen Kim Hai Bui, An Ngo, Nam D. Hoang, Yueqi Song +9 more
benchmarkcode/data
Abstract · Preprint, 150 pages

Vision-language models can get individual judgments right while struggling to recover and consistently use the exact properties of a 3D object. To characterize this gap, we introduce BedBench, a controlled benchmark of thirteen procedurally generated tasks spanning individual judgments and composite answers. Across twelve open-weight and four frontier models, per-color counting exposes a pronounced weakness: humans achieve 93.3% accuracy, compared with 7.2% for the best model. Targeted fine-tuning raises one model's accuracy from 3.0% to 68.6%, but reveals a further difficulty: its answers about the most common colors remain inconsistent with its own per-color counts. Supplying those counts as text improves its judgments, showing that recovering a property does not mean using it consistently. Reasoning, scaling, and added visual cues each improve individual judgments, but none makes exact answers reliable. To locate the difficulty, we probe Qwen3.5-27B: exact per-color counts remain recoverable at the language model's input, yet the model's answers rarely recover them (65.0% for the probe vs. 3.0% for the model). Image-removal controls further show that some apparent successes require no visual evidence. These findings motivate evaluating 3D reasoning by exact correctness, consistent use, and image dependence, rather than accuracy alone. Code and dataset are publicly available: https://github.com/leduckhai/BedBench

0.64

FDSML: Formula-Driven Supervised Metric Learning with Parameter-Aware Triplet Loss

cs.CV cs.AIcs.LGcs.NE ·2026-10-29 ·Hina Otake, Shinei Arakawa, Keitaro Tanaka, Yoshihiro Fukuhara, Hirokatsu Kataoka +1 more
publishedcode/data
Abstract · Accepted to WACV 2027. 24 pages (11 main + 13 supplementary). Project page: https://shineiarakawa.github.io/fdsml/

Formula-Driven Supervised Learning (FDSL) enables vision pre-training without real images, but existing methods cast pre-training as classification on discretized labels, discarding continuity in the generative process. We propose Formula-Driven Supervised Metric Learning (FDSML), which derives relational supervision from image-generation parameters beyond discrete class labels. FDSML combines FineFractalDB, our refined fractal dataset with higher intra-class variation and more stable structures, with a Parameter-Aware Triplet (PA-Triplet) Loss that assigns positives and negatives according to distances in the underlying Iterated Function System parameter space. This objective trains embeddings to respect parameter-derived relative similarity orderings. Experiments show that FDSML improves transferability to image retrieval while remaining competitive on downstream classification and improving robustness under several corruption settings. On ResNet-50, Recall@1 improves from 65.6% to 74.0% on Stanford Online Products and from 56.2% to 78.4% on Stanford Cars over FDSL pre-training. These results suggest that parameter-derived relational supervision provides a stronger synthetic pre-training signal than label-based classification for transfer beyond recognition.

0.63

Comprehensive empirical evaluation of feature extractors in computer vision

cs.CV cs.LG ·2026-10-29 ·Murat Isik
publishedhighly cited
Abstract · 26 Pages, 9 Figures, 6 Tables

Feature detection and matching are fundamental components in computer vision, underpinning a broad spectrum of applications. This study offers a comprehensive evaluation of traditional feature detections and descriptors, analyzing methods such as Scale Invariant Feature Transform (SIFT), Speeded-Up Robust Features (SURF), Binary Robust Independent Elementary Features (BRIEF), Oriented FAST and Rotated BRIEF (ORB), Binary Robust Invariant Scalable Keypoints (BRISK), KAZE, Accelerated KAZE (AKAZE), Fast Retina Keypoint (FREAK), Dense and Accurate Invariant Scalable descriptor for Yale (DAISY), Features from Accelerated Segment Test (FAST), and STAR. Each feature extractor was assessed based on its architectural design and complexity, focusing on how these factors influence computational efficiency and robustness under various transformations. Utilizing the Image Matching Challenge Photo Tourism 2020 dataset, which includes over 1.5 million images, the study identifies the FAST algorithm as the most efficient detector when paired with the ORB descriptor and Brute-Force (BF) matcher, offering the fastest feature extraction and matching process. ORB is notably effective on affine transformed and brightened images, while AKAZE excels in conditions involving blurring, fisheye distortion, image rotation, and perspective distortions. Through more than 2 million comparisons, the study highlights the feature extractors that demonstrate superior resilience across various conditions, including rotation, scaling, blurring, brightening, affine transformations, perspective distortions, fisheye distortion, and salt-and-pepper noise

0.61

ElasticFormer: Detecting Objects in HRW Shots via Elastic Computing Vision Transformer

cs.CV ·2026-10-29 ·Wenxi Li, Jingchen Huang, Chenyang Lyu, Moran Liu, Haozhe Lin +2 more
top venuebenchmarkcode/data
Abstract · Published at CVPR 2026. 14 pages including supplementary material. Code: https://github.com/Purecoo00/ElasticFormer

Recent advances in gigapixel-level imaging have brought High-Resolution Wide shots to the forefront of research. However, these images present significant challenges: extreme sparsity of foreground, gigapixel-level resolutions and diverse target counts. This makes traditional close-up detectors inaccurate and slow as they are overwhelmed by the background. Although previous research has explored sparse backbones, their fixed sparsity patterns lack the adaptability required to handle diverse target numbers. To address this, we introduce ElasticFormer, a sparse backbone that dynamically allocates computational resources based on foreground proportion. After scoring windows based on variance, proposed ElasticSelector module will predict the foreground proportion for top-k selection. The mechanism guides the model to select target-containing windows, scaling resources in areas where objects are clustered. We introduce a novel loss function combined with the 3-phase training strategy for ElasticSelector, allowing it to function properly when bounding box annotations are missing. A WSOD study is carried on PASCAL VOC 2007 to evaluate its extensibility. Further, ElasticNet is created to verify its backbone-agnostic nature. In experiments on the PANDA gigapixel benchmark, ElasticFormer reduces backbone FLOPs by 80% while achieving a significant improvement in AP50 when compared to fixed-ratio sparse methods.

0.60

Intrinsic Distribution Matching

cs.CV cs.AI ·2026-10-29 ·Deyuan Liu, Yansen Han, Kaiwen Zheng, Jun Xie, Jiacheng Liu +5 more
code/datacommunity pick
Abstract

One-step and few-step text-to-image generation must accurately render the details specified in the prompt despite limited opportunities for refinement. Distribution matching distillation (DMD) uses score differences to bring the generated distribution closer to a reference distribution. Following prior work, we estimate this difference with internal velocity branches, avoiding a separate fake-score network. We ask whether an explicit moment constraint improves this learned score supervision. A decomposition of the reverse-KL field into a Gaussian moment component and a non-Gaussian residual identifies statistics that can be constrained explicitly. We introduce Intrinsic Distribution Matching (IDM), which retains the full DMD correction and matches feature means and per-coordinate standard deviations using the generator's exponential-moving-average (EMA) copy, requiring no external feature encoder. Experiments use full-parameter and low-rank adaptation (LoRA) tuning on Z-Image-Turbo 6B and Qwen-Image 20B. At two network evaluations, IDM reaches 94.25% and 90.20% on Chinese and English LongText-Bench, the highest scores among the compared methods at this budget. Controlled ablations show gains from the added moment loss, while generator EMA features perform competitively with an external DINOv2 encoder across the five reported metrics. Code is available at https://github.com/inclusionAI/IDM.

0.59

BiHashFormer: Hash-Driven Dual-Branch Transformer for Efficient Object Detection in HRW Shots

cs.CV ·2026-10-29 ·Jingchen Huang, Wenxi Li, Chenyang Lyu, Moran Liu, Haozhe Lin +1 more
top venuebenchmarkcode/data
Abstract · Accepted at NeurIPS 2026. 34 pages including supplementary material and the NeurIPS checklist. Code: https://github.com/Purecoo00/BiHashForm…

Recent advances in gigapixel-level imaging have brought High-Resolution Wide shots to the forefront of research. However, these images present significant challenges: extreme sparsity of foreground, gigapixel-level resolutions and interleaving of foreground and background. This causes traditional detectors and attention mechanisms to be hindered by background, resulting in inefficiency and inaccuracy. To tackle this problem, we propose BiHashFormer, a dual sparsified branch transformer built on hashing. RoI Selector will predict the foreground proportion for top-k selecting target-containing windows. The compression branch processes windows selected and refines attention into HashAttention by discarding half of the key-value pairs, reducing cost and improving efficiency. The compensation branch uses HashMiner to perform low-cost hash searches in the remaining regions, recovering features of missed objects. Experiments confirm that the selection relationship in HashAttention is one-way and reveal the preference of different queries for keys. In experiments on the gigapixel benchmark PANDA, BiHashFormer reduces 37.5% backbone FLOPs while improving AP50 to 81.0%. Moreover, this parallel dual-branch structure allows the model to avoid inference delays caused by inter-branch dependencies, while achieving much greater stability than single-branch models, specially with low rates of RoI selection.

0.58

COGS: A Diagnostic Benchmark for Attribute-Aware Object-Level 3D Gaussian Splatting Perception

cs.CV ·2026-10-29 ·Zheng Xiao, Yifan Liu, Weijian Deng, Jiashu Li, Ziqin Liu +2 more
publishedbenchmarkdatasetcode/data
Abstract · Accepted at WACV 2027, Datasets and Benchmarks Track. 20 pages, including references and supplementary material

3D Gaussian Splatting (3DGS) is increasingly explored as an input representation for 3D perception. Yet it remains unclear how current 3D architectures exploit heterogeneous Gaussian attributes. This ambiguity makes accuracy hard to interpret and limits principled architecture design for 3DGS perception. However, existing object-level 3DGS datasets lack the appearance variation or artifact control needed for systematic attribute-level analysis. We introduce COGS, a large-scale 3DGS dataset with 15,675 artifact-filtered and appearance-rich objects across 40 categories. Built on COGS, we establish COGS-Bench for object classification and attribute-usage diagnosis, complemented by external validation. Our benchmark shows that Gaussian attributes are not uniformly beneficial: selected subsets can outperform the full attribute set, and the effects of appearance, opacity, and covariance-related attribute combinations vary substantially across architectures. We further study this finding with HALO, a reference model that preserves attribute-specific evidence before hierarchical aggregation. Experiments on COGS-Bench and external 3DGS datasets demonstrate that COGS supports systematic diagnosis of 3DGS attribute usage, while HALO highlights the value of role-preserving attribute organization. Our dataset is available at https://huggingface.co/datasets/XiaoZhengCV/COGS.

0.56

NeurEO: dissecting Earth observation embeddings with computational neuroscience

cs.CV ·2026-10-29 ·Thijs L van der Plas, Jacob JW Bakermans, Ioannis N Athanasiadis
Abstract · Originally published at the EurIPS 2025 REO workshop

Earth embeddings are compressed representations of rich Earth observation data that enable downstream models to tackle a wide range of complex geospatial tasks. Yet, this increase in performance comes at the cost of decreased interpretability, as the semantic meaning of Earth embeddings is unknown. To address this challenge, we borrow methods from the field of computational neuroscience, which aim to interpret activity in (biological) neurons. Here, we adapt and apply these methods to assign meaning to AlphaEarth embeddings. We analyse embedding features both as stand-alone data, equivalent to spontaneous neural activity, and in relation to land cover and elevation correlates, equivalent to stimulus-response neural activity. We demonstrate correlational and multi-scale structure in the embedding features. Embedding features can display highly specific and directional tuning to land cover and elevation, such as a `North vs South woodland edge' response. When we remove the main trends of combined land cover responses (like water versus land tuning), we discover fine-grained representational details with increased effective dimensionality. Together, our analyses start to unravel what these Earth embeddings represent, which is critical for developing trustworthy downstream applications.

0.54

DOSE: Dominance-Guided Ordinal Severity Estimation from Sparse Relative Annotations

cs.CV ·2026-10-29 ·Tushir Sahu, Hideaki Hayashi, Hajime Nagahara
publishedcode/data
Abstract · Accepted to WACV 2027. 8 pages, 4 figures. Adjust the counts if you add the supplementary material

Medical severity estimation commonly relies on absolute grade annotations, which are fundamentally ordinal, yet existing methods often treat the grades as independent categories. Learning-to-rank methods, which can be trained using relative annotations based on severity comparisons of image pairs, can provide continuous severity rankings but cannot directly predict discrete severity grades. We propose Dominance-guided Ordinal Severity Estimation from Sparse Relative Annotations (DOSE), a framework for learning ordinal medical severity from sparse pairwise comparisons without using absolute annotations. DOSE aggregates the observed comparisons through a global comparison graph to estimate a dominance score for each image. Differences between dominance scores set cumulative margins in the embedding space, so that pairs with larger estimated severity differences are separated more strongly, and also generate pseudo-pairs that supervise a scalar severity head alongside the observed comparisons. The learned continuous severity axis is converted into discrete grades via label-free score calibration, with tie-anchored segmentation as the main calibration method. In the experiment, DOSE consistently outperformed pairwise-ranking baselines on ACNE04, IDRiD, and LIMUC under different pair budgets. Code is available at https://github.com/TushirSahu/DOSE

0.53

Video Generation Meets Agent Harness: Agentic Video Generation, Agentic World Modeling and Recursive Self-Improvement

cs.CV ·2026-10-29 ·Liaoyuan Fan, Yuhao Zheng, Dongxing Mao, Hangyu Ran, Siyuan Hu +16 more
code/datacommunity pick
Abstract · website: https://agentic-videogen.github.io/ github: https://github.com/agentic-videogen/awesome-agentic-videogen

Video generation is shifting from end-to-end synthesis toward agentic generation, where planning, tool use, and feedback enable coherent, controllable multi-shot videos. In parallel, video world models provide action-conditioned observations and predicted futures, serving as interactive environments for agents. We study their intersection through the agent harness: an executable runtime that constructs context, exposes tools, maintains state, routes feedback, and manages budgets around video models. This report examines three directions. First, agentic video generation, where harnesses coordinate scripting, shot planning, asset management, production, and editing, alongside the evaluation challenges of these systems. Second, agentic world modeling, where harnesses mediate observation, control, execution, and state. We discuss hybrid designs in which game engines or agent-written programs maintain world state while video models render observations, and how agents and world models can mutually evaluate and improve one another. Third, recursive self-improvement (RSI) in video model development, distinguishing output revisions, updates to harness state, generator parameters, and evaluators, and changes to the improvement procedure itself. We identify how these improvements feed back into subsequent development. Across these directions, we analyze essential agent capabilities, including audiovisual perception, long-horizon planning, memory, code-based interaction, feedback-guided control, skill reuse, and multi-agent coordination. We conclude with challenges in autonomous video production, lifelong memory, programmable hybrid control, autonomous model development, provenance, cost accounting, and evaluation linking video quality and task performance to sustained improvements in models and harnesses.

0.52

SynthFaces: Balanced Large Scale Human Dataset

cs.CV ·2026-10-29 ·Sergey Karpukhin, Denis Bobkov, Kirill Morozov, Aibek Alanov
publishedcode/data
Abstract · accepted at BMVC 2026 github: https://github.com/ControlGenAI/SynthFaces

The pursuit of fair and equitable AI in human-centric tasks is critically hindered by the inherent biases and privacy risks of real-world datasets, which offer imbalanced representations of identity and often contain sensitive personal information. Consequently, state-of-the-art generative models trained on this skewed data exhibit unfair performance, particularly in identity-preserving generation. In this work we propose a pipeline for controllable generation of dataset with diverse faces for generative models training and finetuning. We achieve this by expanding foundational Arc2Face diffusion model with a flow matching generative model in the manifold of identity embeddings. We demonstrate that such approach achieves both goals of generative fidelity and label diversity not present in real human-centric datasets. As a result we collect and further contextualize a highly balanced identity-centric dataset of humans named SynthFaces. By design, our released images and identities are synthetic, reducing direct reliance on real personal images; we empirically check identity separation with a nearest-neighbor non-leakage metric. We purpose it for training and tuning of fair models for various downstream tasks, and showcase its usability in finetuning an existing adapter-based method for personalized generation.

0.52

Incentivizing Self-Learning in Large Language Models via Reinforcement Learning

cs.CL ·2026-10-29 ·Lingkai Kong, Yuzhe Gu, Haiteng Zhao, Zijian Wu, Youqing Fang +4 more
community pick
Abstract

Large language models (LLMs) draw on parametric knowledge and retrieval-augmented generation (RAG) to solve knowledge-intensive tasks. Yet problem solving in unfamiliar domains calls for tightly coupled knowledge acquisition and reasoning, beyond factual retrieval and aggregation or learning solely from a preassembled context. We study self-learning -- autonomously retrieving, studying, and applying knowledge from long-form reference materials on the fly. To investigate this capability, we formulate a textbook-based mathematical problem-solving task. LLMs must navigate textbook exposition, learn and apply relevant principles, and construct proofs for exercises, with reference solutions withheld. We introduce a BookNavigator interface to support this human-inspired study process and incentivize self-learning through reinforcement learning (RL) with fine-grained rewards. Experiments show that our method improves performance across in-domain and out-of-domain textbooks and subjects, raising Intern-S2-Preview-35B's overall proof accuracy from 51.5% to 79.5% and outperforming RL-trained textbook-free and fixed-retrieval baselines. These gains extend to complex long-context benchmarks involving different task types, with improvements over the base model of 1.7 and 2.3 percentage points on BrowseComp-Plus and CL-bench, respectively. Together, these findings suggest that RL can foster a domain-agnostic self-learning strategy, highlighting the potential of self-learning from long-form corpora to strengthen broader problem-solving capabilities on complex, long-context tasks.

0.50

WorldProbe: When Video World Models Fall Apart under Long-Horizon Spatial Exploration

cs.CV ·2026-10-29 ·Minqi Wang, Jie Peng, Hejun Dong, Ruchang Yao, Puyi Wang +6 more
benchmarkcode/data
Abstract · 19 pages, 9 figures

Video world models generate future visual observations from camera or action inputs, with the goal of synthesizing persistent 3D worlds that can be explored across time and space. Existing benchmarks typically evaluate short interactions or rely on camera trajectories fixed before generation. The performance of video world models under long-horizon exploration with extensive camera motion remains under-explored. We introduce WorldProbe, a diagnostic benchmark that uses a goal-conditioned multimodal agent to adapt camera actions online during generation. WorldProbe comprises 164 test cases across four probing families. Across 10 state-of-the-art models, we generate more than 1,600 rollouts totaling approximately 800 minutes, with horizons up to 180 seconds. Our evaluation across four dimensions reveals both the degradation of local measures and the emergence of long-range inconsistencies. We find that recent models sustain visual and content quality for longer, while long-range memory and evolution remain weak. These results expose a gap between locally stable generation and persistent world consistency that is largely missed by existing benchmarks. Code and benchmark resources are available at https://github.com/Untitled-W/WorldProbe.

0.49

Glance: The Generator Is Its Own Reranker

cs.CL cs.IR ·2026-10-29 ·Adilet Metinov, Chingiz Arziev
code/data
Abstract · 12 pages, 2 figures, 6 tables. Adapters and code: https://huggingface.co/airun-ai/GlanceRAG-MiMo-9B-Reranker

Retrieval-augmented generation commonly serves separate models for reranking and answering, duplicating model residency and passage processing. We introduce Glance, which equips the generator with a lightweight decision adapter and pointer head to score candidate passages from a shared prefill. Short relevance probes branch from the cached context, producing one score per passage without separately prefilling each query-passage pair. A single resident 9B backbone supports generation, general decisions and reranking, with approximately 0.1 GiB of additional memory per decision skill. After 39 minutes of MS MARCO fine-tuning, Glance achieves nDCG@10 of 0.741 on SciFact and 0.345 on NFCorpus, compared with 0.721 and 0.326 for bge-reranker-v2-m3. On 300 HotpotQA questions, its selected context improves answer F1 by 0.045 over cross-encoder-selected context under the same base generator (paired bootstrap 95% CI [0.014, 0.076]). In a scaling study, reranking quality keeps rising up to 200 candidates (43k tokens) scored from one shared prefill, where it exceeds the cross-encoder by 0.058 nDCG@10, and a probe-only activated adapter lets generation continue from the original passage cache. These initial results demonstrate generator-based reranking with one resident backbone; evaluation currently covers two retrieval datasets, one QA dataset and one backbone family.

0.48

UGR: Unified Generative Ranking for Heterogeneous E-commerce Recommendation

cs.IR ·2026-10-29 ·Zonghe Shao, Huabin Chen, Sitong Lu, Fangmei Zhu, Yingcai Ma +5 more
hot direction
Abstract

List-wise ranking in e-commerce recommendation must balance user utility and commercial value across organic content, ads, and candidates from other businesses, while modeling candidate interactions and sequential dependencies. Existing generative ranking methods typically learn from logged exposure sequences and operate on truncated candidate sets from a single business. Exposure-sequence learning may not fully exploit post-exposure feedback signals. Meanwhile, generative ranking over heterogeneous, multi-business candidates remains less explored, particularly when recommendation scores and advertising value must be jointly modeled under generalized second-price (GSP) auctions. We propose Unified Generative Ranking (UGR), which performs generative ranking over the full request-level multi-business heterogeneous candidate set, typically comprising around a thousand candidates in production, and directly generates multi-business ranking lists through candidate-aware autoregressive decoding. To go beyond exposure-sequence imitation, UGR introduces complementary feedback supervision through weighted supervised fine-tuning and permutation preference optimization, which respectively leverage observed post-exposure feedback and feedback-derived preferences among permutations. UGR uses a training-time advertising debiasing tower to model historical ad exposure effects and incorporates current ad value during serving while preserving the existing allocation and GSP pricing rules. Multi-value sequence retrieval produces candidate ranking lists, and the existing downstream evaluator selects the final list for exposure. Offline experiments demonstrate consistent gains in ranking quality, while online A/B tests on Taobao's Guess You Like recommendation service yield improvements of 1.76% in Core IPV, 3.29% in Transaction Volume, and 2.10% in GMV over the production baseline.

0.48

rTaste: Do Research Agents Have Taste? Evaluating Early Stage Research Judgment

cs.CL cs.AI ·2026-10-29 ·Art Kanke
code/data
Abstract · 27 pages, 1 figure, 9 tables. Code and data: https://github.com/ArtKanke/rTaste

Do research agents have taste, the judgment of which experiments are worth running before their outcomes are known? We give six frontier models seven recent NLP papers. Each model, working as a research agent alone or as a team, proposes minimum viable experiments (MVEs), forecasts whether each will pass a threshold it sets and ranks them. A separate executor then attempts every MVE, including those ranked last, and three model auditors judge the results. Of 415 MVEs, 194 are judged sound, valuable and new, but fewer than half of these (73) are also judged beyond the seed paper and non obvious. The agents' first choices match the auditors' most valued experiment more often than a random pick, but are not shown to improve on choosing the experiment written first. Their forecasts beat a base rate but are not measurably more accurate than those of a fresh instance of the same model given only the paper and registrations. Teams cost more without a clear overall improvement in proposal quality. The agents produce many worthwhile experiments, but the evidence that their rankings improve the choice of which to run remains limited. The evaluation harness and a dataset of all 415 MVEs, with their forecasts and audited outcomes, are available at https://github.com/ArtKanke/rTaste.

0.47

CASE: Concept-Aware Segmentation from Positive and Negative Visual Exemplars

cs.CV ·2026-10-29 ·Yi-Jen Tsai, Yen-Yu Lin, Chien-Yao Wang
publishedbenchmarkcode/data
Abstract · Accepted to ACCV 2026

Few-shot semantic segmentation typically identifies target objects based on positive support examples of a single semantic class. In practice, however, positive cues alone are often underspecified: for instance, "red cars" may be confused with cars of all colors, "vehicles" may include cars that should be excluded, and "material-defined objects" may span multiple categories. Fulfilling precise requests requires positive exemplars to provide inclusion evidence and negative exemplars to act as intent-defining constraints for exclusion and disambiguation. We formulate this setting as Specified Visual Prompt Segmentation (SVPS), where the target concept is specified by both positive and negative visual prompts. In this work, visual exemplars refer to support-image regions specified by bounding-box prompts. To study SVPS, we introduce CASE (Concept-Aware Segmentation from Positive and Negative Visual Exemplars), a benchmark built on COCO and PACO. CASE encompasses six category- and attribute-level task families: exclusion, isolation, union, attribute exclusion, attribute isolation, and attribute union. We evaluate SAM 3 under this setting and find that, despite its powerful positive-only few-shot segmentation capability, frozen SAM 3 does not consistently treat negative visual prompts as exclusion evidence. To address this, we present Concept Prompt Adaptation (CPA), a conservative LoRA-based adaptation baseline that improves positive-negative prompt interpretation on CASE while remaining competitive under positive-only few-shot segmentation. Code is available at https://github.com/WongKinYiu/FSS-CASE.

0.47

A Unified Benchmark for Aerial-Enhanced Autonomous Driving: The KITScenes Aerial Dataset

cs.CV ·2026-10-29 ·Daniel Lengerer, Richard Schwarzkopf, Jonas Merkert, Fabian Immel, Rupert Polley +6 more
publishedbenchmarkcode/data
Abstract · Accepted at the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV 2027)

Aerial imagery provides complementary spatial context for autonomous driving by exposing road structure beyond the immediate perception range of onboard sensors. Yet map construction from aerial imagery, aerial-assisted online HD mapping, and cross-view localization are typically studied on separate data and benchmarks. We introduce KITScenes Aerial, an extension of the KITScenes Multimodal dataset that augments synchronized onboard sensor data, vehicle poses, and rich Lanelet2 map annotations with spatially associated aerial imagery and higher-resolution neural representations, providing a common basis for a variety of research tasks. Our mapping evaluation focuses on all 53 KITScenes map classes across different model designs, spatial ranges, and aerial resolutions, while retaining compatibility with the widely used three-class setting. Representative baselines show that aerial context is particularly beneficial for fine-grained and extended-range mapping, more than doubling 53-class mAP at the larger mapping extent. Higher-resolution aerial imagery yields a further 4.3 mAP (+24 %) improvement, with gains extending to challenging map elements such as traffic lights. We additionally evaluate cross-view localization under geographically disjoint conditions, where an adapted fine-grained feature-matching baseline achieves a mean translation error of 5.13 m. KITScenes Aerial thus enables unified study of aerial and ground-level information across complementary mapping and localization tasks. Project page: https://kitscenes.com/aerial

0.46

Robust LIDAR Sensor-to-Sensor Domain Generalization via Accumulated-Point Scene Completion

cs.CV ·2026-10-29 ·Christoph B Rist, Larissa T Triess, Markus Enzweiler, Dariu M Gavrila
published
Abstract · 2026 IEEE Intelligent Vehicles Symposium (IV)

We study domain generalization for LIDAR semantic segmentation. Deep neural networks achieve strong performance on 3D point clouds, yet real-world LIDAR exhibits sensor-specific sampling patterns that hinder transfer across datasets. In many applications, performance on completely unseen sensors is crucial. A model should preferably work on a new setup without prior access to its data or labels. To this end, we introduce a source-only training scheme that adds an occupancy prediction auxiliary objective. We aggregate source-domain scans into a simple world model and use it as the supervision signal, replacing single-scan standard segmentation training. A key property of our approach is that it does not alter the input scan. During training and testing the network consumes the same single-frame LIDAR input, only the supervision signal is enriched via the occupancy objective. At deployment, there is no runtime overhead as the auxiliary head can be dropped for pure segmentation, or kept to obtain scene completion alongside semantic segmentation. In an experimental evaluation across distinct LIDAR datasets, we demonstrate that a model's generalization to new domains can be significantly affected by how it is supervised. In certain setups, performance on unseen domains improves without any explicit domain adaptation (no target scans, no target labels, no sensor priors) and without prediction refinement. At the same time, our method maintains competitive source-domain segmentation performance and incurs no additional inference cost.

0.46

PriorMaskMap: Robust Online Vectorized Map Construction with Biased Priors

cs.CV cs.RO ·2026-10-29 ·Haoming Xu, Wei Li, Yu Hu
publishedcode/data
Abstract · Accept by ECCV2026

Online vectorized map construction provides autonomous vehicles with essential, real-time semantic and geometric scene understanding. While leveraging prior maps presents an opportunity to improve accuracy, they are often biased due to outdated information or temporary road changes. Existing methods, which typically treat priors as fully reliable or only consider limited noise patterns, suffer severe performance degradation when these priors are imperfect. To address this, we propose PriorMaskMap, a robust framework for prior-guided vectorized map construction. Its first innovation is a Prior-Aware Confidence Estimator (PACE) that evaluates the input rasterized map against current visual observations. It assigns confidence scores to regions classified as reliable, erroneous, or blank, creating a foundational mask that guides two subsequent branches. In one branch, the Dual-Prior Query Generator (DPQG) selectively filters and embeds both rasterized and vectorized priors to initialize informative queries from trustworthy map structures. In the other, the Confidence-Guided Prior Fusion (CGPF) module spatially propagates and refines the mask, using the refined version to govern the fusion of BEV features with the prior map, thereby providing rich contextual representations for the map decoder. Evaluation on the nuScenes dataset shows that PriorMaskMap achieves superior mapping precision over existing approaches and exhibits marked robustness when provided with priors containing apparent errors. Our code is released at https://github.com/healenrens/PriorMaskMap.

0.46

Dynamic Skinning: Kinematics-Driven Cartoon Effects for Articulated Characters

cs.GR ·2026-10-29 ·Damien Rohmer, Karim Salem, Niranjan Kalyanasundaram, Victor Zordan
publishedcode/data
Abstract · 10 pages, 9 figures. Author's accepted version of the article published in IEEE Transactions on Visualization and Computer Graphics (TVCG), …

We present an extension to traditional rig skinning, like Linear Blend Skinning (LBS), to produce secondary motions that exhibit the appearance of physical phenomena without need for simulation. At the core of the technique, we call dynamic skinning, is a set of deformers which offset position of individual vertices as a function of position derivatives and time. Examples of such deformers create effects such as oscillation in response to movement and the appearance of wave propagation, among others. Because the technique computes offsets directly and does not solve physics equations, it is extremely fast to compute. It also boasts a high degree of customizability which supports a desirable artist workflow and fine level of control. Finally, we showcase the technique in a number of scenarios and make comparisons with the state of the art.

0.46

NIF: Normal-Intersection Fields for Geometrically Consistent Oriented Box Refinement

cs.CV ·2026-10-29 ·Shaokai Zheng
code/data
Abstract

Accurate oriented detection requires boxes to follow object boundaries, yet small errors in the coupled center, size, and angle estimates can displace long boundaries even when an object is correctly classified. We propose NIF, a plug-in refinement branch that represents box geometry with normal-intersection fields. Given a coarse oriented box, it samples local features at 20 boundary anchors and predicts signed distances along their outward normals. These observations have an exact affine representation in five rectangle parameters; a differentiable projection uses error-aware reliability weights to recover a geometrically consistent box. Tangential pooling preserves the normal-position profile, while the shared projection adds 0.132M parameters. On DOTA-v1.0, NIF obtains 77.98%, 79.03%, and 79.32% mAP with Oriented R-CNN, Strip R-CNN, and LSKNet-S, respectively; LSKNet-S gives the strongest single-scale result among the compared methods. NIF also reaches 68.19% mAP on DIOR-R and improves matched FAIR1M-v1.0 and DIOR-R baselines by 0.30% and 0.71%, respectively. On HRSC2016, it raises mean IoU from 0.867 to 0.888 while reducing field error by 24.5%. Finally, with four different detector frameworks frozen, NIF improves relative AP75 by 1.55-4.16% and COCO-style mAP by 0.37-3.76%, without detector retraining. Upon acceptance of this paper, we will publicly release the code at https://github.com/zsk466991125-collab/Normal-Intersection-Fields.

0.45

YOLOF-MLD: A Multi-Level Distillation Framework for Efficient Object Detection

cs.CV ·2026-10-29 ·Xiaolong Wu, Liming Zheng, Zitong Liu, Yue Zheng
code/datahot direction
Abstract · 4 pages

Efficient object detection on resource-constrained edge devices is essential for safety-critical applications such as fire-smoke monitoring. However, current lightweight frameworks, such as YOLO26, still rely on the parameter-dense multi-branch Feature Pyramid Network (FPN), preventing full exploitation of their lightweight advantage. Replacing FPN with YOLOF's single-input-single-output (SiSo) neck is a promising solution but causes substantial accuracy loss. To alleviate this, we propose YOLOF-MLD, a CPU-efficient multi-level distillation (MLD) framework that combines a YOLO26n backbone, a SiSo neck, and MLD. It aligns features at the backbone via sliced-semantics matching, neck with receptive-field matching, and response levels. On D-Fire and Annotated Fire-Smoke (AFS) datasets, YOLOF-MLD reduces the parameter count of YOLO26n by 35.7\% and achieves a 62.3\% speedup on Raspberry Pi 5, attaining 71.6% and 59.9% mAP@50, respectively, compared with 75.5% and 60.0% for YOLO26n. These results demonstrate a competitive accuracy--latency trade-off for edge deployment. Our code is available at https://github.com/XiaolongWU111/YOLOF---MLD/blob/main/README.md.

0.45

XLaw-JP-27B: A Lightweight Open Model Specialized for Japanese Law

cs.CL ·2026-10-29 ·Andrew Shin
code/data
Abstract · https://x-law.jp/

We present XLaw-JP-27B, an open-weight model specialized for Japanese law that runs on a single 48\,GB GPU under the Apache~2.0 licence. It adapts Qwen3.8-27B with hybrid retrieval over all current statutes and 31k case summaries, a light QLoRA fine-tune on its own verified answers to past bar-exam questions (self-distillation), a contract-specific LoRA adapter, and four-sample voting. We evaluate it on eight Japanese legal tasks---the multiple-choice section of the bar examination in 2024, 2025 and 2026, statement-level bar questions (JBE-QA), statute questions (LawQA-JP), the patent attorney exam, and contract review and drafting built from government model contracts---against open models of 12--33B under identical prompts and retrieval. XLaw-JP-27B ranks first on six tasks, ties on one and is third on one (mean rank 1.3 against 2.8--3.1 for the baselines). It is the only system that passes the bar exam in all three years (129, 120 and 130 of 175, meeting every subject minimum; no test-year material was used in training), while open Japanese 32--33B models fail at least one year. A controlled study of each component shows that retrieval, voting and self-distillation help, and that thinking mode, retrieval-grounded synthetic tuning, self-verification, per-statement decomposition and more self-distillation do not. The largest remaining gaps to frontier models are in contract review and patent law. The model is available as a free web service at https://x-law.jp/ with the weights available at https://huggingface.co/X-LawJapan/xlaw-jp-27b under Apache~2.0.

0.44

AquaProbe: Efficient Adaptation of Frozen Vision Models for Underwater Images

cs.CV cs.LGcs.RO ·2026-10-29 ·Md Shihabul Islam Shovo, Nishi Kanta Paul, Trung Dong, Xiaomin Lin
top venue
Abstract · 10 pages, 4 figures, 3 tables

Reliable visual perception is essential for safe autonomous underwater exploration. Yet vision models pretrained on terrestrial images become less reliable underwater, where turbidity, backscatter, and wavelength-dependent attenuation alter scene appearance. Existing adaptation methods either spread adapters equally across the encoder depth or select layers using the divergence between two fixed domains. In this paper, we introduce AquaProbe for measuring how the location of the domain gap changes with water type and imaging range. AquaProbe separates newly introduced divergence from that accumulated through a frozen vision encoder using paired clean and physically rendered images and allocates a fixed LoRA budget according to the measured profile. The segmentation mean Intersection-over-Union (mIoU) for the proportional LoRA rank allocation, based on the measured profile, is 70.1; this is a better result than uniform LoRA rank allocation by 1.9 mIoU points while having the same parameter budget of 0.59 million adapter parameters. The three-way router, conditioned on estimated condition clusters, achieves 70.6 mIoU on the standard SUIM split and 63.4 AUDRC under held-out degradation.

0.44

Parametric Gaussian Noise Mixer (PGNM): Unified Synthesis of Camera Noise and Film Grain

cs.CV ·2026-10-29 ·Mingyang Song, Martín Rodríguez Muñoz
code/data
Abstract · Technical report, not submitted to any conference. 49 pages (9-page main paper, 40-page supplementary material), 29 figures, 33 tables. Code…

Noise synthesis is commonly approached through physical models of image formation or neural networks trained to generate noisy images. We explore a middle ground: an explicit statistical model fitted directly from paired clean and noisy images. A normalized convolution correlates a Gaussian field across pixels and color channels, and a lookup maps each Gaussian value to a residual given the clean intensity. Because the field stays standard Gaussian, the two components are fitted separately, without reproducing the physical process or training a neural generator. With one formulation for camera noise from SIDD and simulated film grain from Newson, the model has the lowest average distribution and correlation errors on SIDD among the evaluated methods and remains competitive on film grain. A denoiser trained on its noise comes closest to one trained on real noise, and a 1080p frame renders in 10 ms on GPU.

0.44

Scaling Mathematical Proof Supervision from Structured Textbooks

cs.CL cs.AI ·2026-10-29 ·Youqing Fang, Zhouqi Hua, Yinhao Tang, Yuzhe Gu, Lingkai Kong +5 more
community pick
Abstract · 35 pages, 18 figures, 14 tables

Mathematical supervision for language models is typically flattened into isolated problem-solution pairs, discarding the curricular organization and evidential structure of the original sources. We introduce TOME-Math (Textbook-Oriented Mathematical Evidence), a large-scale corpus that converts 2,118 textbooks spanning the core areas of university-level mathematics subjects into approximately 380K structured proof-supervision records while preserving chapter order, proposition-proof links, dependency references, and source provenance. Leveraging these structural signals, we explore curriculum learning across multiple mathematical subjects and characterize scaling patterns with supervision volume and source breadth. At full scale, fine-tuning on TOME-Math raises the unweighted mean score across our held-out evaluations and established public benchmarks by 16.7 percentage points over the base model, while exceeding all three public instruction-data baselines evaluated in aggregate. Our findings highlight structured textbooks as both a scalable source of mathematical supervision and an empirical framework for studying the impact of data organization on mathematical reasoning. We will release the corpus to facilitate further research on structured mathematical supervision.

0.44

PALATE: Pretrained Autoregressive LATent Embeddings

cs.IR cs.LG ·2026-10-29 ·Sizhang Zhao, Praveen Gondhi, Bin Li, Utsaw Kumar
publishedhot direction
Abstract · Accepted to GRAIL @ CIKM 2026. 5 pages, 2 figures, 3 tables

Local-commerce and food-delivery marketplaces run large ads businesses: across hundreds of thousands of stores, merchants pay to be surfaced on browse and search surfaces, and a ranker decides which ads each consumer sees. Ranking well demands strong per-consumer personalization, yet a consumer's raw interaction history is sparse and high-cardinality. Our production ranker had personalized only through hand-engineered signals: aggregate engagement counts, categorical embeddings, and order-agnostic pooled history, with no learned per-consumer sequence representation. We learn one by training a GPT-style generative model of the consumer's engagement sequence to predict their next few engagements, and read its last-token hidden state as the consumer embedding. That single frozen vector becomes one shared personalization input for every ad surface: a daily job refreshes it, and each ranker consumes it as an ordinary dense feature. Offline it is the largest single-feature gain in our ranking-AUC ladder, a relative Eval-AUC gain of +0.59% on a browse-driven feed and +0.57% on a query-driven search surface. Online, it shipped as one component of bundled ranker retrains on both surfaces: Discovery (revenue +0.79%, conversion rate +1.17%) and Search (revenue +0.94%, conversion rate +1.50%). Three design choices adapt the generative objective to a repurchase-heavy marketplace, and are the core of our contribution: an exploration-weighted loss that up-weights businesses new to the consumer over ones already engaged; a two-stage curriculum: a click pretrain that keeps the representation broad and exploratory, then a conversion fine-tune that sharpens purchase intent; and a designed negative set that pairs a fixed popular-business pool and businesses shown but not ordered with random and in-batch negatives.

0.44

Inter-Frame Predictive Contour Coding for Semantic Map Compression

eess.IV cs.CV ·2026-10-29 ·Runyu Yang, Hyomin Choi, Fabien Racapé, Ivan V. Bajić
publishedcode/data
Abstract · 4 pages, 4 figures, to be presented at the 2026 IEEE International Conference on Visual Communications and Image Processing, Singapore, Dece…

Dynamic semantic maps have become an important intermediate representation for many computer vision applications. Efficient lossless compression of such maps is essential for reducing storage and transmission costs. Existing contour-based semantic map compression methods mainly exploit spatial redundancy within individual frames while largely ignoring the strong temporal correlation between adjacent semantic maps. In this paper, we propose an inter-frame predictive contour coding framework for lossless semantic map compression. The proposed framework first predicts blob positions and label indexes from the reference frame and then performs contour prediction using corresponding chain-code units in the reference contours. A mode selection mechanism is further employed to choose between intra and inter coding according to the estimated coding cost. By exploiting temporal contour redundancy while preserving the advantages of contour-based representation, the proposed framework significantly improves coding efficiency. Experimental results on the VSPW dynamic semantic segmentation dataset demonstrate that the proposed method consistently outperforms state-of-the-art lossless semantic map compression methods. Compared with the recent context-adaptive extended chain coding framework, the proposed inter-frame coding achieves an average bitrate reduction of 3.0\% without significant increase in the average encoding time, demonstrating an effective trade-off between compression performance and computational complexity. The source code is made publicly available at https://github.com/InterDigitalInc/LosslessSegmentationMapCompression.

0.43

Beyond Ground Truth: A Heuristic Unsupervised Quality Assurance Engine for Diagnosing Calibrational Blindness in Dental Segmentation

cs.CV ·2026-10-29 ·Fatematuj Johura Mim
top venue
Abstract · Accepted at 11th IEEE Asia-Pacific Conference on Computer Science and Data Engineering (IEEE CSDE 2026), Cox's Bazar, Bangladesh. Copyright …

In real-world clinical settings, dental X-ray segmentation using deep learning methods suffers from reliability issues. Even when the segmentation quality is bad, deep learning models show high confidence in their predictions (calibrational blindness). Moreover, in real use, doctors do not have ground-truth masks to check their accuracy. To address this problem, we introduce a simple quality-checking engine called HUS-QA with three stages. We train it on the PRAD (Periapical Dental X-ray) dataset to build our SimCLR self-supervised ResNet-18 U-Net model, which generates the segmentation mask and then evaluates its quality. Our HUS-QA checks the segmentation mask quality without a ground-truth mask. Prediction confidence, edge sharpness, shape connectivity, and area-size penalty are the simple indicators considered in this research. Results show that standard confidence methods fail (r = -0.1394, p = 1.7864e-3, ROC AUC = 0.4275). By contrast, our HUS-QA shows a positive correlation with actual Dice scores (r = 0.4647, p = 3.7510e-28) and achieves a strong ROC-AUC of 0.8405. Therefore, HUS-QA serves as a practical tool for clinical quality control.

0.43

Do Open Language Models Invent Scripture? BILAL-BENCH: Benchmarking Islamic-Source Fabrication in Bengali

cs.CL ·2026-10-29 ·Sabah Nushra, Md Kishor Morol
publishedbenchmarkcode/data
Abstract · 15 pages, 6 tables. Accepted at the 7th Muslims in ML (MusIML) Workshop at NeurIPS 2026. Code and dataset: https://github.com/Meowthhh/BILAL…

When a language model invents a Qur'anic verse or hadith, it gives false authority to text that readers may treat as sacred. Existing Islamic hallucination benchmarks are largely Arabic and often closed-ended, while Bengali resources mainly study general-domain hallucination or factual knowledge. We introduce BILAL-BENCH, a 240-question expert-verified benchmark for open-ended Islamic-source fabrication in Bengali. It covers Qur'anic attribution, hadith, fiqh, contemporary issues, and adversarial false-premise prompts, and uses a two-axis rubric that separates source errors from assertiveness. We evaluate seven open-weight models from 1B to 120B on 1,680 double-annotated responses. Six of the seven models confidently fabricate on 79.3-95.3% of source-production questions, with no monotonic improvement from scale. Hadith provenance is the weakest category across models: only two of 315 responses are correct and correctly sourced. The two 8B BanglaLLaMA variants have similar confident-fabrication rates but different error profiles. Models vary far more in whether they withhold a claim they cannot support than in source accuracy: appropriate abstention or false-premise correction ranges from 6 to 46 responses across the benchmark, and does not track parameter count.

0.43

AI Cosmos: An Agentic Infra for Situated Learning

cs.CL ·2026-10-29 ·AI Cosmos Team, Jinfeng Zhou, Yuxuan Chen, Fangming Chang, Mingyu Su +30 more
code/data
Abstract · https://github.com/THU-AICosmos

Large language models (LLMs) can improve task performance during learning, yet these gains do not necessarily reflect learners' ability to thereafter apply the knowledge and skills independently. We present AI Cosmos, an agentic infrastructure for situated learning that organizes learning around two complementary principles: scenario-based participation and cumulative assets of practice. AI Cosmos organizes learners and role-specialized agents within learning scenarios that integrate task-relevant knowledge and, when needed, executable workbenches. An asset-memory loop preserves the processes and outcomes of practice as reusable assets with explicit provenance and dependencies. We evaluate scenario-based interaction through two preregistered randomized controlled studies examining the acquisition of legal reasoning and programming competence, holding the same underlying LLM across conditions. AI Cosmos achieves greater learning gains than a conventional LLM in legal reasoning and higher mastery of unseen programming problems than an LLM-as-Teacher baseline.

0.42

Greenscapes: Optimized Vegetation Configurations in Cities

cs.GR cs.CV ·2026-10-29 ·Patryk Bartkowiak, Marcin Marlewski, Paweł Wójcik, Jae Joong Lee, Dominik L. Michels +3 more
code/data
Abstract · 18 pages (11 main paper + 7 supplementary material), 15 figures, 5 tables. Project page: https://greenscapes-project.github.io/ Code: https:…

Vegetation plays a central role in cities, serving functional roles such as balancing shade and sky exposure, shaping the microclimate, carbon sequestration, and providing acoustic screening. Furthermore, aesthetic choices in landscape styles (formal, modernist, picturesque) are important for citizens' well-being and should be considered in the design of urban landscapes. In this paper, we present a novel method for placing vegetation, ranging from shrubs to large trees, in 3D urban scenes. Our method uses individual lots as input and employs a conditioned Vision-Language Model (VLM) to generate a top-down rendering of vegetation, placing large trees, small trees, and shrubs on lawn areas. This rendering is then converted into a 3D scene by detecting individual plants and selecting appropriate 3D assets. To refine the proposed configurations, we define an optimization loop that keeps the VLM-generated positions fixed after a deterministic feasibility repair and uses CMA-ES to optimize per-instance scale transformations along with discrete vegetation asset substitutions. To define the objective function for the optimization, we evaluate the generated 3D scenes with efficient surrogate models of shade and path-based acoustic exposure. Furthermore, we render in-scene views at each optimization step to compute view-dependent perceptual descriptors, enabling the generation of both aesthetic and functional vegetation layouts for urban scenes.

0.42

DINOv2 for PET: A Pre-Training Domain Benchmark Across Classification, Segmentation, and Prognosis

cs.CV ·2026-10-29 ·Roxanne De Maeyer, Yannick Brackenier, Vicky De Ridder, Simon DeKeyser, Maarten Larmuseau +1 more
publishedbenchmarkcode/data
Abstract · 8 pages + references, MICCAI 2026, CaPTion workshop

DINO (self-DIstillation with NO labels) has become a stan- dard for self-supervised pre-training, producing features that generalise across datasets and downstream tasks. In medical imaging it has been adopted mainly for radiology (X-ray, CT and MRI), where even natural- image DINO features transfer well. In the literature, these features have been reported to transfer worse to functional imaging such as PET. Yet PET-specific DINO models have barely been introduced, and even less validated across multiple tasks and datasets. We present PET-DINOv2-2D, the first DINOv2 model pre-trained on 2D PET, and benchmark it against open-source encoders: the origi- nal DINOv2 (pre-trained on the LVD-142M natural-image dataset) and OmniRad (pre-trained on CT, MRI and ultrasound). Using the same lightweight heads for every backbone, and despite far less pre-training data, PET-DINOv2-2D matches or outperforms both on three down- stream PET tasks: normal/abnormal classification, lesion segmentation and recurrence-free survival prognosis. Its advantage is largest when few labels are available, suggesting that PET pre-training gives more transferable, label-efficient features. Code is made available at https://github.com/nuclivision/pet-dinov2-2d/.

0.42

Ambisonics: Introducing an Accessible, Distributed Approach

cs.SD eess.AS ·2026-10-29 ·Thomas A Rushton, Romain Michon, Tanguy Risset, Clemens Wegener, Pierre Lecomte
publishedunusual
Abstract

This paper presents an accessible, distributed approach to higher order ambisonics using a network of low-cost microcontroller-based devices. Leveraging recent advances in embedded computing, networked audio, and precise time synchronization, the system offers an alternative to conventional centralized architectures that typically require expensive multichannel interfaces and high-performance digital signal processing hardware. The proposed architecture combines an audio plugin for soundfield design with a modular network of ambisonics processing units, which, exploiting the parallelisable nature of ambisonics decoding, are each responsible for a subset of loudspeaker outputs. Synchronization is achieved through the IEEE 1588 Precision Time Protocol, complemented by adaptive clock correction to maintain sub-microsecond alignment between devices. A third-order ambisonics implementation demonstrates the viability of the approach, with potential scalability to higher orders and larger arrays. Overall, this work contributes to the development of frugal, distributed solutions for immersive audio, expanding access to ambisonics-based systems for a wide range of applications.

0.41

Mind The Gap (MTG-AD): Dataset and Fine-grained Evaluation for End-to-End Inline Audio Descriptions

cs.CV cs.AIcs.HC ·2026-10-29 ·Subhashini Venugopalan, Jenny Tan, Taylor Roper, Jimmy Tobin, Anton Kast +3 more
top venuebenchmarkdatasetcode/data
Abstract · NeurIPS 2026 Evaluations & Datasets Track, camera ready

Audio descriptions (AD) are essential for making visual media accessible to blind and low-vision (BLV) users. While Multimodal Large Language Models (MLLMs) offer a scalable solution for generating audio descriptions on-demand, their performance on ``inline'' audio descriptions -- which must fit within existing silences in a video -- remains under-explored, particularly for diverse user-generated content (UGC). We present a comprehensive investigation into MLLM-generated inline audio descriptions. First, we introduce MTG-AD, a professionally annotated dataset of 36 hours of videos, each averaging $\sim$5 mins. and 8k+ descriptions in total, co-created by paired blind and sighted AD professionals to provide a high-quality benchmark for this task. Second, we propose MTG-AD-Eval, an evaluation framework based on audio description expert guidelines that provides fine-grained, reference-free actionable metrics for model improvement. Crucially, our framework evaluates the end-to-end task: it renders the generated scripts into audio tracks to assess their fit within the original video's natural silences, minimizing disruptive overlaps. Our analysis across frontier MLLMs reveals that while models accurately describe visual events and substantially reduce speech overlaps, gaps persist in narrative pacing and flow. Finally, we empirically validate that our automated and reference-free metrics correlate significantly with human quality ratings from sighted raters and 7 BLV participants, finding that while users narrowly prefer human-authored audio descriptions, they find MLLM-generated audio descriptions highly beneficial for comprehension. We provide actionable system design guidance and release MTG-AD at https://hf.co/datasets/google/mind-the-gap-inline-ad and MTG-AD-Eval at https://github.com/google/mtg-ad.

0.40

BRIDGE: Learning Brain-Gut Metabolic Evidence for Alzheimer's Disease Diagnosis

cs.CV ·2026-10-29 ·Pengli Zhu, Ruxi Chang, Tianliang Yao, Yu Jiang, Shengyuan Liu +5 more
code/data
Abstract · 10 pages, 6 figures

Alzheimer's disease (AD) diagnosis may benefit from gut metabolic information associated with brain-gut interactions, yet existing models largely analyze cerebral glucose metabolism alone. Although whole-body fluorodeoxyglucose positron emission tomography (FDG-PET) can provide paired brain-gut observations, routine brain examinations rarely include gut imaging, limiting direct use of this information for diagnosis. In this paper, we present BRIDGE (Brain and Gut Representation Integration via Distribution-Guided Enhancement), a two-stage framework for transferring brain-gut metabolic knowledge to brain-only diagnosis. Specifically, BRIDGE integrates a conditional brain-to-gut latent transport module and a disease-evidence-guided cerebral pathway to capture subject-specific cross-organ correspondence and disease-relevant cerebral representations. For the cerebral pathway, a large language model (LLM) helps organize literature-derived descriptors into human-approved semantic anchors that guide feature learning. Additionally, we introduce a counterfactual-guided residual fusion mechanism, which uses matched and non-matched virtual-gut evidence to supervise a confidence gate and dynamically regulate the contribution of transported cross-organ information. At inference, BRIDGE exploits complementary gut-related context using brain PET alone, without measured gut PET. Extensive experiments across public and private cohorts demonstrate that BRIDGE outperforms brain-only and cross-organ baselines, achieving AUC gains of 6.0% and 5.2% over its brain-only counterpart for CN/MCI and CN/AD diagnosis, respectively. Notably, its ability to transfer complementary brain-gut metabolic context while preserving brain PET as the sole inference input highlights its potential for clinically practical diagnosis. Our code is available at https://github.com/CUHK-AIM-Group/BRIDGE.

0.40

VoxLock: Target-Speaker Extraction for Voice Agents Without an Enrollment Step

eess.AS ·2026-10-29 ·Arbaz Siddiqui
benchmarkcode/data
Abstract · 4 pages, 4 tables. Model, code and per-file scores: https://github.com/arbazsiddiqui/voxlock

A voice agent's speech recognizer transcribes everyone near the microphone, including people other than the caller. I fine-tuned a public causal target-speaker extraction model (REAL-TSE, 27M parameters) on one GPU to work from a voice sample cut from the call itself, and released it as VoxLock. I score it by what a recognizer writes on real recordings. On Krisp's 265-recording benchmark it reaches 4.95% WER against 23.21% unprocessed and lets 9 background words into the transcript over 85 minutes of background-only speech. Its WER is not significantly different from Krisp's best filters and is significantly lower than the other commercial and open systems I tested. On 400 AMI meeting windows built and frozen before scoring, it cuts WER from 42.95% to 10.63%. The 8-bit model runs a live call on 77% of one CPU thread with 24 ms of algorithmic latency. Model, code and per-file scores are open.

0.40

STRATA: Reinforcement-driven Active Sampling with Tri-paradigm Adaptation for Efficient 3D Spatial Transcriptomics

cs.CV ·2026-10-29 ·Junchao Zhu, Ruining Deng, Chongyu Qu, Zhengyi Lu, Yanfan Zhu +5 more
code/data
Abstract

Spatial transcriptomics (ST) enables joint characterization of tissue morphology and gene expression. However, 3D volumetric ST acquisition remains expensive. AI-driven methods offer a low-cost alternative but remain limited in accuracy, while recent 2D section-wise ST acquisition still rely on dense grid sampling, resulting in redundant measurements and undersampling of biologically informative regions. To bridge the gap, we propose STRATA, a reinforcement-driven active sampling framework with tri-paradigm adaptation for reconstructing dense 3D transcriptomic maps from sparse voxel-wise ST acquisition. STRATA constructs a 3D graph integrating H&E morphology with single-cell priors and employs a reinforcement learning agent to estimate morphological sparsity and transcriptomic informativeness, enabling budget-aware selection of informative spots. Given these sparse measurements, an adaptive tri-paradigm predictor combines retrieval, regression, and cross-layer imputation through lightweight fusion, dynamically weighting their contributions. Experiments across datasets demonstrate improved sampling efficiency and accuracy, particularly under low-budget settings. Code is available at https://github.com/Junchao-Zhu/STRATA.

0.40

When the Top Answer Decides: Gender Bias in Decision Models

cs.CL cs.CY ·2026-10-29 ·Yeb Havinga
code/data
Abstract · 19 pages, 6 figures. Code, stimuli and results: https://github.com/yhavinga/decision-model-gender-bias

Decision models are language models built to route, classify and triage: they return option probabilities from one forward pass, and an application acts on the top answer. They are offered off the shelf, trained for the decision format rather than for fairness or any one application, and we found no bias measurement in their cards. We test TypeSafe's closed Jev and five open decision models on gender-swapped yes/no decisions about people (Tamkin et al., 2023), BBQ, StereoSet and the gender-bias studies of Fulgu and Capraro (2024). Five of the six give women significantly more favourable answers than men (+0.05 to +0.10 in log-odds), and a gender swap changes 1-3.5% of top answers (Jev's partly through its call-to-call variation). The makers' recommended confidence gate removes many biased commitments but raises the stereotyped share of those it keeps in ambiguous BBQ gender questions (granite-switch 0.76 to 0.85, pplx 0.81 to 1.00). Probes show gender preferences in the same models: all six rate violence for a greater good as more acceptable against a man than against a woman, and two default to "girl" when forced to guess a child's gender without a cue. Decision-level and probability-level measures differ, so audits should read a model as its application does.

0.39

Elucidating the Design Space of LM Data Attribution

cs.CL ·2026-10-29 ·Steven Yu, Sheng-Yu Wang, Kimia Hamidieh, Antonio Torralba, David Bau
benchmarkcode/data
Abstract · Project Page: https://catt.baulab.info/

Data attribution methods for Language Models (LMs) are typically evaluated using downstream tasks like fact retrieval. However, existing evaluations vary multiple dataset properties simultaneously, making it impossible to isolate exactly why an attribution algorithm fails. To diagnose these failures, we explore the design space of LLM data attribution, with a primary focus on evaluation. We introduce the Controlled Attribution Benchmark (CATT) for fact tracing, in which we fine-tune LLMs on entirely fictitious knowledge, providing unambiguous ground truth for attribution. This synthetic setup allows us to systematically alter one dataset configuration at a time, such as document redundancy and lexical distractors. With this configurable evaluation, we demonstrate that redundancy and distractors consistently degrade attribution performance. Rather than proposing a new estimator, we examine two design choices that are underexplored: the loss used to compute gradients and how per-token attribution scores are aggregated into a document score. We find that the true causal evidence within a training document is highly localized, but standard methods dilute this evidence into noise by averaging attribution scores across the entire document. We find that replacing standard cross-entropy loss gradients with the negative log-odds gradients surfaces the critical tokens from other uninformative ones. We can then aggregate only the top token scores to improve ground truth recall by 35 to 72 percentage points across multiple baselines on our diagnostic benchmark. The negative log-odds correction also generalizes to FTRACE-TREX, improving counterfactual retraining performance for every method we test, by up to 4.9x.

0.39

Retrieve-then-Disambiguate: Scalable Reviewer Discovery

cs.IR cs.DL ·2026-10-29 ·Vicente Amado Olivo, Joe Petrecca, Wolfgang E. Kerzendorf
code/data
Abstract

The growing number of scientific submissions is straining the peer review process, while a large pool of relevant researchers remains unidentified and is therefore not considered as potential reviewers. We formulate the discovery of these researchers as a retrieval problem: given a submission, we retrieve and rank researchers based on domain similarity. We present retrieve-then-disambiguate, an open-source algorithm that identifies domain relevant researchers for a query submission by retrieving similar publications from the scientific literature and resolving ambiguous author names into unique researcher identities. We evaluate the algorithm on 500 astronomy and physics publications. We use ORCID-linked first authors, coauthors, and cited researchers as proxy labels to evaluate the recovery of researchers with known connections to each query publication. When retrieving the top 500 most similar papers, the algorithm recovers roughly 68\% of first authors, 60\% of coauthors, and 45\% of cited researchers. Our code is available at https://github.com/deepthought-initiative/Retrieve-then-Disambiguate.

0.39

Grounding Image Composition with Label-Conditioned Visual Evidence

cs.CV ·2026-10-29 ·Armin Dadras, Tingyu Lin, Robert Sablatnig, Franziska Proksa, Markus Seidl
publishedbenchmark
Abstract · Accepted at WACV 2027

Image composition labels such as center, diagonal, or rule of thirds describe visual structure rather than object categories. KU-PCP provides image-level labels and class-specific geometric elements, but does not show which visible regions people associate with each of its nine classes. We introduce KU-PCP Elements (KU-PCPE), a label-conditioned region-evidence benchmark for photographic composition. Given an image and a label already judged applicable, the task is to segment the visible regions that explain why the label applies. KU-PCPE contains 4,093 single-annotator training masks and a 1,327-task expert test split with three masks per image-label task. The annotations also include exploratory segment-versus-line ratings that describe whether region or semantic-guideline evidence better explains the label. A held-out expert reaches .867 Dice against the other experts, while the strongest deployable model reaches .664, a proposal oracle .672, and the best vision-language model (VLM) proposal selector .395. KU-PCPE exposes the gap between recognizing a composition label and grounding the human-aligned visual evidence that makes it plausible in a particular image.

0.39

ProdCrawl: A Replayable Test Collection for Focused Crawling

cs.IR ·2026-10-29 ·Manuel Rey-Escobar, David Otero, Javier Parapar
code/datahot direction
Abstract

A focused crawler searches for web pages on a specific topic under a constrained download budget. Since it can only fetch a limited number of pages, its success depends on choosing which links to follow. Focused crawlers are usually evaluated on live websites by harvest rate, the fraction of downloaded pages that are relevant. This has two problems: on the one hand, websites change over time, so a crawl cannot be repeated under the same conditions; on the other hand, there is no ground truth for which pages are relevant. As a result, comparisons between crawlers are unreliable. We present ProdCrawl, a replayable test collection for focused crawling. It comprises static mirrors of 621 company and organization websites, about 1.1 million pages, with each page labeled as a product page or non-product page based on a majority vote among three large language models. Any crawler can be run on exactly the same sites and scored against the same labels. We use ProdCrawl to compare simple baselines (breadth-first, depth-first and text classifiers) with an approach that scores each link by searching for similar link texts whose target pages have already been labeled. Our results show that text classifiers and keyword matching do not significantly outperform breadth-first crawling, whereas neural embeddings, which compare link texts by semantic meaning, outperform all simple strategies. Because every run evaluates over identical pages and labels, these results are exactly reproducible. The collection, codebase, and models are publicly released to support experimental replication and fair crawler comparison.

0.39

Efficient LLM-based Subgraph Retrieval for Multi-Hop Knowledge Base Question Answering

cs.IR ·2026-10-29 ·Duanyang Yuan, Sihang Zhou, Xiaoshu Chen, Haoyuan Chen, Ke Liang +1 more
top venuebenchmark
Abstract

Multi-hop Knowledge Base Question Answering (KBQA) aims to find answer entities in the knowledge base that are multiple hops away from the entities in the question. Information retrieval-based (IR-based) methods extract a pivotal subgraph from the entire KB to locate candidate answers and then evaluate their plausibility through semantic matching with the question. However, we observed that the extracted subgraphs often include nodes that are weakly related or irrelevant to the question. Without a proper node filtering mechanism, the number of irrelevant nodes grows as the number of hops increases, leading to excessive consumption of computational resources. To address these challenges, this study introduces an efficient LLM-based subgraph retrieval method for multi-hop knowledge base question answering, M-ER. The framework leverages Monte Carlo Tree Search (MCTS) to transform subgraph exploration into a tree-structured search process. During the MCTS selection phase, nodes that are highly relevant to the question are prioritized for inclusion in the subgraph, eliminating the need to traverse all nodes in the KB. The framework further incorporates a large language model (LLM) to refine the search direction, ensuring that exploration remains focused on nodes relevant to the question. In addition, selected nodes are quantitatively scored, and these scores are fed back into the node selection process to effectively filter out irrelevant candidates, thereby improving the quality of the subgraph. This mechanism not only narrows the search space but also enhances the overall efficiency of multi-hop KBQA. Experiments on the WebQSP benchmark demonstrate that M-ER achieves 78.88\% on the Hits@1 metric, while also improving computational efficiency. These results not only validate the effectiveness of M-ER, but also offer a viable technical path to balance performance and computational efficiency.

0.38

PhysInsert: Physically Plausible Object Insertion with Geometric Priors and Bidirectional Object-Scene Interactions

cs.CV ·2026-10-29 ·Yue Gong, Le Zhang, Rui Chen, Song Chang, Xiaochao Qu +3 more
datasetcode/data
Abstract · 51 pages, including supplementary material. Project page: https://arthuring.github.io/PhysInsert/

We present PhysInsert, a framework for physically plausible reference-based object insertion that jointly adapts objects and their surrounding scenes. Existing methods often rely on masked inputs, generation-based pairs, or training pairs with limited viewpoint variation, leaving pose and illumination adaptation and bidirectional object-scene interactions insufficiently constrained. We construct the Insert World dataset with strictly aligned Blender-rendered counterfactual pairs and object references across viewpoints and lighting conditions. Building on these data, Geometry-Guided Adaptation couples a VGGT geometry stream with image and text streams under target-scene feature supervision to guide spatially consistent insertion. Complementary 2D Object-Scene Spatial Supervision jointly supervises object occupancy and environmental changes while grounding spatial instructions in object regions through image-text attention. We introduce PhysInsert-Bench with dedicated cross-environment and cross-view cases requiring illumination and pose adaptation beyond appearance copying. Experiments on PhysInsert-Bench and DreamBooth demonstrate improved image fidelity; a user study further shows more accurate placement and improved lighting integration, perspective, and pose plausibility.

0.38

EPC-RawDC: Equivariant Physics Constrained 3D Diffusion with Raw Data Consistency for CBCT to CT Image Synthesis

cs.CV cs.AIphysics.med-ph ·2026-10-29 ·Alzahra Altalib, Chunhui Li, Alessandro Perelli
code/datahot direction
Abstract · 10 pages, 4 figures, 8 tables

Cone beam computed tomography (CBCT) is widely used for image guided and adaptive radiotherapy, but scatter, noise, reconstruction artifacts, and inaccurate Hounsfield units limit its quantitative use. Recent diffusion approaches for CBCT to CT synthesis remain largely image driven, with limited use of the underlying acquisition physics. We propose EPC-RawDC, a conditional 3D latent diffusion framework that combines volumetric generative modeling with two complementary constraints derived from the same scanner specific CBCT forward model. First, geometry equivariant projection constraint exploits the rotational equivariance of the CBCT forward operator by supervising the synthesized CT through correspondingly transformed reference projections. Second, raw data consistency uses noisy projection measurements of the individual acquisition as scan specific supervision during training. The framework was evaluated on two independent clinical cohorts with distinct scanner geometries and projection sources. For JUST, the physics constraints used measured clinical raw CBCT projections acquired with an Elekta system, whereas for SynthRAD2025 the projection data were simulated from the reference CT using a Varian geometry with Poisson photon statistics. EPC-RawDC achieved an SSIM of 0.97 in both cohorts and improved PSNR over DDPM by 2.66 dB on SynthRAD2025 and 2.10 dB on JUST, while reducing mean MAE by 20.3% and 19.0%, respectively. On SynthRAD2025, projection consistency and equivariant errors were reduced by 55.0% and 41.4% relative to DDPM. These results demonstrate that combining a volumetric diffusion prior with scanner geometry and scan specific projection measurements improves quantitative CBCT to CT synthesis across both simulated and measured projection domains. Code available after publication at https://github.com/ALZAHRAALTALIB/EPC-RawDC.

0.37

Grammar Control for Language Learning with Sequential Monte Carlo and Small Language Models

cs.CL ·2026-10-29 ·Kate McCurdy, Jacob Hoover Vigly, Thomas Hikaru Clark, Karolina Stańczak, Yahya Emara +4 more
top venue
Abstract · EMNLP 2026

Second language learners benefit from exposure to level-appropriate grammar in interactive language use. While prompted frontier language models can usually produce target grammar structures, less resource-intensive models that can be run on device often fail in this task. We use sequential Monte Carlo (SMC) to steer dialogue turn generation in small lan- language models toward target grammar constructions from an expert-developed pedagogical resource. With a minimal set of hand-coded potential functions, SMC achieves near-perfect coverage of target grammar, enabling small models to outperform state-of-the-art results from models 140 times larger. We also use code synthesis with frontier models to generate new potentials, showing that our method scales to hundreds of target grammar skills with no need for additional training or supervised data.

0.37

HALO: Human-Grounded Granular Evaluation of Proactive Agents

cs.CL ·2026-10-29 ·Wenqi Pei, Xinyu Wang, Jinrui Liu, Cynthia Zhuge, Jie Sheng +3 more
datasetcode/data
Abstract

Proactive agents must decide whether, when, and how far to assist without an explicit request. Existing evaluations often rely on automated scores or binary verdicts, offering limited human diagnosis of evidence use and intervention quality. To address these gaps, we introduce HALO, Human-Grounded GrAnular EvaLuation of PrOactive Agents. HALO-Arena covers 4 proactive need types, Dependency, Blockage, Risk, and Opportunity, across 10 everyday domains and 50 personal environments. Scenarios require agents to retrieve evidence from structured databases or unstructured documents before judging whether assistance is warranted and choosing its scope. On HALO-Arena, we construct HALO-8K, a dataset of 8,000 responses from 10 models across 400 scenarios, annotated by 20 experts for action level, acceptability verdicts, and human diagnosis. For rejected responses, human diagnosis covers timing, intensity, and other issues. Recorded reasoning tokens and retrieval rounds measure the effort behind these responses. The best proprietary and open-source models achieve 72.88% and 56.63% acceptability, respectively. Our central finding is that over-proactivity is the dominant observed failure mode, with Premature timing or Excessive intensity accounting for 61.20% of rejected responses. Further analyses yield complementary findings on the balance between intervention and restraint, evidence acquisition, and the calibration of timing and intensity.

0.36

HairDeck: Animation-Ready Stylized Hair Cards from a Single Image

cs.CV cs.GR ·2026-10-29 ·Yuze He, Yixiang Wei, Wang Zhao, Chuhao Chen, Yanning Zhou +1 more
code/datacommunity pick
Abstract · 27 pages, 23 figures

The hair of stylized characters in games and avatars is built from hair cards: artist-placed ribbons whose geometry is the visible hair, animated by spring bones. Existing image-based methods produce realistic hair or a single fused mesh, while CHARM, the only generator of stylized hair cards, decodes a hairstyle autoregressively as a long sequence of control points. Observing that a card hairstyle is an unordered set of artist-defined primitives, we propose HairDeck, which generates it as one: each card is a single token, and a set diffusion transformer denoises all cards together. From a single image, HairDeck generates the cards and the spring-bone chains that drive them, yielding an animatable asset that exports to ready-to-use downstream formats. On CHARM's own test set, it lowers Chamfer distance by 44% and card-orientation error by more than half while running 33 times faster, and from single images it outperforms CHARM, image-to-3D pipelines and strand-based methods. Project page: https://hyzcluster.github.io/HairDeck

0.36

Zonal-Aware Prostate Cancer Detection and Classification via Hadamard-Fused Mamba-UNet

eess.IV ·2026-10-29 ·Emadeldeen Hamdan, Gorkem Durak, Muhammed Enes Tasci, Halil Ertugrul Aktas, Fergan Bol +12 more
published
Abstract · Published in SPIE Medical Imaging - Computer-Aided Diagnosis Conference

Accurate clinically significant prostate cancer (csPCa) detection from prostate MRI requires a detailed understanding of the prostate anatomical context. Deep learning approaches have achieved promising results in zonal and lesion segmentation; however, the joint analysis of prostate zonal anatomy and lesion morphology in a unified framework remains relatively underexplored. In this work, we propose ZH-Mamba, a multi-stage zonal-aware csPCa detection framework consisting of two H-Mamba UNets models for prostate zonal segmentation and lesion localization, followed by an anatomy-aware Convolutional Neural Network classifier for csPCa detection. The proposed model leverages a learnable Hadamard Transform-based fusion mechanism to enhance zonal and lesion feature representations without increasing computational complexity. The learned fusion patterns consistently emphasize prostate regions, suggesting that the proposed fusion captures anatomically relevant feature representations. We demonstrate a strong zonal segmentation performance on the PI-CAI dataset, reaching Dice scores of $\textbf{0.95}$, $\textbf{0.85}$, and $\textbf{0.95}$ for whole gland (WG), peripheral zone (PZ), and transition zone (TZ), respectively, providing robust anatomical priors for downstream csPCa detection. Experimental results on an independent internal dataset confirm the generalization capability of the proposed zonal segmentation model. For lesion localization, the H-Mamba UNet achieves a competitive Dice score of $\textbf{0.70}$. Integrating the zonal masks and lesion localization, the proposed ZH-Mamba framework achieves an AUC of $\textbf{0.86}$, Average Precision ($\mathrm{AP}_{\mathrm{Patient}}$) of $\textbf{0.72}$, and F1 score of $\textbf{0.76}$, demonstrating the effectiveness of zonal-aware csPCa detection.

0.36

SymKL: A Unified Representation-Space Measure of Signal Degradation

cs.CV cs.AI ·2026-10-29 ·Dhananjaya Jayasundara, Sudarshan Rajagopalan, Vishal M. Patel
code/dataunusual
Abstract

Quantifying degradation between a pristine signal and its degraded observation is fundamental to signal processing, yet existing full-reference measures are inherently modality-specific. We introduce SymKL, a unified full-reference degradation measure based on implicit neural representations (INRs). An INR is fitted once to the pristine reference and characterized by its diagonal Fisher information, which captures the local information geometry of the learned representation. Evaluating the Fisher information with the pristine reference and a degraded observation yields two parameter-sensitivity profiles whose symmetric KL divergence defines the degradation score. The proposed formulation requires only the reference and degraded observation, without knowledge of the corruption process or degradation severity, making it applicable across images, audio, and 3D data. We further derive an exact closed-form expression for the diagonal Fisher of SIREN, enabling efficient computation. Experiments on image, speech, and volumetric 3D benchmarks demonstrate that SymKL accurately tracks degradation severity, remains monotonic where conventional metrics saturate or fail, and provides a unified modality-agnostic measure of signal degradation. Code will be made publicly available after the review process.

0.36

MorphSIG: Subject-Driven Image Generation via Decoupled Anchoring and Feature Transport

cs.CV ·2026-10-29 ·Hailong Yan, Yongrui Zhang, Xiangtao Zhang, Le Zhang
benchmark
Abstract

We propose MorphSIG, a training-free framework that reformulates subject-driven image generation (SIG) as pseudo-video feature transport. MorphSIG contains two core modules: Progressive Decoupled Anchoring, which uses frequency-domain decoupling and structural scrambling during early denoising to break structural locking and enable a gradual transition from free composition to identity alignment; and Pseudo-flow Guided Feature Transport, which constructs a latent-space pseudo-flow field to mimic Image-to-Video temporal coherence and transport subject semantics precisely. To address stylistic homogeneity and the lack of ground truth in existing benchmarks, we further introduce a benchmark with diverse stylized subjects. Extensive experiments show that MorphSIG significantly improves pose diversity and subject fidelity over baselines without additional training. Overall, MorphSIG bridges static generation and dynamic propagation, providing a cost-effective paradigm for high-fidelity subject consistency.

0.35

Multi-level Constrained Learning for Underwater Image Enhancement

cs.CV ·2026-10-29 ·Wangtao Bao, Liping Xie, Kanjian Zhang, Huimin Lu
code/data
Abstract

Underwater image enhancement aims to correct color and recover detail degraded by wavelength-dependent attenuation, scattering, and noise. Paired-reference supervision constrains output similarity, but does not by itself specify the color mapping, the strength of local detail corrections, or consistency with the degraded observation. We propose MCL-UIE, a paired-supervised framework that addresses these requirements by constraining restoration operations and the training objective. Monotonic Chromatic Transform (MCT) restricts global color mapping in log-LMS space through positive diagonal entries and bounded cross-channel coefficients. Uncertainty-Gated Wavelet Reconstruction (UGWR) modulates high-frequency residuals with a learned confidence gate calibrated against normalized wavelet reconstruction errors. During training, Reference-Anchored Degradation Replay (RADR) renders the enhanced image and its paired reference using shared latent degradation proxies, encouraging both to reproduce the input. The three components share a restoration network that combines structured color correction, locally controlled detail recovery, and reference-anchored replay. MCL-UIE ranks first in thirteen of sixteen full-reference dataset-metric comparisons and produces competitive visual results on no-reference benchmarks. Objective ablations indicate that confidence calibration is most effective when combined with replay supervision. RADR is used only during training and adds no inference-time computation. Our code is available at https://github.com/Wangtao-Bao/MCL-UIE.

0.35

Can LLMs Ignore What They Know? Measuring Knowledge Leakage in Persona-Based User Simulation

cs.CL ·2026-10-29 ·Sangmin Lee, Sangwon Lee, Sungbeom Choi, Jeeseung Park, Kibeom Myoung
published
Abstract · 13 pages, 2 figures. Accepted at UserSim @ NeurIPS 2026 (non-archival workshop)

When persona-conditioned responses are read as respondent-level simulations, each response should reflect the described persona's knowledge, and the respondent written to know nothing must act as if the model's knowledge were not there. We test whether it can, and find persistent favourable-item real-twin rating separation under the tested low-knowledge and explicit-ignorance prompts. We define the resulting problem (knowledge leakage: a response bias that arises when knowledge the real user could not have enters the simulated response) and present a methodology for measuring it quantitatively. The core instrument pairs real items with fabricated twins matched in surface form: screened fabricated variants designed to minimize prior item-specific knowledge. The paired difference therefore provides a diagnostic contrast for knowledge leakage even in rating tasks where no answer is correct. In a rating experiment across three knowledge-handling arms and five API endpoints from two model families, we find, first, that the personas we test do not fully suppress real-item existence and familiarity signals: a favourable-item real-twin rating contrast remains under an explicit in-context statement that the persona has never heard of the item. Second, the unfavourable-item contrast reverses sign across domain groups: pooling all unfavourable-content items across domains the contrast sits near zero (+0.077 [-0.057, +0.204]), yet the same estimator in the same conditions returns -0.359 where the rating scale asks about an item's documented hazard and +0.438 where it asks about a company or an organisation. A pooled analysis of unfavourable-content items against zero therefore does not detect the opposing domain-group differences. All estimates are conditional on the adequacy of twin construction and screening; human low-knowledge baselines remain necessary for population-level interpretation.

0.35

When to Assist and What to Suggest: Joint Gating and Memory-Grounded Hint Generation in One Compact Decoder

cs.CL cs.AIcs.HC ·2026-10-29 ·A. J. A. Jubayer Talukder, Mohammad Ariful Haque
theory
Abstract · 20 pages, 10 figures, 13 tables; supplementary material included (Sections S1-S8)

Proactive assistants offer short, memory-grounded hints during a conversation, deciding when to assist and what to suggest. Current systems split this between a gate and a larger generator; we evaluate both together on 28,058 decision points. Stage-wise generation scores are oracle-gate ceilings: behind a real gate, end-to-end exact match is 0.2619 against a ceiling of 0.2971, and end-to-end precision is 0.2387, leaving substantial headroom. A gate pays off through precision and compute, whose saving falls from 4.39x with an 8B generator to 1.38x with a 1B one. The memory profile drives hint quality: removing it costs 24.7 exact-match points (rescaled BERTScore 0.57 to 0.35), while retrieval and a larger generator add nothing on exact match. One 1B decoder can do both, gating from an intermediate layer and generating in the same pass. At its final checkpoint it matches or exceeds the two-model pipelines on automatic end-to-end measures with 1.31-1.76x less wall-clock time than the matched 1B pipeline; across three seeds, its deeper exit ties the separate gate. A human evaluation (1,144 ratings over 32 dialogues) supports the memory effect and rates the single model level with the matched 1B pipeline and below the 8B one. Graded metrics and LlamaPIE's native judge track human ratings more closely than exact match, which we report as a strict lower bound.

0.35

Open-set multi-view seahorse re-identification from underwater photograph sequences

cs.CV ·2026-10-29 ·Tristan Fauvel, Sylvie Louisy, Patrick Louisy, Thomas Fauvel
Abstract

Automated animal re-identification from photograph sequences can facilitate minimally invasive Capture-Mark-Recapture studies at scale. Yet, photographs taken during an observation session can vary greatly in quality, not always providing the necessary information for accurate re-identification. In this paper, we present new methods for taxonomic classification, sex determination, and open-set re-identification from image sequences. We train and evaluate our models on a dataset of 35,914 photographs of 474 individual seahorses observed in the wild. We demonstrate that our approach, by combining information from multiple photographs, yields state-of-the-art performance.

0.35

Conformal Prediction-Aware DOA Tracking

eess.AS ·2026-10-29 ·Maya Veisman Barness, Sharon Gannot, Bracha Laufer Goldshtein
published
Abstract · Presented at the 2026 Israeli Conference on Systems and Electrical Engineering (ICSEE 2026)

We introduce an uncertainty-aware DOA-tracking framework combining deep posterior estimation, conformal-prediction calibration, and temporal smoothing. The approach enhances stability against reverberation and burst-like directional noise, yielding more reliable trajectories and greater robustness than conventional DOA-tracking methods.

0.34

FieldGuideRAG: Image-Grounded GraphRAG for Fine-Grained Recognition with Vision-Language Models

cs.CV cs.IR ·2026-10-29 ·Jiacheng Ding, Yanliang Qi, Cong Guo, Peihua Mai
Abstract

We present FieldGuideRAG, a training-free GraphRAG method in which a vision-language model, or VLM, recognizes fine-grained species by comparing the query with real photographs stored on the nodes of a knowledge graph. Multimodal GraphRAG systems describe each entity in text, and text conveys little of the visual detail that separates look-alike species. Our graph stores one photograph on every species node and links each species to its relatives, its geographic range and its ecological facts. A frozen VLM verifies the query against the photograph of each candidate node. FieldGuideRAG fuses the verification scores with the range edge and answers questions about the organism from the fact edges of the verified species. On CUB-200, a Wikidata graph and four iNaturalist domains, the node photograph improves verification over the species name on every graph, while text descriptions add little. Verification of the candidate pool exceeds a SigLIP retriever by up to 16 points, the range edge adds 7 to 19 points on iNaturalist, and answering from fact edges is correct on up to 89.5% of 28,820 ecological questions. The verification scores also detect queries whose species is missing from the graph. Graphs, prompts and code will be released.

0.34

Robustness of Chatbot LLM Judgments of Presidential Actions and Hypothetical Reforms

cs.CL cs.CY ·2026-10-29 ·Oleg Vasilyev
datasetunusual
Abstract

Chatbots are increasingly consulted about public affairs, and their judgments of policies are useful only if they are robust. We measure this robustness in the setting of an ordinary user, through consumer chat interfaces. For 30 days, five chatbot configurations (Mistral, Gemini Flash, Claude Sonnet, and Claude Opus at two effort settings) were asked twice a day to score the long-term effect on the United States of 40 actions of the last four presidential administrations and of 15 hypothetical reforms, each with a confidence and a short comment. The judgments are robust: all configurations agree on the sign and the ranking of the administrations' mean scores, and two different configurations asked on the same day give opposite signs to the same action in about 4% of cases. Responses from different days differ more than same-day responses (which shared the order of the items); the difference grows slowly over the month, except where the service itself changed. We identify three such changes, which shifted the judgments while leaving little or no trace in the interface. Treating the comments as a retrieval task, we find them specific to each item and consistent across runs and chatbots, with lexical and embedding similarities agreeing. We release the dataset of 16,500 judgments.

0.34

InfantNet: A Large Scale Dataset for Infant Body Pose and Shape Estimation

cs.CV ·2026-10-29 ·Zhipeng Hui, Jiahao Wang, Xun Dong, Guofeng Zhang, Xingrui Wang +11 more
benchmark
Abstract

Infant pose and shape estimation is essential for applications in childcare, developmental monitoring, and medical diagnosis. However, existing methods and datasets are largely designed for adults, and direct transfer to infants fails due to substantial differences in body proportions, articulation limits, and frequent self-occlusion. To address this gap, we introduce InfantNet, the largest real-image infant dataset to date, comprising 108,902 RGB images of infants aged 6-18 months. Each image is annotated with 2D keypoints, and a curated subset of 11,642 images additionally includes 3D pose and shape annotations with full SMIL parameters, verified through multi-view rendering and reprojection-error filtering. We use an iterative annotation pipeline to ensure high fidelity across both 2D and 3D labels. InfantNet establishes a large-scale, comprehensive benchmark for infant 2D keypoint detection and 3D pose-and-shape recovery. Baseline experiments demonstrate that state-of-the-art adult pose estimators do not generalize well to infants, whereas fine-tuning on InfantNet yields a consistent improvement. The gains are even more pronounced for 3D pose and shape estimation, where SMIL-based estimation demonstrates robust performance with accuracy comparable to results reported for SMPL-based models. By releasing the InfantNet dataset and benchmark, we provide a vital resource for advancing infant pose analysis and related healthcare applications.

0.34

Qistas: A Pre-Registered, Portable Protocol for Fairness Evaluation of Large Language Models

cs.CL cs.CY ·2026-10-29 ·Nawar S. Alseelawi, Mustafa S. Aljumaily
positioncode/data
Abstract · 19 pages, 1 figure, 5 tables. Protocol paper; no model results. Pre-registration: https://osf.io/crg5n. Reference implementation: https://gi…

Fairness benchmarks for large language models (LLMs) are overwhelmingly English-born and American in their social categories; translated versions carry those categories into societies where they do not apply, and almost none fix their hypotheses before the models are run. We present Qistas (Arabic qistas, "the just balance"), a protocol for evaluating how LLMs treat the social, confessional, and regional constituencies of a society, portable across countries and instantiated first for Iraq. In its pilot form it measures only what needs no verdict on contested truth: procedural symmetry across matched counterfactual pairs, differential refusal, stereotype reliance under ambiguity, and terminological framing of documented events. Every measurement is paired by construction and analysed with paired machinery (sign-flip permutation tests, exact McNemar tests, template-clustered bootstrap intervals, Holm-corrected families), with a power analysis that states, before any data exist, the smallest effect each axis can detect after the design effect of template clustering. We release the authoring apparatus (a pruned 3x5 taxonomy with four cells closed by design, per-type checklists, a cumulative rulings log, a structural audit, and a human review notebook) and report its first field test: a semantic review of a 58-item seed bank in which 24 of 25 template groups failed at least one check. The defect patterns it exposed (evaluation criteria leaking into prompts, token-level swaps that break Arabic agreement, epistemically unbalanced pairs, single-theme refusal probes, unwritten dialect twins counted as coverage) are, we argue, general hazards of counterfactual fairness benchmarks. This is a protocol paper, not a results paper: no model has been evaluated, and the accompanying pre-registration commits the hypotheses, metrics, exclusion rules, and analysis plan that the pilot will follow.

0.34

PicturePerfect: Diffusion-based Same-identity Facial Replacement

cs.CV ·2026-10-29 ·Manik Dhingra, Yashesh Savani, Mahesh Reddy, Antoine Mercier, Jiahe Liu +1 more
publisheddataset
Abstract · Camera-ready version (ECCV 2026 workshop - Third Workshop on Foundation and Generative Models in Biometrics)

With the growing success of diffusion models in computer vision, we explore their potential for same-identity facial replacement in photographs. Specifically, we propose a diffusion-based method, built on top of a pre-trained text-to-image model, that takes as input a portrait image of a person and a second reference image of the same individual, potentially captured under different conditions. The goal is to seamlessly replace the input face with the reference face, while keeping the background intact. Surprisingly, despite the clear real-world utility of this task, no recently published work has directly addressed face replacement in this specific setting. To support this goal, we construct a large dataset of image pairs depicting the same person under varying facial expressions and poses. Experimental results demonstrate that our approach produces more realistic and identity-consistent results than existing face reenactment models.

0.34

Telling Is Not Seeing: Textual Personas Rarely Make Vision-Language Models Answer Like the People They Describe

cs.CV cs.AI ·2026-10-29 ·Jiawei Chen
hot direction
Abstract

Vision-language models (VLMs) are increasingly prompted with personas, such as "You have low vision", to stand in for people in user studies. We ask whether a persona makes a model see like the person it describes. We compare persona answers with measured people with low vision, color blindness or hearing loss, and with simple rules that use only the facts the persona states. The persona never beats these rules, and it erases the real differences between groups and individuals. Describing blur leaves recognition unchanged in all 14 models we test, while blurring the photo lowers it as it does for people. Inside open models, the description changes neither the image information nor the answer probabilities. Yet the models know the effects they fail to show: they can estimate how blur and illusions affect people. Telling is not seeing: a persona makes a model say what it knows about people, not see what they see.

0.33

Does the Citation Matter? Quality, Grounding, and Physician-Rubric Accuracy of Google's AI Overviews for Medical Information

cs.IR cs.CLcs.CYcs.HC ·2026-10-29 ·Sam Agnoli, Brigid Reilly, Lily Barna, Emőke-Ágnes Horvát
benchmarkcode/data
Abstract · 12 pages, 5 figures, 10 tables. Code and data: https://github.com/SamAgnoli/AIOverviews

Google's AI Overviews (AIOs) are the first result users encounter when searching the web, shifting the work of selecting and evaluating sources from the user to Gemini. This may be especially consequential for health information, given that Google's traditional algorithm was designed to prioritize quality sources, and inaccurate information carries health risks. We evaluated 3,611 benchmark health questions, comparing each AIO against that question's top five traditional results on source selection and quality. We also evaluated the AIOs' statement grounding in their cited sources and how accurate AIOs were against physician-written rubrics. Overviews appeared for 96.4% of questions but drew on substantially different evidence than traditional search: only 44.4% of the top five results were cited by the corresponding AIO, which instead overrepresented commercial health publishers (e.g., WebMD, Healthline) at the expense of academic and government sources (e.g., Cleveland Clinic, National Health Service). AIO-cited pages scored lower than traditional results on four of six well-established credibility criteria. The grounding of AIO statements to their sources was poor: after correcting for classifier error, an estimated 56.7% of cited statements were supported by their cited page, and a further 20.8% of all statements were factual claims with no citation at all. When AIOs were graded against physicians' rubrics, they earned 38.2% of all available points (e.g., accuracy 54.2%, completeness 21.6%), and among the subset of AIOs whose physician rubric anticipated a specific inaccurate or potentially harmful claim, 22.2% of AIOs contained the harmful content. Notably, the accuracy of AIOs was unrelated to the grounding of their statements. We recommend AIO literacy interventions for users and holding AIOs at least to Google's own health-ranking standards.

0.33

RFW-A: Agentic Procedural Generation of 400K Articulated Assets over 500 Categories

cs.CV ·2026-10-29 ·Yuanbo Lyu, Changjian Pan, Saining Zhang, Yi Yang, Peishuo Li +17 more
code/data
Abstract · 37 pages, 19 figures. Project Page: https://brain-labbb.github.io/RFW_A_web/

High-quality articulated assets are essential for embodied perception, simulation, and manipulation, yet constructing them at scale remains difficult. Manual annotation and procedural modeling require substantial human effort. Recent agentic systems automate asset construction through program synthesis, but repeat costly reasoning and refinement for every new asset. Our key observation is that assets within a category can share construction logic despite variations in structure and appearance. Agentic synthesis and procedural generation are therefore complementary: agents can construct reusable programs, while procedural execution reuses the resulting logic across many samples. We therefore present RFW-A, which combines the two in a new articulated-asset generation paradigm. Starting from existing articulated programs, RFW-A organizes reusable components and their dependencies into category-level generators. Agents test and refine these generators across sampled configurations and motion states. Once constructed and validated, the generators produce diverse assets through procedural execution with no further model calls or token charges. The estimated API cost of generator authoring averages $10.62, amortizing to $0.05 per asset over 200 assets, versus $1.24 for the sampled per-asset agentic synthesis baseline. Using RFW-A, we built RFW-A-400K, containing over 400K sim-ready articulated assets across more than 500 categories. Experiments show high structural and kinematic validity and stable execution across physics simulators. Beyond asset validity, training with RFW-A-400K improves part-level articulation estimation on seen and unseen categories, while policies trained in simulation with retrieved and program-edited assets transfer zero-shot to real articulated objects.

0.33

Mage-VL-Cowork: A Full Training Recipe to Learn Hybrid Computer Use From Evolved Harnesses

cs.CL cs.LG ·2026-10-29 ·Xiaoyi Zhang, Haoqing Wang, Xinjie Zhang, Xiang Ming, Xing Wang +17 more
code/datacommunity pick
Abstract · Project page: https://microsoft.github.io/Mage

Effective computer-use agents (CUAs) should seamlessly combine visual GUI interaction with code and terminal execution within a single policy. We present Mage-VL-Cowork, a 4B hybrid CUA trained from a general multimodally aligned checkpoint through an end-to-end, progressive five-stage curriculum across midtraining, SFT, and RL. To power this curriculum at scale, Mage-CUA-Forge replaces per-task agent construction with reusable, skill-conditioned generation workflows and application-specific preparers, synthesizing 293,806 desktop and web tasks across 6,915 skills while reducing recurring per-task expenditure by 82.6%. Rather than directly distilling a default teacher, we evolve the teacher harness and instantiate specialized CUA harnesses to yield high-quality trajectories alongside stage-specific supervision. Mage-VL-Cowork-4B finally climbs to 66.0% on OSWorld-Verified through different training stages with more efficient reasoning cost, achieving state-of-the-art performance at its parameter size and pushing forward the frontier of on-device intelligence.

0.33

What Survives Label Downsampling? Exact Region Survival for Coarse Segmentation Targets

cs.CV ·2026-10-29 ·Zhiyuan Zhang, Yukun Ou, San Tek Choi
Abstract · 25 pages, 13 figures, including supplementary material

Segmentation networks that predict at an output stride are trained against label maps downsampled by strided sampling, nearest-neighbor resampling or majority pooling. These methods label each coarse grid cell on its own, so a region that wins no cell is absent from the target and receives no positive label. We present MatchDown, which builds the coarse target from a minimum-cost matching between regions and the grid cells they occupy. It keeps as many regions as any one-region-per-cell target can, certifies how many losses are unavoidable, and with priority tiers (MatchDown-P) keeps every evaluated region (a region that a metric scores) of at least one cell's area. On COCO, Cityscapes and BBBC038 at factors 2 to 16, MatchDown-P loses exactly the minimum number of regions, whereas the downsampling used by StarDist, nnU-Net and connectomics pyramids loses four to five times as many on COCO at factor 8. Keeping a region in the target and reconstructing its shape are different problems: we show that no region lost by majority pooling can be reconstructed above IoU 0.5 by upsampling any coarse target, yet on BBBC038 a star-convex head trained on MatchDown targets segments 35% of the nuclei that majority pooling loses above IoU 0.5, against 15% when trained on majority targets and 24% on StarDist's strided target. On two nucleus datasets, with thresholds cross-validated for count error, this head has instance accuracy at IoU 0.5 (Acc50) 2.6 to 4.2 points above the strided target and at most 2.8 points below cell-center sampling, and a lower count error than the strided and cell-center targets on BBBC039. On BBBC039, an affinity head, which outputs one label per cell, lowers count error relative to majority targets by 9.3 nuclei per image at 7.2 points lower Acc50. MatchDown runs once, before training, in place of the label-downsampling step.

0.32

Beyond IPS: Reliable Counterfactual Evaluation in Multi-Stage Ad Systems without Logged Propensities

cs.IR ·2026-10-29 ·Mohsen Malmir, Mohamed A. Radwan, Houssam Nassif, Murat Ali Bayir
Abstract

Offline evaluation of ad-ranking policies in multi-stage delivery systems is challenging because action propensities are often unavailable, limiting inverse propensity scoring (IPS). Reward-model-based estimators such as the Direct Method (DM) avoid propensities but can be biased when candidate policies induce impression distributions that differ from the logged data. We formalize DM-based counterfactual replay and propose mixture-weighted reward modeling, which reweights logged data toward a mixture of candidate-policy distributions using density ratios from a multiclass domain classifier. We bound each candidate's value-estimation error by overlap, cold-start, and replay terms, and derive an asymptotic rule that uses the cold-start mass to decide whether to trust offline selection or route candidates to online A/B testing. On an ads-delivery simulator and on production traffic from a large-scale CTR ranking system, the mixture-weighted reward model improves policy selection over its unweighted counterpart in both settings, with the largest gains at low overlap in simulation. Combined with a doubly robust correction, it attains the highest Top-1 accuracy in production and at low and medium overlap in simulation.

0.32

BRACHIO: Broad Representations Across Centres for Head-and-neck Imaging Oncology (HECKTOR 2026)

cs.CV ·2026-10-29 ·Roxanne De Maeyer, Ine Dirks, Yannick Brackenier, Simon DeKeyser, Jef Vandemeulebroucke +1 more
published
Abstract · HECKTOR Challenge 2026, MICCAI, head and neck cancer, PET imaging

The HECKTOR 2026 challenge benchmarks machine learning for head and neck cancer, where teams delineate primary and nodal tumour volumes, perform TN staging, and predict recurrence-free survival (RFS). This report describes the submissions for the HECKTOR challenge of team nucli. As the three tasks span both dense and global information from multicentric PET/CT scans, we utilise self-supervised pre-trained encoders, training a 2D DINOv2 encoder (PET-DINOv2-MIPs) on HECKTOR 2026's PET scans. Our final test submission ensembles a DynUNet trained from scratch with a SwinUNETR whose Swin trunk was initialised from a SimMIM-pre-trained checkpoint for segmentation, reaching a mean Dice of 0.5409. For TN staging, GTVp/GTVn geometric features from the segmentation masks are fit to stage probabilities by logistic regression (LR) and fused with PET-DINOv2-MIPs deep features (PCA + LR) via a second-level LR, reaching balanced accuracies of 0.3962 (T-stage) and 0.6432 (N-stage). For the RFS prognosis task, we fine-tune the last two blocks of the PET-DINOv2-MIPs ViT-S encoder jointly with an MLP head under a CoxPH loss, concatenating clinical covariates with the imaging features, reaching a test Harrell's C-index of 0.6577 and a weighted challenge score of 0.5802. The prognosis result is comparable to the best entry of the previous edition and consistent with our internal cross-validation, while our segmentation result is well below it and limits the staging results that depend on it.

0.32

TaskFrame: Connecting Task-Level Planning with Frame-Level Control for Game Agents

cs.CV cs.AI ·2026-10-29 ·Guangzhi Wang, Junfu Pu, Yiran Wang, Xingyilang Yin, Honghao Chen +8 more
Abstract

Building general game agents requires both high-level task reasoning and low-latency action control in rapidly changing visual environments. These capabilities favor models with substantially different computational profiles, making them difficult to realize within a single model. In this work, we present TaskFrame, a hierarchical planner-controller framework that connects task-level planning with frame-level keyboard-and-mouse control. A vision-language task planner decomposes a high-level goal into an ordered sequence of atomic instructions and is optimized through supervised learning and dependency-aware GRPO. The hidden states of special goal tokens that terminate each atomic instruction provide continuous goal embeddings that are cached and reused across control steps. A causal Transformer controller combines the embedding of the active goal with recent observations to predict discrete keyboard and mouse-button actions and continuous camera motion at every game step. We train the system in three stages: planner training, controller pretraining, and joint alignment. Experiments in Minecraft show that TaskFrame is competitive with general-purpose vision-language models on instruction decomposition, and that in closed-loop evaluation the agent achieves a higher combat success rate than the evaluated baselines.

0.32

Regional Anchor Routing for Multi-Object Motion Transfer

cs.CV ·2026-10-29 ·Yuze Li, Fuchen Long, Xinlong Sun, Cheng Yan, Xinyu Zhang +1 more
code/data
Abstract · Project Page: https://ethan-li123.github.io/RART_page/

Compositional motion transfer requires a model to extract motions from multiple reference videos, and apply each motion to a user-specified object in one target scene. Existing methods normally struggle to reliably assign different reference motions to the correct target objects, thus a motion intended for one object may affect another. Meanwhile, some motion-specific tuning approaches are still facing another issue: they must run a new optimization whenever the reference video changes, making repeated transfers inefficient. To address these limitations, we propose RART, a framework that unifies object-wise motion extraction, target-region anchoring, and frame-aligned injection into a frozen video Diffusion Transformer for multi-object motion transfer. For each reference motion, RART extracts temporally ordered features from the masked reference video, and then conditions them on the target first-frame mask to form mask-anchored motion tokens. Frame-aligned motion cross-attention allows each generated frame to read only the motion tokens at the corresponding time slots, composing all object motions in one denoising process. During training, a motion-attention routing loss encourages each target region to attend to its corresponding motion tokens and reduces interference between objects. Therefore, RART can handle novel reference-target combinations without per-video optimization. Experiments show that RART achieves accurate motion preservation, efficient inference, and strong compositional generalization.

0.31

Accurate MRI-free PET Quantification and Reliability Assessment via CT-to-MRI Synthesis and Conformal Prediction

cs.CV ·2026-10-29 ·Jungwoo Seo, Seung Yun Choi, Kun Ho Lee, Seun Jeon, Byoung Seok Ye +1 more
Abstract · 28 pages, 8 figures, including supplementary material. Submitted to Medical Image Analysis

Standardized uptake value ratio (SUVR) quantification of amyloid PET commonly relies on a subject-specific T1-weighted MRI, yet a usable MRI is not always available. We therefore combined CT-to-MRI synthesis with conformal prediction for MRI-free PET quantification with reliability assessment. A slice-consistent Brownian Bridge Diffusion Model was strengthened with multi-dataset training and data-source conditioning to enrich the anatomical prior learned from MRI, and with expanded multi-slice CT context to exploit the through-plane continuity of subject-specific anatomy in CT. Synthetic MRI was used for FreeSurfer-based anatomical parcellation and PET quantification, while deep-ensemble uncertainty was calibrated with conformal prediction to estimate a plausible range for the corresponding real MRI-based SUVR. On $^{11}$C-PiB PET from OASIS-3, the proposed method achieved the highest synthesis and parcellation fidelity among the compared methods. Mean cortical SUVR from the proposed synthetic MRI agreed closely with the real MRI-based reference (MAE $0.016 \pm 0.015$, Pearson $r = 0.9987$) and yielded the same amyloid-positivity classification as the real MRI in all 87 test subjects at the prespecified cut-off. Conformal prediction achieved empirical coverage close to the prespecified levels, and at 90% target coverage, 85 of 87 subjects had prediction intervals that supported an unambiguous amyloid-positivity decision. Zero-shot evaluation on an independent external cohort showed that the proposed CT-to-MRI model could support PET quantification without retraining, although conformal calibration required target-cohort recalibration to recover the intended coverage. These findings support CT-to-MRI-based MRI-free PET quantification as a promising approach for providing accurate SUVR measurements and an explicit assessment of their reliability.

0.31

RADIOlogy: A Single-Pass Approach to Training-Free Zero-Shot Anomaly Detection

cs.CV ·2026-10-29 ·Joaquín Figueira, Giacomo D'Amicantonio, Egor Bondarau
publishedcode/data
Abstract · Accepted to WACV 2027. Code available at https://github.com/jacomof/RADIOlogy

Current zero-shot anomaly detection methods often rely on complex prompt-learning techniques or computationally expensive sliding-window strategies to achieve fine-grained visual alignment to detect small-scale anomalies. This reliance stems from the fundamental limitation of standard CLIP architectures, where [CLS] token attention and spatial patch pooling struggle to capture subtle, localized anomaly features. While prompt-learning approaches aim to bridge this gap, they typically require a large number of ground-truth segmentation masks from auxiliary datasets, which are often unavailable in specialized domains. In this work, we show an emerging property of multi-head attention pooling in vision-language models: remarkable zero-shot anomaly localization and detection right out of the box. We translate this property into a single-pass anomaly detection technique, RADIOlogy, that does not require dense, windowed execution and significantly outperforms other training-free baselines. Instead, the method relies on a simple contrastive scoring approach that compares the average similarity between vision features and a suite of normal and abnormal prompts. Furthermore, we leverage the agglomerative features of RADIO models to enhance pixel-level anomaly segmentation using a dense anomaly refinement pipeline. Finally, we show how the vision-language family of training-free zero-shot anomaly detection techniques can be applied successfully to the medical domain with minimal modifications. Code is available at https://github.com/jacomof/RADIOlogy.

0.31

Spatial Atlas-Informed Loss Ensembles for Whole-Body PSMA PET/CT Metastasis Segmentation

cs.CV ·2026-10-29 ·Grant N. Chau, Bashirul A. Biswas, Biratal Wagle, Matthew E. Maeder, Michael S. Leapman +2 more
Abstract · 21 pages, 6 figures

Existing whole-body PSMA PET/CT prostate metastasis segmentation methods generally rely on images alone, without explicitly incorporating where metastases tend to occur. We investigate whether a population metastasis atlas can improve lesion detection when integrated into model training. We constructed voxel-wise mean and standard deviation maps from 375 expert-annotated training scans and used them to weight the cross-entropy loss of an nnU-Net. The additional penalties for missed lesions and false positives depend on spatial atlas statistics and diminish as predictions become confidently correct. Seven configurations were compared on validation data; three were combined by majority voting into SAILS (Spatial Atlas-Informed Loss enSemble). We compared SAILS with a baseline nnU-Net and inference-time atlas fusion on AutoPET~III (110 disease-positive cases) and Deep-PSMA (100 external cases). No atlas or registration is needed at inference. Results: On AutoPET~III, SAILS improved whole-body Dice (0.531 vs. 0.520; p=0.006), reduced missed-lesion volume (9.5 vs. 12.1 cc/case; p<0.01), and increased pooled lesion sensitivity (0.857 vs. 0.843; p=0.004). On Deep-PSMA, SAILS detected 126 additional reference lesions (sensitivity 0.769 vs. 0.751; p<0.001), without a significant whole-body Dice change (-0.009; p=0.36). High-risk bone Dice improved in both cohorts (0.576 vs. 0.552 and 0.667 vs. 0.653; both p<0.01). False-positive volume increased by 2.3 and 18.2 cc/case, respectively (p<0.001). Atlas-weighted training improves lesion detection, especially in high-risk bone, without test-time atlas registration, but increases false positives.

0.31

AB-Attention: Accurate Block Attention for Accelerating Video Diffusion Transformers

cs.CV ·2026-10-29 ·Bin Xu, Mo Zhang, Zhengfeiyang Zhang, Xinfeng Yang, Yixing Ma +4 more
code/data
Abstract · 20 pages, 5 figures, 3 tables

Video diffusion transformers generate high-quality videos, but the quadratic cost of 3D full attention dominates their inference. Dynamic block-sparse attention reduces this cost by computing exact attention only on key/value blocks selected from current activations. However, its fidelity degrades under aggressive sparsity: blocks are scored from coarse summaries or formed by costly global clustering, pruned blocks vanish from the softmax, and whole tiles are kept or dropped even when only a few of their keys matter. To address these limitations, we propose Accurate Block Attention (AB-Attention), a training-free method that accurately approximates dense attention at high sparsity. Its Intra-Block Clustering Selector (IBCS) estimates block importance faithfully yet cheaply by clustering queries only within each query block, without permuting tokens. Sampling with Inclusion-Probability Weighting (SIP) then samples blocks in proportion to these estimates and reweights them by inverse inclusion probability, so that one pass estimates dense attention rather than truncating it. Finally, its kernel gathers key/value blocks much smaller than a tile into full tensor-core tiles, refining selection at no extra tensor-core cost. On Wan2.1 and HunyuanVideo, AB-Attention skips 89--92\% of attention, whereas the baselines compute 1.8--3.6$\times$ as much, yet it still improves PSNR by up to 4.4~dB over the strongest baseline and achieves up to 2.23$\times$ end-to-end speedup. Our code is available at https://github.com/InfRec-Cardinal/AB-Attention.

0.31

LiDAR-SHIELD: A Fully Dockerized Adversarial Robustness Assessment Framework for LiDAR-Based 3D Perception Models

cs.CV cs.AI ·2026-10-29 ·Alexandros Gkillas, Ioulia Kapsali, Christos Anagnostopoulos, Aris S. Lalos
code/data
Abstract

LiDAR-based 3D perception underpins autonomous driving and robotics, yet deploying these models safely requires stress-testing them against adversarial inputs before real-world use. Although state-of-the-art adversarial attacks exist, their implementations are often fragmented across individual repositories, each tied to a specific architecture or task and requiring manual environment reconstruction and model-specific integration code. We present LIDARSHIELD, a unified and fully Dockerized framework that reverses this workflow: rather than installing the target model inside the toolkit, LIDARSHIELD exposes any model through three lightweight HTTP endpoints, namely /info, /forward, and /gradient As a result, connecting a new model or dataset requires only a single loader function or adapter. The framework supports both 3D semantic segmentation and 3D object detection through the same engine-service protocol, integrating a task-aware suite of state-of-the-art attacks, including PGD and Square Attack for segmentation, and point perturbation, point removal, point injection, and LiDAttack for detection. This enables systematic exposure of corner cases and vulnerabilities across heterogeneous LiDAR backbones. The LiDAR-Shield codebase is publicly available at https://github.com/alexandrosgk/LiDAR-shield.

0.31

Agentic LLM reasoning in pediatric diabetes outcome prediction from multimodal medical records

cs.CL q-bio.QM ·2026-10-29 ·Daniel Palacios, Guantong Qi, Anna Kadi, Natali Borrego, Sukru Aras +6 more
Abstract · 18 pages main text plus 40 pages Supplementary Information

Large language model (LLM) agents are proposed for clinical prediction but are untested in pediatrics, where cohorts are small and outcomes are uncommon. Using records from 2,489 youth with type 2 diabetes (T2D) and 4,018 with type 1 diabetes (T1D), spanning structured data, note-derived social determinants and, in T1D, continuous glucose monitoring data, we predicted seven T2D outcomes, four also in T1D, at 24 months from 18 months of follow-up. CEDAR (Clinical Evidence-Driven Agent for Risk-prediction), an evidence-citing, self-auditing LLM pipeline, outperformed trained machine learning in both cohorts (pooled area under the receiver operating characteristic curve 0.815 versus 0.656 to 0.700 in T2D and 0.756 versus 0.603 to 0.631 in T1D), most on the sparsely screened outcomes. Only 4 of 70,246 values it cited did not match the chart. Reading only the evidence and reasoning cited by CEDAR, two pediatric endocrinologists reached the same prediction as CEDAR in about nine of ten cases, agreeing with each other on 90.5%.

0.31

DyCraft: An Agentic System for Long-Horizon 4D Scene Generation

cs.CV ·2026-10-29 ·Jeff Tan, Hang Dai, Zeqian Long, Yanzhe Lyu, R. Kenny Jones +3 more
code/data
Abstract · Project page: https://dy-craft.github.io/

While frontier coding agents can generate high-quality 3D scenes and articulated assets, they struggle to create coherent long-horizon animations. When equipped with general-purpose harnesses, they frequently produce physically implausible outputs, generating objects that interpenetrate, lack physical support, or move unnaturally. Such bare agents struggle to repair these errors because they default to scripting the full animation in one pass, making it difficult to verify their work and disentangle complex spatiotemporal constraints. Drawing on the practice of human animators, we introduce DyCraft, a system for animated 3D scene generation that breaks animation into steps that coding agents can more reliably solve. DyCraft first expands the user's prompt into a detailed plan and builds a static scene. It then creates a 3D storyboard of key animation poses, where each frame is a static 3D scene that the agent can build and debug in isolation. Storyboarding decomposes the problem into pairs of verified start and end states, allowing the agent to author local motion between them with a collection of motion authoring tools. We evaluate DyCraft on 24 long-horizon, multi-shot animation tasks covering humanoid robots, human and animal characters, and physical mechanisms. Compared with a native GPT-6 agent in Codex and prior agentic baselines, DyCraft substantially reduces physical violations and generates higher-quality animated 3D scenes that are more often preferred by human raters.

0.31

AdMuse: Skill-Guided Creative Advertising Image Generation

cs.CV ·2026-10-29 ·Minyan Luo, Yuxin Zhang, Meng Zhao, Xincan Wang, Fuzhang Wu +4 more
Abstract

An outstanding creative advertising image should both attract attention and communicate a product's positioning. How to choose a persuasive conceptual bridge before rendering begins is challenging for advertising generation. We present AdMuse, a skill-guided multi-agent framework that distills high-quality advertisements into reusable creative skills for cross-domain advertising strategy. Rather than retrieving isolated exemplars, AdMuse abstracts recurring bridge patterns into a skill library that can guide new product briefs. At generation time, AdMuse coordinates skill routing, multi-candidate ideation and selection, and VLM-guided visual refinement: diverse skills propose alternative concepts, relative judgment selects the most promising direction, and vision-language feedback helps realize the selected intent in the final image. This design treats advertising creativity as a structured decision process rather than a single prompt-writing step. Experiments across diverse product and brand positionings show that AdMuse produces visually compelling advertisements that faithfully communicate product positioning.

0.31

Wizard: Manipulate Any World in Real Time via Agentic Control

cs.CV ·2026-10-29 ·Yuzhe Wu, Zijun Lin, Yeying Jin
datasetcode/datacommunity pick
Abstract · Project Page: https://wyzardwu.github.io/wizard

Recent advances in video world models have enabled the generation of increasingly realistic and interactive virtual environments, opening new possibilities for player-customizable gameplay. Agentic frameworks further advance this direction by decoupling explicit game logic and controllable world representations from visual generation. However, existing approaches support only simple game settings and predefined motion patterns, generalize poorly to unseen environments and characters, and offer limited real-time interactivity. In this paper, we introduce Wizard, an agentic game world model that enables real-time, precise control over character motions while generalizing zero-shot across novel scenes, characters, and actions. To achieve this, we construct a diverse dataset spanning 750 scenes and 178 characters, incorporating skeleton and geometry information as structured representations. We leverage these representations as intermediate proxies to condition a video generation model, together with an adaptive masking mechanism that improves alignment between the conditioning signals and generated video rollouts. Furthermore, we introduce an agentic framework that explicitly maintains and updates game states, allowing players to freely configure their virtual worlds and exercise fine-grained control over character motions while preserving visual quality and gameplay consistency. By combining agentic control with structured proxy-based generation, our framework unlocks new possibilities for player-driven game creation with real-time, generalizable, and mechanically consistent world simulation.

0.31

UniTele: Learning Device-Conditioned Degradations for Unified Smartphone Telephoto Super-Resolution

cs.CV ·2026-10-29 ·Yuchen Zhang, Aojie Li, Wei Cheng, Xianfang Zeng, Runxing Liu +3 more
benchmark
Abstract

Super-telephoto imaging on mobile phones suffers from complex, device-dependent degradations that are poorly captured by conventional handcrafted pipelines. Collecting large-scale aligned training pairs across different phones and diverse scenes, however, is both costly and difficult to scale. We therefore study whether device-specific degradations can be learned from only a small number of physical captures and transferred for training a unified super-resolution (SR) model. Our key idea is to factor degradation learning into two learnable components: one captures the device-specific imaging characteristics under well-resolved conditions, while the other captures the additional information loss and artifacts that arise in super-telephoto imaging. We build a paired benchmark across three phone brands (Huawei, Apple, and vivo) by photographing the same content at two display sizes and zoom settings. The larger displayed image is captured at the phone's maximum optical zoom setting, while the smaller image is photographed directly at the phone's highest native zoom setting through the phone's native imaging pipeline. The benchmark contains 519 training and 60 held-out pairs. We use the benchmark to learn device-specific degradation, then apply it to external images to synthesize thousands of additional pairs to train our UniTele SR model. Relative to the pretrained Qwen and FLUX backbones before adaptation, the full UniTele training recipe substantially improves paired restoration fidelity and perceptual similarity on the held-out benchmark and yields markedly more stable and visually faithful restorations.

0.31

Confidence Filtering in Video-Based Gait Assessment: Measurement Error versus Selection Shift

cs.CV q-bio.QMstat.ME ·2026-10-29 ·David Regev, Eran Beeri Bamani
Abstract · 6 pages + 2-page appendix, 4 figures, 2 supplementary figures, 2 tables. Submitted to the 2027 IEEE International Conference on Rehabilitati…

Video-based gait assessment often keeps only high-confidence pose observations, which can lower measurement error but change the movement assessed. We compared confidence selection with budget-matched random selection in a Parkinson-related cohort (n = 96) and an NF2-related replication cohort (n = 18). In the Parkinson-related cohort, confidence selection produced a small but consistent shift in the observed movement summary (+0.05 SD, 95% CI 0.03 to 0.07), while held-out Mini-BESTest error stayed flat (-0.04 points, -0.13 to 0.05); a policy preserving motion-group proportions reduced the shift, while NF2 showed no shift and its predictor did not beat a constant baseline. Because this shift mixes measurement error with selection shift, we decompose and simulate both on the cohort's window structure. Filtering lowers local error when error depends on confidence, but biases the endpoint once confidence tracks movement more than error. In the Parkinson-related data, confidence was only weakly tied to movement (rho = 0.14). There, filtering is harmless if well-tracked windows are unbiased. If instead systematic error only redistributes within participants, filtering can add up to 0.58 degrees. Unchanged clinical error from such small-cohort predictors therefore cannot show that quality filtering preserves the assessed movement.

0.30

Mastering Multi-State Dynamics: Continuum Damage and Viscoplastic MPM for RL-Based Dexterous Manipulation

cs.GR ·2026-10-29 ·Yusen Feng, Shangbo Luan
Abstract

Robotic manipulation of multi-state materials needs a simulator that handles large deformation, topology change, and two-way contact between tools and materials while staying fast enough to drive online reinforcement learning. We present MG-CD-MPM, a tool-centered localized extension of CD-MPM. The implicit Ginzburg-Landau phase-field damage system is solved on a local patch by preconditioned conjugate gradients using a geometric-multigrid V-cycle as the preconditioner. The background MPM discretization remains uniform: only the damage solve is localized. Two-way rigid-tool coupling converts MPM contact impulses into forces and torques on the tool, exposing material resistance directly to the policy. The resulting Gym-like simulator covers soft solids, brittle materials, granular media, and viscoplastic or semi-fluid media within a unified per-particle constitutive layer. We demonstrate cutting, scooping, and shoveling in six scenes, combining scripted tool motions with PPO policies trained on two cutting tasks.

0.30

CASH: A Capability-Aware Subsidy Harness for Advertiser Subsidy Allocation

cs.IR ·2026-10-29 ·Minmao Wang, Huihuang Lin, Likang Wu, Hongke Zhao, Shijie Yi +1 more
Abstract

Advertising platforms subsidize advertisers through billing discounts, and Advertiser Subsidy Allocation (ASA) chooses a subsidy level for every account to maximize total discounted revenue under a platform-wide subsidy-rate constraint. In practice, operators configure these subsidies from business rules, advertiser tiers and operating goals. Such configuration draws on rich business experience, yet as the advertiser base grows, tier-level policies can neither resolve how individual accounts respond across subsidy levels nor allocate a shared budget at the granularity of single accounts. We propose CASH, a Capability-Aware Subsidy Harness for ASA. CASH pairs a knowledge-guided large language model (LLM) with a data-driven numerical response model, learns from realized outcomes where each is locally reliable across accounts and subsidy levels, and turns this reliability into explicit capability licenses that govern source selection, agent review and budget allocation. Its Expected Policy Coverage Gain (EPCG) directs a limited feedback budget toward evidence expected to expand high-value reliable coverage, and the resulting feedback further improves the numerical model and updates capability estimates. In a one-month online experiment with CASH serving 60% of a large-scale advertising platform's traffic, CASH raises revenue per account by 15.35% and ADVV by 19.86% over Control at an essentially unchanged subsidy rate, with further gains in revenue quality, LTV and retention.

0.30

Dreams Are Cheaper Than Flights: Curated World-Model Data Improves Aerial Detection

cs.CV cs.AI ·2026-10-29 ·Rugved Katole, Vedant Patil, Jenna Kline, Christopher Stewart
benchmark
Abstract · 6 pages, Corl 2026

Robot perception is increasingly data driven, yet field data is slow, costly, and sometimes disruptive to collect. Video world models offer a cheaper source: seeded with real frames, they extend a recorded scene into new footage with the deployment's sensor, altitude, and environment. We ask whether such synthetic data matches the utility of real field data and whether detectors trained with it perform in the field. Seeding a 7B-parameter world model with wildlife drone video, we more than double the training set with curated frames. On held-out videos from the same drones, mAP@50--95 rises 4.3\% when these frames are added to real data and 10.3\% when they replace it, and up to 33.5\% on an urban benchmark. Although an expert rejects an estimated 87.5\% of generated clips, the data costs about \$1,100 of GPU time, against more than \$3,200 of field labor alone for a less useful real training set. Synthetic data thus shifts data costs from fieldwork to compute.

0.30

PFA: Attention Can Be Progressively Focused in Long-Sequence Generative Recommendation

cs.IR ·2026-10-29 ·Pengjie Zhou, Weiqin Yang, Haibo Xing, Hao Deng, Jinxin Hu
code/data
Abstract · 11 pages, 3 figures, 7 tables

Long user behavioral sequences improve the performance of generative recommendation (GR), but computing full attention over long histories during autoregressive beam search decoding incurs substantial online inference latency. To accelerate long-sequence decoding, existing approaches typically rely on static history compression or dynamic top-k sparse attention; however, the former risks losing critical semantics, whereas the latter incurs significant overhead from repeated token selection and non-contiguous memory access. In this work, we thoroughly analyze attention patterns during GR decoding and reveal two key properties: (i) progressive attention focusing, where attention progressively focuses on a small subset of critical historical tokens; and (ii) cross-step ranking consistency, where their importance rankings remain relatively stable across subsequent decoding steps. Motivated by these observations, we propose Progressively Focused Attention (PFA), an efficient sparse decoding framework for long-sequence GR. PFA retains full attention at the initial decoding step to explore global historical preferences. It then establishes a shared importance ranking through cross-beam weighted aggregation and reorders the key-value cache of the selected tokens into a contiguous memory layout. Subsequent decoding steps reuse this ranking and geometrically reduce the token budget, preserving critical semantics while lowering computational and memory access overhead. Extensive experiments on two public benchmarks and a large-scale industrial dataset demonstrate that PFA outperforms state-of-the-art baselines. A 7-day online A/B test on a large-scale e-commerce advertising platform further confirms that PFA delivers substantial business gains while maintaining low inference latency. Our code is publicly available at https://anonymous.4open.science/r/PFA-review-088D.

0.30

ARBITRA: Robust Document-Level ATT&CK Extraction via Multi-Expert Argumentation

cs.CL cs.CR ·2026-10-29 ·Youngjin Jin, Seungwon Shin
Abstract · 13 pages of main paper, 21 pages total, preprint

Cyber threat intelligence (CTI) reports describe adversary behavior in natural language, but converting them into structured MITRE ATT&CK mappings remains difficult: existing methods decompose reports to sentence-level granularity for extraction, while modern LLMs can work with full reports in a single pass but with noticeable performance degradation. We present ARBITRA, which orchestrates multiple role-specialized LLM experts that propose candidate ATT&CK entities from the full report, which are then verified through a per-candidate quantitative bipolar argumentation framework (QBAF) grounded in document evidence and rationales. We additionally construct reasoning-specialized corpora for fine-tuning so experts can act as effective participants during this verification process. Across ATT&CK extraction benchmarks, ARBITRA reaches a stronger precision-oriented operating point than sentence-level pipelines, recent LLM-based extraction systems, and frontier models including GPT-5.5. Under adversarial perturbations simulating conflicting TTP statements, cross-report contamination, and semantic poisoning, ARBITRA limits mean precision degradation to roughly half that of frontier-model baselines for a more robust extraction.

0.30

L3G-World: Real-Time Infinite World Generation with Latent 3D Gaussian Memory

cs.CV ·2026-10-29 ·Jinhyeok Choi, JoungBin Lee, Jongmin Lee, Minkyeong Jeon, Jiyoung Kim +3 more
code/datacommunity pick
Abstract · Project page: https://cvlab-kaist.github.io/L3G-World/

Interactive world models generate visual content conditioned on user actions, yet maintaining spatial consistency over long-horizon exploration in real time remains challenging. Existing approaches rely on either temporal context or explicit 3D memory, both of which scale poorly because memory grows with the length of the generated stream. Moreover, explicit 3D memories condition latent video generators by re-encoding RGB renderings, which precludes real-time generation. In this paper, we introduce a Latent 3D Gaussian Memory (L3G) that combines the efficiency of feed-forward reconstruction with a compact, rapidly renderable scene representation. Building on this memory, we propose L3G-World, in which a memory updater and a video diffusion transformer (DiT) alternate autoregressively to generate spatially consistent worlds. The memory updater spawns Gaussians in newly observed regions and refines existing ones, so that the memory grows with scene extent rather than stream length. Each Gaussian further stores a VAE latent feature, allowing the memory to be rendered directly into the DiT's latent space without VAE re-encoding. Experiments show that L3G-World achieves state-of-the-art generation quality and long-horizon spatial consistency, with up to $38.6\times$ fewer 3D primitives, $3.2\times$ faster memory updates, and $359\times$ faster memory rendering relative to the most efficient prior 3D-memory baseline for each metric, enabling real-time world generation at 9 FPS.

0.30

PathoJev: Efficient Pathology Decisions from Shared Visual Context

cs.CV ·2026-10-29 ·Chengyang Zhang, Wenchuan Zhang, Bo Li, Xinyu Liu, Jiaming Yang +6 more
code/data
Abstract

Pathology workflows require multiple visual judgments about the same slide, from identifying tumor type to assessing grade and invasion. Generating and parsing a separate textual response for each structured question adds latency between visual evidence and usable decisions. Building on the System One paradigm, we introduce PathoJev, a pathology model that directly predicts answer probabilities for whole-slide and region-level assessment. PathoJev encodes each slide once and evaluates independent question suffixes in batches over the shared visual context. Reusing the hybrid backbone's recurrent states further limits the computation that grows with slide-context length for each additional question. Across three whole-slide benchmarks, the 9B and 27B models achieve mean accuracies of 77.2% and 78.1%, exceeding SlideChat by 5.4 and 6.3 percentage points. Both scales also maintain competitive accuracy on five ROI benchmarks. On WSI-Bench, the 9B model runs 9.1x faster than SlideChat. Shared context accelerates eight-question workloads by 6.9x and 5.9x over separate prefix encoding, rising to 11.2x and 8.9x for 64 questions. These results provide a reusable probabilistic interface for combining multiple visual judgments in whole-slide and local tissue assessment. Code: https://github.com/ChyaZhang/PathoJev.

0.30

Youtu-Parsing-Omni: One Encoder, One Schema, All Modalities

cs.CV ·2026-10-29 ·Yunhang Shen, Zhangquan Chen, Xiaotian Li, Zhongpeng Cai, Qing Zhou +25 more
code/data
Abstract

We present Youtu-Parsing-Omni, an omni-modal parser that converts documents, images, audio, and video into structured records grounded in space and time, providing a perception layer for agents that read, locate, and cite evidence across modalities. Its design rests on one encoder for every modality and one schema for every task. Youtu-Omni-Encoder, initialized from a pre-trained language model to strengthen text reading, encodes all modalities with shared weights in a unified space-time coordinate system, and a few fusion layers let video frames attend to concurrent audio, bringing cross-modal fusion and temporal alignment into perception itself. OmniSchema unifies the outputs of all ten parsing tasks, allowing a single self-evolving post-training procedure, OmniSchema-Aware On-Policy Distillation (OSAD), to provide a shared token-role-based objective across tasks. In each round, the model is trained on its own rollouts against token-level targets from a reference-conditioned copy of itself, which is then refreshed with the trained weights, so that the parser and its supervision co-evolve. With a single 5 B-parameter checkpoint, Youtu-Parsing-Omni achieves state-of-the-art performance on OmniDocBench v1.6 (96.96 Overall) and leads all open-weight models on OmniParsingBench (75.08 Avg.), with its advantage concentrated on audio and video. It also achieves the best overall chart-parsing score (95.02) and the highest exact chemical-fingerprint agreement among compared systems, and matches a dedicated optical-music model in character-level transcription of typeset sheet music while staying far ahead of general-purpose systems. Code and model weights are publicly available.

0.30

Read, Allocate, Route: Role-Coupled Tokenization for Pixel-Space Generative Transformers

cs.CV cs.AIcs.LGcs.NE ·2026-10-29 ·Jiongfan Yang, Weichen Fan, Haiwen Diao, Zhongang Cai, Ziwei Liu
community pick
Abstract · 22 pages, 10 figures; under review at ICLR 2027

Pixel-space generative Transformers typically use one token per image patch, so capturing finer detail requires more tokens and makes global attention increasingly expensive. We observe that finer detail does not necessarily require more global communication: low-frequency structure benefits from global exchange, while high-frequency detail can be modeled locally. Based on this insight, we introduce SpecSlot, a pixel-space rectified-flow Transformer that decouples output detail from the length of the global attention sequence. SpecSlot expands each spatial anchor into multiple slots, each assigned a disjoint band of DCT coefficients; this separates anchor density from spectral subdivision. Only the lowest-frequency slot from each anchor participates in global attention, while all slots interact locally. This allows the model to increase spectral detail without enlarging the global sequence. A spatial head further adds a local correction to the spectral prediction, while the generative state, target, and loss remain in pixel space. On ImageNet-256, SpecSlot achieves a comparable final FID to a FLOPs-matched JiT control while converging faster. On ImageNet-512, it improves FID by 2.84 over the control with 3.9% fewer FLOPs, and comes within only 0.69 FID of JiT-B/16 at 38.2% lower cost. On text-to-image generation, SpecSlot surpasses JiT-B/8 on DPG-Bench by 1.06 at 40.6% lower forward cost.

0.29

Semantic Mismapping in Generalized Category Discovery: Evaluation and Mitigation

cs.CV cs.AIcs.LG ·2026-10-29 ·Zitai Wang, Qianqian Xu, Cong Hua, Xi Chen, Zhiyong Yang +2 more
theory
Abstract

Generalized Category Discovery (GCD) aims to perform old-category recognition and novel-category discovery in the same unlabeled set. Early methods address both objectives through clustering and evaluate the exchangeable predicted indices with global Hungarian (GH) matching. Recent methods adopt trainable parametric heads, using supervision for old-category recognition and discovery-oriented objectives for new categories. In this way, the output space becomes partially exchangeable: old indices are semantically anchored, while new indices remain exchangeable. However, these methods still apply global matching, which can reassign anchored old indices to other categories and hide some prediction errors. Empirically, we are surprised to observe that this reassignment can induce an overestimation of new-category accuracy by up to 25.63 points. We term this phenomenon semantic mismapping and further derive a permutation-cycle condition that characterizes when semantic mismapping can increase the global matching score. To fix this issue, we propose Semantics-Preserving Evaluation (SPE), which preserves the supervised semantics of old indices and restricts Hungarian matching to the new region. Furthermore, we propose Mismapping-Mitigating Optimization (MMO) to alleviate the prediction errors that can give rise to semantic mismapping, combining a region-balance loss (RBL) for old/new routing with a novel-diversity loss (NDL) that encourages orthogonality among new prototypes. Theoretically, we show that minimizing the two losses can tighten an upper bound on spurious accuracy gains from semantic mismapping, with controllable residual terms. Empirical results under both GH and SPE demonstrate the effectiveness of our proposed method.

0.29

RoFiGR: Coarse Routing, Fine Refinement via Level-Adaptive Distillation in Multimodal Generative Recommendation

cs.IR ·2026-10-29 ·Ji Dai, Zhipeng Wei, Zhuang Zhuang, Rongfeng Guo, Jie Chen +2 more
hot direction
Abstract · 12 pages, 10 figures, 10 tables

Generative recommenders predict an item's semantic identifier (SID) one code at a time. Early codes select broad branches, so a routing error can exclude relevant items before fine-grained prediction begins. In multimodal recommendation, these coarse decisions must reflect collaborative preferences as well as item semantics. We introduce RoFiGR, a two-tree recommender that uses collaborative information in both SID construction and generation. It aligns cumulative L1-L2 code vectors with frozen collaborative item embeddings and explicitly encodes item distance and source level in the serialized history. During generation, a sequential teacher supervises legal branches. Fixed coarse-level weights maintain routing supervision, while a bounded adaptive residual allocates additional weight according to teacher-student demand. On Instruments, Arts, and Games, RoFiGR improves all five ranking metrics over the evaluated external baselines. Five-seed paired tests with Holm correction support the main NDCG@10 improvements. Comparisons with alternative collaborative supervision objectives and tokenizers support the proposed design. Matched task and decoding analyses distinguish the effects of training exposure, stream fusion, and candidate budgets.

0.29

Agent-Wise Test-Time Normalization for Camera-Based Cooperative Perception with Corrupted Collaborators

cs.CV cs.AI ·2026-10-29 ·Yu Wang, Shaohua Wang, Yicheng Li, Kecheng Sun
Abstract · 11 pages, 8 figures

Intermediate-fusion cooperative perception lets a vehicle detect objects beyond its own field of view by fusing features shared by nearby agents. When only a collaborator's cameras are degraded by fog, motion blur or exposure errors, the features it shares drift away from the training distribution while the ego input stays clean, and cooperation can become worse than not cooperating at all. Test-time batch-normalization adaptation, the standard training-free remedy, estimates statistics over the pooled batch of all agents and therefore normalizes both the clean ego and the corrupted collaborator with a mixture of two distributions. We show that this mixing harms even the clean ego: when the ego shares BN statistics with a corrupted collaborator, its detection from its own features drops by 3.58 to 6.68 AP@0.5 points although its cameras are untouched. We propose AgentNorm, an agent-wise test-time normalization that i) estimates the normalization statistics of every collaborator from its own current observation and ii) anchors the ego to its source statistics, so that the ego path remains exactly the original model. AgentNorm has no learnable parameters, needs no gradients or labels, is independent of the order of test frames, and runs locally on each agent without changing the transmitted features. On camera-based OPV2V with four fusion checkpoints and collaborator-only corruptions, AgentNorm raises AP@0.5 under severity-5 fog from 29.42-37.25 to 43.32-58.86 and outperforms pooled normalization by 5.53-12.94 points under fog and motion blur, at an extra latency of 2.3 ms per frame. On the real-world DAIR-V2X-C dataset, it outperforms pooled normalization in all ten infrastructure-corruption settings. We also report the cases it does not fix: occlusion, which removes image content rather than shifting feature statistics, and the case in which all agents share the same shift.

0.29

Spectral Invariant Fusion and Refinement for Domain Generalized Semantic Segmentation

cs.CV cs.AI ·2026-10-29 ·Yansong Bai, Yunzhi Zhuge, Xinzhuo Yu, Lu Zhang, Jin Zhan +2 more
top venue
Abstract

Domain generalized semantic segmentation is challenging because it requires accurate dense prediction on unseen domains without any access to target-domain data during training. Existing methods improve transferability through data diversification, invariant learning, or direct use of foundation models, but they often struggle to simultaneously suppress domain-sensitive interference and preserve fine-grained spatial cues. In this work, we propose SIFR, a frequency-aware dual-stream framework that tackles domain shift at both the fusion and decoding stages. SIFR uses spectral modeling as a unified principle to regulate both heterogeneous prior interaction and dense decoder refinement. Instead of directly fusing frozen CLIP and DINOv2 features, SIFR performs spectral-guided invariant fusion to filter out domain-sensitive responses while preserving complementary semantic and structural information. In addition, SIFR incorporates a spectral decoder refinement module that stabilizes multi-scale decoder features through low-frequency statistics-based normalization and adaptive frequency selection, thereby reducing the propagation of residual domain bias in dense prediction. This design enables SIFR to jointly improve domain robustness and segmentation precision under severe distribution shifts. Extensive experiments on multiple DGSS benchmarks, including synthetic-to-real, real-to-real, and adverse-weather settings, demonstrate consistent improvements over previous state-of-the-art methods.

0.29

Don't Lose the Thread: Empowering Long-Horizon LLM Agents with Cognitive Resource Self-Allocation

cs.CL ·2026-10-29 ·Zheng Zhu, Xiaozhuan Liang, Xi Chen, Bei Yu, Jiaya Jia
publishedbenchmarkcode/data
Abstract · Accepted at COLM 2026. Code is available at https://github.com/JIA-Lab-research/CORAL

Agents powered by large language models (LLMs) have demonstrated remarkable progress in solving complex reasoning tasks. However, LLM agents often falter on long-horizon tasks due to cognitive overload, as their working memory becomes cluttered with expanding and irrelevant information, which dilutes their attention and hinders effective planning and reasoning. To mitigate this challenge, we introduce Cognitive Resource Self-Allocation (CORAL), a novel reasoning paradigm that empowers agents to proactively optimize their context. Implemented as an agent-callable working memory management toolset, CORAL allows an agent to checkpoint its task progress and verified facts, and adaptively initiate a new problem-solving episode by resetting cluttered context while preserving only the checkpointed information, enabling it to resume reasoning from a clean state. We further enhance the agent's checkpointing capabilities using a Multi-episode Agentic Reinforced Policy Optimization algorithm. On the GAIA benchmark, CORAL with an 8B model achieves 42.8% average accuracy, with particularly strong gains on the most challenging long-horizon tasks (Level 2 and Level 3). Crucially, these gains come at a substantially lower inference cost: CORAL consumes 6-7$\times$ fewer input tokens than a ReAct agent. It also generalizes to multi-hop Knowledge QA benchmarks and to the considerably harder BrowseComp, where it improves a strong proprietary agent by 8.0 points.

0.29

DEQ-Stereo: Deep Euilibrium Video Stereo Matching via Adaptive Convergence

cs.CV ·2026-10-29 ·Kiljoon Han, Kyumin Hwang, Wonhyeok Choi, Jeonghoon Kim, Sunghoom Im
published
Abstract · Accepted at ACCV2026

Iterative refinement models have significantly improved stereo matching accuracy, but their rigid, fixed update schedules incur heavy computational costs and a strict trade-off between speed and accuracy. This challenge is magnified in video streams, where traditional online methods resort to reducing feature channels or relying on expensive temporal warping. To address this, we propose DEQ-Stereo, a novel video stereo matching framework based on the Deep Equilibrium (DEQ) paradigm. Instead of unrolling a fixed number of iterations, DEQ-Stereo reformulates the refinement as an implicit fixed-point problem, adaptively terminating once convergence is met. To ensure fast and stable intra-frame convergence, we introduce the Convolutional Simplified Gated Unit (ConvSGU), which eliminates unnecessary non-linearities by modeling the latent state as a gated, element-wise sum of inputs. Furthermore, to exploit cross-frame temporal coherence, we propose Temporal State Transfer (TST), a warm-start strategy that propagates the converged hidden state to the next frame without requiring explicit pose or warping priors. Extensive evaluations demonstrate that DEQ-Stereo delivers state-of-the-art performance with a superior accuracy-to-latency trade-off.

0.29

Try Before You Evolve: Ranking-Scored Trials for Recursive Self-Improvement of Recommender Systems

cs.IR ·2026-10-29 ·Xu Huang, Jiawei Feng, Jiancan Wu
Abstract

Recommenders on Web platforms such as e-commerce and music streaming need their training recipes re-tuned for each dataset. Large language model (LLM) agents can now rewrite these recipes, enabling recursive self-improvement (RSI) of recommender systems. Existing agents propose one edit per round, fully retrain the recommender, and keep the edit only if its ranking quality improves. Unlike programs or patches that can be scored by execution, a recommender edit changes how the model is trained, which raises two challenges. First, ranking feedback is delayed: an edit's ranking quality is known only after full retraining, so much of the retraining budget goes to edits that are later discarded. Second, the signals available before retraining are unreliable: sampled training losses can look solved while full-ranking performance remains poor, and they are not comparable across edits that change the loss or sampler. We propose RecRSI, which generates diverse candidate edits from a diagnosis of the current model and selects one for full retraining using early ranking feedback: each candidate receives a short trial, the first steps of its full training run, scored by full-ranking NDCG@10 on held-out probe users. With two sequential recommenders on three datasets and the same number of full retrainings, RecRSI significantly outperforms existing self-evolving agents in every setting, and on both Amazon datasets reaches in its first round a ranking quality that none of them reaches in five.

0.29

Kinematic Causal Attribution: Black-Box Explanations of Motion versus Geometry for 4D Point-Cloud Video Models

cs.CV ·2026-10-29 ·Zerui Wang
Abstract · 27 pages, 6 figures, 14 tables

4D perception models, which classify 3D point-cloud videos, are being built into robots, assistive monitoring and AR, yet before deploying one a practitioner cannot check which cue it relies on: how a body part moves, or what shape it has. 3D attribution is single-frame, video attribution is confined to RGB pixels, gradient maps need white-box access, attention is absent from state-space backbones, and point-level perturbations assume a point correspondence across frames that real depth and LiDAR data lack. We present Kinematic Causal Attribution (KCA), a black-box, interventional attribution method that queries only the model's outputs. Its freeze operator removes a region's motion while keeping its shape, and a shape-free proxy removes the shape while keeping the motion. Together they span a motion$\times$geometry cube for every tracked region, whose order-symmetric Shapley decomposition is complete and whose interaction is measured rather than assumed away. The operators rewrite per-frame point sets, so they need no correspondences, and a context-averaged estimator handles structures that span several regions. We validate KCA on the official PSTNet, P4Transformer and Mamba4D trained on MSR-Action3D. The decisive test is an audit: two binary tasks are built from real clips whose classes differ only in the motion or only in the shape of the same arm, and the motion share of the arm's importance follows the defining cue on all three backbones (0.83 vs. 0.60 on P4Transformer, 0.86 vs. 0.30 on PSTNet, 0.92 vs. 0.61 on Mamba4D), a distinction no single attribution map can make. The split costs no faithfulness: the necessity maps have the lowest deletion AUC on P4Transformer and PSTNet, tie with the best baseline on Mamba4D, and lose their correlation with the trained model's once the weights are randomised. Skeleton ground truth, re-sampling stability and cost complete the validation.

0.29

Single-Annotation, Multi-Hypothesis Segmentation with Semantic-Spatial Residual Banks

cs.CV cs.LG ·2026-10-29 ·Armin Dadras, Robert Sablatnig, Franziska Proksa, Markus Seidl
published
Abstract · Accepted at WACV 2027

Multi-rater segmentation datasets show that a single image can admit several valid masks, yet most training sets expose only one annotation per image. We study single-annotation multi-hypothesis segmentation: each training image exposes one fixed mask, while the model must return a finite bank that covers held-out human annotations at test time. Our experiments reveal a representation tradeoff. Banks learned directly from images recover variations in local boundaries and extent but lose structural hypotheses, whereas frozen semantic features preserve structural grouping but underrepresent spatial variation. We introduce the Semantic-Spatial Residual Bank (SSRB), which freezes a semantic bank trained with balanced assignments and learns a residual from the image for each hypothesis. Across five medical endpoints, SSRB improves coverage of held-out masks at $K=8$ over a Probabilistic U-Net under the same one-mask constraint. On LIDC and LIDC-hard, it also produces substantially more distinct useful masks. On BSDS, it preserves the semantic bank's performance in grouping regions. Its advantage grows as more masks are returned, showing that residuals learned for corresponding heads improve coverage of local annotation variation while preserving how regions are grouped.

0.29

Fight World Model: Customizable Combat through Visual Interaction Proxies

cs.CV ·2026-10-29 ·Jiahua Li, Zixu John, Tom Zhong, Fuping Wu, Tianhao Xu +3 more
Abstract · Tech Report

A playable, programmable game world must maintain persistent state, enforce executable rules, and support extensible gameplay mechanics. Existing video world models can generate plausible combat, yet struggle to keep hit reactions and health updates consistent with game rules. We introduce Fight World Model (FWM), which separates rule-governed combat execution from video rendering. An executable engine maintains state and resolves interactions. A deterministic compiler encodes these outcomes and attempted actions in a spatially and temporally aligned interaction proxy. The proxy aligns action and skill windows, brief hit reactions, and persistent health by fighter identity and timing. These cues can be controlled separately to render interactions under modified rules. Identity-bound references preserve character appearance across side switches. We train FWM on FightWorld Corpus, spanning 52 skill effects, 187 character-appearance variants, and 15 arenas. Without retraining the renderer, FWM supports damage scaling, limited-hit immunity, lifesteal, and multi-stage skill mechanisms beyond the native game. It also allows character and arena customization. Experiments demonstrate the effectiveness of FWM, which outperforms existing game world models in action fidelity, hit reactions, health-state consistency, and character appearance preservation. Our project is available at https://fight-world-model.github.io.

0.29

ScienceJudge: Pattern Perturbation Augmentation for Sentence-Level Scientific Claim Verification

cs.CL ·2026-10-29 ·Chuhai Cai, Xin Chen, Min Yang, Feng Jiang
benchmark
Abstract · 13 pages, 1 figure; NLPCC 2026 Shared Task 10 (Track 1)

Sentence-level scientific claim verification determines whether each sentence in an AI-generated scientific paragraph is supported by a given multimodal evidence bundle. Fine-grained reporting errors often preserve fluent and plausible statements while introducing subtle evidence-claim deviations, making them difficult to detect under sparse and imbalanced supervision. We propose Pattern Perturbation Augmentation (PPA), which models reporting errors as label-specific faithfulness breaking transformations and induces reusable error patterns from labeled examples. PPA applies these patterns to supported claims to construct controlled hard negatives and expands their coverage through pattern-level, operation-level, and cross-domain augmentation. Based on these augmented views, we develop ScienceJudge, which integrates PPA-trained vision-language predictors and instruction-guided LLM predictors through class-weighted ensemble inference. On the local development set, ScienceJudge improves the average score from 0.8124 to 0.8692, achieving a Macro-F1 of 0.8466 and a Paragraph Exact Match score of 0.8919. Our submitted system ranks first on the official NLPCC 2026 Task 10 Track 1 leaderboard with a final score of 0.6890.

0.29

VideoMemBench: How Well Do MLLMs Remember in Video Conversations?

cs.CV ·2026-10-29 ·Zhengdong Hu, Chao Wang, Kaituo Feng, Huanjin Yao, Mingfei Lu +2 more
code/datacommunity pick
Abstract

Multimodal large language models (MLLMs) can now process hour-long videos, making extended multi-turn dialogue over a single video a natural mode of use. Sustaining such a dialogue requires more than answering each question in isolation: the model must retrieve and reuse information established earlier in the conversation, both what it has seen and what it has said, an ability we term in-context video memory. Existing video QA benchmarks leave this ability unmeasured: their questions are mutually independent, never reference one another, and each is answerable from the video alone, so nothing ever requires carrying information across turns. We introduce VideoMemBench, in which each sample pairs a single video with a long multi-turn dialogue, whose turns are connected by typed cross-turn dependencies: an Episodic Dependency refers back to a visual episode described in an earlier turn, probing memory of what the model has seen, while an Answer Dependency builds on the model's own earlier answer, probing memory of what the model has said. Every dependency is verified considering two properties: 1) necessity, meaning the question cannot be answered without its earlier turns, and 2) non-leakage, meaning the question does not restate the information from its earlier turns. To enforce these properties, we design a construction pipeline with three LLM roles: a planner, a generator, and a judger, with each dialogue reviewed by human experts. VideoMemBench comprises 326 videos with 2,476 turns, of which 1,387 depend on earlier turns, including multi-parent dependencies and long-range call-back. Among over 20 evaluated open- and closed-source MLLMs, Claude Opus 5 performs the best with an overall score of only 45.7%, highlighting the challenge of in-context video memory. Our dataset and evaluation are released.

0.29

A Low-Rank Space Captures Linear Alignment Between LLM and Human Neural Activity During Naturalistic Language Listening

cs.CL q-bio.NC ·2026-10-29 ·Zach Paris, Emily Finn
code/data
Abstract · 23 pages, 12 figures. Code: https://github.com/thefinnlab/low-rank-alignment

Large language models (LLMs) are increasingly used as computational models of human language processing, yet it remains unclear which components of their representations drive alignment with neural activity. Here, we use ridge reduced-rank regression to identify the brain-predictive subspace of LLM representations: the representational directions that support linear prediction of neural responses. Across three naturalistic language-listening datasets spanning intracranial electrophysiology and fMRI, and six model conditions spanning autoregressive, bidirectional masked, and randomly initialized language models, we find that full linear encoding performance can be recovered using only 1-3% of the original embedding dimensionality. These low-dimensional subspaces generalize across held-out participants and, to a lesser extent, across datasets, suggesting that neural predictivity is concentrated in a compact and partially shared representational space. We next compare these brain-predictive directions with the principal axes of variation in the embeddings. Principal component analysis requires substantially more dimensions to recover neural prediction, demonstrating that the directions most useful for predicting neural activity are distinct from those capturing the most overall variance within LLM representations themselves. Finally, characterizing the learned subspaces allows us to surface interpretable features that are overrepresented in neural activity above and beyond the LLM representations. Together, these results show that LLM-brain alignment is supported by a small, structured subset of model representations and provides a framework for better understanding exactly why these representations align. Code can be found at: https://github.com/thefinnlab/low-rank-alignment

0.28

Decision and System One models in rheumatology: accuracy and calibration on specialty-training examination questions and diagnostic case vignettes

cs.CL cs.AI ·2026-10-29 ·Alfredo Madrid-García, Diego Benavent, Fabian Lechner, Phillip Kremer, Johannes Knitza +1 more
Abstract · 28 pages, 2 figures, 2 tables; includes supplementary material (8 tables, 1 figure)

Objectives: To evaluate the accuracy and calibration of Decision and System One models on rheumatology examination questions and diagnostic vignettes. Methods: Two benchmarks were analysed: 172 rheumatology questions from Spanish specialty-training examinations (expanded RheumaMIR dataset) in Spanish and English, and 70 English diagnostic vignettes from RheumBench with 5-18 candidate diagnoses, 52 of which included the reference diagnosis. Four System One models (Jev 1.13, Clef, Microsoft-Decision-1, Laya) and GPT-6 Luna (OpenAI Decisions API) were compared with GPT-6 Sol and two local generative models. The primary outcome was accuracy; secondary outcomes included calibration, discrimination, accuracy of answers with probability >=0.90 and language differences. Results: In RheumaMIR, Jev's accuracy was 93.0% in Spanish and 91.9% in English, below GPT-6 Sol (99.4%; Holm-adjusted P<=0.005); GPT-6 Luna reached 97.7%, Clef 91.9-93.0% and Microsoft-Decision-1 86.0-87.8%, without detectable language differences. Jev's answers with probability >=0.90 (76-77% of answers) were 98.5% correct. In RheumBench, accuracy was lower: GPT-6 Sol 76.9%, GPT-6 Luna 63.5%, Clef 59.6%, Jev 50.0% and Microsoft-Decision-1 40.4% (each System One model below GPT-6 Sol, Holm-adjusted P<=0.045; chance 9.6%). Jev became overconfident (mean probability 0.659; over/underconfidence index +0.159), and its answers with probability >=0.90 (25% of cases) were 84.6% correct, against 95.5% for GPT-6 Luna (42%). Laya performed near chance in both benchmarks. Conclusion: The best-performing System One models were highly accurate on examination questions but less accurate on diagnostic vignettes, remaining below GPT-6 Sol. Jev's high-confidence predictions were less reliable on vignettes, highlighting the need for task-specific evaluation and calibration.

0.28

KV Cache Quantization for Diffusion Language Models

cs.CL ·2026-10-29 ·Yan Tai, Yunan Ding, Guangzhao Li, Siqi Luo, Chaoyang Zhang +3 more
top venue
Abstract

KV cache quantization is well studied for autoregressive language models, where cached prefix states are exact under causal attention and quantization is mainly a cache reconstruction problem. Diffusion language models (dLLMs) break this assumption. Fast dLLM inference reuses KV states across denoising steps, but bidirectional token representations can change as masked tokens are revealed, so the reused cache may already differ from full recomputation before quantization is applied. We study KV cache quantization under this cache-reuse error. We adapt KIVI and KVQuant to Fast-dLLM-style prefix and dual cache reuse, and introduce TRACE and dLLM-KVQ. TRACE reconstructs feasible dLLM reveal trajectories from prompt-answer pairs, providing calibration states that match iterative masked decoding. dLLM-KVQ uses these trajectories to choose per-layer K-cache entries that remain in high precision according to output- and attention-level behavior metrics, while applying standard low-bit value quantization. Across LLaDA and Dream evaluations on GSM8K, HumanEval, MATH, and MBPP, dLLM-KVQ achieves the highest average score among quantized methods across the reported 4-bit and 2-bit prefix/dual cache settings.

0.28

WorldFlow: Joint Video and 3D Motion Prediction via Large-Scale Pretraining

cs.CV cs.AIcs.RO ·2026-10-29 ·Yuhao Lu, Jiongxu Chen, Hexiang Wei, Jiajin Wu, Juelin Li +7 more
code/data
Abstract

Predicting future physical states stands as a central challenge in embodied intelligence. Existing video world models leave 3D dynamics implicit, action-based formulations remain tied to specific robot configurations, and 3D motion prediction models use flow-only supervision, providing a limited learning signal for physical dynamics. To circumvent these limitations, we introduce \method{}, a generative world model that jointly predicts future video and query-based 3D trajectories. By parameterizing 3D motion as continuous displacement tokens aligned with the video grid, a unified Mixture-of-Transformers (MoT) backbone jointly denoises both modalities under a rectified-flow objective. This formulation couples explicit geometric dynamics with visual evolution, effectively unlocking the physical commonsense embedded within pretrained video models. To empirically scale and validate this unified architecture, we train \method{} on a heterogeneous mixture of 1.44 million object-level samples spanning simulated and real-world domains. We further introduce WorldFlowBench, comprising 1,678 evaluation samples from 11 data sources across simulated and real-world interactions, including spatial reasoning tasks. Across all three evaluation sets, WorldFlow achieves the lowest average displacement error (ADE) and final displacement error (FDE) among the evaluated learning-based methods. Notably, on the challenging SpatialCog set, WorldFlow achieves an ADE of 0.060 m, a 54.2\% reduction compared with the strongest external baseline. It also reduces ADE and FDE by 55.9\% and 68.5\%, respectively, compared with its flow-only variant. Project page: https://worldflow-rbs.github.io/.

0.28

Annotation-Free Representation Learning for Polyp Retrieval in Colonoscopy Videos

cs.CV ·2026-10-29 ·Luca Parolari, Lamberto Ballan, Carlo Biffi, Loic Le Folgoc
code/data
Abstract · Under submission at ISBI 2027

Colonoscopy is inherently comparative: reliable visual assessment of polyps often depends on complementary views acquired under different viewpoints and imaging conditions. Retrieving such views requires polyp tracklet representations, which recent contrastive methods learn from temporal cues but still build from manually annotated bounding boxes. We propose an annotation-free framework that constructs tracklet views directly from raw colonoscopy videos, enabling temporal self-supervised learning on full-procedure videos, including frames without visible polyps. On REAL-Colon, our method achieves state-of-the-art results on tracklet retrieval and re-identification, outperforming both self-supervised and supervised baselines. We then introduce few-shot polyp retrieval, a task closer to clinical use, where a few observations of a target polyp are used to retrieve its remaining tracklets: with only two examples, our method surpasses most competing methods given ten. Finally, we show that a modest number of unlabeled procedures suffices for competitive performance, making on-site model development more practical. Code is available at https://github.com/lparolari/freerep.

0.28

Probing Spatial State Representations in Visual Foundation Models

cs.CV cs.AI ·2026-10-29 ·Baiqiao Yin, Zhongang Cai, Ruisi Wang, Qiwei Liang, Lei Yang +1 more
community pick
Abstract

An observer sees only part of the surrounding world, yet must keep track of what lies beyond the current view. Spatial perception therefore involves more than recognizing individual images: it calls for a persistent spatial state that is updated under self-motion and used to guide action. Do visual foundation models provide representations that support such a state? We study this question through four levels of spatial capability---spatial structure, object persistence, predictive spatial update, and actionability---using frozen-feature probes on ten representations under streaming egocentric observations of static indoor environments. The resulting capability profiles vary across tasks, network layers, and prefix lengths: strong readout of the visible scene does not ensure accurate object persistence or future-feature retrieval. Matched controls separate readout accuracy from reliance on visual history and motion. Native future prediction retains recoverable spatial structure, but a spatial reader trained on real features does not transfer to predicted features unchanged. Finally, goal-dependent candidate selection does not establish reliable continuous action utility: predicted actions respond to goal changes, yet supplying the correct goal does not reliably improve progress. Across the four levels, object persistence and actionability build on camera-pose readout, whereas feature prediction builds on visible-scene readout. Together, these findings motivate evaluating spatial representations not only by what can be decoded from them, but by whether that information is maintained beyond observation, updated predictively, and used for goal-directed action.

0.28

Neural Blind Deconvolution for Image Scanning Microscopy

cs.CV ·2026-10-29 ·Eliot Guez, Nathan Buskulic, Lisa Cuneo, Luca Calatroni, Giuseppe Vicidomini
Abstract · 4 pages, 2 figures

Image scanning microscopy (ISM) acquires one scanned image per detector element, which are then fused into a single image by multi-image deconvolution to obtain a super-resolved image. Its accuracy, however, depends on point spread function (PSF) models that sample- and system-induced aberrations often invalidate. Since differences between the scanned images encode aberration information, we propose a training-free blind Deep Image Prior (DIP) framework that jointly estimates the specimen and the detector-dependent PSFs. The PSFs are parametrised through the detector geometry and shared excitation and emission PSFs, reducing the unknowns while preserving the physical ISM model, and a Gaussian pre-initialisation stabilises the blind optimisation. On experimental tubulin data, blind DIP recovers finer filamentous structures than multi-image deconvolution and DIP with fixed theoretical PSFs, while adapting the PSFs to the measurements.

0.27

Testimony from Timelines: Evidence-Grounded Significance Testing of Temporal Clinical Hypotheses

cs.CL cs.IR ·2026-10-29 ·Chengyang He, Wenlong Zhang, Koduvayur Subbalakshmi, Ping Wang
benchmark
Abstract

Ask a language model whether sales tend to dip after a product recall, or whether one side effect tends to follow another after a vaccine, and it answers in a fluent sentence that reads like a finding, with no test behind it. Assertions about how events unfold across many cases fill the Web, and each is a temporal hypothesis that only a significance test over the right records can settle. The verifier must specify the null, select the test and decide which records bear on the hypothesis, and any of these choices can go wrong while the conclusion still comes out right. Yet benchmarks for language-model data analysis grade only the conclusion, with the relevant data handed over. We study the problem in health, where a wrong verdict is costly, and present STRATA, a benchmark that grades every decision on the way to a verdict: 8,735 hypotheses over the clinical timelines of vaccine adverse-event reports, with gold plans, evidence sets, p-values and verdicts derived under a released codebook and reproduced by a deterministic oracle. We also propose ART (articulate, retrieve, test), a staged paradigm that exposes each decision as an inspectable output. Extensive experiments with five model configurations show that systems fail mostly at choosing which records count, a verdict-evidence gap that accuracy alone conceals. ART closes most of this gap and brings a 31B open model to within 0.01 of the frontier models under the same paradigm; the frontier models succeed once given the codebook.

0.27

Edited Endpoints as Partial Goals for Image-to-Video Generation

cs.CV ·2026-10-29 ·Sihyun Kim, Sanghyun Jo, Kyungsu Kim
code/data
Abstract · 34 pages, 8 figures, 28 tables. Project page: https://wannagetacat.github.io/Edited-Endpoints-I2V/

Edited endpoint images make the target of an image-to-video transition explicit and can improve transition completion, but also preserve context in the source image's layout. Full-frame endpoint conditioning turns this preserved context into requirements on the video's final state, potentially inducing temporal artifacts. We measure discontinuity with Hard-Cut (HC), late background reversal with Background Motion Reversal (BMR), and motion suppression with Dynamic Degree (DD). These artifacts appear across five video models spanning training-free and learned endpoint conditioning, in model-specific forms. We study Partial, a selective conditioning scheme that constrains only an edit-relevant region estimated from image-editor attention and leaves the remaining terminal content to the video model. Across three backbones on TC-Bench-I2V, Partial reduces temporal artifacts relative to full-frame conditioning, with no significant difference in Transition Completion Ratio (TCR); on Wan2.2-Turbo, uniformly weakening conditioning does not reproduce these reductions. A donor-context intervention keeps full-frame conditioning and replaces only the context outside the edit-relevant region with the final frame of a Partial video. It lowers HC and BMR on HunyuanVideo-1.5 and lowers BMR while restoring DD on Wan2.2-Turbo, with no significant difference in TCR, so constraint extent alone does not account for these artifacts. Together, these findings suggest that edited endpoints should serve as partial future goals, although edit relevance does not guarantee future validity. Code and data will be released at https://github.com/wannagetacat/Edited-Endpoints-I2V.

0.27

Joint Rigid Alignment and Morphable Reconstruction from Partial 3D Face Scans

cs.CV ·2026-10-29 ·Eli A. Underwood, Nicolas Rondan, Angel V. Peterchev, J. Matias Di Martino
code/data
Abstract · 9 pages, 4 figures, 4 tables

Aligning a partial, real-world 3D face scan to a canonical representation requires solving two coupled problems at once: estimating the rigid pose that brings the scan into a common frame, and completing the facial shape in regions the scan never observed. Existing pipelines typically decouple these steps, running a generic point-cloud registration front end followed by a separate non-rigid fitting or deformation stage. Building on the GeoTransformer superpoint-matching architecture, we propose a single network that predicts both a rigid pose and a shape-complete point cloud in one forward pass: a cross-attention regressor, conditioned on transformer point features and predicted coarse correspondences, maps a partial scan to per-patch coefficients of a 3D morphable face model learned offline from a corpus of 8,000 identities. The resulting morphed template is then used, in the same forward pass, as the target for fine-grained Sinkhorn matching and rigid pose estimation, supervised purely by geometric alignment to ground-truth morphed geometry, so no hand-labeled coefficients are required. Against six classical and learned non-rigid registration baselines on held-out identities, our method leads the strongest ones by 3-4 times on per-vertex error and baselines that collapse the head onto the visible region by more than an order of magnitude, and is simultaneously the most accurate and fastest method on ground-truth-free scans from two independent real-scan datasets, NoW and NPHM. Code will be made publicly available.

0.27

M-PFT: Angular Margin for Calibrated Prompt-Tuning of Fine-Tuned Vision-Language Models

cs.CV ·2026-10-29 ·Shihab Aaqil Ahamed, Mohamed Ansaff, Nang Khin Myat Lay Phyu, Udaya Sampath K. Perera Miriya Thanthrige, Chamira U. S. Edussooriya +2 more
published
Abstract · Accepted at WACV 2027

Prompt-tuning has emerged as a parameter-efficient technique for adapting large-scale vision-language models (VLMs) to downstream tasks while keeping the pre-trained weights frozen. However, prompt-tuned VLMs often suffer from a base-to-novel dual miscalibration problem, where models typically show underconfidence on base classes and overconfidence on novel classes. Current VLM calibration approaches address this dual miscalibration either separately or jointly, but lack a geometric understanding of the hyperspherical embedding space and thus fail to exploit angular discriminative properties to adjust distorted decision boundaries. To address this, we propose M-PFT, a novel angular-margin approach for prompt tuning that introduces two complementary objectives to promote class discriminability and semantic anchoring. This enlarges the angular margin by 1) minimizing the maximum cosine similarity between tunable class embeddings to reduce embedding collapse and 2) minimizing the angular distance between each tunable embedding and its corresponding zero-shot embedding to preserve semantic geometry. Extensive experiments across 9 prompt-tuning methods with various backbones on 15 diverse datasets show that our approach reduces the average calibration error on novel classes, while remaining competitive on base classes and maintaining comparable accuracy. Further analyses reveal that M-PFT effectively adjusts distorted decision boundaries induced by prompt tuning while preserving the pre-trained embedding space, improving calibration with negligible computational overhead.

0.27

Radiology Guideline-derived Decision Graph for Efficient Clinical Decision Support

cs.CV cs.AI ·2026-10-29 ·Xuewen Zhou, Zhen Chen
Abstract

Radiological risk stratification and management planning are essential downstream tasks in medical imaging. Automating these tasks is challenging because clinical guideline criteria are distributed across prose, tables, and cross-referenced definitions. We present Guideline-derived Decision Graph (GDGraph), a framework that decouples guideline compilation from case evaluation. During compilation, a large language model partitions guidelines into section-based chunks, extracts local decision subgraphs with linked definitions, and aggregates related branches into an executable decision graph. During inference, a fast System One model answers localized clinical questions while deterministic traversal resolves Boolean logic, turning opaque single-pass predictions into a traceable decision path without task-specific training. Across 347 cases spanning renal cyst, adnexal mass, and thyroid nodule guidelines, GDGraph improves System One accuracy from 57.3% to 90.8%, outperforming retrieval-augmented GPT-4o (80.1%) and graph-guided GPT-4o with chain-of-thought (82.7%) at substantially lower latency and cost. The source code will be available.

0.27

Every Token Needs Its Own Teacher: Multi-teacher Adaptive On-policy Self-distillation

cs.CL ·2026-10-29 ·Yifu Luo, Chuang Ma, Zeyu Chen, Enkelejda Kasneci, Qi Chen +1 more
positioncode/data
Abstract · Preprint

On-policy Self-distillation (OPSD) has proven effective for post-training large language models by constructing a teacher conditioned on privileged references. However, existing approaches keep the teacher fixed either throughout training by consistently exposing it to the full reference, or within each response by adaptively varying partial reference exposure only across responses. In this paper, we argue that neither of them is sufficient. Instead, every token needs its own teacher, as the teacher preference can vary token by token within the same response. Through a teacher-intervention pilot study, we validate that the teacher preference differs across tokens, and reveal that a preferred teacher can be identified for each token through majority voting and KL-based selection. Motivated by these findings, we propose Multi-teacher Adaptive On-policy Self-Distillation (MTA-OPSD), the first token-level adaptive OPSD method. Our method constructs a multi-teacher pool consisting of candidate teachers, and enables every token to adaptively select its preferred teacher during training. Experiments with Qwen3 models on mathematical reasoning tasks show that MTA-OPSD consistently outperforms existing OPSD approaches with an average gain of up to $+5.9$ over the base model, highlighting the importance of adapting teacher supervision at the token level. The code is available at https://github.com/xingzhejun/MTA-OPSD.

0.27

Luggage re-identification: A study of strategies towards a real-world application

cs.CV cs.AI ·2026-10-29 ·Quentin Dhôte, Romaric Audigier
publishedbenchmark
Abstract · 7 pages, 4 figures, 5 tables. Accepted at The 2026 IEEE International Conference on Content-Based Multimedia Indexing (IEEE CBMI)

Luggage re-identification (ReID) is a critical task in video surveillance for forensics, security or border management (suspicious or abandoned luggage recognition), as well as luggage management (lost luggage recovery). It enables the association of observations of the same luggage across disjoint camera views in complex environments such as airports, train stations and border crossing. While existing approaches primarily focus on appearance-based cropped image matching under relatively clean conditions, real-world scenarios are characterized by frequent occlusions, cluttered background and crowded scenes which significantly degrade performance. Yet, strategies leveraging the complementarity of multiple shots and views from different video streams is overlooked. In this paper, we study strategies for an end-to-end pipeline that operates from multiple video streams to detect and re-identify luggage. We evaluate different variants on a benchmark dataset acquired in real-world environments to address the questions of selecting and merging different views and shots of the same luggage as well as training on different environment datasets for an overall ReID performance improvement in practical applications.

0.27

UniffordInteractor: Reasoning the Unified Affordance for Human and Multi-Object Interaction

cs.GR ·2026-10-29 ·Chen Su, Xingqun Qi, Xurui Hu, Bingkun Yang, Muyi Sun +2 more
top venue
Abstract · 13 pages, 11 figures. Published in ACM Transactions on Graphics (Proceedings of SIGGRAPH Asia 2026)

Deriving plausible and dynamic Human Object Interaction (HOI) has gained tremendous progress in Embodied AI and Immersive AR/VR. While existing HOI methods have effectively explored human interaction with individual or fixed objects, they still show limitations in complex multi-object interaction scenes. Meanwhile, previous approaches focus more on the contact information held by human-object relation, overlooking the significance of the contact situation of various entities (i.e., object-object). To address these limitations, we propose UniffordInteractor, a Reasoning-then-Interacting framework that leverages unified affordance for realistic and long-horizon human and multi-object interaction modeling. At the reasoning stage, we devise a Unified Affordance Reasoner (UAR) that employs a fine-tuned 3D Multimodal Large Language Model for unified object-centric affordance prediction. At the interacting stage, we propose Hierarchical Interaction Generator (HIG), which formulates the interaction generation as a hierarchical (e.g., root to body to hand to objects) autoregressive prediction, conditioned on the obtained unified affordance. Extensive experiments demonstrate that UniffordInteractor surpasses SOTA counterparts in human and multi-object interaction generation.

0.27

CDBM: Complexity-Diversity Framework for Building Morphology Analysis

cs.CV ·2026-10-29 ·Evgenii Maslov, Alexandra Vabnits, Denis Karpov, Alexey Morozov, Anastasia Antsiferova +2 more
Abstract

We introduce Complexity-Diversity Building Metrics (CDBM), twelve interpretable descriptors for building massings represented as vertical extrusions. They measure how strongly geometric detail is expressed (complexity) and how unevenly it varies across one- and two-dimensional elements (diversity). Computed for 482,241,306 OpenStreetMap buildings and compared with 3D Building Metrics (3DBM) and the LION shape latent on task-specific samples, CDBM is tested on function, region, contextual relevance, real-generated discrimination, human preference and geometry-conditioned generation. With gradient boosting, CDBM best separates generated from mapped massings and best matches human preference; with logistic regression it matches 3DBM on realism. 3DBM leads on the other three tasks. As generation targets, the descriptors show that models struggle to produce strongly expressed and heterogeneous forms. CDBM complements existing indicators.

0.27

Language Models Show Conformity Bias in RAG: When the Majority Endorses a Poisoned Claim

cs.CL ·2026-10-29 ·Mehrdad Ghassabi, Audrina Ebrahimi, Zahra Zojaji, Amin Milani Fard
Abstract

Retrieval-augmented generation (RAG) can expose language models to misleading evidence in retrieved documents. We investigate whether increasing the number of mutually agreeing poisoned documents affects a model's ability to detect misinformation. We evaluate 200 randomly selected questions from SciFact (random seed 38), comparing contexts with one poisoned and six healthy documents against contexts with six mutually agreeing poisoned documents and one healthy document. For Gemma-3-12b, poison-detection rates decrease from 25% (50/200) in the minority condition to 10% (20/200) in the majority condition. Among the 38 questions with different outcomes across conditions, 34 change from detected to undetected, while only four change in the opposite direction (exact McNemar test, p = 6.04e-7). This pattern is consistent with a conformity-like response to a majority of mutually agreeing poisoned documents. Our results provide initial evidence of conformity bias in RAG-based misinformation detection: Gemma-3-12b is substantially less likely to detect poisoned claims when they dominate the retrieved evidence.

0.27

Practical Assessment of Agentic Systems in Acoustics

cs.SD cs.AI ·2026-10-29 ·Arne Hildenbrand, Philipp Diel, Zhuo Cao, Hanno Scharr, Stefan Pischinger +1 more
benchmark
Abstract

In engineering, applications of artificial intelligence have so far been largely limited to surrogate models and optimization tasks. Large language models changed this, as they can write software from a task given in words. Placed inside a harness, a system of components that lets it search for information and execute software, such a model becomes an agent that can take over recurring engineering tasks. This work measures how the model and the elements of the harness shape the performance and behavior of the agent in two common acoustic applications. In the first, the agent detects and locates faults in audio recordings from a vehicle interior with near-perfect accuracy. A benchmark on this task shows that a weak model sets a limit that no harness design can overcome, and that design choices take effect only once the model is capable enough to use them. In the second, a case study on the design of a metamaterial unit cell, the agent reliably finds designs that achieve a specified insertion loss over a given frequency range. The results aim to help the reader understand what each element of an agent contributes and how to design an agent for acoustic tasks.

0.27

Homebrewed SLMs: Autoregressive and Masked Diffusion Language Models Built at Home for Portuguese

cs.CL cs.LG ·2026-10-29 ·Pedro Carvalho Brom
code/data
Abstract · 34 pages, 14 figures, 13 tables. Code and notebooks: https://github.com/pcbrom/homebrewed-slms

Building a language model of one's own no longer requires a data centre. This tutorial trains one from scratch, tokenizer included, on one consumer GPU, and shows how to tell whether it is good for the tasks of a specialist domain. It proposes no new method. It releases a tested Python package and two notebooks in Brazilian Portuguese with autoregressive and masked diffusion generation on one network, a preregistered comparison protocol, legal tasks built from text alone with a Kneser-Ney floor and public yardsticks, reference numbers for models of 11 to 12M parameters, and a catalogue of defects that only measurement revealed. On Brazilian prose and federal laws, at minutes per run, the autoregressive mode peaks at 64 epochs over the same text and then memorises it, while masked diffusion improves up to 256. At their best points on test, the diffusion bound exceeds the autoregressive likelihood by 0.06 bits per byte, within the preregistered minimum effect. On held-out laws and on acts published after training, the model ties a 6-gram floor in likelihood on the same windows, and it loses to the floor at continuing those acts. It beats the floor at detecting tampered text, and fills cross-references more often than the floor and public models of 160M and 630M parameters, only in decrees, a genre absent from training, where those scorers stay near or below chance, and in filling the blank only when each group weighs the same, since per item it ties the 630M model. On laws, the public models, trained with about 5x10^2 to 10^5 times its compute, lead in tampered text and continuation. The blind human reading is preregistered and not part of this version.

0.27

Morphology-Aware Channel-Decoupled Mamba Network for Medical Image Segmentation

cs.CV ·2026-10-29 ·Chenchen Wang
Abstract

Accurate multi-organ segmentation in computed tomography (CT) is essential for computer-aided diagnosis, treatment planning, and disease monitoring. However, substantial variations in organ shape and size, together with low tissue contrast and ambiguous boundaries, complicate the modeling of heterogeneous anatomical structures. To address these challenges, we propose a Morphology-Aware Channel-Decoupled Mamba Network (MCDNet), which integrates adaptive local structural enhancement with channel-decoupled long-range modeling. Specifically, we introduce a Morphology-Aware Channel-Decoupled State-Space (MCSS) block that divides feature channels into separate processing groups, allowing differentiated structural representations to be learned within an efficient state-space framework. Within each group, a Morphology-Aware Enhancement Module (MAEM) combines deformable convolutions with channel attention to adapt local feature extraction to anatomical shape variations. Group-wise state-space modeling then captures long-range spatial dependencies, and channel gating adaptively integrates the resulting features. Experiments on the Synapse abdominal CT dataset demonstrate the effectiveness of MCDNet, achieving an average Dice Similarity Coefficient of 81.61% for multi-organ segmentation.

0.26

Grounding Multimodal Reasoning with Evidence-Ablated Negatives

cs.CL ·2026-10-29 ·Chenxi Liu, Tianyi Xiong, Yanshuo Chen, Heng Huang
publishedcode/data
Abstract · NeurIPS2026

Reinforcement learning with verifiable rewards (RLVR) has become a powerful tool for improving multimodal reasoning. However, standard outcome supervision suffers from credit assignment ambiguity, as a model may receive positive rewards by exploiting language priors or dataset shortcuts rather than grounding its reasoning in the requisite visual evidence. This makes high-quality negative responses critical, since effective negatives should remain plausible under the current policy while exposing the shortcut that should be suppressed. In this paper, we propose EVAN, a multimodal reinforcement learning framework utilizing EVidence-Ablated Negatives to train LMMs against their own language-prior failures. EVAN first identifies highly visually dependent examples using a scaled image information gain metric and preserves data diversity with a mixed-ranking strategy. It then constructs evidence-ablated images by masking the most discriminative image patches, eliciting plausible but visually ungrounded responses from the current policy. During policy optimization, these negative responses are incorporated as bounded auxiliary signals under the clean image, penalizing shortcut learning while maintaining RL stability. Experiments across challenging multimodal reasoning benchmarks demonstrate that EVAN mitigates modality imbalance and improves over strong RLVR baselines. Our code is available at https://github.com/DawnLIU35/EVAN.git.

0.26

MetaRefine: Score Refinement with Multi-Level Features for Image Assessment

cs.CV cs.AI ·2026-10-29 ·Chuanzhi Xu, Chengkun Yue, Haolan Guo, Xuanhua Yin, Boyu Wei +3 more
Abstract

An image's original dimensions, aspect ratio, and composition provide important cues that can affect its technical and aesthetic quality. Image assessment methods use resizing, cropping, padding, or patch sampling to accommodate different input sizes. However, these operations can alter or obscure original geometric and compositional cues. Existing assessment models also encode rich multi-level visual evidence that can be further extracted and exploited to improve existing models. This motivates us to combine visual evidence with original-image geometry for score refinement without changing the original scoring pipelines. In this paper, we introduce MetaRefine, a metadata-aware score refinement architecture whose lightweight residual heads predict score corrections from geometry, multi-level visual features, and the frozen model's original score. Across 100 configurations on 21 datasets, MetaRefine improves PLCC in 88 configurations and SRCC in 85, with average relative gains of 7.55% and 7.10%, respectively. It reduces MAE in 90 configurations and RMSE in 89, with average relative reductions of 23.27% and 21.93%, respectively. The median parameter overhead is only 1.11%. MetaRefine can further improve state-of-the-art (SOTA) assessment models on multiple benchmarks.

0.26

Symmetric Structure and Motivic Transformation in Melodies, Phrases, and Themes

cs.SD physics.soc-ph ·2026-10-29 ·Alfredo González-Espinoza
Abstract

Transposition, inversion, and retrograde are fundamental operations in music theory, from Schoenberg's twelve-tone technique to Forte's set classes and Lewin's transformational theory, yet their use in melodies has rarely been quantified. We read a melodic line through the orbits and stabilizers of its windows of $m = 3-6$ notes under the group generated by these operations and cyclic shift. The window entropy splits into an orbit entropy (motivic vocabulary) and a within-orbit entropy normalized by orbit size (deployment), and the group elements relating windows of one orbit give the motivic transformations. In 852 keyboard pieces spanning four centuries, at least 97% of melodies have lower orbit entropy than surrogates preserving their pitch content, spectrum, or one-step transitions ($m \geq 4$). Periods differ in vocabulary and symmetry types; musical forms differ in deployment. Recurring motifs are transposed 1.6 times more often than in random permutations of their forms, and transposed by a fifth or fourth 2.8 times more often, a preference that weakens from the Baroque to the 20th century and from Bach's fugues to Shostakovich's. Folk phrases and instrumental themes are rarely symmetric (2-10%), but are 1.6-5 times more often symmetric than note rearrangements. The framework extends to any symbolic sequence with a symmetry group.

0.26

CLaW: Codec-Guided Adaptive Latent Watermarking for Traceable Diffusion Image Generation

cs.CV cs.CR ·2026-10-29 ·Huangchao Song, Lingyun Yu, Cong Li, Peiqi Jiang, Mingqi Fang +1 more
top venue
Abstract · Accepted at NeurIPS 2026

With the rapid advancement of text-to-image diffusion models, increasingly realistic AI-generated content has raised serious concerns about misuse and copyright infringement. Digital watermarking offers a promising solution by enabling user-level traceability, yet existing methods remain difficult to deploy at scale, hindered by costly model retraining or detection, poor effectiveness-fidelity trade-offs, and limited robustness to image transformations. To address these challenges, we propose CLaW (Codec-guided Latent Watermarking), a robust and efficient watermarking framework for traceable diffusion image generation with frozen diffusion backbones and inversion-free detection. Specifically, CLaW first uses a pretrained watermark codec to map each watermark message to a latent residual, enabling low-cost sampling-time injection. Then, the residual is injected within a late denoising window, where image semantics are better preserved and the watermark signal is less perturbed by subsequent denoising updates, yielding a better balance between watermark effectiveness and image fidelity. Furthermore, we introduce a decoder-guided adaptive injection mechanism that uses decoding confidence as feedback to dynamically calibrate watermark strength, reinforcing weak watermark signals and improving robustness under image transformations. Extensive experiments show that CLaW preserves high visual fidelity and achieves state-of-the-art robustness under image transformations, enabling reliable user-level traceability.

0.26

PhysCount: Do MLLMs Count Visual Appearances or Physical Entities?

cs.CV ·2026-10-29 ·Lidong Lu, Jiaxin Lv, Liangrui Dong, Xinyi Mao, Lidong Zou +3 more
benchmark
Abstract

Images reveal appearances, but counting often asks about physical objects. A single entity may be visible directly, in several reflections, or through glass, so detecting every visible occurrence does not determine how many entities are present. Existing visual counting benchmarks rarely separate these two operations. We introduce PhysCount, a benchmark for evaluating whether multimodal large language models (MLLMs) can distinguish what appears in an image from what physically exists in the scene. PhysCount contains 546 real-world images and 4,595 human-written questions organized into 651 target-specific groups. For each group, annotators locate the relevant appearances, label how each is observed, and identify appearances of the same entity. Four complementary question families test the enumeration of appearances, the enumeration of physical entities, image-formation-specific counting, and appearance--entity correspondence. We measure both accuracy on each family and the proportion of groups answered entirely correctly. Across 40 MLLMs, counting visible appearances is substantially easier than recovering the identities that connect them. Even the strongest model answers every question correctly in only 13.36\% of groups. Performance declines when one entity produces many appearances or when its appearances cross observation modes. Higher image resolution improves access to visual evidence but brings limited gains in complete-group accuracy, whereas answering related questions jointly produces a marked improvement. These results expose a gap between recognizing what is visible and correctly recovering the physical entities behind those observations.

0.26

MLLM-GS: Multimodal Large Language Model Prior-Guided Fast Novel View Synthesis in Extreme Environments

cs.CV ·2026-10-29 ·Zhiliang Wu, Xinye Zheng, Kun Li, Fei Wang, Hao Guo +5 more
code/data
Abstract

Novel view synthesis (NVS) in degraded scenes aims to reconstruct 3D scenes from multi-view observations affected by various degradations, and to generate clear and photorealistic images at unseen viewpoints. Despite recent advances, existing methods often rely on degradation-specific models, limiting their generalization across diverse extreme environments. Generative models provide rich visual priors, but their outputs often suffer from pixel misalignment and color deviations, which can propagate into the reconstructed 3D scene and compromise reconstruction quality. Furthermore, the high computational and memory costs of 3D reconstruction hinder the practical application of NVS in extreme environments. To this end, we propose MLLM-GS, a multimodal large language model (MLLM) prior-guided framework for fast NVS in extreme environments. MLLM-GS consists of two branches: MLLM prior-guided view enhancement and fast 3D scene reconstruction. The former leverages the rich visual priors of a frozen MLLM to enhance views under different severe degradations and introduces a lightweight correction network to reduce pixel shifts and color distortions in MLLM outputs. With supervision from only a few paired clean views, this branch produces reliable enhanced views for all degraded observations, providing pseudo-GT views for subsequent 3D reconstruction. The latter improves 3DGS efficiency by optimizing rasterization, Gaussian memory layout, and parameter updates, reducing redundant computation, memory-access overhead, and optimization costs. Extensive experiments on multiple degraded scenes demonstrate that MLLM-GS effectively handles various severe degradations and achieves high-quality NVS with significantly lower reconstruction time and memory consumption. Project Page: https://wzlwlm13.github.io/MLLM-GS

0.26

Anatomical Prior-Guided Routing for 3D Lung Tumor Segmentation

cs.CV ·2026-10-29 ·Uihyeon Maeng, Jonghun Kim, Hyunjin Park
published
Abstract · Accepted to ACCV 2026 (Oral), 18 pages, 6 figures

Accurate lung tumor segmentation in computed tomography (CT) is essential for treatment planning, response assessment, and quantitative imaging analysis. However, 3D lung tumor segmentation remains challenging because tumors occupy only a small portion of the whole volume and exhibit heterogeneous appearances across anatomical locations. In patch-based training and sliding-window inference for 3D medical image segmentation, location-dependent heterogeneity is further amplified because cropped image patches provide limited global anatomical context, making it difficult for a model to adapt its feature processing according to the anatomical region from which each patch is extracted. To address this limitation, we propose an Anatomical Prior-guided Routing Module (APRM) for lung tumor segmentation. For each image patch, we obtain anatomical prior information of regional occupancy and deterministically select an anatomical routing gate. The selected gate then performs sparse feature routing over a shared expert adapter pool, enabling anatomy-specific feature adaptation while sharing expert capacity across regions. We evaluate APRM on two datasets with various backbone networks and demonstrate consistent improvements over baseline models. Furthermore, feature visualizations show that APRM produces different feature responses depending on patch location, indicating that the proposed routing mechanism effectively adapts feature processing according to anatomical context.

0.26

Operator-Aware Posterior Sampling for Multivariate Weather Downscaling

cs.CV ·2026-10-29 ·Carlo Saccardi, Rudolf Saathof, Geethu Joseph, Justin Dauwels
code/data
Abstract

Statistical downscaling is the task of reconstructing high-resolution meteorological fields from coarse inputs. A principled approach formulates downscaling as a Bayesian inverse problem, using a diffusion prior over high-resolution fields and incorporating the coarse field only at inference through a likelihood. This decouples the prior from the coarse source, allowing one pretrained model to adapt to different sources. However, two challenges limit this flexibility: (1) when coarse and target fields come from different reanalysis systems, no fixed coarsening operator can capture their relationship; and (2) guidance gradients can differ by orders of magnitude across variables, making a single guidance strength ineffective. This paper addresses both challenges within the guidance loop, leaving the prior unchanged. We learn the forward operator from paired data and normalize the gradient per channel. We show that this normalization corresponds to steepest descent under a block-max norm, allowing a single guidance strength across variables. Our code is publicly available at https://github.com/CarloSaccardi/operator-aware-weather-downscaling.

0.26

PathCowork: Learning from real-world pathology practice to build AI copilots

cs.CV ·2026-10-29 ·Sheng Wang, Jialu Yao, Maria V. Rossi, Shunsuke Koga, Esther Baranov +20 more
code/data
Abstract · 56 pages, including supplementary information. Code, models and demonstration videos: https://github.com/zhihuanglab/PathCowork

Medical AI learns from the answers doctors record, but rarely from how they interpret the evidence to reach them. Here we introduce PathCowork, a framework that turns pathologists' narrated examinations into supervision for visual grounding and procedures that guide diagnostic AI agents. Synchronized gaze, narration and navigation link visual findings to the process of interpreting them. Across 48 sessions from 16 pathologists, the platform captured 1,028 minutes of slide review. Using these data, we trained PathCowork-1T for server deployment and PathCowork-27B for local deployment, with improved visual grounding. Across five organ-specific evaluations, agents combining this grounding with recorded diagnostic procedures achieved primary task scores averaging 17.9 and 24.1 percentage points higher than Claude Code and Codex, respectively, with shorter processing times and lower inference costs. Compared to the state-of-the-art multiple-instance learning AIs, the PathCowork agent-proposed regions achieved higher prostate-grading scores with limited slide labels. PathCowork can also help construct pathology AI benchmarks where questions derived from the recordings identified difficulties in recognizing viewing context and rejecting inapplicable premises. In summary, PathCowork proposes a new paradigm to scale up the data collection and train pathology AI through transparent reasoning from diagnostic sessions, opening a path towards systems that learn directly from routine clinical practice.

0.26

Judged in Company: In-Context Batching Makes LLM Verdicts Depend on Other Items

cs.CL ·2026-10-29 ·Peng Xie, Amr Alanwar
Abstract

To save cost, practitioners often ask a language model to grade many independent items in one prompt. This assumes that verdicts are separable: that the verdict on an item does not depend on the other items in the prompt or on its position among them. Across three judgement tasks (harm in social-media comments, short-answer grading and review star ratings) and five open-weight models with 4-9B parameters, assigning the same 2,000 items to different random batches of 25 changes 11-22% of decisions on harm and grading, 6.5-13.3 points more than renaming the item keys or rewording the instruction, while AUROC against human labels differs from single-item grading only by -0.033 to +0.010; for four models with 14-32B parameters, 6-15% of decisions change, 4.2-9.1 points more. We show that flip rates decompose exactly into components of order, slot and membership, and that under a threshold model they grow with the square root of the variance that the context contributes, so that effects that are small for scores are large for decisions. Most of the latent dissimilarity between partitions comes from the order of the neighbours and the slot. The main effect of which items share the prompt, the part that voting over permutations of a batch cannot remove, accounts for 0.11-0.22 of it, and majority votes over nine permutations still leave 52-60% of the flips. Batched verdicts also differ systematically from single-item verdicts: even the majority of fifteen random batches disagrees with them on 8.4-17.8% of decisions. Editing a single generation shows that both the model's own earlier verdicts and the neighbours' texts carry the effect. On the easy task, star ratings, only 2.9-3.9% of decisions change. We recommend grading ambiguous items one per prompt, recovering most of the cost with a cached prefix, and reporting agreement with single-item grading whenever items are batched.

0.26

FlexTry-On: A Time-Efficient Framework for Multi-Scenario Virtual Try-on

cs.CV ·2026-10-29 ·Phuong Dam, Quyen Nguyen Duc, Son Hoang Minh, Daeyoung Kim
Abstract

Virtual try-on has achieved remarkable progress in recent years, yet most existing methods are designed for a specific scenario, such as cloth-to-person or person-to-person try-on, and often require separate conditioning strategies and processing pipelines for different settings. We present FlexTry-On, a unified and time-efficient diffusion-based framework for diverse virtual try-on scenarios, including cloth-to-person, person-to-person, cross-domain, and in-the-wild settings. FlexTry-On adopts a common conditioning formulation, where a text-aware condition extraction strategy converts heterogeneous garment sources into a unified representation, and a dual-conditioning diffusion transformer jointly models garment appearance and person identity. Extensive experiments demonstrate competitive performance on both cloth-to-person and person-to-person benchmarks while maintaining fast inference. Moreover, the same framework can be directly applied to unseen cross-domain scenarios without requiring separate architectures or scenario-specific processing pipelines. These results highlight the potential of unified and efficient virtual try-on frameworks as a practical step toward more flexible and widely applicable virtual try-on systems.

0.26

Fast rigid rotational alignment of voxelized 3-D density maps in sliced Wasserstein distance

cs.CV math.NA ·2026-10-29 ·Oscar Mickelin, Yunpeng Shi, Amit Singer, Eric J. Verbeke
Abstract

We develop a fast method for the rigid alignment of voxelized three-dimensional density functions over $\mathsf{SO}(3)$ using sliced optimal transport. Our approach efficiently transforms each density map into a tensor of one-dimensional quantile functions, taken from central slices over the unit sphere, such that the Euclidean distance of the transformed volumes approximates the sliced 2-Wasserstein distance. Our formulation is therefore compatible with existing, fast, voxel-based Euclidean alignment algorithms. We provide theoretical analysis and numerical experiments demonstrating that the sliced 2-Wasserstein distance is a robust metric for comparing three-dimensional shapes and for computing their rigid alignment.

0.26

SelfThinker: Self-Evolving Reasoning via Reliability-Verified Intrinsic Visual Thoughts

cs.CV ·2026-10-29 ·Zhangquan Chen, Ruqi Huang
benchmarkcode/data
Abstract · 25 pages, 12 figures

Unified multimodal models can both generate and understand images within a single set of weights, offering an opportunity to use image synthesis as training supervision for visual comprehension. However, existing visual chain-of-thought methods rely on human-annotated helper images during training and require slow image generation at test time. We present SelfThinker, the first self-evolving framework that turns a unified model's self-generated visual thoughts into stronger understanding without human image annotations. Given a visual problem, SelfThinker samples candidate helper images from its generation pathway and uses its frozen understanding pathway to score each candidate via conditional pointwise mutual information. High-utility visual thoughts then refine the generator via flow matching and serve as privileged teacher context for on-policy self-distillation. At inference, SelfThinker reasons directly from the original image in text without generating intermediate images. Although trained only on small synthetic visual-search and navigation datasets, SelfThinker transfers broadly to natural-image and spatial benchmarks, raising the seven-benchmark average of ThinkMorph-7B from 52.7 to 64.0 (+11.3 points). Controlled image-perturbation probes further show that SelfThinker surpasses ThinkMorph-7B by up to 40.4 points and its pretrained BAGEL-7B backbone by up to 23.2 points across fine-grained and spatial perception tasks. Our code is available at https://github.com/zhangquanchen/SelfThinker.

0.26

Beyond CCT: Multi-channel SPD recovery for per-shot camera color correction

cs.CV ·2026-10-29 ·Shuwei Yue, Hongke Wei, Qiao Chen
benchmark
Abstract

Modern cameras correct color via two or three pre-calibrated 3x3 matrices interpolated by Correlated Color Temperature (CCT). This collapses the illuminant SPD to a scalar. LEDs, fluorescents, and mixed sources sit off the Planckian locus, and metameric illuminants with the same chromaticity demand different optimal CCMs that a CCT pipeline cannot distinguish. We recover the illuminant SPD instead using non-negative least squares against a basis dictionary, conditioned on the camera white point and a multi-channel ambient-light sensor (ALS). A closed-form 3x3 CCM is then fit per shot. On a synthetic benchmark of 300 illuminants and 3,100 reflectances, mean angular error drops from 1.89 to 1.20 degrees, with a 1.7x tail reduction on the LED-like off-Planckian group (n=45) where metameric ambiguity is tangible. The gain holds under realistic ALS imperfections and observation noise, and the same off-Planckian advantage carries over to 1,684 real measured SPDs.

0.25

R2QRec: Learning Recommendation Queries from Retrieval Feedback

cs.IR ·2026-10-29 ·Weixin Chen, Shuchang Liu, Hailan Yang, Yuhan Zhao, Li Kang +11 more
hot direction
Abstract

Query-based recommendation separates expressing user needs from locating matching items. Its central learning problem is how to turn item interactions into effective retrieval requests when interaction logs do not annotate queries. We present R2QRec, which uses a fixed retrieval system to construct query supervision and evaluate recommendation outcomes. Retrieval verification first selects generated expressions that recover observed items as supervised fine tuning targets, allowing retrieval behavior to determine the target wording. Policy refinement then evaluates generated query sets by the observed item's rank in their fused recommendation list. Log-rank feedback distinguishes outcomes beyond the recommendation cutoff, while a KL penalty constrains policy change relative to the supervised reference. The stages have distinct roles: verification identifies expressions to imitate, and policy refinement improves the combined ranking produced by newly generated queries. Only the query generator is trained; the retriever and item index remain fixed, and no query annotations or learned reward model are required. Across three public datasets, R2QRec achieves state-of-the-art overall recommendation performance among the evaluated methods. Extensive experiments demonstrate the effectiveness of the proposed design and its generalization across generators and domains.

0.25

EDU-Jev Training System-One Decision Models for Education

cs.CL ·2026-10-29 ·Jiazheng Li, Yulan He
Abstract · 16 pages, 3 figures, 4 tables. Work in progress

Many automated decisions in education involve choosing from a fixed set of options: assigning a rubric score, checking whether an answer meets a marking criterion, or identifying the first wrong step in a solution. Generative language models are increasingly used to make these decisions, but their responses require parsing, longer outputs increase inference costs, and confidence expressed in words does not necessarily reflect the probability of being correct. We present EDU-Jev, a family of single-pass ("System One") decision models at 0.8B, 2B and 4B parameters. Given an assessment context and a question with predefined answer options, the model returns a probability distribution over those options without generating text. A pointer head scores a marker for each option, ensuring the predictions are restricted to the allowed choices. Training uses a single supervised stage with soft cross-entropy against the available target label distributions; calibration is assessed empirically. Across twelve education benchmarks, EDU-Jev 4B achieves an average accuracy of 67.6%, exceeding the strongest of five commercial decision APIs by 9.4 percentage points and the best open 4B decision model by 18.7 points. On four cross-domain education sets, it matches the strongest API, which remains ahead on general decision-making and reasoning tasks. The 2B model processes about 150 decisions per second on a single H100 GPU.

0.25

ACD-GS: Asymmetric Curvature-aware Densification for 3D Gaussian Splatting

cs.CV ·2026-10-29 ·Houqiang Zhong, Tianchi Zhu, Qiang Hu, Xiaoyun Zhang, Li Song +1 more
code/data
Abstract

3D Gaussian Splatting (3DGS) has emerged as a powerful paradigm for real-time photorealistic novel view synthesis. However, its training-time densification remains largely heuristic, often resulting in redundant Gaussian proliferation and inaccurate geometric reconstruction. Existing methods predominantly rely on first-order screen-space gradients to trigger cloning and splitting, which fail to distinguish true geometric discontinuities from high-frequency textures or noisy residuals. To address this limitation, we propose ACD-GS, an Asymmetric Curvature-aware Densification framework for compact and geometry-aware 3D Gaussian optimization. Our key insight is that second-order depth responses provide a more reliable structural prior for density control, enabling asymmetric regulation of cloning and splitting to better capture geometric discontinuities. Building on this, we introduce a multi-view photometric consistency constraint to suppress transient single-view errors, together with a capacity-aware adaptive regulation strategy that balances densification and pruning during training. Unlike post-hoc compression or quantization-based approaches, ACD-GS directly prevents redundant Gaussian generation during training, yielding inherently compact representations. Extensive experiments demonstrate that our method reduces nearly 60% of Gaussian count and storage while maintaining competitive rendering fidelity, consistently outperforming existing methods across multiple datasets. Our code is available in https://waveviewer.github.io/ACD-GS/.

0.25

PRISM: Personalized Item Representations in User History

cs.IR cs.AI ·2026-10-29 ·Renzhi Wu, Zikun Cui, Junjie Yang, Li Sheng, Yongjun Xie +4 more
hot direction
Abstract

Sequential recommenders typically construct history inputs from item identifiers or content features using an embedding map shared across users. The same item therefore yields the same initial embedding representation, even when users have different relationships to the item. We propose personalized item representations in user history, where each item's embedding expresses its relationship to the user. Personalization becomes a property of the history inputs themselves, supplying richer evidence for later sequence modeling. We instantiate this idea with PRISM, which reuses upstream ranking computation to generate personalized item embeddings. An autoencoder jointly compresses all layer states from a user tower, a user--item tower, and same-request candidate mixing into one embedding per scored pair for downstream replay. Offline evaluation on Surface 1, a major recommendation surface at billion scale, reports an average downstream normalized entropy (NE) change of $-0.15\%$, compared with $-0.07\%$ for VISTA at similar history lengths. We further identify an empirical sequence-length scaling law: over history lengths of 26--240 events, NE gain follows a fitted power law with exponent 0.468, increasing from 0.06\% to 0.17\%. Therefore, personalized item representations in user history provide a new scaling dimension for improving recommendation quality.

0.25

DeCoRL: Decomposed Consistency Reinforcement Learning for Multi-Image Composition

cs.CV cs.AI ·2026-10-29 ·Zhiqiang Wu, Shuang Sun, Jiale Zhang, Jing Li, Yibo Chen +2 more
publishedbenchmark
Abstract · NeurIPS

Multi-image composition (MICo) is a critical yet challenging capability for large-scale image editing models, where consistency degrades rapidly as the number of reference images increases. While supervised fine-tuning (SFT) can equip models with basic MICo ability, it provides only implicit supervision on the composed image and often fails to learn reliable object-level correspondence across multiple references. We propose DeCoRL, a reinforcement-learning (RL) framework that improves MICo consistency by decomposing holistic MICo consistency into object-level consistency. DeCoRL first establishes instruction-relevant object correspondences across reference images and localizes the corresponding regions in the generated image. It then uses a VLM-based reward model trained on single-object consistency annotations to produce fine-grained, object-wise feedback, which is aggregated into a multi-object consistency reward to guide RL post-training. We further introduce DeCo-Bench, a benchmark covering compositions from 2 to 6 reference images. Experiments on DeCo-Bench and the public MICo-Bench demonstrate consistent improvements in MICo consistency.

0.25

Toward Complete Speaker Anonymization: a Streaming Cascaded System for Voice Privacy and PII Removal

cs.SD eess.AS ·2026-10-29 ·Stanisław Kacprzak, Thomas Thebaud, Najim Dehak
Abstract · Submitted to ICASSP 2027

We present a cascade streaming voice privacy system, consisting of ASR, NLP, and TTS modules. By leveraging recent improvements in all three aspects of speech and language processing, we show that such a system is capable of achieving a result comparable to state-of-the-art offline voice privacy systems, while providing controllable capabilities of content protection beyond voice. With an estimated utterance-onset end-to-end latency of 717 ms, the presented system preserves speech content with 11% WER on Librispeech, while detecting PII with a 69% SLU F1 score and achieving up to 47.66% EER against a semi-informed attacker.

0.25

Federated Recommender Agents

cs.IR ·2026-10-29 ·Jaehyung Lim, Woojoo Kim, Junyoung Kim, Dongha Kim, Wonbin Kweon +1 more
Abstract · 13 pages

LLM-powered recommender agents use reasoning and memory for personalization, but centralized access to user information raises privacy concerns, while isolated agents cannot learn from others' experiences. We formulate Federated Recommender Agents, a setting where local agents collaborate through natural-language knowledge while keeping interaction histories and personal memories local. We propose FedRecAgent, which enables this collaboration without updating LLM parameters. Its central challenge is determining which reasoning to integrate, since similar behavioral contexts can lead to different choices. FedRecAgent represents local experiences as Condition-Transition-Reasoning (C-T-R) units and hierarchically integrates reasoning by condition and transition. Within this structure, it consolidates related explanations and discriminates reasoning across alternative transitions. Each client maintains semantic and decision memories to capture preferences and conditional choice tendencies, guiding the retrieval and selection of shared reasoning for recommendation. Experiments across three domains under dense and sparse settings demonstrate improved recommendation performance over federated and LLM-based baselines, while ablations support the contributions of personalized memories and structured reasoning integration.

0.25

Diagnosing the Reference Gap: What Visual Scaffolds Can and Cannot Fix in Multimodal Reasoning

cs.CV cs.CL ·2026-10-29 ·Guizhen Chen, Weiwen Xu, Wotao Yin, Anh Tuan Luu, Hou Pong Chan
benchmarkposition
Abstract · 20 pages, 8 figures

Multimodal large language models can recognize visual content yet struggle to reason reliably over it across multiple steps. In counting, a model may see all objects but lose track of which ones it has counted. We argue that such failures, commonly attributed to perception, reveal a reference gap: the failure to bind language to the relevant visual entities, regions, or locations. Motivated by the theory of visual routines, we test whether external references such as numbering can narrow this gap. We introduce ReGap-Bench, a paired benchmark of five visual reasoning tasks in which the scene, question, and answer are fixed and only the reference varies. Results show that task-matched references improve accuracy by up to 25\%, while mismatched references provide little benefit or even hurt performance. These findings identify the reference gap as a bottleneck distinct from visual perception and show that constructing task-appropriate visual references can serve as a training-free visual skill for improving multimodal reasoning.

0.25

Blind Video Denoising for Vocal Tract rtMRI using a Self-Supervised Ensemble

eess.IV cs.CV ·2026-10-29 ·Julian Andres Hernandez-Potes, Tomás Arias-Vergara, Daiqi Liu, Andreas Maier, Paula Andrea Perez-Toro
Abstract · 7 pages, 5 figures

Real-time magnetic resonance imaging of the vocal tract makes speech articulation directly observable, but the trade-off between temporal resolution and signal quality introduces substantial noise and obscures fine articulatory detail. Since clean references cannot be acquired for a moving vocal tract, supervised denoising approaches are not applicable in this setting. We propose a self-supervised denoising framework that combines spatiotemporal super-resolution learning with pseudo-supervision derived from a variance-stabilized non-local means filter. A dual-model fusion strategy balances noise suppression and preservation of high-frequency articulatory structure. Experiments on the USC-75 multi-speaker dataset under natural noise conditions show that the ensemble raises the signal-to-noise ratio by 11.9 dB on unseen subjects while attaining more high-frequency articulatory content in the frames. Among tuned implementations of anisotropic diffusion, block-matching filtering, non-local means and pre-smoothed non-local means, every configuration that achieves comparable noise suppression does so at the cost of high-frequency content or of structural fidelity. Under severe added Gaussian, Rayleigh and Rician noise, peak signal-to-noise ratio improves by 7.6 to 11.8 dB and structural similarity roughly triples. An ablation shows that the variance-stabilized pseudo-supervision is an essential element for denoising, and that the effect is attributable to the hard thresholding contained in the stabilization transform.

0.25

Prior-Guided Single-Exposure HDR Novel View Synthesis via Base-Gain Representation

cs.CV cs.MM ·2026-10-29 ·Armin Shafiee Sarvestani, Sheyang Tang, Jialu Xu, Zhou Wang
Abstract

Single-exposure high dynamic range (HDR) novel view synthesis (NVS) recovers HDR scenes from low dynamic range (LDR) views captured at a single exposure and estimates camera response function (CRF) for LDR re-exposure. Without HDR ground-truth, using single-exposure LDR views as the sole source of direct supervision underconstrains HDR reconstruction and cannot ensure its faithful recovery, motivating the need for learned LDR-to-HDR reconstruction priors. Moreover, single-exposure LDR views supervise CRF only over a limited range, necessitating a learnable prior for reliable re-exposure beyond it. We propose PriorGS, a 3D Gaussian Splatting framework that addresses both challenges by (1) guiding HDR recovery with pseudo-HDR supervision from 2D single-exposure HDR reconstruction methods while anchoring it to input LDR views through a closed reconstruction loop; and (2) regularizing CRF modeling with a learnable prior that guarantees monotonicity and promotes plausible responses. Yet cross-view inconsistencies in independently generated pseudo-HDR targets impair scene reconstruction when HDR is encoded directly as Gaussian's color attribute. We therefore introduce a 3D base-gain HDR representation that factors HDR radiance into an LDR base and an overlaid 3D gain field, promoting robustness to such inconsistencies and enabling 3D HDR appearance editing. Extensive experiments demonstrate PriorGS's substantial superiority over baselines.

0.25

Difficulty Calibration Across Time for Membership Inference in Diffusion Models

cs.CV cs.CRcs.LG ·2026-10-29 ·Michael Joseph, Yedid Hoshen
code/data
Abstract

Modern diffusion models are trained on massive datasets that often include copyrighted or private images. Membership inference attacks (MIAs) address this by testing whether a given image was in a model's training set. The standard approach scores an image by its reconstruction loss and flags low-loss images as members. This often fails because some images are simply easier to reconstruct than others, so a simple unseen image can score a lower loss than a complex training image. Difficulty calibration separates memorization from how difficult an image is to begin with. We show how to calibrate along two axes of time: training time and diffusion time. First, across training time, we introduce Reference-Model Calibration (RMC), which compares a fine-tuned model's loss against an earlier checkpoint. This cleanly isolates newly memorized data and gives near-perfect auditing, even without text captions. Since earlier checkpoints are not always available, we next present Temporal Self-Calibration (TSC), which calibrates across diffusion time using only a single model. Our key observation is that an image's difficulty persists across noise levels, whereas memorization is localized in time. By comparing the loss at a target timestep against neighboring timesteps from the same model, TSC attenuates image difficulty using the model's own trajectory. Empirically, our methods beat the state of the art by a wide margin, and unlike most existing attacks, generalize across different model parameterizations such as $x$-, $v$-, and $ε$-prediction models. Code is available at https://github.com/MichaelJoseph777/TSC.

0.25

WebOverlay: High-Throughput Vision-Only Distillation for Extreme Aspect-Ratio Web Overlay Detection

cs.CV cs.AIcs.IR ·2026-10-29 ·Jiannan Huang, Jerry Ding, Li Yu, Zhenting Li
Abstract · 9 pages, 2 figures, 2 tables

Web overlays, such as cookie consent banners, newsletter prompts, and age gates, obstruct automated web crawlers and degrade downstream document representations. While proprietary multimodal Large Language Models (MLLMs) can accurately identify overlays, their latency and inference cost preclude web-scale ingestion. Conversely, lightweight vision models struggle with full-page renderings characterized by extreme vertical aspect ratios, where standard isotropic resizing collapses critical text features and causes false positives. In this work, we present WebOverlay, a high-throughput vision-only detector trained via distillation from a frontier vision-language teacher. We establish a dynamic tiling strategy that partitions extreme aspect-ratio web renders into aspect-preserving patches with spatial overlap, coupled with an aspect-ratio-aware boundary adjudication mechanism to eliminate boundary artifacts. To eliminate cross-modal latency overheads, our student model relies purely on visual cues, bypassing DOM tree parsing. WebOverlay significantly outperforms DOM-based heuristics and achieves competitive precision and recall against heavy proprietary MLLM baselines while delivering multiple orders of magnitude higher throughput and substantially reduced deployment costs.

0.25

LucidLyra: Feed-Forward Generative 3D Scene Reconstruction and Joint Understanding via Multi-Task Distillation

cs.CV ·2026-10-29 ·Oliver Hahn, Rodrigo Marcuzzi, Sangeetha Grama Srinivasan, Alessandro Burzio, Sherwin Bahmani +5 more
code/data
Abstract · Project Page: https://research.nvidia.com/labs/sil/projects/lucidlyra

We present LucidLyra, a framework that makes generatively reconstructed 3D scenes directly queryable for semantics and instances in a feed-forward manner from video latents. While generative 3D reconstruction can synthesize virtual environments from a single image, downstream applications in robotics and physical AI require not only visual realism but also access to scene semantics and object identities. Existing 3D understanding approaches, however, require captured multi-view scenes, have limited generalization capabilities, or rely on expensive per-scene processing and optimization. To bridge this gap, our method distills complementary knowledge from a camera-controlled video diffusion model and frozen 2D vision foundation models into a unified model to jointly infer 3D Gaussians encoding geometry, appearance, open-vocabulary semantics, and class-agnostic instance cues. The resulting scenes support novel-view rendering, open-vocabulary segmentation, and object grouping without per-scene optimization, while training requires neither manual 3D annotations nor captured multi-view data. Our method outperforms baselines combining state-of-the-art 3DGS segmentation approaches with generative 3D scene reconstruction on curated synthetic and real-image-conditioned benchmarks, while preserving reconstruction quality. Furthermore, our latent formulation naturally extends to monocular video, reconstructing structured dynamic 3D scenes across time. By turning renderable but unstructured outputs into queryable 3D worlds, LucidLyra enables downstream applications including 3D scene editing, 3D asset mining, and robotic simulation.

0.24

MemoryFusion: Cross-Temporal Memory Learning for Multimodal Video Fusion

cs.CV ·2026-10-29 ·Meiqi Gong, Hao Zhang, Jiayi Ma
code/data
Abstract

Multimodal videos contain rich temporal cues that can facilitate faithful fusion. However, existing video fusion methods mainly rely on short-range temporal interactions, leaving valuable long-term information underexplored. Moreover, inconsistent fusion representations can cause flickering artifacts, undermining temporal coherence. To address these coupled challenges, we propose MemoryFusion, a two-stage framework integrating cross-temporal memory learning with progressive temporal refinement. Considering the complementary roles of adjacent and distant frames, the first stage combines a Short-Term Memory Bank for local temporal guidance with a Long-Term Memory Bank that preserves reliable historical information through a fusion-oriented update strategy. To mitigate temporal fluctuations, a Temporal Smoothing Module leverages soft motion cues for coarse background stabilization. Building upon the resulting fusion prior, The second stage further employs a lightweight 3D residual Temporal Refinement Module to correct remaining temporal inconsistencies while enhancing structural details. Extensive experiments demonstrate that MemoryFusion achieves superior fusion quality and temporal consistency over existing frame-based and video-based fusion methods. The source code is publicly available at https://github.com/Meiqi-Gong/MemoryFusion.

0.24

Mobilizing the Archive: Bitext Extraction at Scale from Archival Documents for Extremely Low-resource Languages

cs.CL ·2026-10-29 ·Chutong Meng, Alec J. Bateman, Antonios Anastasopoulos
Abstract

While mining parallel sentences from web pages has proven effective for low-resource machine translation, the majority of the world's languages have little to no web presence. What they do have, though, are resources that are the result of centuries of efforts by linguists, such as grammars, dictionaries, and other documentation materials. We propose extracting translation examples from such linguistic resources, which offer parallel data for languages under-represented online and could potentially scale to thousands more languages. We develop a pipeline to achieve this, leveraging an Optical Character Recognition (OCR) model as well as a vision language model (VLM), to ensure identified content to be aligned with visual layout information. We demonstrate the utility of the extracted bitexts for bootstrapping translation systems across 20 languages, both by training machine translation models and by applying in-context learning for large language models. For in-context learning, we observe improvements of +15.78 and +6.94 chrF2++ over zero-shot translation for into low- and high-resource languages, respectively.

0.24

GistRank: Gist-Compressed Listwise Ranking with Large Language Models

cs.IR ·2026-10-29 ·Guo Chen, Wenhan Liu, Reed Li, Qianwen Ma, Yu Lu +3 more
Abstract

Large language models (LLMs) have become a powerful tool for listwise relevance ranking, which can improve web search by comparing retrieved passages jointly. However, processing every candidate verbatim and generating a permutation autoregressively can make inference costly, every candidate must be serialized verbatim into the prompt and the input therefore grows linearly with both the number and the length of the passages. While sliding-window ranking handles context limits through repeated re-encoding of overlapping blocks, hard compression techniques sacrifice the fine-grained details required for accurate discriminative judgment. To better understand and mitigate this trade-off, we conduct a comprehensive study of soft compression for LLM-based listwise ranking. Our findings reveal that a vanilla gist representation is ill-suited to ranking for two reasons: strictly causal encoding weakens how thoroughly each gist token absorbs its passage, and the representation is never trained to expose the relative-preference structure that separates closely competing candidates. To this end, we propose GistRank, a three-stage training framework that equips a large language model with gist-compressed passage representations by learning faithful gist tokens under a semi-bidirectional attention scheme, instilling discriminative comparison through pairwise training, and adapting the model to the listwise objective. GistRank collapses each passage into only $K{=}4$ gist tokens, this gives 14$\times$ speedups over one-shot full-text ranking while reaching 59.17 average nDCG@10 over TREC DL19/20 and five BEIR subsets within 0.89 of the strongest full-context baseline. Overall, GistRank provides an effective means to enhance efficiency for searches in real-world environments.

0.24

Look at Where You Think: Localization-Head Alignment for Fine-Grained Perception in Thinking VLMs

cs.CV ·2026-10-29 ·Yuan Wang, Wei Wu, Hongning Wang
Abstract · 23 pages, including references and appendices

Thinking vision-language models reason for thousands of tokens, yet still miss small details in high-resolution images. Zooming methods pay for extra inference passes, and outcome rewards never tell the model where to look at. In Qwen3.5-4B, we find that the bottleneck is selection, not seeing: a few attention heads, which we call localization heads, pick the evidence out of thousands of visual tokens; ablating four of them costs 13.2 points more than ablating four random heads, and their focus fades as the reasoning grows. We therefore propose Localization-Head Alignment (LHA), which makes attention a training signal: whenever the reasoning states a visual fact, these heads learn to look at the focal region rather than the background, as an auxiliary loss in both SFT and RL training. On benchmarks of V*, HR-Bench 4K and TreeBench, LHA lifts Qwen3.5-4B from 74.1% to 76.6% accuracy, the best single-pass 4B result we evaluate, ahead of multi-pass zooming and released 7B models. Alignment alone adds 1.8 points over otherwise identical training while shortening reasoning by 20%, and the gains carry over to Qwen2.5-VL-3B.

0.24

TactoHand: Learning Geometry-Based Dense Hand-Surface Interaction Fields from Video

cs.CV ·2026-10-29 ·Ziheng Xi, Weibo Li, Jiayi Yu, Sipeng Zheng, Wanpeng Zhang +4 more
code/data
Abstract · 23 pages, including references and appendices

Understanding how humans interact with the world requires a representation of physical interactions at the hand surface. Tactile sensing captures these interactions, but collecting tactile data at scale requires specialized hardware. We introduce TactoHand, a model that predicts three dense hand-surface interaction fields---contact, proximity, and 3D force---from monocular video. We derive geometry-based contact and proximity labels and motion-derived force labels from eligible training data within a curated collection spanning approximately 50 million source frames. TactoHand uses a temporal transformer to aggregate visual evidence across frames, while three learned queries decode the corresponding hand-surface fields. A differentiable physics layer further constrains force predictions using observed object motion. Across in-domain and out-of-domain benchmarks, TactoHand improves contact and proximity estimation over prior methods. In addition, physics supervision reduces position and rotation errors in motion rollouts, and the predicted contact and proximity fields improve downstream hand-object reconstruction. Together, these results show a scalable, low-cost path to learning hand-surface interaction fields without collecting tactile measurements. Code will be released at https://github.com/zhenqis123/Tactohand.

0.24

Apriel Voice 27B Preview: An Open Model for Speech, Reasoning, and Tool Use

cs.SD cs.CLcs.LGeess.AS ·2026-10-29 ·Aman Tiwari, Akshay Kalkunte, Akintunde Oladipo, Jash Shah, Tara Bogavelli +26 more
community pick
Abstract

A capable voice assistant must do more than recognize words; it must understand requests, reason over context, sustain conversations, and act through tools. We present Apriel Voice 27B Preview, an open speech-language model that unifies these capabilities within a shared reasoning backbone while retaining text and image input modalities. Built on the post-trained Gemma 4 26B-A4B checkpoint, Apriel Voice directly processes speech and generates both responses and tool calls without requiring a separate dialogue model. We introduce a two-stage training procedure to adapt the model to speech. The first stage establishes the speech interface, while the second jointly adapts the projector and backbone through supervised fine-tuning, text replay, and source-dependent cross-entropy and forward Kullback--Leibler divergence supervision. Automatically constructed speech-conditioned dialogues provide conversational training data, while local teacher serving, overlapping teacher and student computation, and fused full-vocabulary losses enable efficient online distillation. In thinking mode, Apriel Voice achieves a score of 98.90 on Big Bench Audio, 81.13 on Speech-BFCL, and a task success rate of 65.51% on $τ$-Voice, with a mean latency of 1.91 seconds. It also achieves 51.27% English EVA accuracy and 74.13% experience pass@1. Finally, we deploy Apriel Voice 27B Preview on a custom serving platform to support interactive voice interactions.

0.24

CoFiBreak: A Coarse-to-Fine Diversity Reward for Effective Multi-Turn Jailbreak

cs.CL ·2026-10-29 ·Jialin Song, Xiaodong Liu, Weiwei Yang, Mingqian Feng, Chengquan Guo +5 more
Abstract · 10 pages main text, 24 pages total, 8 figures, 12 tables

Automated red teaming trains an attacker large language model (LLM) to elicit harmful responses from a victim LLM, but reinforcement learning can collapse attackers onto a narrow set of prompts when attack success is the only reward. Such a collapse of attack diversity can bias safety evaluation toward repeatedly testing near-duplicate attack patterns rather than the underlying harmful intents. Existing methods often trade off attack success or inference cost for greater content or strategy diversity. We propose CoFiBreak, a coarse-to-fine diversity reward for Group Relative Policy Optimization (GRPO) that improves attack diversity while preserving strong attack effectiveness and low inference cost. The coarse reward broadens which strategies the attacker uses by encouraging strategies underrepresented within each GRPO group, while the fine reward broadens how those strategies are realized by discouraging reuse of the same personas, framings, and output formats after harmful-intent information is removed from prompt embeddings. Diversity rewards apply only to attacks that already exceed an effectiveness threshold, so diversity cannot be gained through ineffective attacks. Since both rewards are computed only during training, CoFiBreak also retains inference efficiency. Trained with Llama3.1-8B-Instruct as the victim, CoFiBreak discovers 1.6-2.0x as many distinct successful fine-grained strategies as the strongest baseline across frontier victims on HarmBench. It reaches 98.8% ASR@10 (attack success rate with 10 generations per intent) on AdvBench with the highest diversity score among all compared methods, which is 2.1x that of the strongest-attacking baseline. The resulting attacks transfer to unseen frontier LLMs, reaching 71.0% ASR@10 on AdvBench with the unseen victim Claude-Opus-4.6. The code will be released.

0.23

A Sequential Decision Framework for Evaluating LLM-Guided Search

cs.IR ·2026-10-29 ·Siqing Huo, Andrew Parry, Debasis Ganguly, Mark Smucker, Charles L. A. Clarke
Abstract

Recent systems applying retrieval-augmented generation (RAG) and "deep research" iteratively generate queries, retrieve documents, and examine their contents, using the resulting information to guide subsequent search steps. When this process is carried out autonomously rather than by a human searcher, the system must make explicit decisions about which search trajectories to explore and what objective the search is intended to optimize. We introduce a simple framework for reasoning about such search policies, providing a principled approach to evaluation. The framework models the search environment as a set of ranked lists and represents behavior using two primitive actions: select, which advances judgment through an existing ranking, and spawn, which generates a new ranking through query generation or reformulation. Each action produces a document to inspect, yields information that updates the search state, and contributes a reward for a task-specific objective. Within this abstraction, a variety of existing retrieval strategies can be expressed as policies operating over a common environment, while common retrieval objectives can be represented through reward and weighting functions over the resulting search trajectory, and the resulting formulation admits a natural reinforcement learning interpretation. We illustrate the approach using high-recall information retrieval (HRIR), where the goal is to identify as many relevant documents as possible with limited judging effort, showing how the framework provides a unified way to describe and reason about strategies for both classical retrieval tasks and emerging LLM-guided search systems. Through experiments, we demonstrate that both the underlying retrieval method and the "deep research" policy employed can materially affect performance.

0.23

CrossWeave: Cross-Modal Scene and Instance Retrieval from Sparse 2D-3D Alignment

cs.CV ·2026-10-29 ·Aadith Warrier, Gnana Prakash Punnavajhala, Siddharth Tourani, Muhammad Haris Khan, Avinash Sharma +1 more
code/data
Abstract · Project Page: https://crossweave3d.github.io/

Cross-modal scene and instance retrieval aims to retrieve a corresponding 3D environment or object given query 2D images, serving as a foundational mechanism for spatial reasoning in robotics and AR/VR. Existing methods achieve this by learning alignment only at the level of whole scenes or objects, which provides no supervision for the fine-grained local structure within them. In this work, we show that supervision on a set of sparse 2D-3D correspondences, without scene- or instance-level contrastive objectives, is sufficient to train a single model for both scene- and instance-level retrieval. Specifically, we propose $\textbf{CrossWeave}$, a novel framework for cross-modal retrieval that learns sparse patch-point alignment, which alone outperforms existing methods. An optimal transport-based aggregator over the aligned features gives further gains by clustering structurally redundant regions, freeing descriptor capacity for semantically distinct local features. CrossWeave achieves state-of-the-art performance on both scene- and instance-level retrieval on ScanNet and 3RScan with a $\textit{single model}$, whereas existing methods require separate models for each task. Our results suggest that whole-scene encoding is unnecessary for cross-modal retrieval and that sparse local correspondence supervision is a more effective and flexible alternative.

0.23

xPRT: Cross-Passage Retrieval Pretraining

cs.IR ·2026-10-29 ·Danush Khanna, Himanshu Singhal, Pranjal A. Chitale, Bishal Santra
Abstract · Under Review

Dense retrievers are a foundational component underlying modern RAG and search systems. Training these models typically requires a curated collection of annotated query-document pairs, which are prohibitively expensive to obtain outside a handful of well-resourced domains. This has motivated a line of work that leverages large-scale unsupervised retrieval pretraining from unlabelled text alone. Most such methods learn by optimizing a language modelling or contrastive learning objective over short spans within a document. This raises the question: how can we extract richer learning signals from a single document for retrieval pretraining? To address it, we introduce xPRT (Cross-Passage Retrieval Pretraining), an unsupervised framework that learns from the key information distributed across an entire document and the connections between its parts. We evaluate xPRT against five state-of-the-art objectives on MTEB English v2, continuing pretraining from a public unsupervised checkpoint while holding the corpus, schedule, and training procedure fixed. Across diverse tasks spanning all task types, we observe that xPRT consistently outperforms the strongest baseline by 3.2% on an average. Notably, in this continual pretraining setting, many baseline methods incur substantial performance regressions, whereas our method delivers the largest complementary gains over the initial pretrained checkpoint. Controlled ablations confirm that xPRT's performance gains stem from its proposed cross-passage signals. xPRT demonstrates that a single unlabelled document is a much richer source of retrieval signal than prior work assumed and offers complementary gains to what current retrievers already capture.

0.23

Decompose, Estimate, Select: Coverage-Aware Context Construction for Long-Form RAG

cs.IR ·2026-10-29 ·He Tu, Aleena Bosco Kurumthottam, Torsten Suel
theory
Abstract

Retrieval-augmented generation (RAG) grounds a language model's output in retrieved evidence, but for complex questions spanning multiple aspects of an information need, relevance-based retrieval may return redundant evidence while leaving some aspects unsupported. We propose a coverage-aware retrieval pipeline that decomposes queries into subquestions, ranks candidates with a coverage-trained bi-encoder, predicts a document\,$\times$\,subquestion coverage matrix with a cross-encoder, and selects documents by greedy maximum-marginal coverage. On TREC RAG 2024, our approach improves $α$-nDCG@10, which rewards newly covered nuggets and discounts redundant evidence, from $0.625$ to $0.702$, approaching the $0.718$ gold-nugget upper bound without sacrificing relevance. We then find that ground-truth oracle retrieval, despite substantially higher retrieval coverage, delivers no more nugget coverage to the answer than our estimated selection. This suggests that answer completeness depends not only on what evidence is retrieved, but also on how effectively the generator uses it. To address this, we introduce a \emph{grounded self-feedback loop} in which the model identifies underused evidence from its first answer and generates again without gold supervision, raising $V_{\mathrm{strict}}$, the fraction of vital reference nuggets fully supported by the answer, by $0.086$ over two passes while largely preserving faithfulness.

0.23

AVID: A Weakly Aligned RGB-Infrared Dataset and Perturbation Benchmark for Tiny Aerial Targets

cs.CV ·2026-10-29 ·Dharini Raghavan, Vishnu Rao, Sethu Selvi S
benchmark
Abstract · 23 pages, 8 figures

Long-range detection of small drones is complicated by their visual similarity to birds and by the limited detail available from ground-based cameras. Combining visible and infrared observations offers complementary information, but their usefulness depends on target size, cross-modal alignment, and image quality. We present AVID, a benchmark for studying these factors using real RGB-infrared bird-drone surveillance footage originating from the Drone Detection Dataset and curated for the IEEE ICIP 2025 Video and Image Processing Cup. AVID comprises 32,040 corresponding frame pairs from 115 videos, with 35,952 existing bird and drone annotations at $320\times256$ resolution. Targets occupy an unusually narrow size range: 68.1\% have a geometric-mean bounding-box side of 8--16 pixels, and 99.0\% meet the COCO small-object criterion. To characterize the geometric difficulty of combining the modalities, we introduce an object-relative alignment ratio and an alignment burden curve. In a sample of 300 frame pairs, 32.3\% have estimated global translations greater than 5 pixels; among the 264 pairs with matching annotations, 16.3\% exceed one median target side. We further introduce AVID-C, which simulates seven families of image degradation at five severity levels independently in RGB, infrared, or both. A fixed-feature analysis examines how scale, degradation, and added cross-modal shifts affect fitted bird-drone class overlap. At 12--16 pixels, the Bhattacharyya error bounds are 0.148 for RGB, 0.198 for infrared, and 0.046 for their combined features. These values describe separation in the specified feature and density models, not detector accuracy. Together, AVID and AVID-C support scale-aware, alignment-aware, and sensor-specific studies of tiny aerial targets.

0.23

Aligning Medical Vision Language Models at the Capability Boundary with Entropy Adaptive Rubric Grounding

cs.CV cs.AI ·2026-10-29 ·Hongyu Bai, Jiaqi Zeng, Chunzheng Zhu
published
Abstract · Accepted at ACM Multimedia Asia 2026 (MMAsia 2026). 10 pages, 4 figures

Medical diagnostic reasoning requires verifiable links between visual evidence and clinical knowledge, yet current medical vision language models are aligned with objectives that reward the wrong behavior. Uniform supervised fine tuning lets easy cases dominate updates while clinically critical boundary cases saturate prematurely. Sparse rewards based only on final-answer correctness cannot assess anatomical grounding or clinical logic, inviting reward hacking through answer shortcuts. We propose MedCBR, a two-stage alignment framework coupling boundary-aware preference optimization with rubric-grounded reinforcement learning. First, we construct counterfactual boundary preferences at the model's capability frontier using empirical pass rates and preference pairs that share reasoning prefixes but diverge at decisive diagnostic branches. Difficulty Weighted Boundary Preference Optimization (DBPO) concentrates learning on this calibrated boundary region. Second, we distill reusable clinical rubrics from boundary trajectories and activate them according to prediction entropy. Entropy Gated Grounded Policy Optimization (EGPO) rewards answer correctness, anatomical grounding, and reasoning fidelity while suppressing rubric noise on reliable samples. A reference policy anchored to the boundary-aligned model couples population-level boundary priors with instance-level uncertainty, making diagnostic reasoning more accurate and clinically grounded. Experiments on six medical benchmarks show consistent improvements over strong general, specialized, and reinforcement learning baselines, reaching 61.2% average accuracy with the largest gains on challenging differential diagnosis tasks. Ablations show super-additive gains: combining the two stages exceeds the sum of their individual improvements.

0.23

PhysAg3D: Bridging 3D Generative and Agentic Models for High-Fidelity and Physically-Interactive Assets

cs.CV ·2026-10-29 ·Bi'an Du, Fanpeng Meng, Jiemin Fang, Chen Yang, Pufan Li +3 more
code/data
Abstract

Generating interactive 3D assets from a single image requires realistic appearance, fidelity to the depicted object, and plausible physical behavior. Yet detailed geometry alone does not determine how an object should move or deform. We introduce PhysAg3D, which couples 3D generative models with agentic reasoning and tool use. A vision-language agent first infers semantic parts, candidate interactions, and physical priors from the image; a 3D generator then produces a textured whole-object mesh from the same image. A hierarchical interaction representation links these revisable priors to mechanisms and the generated surfaces they control, organizing interaction construction and refinement. Guided by this representation and observations of the mesh, the agent uses geometric and physical tools to construct parts, joints, deformable regions, and surface bindings while retaining useful generated geometry and materials. Rendered interactions and execution measurements guide targeted revisions within a fixed budget. The framework accommodates both rigid articulation and compliant deformation. Quantitative and qualitative evaluations on real photographs and external benchmarks demonstrate detailed, image-aligned assets with plausible interactions. Ablations support bounded refinement and source-surface preservation. Project Page: supra-3d.github.io/PhysAg3D.

0.23

PIVOT: Progressive Intervention via Online Teacher Correction

cs.CL ·2026-10-29 ·Zhihao Yang, Boyuan Zheng, Puhui Tan, Yukun Chen, Kuan Zhang +5 more
code/data
Abstract · 19 pages, 9 figures, 6 tables. Project page: https://zhihaoyang30.github.io/PIVOT/

Reinforcement learning from verifiable rewards (RLVR) is limited by sparse outcome-level feedback on hard reasoning tasks: when all rollouts in a group fail, rewards provide no learning signal. Stronger teachers can provide additional guidance, but this raises a central question: how can such signals be incorporated into RLVR to support successful rollouts while remaining learnable for the student? We distinguish two complementary mechanisms: trajectory-level intervention, which improves access to reward-bearing trajectories, and token-level supervision, which provides dense targets to guide student learning. Our controlled experiments show that either mechanism alone can remain insufficient, motivating their combination. We therefore introduce PIVOT, which inserts teacher-generated corrections into student rollouts and applies token-level KL supervision to these segments alongside the RLVR objective. Beyond their combination, we systematically analyze where to intervene and how to supervise, yielding a recipe for effective guidance. Across hard logic games and mathematical reasoning benchmarks, PIVOT consistently outperforms existing methods, nearly doubling mean avg@8 on logic games from 19.5% for the best-performing baseline to 35.6%.

0.23

AKDQ: Adaptive Key-Delta Quantization for Low-Bitrate RVQ-based Neural Speech Codecs

eess.AS ·2026-10-29 ·Xiao-Hang Jiang, Yang Ai, Zhen-Hua Ling, Ji Wu
published
Abstract · SPL accepted

Low-bitrate speech coding is essential for bandwidth-constrained channels and on-device applications, but conventional residual vector quantization (RVQ)-based codecs employ a fixed number of codebooks per frame regardless of content. Adjacent speech frames often exhibit strong similarity, and quantizing them independently with the same scheme results in redundant encoding, thereby wasting bitrate. Therefore, we propose Adaptive Key-Delta Quantization (AKDQ), which replaces the fixed per-frame VQ allocation of RVQ with an adaptive frame-wise key-or-delta policy. AKDQ is a plug-in strategy that requires no modification to the codec's encoder-decoder backbone. Specifically, in AKDQ, each frame is classified as either a key-frame (K-frame) quantized with all VQs, or a delta-frame (D-frame) quantized with only a few VQs on a delta correction from the previous reconstruction, thereby reducing bitrate. The K/D decision is made by a budget-aware causal scheduler through a bit-bank mechanism, keeping the system streamable with no additional lookahead. Experiments on both causal and non-causal codec families demonstrate that AKDQ effectively reduces bitrate redundancy. For instance, on DAC, using AKDQ at 1.5 kbps achieves reconstruction quality comparable to using the baseline RVQ at 2.0 kbps, saving 25% of the bitrate.

0.23

TalkFlow: Structured Latent Flow Matching for Speech Driven Holistic Motion Generation

cs.GR cs.CV ·2026-10-29 ·Xinmu Wang, Fumi Honda, Xiang Gao, Xianfeng Gu, Xiyun Song +1 more
publishedcode/data
Abstract

Holistic co-speech motion generation aims to synthesize coordinated facial expressions, hand gestures, and body movements from speech, which is challenging due to heterogeneous part-specific dynamics, long-range temporal coherence, and cross-part coordination. Directly modeling high-dimensional motion can be unstable, while overly compact or unstructured latent spaces may lose fine-grained motion details. We present \textbf{TalkFlow}, a structured latent flow framework for speech-driven holistic motion generation. TalkFlow learns a continuous three-stream latent representation for upper-body motion, lower-body motion, and facial expressions, preserving part-specific dynamics without forcing heterogeneous motion components into a single latent space. To maintain holistic coordination, lightweight cross-stream interactions are incorporated in both latent representation learning and flow-based generation. Conditioned on multi-level speech features, TalkFlow performs speech-conditioned latent flow matching, learning a continuous transport process from source latents to realistic motion latents. Experiments on the BEAT2 dataset demonstrate strong performance across facial expression accuracy, body gesture quality, and speech-motion synchronization. Qualitative comparisons and user studies further show that TalkFlow generates more natural, expressive, and speech-aligned holistic motion than state-of-the-art methods. The project page is available at https://xinmuwang.github.io/TalkFlow/.

0.23

Read More, Reason Deeper: Latent Workspaces for Long-Sequence Recommendation

cs.IR ·2026-10-29 ·Renzhi Wu, Li Sheng, Zikun Cui, Junjie Yang, Tai Guo +6 more
Abstract

Long-sequence recommendation can improve along two axes: \emph{evidence scaling} exposes longer histories, while \emph{reasoning scaling} adds layers that interpret that evidence for a candidate. Existing models couple them: sequence encoders propagate every added event through depth, whereas one-shot selection prevents later layers from revisiting omitted evidence. We introduce the Latent Workspace Ranker (LWR), which keeps typed evidence in persistent banks and propagates a fixed-size candidate-local workspace. Each layer rereads the banks from its evolving workspace and archives compact post-interaction state, allowing history length and reasoning depth to grow independently. We compare four architectures across various history--depth configurations using evidence scaling at fixed depth, reasoning scaling at fixed 10K history, and their joint compute-efficient frontier. LWR leads all families at long histories, continues improving with reasoning compute, and converts additional training compute into quality gains at the fastest fitted rate on the joint frontier. At 10K history, it improves NE by 0.19% over HSTU while using 69.6% fewer forward GFLOP/request. It also outperforms existing baselines on two billion-scale ranking surfaces. These results show that separating addressable evidence from propagated state makes reading more and reasoning deeper independently controllable and jointly useful.

0.22

How Specific Do I Need to Be? Effects of User Query Underspecification on Deep Research Agents

cs.CL ·2026-10-29 ·Yi Wen, Xiangjue Dong, Bingxuan Li, Hangxiao Zhu, Leo Zhicheng Liu +2 more
Abstract · 16 pages, 8 figures

Deep research agents (DRAs) conduct long-horizon reasoning, search, and literature synthesis to provide long-form answers to complex, domain-specific questions. These agents are typically trained and evaluated on fully specified user queries. However, real-world user queries can often be underspecified. We introduce UNDERSPEC-DR, a framework for systematically evaluating how underspecified queries affect agents' final answer quality and intermediate search behavior. Across three DRAs we evaluated, underspecification substantially reduces how well DRA's answers fulfill users' information needs, even though they remain grounded in cited evidence. Our analysis of intermediate search behavior suggests a possible explanation: DRAs conduct broader searches that draw less on the original user query but rely more on their learned prior during training, which can be misaligned with users' intent. Our findings reveal the vulnerability of current DRAs to underspecified user queries and inform future efforts to build DRAs that proactively recognize when additional user input is needed to reliably complete deep research tasks.

0.22

Towards Full-Text Scholarly Argument Mining using Entity-Anchored Claim Extraction

cs.CL cs.IR ·2026-10-29 ·Bhuvanesh Verma, Andy Lüking, Mounika Marreddy, Leon Lukas Hammerla, Alexander Mehler
Abstract

Argument mining in full-text scholarly articles remains an important yet challenging task. This difficulty stems not only from the need to model long-distance dependencies but also from the substantial effort required to annotate sentence-level argumentative components throughout lengthy documents. To address these challenges, we introduce a novel approach that leverages scientific entity mentions (e.g., tasks, datasets, and models) to identify and structure claims in scholarly literature. We validate our approach through a case study involving 50 full-text research papers on sentiment analysis. We first establish a reliable human gold standard (Krippendorff's $α= 0.80$), identifying 113 claim-bearing sentences and extracting 199 structured claim records from 300 candidate sentences. To address the annotation bottleneck, we employ a majority-voting ensemble of large language models (LLMs) to generate a silver-standard dataset comprising 1,487 claims. Encoders fine-tuned exclusively on the silver-standard dataset outperform those trained on a small human-annotated gold-standard dataset and demonstrate strong zero-shot transfer to NLP papers outside the training corpus.

0.22

Toward Contrast-Free Breast MRI: Improving Cancer Detection With Virtual Contrast Enhanced Imaging

eess.IV cs.CV ·2026-10-29 ·Hongxi Chen, Sunwoo Kwak, Sanyukta Adap, Katerina Dodelzon, Paula Feldman +1 more
Abstract · 5 pages, 2 figures, 1 table. Submitted to ISBI 2027

Although contrast-enhanced (CE) breast MRI is an important tool for cancer detection, intravenous gadolinium administration introduces additional considerations for patients and clinical workflows. AI based virtual contrast enhanced (VCE) MRI could help identify patients for whom acquired CE imaging may be avoidable without compromising diagnostic performance. We developed a framework that combines non contrast (NC) and VCE MRI predictions for cancer detection. On the held-out test set, cancer detection using NC MRI alone achieved an AUROC of 0.865. Combining the predictions from NC and VCE MRI improved AUROC to 0.901. Building on this complementary predictive value, we evaluated patient-level routing to progressively replace acquired CE examinations with the NC plus VCE pathway, prioritizing patients with the most confident NC predictions. This simulation suggests that diagnostic performance can be largely preserved while limiting acquired CE imaging to approximately 40% of patients. Together, these findings demonstrate the complementary value of VCE MRI and suggest its potential to support selective contrast use in breast cancer screening.

0.22

Better Retrievers, Worse Labels: How Incomplete Relevance Judgments Distort Abstention Evaluation in Conversational RAG

cs.CL ·2026-10-29 ·Ivan Kochergin, Vasily Konovalov
Abstract

Conversational RAG benchmarks such as MTRAG and RETECO evaluate retrieval and generation end to end: each system answers from its own retrieval, so abstention must be scored against whether each retrieved context is sufficient. A convenient shortcut reads sufficiency off the relevance judgments and treats a context without gold documents as unanswerable. We re-run the official baselines exactly and test this shortcut on RECOR (RETECO track~2) with four retrievers (nDCG@10 from .18 to .59) and three Qwen3 readers. Two open LLM judges and a panel of four small open models label each context by content, and a model of another family audits 250 contexts blind, with adjudication. The stronger the retriever, the larger the share of no-gold contexts that still contain the answer (28\% to 72\%) and the lower the shortcut's precision (.68 to .10; .70 to .21 under the audit). A reader agrees: Qwen3-4B answers 62--67\% of these false misses correctly, against 23--30\% of no-gold contexts judged insufficient and 1\% with no documents. In most false misses, the answer comes from a hard negative mined for the same question, another turn's gold or a neighbouring chunk of the gold page. On MTRAG, whose judgments come from annotators, precision drops from .72 to .50. Under the shortcut, answerability AUROC is up to .19 too low, and the best sufficiency detector appears to be a retrieval score that never reads the documents. We release code, runs and labels, and recommend scoring refusals against content labels, not the presence of gold.

0.22

Tokenizer Tokens Tell Training Tales

cs.CL ·2026-10-29 ·Nishita Jain, Parth Upman, Shreyank N Gowda
benchmark
Abstract

Tokenizers are usually treated as fixed preprocessing tools, but they are trained on text and can retain traces of that text in their vocabularies. We ask whether a BPE tokenizer can reveal if a candidate dataset was part of its training corpus. The key observation is simple. Rare strings become tokens only when enough corpus frequency pushes them over the merge threshold. This suggests that membership signal should appear in rare tokens used by the candidate dataset, rather than in the order in which merges were made. We formalize this under a threshold model of BPE inclusion, showing that the likelihood-ratio test assigns its largest weights to tokens that are rare in the background but frequent in the candidate dataset. The same model suggests that merge rank and merge-graph position should not add a first-order membership signal once token existence and dataset frequency are known. We instantiate this view as RTAM-Fusion, which calibrates rare-token mass with shadow tokenizers and combines signals across rarity levels. On the tokenizer membership benchmark, RTAM-Fusion improves AUC and balanced accuracy over Vocabulary Overlap while using fewer shadow tokenizers. Existing defenses also continue to leak membership signal against adaptive attackers at production vocabulary sizes. Code: https://anonymous.4open.science/r/RTAM-Fusion-F7EE/

0.22

TAPE: Learning General Audio Effect Representations

cs.SD cs.AIcs.LGeess.AS ·2026-10-29 ·Phillip Long, Yongyi Zang, Qihui Yang, Christodoulos Benetatos, Junfei Liu +3 more
Abstract · Submitted to Transactions of Machine Learning Research; 25 pages, 5 figures, 18 tables

Musicians and mixing engineers often can decipher the signal processing chain they hear on a reference recording, then adapt it to new material. We present TAPE (Transformations of Audio Parameter Embeddings), which estimates this processing from the processed reference audio alone. An encoder trained to identify processors in eight major effect families maps the reference to an embedding. A differentiable digital signal processing (DSP) chain decodes the embedding into editable parameters in one forward pass and applies them to a different unprocessed recording. Its modules are equalization, compression, waveshaping, a small residual network, and reverberation. TAPE achieves lower spectral error than DeepAFx-ST, the previous state of the art for single-step regression, and lower spectral error than per-example search over a chain approximating ST-ITO's, while running 275$\times$ faster than that search. Because the decoder outputs DSP parameters, its chain can also run causally: a C++ implementation with the trained weights processed 128-sample buffers 1.8$\times$ faster than real time, with no missed deadlines.

0.22

FutureWorld: A Plug-and-Play Causal Next-State Predictor for Interactive Video World Exploration

cs.CV ·2026-10-29 ·Guangyuan Li, Kelu Yao, Shuolin Xu, Xinlei Yu, Wei Xing +4 more
code/data
Abstract · 17 pages, 6 figures

Interactive video world models aim to follow user actions while maintaining scene consistency over extended interactions. History-based methods reuse generated latents to preserve scene content, but three limitations of this reuse contribute to error recirculation. 1) The latent generated at each interaction is retained as history, potentially containing blurred details, geometric distortions, or semantic drift. 2) Although retrieved historical content matches the current scene, it can still contain errors that are carried into the next chunk. 3) New outputs then become history for later interactions, potentially reinforcing the same errors through repeated reuse. To limit this cycle, we aim to predict the future scene state from the current generated latent and the next action before generating the next interaction's video chunk. This future-state prior provides guidance for the next chunk, with the aim of reducing the errors carried forward from history. To this end, we propose a plug-and-play Causal Next-Chunk Predictor (C-NCP) with cross-model adaptability and prediction-guided modulation. We integrate C-NCP into four base generators and evaluate the augmented models across three benchmarks for interactive long-video generation and action-conditioned world modeling. Extensive experiments demonstrate reduced error accumulation and improved long-horizon consistency across diverse generation settings. The project page is at https://GuangYuanKK.github.io/futureworld.

0.21

MOSAIC-Flow: Orthogonal-Axis and Status-Conditioned Fusion for Missing-Sequence Completion in Brain Tumor MRI

cs.CV ·2026-10-29 ·Pengli Zhu, Zhihao Peng, Yixuan Yuan
code/data
Abstract · 5 pages, 4 figures

Missing-sequence completion is important for comprehensive brain tumor characterization, but cross-modal fusion can obscure sequence-specific contrast and lose fine-grained three-dimensional tumor detail. To address this challenge, we introduce Modality-aware Orthogonal-axis Status-conditioned Attentive Integration for Completion (MOSAIC-Flow), a unified framework for missing-sequence completion that learns a target-conditioned velocity field while keeping the seven acquired volumes fixed throughout the transport. Specifically, modality-aware stems independently encode each sequence, and orthogonal-axis refinement captures axial, coronal, and sagittal dependencies while retaining the three-dimensional representation. Status-conditioned prototype fusion uses compact modality and availability tokens to route and modulate pooled queries from the fused visual representation, enabling target-adaptive conditioning without dense voxel-to-voxel attention. A conditional flow matching objective transports target-channel noise toward the single missing sequence, while four Heun trajectories are averaged to produce the completion. Evaluation across structural, diffusion, and contrast-enhanced MRI in UCSF-PDGM demonstrates consistent whole-brain and tumor-region fidelity across all eight single-target completion tasks, with the strongest performance on structural sequences. Code is available at https://github.com/CUHK-AIM-Group/MOSAIC-FLOW.

0.21

CAD-WAM: Anticipating Geometry for Text-to-CAD Generation

cs.CV cs.GR ·2026-10-29 ·Zhengyang Li, Weijie Zheng, Yuxin Cui, Sijie Wang, Wenjie Feng +1 more
code/data
Abstract

Generating editable computer-aided design (CAD) models from text requires selecting operations that produce the intended changes to the geometry constructed so far. An incorrect operation can compromise subsequent steps, making reliable multi-step generation challenging. Existing methods based on large language models (LLMs) primarily learn to predict commands, without explicit supervision of the geometric states those commands should produce. Inspired by world-action models, we introduce CAD-WAM, which couples geometry-conditioned action generation with next-state prediction. We define complete modeling operations as actions in a CAD domain-specific language (DSL) and encode surface point clouds of intermediate solids into compact latent states using a pretrained variational autoencoder. An LLM backbone generates the next action from the target description and current geometric state, while an auxiliary head on the same backbone predicts the next state to provide geometric supervision during training. We construct step-wise training trajectories from parametric CAD models in the DeepCAD dataset, pairing each trajectory with an expert-level textual description. At inference, each generated action is executed by a CAD kernel, and the resulting solid is re-encoded to inform the next decision. On the DeepCAD test set, CAD-WAM outperforms the task-specific baselines in geometric accuracy and output validity, with larger gains on models requiring more operations, and produces more accurate geometry than general-purpose LLMs prompted with the same DSL. Ablations further show that next-state supervision reduces mean Chamfer distance by 13.2\% relative to an otherwise identical model trained only to predict actions. Code is available at https://github.com/Lzy-1037/CAD-WAM.

0.21

Li$^{2}$SAO: Lifting Light Single-Rollout Asynchronous RL Optimization via Group-Relative Actor-Critic Bootstrapping

cs.CL cs.LG ·2026-10-29 ·Zichen Bai, Yanli Wang, Yang Liu, Junhao Shi, Xueqin Liu +7 more
code/data
Abstract

Asynchronous reinforcement learning (RL) is becoming a mainstream paradigm for post-training large language models (LLMs), as it improves training efficiency by updating the model as rollouts arrive. Recent work argues that group-wise sampling in GRPO does not naturally fit asynchronous RL training because rollout lengths follow a long-tailed distribution, and proposes Single-rollout Asynchronous Optimization (SAO) to improve stability and off-policy behavior, especially on agentic tasks. However, SAO relies on a closed-source value model obtained from scaling value pretraining. In this paper, we analyze GRPO and LiSAO, our lightweight reproduction of SAO without the pretrained value model, on single-turn reasoning and multi-turn agentic tasks. We find that GRPO outperforms LiSAO on reasoning tasks, while the reverse holds on agentic tasks. Since the value model of LiSAO is not yet well trained , LiSAO lags behind GRPO substantially in early training steps. To address this issue, we propose Li$^{2}$SAO, which Lifts Light Single-rollout Asynchronous Optimization with group-relative actor-critic bootstrapping: while the critic is cold, a group-relative advantage trains the actor and a group-centered return trains the critic, after which training switches to single-rollout GAE. Experiments show that Li$^{2}$SAO catches up with or surpasses GRPO in early training and further raises the later performance ceiling. Code is available at https://github.com/jd-cc-coder/Li2SAO.

0.21

It Takes Two to Probe, Thousands to Retrieve: Putting Vision--Language Probes to the Retrieval Test

cs.CV cs.CL ·2026-10-29 ·Ivana Beňová, Michal Gregor, Marián Šimko
benchmark
Abstract · 12 pages, 2 figures, 3 tables

Vision--language models (VLMs) are tested for fine-grained understanding with image--text matching probes, a model must prefer the correct image or caption over handcrafted negative. In retrieval, the correct item must beat thousands of alternatives. We measure how these two evaluation methods relate by using two probing benchmarks, \svo{} and SugarCrepe, for retrieval tasks. For 100 image and caption queries we judge the top-10 retrievals from full dataset of six VLMs, with human annotators and an LLM assessor. We than compare how well models rank pool of items on the same queries, while changing only where the negatives comes from: the benchmark, a random draw, or the judged-incorrect retrievals of other models. On \svo{} the accuracy of retrieving a positive over handcrafted negatives is significantly higher than for retrieved ones (the correct item is preferred in 85\% against 77\% of the time). Using pool of ten retrieved negatives leads to same accuracy as when one uses pool of 200 to 430 random negatives. On SugarCrepe the order is reversed: its handcrafted negatives are harder than the retrieved ones. Incomplete relevance labels explain most of the low retrieval scores on \svo{}, where the top-ranked item is correct for 54--90\% of the queries under human labels against 6--21\% under the benchmark's labels. However, we detect no association between passing the probe for a query and retrieving a correct item first. The LLM assessor o3 agrees with individual annotators ($κ$ 0.57--0.61) slightly less than they agree with each other (0.64--0.68).

0.21

Benchmarking color judgment in multimodal LLMs with CIE-computable ground truth

cs.CV ·2026-10-29 ·Shuwei Yue, Kaida Xiao, Zhihua Wang, Qiao Chen
benchmark
Abstract

Multimodal large language models (MLLMs) are now used as zero-shot judges of image quality, but their color judgments are rarely tested with colorimetric ground truth. We present a five-task benchmark based on the CIE colorimetric framework: AWB pair judgment, skin-tone bias direction, CCT estimation, CIE Delta E 2000 pair ordering, and color naming under changing illuminants. We test eight MLLM endpoints, including four Qwen3-VL variants, GLM-4.5V, GPT-5.4, Claude-Opus-4.7, and Gemini-3.1-Pro. The models often solve coarse binary AWB judgments, but they remain weak on numeric CCT estimation and on some fine color-difference judgments. On ColorChecker-24 stimuli, the three closed-weight frontier models reach CCT MAE of 715-1056 K, compared with 1358-2133 K for the five open-weight models. A non-ColorChecker control shows that this advantage is partly template-dependent. The results suggest that current MLLMs have useful color words and coarse visual priors, but their color judgments are not reliably calibrated to CIE-computable color differences.

0.21

Drone-PUR: From Visual Clue Discovery to Understanding and Reasoning in UAV Scenes

cs.CV ·2026-10-29 ·Jiahe Wu, Bing Cao, Xingyue Guo, Ruipu Zhao, Beiming Yu +3 more
benchmarkdatasetcode/data
Abstract · 17 pages, 13 figures

The reasoning capabilities of vision-language models (VLMs) can be limited by incomplete or inaccurate visual observations in the general domain. In unmanned aerial vehicle (UAV) scenes, small targets, subtle attributes, and changing viewpoints further complicate perception and understanding. Therefore, we introduce Drone-PUR, a dataset and benchmark organized around visual clue discovery, interpretation, and integration, corresponding to the three capability dimensions of perception, understanding, and reasoning. Reasoning samples are designed to require models to combine multiple visual clues to derive visually grounded conclusions. Its 11 tasks provide 175k training and 20k test samples spanning single images, image sequences, cross-modal images, and multiple viewpoints. All test samples undergo human review. Evaluating 22 general-purpose VLMs reveals persistent limitations in spatial judgment and cross-modal and cross-view reasoning. Across models, stronger perception is associated with better reasoning, while single-task fine-tuning further reveals selective bidirectional transfer between these capabilities. To examine domain adaptation, we develop UAV-PUR from Qwen3-VL through staged fine-tuning with auxiliary supervision. Its 4B and 8B variants improve overall multiple-choice accuracy by 18.33 and 17.52 percentage points over their pretrained counterparts, respectively. The dataset and project resources are available at https://drone-pur.github.io.

0.21

EditAgent: Rethinking Agentic Image Editing

cs.CV ·2026-10-29 ·Haiyang Xu, Tianyu Wang, Jialu Li, Jiteng Mu, Yuanjun Xiong +3 more
benchmarkposition
Abstract

Agentic image editors decompose complex requests into steps, but when a step fails, they mostly retry with the same narrow set of tools and a judge that scores each output by eye directly. We argue that such agents are bounded by two spaces: the action space of what the agent can do, and the observation space of what its judge can see. On our expert-annotated benchmark, a general-purpose editor yields no usable result in over 40% of the cases, and no judge is reliable across edit types. We present EditAgent, which expands both spaces. It routes each step to general editors and specialized methods, and attaches to each candidate a verification record. With these records, off-the-shelf VLM judges outperform finetuned reward models. Candidates are generated in parallel and judged in one listwise call: experts prefer this to sequential refinement in 78.5% of decided comparisons, at no more than half the judge calls. EditAgent improves two open-weight editors on three benchmarks and serves as a data engine for training stronger editors.

0.21

CS-LDM: Clinically Guided Structure-Aware Latent Diffusion for Knee Radiographic Forecasting

eess.IV ·2026-10-29 ·Shaotian Li, Chenhui Wang, Wenhua Wu, Yanqiu Wu, Xiaohan Yu
Abstract

Knee osteoarthritis (KOA) is a leading cause of physical disability and currently has no curative clinical treatment. Early diagnosis and accurate prognosis are therefore vital for managing its progression. Existing methods mainly predict future severity grades and provide limited visual explanation of structural change. Recent generation-guided approaches synthesize follow-up radiographs for progression assessment, but generated images are often judged by predictive accuracy or image realism, so they may satisfy a grading model while remaining morphologically unreliable. We therefore propose CS-LDM, a clinically guided, structure-aware latent diffusion framework that forecasts patient-specific future knee radiographs from the current X-ray and clinically available current KL grade, using LoRA adaptation, current-grade conditioning, and frozen-teacher diagnostic distillation. Experiments on the public Osteoarthritis Initiative (OAI) dataset show that CS-LDM improves future severity-grade consistency while maintaining competitive image fidelity. Further ordinal, OARSI, ROI-level, and stable/progressed analyses show that radiographic forecasting should be evaluated by balancing diagnostic consistency, anatomical plausibility, and structural preservation.

0.20

SonoWorld4D: From One Video to a 4D Audio-Visual Scene via Diffusion Synchronization

cs.CV cs.SD ·2026-10-29 ·Derong Jin, Xiyi Chen, Ming C. Lin, Ruohan Gao
code/data
Abstract · Project Page: https://humathe.github.io/sonoworld4d/

Recent progress in visual scene generation has extended beyond static 3D scenes toward the more challenging problem of generating dynamic 4D scenes from minimal input. However, sound, despite being intrinsically coupled with dynamic objects and events, remains largely unmodeled in prior work. We introduce Video-to-4DAVScene, a new task that turns a video into an explorable, time-varying 4D audio-visual scene, and present the SonoWorld4D framework for this task. Given a monocular video, our method generates a 360 static background, reconstructs explicit motion trajectories of dynamic, sound-emitting foreground objects, and synthesizes a physically-grounded sound field consistent across multiple sources through dual-space audio diffusion synchronization. The resulting 4D scene supports real-time novel-view visual and spatial audio rendering. Evaluation on real-world dynamic scenes and a user study demonstrates SonoWorld4D generates 4D audio-visual scenes with perceptually realistic, physically grounded spatial audio consistent with scene geometry and visual content.

0.20

DTP-OPSD: Reinforcement Learning with Distractor-Aware Temporal-Privilege On-Policy Self-Distillation for Temporal Video Grounding

cs.CV cs.AIcs.LG ·2026-10-29 ·Zhaoyu Chen, Yongwei Nie
code/data
Abstract

Temporal Video Grounding (TVG) aims to localize the start and end times of an event described by a language query in a video. The task becomes particularly challenging when the same video contains multiple repeated, similar, or opposite events. On these hard examples, all responses sampled by Group Relative Policy Optimization (GRPO) may miss the target and receive zero temporal Intersection-over-Union (tIoU). Their group-relative advantages are then all zero, leaving no useful policy gradient signal. To address this, we propose DTP-OPSD, a reinforcement learning framework with distractor-aware temporal-privilege on-policy self-distillation. The student observes the clean video and generates responses on-policy, while a shared-weight teacher observes a privileged video in which the target window is marked by red full-frame borders and same-video distractor windows by blue full-frame borders. For the distillation loss, the teacher evaluates student-generated prefixes and provides dense token-level Jensen-Shannon divergence supervision. This supervision remains informative even when GRPO provides no policy gradient signal. We suppress distillation when the teacher does not outperform the student and strengthen it for groups where the teacher performs better but all student rollouts receive zero tIoU. We also construct a compact hard-interference training set through VLM screening and human verification. With only 300 additional post-training examples, DTP-OPSD achieves the highest mIoU among the compared open-source 8B models on Charades-TimeLens, ActivityNet-TimeLens, and QVHighlights-TimeLens, reaching 55.93, 54.77, and 66.71, respectively. Compared with Pure GRPO trained on the same data, DTP-OPSD improves all 12 reported metrics, with R1@0.7 gains of 1.28, 1.62, and 1.78 points on the three benchmarks. Code will be released at https://github.com/nian1125/DTP-OPSD

0.20

RepConTUR: A Report-Conditioned Volumetric Multimodal Model for Abdominal Tumor Segmentation

cs.CV ·2026-10-29 ·Seunghwan Lee, Sangjoon Park, Jongmin Lee, Yoon Jo Kim, Wonyoung Cho +2 more
code/data
Abstract · Submitted to CVPR 2027

Tumor delineation is required for radiotherapy planning, yet automatic segmentation is limited by errors, predominantly false positives, that distort dose and increase editing time. The radiology report accompanies the scan and localizes disease at the suborgan level, but existing text-promptable models condition on short referring phrases, and most operate in 2D. We present RepConTUR, a single native-3D early-fusion multimodal model that conditions volumetric abdominal tumor segmentation on the narrative report. We extend AbdomenAtlas 3.0 to a suborgan-level promptable segmentation dataset and generate report-based VQA pairs, yielding four report-derived tasks on the same annotations: referring segmentation (RES), report-grounded segmentation (GSEG), report generation (RRG), and visual question answering (VQA). We train RepConTUR, comprising a 3D CT encoder, an early-fusion language model, and a native-3D segmentation module, jointly on the four tasks. Across liver, kidney and pancreas, RepConTUR attains the highest Dice similarity coefficient (DSC), instance-wise precision, and detection F1 among the text-promptable baselines, and adding the report raises mean precision while leaving recall comparable. Against two vision-only specialists, nnU-Net and the report-supervised R-Super, RepConTUR attains the highest value on every in-domain metric for liver and pancreas and the highest mean DSC, recall, precision, and detection F1 over the four out-of-distribution public cohorts. Ablations indicate that conditioning on the narrative report raises detection F1 from 0.499 to 0.528 within the same model, that joint GSEG training raises RES F1 from 0.450 to 0.482, and that the native-3D segmentation module raises GSEG F1 from 0.139 to 0.528 over its 2D counterpart. Code is available at https://github.com/Oncosoft-Research/RepConTUR.

0.20

Segment Anything Model Based Context Guided Mamba Unet for Medical Image Segmentation

cs.CV ·2026-10-29 ·Chenchen Wang
Abstract

Medical image segmentation is crucial for clinical diagnosis and treatment planning, requiring both global contextual understanding and fine-grained structural precision. While the Segment Anything Model (SAM) demonstrates strong generalization capabilities, its direct application to medical images often lacks domain-specific adaptability. Meanwhile, State Space Models (SSMs)-based methods, such as Mamba, efficiently capture long-range dependencies but may struggle with preserving fine anatomical details. To address these challenges, we propose a segment anything model based context guided mamba unet, SAMBA-UNet, a novel framework that integrates SAM's context knowledge with Mamba's hierarchical feature learning to enhance both global representation and local detail refinement. Specifically, we introduce the Hierarchical Context Enhancement Module (HCEM), an attention-based fusion module that deeply incorporates the multi-level features of SAM and Mamba to improve detail perception and segmentation precision. Additionally, we employ LoRA and Adapter-based fine-tuning to adapt SAM to medical imaging tasks, further enhancing its domain-specific segmentation capability and robustness. We conducted a comprehensive study on several medical segmentation datasets, including Synapse, ISIC17, and ISIC18, evaluating our method across multiple metrics. Experimental results demonstrate that SAMBA-UNet achieves state-of-the-art performance, effectively balancing generalization, segmentation accuracy, and computational efficiency. Our code will be released upon acceptance.

0.20

Unlimited OCR Works 2

cs.CL cs.CV ·2026-10-29 ·Shu Wei, Jingjing Wu, Qunyi Xie, Lingshu Zhang, Xiangcheng Du +9 more
benchmark
Abstract

Unlimited OCR ushers the era of long-horizon parsing, substantially mitigating the growth of working memory as document lengthens. However, as document length scales to hundreds of pages, the visual context grows substantially, sharply increasing the computational cost of each decoding step and limiting the scalability of ultra-long document parsing. Inspired by human book transcription via localized focus without re-scanning preceding pages, we introduce Unlimited OCR 2, a document parsing model that looks only at what it transcribes. Building on Unlimited OCR, we upgrade all the decoder attention modules to Sparse Reference Sliding Window Attention (SR-SWA), bounding active reference tokens to a fixed capacity and thus sustaining constant computational costs across arbitrarily long documents. By combining SR-SWA with Unlimited-Horizon Decoding Paradigm, Unlimited OCR 2 enables document parsing over virtually unlimited lengths without prohibitive memory overhead. At the hundred-page scale, Unlimited OCR 2 achieves 7.0$\times$ and 5.6$\times$ per-step speedups over Full Attention and R-SWA, respectively. Furthermore, it delivers state-of-the-art end-to-end performance on both MPDocBench and MarathonDocBench, our newly proposed benchmark targeting hundred-page document parsing. Ultimately, the capacity for uninterrupted, one-shot document parsing signifies a fundamental paradigm shift into the OCR 3.0 era.

0.20

EventSeeColor: Learning to Reconstruct Color from Monochrome Events

cs.CV cs.AI ·2026-10-29 ·Chuanzhi Xu, Langyi Chen, Haoxian Zhou, Zelin Zhang, Haodong Chen +3 more
Abstract

Event cameras capture visual brightness changes under challenging lighting conditions. However, traditional event-based image reconstruction targets grayscale intensity and produces images without scene color. Color distinguishes surfaces with similar intensity. Reconstructing color from events can reveal appearance differences that reconstructed grayscale images cannot express. Although monochrome event cameras do not directly measure color, their event-triggering patterns depend on surface reflectance, illumination, and the sensor's spectral response. These patterns may contain cues that support color reconstruction. In this paper, we present EventSeeColor, which uses event-reconstructed grayscale and aligned events to jointly guide a diffusion colorizer. Grayscale-derived semantic context supports color inference, while region-balanced chroma supervision and local RGB refinement improve color distribution, brightness, and texture. In the combined evaluation on two real event datasets, EventSeeColor achieves the best structural similarity, learned perceptual scores, and chromatic-region color difference among the external methods.

0.20

LoRe-MIM: Lost-or-Replaced Masked Image Modeling for Self-Supervised Representation Learning in Whole-Body PET/CT

cs.CV ·2026-10-29 ·Amartya Bhattacharya, Bashirul Azam Biswas, Biratal Raj Wagle, Michael S. Leapman, Matthew E. Maeder +2 more
Abstract · 25 pages, 9 figures

Deep learning models for whole-body positron emission tomography/computed tomography (PET/CT) hold significant clinical potential for disease detection, treatment planning, and response assessment. However, the time and cost associated with expert annotations remain major bottlenecks in developing robust and generalizable models. Although self-supervised learning (SSL) has emerged as a promising approach to mitigating annotation scarcity, its application to whole-body PET/CT remains relatively underexplored due to the limited availability of large-scale datasets and the challenges associated with learning complementary cross-modal representations. To address these limitations, we curate a heterogeneous, multi-tracer, multi-institutional cohort comprising 6,531 whole-body PET/CT scans and propose LoRe-MIM (Lost-Or-Replaced Masked Image Modelling), a novel self-supervised framework designed to learn multimodal representations through masked image reconstruction. Specifically, LoRe-MIM introduces a masking strategy in which image regions are either replaced with zeros (Lost) or substituted with corresponding complementary imaging modality (Replaced), encouraging the model to capture both modality-specific characteristics and cross-modal relationships. We evaluate the learned representations across multiple downstream tasks, including multimodal PET/CT image reconstruction, patch-based classification into cancer vs. non-cancer, whole-body lesion segmentation, label-efficient training, and robustness assessment. Experimental results demonstrate that LoRe-MIM consistently outperforms the corresponding state-of-the-art baselines across these tasks, highlighting its potential as a scalable and generalizable self-supervised learning framework for whole-body PET/CT analysis.

0.20

Hierarchical Step Decomposition for Training-free Online Video Step Grounding

cs.CV ·2026-10-29 ·Long Zhou, Luca Zanella, Paolo Rota, Massimiliano Mancini, Elisa Ricci
published
Abstract · WACV 2027

Given a set of procedural steps, Video Step Grounding (VSG) aims to identify which steps occur in a video and localize when they are performed. Existing methods are either offline, requiring access to the full video before making predictions, or online, but relying on fixed-length video segments producing noisy, over-segmented outputs. In this work, we propose Hierarchical Step Decomposition (HSD), a training-free framework for online VSG that leverages the semantic understanding capabilities of Large Multimodal Models. HSD addresses two central challenges in online grounding: constructing semantically coherent temporal segments from a streaming video, and bridging the semantic gap between coarse procedural step descriptions and visual observations. To construct elementary temporal units, HSD incrementally groups incoming segments into semantically consistent events, using an anchor caption that is continuously validated against the visual stream. Based on these events, we perform zero-shot procedural step classification and selectively refine uncertain predictions by decomposing candidate steps into finegrained substeps that capture more nuanced visual evidence. By operating on semantically coherent events rather than every fixed-length segment, HSD reduces redundant computation and produces more stable predictions. Experiments on HT-Step, Ego4D Goal-Step, and CrossTask show that HSD substantially outperforms state-of-the-art methods, overall achieving better step recognition, more accurate temporal localization, and faster inference.

0.20

Towards Multilingual End-of-Turn Detection using Frozen Speech Encoders

eess.AS ·2026-10-29 ·Rini Sharon, Kadri Hacioglu, Andreas Stolcke
Abstract

A spoken dialogue system must judge when the user has yielded the floor. This end-of-turn (EoT) detector is normally trained and tuned afresh for each language, which is costly to scale to new languages. We instead build one on a frozen multilingual encoder, training only a small fixed-size head. Using five CALLHOME languages, we investigate three questions. First, how well a head trained on one language transfers zero-shot to the others. Second, how much target-language adaptation is required to recover full task performance. Third, whether a single fixed decision threshold holds across languages without per-language tuning. We find that the head transfers well, even between tonal and pitch-accent languages, and that a few hours of audio recover nearly all the loss, matching in-language training. A frozen multilingual encoder with one lightweight head is a practical, language-agnostic baseline for EoT detection.

0.19

Coupled or Independent? Fault Coupling in Agentic Medical Vision Systems

cs.CV cs.CR ·2026-10-29 ·Nishi Kanta Paul, Md Shihabul Islam Shovo, Md Mostafijur Rahman, Jamil Fayyad
Abstract · 9 pages, 3 figures, 5 tables

Vision-language agents equipped with tools expose intermediate outputs for inspection and comparison with the model's reasoning, which may offer safety benefits over single-shot models. This paper tests that potential benefit when the sensor reading itself is untrusted, so that the tool and reasoning channels share a common visual fault. We consider a dermatology agent based on segmentation, contrastive retrieval, and an iterative vision-language reasoner and define a coupling taxonomy that distinguishes single-channel errors from joint errors that agree or disagree. On PAD-UFES-20, we show that typographic injection produces detectable directional retrieval capture, whereas physical corruption does not. Crop-first retrieval preserves the payload at a readable scale and produces 55% joint wrong agreement across all cases (71% of faulty cases); full-image retrieval produces 35% (43%). Under a binary decision, if both channels are wrong, their predictions must agree, so hard-label channel comparison cannot detect joint failures, limiting disagreement recall to 0.29 (crop-first) versus 0.57 (full-image). The direction of the order effect is reproduced on the independent HIBA test set.

0.19

FFAI: Flow Matching with Feedback Training and Adaptive Initialization for CT Reconstruction

cs.CV ·2026-10-29 ·Jigang Duan, Heran Wang, Genwei Ma, Xing Zhao
benchmark
Abstract

Although low-dose, sparse-view, and limited-angle CT involve distinct degradation mechanisms, all three pose reconstruction challenges due to limited measurement information and can therefore be addressed within a unified framework. In iterative image restoration with flow matching, prediction errors may cause inference states to deviate from the prescribed training path, while a fixed initialization may be unsuitable for inputs with varying degradation severity. We propose a unified conditional flow matching method that conditions on the initial reconstruction and combines feedback-aware training with degradation-adaptive initialization. During training, a teacher network simulates inference updates to generate shifted states, from which a student network learns to recover high-quality reference images, improving its ability to handle prediction errors. During inference, the starting time is selected based on the noise level or extent of missing projection data, and the noise added to the initial state is adjusted accordingly to adapt the starting point to the input. Experiments on public benchmark datasets show that the proposed method achieves better quantitative results than the compared methods across all three reconstruction tasks, effectively suppressing noise and artifacts while preserving anatomical structures and details.

0.19

Uncertainty-aware Coarse-to-Fine Main Pancreatic Duct Segmentation for Width Measurement in 3D MRCP

cs.CV ·2026-10-29 ·Zipei Zhang, Jaakko Sahlsten, Mikko Saukkoriipi, Johanna Ronkainen, Mona Lehto +3 more
Abstract

Assessment of the main pancreatic duct (MPD) from volumetric magnetic resonance cholangiopancreatography (MRCP) scans is an essential task for ductal dilatation diagnosis and pancreatic cancer analysis. However, manual assessment is affected by reader variability in which automated analysis may assist. Nevertheless, fully automated and uncertainty-aware MPD analysis from volumetric MRCP remains underexplored area of research. In this study, we propose an automatic and uncertainty-aware deep learning-based pipeline for MPD segmentation and maximum-width analysis from MRCP scans. A coarse localisation model first identifies the ductal region-of-interest, followed by a fine model that segments the MPD within the region. The dataset consists of 256 MPD-segmented MRCP scans from 69 patients. A randomly selected holdout test subset of 14 patients with 55 MRCP volumes was used for evaluation. We employed a five-model cross-validation ensemble for improved performance and uncertainty estimation. The best MPD segmentation performance was 0.634 DSC, 0.744 precision, and 0.593 recall. The mean absolute error for maximum MPD width prediction was 0.355~mm and dilated MPD ($\geq 5$~mm) was detected with 75.0\% sensitivity and 74.4\% specificity. These novel results suggest that automated MPD segmentation can be used to aid MRCP-based duct measurements.

0.19

What Makes Medical Examination Items Difficult? Using Large Language Models to Explore Patterns Associated With Empirical Item Difficulty

cs.CL cs.AI ·2026-10-29 ·Xiaomeng Xiong, Marcus Walker, Yanlin Jiang
Abstract

Research on medical item difficulty has largely focused on prediction rather than on identifying interpretable patterns associated with empirical difficulty. This study used empirical difficulty labels to guide large language models (LLMs) in inducing difficulty rubrics and examined how rubric guidance affected LLM difficulty judgments. Using training items and their empirical difficulty estimates from the United States Medical Licensing Examination (USMLE) and the Physician Assistant National Certifying Examination (PANCE), Claude Opus 5 induced separate rubrics for absolute and pairwise difficulty judgment. None of the induced rubrics treated surface features, such as stem length, option count, or images, as major sources of difficulty. Several patterns emerged across both examinations, including cue convergence, misleading cues, and option competition, although their specific forms differed. Retrieval of knowledge that could not be derived from the stem was particularly prominent in the PANCE rubrics, whereas the USMLE rubrics placed greater emphasis on integrating multiple clinical cues. We compared commercial and open-weight LLMs on held-out items under unguided and rubric-guided conditions. Rubric guidance changed how models used the difficulty scale but did not consistently improve performance. In absolute judgments, guidance encouraged broader use of the scale and generally improved coarse-grained agreement, without consistently improving rank ordering. In pairwise comparisons, its effects varied across models and examinations, and judgments became more sensitive to presentation order. These findings suggest that the value of label-guided rubric induction may lie more in generating interpretable, testable hypotheses from empirical difficulty data than in reliably improving difficulty judgments.

0.19

Let LiDAR Decide Where to Look: Geometry-Measured Token Selection for Efficient Sparse Latent World Models

cs.CV ·2026-10-29 ·Zeping Zhang, Zhuoya Zhao, Robert Laganière
Abstract

Vision-centric autonomous driving models encode thousands of image tokens per frame, yet most depict sky, buildings, or empty road---content irrelevant to planning. Worse, these tokens dilute the planner's attention: we show that randomly discarding 75% of tokens already outperforms the dense model, because the planner can concentrate on the remaining driving-critical content. Building on this finding, we present Geometry-Aware Token Selection (GATS), which uses LiDAR and radar point projections to select a small fraction of image tokens before the vision backbone runs. Because the selector operates on measured 3D geometry rather than visual features, the backbone encodes only the selected patches, cutting end-to-end FLOPs by up to $18\times$. Each kept token receives a 3D anchor from its LiDAR point and is fed to a self-supervised latent world model that predicts future features and 3D displacements without box annotations. At 25% budget, GATS achieves a 3-sec L2 displacement error of 0.520 m on nuScenes and an EPDMS of 88.9 on NAVSIM---surpassing the dense baseline (0.573 m / 85.9) at $6.6\times$ fewer total FLOPs. At an extreme 10% budget (24 GFLOPs, $18\times$ reduction), it still beats the dense baseline on both nuScenes (0.537 m vs. 0.573 m L2) and NAVSIM (87.9 vs. 85.9 EPDMS). On 739 intent-critical scenarios, GATS covers 95.8% of hazardous vehicles and achieves a 0.09% collision rate---73.5% lower than the dense model (0.34%).

0.19

Segment Anything Pancreas: A Unified MRI Segmentation Model for Pancreatic Cancer Radiotherapy

cs.CV ·2026-10-29 ·Zach Eidex, Mojtaba Safari, Kundan Thind, Parag J. Parikh, Zhen Tian +1 more
Abstract

Background: Limited tumor annotations and incomplete labels across MRI datasets hinder unified segmentation of pancreatic tumors and organs at risk (OARs) for MRI-guided radiotherapy. Purpose: We propose Pancreas-MedNeXt, a unified model that segments the whole pancreas, its subregions, tumors, and surrounding OARs from one MRI volume. Methods: The model uses a shared MedNeXt encoder-decoder with supervised whole-pancreas and subregion pretraining, followed by multitask fine-tuning. Three output heads predict the pancreatic subregions, tumor, and OARs (stomach, duodenum, kidneys). An annotation-aware loss allows training on partially labeled data without treating unannotated structures as background. Pancreas and tumor were evaluated by three-fold cross-validation on 142 PANTHER patients (92 diagnostic, 50 MR-Linac), subregions and OARs on Cyst-X and AMOS test sets, and pancreas and OARs on 33 external T1-weighted TotalSegmentator scans. Comparisons with nnU-Net, MedNeXt, VISTA3D, and Swin UNETR used the Dice similarity coefficient (DSC) with paired bootstrap 95% CIs. Results: Pancreas-MedNeXt achieved the highest mean pancreas and tumor DSC on diagnostic and MR-Linac MRI. Tumor DSC was 58.41% and 42.55%, exceeding task-specific MedNeXt by 4.94 (95% CI: 0.28-9.45) and 5.89 (95% CI: 2.37-9.72) percentage points, respectively. Whole-pancreas DSC was 85.73% and 70.66%. Pretraining improved head, body, and tail DSC by 2.01, 3.00, and 3.78 percentage points over training from scratch. It also achieved the highest DSC for the pancreas, duodenum, and right kidney on AMOS, and for all five organs on external MRI, exceeding the next-best method by 4.92-8.85 percentage points.Conclusion: A single Pancreas-MedNeXt model segments tumors, pancreatic subregions, and OARs on diagnostic and MR-Linac MRI, which could streamline contouring for pancreatic cancer radiotherapy.

0.19

Reading-Level Verification of Kanji Words in Japanese Financial Speech Synthesis

cs.SD cs.CLeess.AS ·2026-10-29 ·Cheng Zhang
Abstract · 6 pages, 4 figures, 7 tables

Few-shot voice cloning makes it cheap to produce daily Japanese financial commentary with several synthetic speakers, but kanji words with more than one reading, such as the word for "market" (shijo vs. ichiba) or "overall trend" (taisei vs. oze), are still misread. A common quality gate transcribes each synthesized sentence with automatic speech recognition (ASR), converts both the input and the transcript to kana, and regenerates on mismatch. We show that this gate is unreliable in both directions: when the audio contains a misreading, the ASR language model frequently writes the expected kanji back, hiding the error, and when the audio is correct, ordinary recognition errors trigger needless regeneration. We propose a reading-level check that scores the expected reading and its confusable alternatives with a connectionist temporal classification (CTC) hiragana recognizer and flags a word when an alternative explains the audio better. On 540 utterances from three voice-cloned production voices covering 30 financial terms, with misreadings injected under controlled conditions, the transcript-based gate detected 78.9% of misreadings at a 44.4% false-alarm rate, whereas the proposed check detected 91.7% at 5.0% (AUC 0.990). On natural kanji input, it caught a front-end error that the transcript-based gate missed for one voice, and reduced regeneration requests from 38.9% to 6.1% of sentences.

0.19

A learning-based method of automatic music transcription

cs.SD ·2026-10-29 ·Phu-Anh-Tai Do, Huu-Hoang Nguyen, Duc-Hoang Nguyen, Thanh-Hai Le
Abstract · 19 pages, 14 figures

Automatic Music Transcription (AMT) seeks to convert acoustic music recordings into symbolic musical notation, a process that remains remarkably difficult for polyphonic piano due to concurrent polyphony, harmonic partial interference, expressive timing, and severe pitch activation sparsity. In this paper, we formulate polyphonic piano transcription as a frame-level multi-label classification problem across the standard 88 piano pitches (MIDI notes 21 to 108). Raw audio waveforms are processed using the Constant-Q Transform (CQT) to construct a geometrically spaced log-frequency time-frequency representation that aligns with the chromatic musical scale across seven octaves (252 frequency bins per frame). We systematically investigate recurrent sequence models, contrasting unidirectional Long Short-Term Memory (LSTM) with Bidirectional LSTM (BiLSTM) backbones, and propose two architectural adaptations tailored to recurrent temporal representations: a modified additive attention module and a modified multihead attention mechanism equipped with hyperbolic tangent activation and head-wise summation. Furthermore, to mitigate the severe class imbalance between inactive and active notes (approximately 41:1), we evaluate Weighted Binary Cross-Entropy (WBCE) alongside standard Binary Cross-Entropy (BCE). Rigorous empirical evaluations on the MAPS and MAESTRO datasets demonstrate steady, compounding gains from bidirectionality and attention mechanisms. Our proposed Multihead Attention-BiLSTM architecture attains an F1-score of 88.20%, an accuracy of 78.90%, a precision of 0.89, and a recall of 0.86, substantially outperforming the baseline unidirectional LSTM (F1-score of 71.20%) while maintaining a compact parameter budget of 4.43M. Qualitative piano-roll visualizations confirm that our model generates faithful transcriptions even in complex harmonic passages.

0.19

Seeing and Imagining as One: Unified Video Generation and 4D Perception for Embodied World Modeling

cs.CV cs.RO ·2026-10-29 ·Haitang Feng, Zhuoman Liu, Zhennan Wang, Sheng Wang, Juelin Li +3 more
code/data
Abstract · Project page: https://gensight-rbs.github.io/

Embodied world models should not only imagine plausible futures, but also perceive the geometric, semantic, and motion structure of the evolving world. We present GenSight, a unified embodied world model that brings future video generation and dense perception into a single pretrained video diffusion transformer. GenSight uses a unified spatiotemporal representation for video, depth, embodiment segmentation, and first-frame-relative 2D flow. Built on a shared DiT backbone, GenSight uses lightweight task modulation to predict all modalities within the same architecture. All tasks are jointly optimized within a unified velocity prediction space. At inference, GenSight supports two complementary modes: multi-step future video generation with aligned dense predictions and single-step dense perception from an observed video. By unifying modality representation, model architecture, and prediction space, GenSight avoids separate modality-specific pipelines and couples appearance, geometry, semantics, and motion in a shared representation. It promotes cross-modal consistency and allows dense perception to benefit from the spatiotemporal priors learned through video generation. Trained on embodied manipulation data spanning multiple robot embodiments and human-hand interactions, GenSight demonstrates strong performance in both video generation and dense 4D perception, highlighting the effectiveness of unifying seeing and imagining for embodied world modeling.

0.18

Sparse Autoencoders for Expert-Assisted Interpretation of Histopathology Foundation Model Embeddings

cs.CV ·2026-10-29 ·Šimon Šmída, Wanda Benesova
Abstract

Histopathology foundation models learn rich embedding representations, but their internal structure remains difficult to inspect. Concept-level interpretability is particularly relevant in computational pathology because pathologists reason through recurring morphological and tissue-level patterns, including tissue architecture, cellular morphology, staining variation, and artifacts. In this work, we investigate whether sparse autoencoders (SAEs) can decompose embeddings from histopathology foundation models into sparse latent features suitable for expert-assisted interpretation. We train vanilla and Batch-TopK SAE variants on NCT-CRC embeddings, using CONCH as the primary foundation model and UNI2-h as an additional robustness evaluation, and characterize the resulting latent features through reconstruction fidelity, sparsity, coherence, label enrichment, and downstream class-signal preservation. The learned sparse representations preserve most downstream tissue-class information, and their latent features retrieve more coherent and label-concentrated patch groups than raw embedding dimensions or PCA components. Discussions with pathology experts further motivated a WSI-level latent visualization workflow that places feature activations into larger tissue context. Overall, the additional UNI2-h evaluation suggests that the observed behavior is not specific to CONCH, supporting the use of sparse autoencoders for expert-assisted analysis of histopathology foundation-model embeddings, while emphasizing that latent features should be treated as candidate concepts requiring pathological assessment rather than automatically validated histopathological concepts.

0.18

Can Image Agents Complete Visual Commissions? A Benchmark and Modular Framework

cs.CV ·2026-10-29 ·Tong Li, Lishuai Gao, Qi Zeng, Yifan Li, Cong Wei +2 more
benchmark
Abstract

Visual commissions require image agents to integrate heterogeneous source materials, preserve requirements across dependent stages, and produce coordinated deliverables. Existing evaluations examine specific capabilities and execution settings, leaving end-to-end commission completion less understood. We introduce Visual Commission Bench (VCB), comprising 600 tasks across 12 categories and three complementary tiers, from capability-focused tasks to complete commissions. VCB uses 5,039 requirement checks and separate visual quality ratings to assess content correctness, visual realization, constraint preservation, and delivery completeness without prescribing an internal workflow. We present V-Protocol, a reference agent that dynamically composes reusable task, cognitive, and expression modules at inference time. Task modules organize delivery, cognitive modules derive content, and expression modules guide visual realization. The modules exchange structured intermediate artifacts, while a shared execution harness tracks artifact revisions and dependencies, manages tool execution, and checks completion. With the same controller and image generator as the agent baselines, V-Protocol achieves 69.0\% strict task success on VCB, compared with 42.0\% for SCOPE. On Mind-Bench, strict accuracy increases from 39.2\% to 45.4\%.

0.18

Self-training for cross-centre coronary vessel segmentation in X-ray angiography using pre-contrast frame subtraction

cs.CV cs.AI ·2026-10-29 ·Danial Moafi, Aida Moafi, Mostafa Aryafar, Ashkan Sami, Maryam Aryafar
Abstract · 7 pages, 2 figures, 3 tables

Deep learning models for coronary vessel segmentation in X-ray angiography often perform poorly on data from centres that were not represented in training. Apart from differences in equipment, the labels of the largest public dataset (ARCADE) are incomplete, as only the segments defined in the SYNTAX score are annotated. In this study, we evaluated whether self-training with pseudo-labels can improve performance on external data without additional manual annotation. Pseudo-labels were generated from two sources, namely the predictions of an nnU-Net model trained on ARCADE and a contrast map obtained by subtracting a pre-contrast frame from the same sequence. The nnU-Net predictions were extended along the contrast map, and pixels detected only in the subtraction image were assigned an ignore label and excluded from the loss. Six models were trained and tested on ARCADE and on two external datasets (XCAD and XACV). All first-round models initialised with the baseline predictions performed better than both the baseline and a control model fine-tuned for the same number of epochs. The model trained on the XCAD pool improved the macro-averaged Dice score by 2.7 points (95% CI 2.3-3.2), while the control decreased it by 0.6 points. A second round of self-training did not lead to further improvement. When an external model trained on 50 fully annotated images was used to generate the initial labels, the XCAD Dice score increased from 0.623 to 0.745 without any XCAD labels, although performance on XACV did not improve because large main vessels were missed. These results suggest that the benefit of self-training depends largely on the sensitivity of the model used to generate the initial pseudo-labels.

0.18

LAPIS: Clinically Constrained Patient-Identity Residual Learning for Chest X-Ray Anonymization

eess.IV ·2026-10-29 ·Nishi Kanta Paul, Md Shihabul Islam Shovo, Désiré Sidibé
Abstract · 5 pages, 3 figures, 2 tables. Submitted to IEEE ISBI 2027

Biometric information in chest radiographs limits their release for research, as patients can be re-identified. We propose LAPIS, which jointly adapts a spatial chest-radiograph latent space so that patient-discriminative information concentrates in a compact, replaceable channel residual while most clinically relevant content stays in the complement. The residual is defined by an exactly semi-orthogonal basis, a spatial adversary suppresses identity leakage from the complement, and replacement-aware training teaches the decoder cross-patient residual swaps under output identity, clinical and realism constraints, so that replacing the residual yields a valid image. At release, the residual is replaced by a mixture of donor residuals. On patient-disjoint model, release and test sets from ChestX-ray14, we evaluate attackers adapted to the anonymized setting, a structural source-retrieval attack, and pathology classifiers trained on released images and tested on real ones. LAPIS reduces worst-case cross-study linkability from 0.961 to 0.791 AUC at a macro AUROC cost of 0.012 relative to reconstruction; a deformation-based baseline, however, remains better at structural source retrieval.

0.17

Efficient Covariance-Aware Vicinal Latent Diffusion for Continuous Conditional Image Generation

cs.CV ·2026-10-29 ·Xin Ding, Shuo Huang, Zhenglong Ding, Sen Zhang, Nenglun Chen +3 more
Abstract · 14 pages

Continuous Conditional Generative Modeling (CCGM) aims to synthesize high-dimensional data conditioned on continuously valued scalar labels. However, existing continuous diffusion models face severe data sparsity under imbalanced and long-tailed label distributions, which can lead to visual artifacts and condition inconsistency. Moreover, inference-time classifier-free guidance and dense pixel-space integration introduce substantial sampling overhead, limiting practical deployment. To address these challenges, we propose CAViLD, an efficient latent diffusion framework for continuous conditional generation. CAViLD first constructs a purified condition-aware latent space through density-adaptive vicinal learning, auxiliary semantic regularization, and NIQE-guided negative data augmentation. It then internalizes the conditional guidance direction through an EMA teacher, enabling single-branch conditional inference, and introduces covariance-aware progressive distillation for label-dependent probability-flow ODEs, supporting sampling down to four NFEs. Experiments on RC-49, UTKFace, and Steering Angle demonstrate a favorable trade-off among generation fidelity, continuous condition consistency, and sampling efficiency.

0.17

Localizing Concepts in Visual Autoregressive Models

cs.CV cs.AI ·2026-10-29 ·Nanxiang Jiang, Yang Liu, Yuanhao Wang, Min Xu
Abstract

Understanding the internal routing of visual concepts is crucial for the safe and controllable adaptation of generative models. While concept localization has been widely studied in diffusion models, the emerging paradigm of next-scale visual autoregressive models remains largely unexplored. In this paper, we introduce LoCo, the first model-agnostic framework to localize where and when specific concepts emerge within autoregressive models. Driven by the native coarse-to-fine nature of next-scale generation, our method precisely maps conceptual knowledge across three distinct dimensions: Layer, Scale, and Position. To systematically localize and evaluate concept routing without context bias, we propose LoCoBench, a comprehensive dataset spanning 10 diverse categories. Extensive probing on image and video autoregressive models like Infinity, HunyuanImage-3.0, and InfinityStar shows that the localized positions are both interpretable and causally related to concept emergence. Building on these insights, we apply our localization method to three key applications: concept erasure, model personalization, and adversarial concept injection. Experiments demonstrate that our targeted intervention achieves state-of-the-art performance, substantially reducing computational overhead while preserving benign utility. Overall, our findings offer insights into how conceptual knowledge is routed during autoregressive generation, introducing a practical pathway for more interpretable, efficient, and secure adaptation.

0.17

Bridging Neural Representations and Symbolic Semantics in Audio Classification

cs.SD cs.AI ·2026-10-29 ·Avishek Das, Pascal Hitzler
Abstract

Deep audio classifiers can achieve strong predictive performance, but their hidden representations remain difficult to interpret. Existing audio explainability methods often focus on input-level attribution, such as waveform segments, spectrogram regions, or acoustic features, and provide limited insight into what hidden neurons encode. Interpreting hidden neurons at the concept level can provide a more meaningful view of what an audio model learns internally. This paper investigates whether hidden neurons in audio classifiers can be associated with human-understandable symbolic acoustic concepts. A model-agnostic neurosymbolic framework is presented for analyzing the final hidden layer of audio classifiers, containing 32 neurons, through symbolic concept induction. The framework is evaluated on two audio classification settings: music genre classification using GTZAN dataset and musical instrument classification using IRMAS dataset. Results show that several hidden neurons can be associated with meaningful induced concepts, and most confirmed neuron-concept pairs are statistically supported using the Mann-Whitney U test. These findings suggest that symbolic concept induction is a promising direction for explaining latent audio representations beyond input-level attribution.

0.17

EmpDPO: Enabling Emphasis Control in TTS via Direct Preference Optimization

cs.SD eess.AS ·2026-10-29 ·Haonan Chen, Yunyun Wang, Megan Wei, Zeyu Jin
Abstract · 5 pages, 3 figures. Submitted to ICASSP 2027. Audio samples: https://chaonan99.github.io/empdpo-demo/

Word-level emphasis conveys meaning that text alone cannot express and is therefore essential for a practical text-to-speech (TTS) system. In this work, we study how to add emphasis control post hoc to a pretrained TTS model that has never been trained with emphasis annotations. We present EmpDPO, a simple, efficient, and effective two-stage recipe that adds emphasis control to a multilingual zero-shot diffusion TTS model through supervised fine-tuning followed by direct preference optimization (DPO). The training signals for both stages come from an off-the-shelf multilingual emphasis detector, so that no manual emphasis annotation or task-specific reward design is needed. We find that DPO can substantially strengthen emphasis control with only a few thousand preference pairs and training steps. Objective and subjective evaluations against strong public systems show that our model produces perceptible emphasis while preserving naturalness and overall quality.

0.17

ROAD: Modern Adversarial Training for Simple One-Step Distillation at Scale

cs.CV cs.LG ·2026-10-29 ·Xiaoou Li, Shan You, Tao Huang
Abstract · 37 pages, 12 figures, 7 tables

One-step distillation turns pretrained diffusion models into fast generators, but strong results often require substantial training machinery and compute. Pure adversarial distillation offers a simpler route, yet has struggled at modern text-to-image scale. To address this limitation, we propose ROAD (Representation-grounded One-step Adversarial Distillation), a one-stage framework that distills text-to-image models with one adversarial objective and a lightweight representation-space critic. First, we replace the classical convolutional head with a Lipschitz-controlled Transformer head with token-level text conditioning, which matches the semantic tokens of modern visual encoders. Second, we observed a fixed random feature mixing leads to long-horizon degradation. We attribute it to generator-critic co-adaptation around one realization-specific view, and periodically re-draw the mixing to keep the randomization effective. Third, for our dense and vector-valued logits, we introduce Jacobian R1 which penalizes the full input-logit Jacobian, overcoming pooled R1's limitation of neglecting most of logit sensitivities. A single adversarial objective then suffices: in one step, ROAD reaches 11.03 FID on SDXL and 13.73 on SD3.5-M on COCO-10K, surpassing the multi-step teachers at a fraction of distribution matching's training compute.

0.17

HiFOR: Hierarchical Foresight with Action-Guided World Grounding for Vision-Language-Action Model

cs.CV cs.RO ·2026-10-29 ·Rongxing Ding, Hongyu Qu, Kehong Wu, Xudong Wang, Fan Zhang +2 more
Abstract

Vision-Language-Action (VLA) models have demonstrated promising capabilities in robotic manipulation by integrating visual perception, language understanding, and action generation. Recent advances have further incorporated future prediction to enhance the understanding of environmental dynamics and task progression. However, existing methods often model short-term dynamics and long-term task objectives separately, while directly incorporating predicted futures into action policies without sufficiently considering their relevance to current action requirements. To address these limitations, we propose HiFOR, a hierarchical foresight framework with action-guided world retrieval. Specifically, Hierarchical Foresight Prediction jointly predicts short-term state evolution, semantic milestones, and final goal states through a Hierarchical Dependency Mask (HDM), establishing dependencies across temporal horizons. Building upon these predictions, Structure-Aware Latent World Modeling organizes future latent representations along the horizon, view, and spatial dimensions to facilitate fine-grained retrieval. Furthermore, Action-Guided Factorized World Retrieval adaptively retrieves relevant future information from the structured latent world based on evolving action representations during denoising, providing targeted guidance for action generation. Extensive experiments on two simulation benchmarks and six real-world robotic manipulation tasks demonstrate the effectiveness of our approach.

0.17

TokenRepel: Generating Diverse Image Sets Without Compromising Quality

cs.CV ·2026-10-29 ·Qingtao Yu, Zhen Wang, Ning Lai, Boyang Hu, Hongdong Li +1 more
published
Abstract · NeurIPS 2026

We present an inference-time approach that enhances the visual diversity of image sets generated by text-to-image models, without compromising the average image quality. Our method leverages the intrinsic geometry of hidden representations in pretrained models to diversify images generated from the same prompt. First, we introduce a token repulsion mechanism that increases the angle between feature tokens across samples. It takes a negative Riemannian gradient step on a von Mises-Fisher mixture distribution of tokens on the unit hypersphere. Second, to preserve the quality of each output image, we propose a token density preservation mechanism that projects the repulsion-induced prediction change onto a half-space aligned with the model's score direction. This encourages the updates to remain within the high-density regions learned by the generative model. We further extend this mechanism to the output of each intermediate layer, enabling finer-grained iterative guidance. We evaluate our method on three types of pretrained text-to-image models. Experiments show substantial gains in visual diversity under identical prompts, while maintaining a comparable level of image fidelity.

0.17

SCOPE: Semantic COdec Pre-Editing for Object-Preserving Compression

cs.CV ·2026-10-29 ·Silvia Dani, Leonardo Galteri, Marco Bertini
Abstract

Image compression is increasingly used before machine vision systems, yet standard codecs are mainly optimized for human perception rather than downstream task accuracy. We propose SCOPE (Semantic COdec Pre-Editing), a conditional image pre-editor that modifies images before compression to improve the bitrate and detection accuracy trade-off. The model is conditioned on the codec family and compression severity, allowing a single trained network to adapt to multiple compression settings without retraining or loading codec-specific weights. Instead of uniformly preserving the whole image, the pre-editor learns to protect task-relevant regions while simplifying less important background content to improve compressibility. The resulting rate-accuracy curves show consistent gains across different codec families and detector architectures, together with strong transferability to unseen learned and wavelet-based compression families.

0.16

Evaluating Cultural Knowledge Editing in Vision-Language Models

cs.CV cs.CL ·2026-10-29 ·Malvina Nikandrou, Georgios Pantazopoulos, Ioannis Konstas, Alessandro Suglia
benchmark
Abstract

Current Vision-Language Models (VLMs) frequently fail to recognize culture-specific concepts, especially from underrepresented regions. Effectively injecting this long-tail knowledge into existing models remains a challenge. In this work, we investigate knowledge editing as a targeted alternative by introducing CulturalBind, a benchmark comprising 3.2K images of 1K concepts from 6 domains, designed to evaluate culture-aware cross-modal knowledge as well as concept-level knowledge editing in VLMs. Pre-edit evaluation reveals that textual and visual knowledge is decoupled: models often demonstrate text-based knowledge of cultural concepts but fail to identify them from images. We then evaluate editing methods in single and sequential scenarios. Although current methods successfully correct individual visual recognition errors, they exhibit low cross-modal portability, failing to integrate edits into the model's broader parametric knowledge. Furthermore, performance degrades rapidly under sequential editing, highlighting the need for comprehensive adaptation approaches that better consolidate multimodal knowledge over time.

0.16

Calibrating Attention-Based Token Pruning for Efficient Multi-Image Synthesis in UMMs

cs.CV ·2026-10-29 ·Zimeng Wu, Qijian Tian, Penghui Du, Changqian Yu, Jiaxin Chen +1 more
Abstract

Unified Multimodal Models (UMMs) enable powerful multi-image synthesis but incur substantial computation from long visual contexts. While token pruning offers a direct route to accelerating inference, existing methods are primarily adapted from understanding tasks, suffering from insufficient reference image exploration, confused conditional guidance, and pooling-induced ambiguity that degrades generation quality. To address these issues, we propose CATP, the first training-free context-compression method tailored to UMM generation, which \textbf{C}alibrates the discrepancies of general \textbf{A}ttention-based \textbf{T}oken \textbf{P}runing for better quality-efficiency trade-off. Concretely, building on independent layer-wise pruning during denoising, we introduce Guidance-Consistent Pruning (GCP) to synchronize pruning decisions across conditioning branches, and a Taylor-Calibrated Attention Metric (TCAM) that derives a theoretically grounded second-order correction for pooling-induced bias. In addition, a Fast Scoring Operator (FSO) fuses the calibration into tiled online Softmax to reduce pruning cost. Extensive experiments across representative UMMs and benchmarks demonstrate that CATP consistently outperforms state-of-the-art token-reduction methods, achieving up to 2.09$\times$ attention-core speedup.

0.16

UFMF: Radar-Guided Kinematic Priors for Real-Time 4D Occupancy Forecasting

cs.CV ·2026-10-29 ·Zeping Zhang, Zhuoya Zhao, Robert Laganière
Abstract

Real-time 4D occupancy forecasting requires reliably extrapolating both ego motion and surrounding actor motion under strict latency budgets. Recent occupancy world models improve long-horizon fidelity through generative decoders and 3D volumetric modeling, yet their computational cost makes real-time deployment challenging. Light-weight BEV forecasters meet latency requirements but often blur or suppress fast-moving actors at longer horizons. We present UFMF (Ultra-Fast Multi-modal Forecaster), a real-time forecaster that keeps an attention-free diagonal state-space temporal core while injecting targeted kinematic and geometric structure where lightweight models typically fail. A radar-guided velocity-aware dynamic prior (VADP) decomposes future displacement into a linear extrapolation plus a learned horizon-dependent residual. The resulting per-cell flow fields transport a compact dynamic latent subspace during rollout. Two auxiliary training constraints complement this branch: world-frame dynamic supervision enforces inter-horizon trajectory coherence, and class-conditioned vertical-profile supervision regularizes BEV-to-3D lifting. UFMF operates at 122 FPS on a single RTX 4090 and reaches 0.45 fast-displacement mIoU on nuScenes, outperforming recent occupancy world models at a fraction of their computational cost.

0.15

IndicIPR-QA: A Benchmark for Extractive Question Answering on Indian IPR Laws

cs.CL ·2026-10-29 ·M. Mohaiminul Islam, Muhammad Abulaish
benchmark
Abstract

Legal Question Answering (QA) over statutory texts is a challenging task in Natural Language Processing (NLP), especially in low-resource settings with complex legal systems. It requires answers that are precise and grounded in the law. However, to the best of current knowledge, no existing extractive QA benchmark provides exact answer spans from statutory texts across all seven Indian Intellectual Property Rights (IPR) Acts. This paper presents IndicIPR-QA1, an extractive QA benchmark for Indian IPR laws. The proposed dataset contains 4,462 QA pairs from 1,116 statutory contexts across seven Acts. Questions and answers are generated using GPT-4o and validated through nine automatic quality checks, single and multi-model LLM evaluations, and a 200-pair human audit. Section-level partitioning prevents source-context overlap across the training, validation, and test sets. Further analysis examines answer span length, lexical overlap, and multi-clause structure across the corpus and individual Acts. BM25 achieves a token-level F1 score of 20.21, while fine-tuned InLegalBERT reaches 85.02 under a unified sliding-window inference protocol. The large performance gap between BM25 (20.21 F1) and InLegalBERT (85.02 F1) shows that simple word matching is not enough to answer these legal questions. Performance also varies across the seven Acts, indicating substantial differences in answer-extraction performance.

0.15

Trajeform: Embedding Crowd Data

cs.GR cs.LG ·2026-10-29 ·Marilena Lemonari, Panayiotis Charalambous, Yiorgos Chrysanthou
Abstract · 14 pages, 17 figures, 5 tables

Crowd simulation aims to generate agent behaviours that are rich, natural, and appropriate to their surrounding environment. Recent advances in Machine Learning have enabled such behavioural patterns to be learned directly from trajectory data, allowing simulation models to capture increasingly complex crowd dynamics. However, the capabilities of these systems remain strongly dependent on the type and diversity of the data used during training. When evaluating how well crowd simulators generalise to real-world datasets, most existing metrics focus primarily on trajectory-level differences rather than behaviour-level similarity. In this paper, we present Trajeform, a framework that trains a reconstruction VAE (Variational Autoencoder) on real trajectory data to learn a meaningful continuous embedding that preserves behavioural characteristics of real crowds. The trained encoder can then be applied to synthetic trajectories generated by simulators and used to compare the resulting embedding distributions. Additionally, the learned embedding can be used to analyse supplementary data, such as trajectories corresponding to a simulator's principal behaviours or captured data from controlled environments (e.g., cycling), providing insight into their ability to span the real behavioural embedding space. The framework compares data from three simulators (CCP, modCCP, and RVO) against ground-truth datasets and is evaluated quantitatively and qualitatively using latent-space analysis, distributional metrics, and trajectory generation through sampling from the aggregated posterior distribution. The results indicate that the learned latent space captures meaningful trajectory dependencies found in the real data, making the framework useful both as an additional measure of synthetic data believability and as a tool for informing the training of future trajectory models.

0.15

How I Aligned Your Model: Scaling Up Multilingual Safety Training with Templates

cs.CL cs.AIcs.LG ·2026-10-29 ·Max Neuwinger, Hanna Yukhymenko, Imanol Schlag
Abstract

Safety alignment for Large Language Models is heavily focused on English with limited multilingual representation, particularly damaging low-resource language performance. Related open training datasets are often limited in size or do not follow region-localized harmful topics to broaden the scope of alignment beyond cultures. We propose a scalable template-based augmentation pipeline to produce massive multilingual refusal and benign corpora, extracting safe templates from harmful prompts, translating them and then filling in the slots with culturally authentic entities, both harmful and benign. As a result, we release MSafeTemp, a validated corpus spanning 60 languages useful for multilingual safety alignment and evaluation. Using the dataset, we show that English-only evaluation misjudges multilingual safety: for several models, attacks that fail in English succeed once translated, while culturally localized variants expose further failures that translation alone misses. We also confirm that augmenting safety training data through templates allows for higher diversity and quality of alignment, resulting in improved safety performance beyond English. MSafeTemp and the respective pipeline are aimed at facilitating research and development of safer and more responsible models for less-represented cultures.

0.15

sDMD: Sinkhorn Distribution Matching Distillation

cs.CV cs.LG ·2026-10-29 ·Jaihoon Kim, Lingxiao Li, Soshi Shimada, Roshni Cooper, Minhyuk Sung +1 more
code/data
Abstract · Project Page: https://sinkhorn-dmd.github.io/

Distribution Matching Distillation (DMD) distills multi-step diffusion models into few-step generators by minimizing reverse KL divergence. Despite its strong performance, DMD guides samples by local density alone, which introduces two limitations: teacher scores can be unreliable in regions of low teacher density, and the mode-seeking behavior of reverse KL can collapse the student onto a subset of the teacher modes. We first connect DMD to Wasserstein gradient flow, showing that its gradient follows from a fixed-point formulation with integral KL (IKL) as the energy functional. This perspective motivates augmenting IKL with an optimal transport term that addresses these limitations by guiding outlier samples toward the target and promoting mode coverage. We realize this term using Sinkhorn divergence estimated from finite batches. In practice, however, errors in the estimated Sinkhorn velocity can perturb in-domain samples for which IKL guidance is already accurate. To this end, we introduce sDMD, which adaptively activates the Sinkhorn correction based on discriminator outputs, focusing on outlier samples where IKL may be unreliable. With 2--4 sampling steps on FLUX.1-dev, Qwen-Image-20B, and Wan2.1-14B, sDMD outperforms distillation baselines on GenEval, DPG, WISE, and VBench while improving sample diversity over DMD2.

0.15

DiFuse: Fusing Native Diffusion Representations for Spatial Reasoning in Unified Multimodal Models

cs.CV ·2026-10-29 ·Ruixin Fan, Weining Wang, Jiahui Sun, Tailai Chen, Jing Liu
Abstract · 20 pages, 6 figures

Unified multimodal models (UMMs) jointly support visual understanding and image generation, yet information flow between these two capabilities remains largely asymmetric: understanding representations routinely condition generation, whereas generative representations are rarely reused for multimodal reasoning. In this work, we investigate whether intermediate representations from a UMM's native diffusion branch can enhance spatial reasoning without compromising general multimodal understanding. We propose DiFuse, which selectively incorporates intermediate diffusion representations into the understanding pathway of UMMs. Through systematic layer-wise analysis, we identify which diffusion representations are most informative and where they can be integrated with minimal interference to semantic processing. Guided by these findings, DiFuse spatially aligns diffusion features with visual tokens and injects them into selected multimodal decoder layers via residual fusion. Extensive experiments on UniWorld-V1 show that, DiFuse consistently improves spatial reasoning over a matched SFT baseline while maintaining broadly comparable performance on general multimodal benchmarks. Further analyses show that these gains depend critically on the choice of diffusion representations, injection positions, and the preservation of spatial correspondence between diffusion features and visual tokens. Results on OmniGen2 further demonstrate the transferability of DiFuse across unified multimodal architectures. Overall, our findings show that native diffusion representations can provide complementary spatial cues for multimodal reasoning when selectively integrated in a structure-preserving manner.

0.14

Lightweight Underwater Acoustic Vessel Classification for Embedded Deployment

cs.SD ·2026-10-29 ·Abishek Soti, Thura Pyae Sone, Naqib Ibnul, Jim Huang, Sergio Chevtchenko +3 more
Abstract

We propose a Raspberry Pi-based underwater vessel classification system for edge deployment under limited connectivity, bandwidth and power in offshore acoustic surveillance. Using ShipsEar, we compare six classifiers on a Raspberry Pi 5 through a complete audio-to-prediction pipeline. We assess classification performance, processing latency, CPU utilisation, memory consumption and monitored power. Supporting experiments examine acoustic feature selection, statistical pooling and Transformer patch configurations. The two-layer convolutional neural network (CNN) achieved a mean macro-F1 of \textbf{(99.14 \textpm{} 0.32)\%} across five random seeds. The radial basis function support vector machine (RBF-SVM) achieved \textbf{98.83\%}, with a median end-to-end latency of 28.109~ms compared with 120.655~ms for the CNN. This favourable trade-off motivated RBF-SVM deployment on the Raspberry Pi Zero 2 W. It maintained the same classification performance with a median end-to-end latency of \textbf{349.09~ms} and a sustained throughput of \textbf{2.962 classifications/s}, or \textbf{14.8} times the rate required for consecutive five-second audio segments.

0.14

MRC-MMAgent: Reinforcement Learning for Multi-Role Collaboration in Multimodal Tool-Using Agents

cs.CV ·2026-10-29 ·Shengqin Wang, Jie Jin, Xiaofan Li, Wentao Yan, Yu Cheng +6 more
Abstract

Multimodal tool-using agents augment vision-language models with access to external information and capabilities, enabling them to solve knowledge-intensive and visually complex tasks that cannot be addressed using visual context alone. However, existing reinforcement learning pipelines largely train single-role agents to search and answer using terminal rewards, overlooking how humans inspect retrieved evidence, reformulate queries, and revise conclusions during search. Consequently, they fail to optimize this self-corrective process over complete interaction trajectories. We propose MRC-MMAgent, a multi-role collaborative reinforcement learning framework that incorporates trajectory assessment and continuation directly into the executable agent loop. First, the framework establishes a multi-role collaborative loop that integrates solving, inspection, and continuation, enabling the agent to acquire evidence through tools, evaluate the preceding trajectory, and resume problem solving when the current answer is insufficiently supported. Second, we introduce a role-coupled learning objective that combines task success, interaction validity, and successful correction with full-trajectory policy optimization across the decisions generated by all roles. Extensive experiments show that, using only the original baseline training set without constructing additional datasets, our method consistently outperforms the corresponding base models and achieves leading or highly competitive performance across benchmarks.

0.14

RADI3N: GeometRic ImAge EDiting via ParametrIc 3D CoNtrol

cs.CV ·2026-10-29 ·Abdul Wasi, Akhil Gorugantu, Mahesh Bhosale, Anjan Dutta, Junsong Yuan +2 more
Abstract

Recent diffusion-based methods have substantially advanced image editing, yet parametric 3D control at the part level remains challenging. Existing approaches either operate on whole objects with manual rigging, or revert to 2D handles at the part level. They also struggle to model appearance and texture under large transformations, leading to unrealistic outputs in geometry-intensive edits. To address these limitations, we introduce RADI3N, a framework for semantically-grounded part-aware object manipulation with parametric 3D control. Our CS-ARAP deformation module suppresses rotation-field discontinuities at part boundaries via neighborhood covariance smoothing, producing structurally consistent deformations under complex part-level edits. We further introduce an inversion-free selective texture transfer mechanism that suppresses high-frequency positional biases and scales up low-frequency semantic cues within cross-image attention. This eliminates strict spatial bias and supports semantic matching, allowing source appearance to follow the edited part onto its new geometry. Experiments demonstrate that RADI3N achieves state-of-the-art performance in geometric fidelity, viewpoint consistency, and controllability.

0.14

Prompts and probabilities in the linguistic assessment of language models: Evidence from semantic focus-sensitivity

cs.CL cs.AI ·2026-10-29 ·Robert Pasternak
position
Abstract

As language models (LMs) have improved, an increasingly common means of evaluating their linguistic competence is through explicit prompting (e.g., "Which sentence is more acceptable?"). Prior work indicates that such prompting underperforms compared to standard token-probability-based methods in smaller models, but improves in larger models. We test this in a new domain: semantic focus-sensitivity, responsible for contrasts like "Lucy only SWAM yesterday" vs. "Lucy only swam YESTERDAY". Testing models of various sizes using text-based focus markers like boldface and italics, we find universal above-chance performance for token probabilities. Prompting results mostly align with previous findings, but in some larger models they outperform probabilities by a wide margin. We propose that this is because LMs are not prone to produce out-of-the-blue focus marking, which muddies results from token probabilities in a way that does not affect prompting. We test this by replacing standard focus markers with randomly generated ones; token probabilities still beat chance but show significantly greater performance decrease compared to prompting. These results suggest that metalinguistic prompting, while non-trivial for LMs, can also be more robust when testing stylistically atypical data. We argue that this has important consequences for the assessment of LMs' language competence, even if LMs do not perform metalinguistic judgments through human-like introspection.

0.14

The Importance of Speech Genre for End-of-Turn Detector Training

eess.AS cs.AI ·2026-10-29 ·Rini Sharon, Kadri Hacioglu, Jeena Prakash, Andreas Stolcke
code/data
Abstract

End-of-turn (EoT) detection has been studied mostly through features, models, and modalities, while the speech genre of the training data has drawn far less attention. We isolate this factor. Using one reproducible annotation pipeline, we build a matched pair of corpora, multi-turn human-human conversational and single-turn human-machine, and train the same streaming ConvLSTM EoT detector on each. Across four test sets the conversational model reaches up to +0.45 higher F1, driven by fewer false alarms on the mid-turn pauses and hesitations that are rare in single-turn corpora. This conversational-training effect is not specific to frame-level decoding. Under a matched complete-or-not test it also appears at the utterance level, across two encoders, but only on conversational test data. Testing only on single-turn speech, as most prior work does, therefore hides the benefit of conversational training data. Against open-source baselines, our 1.8M-parameter model attains the highest F1 on every test set.

0.13

AVPS-360: A Spatial Audio-Visual Benchmark for Instance Segmentation and Tracking of Sound-Producing Persons

cs.CV ·2026-10-29 ·Yifu Sun, Parthasaarathy Sudarsanam, Tuomas Virtanen, Archontis Politis
benchmark
Abstract

Audio-visual instance segmentation (AVIS) segments and tracks individual sound-producing objects in videos. Existing benchmarks predominantly use monaural audio, which lacks inter-channel cues for spatially distinguishing concurrent sound sources. We introduce AVPS-360 (Audio-Visual Person Segmentation 360), a spatial audio-visual benchmark bridging AVIS and sound event localisation and detection (SELD). Built upon STARSS23, it pairs first-order Ambisonics audio with synchronised 360° video across 230 recordings totalling 10.5 hours. Recordings average 164.0 seconds. The dataset captures intermittent sound production in real-world interactions dominated by speech, alongside laughter, clapping, and footsteps. It provides 31,678 manually reviewed instance masks linked to source identities and sound-event labels. We also present a baseline that integrates spatial audio with visual features for frame-level segmentation and identity association across frames. Experiments on AVPS-360 show that existing AVIS methods struggle with detection and identity association in long recordings with intermittent sound production, while our baseline substantially improves performance in both aspects.

0.13

Integrating Heterogeneous Echocardiography AI Outputs in Language Space: Learnable Late Fusion with a Supervised LLM

cs.CL cs.AI ·2026-10-29 ·Ryo Takizawa, Tempei Kabayama, Chiemi Yamazaki, Shun Kitamura, Ryo Matsuoka +3 more
Abstract

View- and target-specific echocardiography AI models capture useful but complementary information and may not achieve sufficient predictive performance in isolation. A learned ensemble is a natural way to combine these outputs, but heterogeneous models typically require each output to be converted into a binary decision, probability, confidence score, or fixed-length feature vector. We developed a supervised large language model (LLM)-based late-fusion module that directly aggregates view-specific predicted finding statements and predicted measurements in language space and learns a case-level fusion function. In patient-level three-fold cross-validation of 859 examinations from 646 patients, the proposed method integrated heterogeneous outputs within a single framework, performed comparably to the best-performing reference strategy evaluated, and outperformed a general-purpose LLM without task-specific training. These results suggest that a supervised LLM can combine multiple upstream models without model-specific output conversion or feature engineering that requires medical knowledge.

0.12

Array-Agnostic High-Order Ambisonics Encoding via Diffusion Posterior Sampling

eess.AS eess.SP ·2026-10-29 ·Amit Milstein, Nir Shlezinger, Boaz Rafaely
Abstract

Ambisonics provides a device-independent representation of three-dimensional sound fields; however, extracting these coefficients from practical microphone array recordings remains strongly dependent on array geometry, acoustic transfer functions, and spatial sampling resolution. This pipeline is typically divided into two distinct tasks: encoding, which entails reconstructing the physically resolvable spatial components, and upscaling, which requires generating higher-order coefficients. Model-based approaches often fail on irregular arrays due to ill-conditioned transfer matrices, whereas conventional neural encoders remain strictly tied to the specific acquisition setup seen during training. Furthermore, existing upscaling methods usually assume ideal low-order inputs, ignoring device-specific array artifacts. We present ADEUPS, a unified framework that couples an array-independent Ambisonic diffusion prior with an explicit physical acquisition model during inference, enabling zero-shot encoding and upscaling across diverse array configurations. We train our prior using an order-masking strategy and perform inference via score blending, which we demonstrate to be crucial for solving the underlying inverse problem. Through extensive numerical evaluations across various array geometries, we show that ADEUPS consistently outperforms task- and array-specific baselines. Finally, we conduct subjective listening tests to confirm the perceptual effectiveness of our approach.

0.12

Fewer Heads, Fewer Memories: Pruning Redundant Attention Heads Suppresses Verbatim Memorisation

cs.CL cs.LG ·2026-10-29 ·Aryan Chowdhury, Riza Batista-Navarro
benchmark
Abstract · 7 Pages, 5 Tables, 1 Figure

Large language models have been shown to memorise and reproduce verbatim sequences from their training data, raising concerns about privacy leakage, intellectual property exposure, and the limits of generalisation. Recent theoretical work suggests that attention heads may contribute to this behaviour by functioning as associative memory mechanisms, with memorisation capacity scaling with the number of heads. At the same time, empirical work has shown that many attention heads are redundant and can be pruned with limited impact on downstream performance. Despite these parallel findings, the effect of pruning redundant heads on memorised-sequence reproduction remains unexplored. In this work, we demonstrate that zeroing out a substantial fraction of attention heads prior to inference can significantly suppress verbatim memorisation while preserving, or even improving, language modelling performance on the benchmark used to identify redundant heads. Evaluating GPT-2-XL(1.5B) on known memorised sequences, we find that pruning 729 task-redundant heads reduces the memorisation count by 91.5%, while improving WikiText-103 perplexity by 0.92. These findings suggest that attention-head redundancy can be exploited to mitigate memorisation at inference time, without retraining or modifying the training data.

0.12

ShallowDefend: Harnessing Shallow Safety Alignment for Multi-Turn Defender

cs.CL ·2026-10-29 ·Qiyao Ma, Xiaodong Liu, Weiwei Yang, Mingqian Feng, Chengquan Guo +7 more
Abstract

Safety alignment has enabled large language models to reliably refuse harmful requests directly, yet the same request distributed across a multi-turn conversation still succeeds far too often. This brittleness has a mechanistic root: Alignment is shallow, with safety behavior concentrated in a model's first few output tokens, and existing defenses inherit the weakness, supervising the surface form of the response without ever giving the model an explicit safety decision to learn. We take the opposite view of shallowness: if a short span determines the trajectory of an entire response, that span is the highest-leverage place to intervene, and the defender can occupy it first. We propose Assess $\to$ Decide $\to$ Respond, a framework in which the defender assesses the user's intent, commits to an explicit, parseable safety decision, and only then generates a reply conditioned on it, with the commitment trained by reinforcement learning at the payload turn. Because the decision is a token span, the reward can be a rule: a self-verified reward checks the committed label against the conversation's known nature, requiring no judge and no annotation, while the same format also supports an LLM-as-judge reward over a graded decision space with safe redirection. Across two model families and fourteen benchmarks spanning safety, utility, and helpfulness, our defenders achieve the strongest multi-turn safety in a controlled comparison, maintaining their lead in safety while preserving utility at a modest cost to helpfulness. A per-token KL analysis and interventional probes confirm why this works: training concentrates its effect on the brief deliberation that sets a response's trajectory, and the decision committed there is a causal control point. The very shallowness attackers exploit becomes the mechanism of defense.

0.12

Adaptive Personalization Under Preference Epistemic Uncertainty

cs.CL ·2026-10-29 ·Priyanka Dey, Rajarshi Das, Vikas Bahirwani, Qi Zeng, Julia Kruk +1 more
benchmark
Abstract

Human preferences change over time: an established preference may remain valid over time, become uncertain as context shifts, or be explicitly revised. Effective personalization requires conversational AI to adapt its response strategy to evolving preferences rather than relying on passive memory retrieval. We formalize this capability as adaptive personalization under epistemic uncertainty. A human preference study further confirms that people favor state-contingent response strategies as preference uncertainty changes. To evaluate whether language models mirror this adaptation, we create a multilingual benchmark spanning 250 personas, 5 languages, and 25K conversational turns that decouples strategy selection from response execution. Across eight closed- and open-weight models, we uncover a recognition-generation gap: models reliably identify appropriate strategies when selecting among candidate options, but collapse during open-ended generation, suffering a 35-46% drop in strategy calibration. Crucially, this failure is systematic: models default to exploit-first behavior, over-relying on existing preferences regardless of preference epistemic uncertainty. We show this failure is addressable: while prompting interventions fail, targeted training that supervises models' internal strategy representations toward state-conditional human preferences narrows the user acceptability generalization gap by a third on out-of-distribution personas and conversational styles.

0.12

Upcycling Language Models for Multiple Specialized Domains

cs.CL cs.AI ·2026-10-29 ·Mingzi Cao, Xi Wang, Nikolaos Aletras
Abstract

Effective multi-domain adaptation of large language models requires prioritizing relevant domain knowledge while minimizing interference from irrelevant domains. Existing approaches commonly use Low-Rank Adaptation (LoRA) to train a separate adapter for each domain and combine these adapters within one model. However, they either weight adapters equally regardless of the input domain or dynamically weight adapters while still activating all of them including irrelevant ones. Additionally, while efficient, their low-rank structure limits performance. In contrast, upcycling is a model expansion method that adds experts and a router to a dense Transformer layer offering larger capacity than LoRA approaches. In this paper, we propose Domain-Specialized Upcycling (DS-Upcycling) for multi-domain adaptation. We introduce a new domain-aware supervision that allows the router to effectively learn to distinguish domains by selecting the appropriate experts. During fine-tuning on each new domain, we update only the corresponding expert and router while freezing all other parameters. Experiments show that our method achieves the highest average performance across the adapted domains compared to baselines while retaining general performance. With just 256 examples per domain for router supervision, DS-Upcycling achieves 90-98% routing accuracy after adaptation compared with 43-70% for routers trained without our approach.

0.12

Textual Response-Aware Guidance for Zero-Shot Adversarial Robustness of Vision-Language Models

cs.CV cs.AIcs.LGcs.MM ·2026-10-29 ·Yubo Cui, Xianchao Guan, Zheng Zhang
Abstract

Vision-language models (VLMs) exhibit strong zero-shot generalization but remain vulnerable to adversarial attacks. Existing adversarial fine-tuning methods often use pre-trained model guidance to preserve image--text alignment while improving robustness. Our analysis of visual representation drift reveals complementary limitations of two representative forms of guidance. Matching visual features constrains all directions equally, providing broad coverage without semantic emphasis. Matching prediction distributions prioritizes semantically relevant directions, but remains limited by the candidate texts and prediction probabilities. These limitations can hinder preservation of pre-trained image--text relationships and degrade zero-shot performance. We therefore propose Textual Response-Aware Guidance (TRAG), which constrains visual representation drift using the pre-trained text encoder's responses to local variations in input embeddings. TRAG constructs anchor-level constraints from the response directions and magnitudes around semantic anchors, then aggregates them using pre-trained image--anchor relevance, yielding an image-adaptive constraint. Extensive experiments across architectures and tasks demonstrate state-of-the-art clean and robust performance. Further analysis shows that TRAG helps preserve pre-trained image--text relationships.

0.11

Red-Teaming Large Language Models in Low-Resource African Languages: A Case Study of Yoruba and Nigerian Pidgin

cs.CL cs.AI ·2026-10-29 ·Paula Chikaodinaka Anyanwu, Valdini Douglace Lemofouet, Blessing Ngozi Uzor, Ifeoluwa Dorcas Awotunde, Danielle Blanche Kapsa +5 more
Abstract

As large language models (LLMs) become more widespread in Africa, their safety in low-resource languages remains largely untested since most red-teaming work focuses on English. This study evaluates six state-of-the-art LLMs using 400 explicit adversarial prompts in Yoruba and Nigerian Pidgin, categorised into four harm categories. The responses were scored by native-speaker annotators and reliability was assessed using Cohen's $κ$. While models that rejected up to 69% of English prompts rejected as few as 0% in Yoruba, this was not because they were safer, but because they failed to understand the prompts. Attack success rates increased for most models in the Nigerian languages, with GPT-4o Mini demonstrating the most significant cross-lingual safety discrepancy (+0.23 in Yoruba). When models did engage with adversarial prompts, many of the responses were harmful, with rates of harmful content reaching 0.63 in English and 0.50 in Nigerian Pidgin. The most difficult harm category to detect was cultural stereotypes and discrimination (ASR = 0.44 in Nigerian Pidgin). Comprehension in Yoruba was low across most models, and this directly suppressed apparent attack success without indicating actual safety. Gemini 2.0 Flash was the most consistent model across languages. Despite its explicit training in African languages, Aya 101 had the highest attack success rate in English (0.60) and one of the lowest Yoruba comprehension rates (0.28). These results demonstrate that English-centric safety alignment does not transfer to Nigerian languages, highlighting the urgent need for language-specific safety evaluations.

0.11

Cord: Confidence-Anchored Order Residual Decoding for Masked Diffusion Language Models

cs.CL cs.AI ·2026-10-29 ·Chengwei Zhou, Zhaoyan Jia, Massoud Pedram, Gourav Datta
Abstract

The decoding order of masked diffusion language models (dLLMs) can substantially affect generation quality, yet common decoders typically apply the same reveal rule throughout generation. Confidence provides a strong default, but it measures how decisively a position can be predicted now, not how useful revealing that position will be for subsequent decoding. We propose Cord, a lightweight prompt-conditioned policy that retains confidence decoding as a stable reference and learns when and how to depart from it through geometric residuals over generation direction, proximity to revealed context, and two-sided infilling. The residuals adapt to both the prompt and the evolving partial response. Rather than sampling a new policy action at every decoding step, Cord samples a single low-dimensional coefficient matrix per rollout, which induces an entire state-dependent reveal trajectory. This yields a contextual-bandit formulation trained directly from sequence-level rewards, without step-level credit assignment or backbone updates. Across three dLLMs and five reasoning, code, and constraint-satisfaction benchmarks, Cord improves average task success from 32.0% under confidence decoding to 48.2%, and further improves over the best per-domain fixed residual. Behavioral audits show that the learned residuals induce substantial, task-dependent changes in reveal locality and two-sided infilling. Cord adds only $0.53$M trainable parameters, requires no additional backbone forward passes, and incurs only 5.39% inference latency overhead.

0.11

GROME: GROunding Factorized MEmory into Video World Models

cs.CV cs.LG ·2026-10-29 ·Fang Yu, Xiaoxue Wu, Haiyu Zhang, Yangfu Li, Botong Zhao +4 more
Abstract

Long-horizon consistency is essential for video world models, yet it becomes difficult to maintain once observed content falls outside the context window. Existing memory mechanisms either retrieve history as implicit context, which remains constrained by context capacity and entangles static and dynamic information or consolidate history into explicit global representations, which struggle to model evolving dynamics and accumulate geometric errors. To overcome these limitations, we introduce GROME, a novel factorized memory framework that augments video world models for long-horizon coherence across both static scenes and dynamic entities. GROME operates via three synergistic components: 1) Static-Dynamic Factorization decouples historical evidence into static and dynamic components, preserving static scene geometry and dynamic entity states separately; 2) Local Memory Retrieval selects spatially aligned static history to eliminate geometric drift alongside the latest dynamic entity states to ensure temporal continuity; 3) Training-Free Memory Injection seamlessly grounds the factorized memory into video world models via their native conditioning pathways. Experiments on long-horizon revisit scenarios demonstrate that GROME substantially enhances scene persistence and dynamic entity fidelity over existing methods, all while preserving the visual quality and controllability of the frozen backbone without additional training. Beyond self-rollouts, GROME readily accommodates external observations as plug-and-play memory, unlocking flexible and controllable video continuation and scene exploration.

0.10

Latent Path Exploration via Potential-Guided Subtrajectory Balance

cs.CL cs.LG ·2026-10-29 ·Xiaoling Zeng, Qi Zhang
Abstract

Continuous latent reasoning enables language models to perform intermediate computation without generating explicit reasoning traces. Advances in stochastic latent reasoning allow multiple paths to be sampled in continuous space, but stochasticity alone does not ensure effective exploration. A key challenge is guiding stochastic transitions between intermediate latent states toward useful reasoning paths. We introduce SubLAT, which couples a relative answer potential with Subtrajectory Balance to learn stochastic latent exploration. The relative potential evaluates the answer support of sampled states against replacement states, with a correctness signal at the terminal state. Subtrajectory Balance relates potential differences to forward and backward transition densities across subtrajectories of different lengths. The model is jointly trained with answer supervision, CoT supervision, and representation distillation. Experiments on arithmetic and deductive reasoning show increasing solution coverage with larger sampling budgets across in-domain and out-of-domain benchmarks, as well as improvements on several structural OOD benchmarks. Trajectory analyses reveal progressive divergence among sampled latent paths and higher mean relative potentials for correct trajectories across arithmetic benchmarks.

0.10

Discovering Interpretable Acoustic Features in Autoregressive Voice-Cloning TTS Models

eess.AS cs.CLcs.LGcs.SD ·2026-10-29 ·Anuj Diwan, David Harwath, Eunsol Choi
Abstract · Under review

Voice-cloning TTS models synthesize an unseen speaker's voice from brief speech prompts, but it remains unclear which acoustic properties they represent and use. We introduce a feature-based interpretability framework for autoregressive voice-cloning TTS models, analyzing features from sparse autoencoders (SAEs), PCA, and random directions via sparse probing, steering, and quantitative auto-interpretability. We focus on Qwen3-TTS and extend probing and steering to $5$ additional models to test generalization across architectures. These models encode concepts like pitch, energy, noise, and accent, but sparse prediction and steering are effective only for some concepts. Strong probing does not always translate to steering or vice-versa. SAEs are useful for sparse probing of human-labelled concepts and novel-concept discovery, while PCA is competitive for dense probing and steering. Auto-interpretability finds $684$ automatically validated SAE features in Qwen3-TTS covering speaker clusters, speaking style, and recording conditions beyond known concepts. Overall, we perform a comprehensive interpretability study of autoregressive voice-cloning TTS models.

0.10

UniRA: Unified Representation Alignment and Auto-Encoding for Efficient Diffusion Generation

cs.CV ·2026-10-29 ·An Lanji, Dawei Liu, Jin Li, Haoran Xu, Mei Chen +1 more
Abstract · 7 pages, 5 figures; submitted to ICML2027;

Diffusion transformers (DiTs) have become the dominant architecture for high-fidelity image synthesis, yet they suffer from slow training convergence and high computational cost. Two recent lines of work address this from complementary angles: Representation Alignment (REPA) accelerates DiT training by aligning intermediate features with a pretrained external representation model (e.g., DINOv2), while Representation Auto-Encoders (RAE) replace the pixel-space VAE with a compact representation-space tokenizer learned from self-supervised features. However, these approaches are developed and applied in isolation: REPA still relies on a pixel-space VAE for latent diffusion, and RAE does not exploit the alignment signal that accelerates the diffusion backbone. We present \textbf{UniRA}, a unified framework that couples representation alignment and representation auto-encoding into a single objective. UniRA learns a representation-space tokenizer whose output serves simultaneously as the diffusion target and the alignment anchor, so the tokenizer, the diffusion backbone, and the external representation model are co-optimized. We give a theoretical result showing that this joint optimization preserves the representation quality of the external model while accelerating diffusion convergence, and we derive a principled weighting that balances the reconstruction, alignment, and generation objectives. On ImageNet class-conditional generation at 256*256 and 512*512, UniRA matches or surpasses state-of-the-art DiTs (SD3, REPA, RAE) in FID while reducing training steps by up to 5*. We further show that the unified representation space supports improved zero-shot classification and linear probing, confirming that generation and understanding benefit from a shared, aligned latent space.

0.09

EgoHSI3R: Online 4D Hand-Scene Interaction Reconstruction from Egocentric Video

cs.CV ·2026-10-29 ·Mingzhi Pei, Zhangchen Ye, Ruoqu Chen, Liu Cao, Mengdi Xu
Abstract

Egocentric videos offer an abundant source of hand-scene interactions for embodied learning, which requires recovering hands, camera, and scene consistently in a shared 3D world. Existing methods focus on either camera-frame hands or world-space hand trajectories, without jointly recovering the scene pointmaps or interaction information. Building a complete reconstruction from such specialized predictions therefore typically requires per-sequence optimization. We propose EgoHSI3R, a unified feed-forward framework for online 4D hand-scene interaction reconstruction from egocentric monocular video. To jointly recover these components, it maintains a recurrent scene representation that accumulates geometric evidence over time and provides a shared world space and scale. We introduce hand-centric prompts that fuse pretrained hand priors and local scene context, while MANO vertex queries attend to scene features to predict dense per-vertex contact. These designs enable a single network to jointly estimate camera motion, dense scene geometry, world-space bilateral MANO hands, and per-vertex hand contact. Despite operating causally, EgoHSI3R achieves the lowest world-space hand joint error on ARCTIC among the evaluated methods, while offering faster inference than existing world-space hand reconstruction pipelines and competitive camera and scene reconstruction accuracy.

0.09

VEvo-Harness: Causally-Guarded Self-Evolving Harness for Agentic Image Generation

cs.CV ·2026-10-29 ·Yuan Wang, Yuhang Yu, Mengting Chen, Tianrun Xu, Zihao Pan +2 more
Abstract

The central bottleneck in visual generation and editing has migrated from achieving pixel-level fidelity to satisfying densely constrained user specifications through increasingly agentic workflows. Prevailing agentic generation pipelines remain shallow generate-verify-rewrite loops treating VLM judgments as coarse visual feedback, either discarding repair experience or admitting ungated updates that poison skill library. With memorize and govern periods left open, they degrade into ungoverned iterators of repeated trial-and-error rather than continual self-evolving harness. Towards this issue, we present \textbf{VEvo-Harness}, the first causally-guarded continual self-evolving harness system for agentic visual generation. VEvo-Harness introduces four coupled innovations: (1) \textit{a visual traceback} that densifies sparse judgments into root-cause attributions with a hierarchical verification tree; (2) \textit{a visual continual governance} that admits only causally valid skill patches through counterfactual dual-arm image generation replay; (3) \textit{a lifecycle-managed harness state} that hierarchically distills root-cause-anchored episodes across working, experience, and pattern tiers into a persistent skill library of routable and causally verified capabilities; (4) \textit{a heterogeneous sandbox policy} that orchestrates a diverse action space of repair primitives, dynamically routing cost-aware interventions based on root causes to select the most effective strategy. VEvo-Harness surpasses state-of-the-art agentic methods and rivals proprietary generators (\textit{e.g.}, GPT-Image, Nano-Pro) while maintaining the continual harness evolution and generalization.

0.09

Learning a Motion Manifold to Constrain Ambiguity in Unsupervised 4D Medical Image Interpolation

cs.CV ·2026-10-29 ·Haojin Li, Huazhu Fu, Heng Li, Jiang Liu
Abstract

Unsupervised 4D medical image interpolation reconstructs intermediate anatomical states from temporally sparse observations without intermediate supervision. Distinct motion fields may attain comparable endpoint agreement while producing different intermediate anatomy, giving rise to motion ambiguity. This ambiguity is difficult to constrain when many independently adjustable motion variables must be estimated from the observed endpoints. We develop a framework that learns a motion manifold to restrict candidate motion fields while retaining sample-specific variation. Cross-Sample Spectral Basis Learning extracts recurring motion structure from endpoint-derived dense estimates and encodes it in a shared spectral basis. Sample-Conditioned Cayley Basis Reorientation uses predicted motion coordinates to rotate the effective motion directions for each endpoint pair without increasing the number of independent coordinates. Their composition defines a nonlinear family of complete velocity fields across endpoint pairs, accommodating anatomical and motion-state differences within a restricted representation. Integrating each predicted field supports bidirectional reconstruction at arbitrary intermediate times. Theoretical analysis establishes sufficient conditions under which endpoint observations distinguish nearby motions within the learned family. Experiments on ACDC and 4D-Lung demonstrate state-of-the-art performance among the evaluated methods trained without intermediate volumes and competitive results against methods using intermediate supervision.

0.09

SpeakerTurn: Target-Speaker Semantic Turn Detection in Multi-Speaker Environments

eess.AS ·2026-10-29 ·Nan Li, Hetao Shan, Jintao Liang, Jiahao Yan, Longbiao Wang +2 more
Abstract · 6 pages, 2 figures, 3 tables

Semantic completeness alone cannot determine when an assistant should respond in a multi-speaker environment: a completed utterance may belong to an interfering speaker, while the designated user is still speaking, acknowledging, or requesting time. We therefore study semantic turn inference conditioned on speaker identity. We propose SpeakerTurn, a framework that jointly predicts the target transcript and turn state from a speech mixture and an independent target-speaker reference. Built on CocktailASR-1, it links semantic decisions to the corresponding target content through joint supervision, distinguishing complete, incomplete, backchannel, wait, and target-absent states, with a separate interruption task. We also examine whether structured chain-of-thought (CoT) supervision improves decisions by representing speaker presence and semantic cues. On a synthetic multi-speaker test set, SpeakerTurn achieves five-state macro F1 of 0.811 and target-transcription character error rate (CER) of 4.20%, correctly resolving all ten reference-swapped interruption pairs. Joint-format transcription and native-interface recognition nevertheless behave differently, and CoT improves validation transcription without improving turn macro F1. These results show why target attribution, semantic turn discrimination, and recognition retention require separate assessment when adapting speech language models for speaker-conditioned interaction.

0.09

What Makes a Good Audio Prompt for Zero-Shot TTS? Learning Prompt Selection from Synthesis Outcomes

cs.SD ·2026-10-29 ·Arnon Turetzky, Yossi Adi
Abstract

Zero-shot text-to-speech (TTS) models clone a target voice from a short reference clip, commonly referred to as the prompt. In practice, the prompt has to be cut from a longer recording of the speaker, so which part to use must be chosen. To understand what makes a good prompt, we analyze hundreds of thousands of outputs synthesized by four TTS models from candidate prompts of YouTube recordings. The choice of prompt substantially changes intelligibility, naturalness and speaker similarity. Standard predicted-quality scores capture mainly the naturalness part but miss most of the intelligibility gain. We find that the models largely agree on which prompt yields the highest measured speaker similarity, but much less on which prompt yields intelligible speech. We propose a small network that learns from the outputs a TTS model produced to rank a recording's candidate prompts from their audio alone. Without synthesizing anything at selection time, it picks prompts that lower the word error rate on held-out recordings by a third (Qwen3-TTS) to two fifths (F5-TTS) relative to a randomly chosen prompt, more than predicted quality achieves and nearly as much as synthesizing a dozen test sentences from every candidate first.

0.08

Cross-Encoder Anchor Refinement for Zero-Shot Anomaly Detection

cs.CV ·2026-10-29 ·Kaisong Ji, Ziyang Xu, Chang Bian, Yu Zhou, Wenyu Liu +1 more
Abstract · 15 pages, 6 figures

Zero-shot anomaly detection aims to transfer anomaly discrimination from auxiliary training domains to unseen categories while localizing subtle defects. A key challenge is to construct discriminative references that combine contextual representations with complementary local evidence. We present Cross-Encoder Anchor Refinement (CEAR), which connects two visual encoding paths for the same image at the anchor stage, before normal and anomalous prototypes are constructed. An independent convolutional encoder pretrained for industrial product recognition builds an image-specific memory shared across selected vision transformer (ViT) layers. Spatial anchors read from ViT patch features query this memory through cross-attention and undergo residual refinement. Image-conditioned normal and anomalous queries then gate anchor aggregation to construct prototypes for anomaly scoring. Across 13 industrial and medical datasets, CEAR achieves the best performance in 57 of 68 evaluations spanning image-level detection and pixel-level localization. Notably, CEAR surpasses all other state-of-the-art methods in 41 of 44 pixel-level comparisons. Across 11 localization datasets, mean gains in pixel-level maximum F1 score and average precision are 4.45 and 3.79 percentage points, respectively. This study shows how cross-encoder interaction at the anchor stage can integrate complementary visual representations for fine-grained anomaly localization across domains.

0.08

The Generative Gap Between Speculative Drafter and Continuous Diffusion Language Model

cs.CL cs.AI ·2026-10-29 ·Yujian Chen, Hao Jiang, Guang Cheng, Weijian Luo, Tianyang Hu
Abstract

Continuous diffusion language models (CDLMs) and speculative drafters offer efficient language generation but are typically evaluated differently: CDLMs generate independently through iterative refinement, whereas drafters propose text for target-model verification and correction. We compare their architectures without verification, using LangFlow for CDLMs and adapting DSpark's Markov and recurrent-head architectures for standalone generation. All models are trained from scratch for unconditional, single-block OpenWebText generation at lengths 16, 128, and 1024, without target features or verification, using common preprocessing, backbone dimensions, and training-update budgets. We evaluate generation perplexity (GenPPL) with an external language model and distributional similarity to reference text with MAUVE. With default denoising steps equal to sequence length, LangFlow outperforms both DSpark variants on both metrics at lengths 128 and 1024, while DSpark-RNN leads at 16. A conditional KL analysis characterizes dependencies excluded by the Markov head's previous-token conditioning. Recurrent memory removes this restriction and improves both metrics, but the longer-sequence gap persists. Teacher probability matching further lowers Markov GenPPL without observed MAUVE gains. Increasing LangFlow's denoising steps at fixed weights improves both metrics at lengths 16 and 128, changing the length-128 ordering relative to DSpark-RNN. These results establish LangFlow's longer-sequence advantage under the evaluated budgets and highlight iterative full-sequence refinement as a route to improving standalone generation beyond gains from recurrent head memory.

0.07

Multi-View Latent MoE: Letting Every Router Shape Expert Computation

cs.CL ·2026-10-29 ·Zelin Zhao, Junyuan Shang, Tianchi Ma, Shuohuan Wang, Dianhai Yu
Abstract

Sparse mixture-of-experts (MoE) models expand parameter capacity by activating a small subset of experts per token. Expert selection directly affects model capabilities under a fixed active-expert budget. Dense evaluation of all experts reveals weak agreement between single-router rankings and expert-utility evaluation rankings, especially in middle and deep layers. Richer signals for selection and computation incur additional cost: selection based on internal expert activations requires preliminary computation, while processing multiple full representations separately through the same expert repeats its computation. We propose Multi-View Latent MoE to improve expert selection with multiple routers and reuse their preferences to modulate expert computation under a fixed expert budget. Its Multi-Router submodule aggregates sigmoid scores from independently initialized routers. These routers develop highly differentiated expert preferences during training, allowing experts strongly favored by one router or jointly supported by several routers to be selected. Multi-View further uses each selected expert's relative scores across routers to mix learned latent views into its input, reusing routing preferences for both selection and computation. Each selected expert executes one forward pass, preserving the active-expert count, expert FLOPs, and all-to-all communication volume. On a 24-layer 10B-A1B model trained on 708B tokens, Multi-View Latent MoE improves the nine-task downstream mean by 1.18 percentage points over the baseline. Multi-Router alone yields a 0.6-percentage-point gain and improves the rank correlation between router scores and evaluated expert utility, with relative increases of 8.1% globally and approximately 58% in the layer with the lowest single-router alignment. Controlled mixing interventions show the model uses input differences that depend on the expert and token.

0.06

Rethinking Pre-pretraining

cs.CL cs.AI ·2026-10-29 ·Alexander Stern, Tim Rocktäschel, Acyr Locatelli, Laura Ruis
Abstract

We report that allocating compute to synthetic pre-pretraining (PPT) never outperforms pure natural-language training in our compute-matched experiments. This holds across four PPT curricula (k-Dyck, neural cellular automata (NCA), declarative reference corpora, and code) and three training regimes: data-constrained, data-unconstrained, and a high token-to-parameter-ratio regime (881M parameters trained on 50B tokens). When natural language is scarce, repeated epochs over a 100M-token corpus beat every synthetic curriculum on natural-language validation loss, with no consistent downstream accuracy advantage from PPT. We find that PPT gains reported in prior work are sensitive to the weight initialization scheme. In an NCA reproduction, we recover a 5.4% perplexity improvement from PPT under the original initialization scheme. An alternative scheme improves both methods, but the no-PPT baseline more, closing the gap between them. However, when PPT receives additional compute rather than a share of the natural-language budget, gains persist at small natural-language budgets even with improved initialization, but disappear by 1B tokens. Finally, models forget what they learn during PPT in subsequent natural-language training; interleaving the same data throughout training prevents forgetting, but does not significantly improve natural-language performance. Our results suggest limited benefits of synthetic pre-pretraining as a distinct training stage and motivate rethinking how such data are used in language model training.

0.05

Raincheck: Deferred Multimodal Prefill for Efficient Multi-Turn MLLM Inference

cs.CL cs.AI ·2026-10-29 ·Wenjie Huang, Jun Zhang, Hongyaoxing Gu, Haoyang Huang, Tianqi Xu +8 more
Abstract · Withdrawn at the authors' request pending completion of the required corporate disclosure and publication approval process

Multi-turn conversations with multimodal large language models (MLLMs) reuse input context, but long multimodal inputs increase time to first token (TTFT). Query-aware hard pruning reduces prefill tokens but permanently excludes information that later questions may require. We propose Raincheck, a framework for deferred multimodal prefill that decouples initial token selection from later information availability. An existing pruning method selects tokens for initial prefill; Raincheck preserves and progressively prefills unselected tokens in source order using unused compute capacity during decode. Unfinished prefill continues after answer return during available scheduling gaps. Attention visibility control and affine state mapping enable context updates after generation begins in full-attention and hybrid recurrent models. Prefill-decode fusion reduces repeated weight reads. We evaluate Qwen3-VL-8B and Qwen3.5-9B with 30% initial multimodal-token retention. Across both models and four pruning methods, Raincheck recovers on average 77.7% and 60.4% of the quality lost to matched hard pruning on MultiVerse and ConvBench, respectively. On MultiVerse, the SGLang implementation reduces mean first-turn TTFT by up to 24.1% relative to full prefill, with less than 0.5% increase in mean time per output token (TPOT).

Security & Privacy 45

PairWatch: Pair-Aware Anomaly Transformer for Detecting Adversarial Rendezvous in Public Satellite Catalogs

cs.CR ·2026-10-29 ·Young Eun Kwon, Ji Won Yoon
top venuebenchmarkpositioncode/data
Abstract · 5 pages, 4 tables. Accepted at the 35th ACM International Conference on Information and Knowledge Management (CIKM '26), Rome, Italy. Code a…

Documented adversarial rendezvous and proximity operations (RPO) appear in the public TLE catalog, yet existing TLE-based maneuver detectors treat each satellite in isolation - mistaking routine station-keeping for adversarial behavior and missing RPO entirely when the target stays passive. We argue that the right unit is the satellite pair and the right inductive bias is the kinematic symmetry between approach and recede. We instantiate both with PairWatch, a pair-aware AnomalyTransformer whose attention prior is augmented with a sign-symmetric geometric similarity over relative-motion features, and release an open benchmark of 359 windows from public Space-Track data spanning 11 documented RPO events, 100 Clohessy-Wiltshire synthetic adversarial windows, and a five-class hard-negative set culminating in the TerraSAR-X / TanDEM-X helix formation. Under the deployment-relevant criterion of a single threshold separating real RPO from operational negatives - a necessary property for catalog-scale screening - PairWatch is the only model that succeeds, with the property holding in 100% of bootstrap resamples versus 0-19.5% for three neural baselines and under 0.1% for four physics-informed detectors. A per-source analysis exposes a shortcut-learning failure mode of single-number AUROC: simpler baselines attain higher overall AUROC by shortcutting a synthetic-vs-formation contrast with no operational meaning; removing this single cell flips the test ranking entirely (LSTM-AD), and these baselines fail the security-relevant task the benchmark was built to measure.

0.57

Towards Elastic Least Privilege for Agent Planning

cs.CR ·2026-10-29 ·Simon Parkinson, Zhipeng Wang, Mustafa Mustafa
Abstract

General-purpose AI agents require authority that changes between tasks and during execution. Common deployments trade broad permissions that expose unnecessary authority and narrow permissions that raise approval requests. Existing mechanisms can enforce dynamic authority, but agent tool interfaces rarely provide the agent's planner, the component that chooses its actions, a means of comparing the authority required by alternative actions or of seeing the cost of the authority it already holds. Drawing on the rich history of access control research, we propose the concept of \emph{elastic least privilege}, in which an agent begins with a small baseline of authority and may acquire more within an owner-defined envelope. Elevations are leased and measured against a replenishing budget. Their cost increases with power, proximity to mandatory safety invariants, and repeated requests. Costs are visible to the planner, which can prefer lower-privilege choices and release authority it no longer needs. Authority returns automatically to the baseline when a lease expires, when it is released, or when the budget is exhausted, while persistent legitimate demand may prompt an owner-approved change to the baseline. Building on prior risk-priced and budget-aware access control, we redesign these mechanisms for agents. How best to implement the approach and how well it works in practice are open research questions. In this paper, we present how the approach could work, illustrate the model, and set an open research agenda.

0.53

Investigating Data Flows from Android Games

cs.CR ·2026-10-29 ·Judit Borcsiczky, Lukas Moog, Boel Nelson
publisheddatasetcode/data
Abstract · Long version (including appendix) of paper accepted at Nordsec 2026. Accompanying dataset available at https://doi.org/10.7910/DVN/CAV2I2

Online applications can share data with arbitrary parties, without users noticing. In practice, these data flows could constitute privacy violations to individual users, with the lack of transparency leaving users unaware of the unwanted data flows. To understand the spread and scale of this privacy threat, we investigate what data is collected and where it is sent in popular free-to-play Android games. As part of our investigation, we conduct a measurement study where we collect and analyze over 16 GB of network traffic from 50 puzzle games. Additionally, we explore the effectiveness of user-side countermeasures for a selection of the apps. We find that while most countermeasures, such as ad-blocking DNS services, decrease data flows, withholding consent and opting out of data collection can produce inconsistent behavior including increased outgoing traffic. To encourage further analysis, we release our measurements as an open dataset.

0.45

Usable String Matching by Humans through Overlaid Presentation

cs.CR cs.HC ·2026-10-29 ·Enka Blanchard, Ted Selker, Alan T. Sherman, Levi Gabasova, Mehdi Bouhamidi +1 more
published
Abstract · A shorter version of this paper (without the appendices) has been conditionally accepted to the International Journal of Human-Computer Stud…

We explore the ability of humans to compare strings, an error-prone activity which occurs in critical contexts (e.g., bank wire transfers), where it is not always possible to trust a machine to help. We hypothesize that understanding human cognitive and perceptual limitations can improve how people verify whether two character strings are identical. We vary the strings' relative positions and presentation formats to measure comparison speed and accuracy. We present results from two online experiments with 178 human subjects who compared pairs of strings to determine whether the two strings were identical. The strings varied by length (5 characters to small paragraphs), difference (matching versus making a random change or a hard-to-detect one), alignment (vertical versus horizontal), and presentation (Overlay versus Adjacent). Our experiments test whether, by encoding and memory tasks into perceptual tasks, overlay formats can offer advantages in speed and accuracy. Results show that overlay offers a substantially faster method, with improved accuracy for long strings. Nevertheless, mean error rates were at least 4 percent, indicating that even the improved approach may be insufficiently accurate for many verification tasks.

0.44

Evaluating Structural and SAT Attack Resilience in Cryptographic Dynamic Logic Locking via Keystream Capping

cs.CR ·2026-10-29 ·Nahush Tambe, Naseeruddin Lodge, Vineet Chadalavada, Fareena Saqib
benchmark
Abstract · 8 pages and 4 figures

Untrusted third-party entities throughout the semiconductor supply chain expose proprietary hardware intellectual property to reverse engineering, piracy, and overproduction. While logic locking counters this threat by embedding key-dependent logic, conventional static key-gates remain vulnerable to both oracle-guided SAT attacks and structural removal attacks due to their time-invariant netlist signatures. Dynamic schemes such as STREAMLOCK mitigate oracle attacks by continuously masking signals with cryptographic keystream, achieving near 50\% Hamming distance output corruption under incorrect keys with low hardware overhead. However, confining key-gates to primary outputs leaves internal combinational cones vulnerable to single-tap structural bypasses. This work extends dynamic logic locking by distributing key-gates within internal paths prioritized via fault-propagation analysis. The original and internally-hardened architectures are evaluated against structural removal attacks and a bounded, unrolling-based SAT-style attack across three ISCAS'85 benchmarks. Distributing masking internally raises the uninformed attacker's brute-force search space from 6 candidate node subsets to over $10^{45}$ for the largest benchmark. Crucially, empirical analysis reveals that topological relocation alone does not prevent algebraic key extraction when wide circuits expose numerous keystream bits per clock cycle. To eliminate this leak, a keystream-bit reuse cap is introduced to restrict per-cycle exposure independent of circuit width. Under this constraint, SAT-solver tractability encounters an immediate computational cliff, transitioning sharply from rapid under-constrained solving to timeout thresholds without disclosing the secret key. Integrating fault-aware internal placement with keystream reuse capping delivers joint resilience against structural bypasses and bounded SAT attacks.

0.40

Appreciation Teaches Unauthorized Execution: Self-Evolution Degrades the Robustness of Personal Agents to Indirect Prompt Injection

cs.CR cs.AI ·2026-10-29 ·Xinyan Wang, Xiaogeng Liu, Jun Shao, Chaowei Xiao
code/data
Abstract

Self-evolving personal agents act for users and learn from user feedback. However, if an agent obeys an instruction that an untrusted third party hides in content, the third party gains control of the agent. Such indirect prompt injection (IPI) poses severe risks. We investigate whether self-evolution degrades the robustness of personal agents to IPI. To study this question, we design a controlled two-phase methodology. In the adaptation phase, the user asks the agent to read external content and complete the task described in it, a common request in daily use. The agent then learns from the user's feedback and improves itself as designed. In the test phase, a third party writes a similar task into new external content, but the user's test request does not authorize the injected task. Using this methodology, we build multiple tasks with diverse test instances in different domains for a real personal agent. Our extensive experiments reveal that self-evolution substantially degrades robustness to IPI. Specifically, compared with the base agent, which does not evolve from user feedback, the average attack success rate rises by 14.0--48.8 percentage points under skill evolution and by 3.0--39.3 percentage points under reinforcement learning, posing safety risks. The code is available at https://github.com/SaFo-Lab/Evolve-IPI

0.39

From if to Oops: Turning a conditional into a key-recovery oracle

cs.CR ·2026-10-29 ·Gustavo Banegas, Nia Dimitrova, Monika Trimoska
position
Abstract

Deterministic variants of CSIDH remove randomness from the group action, either through implementation techniques or, as in Houben's deterministic CSIDH and OSIDH-LD, by design. Since earlier fault attacks on CSIDH mostly target random point sampling, determinism seems to close that avenue. We show that it opens another: every secret-dependent step occurs at a fixed, public point of every execution, so an attacker can fault the same step repeatedly. Our attack forces the mask that selects which of two lanes computes the next isogeny to zero, for example by skipping one instruction. Whether the output changes reveals whether a secret exponent reaches a chosen threshold, and a binary search over these thresholds recovers the full key with one fault per execution: at most 297 oracle queries for OSIDH-LD and 222 for deterministic CSIDH with CSIDH-512-sized keys. We recover full OSIDH-LD keys in a software fault simulation of the reference implementation, and show on a ChipWhisperer-Lite that the targeted instruction is skipped with probability at least 0.94 at an isolated fault position. Finally, we argue that fault resistance requires giving up determinism, and propose countermeasures relying on randomization.

0.37

The Privacy and Security Risks of Edge LLM Optimization

cs.CR ·2026-10-29 ·Mingchen Li, Ren-Yi Huang, Dumindu Samaraweera, Prashant Shekhar, J. Morris Chang
positionhot direction
Abstract

Edge deployment of Large Language Models requires optimization techniques that reduce memory, computation, and energy costs. However, preserving task performance does not necessarily preserve security. We argue that optimization is an information transformation that can alter representations, decision pathways, and trust boundaries, creating security drift and optimization-induced attack surfaces.

0.36

DYNAS: A Dynamic Network Penetration Testing Framework Using Deep Hierarchical Reinforcement Learning

cs.CR cs.LGcs.NI ·2026-10-29 ·Md. Rakibul Hassan, Majedul Islam, Md. Ashraful Alam, Muhammad Iqbal Hossain, Swakkhar Shatabda
published
Abstract · Accepted at the NSS 2026: 20th International Conference on Network and System Security (submission version)

As network scale and uncertainty complicate automated penetration testing, reinforcement learning (RL) environments must support controlled evaluation under defensive and non-stationary conditions. We present DYNAS (Dynamic Network Attack Simulator), which combines parameterized large-scale scenario generation with stateful intrusion detection responses, explicit scan false-positive and false-negative noise, temporary service unavailability, and duration-based action timeouts. We further introduce Pen-DHRL, a recurrent manager--worker policy that selects persistent subgoals using IDS-conditioned global context. We compare Pen-DHRL with GNN, Invariant, and Attention baselines across four scenario variants and multiple training regimes. Under IDS-disabled training followed by IDS-enabled evaluation, Pen-DHRL obtains approximately 64\% and 24\% higher reward per step on Bridge and Dynamic, respectively, and approximately 26\% and 30\% more captured hosts on VarA and VarB than the next-best baseline. With all defensive dynamics active from the beginning of training, Pen-DHRL achieves positive mean episode returns in all four scenarios, whereas the flat baselines produce negative mean returns in all 12 architecture--scenario combinations; staged curriculum training produces positive mean returns across all 16 evaluated combinations.

0.36

Adaptive Dynamic Security for Operational Technology Against AI-Powered Attacks

cs.CR cs.NI ·2026-10-29 ·Mohamed Chahine Ghanem
Abstract · 9 pages

Operational technology now faces attackers whose tempo is set by AI. We present ADS-OT, an adaptive dynamic security framework modelling defensive posture as a partially observable decision problem. A reinforcement-learning policy adapts monitoring, access, deception, and segmentation under a safety-governance layer, reducing attack-path completion, detection, and containment versus static baselines.

0.34

Syndrome Trellis Decryption in Post-Quantum Cryptography

cs.CR ·2026-10-29 ·Meir Ariel
published
Abstract · 21 pages, 6 figures

We present a novel post-quantum cryptographic framework based on masked high-memory convolutional codes in the Niederreiter paradigm. The public key is derived from a high-memory polynomial parity-check matrix augmented by dense random masking and additional invertible linear transformations. The resulting matrix exhibits strong random-like properties, effectively concealing its algebraic structure and resisting known structural and algebraic attacks. Legitimate recipients retain polynomial-time decoding capability, whereas adversaries appear to be limited to generic information-set decoding with exponential complexity. Under standard cryptanalytic estimates, the proposed construction achieves security margins exceeding those of classical McEliece- and Niederreiter-based systems by factors greater than 2^200. Beyond its enhanced security profile, the construction offers considerable design flexibility, supporting arbitrary plaintext lengths, linear-time decryption, and a uniform per-bit computational cost, enabling efficient scalability to very long messages. Practical implementation is facilitated by a parallel array of syndrome-trellis decoders, dynamically instantiated according to candidate syndrome values, enabling efficient hardware and software implementations with high throughput. Overall, the proposed scheme is a promising candidate for robust, scalable, quantum-resistant public-key cryptography.

0.33

AgentHeal: Evaluating Goal-Preserving Resource Recovery of Web Agents under Access Barriers

cs.CR cs.CL ·2026-10-29 ·Zhiyi Liu, Haohua Du, Haoqi Sun, Haisheng Tan
benchmarkcode/data
Abstract · Code and Dataset are available at https://github.com/Lzzzzzzzzzzy/AgentHeal

Large language model (LLM) agents have demonstrated strong capabilities in automating complex information-seeking and transactional tasks across the open web. However, their real-world reliability remains limited by access barriers, such as dead links, authentication walls, rate limits, and anti-scraping defenses. These barriers block the primary execution path, causing agents to abort prematurely, loop indefinitely, or quietly compromise the user's original constraints. Existing evaluations either implicitly assume that target resources are freely accessible or focus on transient perturbations that simple retries can resolve, failing to capture scenarios where agents must actively reroute around impassable bottlenecks. To bridge this gap, this paper introduces AgentHeal, a comprehensive benchmark evaluating goal-preserving resource recovery, defined as the ability of web agents to detect non-transient access barriers, pivot to legitimate alternative sources, and fulfill the user's exact original objective. Spanning nine barrier families across a counterfactual sandbox and real-website replay, our experiments on leading LLM backbones show that average task success drops sharply from 89.4% to 43.7% under access barriers. Further analysis reveals that agents frequently fall for deceptive look-alikes, fail to escalate to users when necessary, violate access policies, or waste computation on futile retries, providing actionable diagnostics for developing resilient web agents.

0.32

Closing the Applied Cybersecurity Gap in Critical Infrastructure: A Cyber-to-Physical Security Assessment Framework for Water Systems

cs.CR ·2026-10-29 ·Halil Dursunoglu, Kaan Sulkalar
Abstract

The integration of information technology and operational technology in water infrastructure creates pathways through which cyber compromise may affect physical service delivery. This study revisits the Applied Cybersecurity Gap (ACG) framework by evaluating an experimental Applied Cybersecurity Gap Index (ACGI) alongside the Cyber-to-Physical Propagation Rate (CPPR), defined as the proportion of initially compromised trials that produce a specified physical-impact outcome. The expanded evaluation comprises 2,160 synthetic trials across three water-network benchmarks, including EPANET Example Network 3 (Net3), and couples declared cyber-attack progression rules with hydraulic simulations. A source-trained machine-learning model evaluated on 247 compromised Net3 trials achieved ROC-AUC 0.772, PR-AUC 0.775, and F1 0.733, compared with 0.661, 0.642, and 0.704 for a scenario-only baseline. The configuration-cluster bootstrap interval for the ROC-AUC difference was 0.038--0.183. A separate 720-run Net3 hydraulic operating-point sensitivity analysis achieved the study's exploratory baseline pressure criterion in all reruns, preserved 717 of 720 physical-impact labels, and produced ROC-AUC 0.791 with the frozen source-trained model. Additional analyses examine the sensitivity of CPPR to the physical-impact threshold and assumed control-blocking probabilities, and demonstrate that the ACGI--CPPR association is partly induced by the simulation design rather than independently validated. The findings establish a reproducible simulation-based assessment workflow, but do not constitute field calibration of cybersecurity-control effectiveness or prediction of real-world utility incidents.

0.32

RV2GC: Reconstructing the Garbling Boundary in a RISC-V Garbled Processor

cs.CR cs.AR ·2026-10-29 ·Zixiao Feng, Xiaolin Li, Chenglu Jin, Wei Yan, Yong Zhang +3 more
Abstract · 12 pages, 9 figures, 7 tables

Programmable garbled-circuit execution must balance the benefits of cross-program reuse against the cost of repeated work during secure execution. A complete garbled processor provides a fixed circuit for many programs, but its evaluator repeatedly visits processor-wide structures even when only a small part performs the current protected computation. We present RV2GC, which reconstructs the garbling boundary around a fixed library of RISC-V instruction circuits. Public scheduling and operand routing operate outside the circuit topologies; persistent label-encoded state connects independently invoked garbled instruction circuits and instruction-specific garbling-free paths through a common interface. A Conditional-execution Stack and protected commits support secret-dependent branches, bounded loops, and secret-indexed memory access under public scheduling, keeping instruction and memory-address sequences independent of private inputs. Across nine benchmarks, measured ARM2GC-to-RV2GC runtime ratios range from 12.81 to 4,043. In terms of circuit size, the largest listed instruction circuit has 3,386 gates, corresponding to 2.044% and 0.393% of the 256-byte and 4,096-byte Amber processors, respectively. Additional comparisons quantify schedule storage, communication, and the cost of secret-dependent execution.

0.31

DICE: Decentralized Incentive-driven Federated Learning for Efficient & Scalable Customization of Pretrained Models in Web3

cs.CR cs.DCcs.LG ·2026-10-29 ·Hong Lin, Zhongle Xie, Lidan Shou, Ke Chen, Gang Chen
top venue
Abstract

Customizing Pretrained Models (PMs) for downstream tasks demands task-specific data, which burgeoning Web3 applications provide abundantly. Blockchain-powered federated learning (BCFL) offers a secure and privacy-preserving paradigm to customize PMs among self-interested and untrusted Web3 participants. Yet, existing BCFL algorithms face severe efficiency and scalability challenges when applied to large PMs, due to their substantial computational and memory requirements and the hardware constraints of many participants. This paper introduces DICE, a novel BCFL algorithm that addresses the efficiency and scalability challenges of customizing large PMs in Web3 via two sets of synergistic techniques: (i) a dynamic progressive learning scheme and a model slice verification scheme to enable a wider range of honest participants, including those unable to deploy the full model, to engage in customizing PMs, while preserving model utility; (ii) a tree-structured committee consensus protocol and an efficiency-driven incentive scheme to address the dominant bottleneck that hinders existing BCFL algorithms from scaling to a larger number of data holders for customizing data-hungry large PMs, while maintaining security. Theoretical analysis and extensive experimental results on customizing vision and language PMs demonstrate DICE's efficiency, effectiveness, and security.

0.31

InferD: A Honeypot Framework for AI Inference Infrastructure

cs.CR cs.AI ·2026-10-29 ·Hicham Hammouchi, Gabriele Lenzini
published
Abstract · AISec Workshop '26, ACM CCS 2026, The Hague, Netherlands

Large language model serving stacks, commercial inference APIs, vector databases, and agent orchestration gateways are now ex- posed on the public internet and indexed by commodity scanners within hours of deployment. The threat characteristics of this sur- face are only beginning to be measured, and the honeypot frame- works that could measure it were designed for conventional net- work services rather than for AI infrastructure. In fact, emulating an AI service is a distinct deception problem: the observable behavior of the service, including model inventories, token formats, tool-call protocols, and authorization flows, must be reproduced faithfully, while the underlying computation, model inference and tool ex- ecution, must never occur. We present InferD, a low-interaction honeypot framework organized around a single principle we call deception without inference: every request an adversary can observe is answered from static response pools, and nothing an adversary submits is ever executed. InferD decomposes the emulation into four deception planes, transport, service protocol, vulnerability surface, and agentic session, adds a provenance plane that links a planted credential to evidence of downstream model-mediated processing.

0.30

RecHash: Segmented Hash-Entangled Chains for Stateful Sequential Authentication

cs.CR ·2026-10-29 ·Luis Adrián Lizama-Pérez
theory
Abstract · 23 pages, 3 figures, 4 tables

RecHash is a stateful sequential authentication construction based on paired hash-chain openings and an authenticated record history. It develops the decreasing position relation of the 2019 hash-entangled construction, whose exact integer normalization yields a two-coordinate constant-sum antichain. Direct proportional reuse is constrained by integer capacity and multiplicative contraction. The main construction instead assigns each record to a predetermined chain segment, with complementary positions encoding each digest block. All digests fit every scheduled slot, independently of preceding messages. For an $\ell$-bit digest distributed over $k$ equal-length chain pairs, balanced allocation gives $\lfloor L/(2^{\lceil\ell/k\rceil}+1)\rfloor$ slots. A 256-bit digest, 32 pairs, and length 65535 provide 254 application records followed by a renewal certificate, with 2048 bytes of openings per record for 256-bit chain outputs. We give a complete segmented state transition, a worked trace, exact allocation and resource results, and a renewal-free security-composition theorem conditional on authenticated history and an adaptive one-step guarantee. The analysis relates the construction to constant-sum encodings and standardized hash-based signatures; a concrete post-quantum reduction and security across renewal remain open.

0.29

LiHA: Worse is Better for Lightweight Hash-based Source Authentication

cs.CR cs.NI ·2026-10-29 ·Negin Dehghanchaleshtori, Homit Dalia, Himanshu Chandra, Rohan Tripathi, Marc E. Fiuczynski +1 more
position
Abstract

The omission of source IP authentication remains a fundamental vulnerability in the Internet. Existing approaches rely on routing plausibility, topology assumptions, per-flow state, or wide deployment. These requirements have kept strong source authentication out of the data plane. As simplicity is the key to deployability, we present LiHA, a lightweight source authentication mechanism built around operations that routers already perform efficiently. LiHA binds secret keys to source prefixes using longest prefix match and authenticates each packet with a single message authentication code (MAC). The data plane requires no handshakes, per-flow state, or per-packet control plane access. A sender performs one prefix lookup and one fixed-size hash. The verifier repeats the same operations and forwards the packet only if the MAC matches. LiHA is designed for incremental deployment. It primarily protects against off-path attackers that spoof source prefixes they do not control. An on-path attacker can replay valid authenticated packets until the key rotates. LiHA does not eliminate every denial-of-service threat. Instead, it confines accepted traffic to authenticated source prefix space within participating networks. We argue that simpler source authentication with clear limits is more likely to be deployed. We implemented LiHA in three environments: as a set of user-space libraries, as eBPF programs, and in an FPGA. We found that the user-space implementation had very high overheads. The eBPF implementation got reasonable single thread performance even with a software based cryptographic hash. We found that only the FPGA could obtain 100 Gbps single thread performance, but that was only when bypassing the kernel using a DPDK user-level networking stack.

0.29

Prevented is Not Protected: What Attack Success Rate Hides about RAG

cs.CR cs.CLcs.IR ·2026-10-29 ·Chahana Dahal, Zuobin Xiong
code/data
Abstract

Retrieval-Augmented Generation (RAG) is vulnerable to knowledge poisoning, where adversaries manipulate retrieved evidence to influence model outputs. Existing defenses are primarily evaluated using attack success rate (ASR), which obscures defense-induced failures and the mechanisms behind apparent attack prevention. To address this issue, this work introduces \textbf{Aligned RAG Evaluation}, a matched-query protocol that jointly examines clean and poisoned responses, with and without defense. By combining matched conditions analysis, LLM-judge endorsement, and retrieval dependence controls, the proposed framework distinguishes prevented from introduced failures, target mentions from LLM endorsement, and single-passage influence from evidence composition. Experiments on HotpotQA and Natural Questions reveal that existing defenses can introduce unexpected target-positive answers, although they can prevent some attacks, while evaluation criteria can also change defense rankings. Our findings expose critical blind spots in ASR-based evaluation and motivate a more comprehensive assessment of RAG poisoning defenses. The code and the evaluation artifact are available at https://github.com/chxhana/aligned-rag-eval.

0.29

Systematic Evaluation of Adversarial Jailbreaking Attacks and Defenses Across Seven Large Language Models

cs.CR ·2026-10-29 ·Md. Robiul Islam Niloy, Abu Sayed Md. Mostafizur Rahaman
code/data
Abstract

Large language models (LLMs) deployed in safety-critical settings must resist adversarial inputs designed to elicit harmful outputs. We present a systematic empirical evaluation of five adversarial jailbreaking attack families-direct requests, roleplay-based jailbreaking, hypothetical framing, instruction override, and obfuscation-across seven open-weight LLMs served locally: Llama-3-8B, Llama-3.1-8B, Llama-3.2-3B, Mistral-7B, Qwen2.5-7B, Gemma2-9B, and Phi-4-14B. We further evaluate four defense mechanisms-keyword blacklisting, self-reminder prompting, paraphrase-based sanitization, and perplexity-based filtering-against two high-ASR models (Mistral-7B and Llama-3.1-8B) on a 20-prompt subset. Across 250 prompts per model, a content judge counted obfuscation attacks as harmful (HARMFUL or PARTIAL responses) at a pooled rate of 62.3%, the highest of any family, although only 7.1% of responses were rated HARMFUL. Hypothetical framing was the highest family for three of seven models. Obfuscation succeeded at nonzero rates on all seven. The content judge has not yet been validated against human labels. We find that model size does not significantly predict adversarial robustness (Spearman Rho =-0.64, p=0.12, n=7). A rule-based judge suggested a large gap between the two Llama-3 generations (16.8% and 40.8%); the content judge does not (17.2% and 18.4%). Self-reminder prompting gave the largest reduction, cutting ASR from 50% to 10% on Mistral-7B and from 45% to 5% on Llama-3.1-8B (40 percentage points on each model, against matched no-defense runs, scored with the rule-based judge). Our code and datasets are publicly available at https://github.com/Mdniloykhan/jailbreak-eval-7llms.

0.27

Cybersecurity Practices in Industrial Automation and Control Systems: a Study of Croatian System Integrators

cs.CR eess.SY ·2026-10-29 ·Filip Katulić, Damir Sumina, Stjepan Groš, Tena Velki
Abstract

Given the increasing frequency of industrial cybersecurity incidents, governments, industry, and the research community are all working to find ways to protect industrial automation and control systems (IACSs). A key regulatory effort in the European context is the EU's Network and Information Security Directive 2 (NIS2), which aims to strengthen cybersecurity requirements and improve resilience across critical and important sectors. Nevertheless, translating cybersecurity requirements into practical engineering remains challenging within IACS environments. This paper explores how Croatian IACS system integrators perceive cybersecurity requirements, implementation challenges, and current practices in the industrial environments in which they operate. The findings are based on three focus groups with system integrators working with clients in the energy, chemical and pharmaceutical, and food and beverage sectors. Participants reported limited cybersecurity awareness among IACS owners, as well as perceived differences in practices between smaller and larger companies and between domestic and foreign firms. Participants were divided in their expectations regarding regulatory pressure: while some believed it could improve cybersecurity, others expected only minimal compliance focused on cost reduction. The results of this study provide a basis for developing more practical cybersecurity practices and offer a deeper understanding of the key challenges faced by IACS system integrators and owners when dealing with cybersecurity.

0.27

Power Adjusting Games Under Block-Dependent Pool Payout Schemes

cs.CR cs.DCcs.ITmath.PR ·2026-10-29 ·Mustafa Doger, Sennur Ulukus
unusual
Abstract

Advanced power adjusting block withholding attacks such as PAW and T-PAW have been analyzed so far under a global-reset version of the proportional scheme (GRP), which is rarely used in practice. In this paper, we analyze T-PAW, which contains FAW and PAW as special cases, under the block-dependent payout schemes deployed in practice: Pay-Per-Last-N-Shares (PPLNS), round-based Proportional (PROP) and exponential-decay (EXP). We derive the revenue ratios analytically, by renewal arguments for PPLNS and through the stationary carry process of an open round for PROP and EXP. We further introduce Power Adjusting Hopping (PAH), where an adversary of any size adjusts its power allocation towards the pool according to the blockwork age of the round without withholding any block, and derive its revenue ratios under PROP and EXP. We show that GRP misestimates the T-PAW gain in a scheme-dependent direction: it overstates the gain under PROP up to an order of magnitude and understates it under EXP for the original slush decay rate. PROP is highly vulnerable to PAH, while EXP faces a trade-off: a slow decay invites PAH and a fast decay invites T-PAW. PPLNS is the least vulnerable scheme: it is immune to PAH and it mitigates, but does not eliminate, T-PAW. All analytical results are verified by Monte Carlo simulations.

0.26

Kerkoporta: Injecting Structured-Output Backdoors in Reasoning Coding Agents

cs.CR cs.AI ·2026-10-29 ·Georgios Syros, Blake Bullwinkel, Alina Oprea, Cristina Nita-Rotaru, Amanda Minnich +2 more
code/data
Abstract

Existing large language model (LLM) backdoor attacks operate primarily in chat settings, where the risk is bounded by free-form text outputs. AI agents remove that bound by performing actions and executing commands on the user's machine. We introduce structured-output backdoor attacks against reasoning models in agentic coding settings. A trigger placed anywhere in the agent's context causes the model to emit an attacker-chosen tool call, which the harness executes, directly compromising the system's confidentiality, integrity, or availability. We present Kerkoporta, a methodology that constructs high-quality poison samples, injects the backdoor via supervised fine-tuning, and recovers utility loss with lightweight reinforcement learning. We evaluate on modern, heavily post-trained open-weight models across proprietary and open-source harnesses, focusing on coding for its wide adoption and long-horizon task structure. The attack reaches success rates of up to 96.2% and can be configured to generalize across harnesses or bind to a single target. The resulting models can outperform the clean baseline on standard utility benchmarks, giving victims a direct incentive to adopt them. Overall, we demonstrate that LLM backdoors can have critical security impacts in the context of modern coding agents, and we characterize the poisoning methodology required to insert backdoors that are both effective and stealthy in this setting.

0.26

DualGuard: Comprehensive and Efficient Memory Safety across Sequential and Speculative Execution

cs.CR cs.PL ·2026-10-29 ·Sixuan Dang, Quan Zhou, Kaiming Huang, Jianan Zhu, Anitha Gollamudi +2 more
published
Abstract

Although decades of research have yielded sophisticated mechanisms for enforcing memory safety, speculative execution can subvert these protections through transient memory operations. Conversely, defenses designed specifically for speculative safety, such as Speculative Load Hardening (SLH), cannot enforce memory safety during non-speculative execution. A naive combination of defenses from both domains raises two fundamental challenges: (1) which combination of otherwise fragmented defenses can ensure dual memory safety under speculative and non-speculative executions? (2) whether such combinations introduce redundant protections that incur unnecessary performance overhead? We present DualGuard, a unified framework for dual memory protection under non-speculative and speculative executions. DualGuard systematically analyzes memory behaviors, classifying them into four categories, and further identifies the minimal composition requirements for dual memory safety, which allow us to strategically combine existing mitigations and design new ones to avoid redundant or excessive protections. We implement DualGuard through a combination of static analysis and runtime enforcement within the LLVM infrastructure. Our evaluation demonstrates that DualGuard reduces overhead by 3%-72% compared to naive compositions of fragmented defenses. In some cases, it even incurs lower overhead than approaches that guarantee memory safety solely under either non-speculative or speculative execution.

0.25

SEAL: Sealed Queries for Local Sanitization against Indirect Prompt Injection

cs.CR ·2026-10-29 ·Doohyun Kim, Donghwa Kang, Mingyu Jun, Brent ByungHoon Kang
Abstract · 19 pages

Indirect prompt injection threatens local LLM agents through untrusted tool outputs, motivating defenses that suppress attacks while preserving task evidence under limited inference budgets. We introduce SEAL, which learns a soft prompt and scoring readout offline to sanitize tool outputs using a frozen task model. SEAL constructs scoring queries from trusted initial context and the learned prompt, keeping the reference independent of inspected tool text. It scores tokens without normalization across context tokens and uses preservation regularization to encourage retention of useful evidence during localized deletion. This design requires neither defense-generated tokens nor a separate verifier model. Across AgentDojo and AgentDyn, SEAL suppresses attacks while improving task completion under attack. Specifically, on AgentDojo, SEAL reduces attack success from 34.6\% to 0.1\% and increases Joint---task completion without attack success---from 51.4\% to 77.5\%, while using approximately $127\times$ fewer defense-related tokens than PISanitizer, a state-of-the-art prompt injection defense. Replication on a second backbone with the same training recipe supports applicability across model families.

0.25

Integrating Perception, Comprehension, and Projection for Alert Validation: A Unified Cyber Situational Awareness Framework

cs.CR ·2026-10-29 ·Navodika Karunasingha, Mohan Baruwal Chhetri, Surya Nepal, Cecile Paris, Salil S. Kanhere
Abstract · 17 pages

Security Operation Centres (SOCs) must validate large volumes of alerts with limited analyst resources, yet alerts often lack sufficient context to determine whether suspicious activity is attack related. While provenance-based intrusion detection research has progressed from perception-level signals to comprehension- and projection-level cyber situational awareness (CSA), it remains unclear whether richer CSA consistently improves alert validation. This paper presents the first systematic evaluation of perception-, comprehension-,and projection-level CSA within a unified alert validation framework. Using Captain and Kairos, we generate multi-level CSA and formulate alert validation as a positive-unlabelled multi-instance learning (PU-MIL) task over CSA-derived behavioural hypotheses. Experimental evaluation on the DARPA TC dataset shows that comprehension-level CSA substantially improves validation by incorporating causal attack context. Relative to the Queue and Max. CSG baselines, PU-MIL achieves up to 28.6 and 75 percentage-point gains in Recall@5, and 24.6% and 86.8% reductions in mean positive rank on Cadets and Theia, respectively. Projection-level CSA can yield additional benefits, reducing the mean positive rank on Theia from 22.0 to 8.5 at a 40-minute projection horizon, although its effectiveness varies across projection horizons. Overall, richer CSA does not necessarily lead to better alert validation. Rather, its value depends on when and how it is introduced, motivating adaptive CSA progression that selectively enriches alerts according to validation requirements.

0.25

TextSeal2: Distortion-Free LLM Watermarking with Copula Mixing and Memoryless Races

cs.CR cs.AIcs.CL ·2026-10-29 ·Tom Sander, Pierre Fernandez, Itai Gat, Haiming Bao, Jiaxin Shi +9 more
code/data
Abstract · Code: https://github.com/facebookresearch/textseal

We present TextSeal2, a distortion-free watermark for large language models that keeps its signal under speculative decoding: tokens that replace rejected drafts remain watermarked with the same keys, and the acceptance rate is unchanged. It also preserves output diversity and adds no measurable inference latency. Its predecessor, TextSeal, restores diversity in the Gumbel-max scheme by routing each token to one of two keys via a coin flip. This is suboptimal because the detector averages the scores from both keys without knowing which key was used. TextSeal2 instead merges two or more keys into a single virtual key and couples it to fresh noise through AutoMark's Gaussian copula. Any number of keys, including publicly shared keys, can be merged this way without diluting detection. Under speculative decoding, TextSeal2 verifies keyed drafts with one more exponential race, run on the leftovers of the draft races. By memorylessness, this race reproduces block verification, so the output follows the target model exactly. We evaluate TextSeal2 on three LLMs. At equal output token diversity, TextSeal2 detects 8 to 22 percentage points more 100-token answers than TextSeal at a false-positive rate of $10^{-3}$, and is comparable to textGrain while requiring no post-logits optimization. Under speculative decoding, it detects 18 and 22 percentage points more 200-token answers than TextSeal and textGrain, respectively, at the same acceptance rate.

0.24

Characterizing LLM Agent Behaviors in Module Shadowing Environments

cs.CR cs.AI ·2026-10-29 ·Shuai Yang, Enmeng Lu, Jinyu Fan, Gongce Cao, Yu Chen +5 more
Abstract

LLM agents operating in untrusted workspaces face the risk of untrusted code execution during routine tasks, yet final reports and workspace states alone cannot reliably establish whether execution occurred. We introduce an evaluation framework that codes observation, reasoning, execution, and response behaviors from agent trajectories and aligns them with independent environmental records to reconstruct execution histories and verify final reports. We apply this framework to module shadowing, in which a local file is loaded in place of a trusted module during import, enabling untrusted code execution without adversarial prompts. Across 90 trials with 15 models, the payload executed in 51% of direct shadowing trials and 87% of indirect shadowing trials. Even explicit recognition of the shadowing mechanism did not ensure prevention, as the payload ran in 10 of the 29 trials with recorded recognition. These failures were often omitted from final reports, as 49 of the 62 triggered trials ended without mention of the security issue. Three reports even denied execution, including one from a trial in which independent logs recorded 19 payload executions. Final workspace states also obscured prior execution when agents deleted or overwrote the payload marker in 10 triggered trials. These findings demonstrate that combining behavioral coding with independent environmental auditing enables rigorous and transparent evaluation of agent safety.

0.23

REALM: Identifying Minimal Task Assets for LLM Agents

cs.CR cs.AIeess.SY ·2026-10-29 ·Zihou Ren, Byungchul Tak, Long Wang
Abstract

Large language model (LLM) agents often receive access to more files, accounts, and services than a task requires. Prompt injection turns this excess access into an opportunity to redirect execution toward unrelated resources. Applying least privilege therefore requires identifying the assets needed by a task, including assets whose identities emerge only during execution. We present REALM (Required Execution Assets Linked to a Mission), an approach to identifying and maintaining a task's required assets and their permitted uses. The approach represents required assets as Dependency Graph nodes, resolves prerequisite relationships, and admits newly discovered assets through source-scoped selection. It separates discovering an asset from authorizing its use, while preserving the agent's freedom to choose its computation. The REALM Builder maintains the selected grants and projects Asset Files; the Partial API Manager exposes resource-bound Partial APIs to the Agent inside a Task Container. Across 60 clean tasks and 560 injected cases in AgentDyn, it completes 40.0% and 36.4% of tasks and decreases ASR from 16.6% to 0.2% with qwen3.5-flash as the base model.

0.23

Measuring Duplication, Leakage, and Rendering Dependence in Malware-as-Image Benchmarks

cs.CR ·2026-10-29 ·Stefan Larson, Kevin Leach
Abstract

Classifying malware binaries rendered as images is a popular approach to malware family identification, and papers on the standard benchmarks for this task report accuracies of 97-99%. In this paper we audit ten of these benchmarks for near-duplicate images and test-train leakage, and retrain standard classifiers under naive and content-aware splits. Deduplicating by file hash, which is the standard practice, does not prevent leakage, because distinct binaries render to identical images. Deduplicating on image content reduces top-1 accuracy by up to 8 points and macro-F1 by up to 15 points on the released benchmarks we find leaky. The rendering step also adds duplication, since byteplots of the same binaries contain 19.7% near-duplicates where static feature vectors contain 2.2%, and changing the width at which bytes are wrapped into rows reduces accuracy by 20 to 33 points, although no byte changes. We recommend that content-level duplication and a non-image baseline be reported alongside accuracy.

0.22

"Shall We Approve?" Consequence-Aware Approval for Autonomous Coding Agents

cs.CR cs.AI ·2026-10-29 ·Weiheng Wu, Junkai Chen, Zhensu Sun, Jieke Shi, Yunbo Lyu +1 more
benchmark
Abstract · 24 pages, 9 figures

Coding agents increasingly rely on automated approvers to decide whether a proposed shell command may run. Existing approvers, from rule-based checkers to guardrail models and deployed permission classifiers, base their decisions on the text of the request and command. Yet a command's behavior depends on the code it invokes and the state of the workspace, and systems that observe file changes leave the assessment to the agent or a human after execution. We present Caper, a consequence-aware approval mechanism that pre-executes each command in a snapshotted sandbox, provides an independent approver with a bounded report of its observed effects, commits allowed commands, and rolls back blocked ones with a reason the agent can act on. Compared with six existing systems, Caper achieves the lowest harmful-outcome rate on SABER (6.7% versus at least 14.8%) and YoloFS-Ext, a benchmark we build in which harmful behavior is hidden inside invoked tools (7.8% versus at least 35.3%), while blocking 1.0% of the safe twins of the dangerous tasks in YoloFS-Ext.

0.20

Does Showing Agents the Privilege of Their Tools Reduce Over-Privileged Tool Selection?

cs.CR ·2026-10-29 ·Simon Parkinson, Mauro Vallati, Zhipeng Wang, Mustafa Mustafa
Abstract

When several tools can complete a task, an agent that picks the most powerful one holds more authority than the task needs. Recent work demonstrates that language-model agents do this often, and more so after a tool fails. Approaches to restrict focus so far on either enforcing least privilege outside the model or retraining the model. We ask a simpler question: does it help to tell the agent what each tool's privilege is? To investigate this question, we run all 544 ToolPrivBench scenarios on eight models, three run locally, and five through a hosted API, under five conditions: (1) no signal, (2) a least-privilege instruction, (3) privilege class and cost shown in each tool description, (4) both the instruction and privilege information, and (5) a matched neutral-metadata control. Showing privilege lowers premature escalation on every model that over-privileges at a measurable rate, reducing over-privileged selection by up to 19.3 percentage points (50\%), whereas irrelevant metadata in the same format does not. On most models, however, information alone mainly delays escalation until the standard tools have been tried. Combined with the instruction it more often avoids escalation, and the share of scenarios completed without a risk tool rises on seven of the eight models, reaching 51.1\% safe completion from a baseline of 18.8\% on the strongest-performing configuration. Visible privilege is therefore a cheap complement to enforcement rather than a substitute.

0.18

An Incentive-Aligned Trust-Centric Coordination Layer for Heterogeneous Blockchain Interoperability

cs.CR cs.DC ·2026-10-29 ·Saeed Moradi, Sara Rouhani
Abstract

The interoperability between permissioned and public blockchains is hampered by asymmetric privacy, trust, and governance assumptions. We propose a coordination architecture that decouples cross-domain trust management and privacy enforcement from underlying blockchain protocols, requiring no protocol-level modifications. The framework integrates cryptographic unlinkability mechanisms with incentive-compatible relay selection, combining stake-weighted participation and temporally accumulated reputation to align economic commitment with sustained correctness. We formalize the adversarial model and derive security and privacy properties for cross-domain coordination, demonstrating robustness against Sybil infiltration, collusion, and strategic stake concentration. Simulation-based analysis characterizes key design trade-offs between quorum thresholds, reputation memory, and anonymity parameters, indicating that privacy-preserving coordination can be achieved without degrading responsiveness or incentive stability when parameters are properly tuned.

0.17

Deep Learning Based Behavioral Drift Modeling for Continuous Biometric Authentication

cs.CR cs.HCcs.LG ·2026-10-29 ·Suprava Saha Dibya, Minar, Abdulla Al Noman, Md. Zafir Abid Shresto, Ahmed Sayef
Abstract

This research proposes a two-level keystroke dynamics authentication system that provides defense-in-depth security through continuous behavioral drift monitoring. Level 1 employs fixed-text password authentication using Temporal Convolutional Networks (TCN) with contrastive learning and population-based training (PBT), achieving 3.77% EER for high-security login verification. Upon successful authentication, Level 2 continuously monitors user typing during natural free-text sessions using a 4-Model Deep Learning Ensemble. If suspicious behavioral drift is detected during monitoring, the system re-authenticates the user by returning to Level 1, providing a feedback loop against session hijacking and behavioral anomalies. The system explicitly models long-term behavioral drift through the fixed-text component's PBT mechanism, which adapts to temporal variations in password typing patterns. This research leverages free-text (Buffalo dataset) and fixed-text (GREYC dataset) keystroke dynamics datasets, extracting temporal and spatial features including hold times, up-down latencies, down-down intervals, and trigram patterns. The final system achieves 96%+ accuracy for fixed-text authentication (3.77% EER) and 83.53% True Acceptance Rate (TAR) at 5% False Acceptance Rate (FAR) for free-text continuous authentication (10.05% EER), demonstrating superior performance and strong resistance to impostor attacks through continuous behavioral monitoring.

0.17

Hardware-Aware Model Inversion in Memristive Spiking Neural Networks

cs.CR cs.ETcs.NE ·2026-10-29 ·Pablo Sorrentino, Fernando M. Quintana, Erika Covi, Fatih Turkmen
Abstract · 15 pages, 8 figures

Hardware-accelerated spiking neural networks (SNNs) implemented on memristive crossbar arrays offer significant gains in energy efficiency, while device variability and analog noise are sometimes assumed to hinder information extraction. However, whether these hardware effects improve, preserve, or degrade privacy after deployment remains largely unexplored. Therefore, we ask whether model-inversion leakage characterized on an SNN remains representative after the same model is mapped to a non-ideal memristive substrate, and how hardware non-idealities and reliability faults affecting synapses, neuron thresholds, and spike processing reshape that leakage. To study this problem, we introduce Spiking Domain Model Inversion (SDMI), an inversion method that combines Expectation over Transformation (EOT) with surrogate gradients to optimize inputs through stochastic analog behavior and non-differentiable spiking dynamics. We use SDMI consistently across the SNN model, its memristive SNN, and the same memristive SNN under additional hardware perturbations. Across four datasets, reconstruction leakage does not vary monotonically with model accuracy or hardware variability. At comparable accuracy levels, neuron-threshold perturbations increase reconstruction fidelity, measured by SSIM, by up to 51.8%, whereas spike-level perturbations primarily degrade utility without consistently improving reconstruction. Weight noise, in contrast, reduces leakage under the same inversion procedure. These results show that hardware variability is not inherently privacy-preserving: model inversion leakage in memristive SNNs depends on the hardware state and cannot be inferred from model accuracy alone.

0.14

Accept Changes, Reject Attacks: Robust Memory Management for LLM Agents

cs.CR cs.AI ·2026-10-29 ·Runze Zhao, Dongruo Zhou, Sumit Kumar Jha, Nathaniel D. Bastian, Ankit Shah
Abstract · 24 pages, 9 figures, 6 tables

Since large language model (LLM) agents rely on long-term memory, what they store now shapes what they do later. A memory manager decides how each input changes this memory and which records the agent acts on. It can fail in two ways: keeping a fact that has since changed (outdated memory), or storing a planted passage or injected instruction (memory poisoning). Most prior work studies them separately, yet they call for opposite write decisions, so no fixed write rule handles both. We therefore formalize memory robustness as handling both without doing worse on other inputs. Since accuracy checks only the answer, we measure robustness with evidence-conditioned accuracy (ECA), which also checks that the records the manager selects include all the evidence the answer needs, updated and free of attacks. However, existing memory systems and defenses are not more robust than an untuned manager on our test sets. We improve robustness with a simple one-shot recipe that fine-tunes the manager on episodes synthesized around a single conversation with inserted changes and attacks. On three backbones, it improves every category score on our test sets and more than doubles the untuned manager's overall ECA, though robustness remains incomplete.

0.14

CLIC: Closed-Loop Intent Consistency for Securing LLM-Driven Control in Open RAN

cs.CR cs.AIcs.NI ·2026-10-29 ·Sunder Ali Khowaja, Kapal Dev, Joongheon Kim, Angelos Antonopoulos, Nikolaos Bartzoudis
Abstract · 6 pages, 1 figure, 3 tables

With the progressive capabilities of large language models (LLMs), researchers are finding ways to integrate them into the control loops of Open Radio Access Network (O-RAN) to enable slicing and resource management through user intents. However, integration of LLMs in the O-RAN control loop via the RAN Intelligent Controller (RIC) makes the system vulnerable to \textit{runtime prompt injection threats}, which manipulate control decisions without violating interface authentication, and \textit{residual virtual prompt injection (VPI)} attacks that plant trigger mechanisms to silently steer the behaviour during fine-tuning. Existing studies address only the runtime threat and do not enforce consistency across the full control path, i.e. from the certified A1 intent, through the multi-step LLM trajectory, to the final E2 action. We propose a closed-loop intent-consistency layer (CLIC) that jointly mitigates both threats. The proposed CLIC defines a formal consistency score using a small quantized language model, which is deployed as a Non-RT rApp for model-integrity certification along with a Near-RT xApp for runtime enforcement. We implement CLIC on the OpenAI Cellular (OAIC) platform and evaluate it under domain-adapted JailbreakBench, AdvBench, and VPI workloads. The experimental results show that the proposed CLIC reduces the attack success rate from 0.92 to 0.29 and residual VPI attacks from 0.37 to 0.04, while improving isolation and fairness at slightly higher latency costs, i.e. approximately 27 ms.

0.14

Understanding and Mitigating Role Spoofing in LLMs

cs.CR cs.AI ·2026-10-29 ·Ezra Edelman, Benjamin Edelman, Surbhi Goel, Eran Malach
Abstract

Despite growing capabilities, LLMs remain susceptible to prompt injections: malicious inputs that can manipulate or hijack models. In particular, it has recently been observed that LLMs can exhibit role confusion (Ye et al., 2026), where models attribute the source of text based on its style instead of its labeled role. Malicious text in tool outputs (for instance, a malicious webpage accessed by an agent) can leverage role confusion to impersonate the user, an attack we call role spoofing}. In this work, we seek to understand and mitigate role spoofing. We study how models track "who said what" in a conversation, showing that they use cues from both content (e.g., semantics and writing style) and structure (e.g., control tokens and formatting). We show that LLMs are highly susceptible to role spoofing attacks that combine both fake structure and content. Additionally, in a toy setting, we show that adversarial training which teaches the model to ignore content-based injections does mitigate content-based attacks but leaves the model susceptible to structure-based attacks. This can explain why frontier models, with increased resistance to many injection attacks, can still be susceptible to role spoofing. Finally, we propose a new method for detecting role spoofing using linear probes trained to predict roles in a conversation, and show that it is effective across different settings.

0.12

Safin-1.5: Safety-Native Architecture and Verifiable Reasoning for Recursive Safety Improvement

cs.CR cs.AIcs.CL ·2026-10-29 ·Jingnan Zheng, Cheng Jin, Ming Zhang, Huan Sun, Zhongtian Ma +26 more
Abstract · 29 pages, 5 figures

Recursive self-improvement (RSI) enables large language models to advance their capabilities through successive updates, but these updates can also introduce new risks and weaken existing safeguards. In line with the AI-45° Law, we pursue Safety RSI, in which safety develops as an intrinsic capability alongside general model capabilities through continual risk discovery and repair, while preserving utility and limiting over-refusal. We introduce Safin-1.5, a model built on a safety architecture and trained through recursive safety improvement, with an inference-time harness that extends its capabilities to formal reasoning. Our approach comprises three components: i) The safety architecture combines persistent safety states with hazardous-knowledge disentanglement. The former retains and selectively invokes safety knowledge, while the latter relocates targeted hazardous knowledge into removable experts, enabling its removal with minimal impact on general capabilities. ii) Recursive safety improvement establishes a discovery-evaluation-repair loop. The loop identifies new vulnerabilities and uses experimental feedback to develop repair strategies. It accepts updates only when they improve safety while preserving general capabilities and limiting over-refusal. iii) The formal reasoning harness combines the model's language understanding with symbolic inference and verification. Proof-obligation-guided semantic bridging recovers missing relations in formal representations, enabling symbolic solvers to construct checkable proofs under formalized premises and accepted semantic bridges. Experimental results demonstrate that Safin-1.5 achieves safety and reasoning performance, outperforming frontier models.

0.11

MetaVul: Agentic Harness Discovery for Vulnerability Detection

cs.CR cs.AIcs.SE ·2026-10-29 ·Ira Ceka, Arnav Garg, Ryan Lutz, Ting Zhang, Ratnadira Widyasari +2 more
Abstract · Originally announced September 2026

LLM agents are increasingly used to automate vulnerability detection, but reliable detection is challenging because agents must distinguish subtle, security-relevant fixes from benign code variation. While underlying LLM capability has improved significantly, the accuracy of the agent still depends heavily on the harness: the code that defines the scaffolding around the LLM. Harnesses are typically fixed, and largely hand-written, limiting adaptation to evaluator feedback. Additionally, when harnesses are static, improvements on one dataset or model may not transfer to others. To address these challenges, we introduce MetaVul, a self-improving framework for harness discovery in vulnerability detection. Starting from a seed structured reasoning scaffold, MetaVul uses a workflow-based optimization loop to iteratively refine the harness using feedback from paired vulnerable and patched functions, while preserving the structured reasoning and verdict format. At each iteration, it searches over the history of prior harnesses, with access to their trajectories, logs, and metrics, and it returns a set of Pareto-optimal harnesses. We apply MetaVul to discover agent harnesses for various architectures. A validation set (202 pairs) is used for optimization, while an unseen dataset (430 pairs) is kept as the held-out set. We test whether harnesses learned from one model transfer to others not seen during optimization. Results on both frontier and small LLMs show that the optimized harnesses improve pairwise vulnerability detection up to 23.3 percentage points on the validation set and up to 25.8 points on the held-out set with a Pareto-frontier harness. Harnesses learned with one model also transfer to others. We show that harnesses optimized with one model transfer to another, improving held-out pairwise accuracy by 14.7 points without further optimization.

0.08

Judging Before Propagation: Task-Graph Contextual Integrity for Multi-Agent Privacy

cs.CR cs.AI ·2026-10-29 ·Hengyu Wu, Blaise Delattre, Yang Cao
benchmark
Abstract

Multi-agent LLM systems increasingly rely on persistent memory and inter-agent communication. These capabilities create privacy risks that extend beyond individual interactions: a release appropriate in isolation may become inappropriate when combined with information previously disclosed to a recipient or propagated to downstream agents along the planned workflow. Existing Contextual Integrity (CI) mechanisms for agentic systems typically rely on local judgments that assess each disclosure in isolation, with or without regard to the recipient's history, but do not account for how information propagates downstream. However, compliance at the current boundary does not guarantee appropriateness across the workflow. We introduce Task-Graph Contextual Integrity (TG-CI), to our knowledge the first task-graph-native formalization of Contextual Integrity for multi-agent systems. TG-CI represents a workflow as a Multi-Agent Task Graph (MATG) and extends CI from atomic transfers to stateful, graph-level information flows by jointly considering an item's intended propagation scope and recipients' evolving disclosure states. We operationalize TG-CI through propagation-scoped authorization under two complementary policies: Full-Scope, which requires admissibility across the complete downstream scope, and Auto-Region, which confines each item to its maximal admissible region using lightweight frontier gates. We further introduce MAPLE-Bench, a benchmark of 144 cases across four scenarios that jointly vary MATG topology, disclosure history, compromised-agent placement, and item-level privacy policies. In our primary experimental configuration, TG-CI reduces the mean unauthorized disclosure count by 85.8-94.1% relative to the Local-History CI baseline, while retaining 93.6-96.1% of the baseline's task accuracy, with the two enforcement policies providing complementary privacy--cost--utility trade-offs.

0.07

RiskyPy: Toward Model-Calibrated Prompt Generation for Joint Security and Functionality Evaluation of LLM-Generated Code

cs.CR cs.AIcs.SE ·2026-10-29 ·Zacharie Chenail-Larcher, Daniele Lunghi, Melissa Tessa, Quentin Stiévenart, Jacques Klein +3 more
Abstract

Evaluating approaches that improve the security of LLM code generation requires datasets that jointly assess security and functionality and remain adapted to the capabilities of the models being studied. However, existing benchmarks are costly to produce and are generally not calibrated to a chosen range of model capabilities. We introduce RiskyPy, an automated pipeline that generates security-oriented code-generation datasets calibrated to model proficiency using functionality and vulnerability canaries. We instantiate RiskyPy for Python using domains covering 18 CWEs from the 2025 CWE Top 25, producing 101 prompts. Experiments on Qwen3-4B and Seed-Coder-8B show that the generated tasks remain functionally accessible, with 57-72% of prompts yielding functional code, while only 5-10% of generations are free of Bandit findings, leaving room for joint evaluation of security and functionality of security improvement approaches.

0.06

PrivacyFlow: Privacy in AI Agents through Information Flow Analysis and Consequence-Aware Risk Management

cs.CR cs.AIcs.LG ·2026-10-29 ·Robin Staab, Tom Wollschläger, Vahab Mirrokni, Reza Shokri
benchmark
Abstract

Autonomous LLM agents executing multi-step workflows over personal data often misalign with user privacy expectations and leak sensitive, out-of-context information. Although Contextual Integrity (CI) formalizes this problem, existing benchmarks and defenses rarely operationalize CI for realistic multi-step task sequences in evolving environments. In this work, we first introduce agentic PrivacyBench, a benchmark of 5873 tasks in 986 task sequences with dynamic environment updates and CI-grounded allow-sets for sensitive information. We also introduce PrivacyFlow, a defense that separates deliberative privacy planning (System 2) from task execution (System 1), treating each sharing decision as a structured risk-management problem. Before execution, PrivacyFlow traces the provenance and purpose of each planned data flow, models recipient background knowledge, and simulates what external actors could infer. This allows us to model the downstream consequences of disclosure, including inference attacks and context-dependent harms. It then weighs the remaining disclosure risk against task utility and executes a minimized plan, blocks it, or consults the user. Across six models, PrivacyFlow significantly reduces step-level privacy failures and raises the share of completed task sequences without a detected privacy violation by 22--55 percentage points. We further highlight a new privacy--utility--autonomy tradeoff: allowing PrivacyFlow to proactively prompt the user in case of uncertainty can increase authorized task completion at the cost of shifting decision effort back to the user. Our results suggest that proactive CI reasoning can substantially improve agentic privacy and motivate further study of this tradeoff.

0.06

Provably Secure Diffusion Watermarking Against Removal Adversaries with Batched Detector Access

cs.CR ·2026-10-29 ·De Zhang Lee, Ee-Chien Chang
Abstract

Recent latent-based watermarking methods can embed undetectable watermarks into AI-generated content. However, these approaches are vulnerable to removal attacks as their watermark detection boundaries are exposed, prompting the use of secret long-term transformations to hide these boundaries. We show that such defenses are insecure under a stronger threat model motivated by publicly reachable detectors: an adversary with batched detector access can reconstruct the transformations, permanently re-exposing the decision boundary. To address this, we propose a watermarking scheme based on a hybrid of long-term and one-time secret transformations, which prevents the reuse of structure across samples and keeps the long-term secret hidden. Besides undetectability, we show that this scheme is provably secure against removal adversaries with batched detector access, in the sense that no probabilistic polynomial-time adversary can achieve a non-negligible advantage over an adversary restricted to adding isotropic white-noise with a $\sqrt{2}\times$ larger perturbation budget, and we show this $\sqrt{2}$ factor to be tight. Evaluations on Stable Diffusion 2.1 and XL demonstrate the attack against the existing defense, and show that our scheme preserves generation quality and withstands common perturbations and targeted removal attacks.

0.06

Attack-Resilient Multi-Bit Image Watermarking Framework via Latent Coordinate Sign Clamping

cs.CR ·2026-10-29 ·Marjan Beheshti, Ronghua Xu
Abstract

Watermarking is essential for image copyright protection and attribution, yet existing methods suffer from fundamental trade-offs among post-hoc applicability to pre-existing images, multi-bit capacity, training-free deployment, and adversarial resilience. To address these limitations, we propose a training-free watermarking framework that exploits the overcompleteness of diffusion latent spaces to embed multi-bit payloads into arbitrary images while preserving high visual quality. Additionally, we develop a three-layer decoding pipeline that combines magnitude-weighted majority voting, BCH, and Chase decoding to maximize message-extraction accuracy in the presence of channel noise. To counter active adversaries, our framework integrates a steganalysis-resistant noise camouflage layer and a dual-layer forgery defense that combines perceptual hashing with a Reed-Solomon fuzzy commitment scheme over CLIP features. Empirical evaluations demonstrate that our method applies to both natural and AI-generated images with a scalable $32\text{--}128$-bit capacity. Furthermore, it effectively mitigates advanced removal vectors ($94.6\%$ decode rate against CtrlRegen+ regeneration attacks) and secures robust forgery detection ($98.8\%$ accuracy via joint CLIP and RS fuzzy commitment mechanism).

0.04

Systems & Networking 42

APPlace: Co-Optimizing Device Locations and Access Points for Routing-Aware Analog Placement

cs.AR ·2026-10-29 ·Bingyang Liu, Xiaohan Gao, Souradip Poddar, David Z. Pan
top venuecode/data
Abstract · Accepted to ASP-DAC 2027. 8 pages, 6 figures, 4 tables. Code: https://github.com/bingyang1132/applace

Analog layout automation remains challenging in Analog/Mixed-Signal circuit design, because parasitics and mismatches introduced during layout strongly affect circuit performance. Placement is especially critical among the layout stages, as it constrains the routing solution space and the achievable post-layout performance. Existing analytical placement methods optimize device coordinates under center-based wirelength proxies and leave access point (AP) selection to routing. In analog circuits, however, a routing pin is an extended piece of terminal metal rather than a fixed point, and the best AP on it depends on where the other connected devices are placed, so a coordinate-only placement optimizes for connection points that the router will not use. We present APPlace, a routing-aware analog placement framework that co-optimizes device locations and AP selections during global placement, followed by a legalization stage that preserves the structure established by global placement. Experimental results on four real-world analog benchmarks show that APPlace achieves better overall post-layout performance than representative baselines at comparable layout area. Ablation studies further show that removing AP co-optimization causes up to a 13 dB CMRR loss among the OTA benchmarks and a 23% increase in comparator delay.

0.87

PDScope: Breaking the Prefill-Decode Seesaw for Memory-Bound MoE Inference

cs.DC ·2026-10-29 ·Enda Yu, Dezun Dong, Zhaoning Zhang, Xiaojian Yang, Weiling Yang +4 more
published
Abstract · accepted by PPoPP'27

Local deployment of MoE-based LLMs on CPU-GPU nodes is attractive for cost and privacy reasons, yet limited GPU memory forces most experts into CPU memory. Prefill is then compute-bound and benefits from aggressive GPU loading, while decode is I/O-bound and requires conservative PCIe transfers. These demands cannot be amortized under low batching---typical of local assistants and document tools---forcing phase-agnostic schedulers into a seesaw: favoring prefill's aggressive loading injects unmaskable I/O that stalls decode, while favoring decode's conservative transfers starves prefill's GPU of work. We observe that expert routing is predictable in phase-specific ways, allowing us to decouple the two phases and optimize prediction, scheduling, and execution independently for each phase. PDScope predicts expert activations to guide placement and prefetching, adapts these policies to each phase's bottleneck via cross-layer cost modeling, and compiles decisions into overhead-free pipelines using a static computation graph, cache prefetching, and zero-copy asynchronous I/O. On mainstream MoE models, PDScope achieves 2.5-10.3x higher prefill throughput and 2.1-6.1x higher decode throughput than existing systems, showing that phase-aware co-design unlocks efficient MoE inference under memory constraints and extends naturally to multi-batch settings.

0.55

True 3D Holography: A Communication Service of Tomorrow and Its Requirements for a New Converged Cloud and Network Architecture on the Path to 6G

cs.NI cs.DCeess.SP ·2026-10-29 ·Ingo Friese, Mandy Galkow-Schneider, Louay Bassbouss, Alexander Zoubarev, Andy Neparidze +15 more
publishedhighly cited
Abstract

Research project 6G NeXt is considering true 3D holography as a use case, setting requirements on the communication as well as the computing infrastructure. In a future holographic communication service, clients are widely spread in the network and cooperatively interact with each other. Especially for holographic communication high processing power is required as well. This makes a high-speed distributed backbone computing infrastructure, which realizes the concept of split computing, inevitable. Furthermore, tight integration between processing facilities and wireless networks is required in order to provide an immersive user experience. This paper illustrates true 3D holographic communication and its requirements. Afterward, an appropriate solution approach is elaborated. Here, novel technological approaches are discussed based on a proposed overall communication and computing architecture.

0.44

OPERA: OPamp Design by an Expert-Taught Reasoning Agent, from Specifications to Layout

cs.AR cs.AI ·2026-10-29 ·Vishal Saxena, Soumyajit Mandal
benchmarkcode/data
Abstract · 5 pages, 3 figures, submitted to ISCAS 2027

Analog flows from specifications to a signed-off layout rarely run without manual iteration. Recent large language model (LLM) agents size circuits with optimizer loops over the device dimensions and stop at the schematic. We present OPERA (OPamp design by an Expert-taught Reasoning Agent), in which a general-purpose LLM agent designs CMOS opamps from specifications to post-layout sign-off with open-source tools. First, the designer teaches the agent the gm/ID method: map the specifications to transconductances and minimum currents, choose gm/ID per device role from characterized lookup tables, and size from equations, leaving a few knobs to search and an explainable design. Second, the agent learns the gLayout framework and taught layout practice, and writes a reusable layout generator that closes the layout and extraction loop at designer-approved gates. In SKY130, the agent designed a shared constant-gm bias and six opamps (5T, folded-cascode and telescopic OTAs; Miller, folded-cascode and telescopic two-stage) for fun >= 50 MHz, PM >= 60 degrees, SR 25 V/μs into 1 pF and a minimum dc gain of 30 to 60 dB. All six layouts are DRC-clean and LVS-matched, and their extracted netlists meet every target at all 15 PVT corners ( fun >= 53.7 MHz, PM >= 62.1 degrees, SR >= 27.9 V/μs). The benchmark netlists are released at https://github.com/AMPICLab/AMPICBench.

0.43

Rebuilding a Field's Open-Problem Catalog from the Literature: A Two-Layer Knowledge Graph, and Why Status Must Stay Proposed

cs.DB cs.AIcs.DL ·2026-10-29 ·Madhulatha Mandarapu, Sandeep Kunkunuru
positioncode/data
Abstract · 9 pages, 1 figure, 7 tables. Code and data: https://github.com/samyama-ai/open-problems-as-data

A research field records its open problems in surveys, keynotes and community reports. Those are prose, written every few years, and nobody can ask them who works on a problem now or whether it was settled last year. We argue the problems should be data, and we build the system that makes them so: a hand-curated problem layer carried in unchanged, under an evidence layer rebuilt from a bibliographic index, joined by typed, provenance-carrying links in one graph database and queryable from one language. A language model does four jobs -- per-paper extraction, catalog grounding, per-paper status verdict, and idea generation -- each constrained to a candidate set so out-of-set answers are refused rather than stored. We report a study on two curated catalogs, database research (1,053 problems) and AI research (1,000 problems), each with a prior catalog-only graph. The rebuilt graph answers every question the catalog-only graph answered -- 47 of 47 on databases, 34 of 35 on AI -- plus questions the catalog-only graph refused: recent workers, untouched problems, methods tried, and outside papers bearing on a problem (336 problems, 770 works on databases). The capability we most wanted does not survive its own audit. Of 41 solves and refutes verdicts hand-checked against the papers, 9 are correct, 30 partly correct and 2 wrong; no refutes verdict (0 of 34) disproves an open problem, and only 1 of 7 solves closes the central question. We report that failure in full and draw the design conclusion: a status change must be proposed with its evidence and never applied, making the human reviewer part of the architecture. We close with the data-engineering problems this raises: entity resolution across curated and harvested layers, verifying model status claims, freshness under source quotas, and evaluation without ground truth.

0.40

Caerus: Prefetching by Thinking Globally and Acting Locally

cs.AR ·2026-10-29 ·Jacky Wong, Jacob Alexandrou, Yinting Huang, Luke Panayi, Lev Mukhanov +2 more
top venue
Abstract · Accepted to MICRO 2026

Modern processors rely heavily on data prefetching to bridge the growing performance gap between computation and memory accesses. In offset prefetching, prefetches are issued at an address delta relative to the most recently observed access, selecting deltas that have exhibited high coverage in the recent access-stream history. The Best-Offset Prefetcher (BOP) exemplifies this approach by selecting a single uniformly-issued global offset that delivers timely prefetches. However, its lack of context-sensitive accuracy tracking results in significantly increased DRAM traffic and sometimes disappointing performance. Conversely, Berti, the state-of-the-art offset prefetcher, achieves impressive prefetching performance and accuracy. However, its Program-Counter (PC)-local approach to extracting offsets limits its ability to capture global memory patterns, meaning there are many scenarios where simpler global offset prefetchers, such as BOP, outperform it. Observing the limitations of contemporary offset prefetchers, we present Caerus: an offset prefetcher that employs PC-local filtering on multiple non-overlapping globally derived offsets. Caerus extracts multiple offsets from the global memory trace and maximizes their combined coverage by training each offset only on accesses not already covered by previously learned offsets. The accuracy of each extracted offset is tracked per-PC based on whether it can cover future accesses for any PC. This allows Caerus to address complex delta patterns while maintaining high accuracy. In SPECspeed 2017, Caerus gains a geomean speedup of 8.3%, outperforming Berti by 2.0% while achieving low hardware and DRAM-traffic overhead.

0.40

LLM Agent-Based Design of RISC-V CPUs with FPGA-in-the-Loop Evaluation

cs.AR cs.AI ·2026-10-29 ·Leo Itoh, Junichiro Kadomoto
code/data
Abstract · 5 pages, 6 figures, 2 tables. Submitted to IEEE ISCAS 2027. Code and data: https://github.com/kadomoto-lab/llm-cpu-fpga-loop

Long-run RTL simulation of CPU candidates can take hours, while short diagnostic runs may miss performance degradation that emerges later in execution. This evaluation cost can delay design decisions in LLM-agent-based CPU development. We present a development environment in which LLM agents implement CPU candidates on an FPGA and use on-board performance measurements to guide design decisions. We applied it to the human-supervised development of a single-cycle RV32I core into a three-wide out-of-order RV32IM core. The final design reaches 5.61 CoreMark/MHz at 130 MHz on a Artix-7 FPGA. For the same CPU candidate running 8,000 CoreMark iterations, FPGA evaluation, including implementation, was 9.6x faster than single-threaded RTL simulation. In one case, board measurements revealed long-run performance degradation in a candidate that improved in a short diagnostic run. The result guided candidate selection and led to RTL-based root-cause analysis and a fix.

0.37

When Imputation Changes the Nearest Neighbor: Exact Certification of Similarity Retrieval for Incomplete Sensor Time Series

cs.DB cs.DS ·2026-10-29 ·Jianru Shen
publishedtheory
Abstract · Accepted at IEEE BigData

Sensor time series routinely reach similarity search with missing values, and the retrieved neighbours then depend on how the gaps were filled. We ask which retrieval conclusions hold for every admissible completion of an incomplete query. For a family of window-minimum distances that contains the $L_1$ distance as its separable end we give a pairwise certificate that one candidate is strictly closer than another, a guaranteed candidate set that contains every possible nearest neighbour, and a certificate that a nearest-neighbour label cannot change. The guarantees hold uniformly over the specified box and apply to the true query whenever it lies inside. At the separable end a closed form is known, but beyond it the worst-case gap need not be attained at a corner of the box. We prove a finite reduction to critical values, give an exact interval dynamic programme for the worst-case ranking gap, and a quadratic-time fast layer of valid bounds and feasible witnesses, supported by an exact criterion for when the feature-box relaxation is tight. On three traffic-sensor collections with real gap patterns, linear interpolation changes the nearest neighbour in 15 to 33 percent of the queries. At the full band the feature-box relaxation misses up to 14 percent of the exactly stable boundary pairs, whereas fixed-completion and Monte Carlo checks issue false certificates. The fast layer resolves about 90 percent of all tested boundary pairs without dynamic programming, and on paired network-scale queries the exact method is 82 to 110 times faster than the programme alone. Label certificates transfer to three public benchmarks, and the band width trades forecasting accuracy and certification coverage for lower imputation sensitivity, with dataset-dependent crossovers.

0.37

Energy-Efficient Memristor Accelerator for Depression Detection with a CNN-LSTM Hybrid Model

cs.AR ·2026-10-29 ·Jiaqi Liu, Yun Yu, Fenglin Xi, Shenjian Zhang, Shenghao Liao +5 more
hot direction
Abstract

Early screening of depression requires efficient analysis of audio and speech data, while conventional CPU and GPU platforms incur high latency and energy consumption for CNN-LSTM inference. To address this challenge, we propose a unified resistive RAM (ReRAM) processing-in-memory accelerator for a CNN-LSTM attention workload. The architecture integrates convolutional, recurrent, attention, and multi-layer perceptron (MLP) stages within a crossbar-based framework, reducing intermediate data movement between spatial and temporal branches. CNN operations are mapped through im2col-based vectorization, while LSTM gate computations are reformulated using concatenated input-state vectors and mapped onto ReRAM crossbars. We further provide module-level workload breakdown, parameter sensitivity analysis, and matched-workload context against representative ReRAM/PIM accelerators. Experimental results show 32.1 times and 9.3 times latency reduction over CPU and GPU baselines, respectively, with about 160 times and 140 times energy reduction. Simplified weight-perturbation tests characterize ASLF sensitivity in an untrained pipeline with synthetic spectrogram-shaped inputs. Task-level prediction accuracy and robustness remain to be evaluated.

0.34

CoraLive: Ultra-Low-Power Underwater Vision for Long-term Coral Reef Monitoring

cs.NI ·2026-10-29 ·Sitong Shen, Chenshu Wu
Abstract · 14 pages, 8 figures

Effective coral reef protection and restoration demands continuous, long-term underwater monitoring. However, current solutions largely rely on fragmented, manual labor and costly infrastructure, or are severely limited by constrained battery lifespans. In this paper, we present CoraLive, an ultra-low-power underwater camera system for long-term, continuous coral reef monitoring. CoraLive presents a decoupled underwater-shore co-design that minimizes on-board computation and storage, offloading heavy reconstruction and segmentation tasks to offline shore-side processing. To achieve this, CoraLive introduces a dual-rate sensing scheme combining frequent low-resolution frames with sparse high-fidelity samples and a content-adaptive update mechanism that extracts and stores only compact, task-specific data, eliminating the need to save raw video frames. We implement CoraLive on a custom STM32U575-based dual-camera prototype with device-level computation and storage optimization. Compared with raw recording, CoraLive reduces daily storage by $4.51\times$ and average power by $7.78\times$, extending battery lifetime from 20.4 to 159.5~days while retaining high-accuracy segmentation and reconstruction. Further validated in a real-world underwater deployment, we envision CoraLive as a pioneering step for ubiquitous long-term coral reef underwater monitoring.

0.32

SPECTRUM: A Strategic Framework and Technical Blueprint for European Exascale Research Data and Compute Infrastructure

cs.DC astro-ph.IM ·2026-10-29 ·Sergio Andreozzi, Stefano Bagnasco, Florian Berberich, Tommaso Boccali, Concezio Bozzi +19 more
unusual
Abstract · 8 pages, 3 figures, 2 tables. Preprint of a paper submitted to the proceedings of the 28th Conference on Computing in High Energy and Nuclea…

The HiLumi LHC, the SKA Observatory and LOFAR 2.0 will take High Energy Physics (HEP) and Radio Astronomy (RA) towards exabyte-scale data within the next decade, beyond the capacity of today's infrastructure. The SPECTRUM project responded with two documents built on the same evidence. A Technical Blueprint describes a European compute and data continuum as a capability map, consolidates the gaps found in HEP, RA and related use cases, and recommends technical actions in eight areas. A Strategic Research, Innovation and Deployment Agenda (SRIDA) organises thirteen priorities in four pillars, maps them to investment areas and European funding instruments, and phases them over a roadmap set against the timelines of the three instruments. This paper, by editors and contributors of both documents, summarises them, their consultative method and the impact the SRIDA expects.

0.31

HippocamFLEX: FPGA Prototyping of Hippocampus-Inspired Architecture for Context-Aware Sequence Prediction

cs.AR ·2026-10-29 ·Bhuvanesh S. Kayarabettu, Hamed Poursiami, Keiland Cooper, Babak Shahbaba, Norbert Fortin +2 more
Abstract

Unlike conventional architectures that treat memory as static pattern storage, the hippocampus organizes experience as ordered sequences and retrieves them through temporally compressed replay spanning prior, current, and anticipated states. This capacity of context-appropriate sequence prediction remains absent from conventional deep learning accelerators and existing neuromorphic fabrics. We present HippocamFLEX, a fully synthesizable digital implementation of these hippocampal computational principles, realized with SystemVerilog RTL on the AMD Zynq UltraScale+,XCZU9EG. The design maps the hippocampal trisynaptic pathway onto three pipeline stages: a Dentate Gyrus / Entorhinal Cortex (DG/EC) stage for hypervector proximity matching, a CA3 stage for context-gated recurrent sequence prediction with multi-phase temporal tagging, and a CA1 stage that integrates past, present, and predicted future events into a theta-sequence inspired output. While much brain-inspired hardware relies on analog mixed-signal design for energy efficiency, this work targets FPGA prototyping as a deliberate architectural choice: FPGAs offer rapid design-space exploration without tape-out cost, bit-exact determinism that simplifies verification. The design supports 64 events across 16 contexts and is verified on physical silicon across context-isolation, temporal-ordering, and mid-sequence sliding-window test cases and is inherently scalable to larger event and context sets.

0.31

MRUD: Maximally Redundant Unsigned Digit Number Systems for High-Performance Arithmetic Units

cs.AR eess.IV ·2026-10-29 ·Samaneh Rahimi, Ali Azarpeyvand
Abstract · 22 pages, 15 figures. Submitted to Integration, the VLSI Journal (Elsevier). Under review

Signed digit redundant number systems such as Maximally Redundant Signed Digit (MRSD) enable carry-free addition but introduce significant overhead in multiplication due to the presence of both positive and negative bits. While Unsigned Digit Redundant (UDR) systems have shown promise in radix-2, their extension to higher radices remains unexplored. This work presents the first generalized Maximally Redundant Unsigned Digit (MRUD) number systems for high-radix arithmetic, extending UDR representations from radix-2 to radix-$2^h$ while preserving carry-free addition. The proposed framework includes digit encoding, digit extension, arithmetic operations (addition, subtraction, and multiplication), sign inversion, forward/reverse conversion algorithms, overflow/underflow detection, and hardware architectures for both signed and unsigned MRUD representations. Unlike MRSD, which is inherently limited to signed arithmetic, MRUD supports both signed and unsigned representations, similar to conventional binary systems. Evaluations across three multiplier modes---RRR (Redundant inputs, Redundant output), BBR (Binary inputs, Redundant output), and BBB (Binary inputs, Binary output)---as well as Multiply-Accumulate (MAC) units, using the FreePDK 45nm library, demonstrate that MRUD achieves up to 30\% lower delay, 26\% smaller area, and 87\% lower power than MRSD. In BBR mode, MRUD achieves up to 43\% delay reduction over the best conventional architecture for 32$\times$32 operands, while in Multiply-Accumulate (MAC) units, performing reverse conversion only once at the end yields up to 64\% higher throughput than DesignWare and over 150\% higher than Wallace. These results demonstrate that MRUD enables high-radix redundant arithmetic with significantly improved multiplier and MAC performance, making it an efficient alternative to conventional signed digit redundant number systems.

0.30

Versatile Photonic Encoding Platform with Selectable Back Ends for Automotive Edge Processing

cs.ET ·2026-10-29 ·Isamu Takai, Makoto Nakai, Ayana Mizuno, Atsutaka Miyamichi
Abstract · 21 pages, 10 figures, 3 tables, 4 supplementary tables

The increasing diversity and computational demands of automotive edge functions require hardware architectures that support heterogeneous processing while limiting implementation complexity and system overhead. We investigate a system-level approach in which a common, fixed photonic front end provides a shared physical transformation for multiple processing functions, while application adaptation is handled by lightweight peripheral processing. We propose the Versatile Photonic Encoding Platform (VPEP), implemented using a fabricated silicon-photonic chip with 32 phase-modulated inputs, a passive fixed encoding core, and 17 on-chip photodetector outputs. Without reconfiguring the photonic core, the same chip is evaluated across four representative functions relevant to automotive edge systems: classification and regression for machine learning, analog-to-digital conversion for mixed-signal processing, and photonic information encoding and recovery for communication-related processing. Compared with linear baselines without photonic encoding, VPEP improves the mean macro-F1 score on 12 of 13 classification datasets and the mean coefficient of determination on 9 of 12 regression benchmarks. For analog-to-digital conversion, the best configuration achieves an in-sample normalized root-mean-square error of approximately 0.75% of full scale. For information encoding and recovery, a lightweight nonlinear decoder reconstructs four-character (32-bit) inputs from 17-channel encoded outputs with approximately 94% character accuracy for a 26-symbol set. These results demonstrate the feasibility of cross-application processing with a single fixed silicon-photonic encoding core and suggest a system-level design approach, complementary to reconfigurable photonic architectures, that combines common photonic hardware with lightweight application-specific processing for future automotive edge systems.

0.29

SPARQLcirc: Probabilistic Query Evaluation with Provenance Circuits

cs.DB cs.AI ·2026-10-29 ·Jingcheng Wu, Ratan Bahadur Thapa, Daniel Hernández, Steffen Staab
Abstract · 23 pages, 17 figures, 3 tables. Extended version with full proofs and additional experiments

Probabilistic query evaluation (PQE) computes the probability of every answer of a query over a tuple-independent probabilistic knowledge graph. We study exact PQE for SPARQL over graphs accessed through SPARQL 1.1 endpoints. Shared triples among the derivations of an answer make exact PQE #P-hard in general. Knowledge compilation of the Boolean provenance of every answer yields exact probabilities, and the provenance formulas of different answers can share subformulas. We propose SPARQLcirc, which rewrites a SPARQL query into CONSTRUCT queries that name every gate by its content, so that the results of independent queries form one Boolean provenance circuit for all answers. The rewriting covers joins, unions, projection, filters, OPTIONAL, difference, and transitive closures. SPARQLcirc compiles the circuit and computes the exact probability of every answer by weighted model counting. For the evaluation, we draw random probabilities for the triples of WatDiv and gMark and the statements of Wikidata and evaluate the CONSTRUCT queries on GraphDB, Fuseki, and Oxigraph. The ratio of the end-to-end time of PQE with SPARQLcirc to that of PQE by compiling the provenance expressions of the SPARQL provenance rewriting NPCS decreases with the number of derivations.

0.29

PhOSv1: Efficient Scaling of Model Sleeping on a Single Server for Autoscaling with GPU-Context Deduplication

cs.DC cs.OS ·2026-10-29 ·Tianle Sun, Xingda Wei, Rong Chen
published
Abstract · Accepted by IEEE HPCC 2026

Autoscaling is a key mechanism in model-as-a-service systems, where fast scale-up is essential because the scale-up time directly determines how long a model service remains overloaded, which in turn governs serving quality. The state-of-the-art approach, model sleeping, adopted by mainstream serving engines including vLLM and SGLang, achieves near-zero-cost scale-up by pausing pre-spawned engines on the GPU. However, sleeping is not a free lunch: each sleeping model must keep its GPU context resident, and this cost grows linearly with the number of models, degrading serving performance. Because providers must support many models, they face a dilemma between fast scale-up for all models and high service performance. In this paper, we show that a single server can keep many models asleep without paying this GPU context tax. The key observation is that the contexts of sleeping models are largely duplicated: the driver (CUDA) context can be shared across models, and their GPU kernel binaries overlap heavily due to common operators and libraries. PhOSv1 deduplicates this context by transparently delegating each engine's GPU context allocation to a central daemon through API remoting; therefore, a single server can keep far more models asleep than current engines allow. Our evaluation shows that PhOSv1 matches the scale-up time of vanilla vLLM sleep mode while substantially reducing the GPU memory footprint, improving end-to-end multi-model serving performance.

0.28

Energy-Efficient FPGA Implementation of an End-to-End Cosine-Domain CNN Using High-Level Synthesis

cs.AR ·2026-10-29 ·Sana Fatima, Ab Al-Hadi Ab Rahman, Ibrahim Yousef Alshareef, Farhad Khodadadi, Nuzhat Khan
Abstract · 9 pages, 11 figures. Currently under review at the VLSI Integration Journal

The computational cost of conventional Convolutional Neural Networks (CNNs) limits their deployment in low-power devices for high-speed edge AI applications. This paper presents an end-to-end cosine-domain CNN (CD-CNN) architecture for energy-efficient FPGA inference using high-level synthesis (HLS). The proposed architecture performs intermediate feature extraction using real-valued DCT-domain representations, avoiding complex-valued arithmetic and intermediate inverse transforms between spectral layers. The design is implemented on a Xilinx Kintex UltraScale+ XCKU5P FPGA and evaluated against a fully spectral Fourier-domain CNN (FD-CNN) under consistent network and implementation constraints. Experiments on the MNIST and 94-class ASCII datasets show that the CD-CNN reduces normalized resource utilization and improves inference throughput on both datasets. Its energy efficiency reaches 112.91 frames/J for MNIST and 101.27 frames/J for ASCII, corresponding to improvements of $1.51\times$ and $2.76\times$, respectively, over the fully spectral FD-CNN architecture. Furthermore, compared to the previous GPU-based fully spectral CD-CNN implementation, the proposed end-to-end FPGA implementation improves energy efficiency by $12.96\times$. These results demonstrate the potential of real-valued spectral processing in the DCT domain for energy-efficient edge AI deployment.

0.28

Quantum amplification of improbable futures

cs.ET stat.CO ·2026-10-29 ·Peter Maurice Catt
Abstract · 11 pages, 2 figures, 2 tables, 4 supplementary tables, graphical abstract

Some consequential futures consist of entirely plausible steps arranged into an unlikely sequence, such as ordinary weather anomalies combining into an extreme event or local equipment failures propagating into a widespread outage. Because the sequence as a whole carries little probability, it may be rarely observed in simulation. Quantum mechanics offers a way to change those odds: ideal amplitude amplification raises the chance of drawing such futures while preserving their relative likelihoods. For a five-state process six steps ahead, 13,888 of 15,625 futures are rare, yet together they account for about one draw in 40. We built the complete circuit in IBM's Qiskit and checked every part against an exact enumeration; four simulated rounds raised the share of rare futures drawn from 2.5% to 98.0%. On IBM's Kingston quantum computer, a two-step reference circuit that marks rare futures from a precomputed list drew them in 68.4% of shots against 43.5% for its control (z = 50); at three steps (about 1,170 two-qubit gates) the predicted signal reversed sign. Because this model's probabilities are known, every result could be checked against exact values; the mechanism is intended for simulators whose probabilities are hard to calculate, such as climate, epidemic and financial models.

0.28

Agentic LinkDreamer: Adaptive Neuro-Fuzzy World-Model Decision Making for Intermittent V2X Cooperative Perception

cs.NI ·2026-10-29 ·Zhaoyu Liu, Dongmin Wang, Liu Cao, Jijing Cai, Kai Fang +2 more
Abstract · This work has been submitted to Elsevier for possible publication. 38 pages, 5 figures, 3 tables

Intermittent vehicle-to-everything (V2X) communication can remove collaborator information needed for cooperative perception, degrading object detection during link outages. Existing recovery methods reconstruct missing features but rarely determine when to trust predictions, reuse historical observations, or request fresh information. This paper proposes Agentic LinkDreamer, a soft-computing framework integrating a latent world model with an interpretable neuro-fuzzy decision agent. It maintains causal collaborator memory and uses a relative-motion-conditioned recurrent world model to predict unavailable bird's-eye-view (BEV) features. Prediction uncertainty, perceptual complementarity, memory freshness, communication quality, and residual communication budget form a compact decision state. A differentiable Takagi-Sugeno controller selects among five modes: ego-only perception, motion-compensated historical feature reuse, latent imagination, low-rate temporal-summary refresh, and high-rate collaborator-feature refresh. The policy balances detection risk, communication overhead, latency, predictive uncertainty, and action switching. After link recovery, reflection compares predicted and received features to update uncertainty calibration and causal memory. For reproducible evaluation without assuming unavailable physical-layer traces, we construct a geometry-derived V2X degradation protocol using recorded trajectories and LiDAR-derived scene geometry from V2X-Real. On the real-world V2X-Real V2V split, Agentic LinkDreamer achieves 90.95 AP@0.5 on outage frames and recovers 92.9% of the cooperative-perception performance gap. At a high-performance operating point, it remains only 0.65 AP below the cooperative upper reference while reducing communication volume by approximately 31% relative to continuous high-rate cooperation. This improves resilience under V2X outages.

0.27

LLM Fine-Tuning Workloads: Characterization, Challenges, and Opportunities for Serverless AI

cs.DC ·2026-10-29 ·Radu Nicolae, Daniele Lotito, Alexandru Iosup
published
Abstract · Technical report; extended version of the paper published at the Workshop on AI and Serverless (WoAIS '26), co-located with ACM/IFIP Middlew…

Serverless execution of Large Language Models (LLMs) is becoming important, reflecting their wide societal use. An increasingly present stage in LLM pipelines is fine-tuning, which specializes trained models and improves inference performance. Serverless operation of LLM fine-tuning workloads shifts operational concerns from the user to the provider. For these new workloads, providers face important operational complexities, such as performance issues, resource waste, and even service failure due to misconfiguration. However, optimizing such workloads is still at an early stage, and, relevant to this work, relatively little is currently known about how LLM fine-tuning workloads run in practice. In this work, we address the knowledge gap by characterizing the execution of LLM fine-tuning workloads on clusters equipped with different GPUs, running LLM models at various scales. We analyze IBM's LLMFineTuningBench, a large-scale, diverse profiling dataset of 30,920 fine-tuning experiments that records both successful and failing configurations. We provide quantitative insight into throughput (expressed in tokens/s), scaling, failure, and potential for performance optimization for four commonly used GPU types, in setups of 1 to 32 GPUs. We find that (1) job feasibility (run or fail) depends on workload properties, with failure ratios rising from 52% for small batch sizes (16-64) up to 95% for the largest (1,024); (2) throughput grows logarithmically with batch size; (3) scaling is near-linear when the workload is deployed intra-node, and optimizations such as RoCE substantially improves efficiency for inter-node deployment (e.g., 0.72 vs. 0.26 efficiency without RoCE); and (4) system-level optimizations can outweigh application-level ones (RoCE, up to 2.9$\times$, vs. Fast MoE, 1.7$\times$).

0.27

LAMP: LLM-Driven Analog Placement through Adaptive Tool Composition

cs.AR ·2026-10-29 ·Zecheng Xu, Guanhua Chen, Yumao Wu, Keren Zhu
Abstract · 5 pages, 3 figures, 1 table

Analog placement requires circuit-specific trade-offs among area, matching, routability, and electrical performance. Balancing these trade-offs through predefined objectives and fixed placement flows remains difficult. In this paper, we propose an analog placement framework in which a large language model (LLM) acts as a layout engineer through composable placement components. These components support constraint management, placement and legalization, and lightweight evaluation of routability and electrical performance. Using multimodal checkpoints of the initial and evolving design, the LLM selects and combines components to construct and revise its own placement flow. It also adapts the placement hierarchy by reopening modules or partitioning and regrouping devices, allowing local layouts to respond to global design needs. An experience skill provides reusable design knowledge to guide these decisions. Evaluation on two SMIC 180 nm circuits shows that the framework reduces placement area and improves most measured post-layout performance metrics.

0.26

Quantifying the Relay-Coverage Trade-off in Mobile Multihop Flooding

cs.NI cs.LG ·2026-10-29 ·Johannes F. Loevenich, Thies Möhlenhof, Roberto Rigolin F. Lopes
unusual
Abstract

Network-wide flooding in mobile multihop networks selects relays from locally maintained topology information. Mobility makes this information outdated, leaving two basic responses: add redundant relays or refresh the topology and re-select relays more frequently. How these mechanisms trade reliability for overhead, however, is not well quantified. We quantify this trade-off on a public trace of 157 vehicles using Optimized Link State Routing version 2 (OLSRv2) and exact minimum-cost relay sets. Twofold coverage increases median relay-cover lifetime by only 2.3x, rather than about an order of magnitude as predicted by an independent-hazard model. The limitation is structural: 43.6 % of two-hop neighbours have only one possible relay, and newly appearing neighbours invalidate existing covers. Moreover, the delivery gain of a second relay falls from 0.061 to 0.016 when relay selection uses instantaneous link state, showing that redundancy mainly compensates for outdated topology information. Relay coverage and topology refresh therefore form a measurable redundancy-freshness trade-off rather than independent robustness mechanisms. In our trace, twofold coverage has lower overhead below about 100 bit/s of application flooding per node, whereas fresher topology information has lower overhead above it. This makes relay redundancy a load-dependent design choice rather than a fixed heuristic robustness setting.

0.26

Learning to Formulate: Agentic Hardware-Software Co-Design of AI Infrastructure through Progressive Optimization and Verification

cs.AR cs.AIcs.DC ·2026-10-29 ·Yi Ding, Li Shang
Abstract

AI infrastructure, the chips and software that train and serve AI models, is designed in layers. Layers make the system manageable, but they also fix which choices an optimizer can consider. Gains that need the chip and its software to change together are excluded before optimization begins: the formulation leaves them out. Language-model agents make it possible to revisit these boundaries, but an agent's proposal cannot show by itself that a design is correct or optimal. We present Formulaic, a flow that divides the work. The agent writes the formulation, a mathematical solver optimizes within it, and verification and measurement check its assumptions and results. The agent refines the problem step by step, from the computation graph toward the chip. Each step is a template: a family of programs with the conditions under which they are correct. Wall relaxation searches the chip by pricing the hardware limits that bind the software. This paper takes a first step, through separate studies rather than one end-to-end run. On NVIDIA GB10, templates the agent wrote run complete Qwen3-8B layers $1.075\times$ faster than SGLang. In a simulated Ventus GPU study, joint design cuts the worst-scenario cycles of Qwen2.5-0.5B to 0.428 of the reference, against 0.934 for software alone. On a five-step GEMM refinement chain, reusing proofs makes checking $2.51\times$ faster.

0.26

An Exact Parallel Dense Gauss-Seidel Algorithm with Trade-Offs between Parallel Execution Time and Preprocessing Time

cs.DC cs.DS ·2026-10-29 ·Mugurel-Ionuţ Andreica, Pantelimon George Popescu, Eliana-Dina Tîrşa
Abstract · This manuscript was originally written in July 2013, and the names and affiliation of the authors are reflecting the state at that time. Rec…

The Gauss-Seidel iterative method is commonly used for solving systems of linear equations because of its good convergence properties. However, the Gauss-Seidel algorithm cannot be easily parallelized because of the data dependencies existing at each iteration. The best parallel execution time for dense systems of equations is $O(n)$ given that enough processors are available and no preprocessing is performed. In this paper we propose a new parallel Gauss-Seidel algorithm which is capable of achieving better parallel runtimes at the expense of more preprocessing time. The algorithm can be tuned in order to provide convenient trade-offs between the preprocessing time and the best parallel runtime achievable by the algorithm.

0.24

Nearly-Optimal Time-Sample Trade-off in Consensus Dynamics

cs.DC cs.DSmath.PR ·2026-10-29 ·Francesco d'Amore, Joel Rybicki, Nobutaka Shimizu, Takeharu Shiraga, Robin Vacus
theory
Abstract

How does the number of sampled opinions affect the time required to reach consensus in distributed consensus dynamics? We study the synchronous $h$-Majority, in which every node independently samples $h$ nodes uniformly at random with replacement in each round and adopts a most frequent sampled opinion, breaking ties uniformly at random. We show that $h$-Majority achieves the nearly-optimal trade-off between the sample size and the consensus time, among a broad class of consensus dynamics. In particular, we prove the following results. For every sample size $h \ge 3$ and for every initial configuration, we prove that the process reaches consensus with high probability within $$\widetilde{O}\left(1+\min\left\{\frac{1}{\min\{h^2p_{\max},1\}},\frac{\sqrt{n}}{h}\right\}\right)$$ rounds, where $n$ is the number of nodes and $p_{\max}$ is the largest initial fraction of nodes supporting any single opinion. This is the first bound on the consensus time for $h$-Majority that holds for every $h\ge3$ and every initial configuration. When the initially most popular opinion has a sufficiently large lead over every other opinion, we show that all nodes adopt this opinion with high probability within $O(\log n/\min\{h^2p_{\max},1\})$ rounds. We establish a matching lower bound up to polylogarithmic factors for any fixed update rule that treats all opinions symmetrically and that selects either a sampled opinion or the node's current opinion.

0.24

Characterizing Performance Tradeoffs in Heterogeneous GPU-Wafer-Scale Accelerator Systems

cs.DC cs.AR ·2026-10-29 ·Ziyu Hu, Zhijing Ye, Jiamin Wang, Zhiqing Zhong, Zhendong Mi +4 more
Abstract

As AI accelerator architectures become increasingly diverse, integrating different architectures into a unified system is emerging as a new heterogeneous computing paradigm. GPUs and WSEs differ substantially in compute, memory, and communication organization, creating both architectural complementarity and additional overheads, yet how to translate this heterogeneity into practical performance gains remains unclear. To address this gap, we propose a hierarchical benchmarking methodology to characterize GPU--WSE heterogeneous execution and the conditions for exploiting complementarity, together with an analytical performance model that identifies when heterogeneity becomes a bottleneck and when it yields system-level performance gains.

0.23

Sandman: A Sandbox Service for Production Agentic Reinforcement Learning

cs.DC cs.AIcs.LG ·2026-10-29 ·Yiming Yan, Hongyang Wang, Wenbo Chen, Xuan Wang, Yichuan Li +10 more
code/data
Abstract · 28 pages, including appendix

Agentic reinforcement learning (RL) needs stateful environments for tool execution and evaluation. Training jobs request thousands of environments in bursts that can overwhelm shared cluster services, while cancelled or timed-out requests can leave resources running. We present Sandman, a sandbox service built and operated at Amazon for large-scale agentic RL workloads in production on shared Kubernetes. A small daemon mounted into unmodified task images preserves files and processes across calls. Sandman paces creation requests, checks sandbox readiness, and retains cleanup responsibility after callers time out or cancel. Production deployments created 3.1 million sandboxes for 122 coding-agent training runs in two weeks. Peak production concurrency was 11,858 sandboxes. In one 10,000-request multi-client burst, Sandman made half the sandboxes usable in 116 seconds and all in 308 seconds. Full delivery took 44% less time than a tuned configuration of agent-sandbox, an open-source sandbox manager for Kubernetes. Sandman left no sandboxes at client exit. On software-engineering task images, median no-op command latency was 2.0 ms, compared with 14.3 ms for agent-sandbox and 203.5 ms for OpenSandbox. We also measure provisioning costs and sandbox delays within training runs, and report design trade-offs and operational lessons from production use and comparisons with three open-source systems. We will open-source Sandman's daemon, provider and pool.

0.23

Multistatic ISAC with Deep Learning-Based Clutter Suppression for 5G and Beyond

cs.NI cs.AIcs.LGeess.SP ·2026-10-29 ·Yalin E. Sagduyu, Tugba Erpek, Sastry Kompella
Abstract

Integrated sensing and communications (ISAC) systems using 5G and beyond signals of opportunity enable joint wireless sensing and communication using shared infrastructure and spectrum resources. While multistatic ISAC can increase sensing coverage with spatially distributed receivers, dense and dynamic clutter can significantly degrade bistatic range and Doppler estimation, leading to errors in target localization and velocity estimation. This paper presents a multistatic ISAC framework with clutter suppression for high-fidelity estimation of target position and velocity. Distributed clutter is modeled with randomly located scatterers with varying Doppler spreads and radar cross sections to capture dense and dynamic clutter effects. A broad range of conventional signal processing baselines are evaluated for clutter suppression, showing limited robustness under strong clutter conditions. To address this limitation, deep learning-based clutter suppression methods are introduced to improve robustness against dense and dynamic clutter by employing residual and attention-enhanced deep neural network architectures that directly enhance target responses from clutter-contaminated observations. These methods substantially outperform conventional approaches in target position and velocity estimation across diverse clutter conditions in multistatic ISAC.

0.22

NORmal: Architecting 3D Ferroelectric NOR Flash as Main Memory -- From Co-Fabricated BEOL Arrays to 200B-Class Multi-Agent LLMs in a Smartphone

cs.AR ·2026-10-29 ·Zhuoran Li, Zhuohang Bian, Yibo Zhao, Xuxiao Yang, Youwei Zhuo
Abstract · 13 pages

Smartphone agents increasingly work as teams, in which a large planner directs parallel workers across long sessions. Running such a team on the device requires 200B-class models and contexts that exceed phone DRAM many times over, yet existing tiers treat flash as storage: UFS starves per-token weight streaming, and mobile High-Bandwidth Flash (HBF) needs a redesigned SoC and cannot absorb continuous KV writes. We show that a non-volatile memory can join an unmodified LPDDR6 bus as a plain rank only if its access latency fits the DRAM timing window, a condition no NAND tier meets. We present NORmal, to our knowledge the first main-memory architecture for 3D ferroelectric NOR flash. We have fabricated and experimentally characterized back-end-of-line (BEOL) 3D FeNOR arrays that co-integrate dense 3-bit MLC and fast, high-endurance in-place-writable SLC with sub-7.5 ns sensing. NORmal turns these co-fabricated arrays into a byte-addressable, deterministic-timing load/store rank on an unmodified LPDDR6 interface through an under-array DRAM-emulation bridge, staggered bank reads with zero-cycle weight swizzling, and posted SLC writes. Across this rank and DRAM, NORmal places agent state by mutability and lifetime rather than by hotness, serving 200B-class models and persistent planner contexts in place while dynamically partitioning subchannels across agent phases. In serving-level simulation on production agent traces, NORmal is 8.9$\times$--28.0$\times$ faster than UFS offloading, within 0.4% of an unlimited-DRAM ideal, and 2.8$\times$--6.3$\times$ faster than HBF on sessions beyond 256K tokens, with 8.1$\times$--17.8$\times$ less memory energy per token.

0.21

How Far Are We From On-Device VLM Inference? A Measurement Study on Edge NPUs

cs.PF cs.AR ·2026-10-29 ·Huan Yang, Shiqi Jiang, Yu Mao, Chuanyue Xiong, Deyu Zhang
Abstract · 26 pages, including references and appendices

Diverse optimizations spanning model architecture, inference runtime, and hardware acceleration have been proposed for on-device vision-language model (VLM) deployment on edge NPUs. Still, cross-layer coordination of VLM execution on edge NPUs remains largely underexplored. This paper presents an empirical measurement study across six representative VLMs, three real-world edge platforms, and ten multimodal benchmarks. Our evaluation drills down from end-to-end application-level latency and thermal traces down to low-level kernel-scale execution behaviors, challenging several widely held community assumptions. First, softmax dominates attention costs; even lookup-table-based softmax optimizations cannot eliminate pipeline bubbles that severely underutilize NPU matrix units. Second, theoretical speedups from linear attention vanish under real-NPU execution due to prohibitive tensor-rearrangement overhead. Third, no hardware backend universally dominates autoregressive decoding: attention-tensor-shaped parallelism dictates NPU efficiency and shifts the CPU-NPU performance balance. We further find task-specific Sweet Resolutions for VLMs, which deliver meaningful accuracy gains at only modest latency overhead. Complementing these findings, we provide actionable cross-layer recommendations for hardware architects, runtime developers, and model designers.

0.21

Can We Rely on Language Models to Orchestrate Services across the Edge-Cloud Continuum?

cs.DC ·2026-10-29 ·Zhenghao Wu, Ruihai Dong, Reza Farahani, Hamed Haddadi, Omer Rana +1 more
Abstract

Service orchestration across the edge--cloud continuum increasingly uses language models (LMs) to translate intents and operational evidence into scheduling, configuration, diagnosis, and remediation decisions. Yet, useful model outputs do not by themselves establish dependable orchestration: reliability also depends on where the LM enters the decision process, what authority it is granted, how actions are constrained, and what system-level evidence supports the claimed effect. Existing surveys provide limited synthesis of the links between continuum-specific orchestration decisions, LM responsibility, operational authority, assurance controls, and supporting empirical evidence. We systematically review 39 primary studies and develop a five-dimensional analytical framework for connecting system context, orchestration decisions, LM agency, operational safeguards, and evaluation evidence. The synthesis reveals predominantly hybrid designs in which LMs support interpretation, candidate generation, and mitigation planning, while downstream components commonly retain feasibility enforcement and execution authority. Bounded closed-loop remediation is emerging, but continuum-specific constraints remain unevenly represented in decision formulations, and production-scale evidence across heterogeneous edge--cloud environments remains limited. These findings suggest that reliance on LM-enabled orchestration depends on aligning decision constraints, operational authority, safeguards, and evaluation evidence with the claimed system effect.

0.19

Tessera: Fast Join-Aggregates over Skewed Sparse Relations in Database Engines

cs.DB ·2026-10-29 ·Gyuyeong Kim
Abstract

Join-aggregate queries count or sum matching rows for each pair of values, often over skewed sparse relations in which a few values are frequent and most pairs of values never co-occur. Conventional execution plans materialize billions of intermediate join rows, taking minutes to tens of minutes per query. Recent matrix-based solutions reformulate these queries as matrix multiplication to reduce or avoid this materialization. However, they remain slow because they either store result cells for non-matching pairs or still materialize intermediate join rows for infrequent values. We present Tessera, a skew-aware join-aggregate operator for database engines that avoids both sources of overhead. Specifically, Tessera computes a sparse matrix product directly over the engine's rows, emits only nonzero cells as result rows, and prunes values that cannot meet the query's threshold before multiplication begins. It also offloads the dense block formed by pairs of the most frequent values to a CPU matrix accelerator when that block is dense enough. Experimental results show that Tessera is faster than the fastest existing system by 2.8x to 18.8x. In addition, Tessera outperforms DuckDB's execution plans by up to 1,802x.

0.19

AI in the EGI e-infrastructure -- Past experiences and future strategies

cs.DC cs.AIcs.CY ·2026-10-29 ·Gergely Sipos, Sergio Andreozzi, Marco Rorro, Andrea Manzi, Tiziana Ferrari
Abstract · 8 pages, 1 figure. Preprint of a paper submitted to the proceedings of the 28th Conference on Computing in High Energy and Nuclear Physics (…

The EGI Federation has provided federated e-infrastructure services to European and global science for over 15 years. We review EGI's early experiences with AI in the interTwin and iMagine Horizon Europe projects, which built digital twins in physics and environmental sciences and an image-analysis platform for aquatic research. We then examine how the AI landscape has changed since 2023 and argue that the focus of AI in science has shifted from model training to adapting and applying AI to specific domains, projects and questions. Building on these lessons, we present the AI strategy emerging from the EGI Federation Strategy 2026-2030, organised around three enablers: (1) an AI-ready infrastructure connecting compute, data and application resources across the continuum; (2) human skills to know what is possible with AI and how to implement it; and (3) trust, to run AI with confidence even in high-risk environments. We describe the projects implementing each enabler, including federated LLM inference, AI-ready data platforms and AI competence centres, and discuss a division of labour with EuroHPC's AI Factories, RAISE, research infrastructures and EOSC, in which EGI specialises in the customisation and long-term operation of AI for research.

0.18

Rimuru: Unified Memory Management for Modern LLM Serving via Elastic Expert Offloading

cs.OS cs.LG ·2026-10-29 ·Zhiyuan Fang, Dingyang Li, Jianfeng Zhong, Bin Wang, Weixi Zhu +4 more
Abstract

Modern LLMs strain device memory through large model weights and attention states, particularly in resource-constrained scenarios. Existing systems separately provision memory for model weights and attention states, imposing fixed fixed boundaries that cannot adapt to the changing memory demands of workloads. Under heavy workloads, insufficient state capacity limits request concurrency and reduces throughput; under light workloads, idle memory in one region cannot be reallocated to another, wasting device memory. Meanwhile, modern LLM architectures introduce heterogeneous tensors with diverse layouts and lifetimes, further complicating memory management. These problems motivate unified memory management for heterogeneous tensors. However, a shared pool alone does not make memory directly reusable across tensors with different allocation granularities and execution layouts. This paper presents Rimuru, a modern LLM serving system that unifies memory management through elastic expert offloading. Rimuru establishes a shared page pool to enables elastic memory sharing across heterogeneous tensors. To accommodate different allocation granularities, Rimuru introduces memory views that allow each tensor type to interpret the shared memory space using its own page size and page table. Furthermore, Rimuru combines huge-page alignment with a multi-level page allocator that splits and merges pages, enabling cross-type capacity reuse. A page-table-aware fused MoE kernel directly accesses non-contiguous expert weights, avoiding copy overhead. Building on these mechanisms, Rimuru employs a workload-aware adjuster that dynamically reallocates pages across tensor types according to runtime demands. Implemented on top of vLLM, Rimuru remains compatible with modern LLMs. Evaluations show that Rimuru achieves up to 19.03x baseline throughput and reduces TTFT by 98.1% on average.

0.18

FedMaJ: Ordinal Meta-Aggregation for Federated Learning Using Majority-Judgment Voting

cs.NI ·2026-10-29 ·Driss Choukri, Essaid Sabir, Halima Elbiaze, Abdelkrim Haqiq
Abstract · Submitted to IEEE ICC 2027

Federated learning traditionally relies on a fixed aggregation rule despite shifting client state and wireless dynamics. We present FedMaJ, a framework that dynamically selects aggregation rules via Majority Judgment voting. During voting rounds, clients evaluate candidate models across performance, resource utilization, channel condition, and fairness metrics. The server adopts the majority-preferred rule and retains it until the next vote, scheduled dynamically based on accuracy trends. Evaluated in over-the-air federated learning with explicit energy and resource cost modeling, synthetic and Fashion-MNIST experiments show that FedMaJ achieves competitive learning performance and superior aggregation-influence fairness. Under extreme-evaluation perturbations, Majority Judgment yields significantly more stable rule selections than mean-based scoring, establishing ordinal voting as a robust control layer for adaptive aggregation.

0.17

How Global Is Anycast?: Evaluating Performance Gaps

cs.NI ·2026-10-29 ·Mohammad Shaikh Sadid Khan, Karishma Setya, ASM Rizvi
Abstract · 9 pages, 9 figures, 1 table

Anycast networks aim to reduce latency by directing users to nearby service sites. However, users can still experience high latency even when accessing services via anycast. Some anycast deployments are localized and intended to serve a specific geographic region, while others are globally distributed to support users worldwide, with the expectation of consistently low latency across locations. In practice, latency can remain poor for certain users due to factors such as routing inefficiencies and peering relationships. These inefficiencies are often more pronounced in regions such as Africa, Asia or South America, where Internet connectivity is generally less mature than in North America or Europe. Despite this, there has been no in-depth analysis of anycast performance in these continents. In this paper, we examine the latency experienced by users in these regions when accessing anycast networks. We analyze more than 10,000 anycast prefixes, quantify the severity of latency problems, and show that clients in Africa, Asia, and South America experience substantially higher latency than those in North America. We also investigate the underlying causes of this performance gap.

0.15

Only Accepted Tokens Count: An Energy Accounting of Speculative Decoding Under Latency Targets

cs.DC cs.AIcs.LGcs.PF ·2026-10-29 ·Bole Ma, Ayesha Afzal, Jan Eitzinger, Harald Koestler, Gerhard Wellein
Abstract

Large language model serving is limited by facility power. Energy per token determines how many tokens a datacenter can serve, while a latency target bounds how quickly each user receives them. During decode, every weight is read from memory for every generated token. Batching amortizes this read across users but makes no individual user faster. Speculative decoding does: a drafter proposes several future tokens, and the target model verifies them in a single forward pass that reuses the same weight read. Because every drafted token must be verified whether or not it is accepted, the energy speculation saves depends on how many drafted tokens are accepted. This saving is hard to pin down. It varies across target models from different attention families and across drafting schemes, and it depends on the text being served. Every drafter we evaluate accepts fewer tokens on conversation than on code, and every trained drafter accepts fewer on Romansh, a low-resource language, than on English. We measure speculative decoding under production serving conditions (SGLang, CUDA graphs enabled) on H100 and H200 GPUs, at batch sizes up to 64 across 5 corpora. On dense models, verification, not drafting, dominates the energy cost. The saving shrinks as the batch grows, because batching has already shared the weight read that speculation would divide among the accepted tokens. At large batch sizes, block drafters push verification past the balance point between arithmetic and memory traffic into the compute-bound regime, where the saving becomes a loss, and a drafter with low acceptance raises energy per token. Meeting a tighter latency target with speculation then costs additional energy. Larger batches lower energy per token without speculation. Speculation adds to that saving only when acceptance is high, and on the model we tested it leaves little headroom for latency-preserving DVFS.

0.14

TRACEs: Fine-Grained Embedded Provenance for Distributed Scientific Workflows

cs.DC cs.CE ·2026-10-29 ·Ty Anderson, Kin Ng, Rajiv Mayani, Jay Lofstead, Ewa Deelman +1 more
Abstract

Large-scale scientific workflows run on distributed HPC resources, making it hard to explain how scientific data products are generated and why similar workflow executions can yield very different results. Existing systems usually focus on either distributed workflow execution or provenance management and rarely treat fine-grained provenance as an intrinsic property of distributed scientific data artifacts. We present TRACEs (Traceable Runtime for Advanced Containerized Executions), a provenance architecture for distributed containerized scientific workflows that embeds data lineage and execution context directly in every generated data artifact. By decoupling provenance from workflow orchestration, TRACEs generalizes artifact-centric provenance from local container runs to distributed workflows while remaining compatible with existing workflow management systems. We evaluate TRACEs on a seven-stage E. coli variant-calling workflow distributed across multiple HPC nodes. Our results show 100% provenance completeness under distributed execution, about 59% less intermediate storage via sparse IMG containers, and runtime and zero-copy I/O throughput comparable to fixed-size SIF containers, plus provenance-guided explanations of workflow variability that trace a 200-fold difference in variant counts to differences in input sequencing data without re-executing the workflow.

0.12

Federated Agentic AI for Interference-Aware Adaptive Spectrum Coordination

cs.NI ·2026-10-29 ·Tomer Baziza Paz, Yaniv Cohen, Kobi Cohen
benchmark
Abstract · 13 pages, 6 figures

Emerging intelligent wireless networks increasingly rely on autonomous AI agents to adapt spectrum usage to dynamic and heterogeneous operating conditions. However, collective learning is challenging when agents experience different local interference environments and cannot exchange raw observations due to communication-overhead and privacy constraints. We propose Federated Agentic Spectrum Coordination with Interference Awareness (FASCIA), a federated multi-agent deep reinforcement learning framework for adaptive spectrum coordination. In FASCIA, each network manager acts as an autonomous AI agent that perceives its local interference environment, accumulates experience in a persistent local memory, and learns an adaptive spectrum coordination policy through local interactions. Coordination among agents is achieved through periodic federated aggregation of locally trained models, without exchanging raw sensing data or experience. Since spectrum decisions made by one agent alter the environments observed by others, local learning is inherently heterogeneous and coupled. FASCIA employs proximal regularization to mitigate the resulting model drift and explicitly accounts for inter-carrier interference arising from partially overlapping channels. FASCIA is the first federated agentic AI framework to address this challenging interference-coupled spectrum coordination setting. Extensive simulations demonstrate that FASCIA achieves performance close to the centralized benchmark while avoiding raw-experience exchange and maintaining effectiveness across heterogeneous network configurations.

0.11

Analysis of Triggered Packet Streams: Stochastic Bounds on System Time under General Triggering Delays

cs.PF math.PR ·2026-10-29 ·Mehran Rahnamania, Michel Mandjes, Farid Ashtiani
theory
Abstract

Many emerging communication systems increasingly involve devices that transmit packets in multiple causally dependent stages as part of an internal workflow, rather than waiting for receiver feedback. To analyze such systems, we recently introduced the $\mathrm{M^T/G/1}$ queue, in which primary customers arrive according to an exogenous Poisson process, and each triggers a secondary customer to join the queue after an independent random delay, called the \textit{triggering delay}. Our prior heuristic approximations assume identical exponential service times, and our matrix-analytic framework assumes exponential triggering delays. However, no exact analysis is available when both distributions are general. In this paper, we derive stochastic bounds, in the usual stochastic order, on the system-time distribution of each customer class under general triggering-delay and service-time distributions. The service times within each primary--secondary pair may be generally dependent. Using sample-path coupling and level-crossing arguments, we show that the stationary workload is stochastically sandwiched between those of two classical queues: a two-class $\mathrm{M/G/1}$ queue with independent Poisson arrivals (lower bound), and a batch-arrival $\mathrm{M^{[2]}/G/1}$ queue where each primary--secondary pair arrives simultaneously (upper bound). Notably, these bounds do not depend on the triggering-delay distribution, and they are asymptotically tight in the sense that the stationary workload converges in distribution to that of the lower- and upper-bound queues as the triggering delays diverge to infinity and converge to zero in probability, respectively. We translate these workload bounds into system-time bounds for each customer class. Simulation experiments support these bounds.

0.08

SeQUeNCe.DQC: Full-stack Simulation of Distributed Quantum Computing

cs.NI cs.DCquant-ph ·2026-10-29 ·Sanjay Suresh, Caitao Zhan, Joaquin Chung, Rajkumar Kettimuthu
benchmark
Abstract · 6 pages, 6 pages

Distributed quantum computing (DQC) connects multiple quantum processing units to overcome the scale limits of monolithic hardware. We introduce SeQUeNCe$.$DQC a new top-level module within the SeQUeNCe package that compiles monolithic circuits into topology-aware execution plans and coordinates their execution using SeQUeNCe's existing quantum-network modules and protocols. We validate SeQUeNCe$.$DQC by reproducing a two-node trapped-ion Grover experiment in simulation and comparing success probabilities with the reported 71% benchmark. We then evaluate how compiler strategy, topology, and memory capacity jointly govern distributed circuit execution time and success rate. In these experiments, our hybrid compilers successfully identified mixed TeleGate/TeleData execution plans that outperformed pure gate-teleportation strategies. By capturing the physical overhead of distributed quantum primitives, SeQUeNCe$.$DQC provides a rigorous platform for co-designing DQC software and hardware architectures.

0.07

Enabling Dynamic Resource Scaling in Quantum Circuit Simulations

cs.DC ·2026-10-29 ·Íñigo Aréjula-Aísa, Sergio Iserte, Petter Sandås, Ricard S. Raigada-García, Oliver Brown +2 more
Abstract

Studies of quantum algorithms often simulate se- quences of circuits whose number of qubits changes from one circuit to the next, such as instances of increasing size of the same problem. Since each additional qubit doubles the state- vector, every circuit needs a different number of nodes, which a fixed allocation on a shared High-Performance Computing (HPC) system cannot provide: a job sized for the largest circuit leaves most of its nodes idle, and one job per circuit adds scheduler interactions and queue waits. We make the QuEST state-vector simulator malleable with the Dynamic Management of Resources (DMR) framework, so that a single job resizes between circuits. Because no state-vector exists between circuits, reconfigurations require no data redistribution and no changes to the simulation kernels, and a checkpoint holding the index of the next circuit is enough to resume. A right-sizing policy derives the number of nodes from the number of qubits, and the circuit_executor API hides reconfigurations from the user. We evaluate the reconfiguration overhead on MareNostrum 5 and show that a single job can run a batch of circuits of different sizes, adapting its allocation to each of them with negligible overhead.

0.05

Software & Programming 53

PR4Code: A Pull Requests Dataset for AI Code Generation

cs.SE ·2026-10-29 ·Benedetta Donato, Leonardo Mariani, Daniela Micucci, Oliviero Riganelli
publisheddataset
Abstract · Published in IEEE Access

AI-driven code generation is evolving from handling small, localized code-completion tasks toward supporting the implementation of full software features. To support this transition, this paper introduces PR4Code, a dataset of 4,508 Java and 8,831 Python curated Pull Requests (PRs) collected from GitHub, each enriched with metadata, commit histories, and detailed code changes. Unlike resources focused on isolated code fragments, PR4Code captures the feature implementation process, delivering a large and diverse collection of real-world development tasks that can be used to assess AI-driven solutions. Our analysis reveals substantial variability in PR structure, commit granularity, and textual content. In addition, we release scripts to regenerate and update the dataset, ensuring reproducibility and maintainability.

0.63

Arithmetical Characterizations of Two Intuitionistic Logical Principles

cs.LO ·2026-10-29 ·Kaito Ichikura
publishedtheory
Abstract · In Proceedings NCL'26, arXiv:2610.34489

This paper investigates the role of negation in arithmetic by analyzing subsystems based on subminimal logic. We introduce NA^PR, a system of subminimal arithmetic equipped with all primitive recursive functions, and answer the following four questions: (1) Can we develop reasonable arithmetic within NA^PR? (2) Can we characterize logical rules Absorption of Negation (An) and Ex Falso Quodlibet (ECQ) by arithmetical principles? (3) Are there strict differences in deductive power among NA^PR, MA^PR, and HA? (4) Do MA^PR and MA, and NA^PR and NA have a similar mutual interpretability relationship as HA? We show the following: First, Heyting Arithmetic (HA) is shown to be a conservative extension of NA^PR for positive formulas. Based on NA^PR, we prove that (ECQ) is equivalent to the definability of disjunction, and that (An) is equivalent to a variant of the least number principle. Furthermore, using the constant domain neighbourhood semantics, we show that the systems NA^PR, MA^PR (written as NA^PR+(An)) and HA (written as NA^PR+(An)+(ECQ)) form a strictly increasing hierarchy of strength. Finally, we demonstrate the essential role of primitive recursive functions in weak arithmetic using classical countermodels.

0.54

Unraveling the Efficacy of Neuron Coverage Criteria on Testing Large Language Models

cs.SE ·2026-10-29 ·Xiaoning Ren, Jiayang Song, Chongyang Liu, Jie Li, Lei Ma +1 more
top venue
Abstract · Published in ACM Transactions on Software Engineering and Methodology (TOSEM), 2026. 35 pages

Neuron coverage criteria aim to assess test suite quality and guide test input generation, but their effectiveness for Transformer-based language models remains unclear. We evaluate six criteria, including four established and two recent criteria, across encoder-only, decoder-only, and encoder-decoder models. Our evaluation examines error-revealing capability, distribution diversity, sample fairness (category diversity), and guidance for test input generation. The results reveal uneven effectiveness. Erroneous samples provide only 0.32% additional coverage over noise samples, indicating limited sensitivity to adversarial errors. By contrast, coverage criteria effectively assess distribution diversity and sample fairness in 26 and 30 of 33 configurations, respectively. Incorporating neuron-wise coverage guidance into test input generation increases adversarial sample production by only 4.56%. These findings suggest that existing coverage criteria can serve as an antidote for assessing diversity but remain largely a placebo for revealing adversarial errors, particularly when used individually. Their practical value therefore depends on whether these limited benefits justify the computational overhead of testing large Transformer-based models. Our study highlights the need for more robust and efficient coverage criteria tailored to these models, with stronger error-revealing capability and greater value for test generation.

0.53

Block-exchange descent for insertion: commuting projections and minimum masks

cs.FL math.CO ·2026-10-29 ·Haobo Ma, Wenlin Zhang, Charles E. Hughes
theorycode/data
Abstract · 22 pages, 10 tables. Software and verification data: https://github.com/the-omega-institute/insertion-degree-descent/releases/tag/v1.0

For fixed nonempty words $u,v$, the minimum insertion degree of a shuffle is the least number of inserted-source runs among its representations. We prove that, after a block exchange from a minimum mask of degree $r$, the minimum over masks with commuting source projections is exactly $r-1$; a larger decrease occurs precisely when a noncommuting mask has fewer runs. If either source is unary, every such exchange is exact. We prove an explicit repair for an infinite periodic degree-three family. Computer-assisted results establish first-mask descent through combined length seventeen over every alphabet, degree-three first-mask descent through length thirty, and a uniform repair of all degree-three failures through length thirty, including all further letter identifications. Counterexamples to arbitrary, first, and last minimum-mask choices explain the need for alternative masks. Finally, we prove that minimum iteration depths have no gaps for arbitrary languages. General block-exchange descent and Singleton No-Gap remain open.

0.50

Uniform Non-Deterministic Semantics for Ciuciura's $mbC^n$-Hierarchy

cs.LO ·2026-10-29 ·Ela Drozdowska, Pawel Pawlowski
publishedtheoryunusual
Abstract · In Proceedings NCL'26, arXiv:2610.34489

We introduce a uniform non-deterministic semantics for the hierarchy of paraconsistent systems mbC^n introduced by Ciuciura. The proposed construction generalizes Omori's three-valued semantics for mbC^1 to all systems of the hierarchy. We prove soundness and completeness of the semantics with respect to the corresponding axiomatizations and show that the consequence relations of the systems mbC^n cannot be characterized by finite deterministic matrices.

0.50

Tableaux Rules for Theories of Definite Descriptions in Positive Free Logic with Some Philosophical Remarks

cs.LO ·2026-10-29 ·Nils Kürbis
published
Abstract · In Proceedings NCL'26, arXiv:2610.34489

While Russell's theory of definite descriptions, expressions of the form 'the F', is almost incontrovertible in classical logic, in free logic matters stand differently. Several alternatives have been proposed. In Lambert's hierarchy, the weakest theory is his minimal free theory of definite descriptions (MFD), which only says that anything is the F iff it is the unique existing F, the strongest one is a chosen object theory in which all improper definite descriptions refer to the same object (FD2). Intermediate is Lambert's FD1, which adds ix(x = t) = t as an axiom to MFD. MFD is very weak. FD2 is very strong. One might hope to find intermediate systems in which ix(x = t) = t is provable from more fundamental principles. The present paper considers one avenue where one might hope to find them and closes it. Two theories, FDA and FDH, will be considered that are motivated by considerations from proof-theoretic semantics. The former turns out to be equivalent to FD2, the latter to have unreasonable consequences. A third theory, similarly motivated, is also not viable. The philosophical upshot of this is briefly discussed.

0.50

Manual binding and semi-automated interface generation for efficient simulation backends with user-friendly frontends: A~demonstration for epidemic use cases

cs.SE ·2026-10-29 ·Maximilian Betz, Daniel-Michel Richter, Daniel Abele, René Schmieding, Henrik Zunker +1 more
Abstract · 27 pages

Challenges and chances of mathematical software are directly entangled with the programming language used. Depending on the background of developers and users, conflicts between different goals may arise. While several, e.g., scripting languages allow a low barrier and user-friendly entry to mathematical modeling and simulation, corresponding implementations will most likely not allow timely outcomes for detailed simulations. In infectious disease modeling, while several large-scale software frameworks rely on C or C++ implementations, most models are implemented in Python or R. While the latter allow for fast prototyping and more interactive development, it has been shown for several applications that C or C++ implementations outperform plain Python or R implementations by factors between one and three orders of magnitude. In particular, in times of pandemics, the urgent need for swift estimation of the system's dynamics makes the use of large-scale models in these noncompiled languages prohibitive. In order to create synergies and close the gap between scientists from the application and the computational domain, binding libraries allow to execute entire, compiled software frameworks from languages such as Python. However, with iterative model development, extension, or addition of new models, implementing bindings can represent another relevant barrier to developers of the C or C++ backends. In this paper, we show how a large-scale software framework with a multitude of infectious disease models can be efficiently exposed to Python. In addition, we suggest a semi-automated way to generate interface code by analyzing generic C++ code through abstract syntax trees, enabling a modularized approach working with intermediate representations.

0.49

On the Reconstruction of Orthomodular Logic via Restricted Negation

cs.LO ·2026-10-29 ·Tomoaki Kawano, Ryo Kashima
published
Abstract · In Proceedings NCL'26, arXiv:2610.34489

Orthomodular logic has long been studied as a fundamental logical framework within quantum logic. In addition to its lattice-theoretic foundations, various formulations have been proposed that incorporate specialized implication connectives and more complex logical connectives. In this study, we reconstruct orthomodular logic by introducing a connective referred to as restricted negation, which has received comparatively little attention in the existing literature. We demonstrate that this connective provides a natural and expressive basis for formulating the core principles of orthomodular logic. Furthermore, as a main result, we construct a new proof system for orthomodular logic founded on this connective.

0.48

Paraclassical Logics with Two Negations

cs.LO ·2026-10-29 ·Pawel Pawlowski, Pietro Vigiani
publishedhot direction
Abstract · In Proceedings NCL'26, arXiv:2610.34489

We study three paraclassical propositional logics with two negations, CLoaN, CLouN, and CLauN, obtained by pairwise combining the one-negation systems CLuN, CLaN, and CLoN. CLuN's negation is glut-tolerant and paraconsistent, CLaN's negationis gap-tolerant and paracomplete, and CLoN's negation is both paraconsistent and paracomplete. We provide sound and complete two-valued non-deterministic semantics for each of the two-negations systems. We then focus on CLauN, the system containing the primitive dual pair of CLaN and CLuN negations. In this setting, we show deductive relationships between different negations expressible in CLauN; we show two diferent ways to recapture classical logic by means of translations, locally and globally; and we show a duality results for the multiple conclusion consequence relation of the logic.

0.48

Merging and Evolution of Data Bases: A Many-Logic Approach

cs.LO cs.DB ·2026-10-29 ·Manuel A. Martins, Abilio Rodrigues
published
Abstract · In Proceedings NCL'26, arXiv:2610.34489

A new approach to information-based logics using many-logic modal structures (MLMSs) was recently proposed by Martins et al. In such a setting, it is possible to have worlds with different underlying logics within the same frame, all anchored to a base lattice that contains the semantics of each logic as a down-complete sublattice. In this paper, we take the lattice L6, defined by the six-valued semantics of the logic of evidence and truth LET + K , as our base lattice. We also consider four-, three-, and two-valued formal systems associated with down-complete sublattices of L6. These logics are capable of representing paracomplete, paraconsistent, and classical contexts with six, four, three, and two informational scenarios, respectively. As applications of MLMSs, we will discuss the merging and temporal evolution of databases whose underlying logics are two-, three-, four-, and six-valued. We will also investigate inconsistency measures in this context.

0.47

Normalisation for a Neutral Free Logic with Definite Descriptions

cs.LO ·2026-10-29 ·Yaroslav Petrukhin
publishedtheory
Abstract · In Proceedings NCL'26, arXiv:2610.34489

In this paper, we present a natural deduction system for a strong Kleene-based neutral free logic with definite descriptions. The system combines a three-valued propositional basis with free quantification, identity, existence, and rules for definite descriptions. The main result is a normalisation theorem for a proof-theoretically equivalent restricted formulation of the system; we also establish a corresponding modified subformula property for its fragment.

0.47

Early Prediction of Cross-Service Implementation Scope: Evidence and Limits from Multiple Microservice Projects

cs.SE ·2026-10-29 ·Zhoumaocuo, Xiaozhou Li
code/datahot direction
Abstract

Microservice architectures aim to localize change, yet feature requests may require implementation across service boundaries. We investigate how well such downstream scope can be anticipated before implementation begins. Using issue--pull-request--commit traceability and implementation evidence corrected through targeted human adjudication, we reconstruct 866 repository-traceable feature requests from 13 open-source microservice projects and map their implementations to validated service boundaries. Cross-service changes occur in 101 requests (11.7%) and exhibit broader implementation footprints, particularly higher code churn, but no clear difference in lead time. Pre-implementation evidence provides limited cross-project predictive signal: the strongest leave-one-system-out model reaches a project-macro Macro-F1 of 0.484, but its advantage over the majority baseline is not significant after Holm correction and is sensitive to small evaluation folds. A local LLM offers no reliable improvement over classical models. Downstream workload is substantially harder to estimate, as learned models do not consistently outperform training-fold median baselines. A simple service-mention heuristic also performs comparably to learned scope triage. These findings show that early feature-request evidence is more useful for lightweight cross-service scope triage than for automated scope determination or precise workload estimation.

0.45

A Basic Modal Logic for the Abstract Treatment of Quantum Mixed States

cs.LO ·2026-10-29 ·Tomoaki Kawano
published
Abstract · In Proceedings NCL'26, arXiv:2610.34489

Quantum logic has developed in various directions since its inception. However, many existing logical approaches either focus only on pure states, or use Hilbert-space models in which the abstract relationship between pure states, mixed states, and probabilities is not made explicit. In this work, we introduce a model designed to handle quantum mixed states abstractly. Furthermore, we construct a proof system for the proposed logic and establish its completeness with respect to the intended semantics.

0.43

On Some Logics Intermediate Between S4.2 and S4.3 Defined by the Notion of Independence

cs.LO ·2026-10-29 ·Maciej A. Hałapacz
published
Abstract · In Proceedings NCL'26, arXiv:2610.34489

Hamkins and Loewe described modal logic S4.2 through the notion of sets of independent buttons - statements that are possibly necessary - and switches - necessarily contingent statements. In this article we investigate logics intermediate between S4.2 and S4.3 that are defined through a different notion of independence. We provide an axiomatic description of such logics, state some basic facts and prove various completeness and decidability results.

0.43

Cyclic Nonassociative Bilinear Logic With Additive Constants as Extension of Full Nonassociative Lambek Calculus

cs.LO ·2026-10-29 ·Paweł Płaczek, Cheng-Syuan Wan
publishedtheory
Abstract · In Proceedings NCL'26, arXiv:2610.34489

Cyclic Linear Logic (CyMALL, CyBL) is a compromise between Commutative Linear Logic (MALL) and Noncommutative Linear Logic (Bilinear Logic, BL). We do not admit commutativity of the tensor, but we still admit only one De Morgan negation. MALL and its variants are natural extensions of Full Lambek Calculus (FL). Abrusci proved that CyBL is not a conservative extension of FL, if we admit additive constants in these logics. In this paper we show that Cyclic Nonassociative Bilinear Logic (CyNBL) with both additive constants is a strongly conservative extension of Full Nonassociative Lambek Calculus with only one additive constant - the upper bound - and that this holds even in the presence of an arbitrary set of nonlogical axioms.

0.41

Do Open-Source Video Ad Clients Follow VAST 4.3? An Empirical Study of Player Conformance

cs.SE cs.MMcs.NI ·2026-10-29 ·Prashant Chaudhary
code/dataunusual
Abstract · 51 pages, 20 tables, 3 figures. Harness, test vectors and results: https://doi.org/10.5281/zenodo.23581408

Context. Video ads reach streaming and connected-TV apps through VAST documents, and companies count ads from the HTTP requests the player's client library sends. A tracking error can leave playback intact and surface later as disagreeing counts. Objective. To measure how far open-source VAST clients meet the player duties of VAST 4.3, whether verdicts depend on the runtime, what generated inputs add, and when deviations were introduced. Method. A catalog of 108 requirements was extracted from the specification. Six open-source clients in eight configurations ran against 69 crafted test vectors and the IAB's sample tags, under Node.js and in five browsers on three engines, with every request recorded. Grids, fuzzing and a sweep over 36 releases extend the catalog; Google's closed IMA SDK is a comparison. Results. The open-source clients have 78 findings and 3 to 21 must-level deviations per configuration. None of the tested open-source clients that follow redirect (Wrapper) chains fills the required macros in a Wrapper's request; the archived IMA 3.794.0 configuration tested does. One library's timeout throws an exception no caller can catch, ending a Node.js process. In a pod (ads played back to back), one player credits a finishing ad's completion to the next ad, fixed in release 5.3.0. A library can conform under jsdom and deviate in every tested Chrome- and Safari-family browser. Conclusions. The tested clients diverge on the duties other companies use to count and diagnose ads. A shared conformance suite run in real browsers could expose most graded deviations; the harness and vectors are released.

0.41

SKFitter: Symmetry resolved Slater Koster tight-binding models from first principles for s, p, d and f electrons, spin orbit coupling and magnetism

cs.MS cond-mat.mtrl-scimath-ph ·2026-10-29 ·Sundeep Kapila, Bhaskaran Muralidharan, Satadeep Bhattacharjee
code/data
Abstract · Open source software available at https://github.com/sundeep-kapila/skfitter

Constructing accurate and physically interpretable tight binding (TB) Hamiltonians from first principles calculations remains challenging, particularly for materials exhibiting complex orbital hybridization, spin orbit coupling (SOC), and magnetic order. SKFitter is an open-source Python framework for constructing and optimizing symmetry resolved Slater Koster TB models directly from density functional theory (DFT) calculations performed using Quantum ESPRESSO or VASP. The framework supports s to f orbitals, incorporates on site SOC and site resolved magnetic exchange interactions, and accommodates both orthogonal and nonorthogonal basis. A central feature is the symmetry resolved parameterization of orbital manifolds, which allows independent two centre hopping integrals between symmetry allowed orbital groups and across distinct neighbour shells. SKFitter also identifies missing orbital contributions and guide systematic basis set refinement and optimizes parameters through weights. The resulting Hamiltonian can be used to evaluate anomalous (AHC) and spin Hall conductivities (SHC), and topological invariants. We demonstrate the framework using eight representative materials spanning semimetals, semiconductors, transition metals, f-electron compounds, and magnetic systems. Symmetry resolved models achieve band energy errors of 9 to 41 meV within 1 eV of the Fermi level. The fitted Hamiltonians reproduce characteristic features of the SHC response of Pt, the SOC dependence of the AHC of Fe, and the momentum dependent altermagnetic spin splitting of MnTe without any altermagnet specific interaction. For EuMn2Bi2, projection guided orbital refinement reduces the near Fermi level fitting error from 120 to 32 meV. Together, these results establish SKFitter as a flexible and reproducible framework for developing symmetry consistent, physically interpretable TB models of complex materials.

0.40

Solving quantified constraint satisfaction problems with applications to control

cs.LO ·2026-10-29 ·Éric Goubault, Nuwan Herath Mudiyanselage, Sylvie Putot
published
Abstract

Many problems in control can be expressed as quantified constraints, such as robust forward and backward reachability, synthesis of controllers, solutions to motion planning problems. In this paper, we develop a method for computing inner and outer approximations of global solutions to very general quantified constraints, generalizing widely the state of the art on similar problems, usually restricted to at most one alternation of quantifiers or to local solutions. By global solutions, we mean characterizing accurately the entire set of solutions, within some prescribed bounds. For this, our approach uses a set inversion technique relying on oracles for set inclusion using local solutions to related quantified reachability problems. We exemplify the approach throughout the paper on different examples related to control problems.

0.40

Restricted Peirce's Law and Natural Deduction

cs.LO ·2026-10-29 ·Satoru Niki
publishedtheoryhot direction
Abstract · In Proceedings NCL'26, arXiv:2610.34489

The constructive logic JP' is known to be the weakest logic between intuitionistic and minimal logic that satisfies Glivenko's theorem. In spite of this natural and desirable formal property, the motivation for its notion of negation remains rather elusive: it can be characterised by a restriction of Peirce's law or by the double negation of ex falso quodlibet, neither of which is very intuitive. In order to have an alternative perspective on this negation, we study in this paper a couple of natural deduction systems for JP'. We first present a system that is normalisable and satisfies a restricted form of the subformula property. Then a modification of this system is introduced, whose rules satisfy the separation of connectives and the full subformula property. Afterwards we turn our attention to a typed lambda-calculus, with respect to which our natural deduction enjoys the Curry-Howard correspondence, followed by a concluding discussion.

0.40

PyReprism: A Shared Lexical Foundation for Polyglot Code Analysis

cs.SE ·2026-10-29 ·Daniel Ogenrwot, Jorge Gonzalo Delgado Cervantes, Richard Sserunjogi, John Businge
code/data
Abstract · 5 pages, 2 figures, Submitted to SANER 2027 Tool Demo Track

Empirical studies of polyglot repositories apply the same analyses to every language of a corpus, such as separating code from comments, computing metrics, normalizing identifiers, and screening for duplicated code. Assembling them from specialized tools introduces differences in language coverage, measurement conventions, and workflows. We present PYREPRISM, an open-source Python library and command-line tool that separates language-specific lexical processing from reusable analyses: each language declares its lexical surface, which yields a lossless stream of typed tokens, and every analysis is implemented once on that stream. Across 187 built-in languages, language-specific declarations have a median length of 16 lines, and no language overrides a shared analysis. Against a reference parser, comment and string classification reach F1 scores of 0.994 and 0.958 on 1,412 files in 12 languages that were never used to develop the tokenizer. In a study of a polyglot repository, one tool replaces three; with matched inputs and analysis operations, it takes 2.9s instead of 8.7s on one core. We also characterize where lexical analysis falls short, in complexity estimation and clone screening. Tool: https://github.com/unlv-evol/PyReprism Video: https://youtu.be/IXc-NP5t4GU

0.38

Exact Avoid-or-Compress Extrema for Dense Reachable Images: Defects Four and Five

cs.FL math.CO ·2026-10-29 ·Qichao Wang, Daoyu Dong
Abstract · 13 pages, 2 figures, 1 table. Computer-assisted upper bounds

Let $S$ be a reachable image of a strongly connected synchronizing automaton and $\varnothing\ne A\subsetneq S$. We determine the maximum shortest avoid-or-compress length over binary and arbitrary finite alphabets for $m=n-|S|=4$ and $m=5$. The exact values are $\min(n-|A|,11)$ and $\min(n-|A|,14)$, respectively. Uniform binary constructions attain both bounds. For the upper bounds, short-pair graphs and actual-image constraints reduce all parameters to finite necessary domains. Exhaustive checks exclude the domains, apart from a single relaxation eliminated by a multipartite overlap lemma. The largest calculation covers 3,206,988 domains.

0.38

When Static Reachability Says No: Checking Joern-Based Negative Verdicts for Python Against Execution

cs.SE cs.CR ·2026-10-29 ·Artur Zięba-Kozarzewski, Bartosz Grzegorzewski
benchmark
Abstract · 8 pages, 5 tables. Artifact: https://doi.org/10.5281/zenodo.23554742

A negative reachability verdict removes a finding from human review. We check such verdicts against execution. Our pipeline, SourceLens, maps Semgrep findings to a Python program's call graph, whose edges a query takes from the calls that Joern links in its code property graph. We ran two command-line programs from the PyCG macro-benchmark, in which the benchmark's manual call graph and Joern disagree, under coverage and recorded the calls between their functions. On these edges, 123 functions of asciinema are unreachable from the program's entry points. The project's own integration script executes 86 of them, and all our runs 113. In Sublist3r, 19 of 24 are executed. The latest Joern release gives the same verdicts. The missing calls go through values, such as classes and functions passed around, or have no call expression, as in with blocks and property reads. In asciinema, four missing edges make 50 of the 113 executed functions reachable. The manual call graphs also lack observed calls, in asciinema many of them property reads and with blocks, which the benchmark may not count as calls. A rule that turns negative verdicts into unknown recovers 62 of the 113 and 15 of the 19, mostly through dunder and private method names. It recovers 0 of the 5 security findings with a negative verdict, all in executed functions. Its selectivity cannot be measured, as only 10 and 5 negative verdicts were not executed, and under library entry-point models it degrades every negative verdict.

0.37

Soei: Structured Observability of Evidence and Intent for Multi-Agent AI Software Engineering

cs.SE cs.AIcs.MA ·2026-10-29 ·Mohammed Nayeem Khan, K S Sarun Kumar
code/data
Abstract · 27 pages, 19 figures, 16 tables. Preprint v1.1

AI coding agents now do long, multi-step and multi-session software work with little supervision. When a developer comes back to that work, the useful question is not what the model said but what was actually done, by whom, and whether it is finished. We study this in three parts. First, we verify Agenttrail, an open-source local observability tool for coding agents, against its source at commit e4ba2da: nine documented constants and mechanisms hold at file-and-line level, two event pipelines are never unified, and its Map service has an unauthenticated local mutation surface. Second, we run a prototype of our system on 45 real coding-agent sessions (102,233 records, 20,577 tool executions, 2,336 human-to-agent episodes). With a blind-audited classifier (precision 83.3% and 96.7%) we estimate that 23.4% (95% CI 17.1 to 35.7) of code-editing episodes end with no independent check after the last edit, and that 57.8% of to-dos the agent marked complete after editing had no proof in between. A 10-minute timing heuristic for session hand-offs recovers 6 of the 24 session pairs that content lineage links. 42% of sessions run longer than 24 hours, 48.1% of transcript bytes lie past a 2 MiB replay bound, and 1,326 records in 16 sessions contain credential-like strings. Third, we propose Soei (Structured Observability of Evidence and Intent): an acknowledgement-aware event schema, a hash-chained local ledger, a W3C PROV-aligned graph, calibrated inference for attribution and lineage, and a verification layer that keeps a claim of completion apart from proof. On the same corpus the ledger is 20% of raw size, holds no content or credentials, verifies at about 126k events/s, detects tampering and replays deterministically. We position the work against concurrent 2026 proposals and define autonomy levels with evidence gates and harness metrics baselined on our data.

0.33

On Łukasiewicz Topological Semantics

cs.LO ·2026-10-29 ·Muhammad Afaq Khan, Manuel A. Martins, Alexandre Madeira, Babar Zaman
published
Abstract · In Proceedings NCL'26, arXiv:2610.34489

In this paper, we propose fuzzy topological semantics for Łukasiewicz valued modal logic, based on the standard Łukasiewicz algebra. Within this setting, formulas are interpreted as fuzzy sets of truth values, while the modal operators are realized through fuzzy interior and closure operators. We define a Kripke-induced fuzzy Alexandroff topology and establish a precise correspondence between Łukasiewicz fuzzy Kripke semantics and its topological counterpart. The proposed topological semantics validates the S4 modal axioms that include reflexivity, transitivity, monotonicity, normality, and the necessitation of truth.

0.33

When Modernization Breaks Working Code: A Single-File Study of LLMs on Legacy PHP

cs.SE ·2026-10-29 ·Afridi Mahmud, Abid Hasan Khondaker, Sadif Ahmed, Md Nafiu Rahman, Rifat Shahriyar
Abstract · 6 pages, 5 figures, 5 tables. Preprint

Pasting a legacy file into an LLM and asking it to "modernize this" is an easy shortcut. But is the result better than leaving the code alone? In this paper, we study this single-file setting with five LLMs, each asked to modernize 100 WordPress 4.0 PHP files to PHP 8.3 while preserving behavior, with no project context or execution feedback. We evaluate every output with 630 hidden behavioral checks and 511 Rector-detected modernization findings. We compare against two baselines: the unchanged originals and the rule-based tool Rector. Often, the answer is no. Untouched, 83 files already pass every check; the best LLM leaves only 59. Each model breaks 32-51 working files while repairing at most 10, whereas Rector breaks 7 and repairs 6. Most of this damage is unnecessary. Only 37 findings (7 percent) require a change under PHP 8.3, yet in compatible files with no required change, the LLMs break 36-60 percent (Rector: 8 percent). The model with the highest modernization score is no more compatible than its closest competitors. Most regressed checks come from files that no longer load at all, often because of type declarations PHP itself rejects. These are errors that need no project context to avoid. Evaluations of LLM-based modernization should therefore measure against a no-change baseline and report modernization and behavioral preservation separately.

0.33

Tree Equivalence of Higher-Order Recursion Schemes Is Undecidable

cs.FL cs.LO ·2026-10-29 ·Mingkun Xiao, Yixuan Sun
theory
Abstract · 31 pages, 3 figures. Rocq artifact: https://doi.org/10.5281/zenodo.23734555

Higher-order recursion schemes (HORS) are finite, simply typed recursive programs that generate potentially infinite trees. Their monadic second-order model-checking problem is decidable. We prove that tree equivalence of deterministic HORS is undecidable and, more precisely, $Π^0_1$-complete under total computable many-one reductions. The lower bound holds for safe, productive schemes of order at most four over the fixed ranked alphabet $Σ_2=\{e:0,d:2\}$, consisting of one nullary and one binary symbol. Adapting the counter-guard witness of Boiret, Piorkowski, and Schmude, we generate all candidate executions of a two-counter machine by producing both continuations at every counter test. An integer weight is nonzero exactly on legal prefixes. Functions at fixed simple types represent the polynomial of growing degree. A shift recurrence updates them without storing a separate step counter, allowing the witness to be implemented by safe applicative HORS rules. At candidate halts, the intermediate schemes generate unary observation subtrees whose lengths encode the two natural-number components of the signed weight. Invalid candidates give matching observations; a legal halt produces a finite discrepancy. Fixed constructor contexts encode these intermediate trees over $Σ_2$. For the upper bound, a finite monotone semantics decides head convergence and supports an algorithm computing exact tree prefixes at any prescribed finite depth, including positions labelled $\bot$.

0.32

Hammering with Vampire in Lean

cs.LO cs.AI ·2026-10-29 ·Jonas Bodingbauer, Laura Kovács
theory
Abstract · 10 pages, 3 figures, 1 table. Tool demonstration paper

We present Vampire-Hammer, a hammer for the Lean interactive theorem prover that discharges goals using the Vampire automated theorem prover and reconstructs proofs checked by the Lean kernel. Vampire-Hammer uses the Lean-auto framework to process goals and premises before dispatching them to Vampire for proof search. Doing so, we rely on standard TPTP problem formats and annotated TSTP proofs that support proof reconstruction. With a 10-second budget on 196,083 eligible Mathlib theorems, Vampire-Hammer solves 21,158 problems not solved by other evaluated tools within the same budget.

0.32

New hierarchical matrices for fast kernel matrix-vector products with Applications to Gaussian processes and kernel ridge regression

cs.MS math.NAstat.CO ·2026-10-29 ·Ritesh Khan, Sivaram Ambikasaran
unusual
Abstract · 27 pages

This article presents two new multilevel hierarchical matrices to perform fast kernel matrix-vector products in $d$ dimensions $(d>1)$, namely $\mathcal{H}_{hw}^{\prime}$, $\mathcal{H}_{hw}^{\prime \prime}$. We introduce a novel hybrid weak admissibility condition to construct these hierarchical matrices. More precisely, at coarser levels of the tree (i.e., larger blocks), we use a weak admissibility condition, wherein \textbf{far-field} and \textbf{vertex-sharing} clusters are considered admissible, while at the finer levels (i.e., smaller blocks), the \textbf{adjacent} clusters are treated as admissible. The admissible blocks at coarser levels of $\mathcal{H}_{hw}^{\prime}$ and $\mathcal{H}_{hw}^{\prime \prime}$ are represented using semi-nested and fully nested bases, respectively. In both hierarchical matrices, the finer levels at which adjacent clusters are treated as admissible are constructed in a non-nested fashion. Thus, they can be viewed as a hybrid of $\mathcal{H}$ and $\mathcal{H}^2$ matrices. We use adaptive cross approximation (ACA) and nested cross approximation (NCA) to construct non-nested and nested bases, respectively. The fast matrix-vector products are performed using the proposed hierarchical matrix formats. To evaluate the performance and usefulness of the proposed hierarchical matrices, we conduct some numerical experiments. The numerical results show that they outperform their uniform admissibility counterparts in terms of storage and matrix-vector product cost, demonstrating the effectiveness of the proposed hybrid formats. We also show that the proposed formats are competitive with the standard (or strong) admissibility-based $\mathcal{H}^2$-matrices. Furthermore, we showcase some interesting applications of the proposed hierarchical matrices.

0.30

Step Recursion: Rational Strides and Exact Inclusion Orders over the Reals

cs.LO math.LO ·2026-10-29 ·Kirill Osipov
theory
Abstract

We study finite-terminating, address-only step recursion on the nonnegative reals, from zero, unit successor and projections. For reduced positive rational strides $s=a/q$ and $t=b/v$, we prove $\mathcal{C}s\subseteq\mathcal{C}t$ if and only if $q$ divides $v$, for both the bounded and unbounded algebras $\mathcal{C}\in{\mathcal{B},\mathcal{U}}$. The bounded compiler preserves earlier global majorants on every auxiliary input. Mixed rational stride sets are classified by least common denominators and supernatural denominators. Primitive-recursive translated-coordinate grid codes characterize the unbounded algebras and yield $\mathcal{U}{1/q}\cap\mathcal{U}{1/v}=\mathcal{U}{1/\gcd(q,v)}$. For bounded intersections, we prove an exact reduction: all corresponding identities hold if and only if the single identity $\mathcal{B}{1/2}\cap\mathcal{B}_{1/3}=\mathcal{B}_1$ holds, equivalently if the Boolean parts of the integer successor and addition algebras under scalar bounded recursion coincide. A Boolean reconstruction reduces general functions to predicates; a clipped fine-grid simulation with a frozen global resource bound identifies the union of predicate precision classes with the addition level. These reductions do not decide the resulting predicate equality, so literal bounded intersections remain unresolved. We also prove that unbounded-program equality is $Π^0_1$-complete and extensional bounded-algebra membership is $Σ^0_2$-complete. The model has no free real arithmetic or Zeno limits; semantic majorant validity is distinguished from syntax transformation.

0.30

STORY2DIFF: A Dataset Linking Real-World User Stories to Code Changes

cs.SE ·2026-10-29 ·Benedetta Donato, Daniele Franchini, Alessandro Greco, Oliviero Riganelli, Daniela Micucci +1 more
datasetcode/data
Abstract · Submitted to the ICSE 2027 Demonstrations and Data Showcase track

Research on requirements-to-code tasks needs examples in which a genuine requirement is paired with the change that implemented it. Existing user-story collections, however, largely contain isolated or synthetic text. To address this limitation, we present STORY2DIFF, a dataset of 7,785 English user stories linked to merged pull requests from January 2017 to July 2026. The dataset is mined from a set of 982 public Java and Python repositories. Each record in the dataset combines the original issue, extracted Connextra elements, the associated pull request and commit metadata, a file-level diff capturing the code changes introduced by the associated pull request, and the affected files before and after the change. The tool we built to mine STORY2DIFF can be used to repeat the automatic extraction of the records, using different parameters (e.g., a different timeframe). The dataset can support a variety of tasks, including requirement discovery, traceability, change- impact estimation, and requirements-aware code generation. Dataset: https://github.com/riganelli/user-story-mining-pipeline/ releases/tag/V4.0.1. Video: https://youtu.be/qOhXUeyeiTQ.

0.29

A Quantitative Interpretation of Useful Call-by-Value

cs.LO ·2026-10-29 ·Pablo Barenbaum, Delia Kesner, Mariana Milicich
Abstract

In this work, we present the first semantic interpretation of useful call-by-value evaluation through a non-idempotent intersection type system called U. One first contribution of this paper is a characterisation of termination of useful call-by-value evaluation via U-typability. That is, a term is typable in system U if and only if it terminates in the useful call-by-value strategy. As a second contribution, we show that system U provides a quantitative interpretation for useful call-by-value evaluation, offering exact step-count information for program evaluation. Even though the specification of the operational semantics of useful evaluation is highly complex, system U is notably simple. As far as we know, system U is one of the scarce quantitative type systems capturing exactly the substitution step-count for variables and abstractions in an open call-by-value strategy. Our third contribution is a termination-equivalence result between standard call-by-value, which may perform erasing reductions, and useful call-by-value, which is non-erasing. To establish this result, we introduce a non-erasing restriction of standard call-by-value, a refinement that serves as an intermediate calculus bridging standard and useful call-by-value. We show that all three calculi have the same termination behaviour, and hence induce the same observational equivalence.

0.29

Fast and Accurate Hydro Unit Commitment

cs.MS math.OCeess.SY ·2026-10-29 ·Shaked Regev, Evan J. R. Brody
theory
Abstract · 5 pages (Conference limit), 5 figures

We introduce a fast novel algorithm to commit hydro-generators. We show that our algorithm can commit 56 generators for 200 days 10,000 times faster than Hexaly, the leading commercial solver for scheduling problems, without sacrificing solution quality. Our method can commit 1792 generators for an entire year in approximately 20 seconds on commodity hardware. Hexaly is unable to solve problems of this size in reasonable time. We attain this runtime improvement by introducing a greedy-heuristic tailored for hydro units. We provide a direction for a possible proof that the error our solver makes decreases with the problem size. This is supported by our experimental results. We prove that our solver's runtime scales log-linearly in the number of generators and time periods.

0.27

Semantic Path Orders are Generalized Weighted Path Orders

cs.LO ·2026-10-29 ·Teppei Saito, Nao Hirokawa
theory
Abstract · 38 pages

Monotonic semantic path orders (MSPOs) are powerful reduction orders, but their effective use in theorem proving and termination analysis raises three issues: efficient comparison of terms, ground totality, and a reduction-pair formulation. Inspired by weighted path orders (WPOs), we introduce generalized weighted path orders (GWPOs) which encompass both WPOs and MSPOs. GWPOs and MSPOs coincide as classes of reduction orders, thereby yielding efficient and ground-total subclasses of MSPOs. We further formulate the reduction-pair version of the GWPO for the dependency pair method, from which a reduction-pair version of the MSPO is obtained. In contrast to the reduction-order setting, the correspondence between GWPOs and MSPOs no longer fully carries over, establishing the GWPO as a framework of independent interest. Effectiveness of the resulting reduction pairs is assessed by standard termination benchmarks.

0.27

Lost in the Latent Space: How Code Embeddings Represent Functional Intent and Implementation Fidelity

cs.SE cs.CLcs.IR ·2026-10-29 ·Lu Sun, Yuheng Huang, Lei Ma, Zhijie Wang
Abstract

Code embedding models have become a key building block for many software engineering applications, yet their internal representations remain poorly understood. Existing evaluations tend to focus on downstream task performance, offering limited insight into whether a single embedding can support both functional abstraction and implementation-sensitive analysis. To address this gap, we present a large-scale empirical study of 14 state-of-the-art code embedding models. We evaluate their representational behavior through three complementary analyses: supervised probing for functional abstraction and implementation fidelity, zero-shot retrieval experiments to assess geometric similarity, and controlled perturbation experiments that measure embedding responses to semantics-preserving code transformations. The results reveal a consistent abstraction-fidelity tension: models that better preserve stylistic and structural signals tend to do so at the expense of semantic generalization, while models with similar retrieval performance can nevertheless respond very differently to the same implementation changes. Overall, our findings suggest that a single universal code representation remains out of reach, motivating more representation-aware evaluation, selection, and deployment of code embedding models in software engineering.

0.26

crush: Extensible Proof Automation for Lean

cs.LO cs.PL ·2026-10-29 ·Mike He, Dexin Zhang, Aarti Gupta
Abstract · 35 pages including appendices, 13 figures

Lean's rich expressiveness supports higher-order functions, inductive datatypes, and user-defined abstractions, making it challenging to translate proof obligations into first-order Satisfiability Modulo Theory (SMT) queries and reconstruct kernel checked proofs. We present crush, an extensible SMT-backed proof automation framework for Lean. To bridge the gap between Lean's higher-order constructs and the solver's first-order logic, crush uses defunctionalization to encode function values and their applications. We formalize this translation for a supported higher-order fragment and prove its soundness. Moreover, crush provides metaprogramming interfaces for extending both Lean-to-SMT translation and the reconstruction of Lean proofs from solver certificates without modifying the translation pipeline. We implement crush as both a tactic and a backend for other third-party automation systems. We show that across four such use cases that target verification and general proof automation, crush discharges more obligations than the competing backends and reconstructs more kernel-checked proofs than the baselines.

0.26

UPPAAi: Towards LLM-Assisted Modeling, Verification, Analysis and Synthesis

cs.SE cs.AIcs.LO ·2026-10-29 ·Peter Dolog, Kim G. Larsen, Marius Mikučionis, Frederik F. K. Rasmussen
Abstract · Submitted to IFS/ETAPS 2027

For more than three decades, UPPAAL has supported the modelling and analysis of real-time and cyber-physical systems. Yet building a suitable model, formalising the intended properties, choosing analysis techniques and interpreting counterexamples remain iterative tasks that demand substantial expertise. Large language models (LLMs) can offer plausible explanations and even solutions, but they are not necessarily correct. We present UPPAAI, an extension of UPPAAL in which users construct, explain, debug and specify models in dialogue with an LLM. The guiding principle is that the user requests, the LLM proposes, explains and orchestrates, and UPPAAL verifies. UPPAAI exposes the UPPAAL editor, queries and verifier through a Tools API. We report on seven case studies, from explaining to a newcomer to learning a controller, covering timed automata, stochastic hybrid automata and strategy learning. LLMs proved to be extremely helpful in not just generating solutions but also pro-actively suggesting relevant UPPAAL features it was not asked about.

0.26

Omission-Safe Conditional Contracts for Evidence-Closed CI Qualification

cs.SE ·2026-10-29 ·Zhiheng Zhang
Abstract · 11 pages, 1 figure, 4 tables

Continuous-integration evidence is conditional: a workflow determines which jobs should run, while an archive records which jobs remain observable. Reconstructing requirements from those observations lets an omission erase its own obligation. We present a qualification method that compiles a pinned workflow into conditional job requirements independently of target presence, retains a fixed capture inventory, and checks source-complete derivation certificates with a separate implementation. For a fixed source, context, scope and binding universe, with sound obligations and applicability judgments that preserve established truth under information extension, withdrawing observation facts cannot create authorization. The implementation handles static matrices, dependency guards and bounded pinned reusable workflows, preserving unknown applicability and separating tolerated outcomes from evidence coverage. Both implementations match the 45 hand-specified authorization controls, reject 180 source-certificate attacks, and exhibit no authorization gain over 704 distinct proper development erasures. An explicitly witnessed 12-job platform run qualifies in both implementations; all 4,095 proper payload-subset erasures block authorization. A frozen collection spans 240 runs in eight workflow families across six projects. All 160 evaluation source certificates independently validate, but unavailable historical execution-source identities leave actual qualification open. Initial gate diagnoses agree on 155/160 runs; a preserved unknown-propagation repair yields 160/160 agreement in a retrospective rerun. The study establishes bounded source extraction and omission resistance, rather than natural-fault prevalence or producer authenticity.

0.26

Countable Nondeterminism in Weighted Programming with an Application to Mechanized Foundations for Probabilistic Program Verification

cs.PL cs.LO ·2026-10-29 ·Sebastian Körner, Oliver Emil Bøving, Christoph Matheja
theory
Abstract

Weighted programming is a framework by Batz et al. for describing mathematical models or optimization problems as imperative programs parametrized over semirings and then analyzing them with program verification techniques like weakest precondition calculi. Applications include reasoning about expected outcomes of probabilistic programs and competitive analysis of online algorithms. We extend weighted programming with (1) countable nondeterminism and (2) user-defined additive rewards, where programs are parametrized over modules instead of semirings. The resulting framework enables reasoning about expected resource consumption of probabilistic programs and about optimal scores of combinatorial games. A technical challenge is that wp-calculi for weighted programs with (1) and (2) over arbitrary modules can be unsound, because it is well-known that continuity properties commonly used in program semantics break in the presence of countable nondeterminism. We study conditions under which a wp-style transformer for reasoning about weighted programs is sound with respect to an operational ground truth defined in terms of weighted decision processes (WDPs) - transition systems that generalize Markov decision processes by drawing transition weights and state rewards from a module. Based on those conditions, we show that a range of concrete instances of the weighted programming framework enjoy sound weakest precondition reasoning. We demonstrate the usefulness of our wp-semantics through case studies for different modules. We mechanized our extended weighted programming framework and our main soundness results in the interactive theorem prover Lean. Instantiating these results for probabilistic programming yields, to the best of our knowledge, the first mechanized proof of a fixed point characterization of minimal expected rewards in Markov decision processes with countable nondeterminism.

0.25

Integer-Quantized Hierarchical Matrices with Adaptive Bit Allocation

cs.MS math.NAcs.PF ·2026-10-29 ·Erin Carson, Xinye Chen, Ritesh Khan
Abstract · 37 pages

We present an adaptive bitwidth allocation method for quantizing low-rank factors in hierarchical matrices in order to decrease storage while meeting a given constraint on their reconstruction error. For the bitwidth allocation we formulate a discrete storage minimization problem and show that an optimal assignment can give larger singular values at least as many bits as smaller ones. We further develop a greedy algorithm that meets the specified max-norm quantization error budget. We investigate a hierarchical block partition in which all of the non-diagonal blocks are represented by quantized low-rank factors and only dense blocks remain dense; in this way a global reconstruction bound related to $η_k$ and $R_k$ is obtained, with the former and latter being bounds on the truncation and reconstruction arithmetic errors, respectively. Extensive simulations, carried out with kernel and BEM matrices under block-relative budgets, indicate that the proposed method achieves a reduction in storage of up to $44.3\times$ when compared to storing the complete matrix densely in FP64, and a reduction of up to $5.70\times$ when compared to the FP64 low-rank factors within the same partition, together with measured reconstruction errors that remained below the maximum norm bounds in all the hierarchical tests.

0.24

Program Bindings Inside Code Models

cs.SE ·2026-10-29 ·Prateek Kumar Rajput, Yewei Song, Jacques Klein, Tegawendé F. Bissyandé
Abstract · 22 pages, 2 Figures, 4 Tables

Code language models complete, review, and repair programs without exposing the program structure encoded in their hidden states. In this paper, we test whether the hidden states of a code model can be described and edited using the information a compiler front end computes. For each token of a Python program, standard analyzers provide its position in the abstract syntax tree, its scope, its definition-use links, and its syntactic category. We fit a decomposition that expresses every hidden state of a frozen model as a sum of one term per token, each combining the token with this structural label, and we replace the model's own states with the decomposition at every layer while it predicts the output of a small program. If the model still answers correctly, the decomposition preserves the information the model uses. We apply this test to 14 open models, seven of them trained for code, and compare against a baseline decomposition of the same size built from token positions. The static-analysis decomposition preserves the model's answers better than the positional one in 10 of the 14 models, including six of the seven code models, and deleting the definition-use links from it removes most of that advantage on the two models tested. Editing the decomposition, for instance rebinding the argument of a call, makes StarCoder2-15B produce the answer of the edited program in 90 of 100 cases, compared with 19 for the positional baseline. The Qwen2.5 family is the exception on several measures. For most of the models tested, standard static analysis therefore provides both a description of what the hidden states encode about a program and a way to change it without retraining.

0.24

A Hierarchy of Composition Procedures for Logic-Based Workflow Patterns

cs.LO ·2026-10-29 ·Radosław Klimek
Abstract

Logic-based workflow patterns capture recurring behavioural structures together with reusable temporal specifications. Their composition raises distinct problems as workflow structures become richer, including instantiation, nesting, conditional execution, repeated occurrences, and traceability of generated formulae. This paper proposes a hierarchy of Pattern Composition ($ΠC$) procedures, from $ΠC_0$ to $ΠC_4$, that organizes these requirements as progressively richer compositional capabilities. The hierarchy separates compositional complexity from the expressive power of the temporal-logic fragment, represented by $TL_j$, yielding a framework $ΠC_i[TL_j]$ for developing logical workflow specifications. We develop activation-aware Pattern Composition ($ΠC_2$), which distinguishes the structural presence of a nested pattern from its execution in a particular workflow context. We formalise its recursive procedure, analyse its properties, and define an evaluation protocol for conditional activation, conservativity under mandatory composition, and specification growth. The hierarchy also characterizes selective occurrence qualification for iterative patterns ($ΠC_3$) and provenance for tracing generated formulae to their sources ($ΠC_4$). Finally, we outline $ΠC_{\mathrm{Assist}}$, an orthogonal mode in which AI proposes and refines pattern expressions subject to the hierarchy's compositional checks.

0.24

Deciding two-variable logic with global and local Presburger quantification

cs.LO math.LO ·2026-10-29 ·Michael Benedikt, Chia-Hsuan Lu, Alex Perry, Tony Tan
Abstract

We look at a two-variable logic over finite graphs that allows Presburger quantifiers: linear arithmetic inequalities between the number of elements that satisfy subformulas. The prior literature shows that closing two-variable logic under Presburger quantification leads to undecidability, while closing under \emph{local quantification} -- the counts are always over neighbors of a given node -- preserves decidability. Here we show decidability of satisfiability for an intermediate logic that allows both local counts and global counts, but where the linear constraints do not mix the two. The proof requires a significant extension of the prior machinery developed for two-variable logic with counting. As a corollary we obtain decidability of satisfiability for an important family of Graph Neural Networks (GNNs). GNNs are a popular graph learning formalism. For example, in a variation of GNNs considered in the past literature, local aggregation is used in each layer to create modified values at each vertex of a graph, and then global aggregation is used in a final layer to obtain a single numerical value for the graph as a whole. Our logic is motivated by the fact that mixing of local counting and global aggregation in GNNs has been shown to lead to undecidability. Our results here show that if global and local aggregation are separated, one regains decidability. Decidability of other static analysis problems, like equivalence of GNNs, follows from the decidability of satisfiability.

0.23

Step Recursion: Exact Membership Spectra, Mixed Strides, and Collapse

cs.LO math.LO ·2026-10-29 ·Kirill Osipov
theory
Abstract

For a strictly increasing $G:\mathbb{N}\to\mathbb{N}$ with $G(0)\geq 1$, we study minimal-basis function algebras generated by recursion along generalized inverses of its iterates. Bounded recursion requires an earlier pointwise majorant; unbounded recursion has no such requirement. If the orbit $G^{[t]}(0)$ has no polynomial upper bound, we obtain exact fixed-function membership spectra, reverse-divisibility inclusion, and indexed intersections that need not be single-stride algebras. Every finite divisibility-downset is realized in both regimes, and mixed strides admit fixed functions expressing any finite monotone Boolean condition on their additive stride monoid. For every generator, the realized bounded spectra form a bounded meet-semilattice: a construction sharing a bounding input coordinate realizes finite intersections without an unrestricted selector. Polynomial orbit growth is equivalent to definability of each of ordinary predecessor, unary doubling, and addition in any fixed unbounded stride algebra. In this case the algebra is ordinary primitive recursion relative to its inverse map, giving full collapse for primitive-recursive $G$. Ordinary predecessor belongs to a bounded stride algebra exactly when the orbit is polynomially bounded and $G$ has bounded successive differences; equivalently, ordinary bounded recursion embeds. A locality obstruction explains the additional bounded requirement, and two primitive-recursive generators with quadratic orbits have opposite bounded collapse behavior.

0.23

On Deciding the Constraint Loop Problem in Dimension Two

cs.LO ·2026-10-29 ·Daria Goliasova, Toghrul Karimov
Abstract

Recently, Guilmant, Lefaucheux, Ouaknine, and Worrell [2] proved decidability of the termination problem for linear constraint loops over two real-valued variables, thus settling a wellknown open problem. Their algorithm is based on a theoretical result that such a loop does not terminate if and only if it has a witness of an infinite run of a specific shape. By analysing the structure of the witness carefully, we show that termination such loops can be decided in polynomial time, improving on the polynomial space bound argued by Guilmant et al. We also present an efficient implementation of the polynomial-time algorithm based on the satisfiability modulo theories (SMT) solver Z3.

0.23

An Empirical Study of Gas Optimization Practices in Solidity Smart Contracts

cs.SE ·2026-10-29 ·Francesco Salzano, Paolo Martino, Simone Scalabrino, Rocco Oliveto, Remo Pareschi
code/data
Abstract

In the Ethereum ecosystem, smart contract execution is governed by a gas model where every computational operation incurs a direct economic cost. Consequently, gas efficiency is a critical requirement for the economic sustainability and widespread adoption of decentralized applications. Despite the importance of cost-effective development, there remains a significant gap between theoretical optimization patterns and the actual practices applied by developers in real-world projects. This paper presents an empirical study investigating gas optimization practices in open-source Solidity repositories created on GitHub between 2018 and 2025. Adopting a Mining Software Repositories approach, we analyzed a dataset of gas-related commits identified through a multi-stage pipeline involving NLP-based filtering and manual qualitative inspection. Based on the analysis of 400 representative commits, we derive an empirical taxonomy of gas optimization strategies consisting of 8 macro-categories and 36 subcategories. Our findings reveal that optimization efforts span multiple levels of abstraction, from low-level arithmetic refinements and bytecode-size reductions to broader architectural changes such as logic internalization and external call reduction. Furthermore, we observe that developers often manage complex trade-offs between execution frequency, safety checks, and computational costs, highlighting that effective optimization is inherently context-dependent. This study provides a practical knowledge base for the design of future developer support tools and advances the understanding of cost-aware software engineering in the blockchain domain.

0.22

Learning Erasing Equal-Length Relational Pattern Languages in the Limit from Positive Data over Binary Alphabets

cs.LO cs.FLcs.LG ·2026-10-29 ·S. Mahmoud Mousawi, Sandra Zilles
Abstract

Relational pattern languages generalize Angluin's pattern languages by requiring the strings substituted for repeated occurrences of a variable to stand in a given relation, rather than to be identical. We study erasing pattern languages under the equal-length relation, where all occurrences of a variable must be replaced by strings of the same length. Over alphabets with at least three letters, this class is known to be learnable in the limit from positive data. We show that over a binary alphabet it is not, which resolves an open question from earlier work. Our proof reveals that non-learnability already holds for the restricted subclass of languages generated by patterns that have exactly three maximal substrings of variables. We then show that patterns with at most two maximal substrings of variables are learnable in the limit from positive data for every alphabet of size at least two. If the target pattern is known to have at most one maximal substring of variables, this can even be achieved by an iterative learner, which updates its hypothesis using only the current example and its previous hypothesis.

0.22

QSoftCore: A Conceptual Foundation for Quantum Software Engineering

cs.SE quant-ph ·2026-10-29 ·Tao Yue, Yuechen Li, Man Zhang
Abstract

Quantum software engineering (QSE) requires understanding application requirements, quantum programs and algorithms, quantum computation paradigms, executions, and deployment. However, relevant quantum computing (QC) concepts are often defined across physical, mathematical, and paradigm-specific levels of abstraction, making it difficult for the QSE community and hence proposed methods to refer to them consistently. To this end, we present QSoftCore, a conceptual model that reorganizes established QC concepts from a software engineering perspective and connects quantum applications, computational problems, algorithms, configurations, programs, and paradigm-specific computation specifications with state evolution, execution results, and backends. Our assessment against QSE literature across key software engineering activities indicates that QSoftCore provides shared QC referents for quantum computations, executions, deployment configurations, backends, and measurement evidence across the QSE lifecycle.

0.20

Functorial Analysis of Group-Alphabet Cellular Automata

cs.FL math.GRmath.COmath.CT+1 ·2026-10-29 ·François Huard, Nathan Dion Larivière
Abstract

Let $G$ be a group and let $n\in\mathbb N$. We consider one-dimensional cellular automata of length $n$ over the alphabet $G$, with periodic boundary conditions and with local rules $φ$ given by words. We construct a functor $F$ from $\mathbf{Grp}$, the category of groups (or any similar algebraic structure), to the category of pointed digraphs, and show its faithfulness. We introduce the image category $\mathcal C_{n,φ}$ of state transition graphs of all such group-alphabet cellular automata, and show $F$ is an equivalence onto this category. Consequently we investigate the translation of categorical structures in $\mathbf{Grp}$ in terms of automata, and give examples of how standard methods of homological algebra may now be applied to automata. Additionally, a graph-theoretical construction of the product in $\mathcal C_{n,φ}$ is given, together with a formula for cycle lengths and multiplicities in this product. When $G$ is abelian, further structural properties of state transition graphs follow, and we contrast these results with non-abelian counterexamples.

0.19

Reachability in Five-Dimensional Vector Addition Systems Is PSPACE-Complete

cs.FL cs.CC ·2026-10-29 ·Weijun Chen, Bo Fu, Yuxi Fu
theory
Abstract · 69 pages,1 figure

Vector addition systems (VAS) are a fundamental model of concurrent computation in which transitions update nonnegative integer counters by adding integer vectors, provided all counters remain nonnegative. Closely related to Petri nets, VAS are the single-state counterpart of vector addition systems with states (VASS), which use finitely many states to restrict the available transitions. A classical simulation of $d$-VASS by $(d+3)$-VAS connects their reachability problems, but the complexity-theoretic picture of fixed-dimensional VAS remains largely unclear. While reachability in $2$-VASS is PSPACE-complete under binary encoding, the best known upper bound for $5$-VAS was only elementary. We prove that every reachable $5$-VAS instance has a reachability witness of length polynomial in its unary size. Consequently, $5$-VAS reachability is NL-complete under unary encoding and PSPACE-complete under binary encoding. We also establish a polynomial factorization of the reachability relation of every $d$-VAS with $d \ge 3$. For $5$-VAS, this strengthens Hopcroft and Pansiot's effective semilinearity of reachability sets to the full reachability relation. For every fixed $d\ge3$, it also yields an $\mathsf F_d$ upper bound for reachability in $(d+3)$-VAS.

0.16

Automated Verification of Hamiltonian Simulation by Quantum Signal Processing

cs.PL cs.LOquant-ph ·2026-10-29 ·Min Lai, Yu-Fang Chen, Shih-Han Hung
Abstract

We present, to the best of our knowledge, the first fully automated verification framework for QSP-based Hamiltonian simulation. Our framework exploits the structure of the simulation algorithm: Hamiltonians are given as weighted sums of Pauli strings, embedded into larger unitaries through block encodings, and transformed into time-evolution operators by quantum signal processing (QSP). Using a symbolic Pauli-sum semantics for ancilla-indexed circuit blocks, our verifier proves that the distinguished all-zero ancilla block is proportional to the target Hamiltonian and infers the subnormalization factor automatically. It then verifies the QSP layer by extracting the implemented polynomial transformation and certifying its approximation error. This structure-aware abstraction is essential for scalability. Composing our verified block encoding and QSP wrapper yields an estimated macro-expanded size of about 1.52 million gates. The verifier certifies the block encoding once and checks the QSP layer through its abstract interface, without expanding the oracle calls. Our implementation verifies block encodings of structured Pauli-sum Hamiltonians with up to 256 system qubits and verifies QSP Hamiltonian-simulation instances by extracting their implemented polynomials and checking the resulting approximation bounds. These results show that fully automated verification can handle common Hamiltonian-simulation circuits at substantial scale when it exploits their algorithmic structure.

0.14

Well-Specified and Well-Tested? Assessing SWE-bench-Style Datasets with a Self-Evolving Agent

cs.SE cs.AIcs.CL ·2026-10-29 ·Zhonghao Jiang, Xuyang Zheng, Yanfu Yan, Zhongxin Liu, Lingfeng Bao +1 more
benchmark
Abstract

Large language models (LLMs) have shown strong performance on software engineering (SE) tasks, driving the need for high-quality datasets for reliable training and evaluation. SWE-bench-style datasets (SWE datasets) are widely used to evaluate LLM-based coding agents on real-world software issues, where task clarity and test coverage directly affect whether instances are solvable and verifiable. However, quality issues persist in existing SWE datasets, and manual annotation is costly and labor-intensive, motivating LLM-based annotation. Existing approaches typically rely on static, overly general rubrics, limiting their ability to handle diverse and nuanced scenarios and often yielding high false positive rates. In contrast, human annotators draw on accumulated experience to make context-aware judgments. Inspired by this, we propose ExLabel, an LLM-based, self-evolving, experience-guided assistant for task clarity and test coverage assessment in SWE datasets. ExLabel maintains an evolving experience bank that distills cross-instance knowledge into stable yet adaptive judgment criteria and reuses relevant experience to guide the assessment of new instances. For systematic evaluation, we construct SWE-TQ, a new benchmark spanning eight programming languages, by cleaning human annotations from SWE-bench and Multi-SWE-bench. Experiments with three backbone LLMs show that ExLabel achieves the best F1 across all settings, with relative improvements of 44.2%/49.9% and 24.0%/18.4% over the best baseline for task clarity and test coverage assessment on the Python/Multilingual subsets, respectively. Moreover, a coding model fine-tuned on SWE-rebench data filtered by ExLabel resolves 66.2% of issues on SWE-bench Verified, outperforming the model trained on data filtered by the best baseline by 14.5%.

0.13

When Requirements Interact: Scaling Long-Horizon Coding via Interdependent Specifications

cs.SE cs.CL ·2026-10-29 ·Yishuo Cai, Minzheng Wang, Liang Zhao, Linghao Zhang, Lei Li +3 more
benchmark
Abstract

To succeed in real-world long-horizon software development, coding agents must satisfy multi-dimensional task specifications, such as interface compatibility and error handling, and do so without breaking existing dependencies. However, current works lack specification in both quantity and diversity, often emphasizing primary goals while ignoring complex constraints. Synthesizing tasks to bridge this gap presents two challenges: naively adding specifications risks reducing tasks to trivial checklists rather than compelling joint reasoning, and verifying these subtle interactions risks overlooking hidden flaws or rejecting valid alternatives. To address these challenges, we introduce CodeSpecWeave, a framework for synthesizing tasks with rich, interdependent specifications. Grounded in real developer seeds, a synthesis agent iteratively develops both coupled task requirements and executable tests, while a review agent audits solution trajectories from multiple models to guarantee test reliability. Using this pipeline, we construct 4,670 training tasks across 2,865 repositories. Applying Reinforcement learning on these tasks significantly improves performance over the base model on DeepSWE-v1.1, largely outperforming existing baselines. Furthermore, we curate CodeSpecWeave-Bench, a rigorous benchmark comprising 100 tasks and 1,044 specifications, which clearly separates frontier models and reveals significant headroom in handling interdependent engineering constraints.

0.07

ExploreRepair: Evidence-Specialized Agents for Vulnerability Repair under Information Constraints

cs.SE cs.CR ·2026-10-29 ·Duc Manh Tran, Ratnadira Widyasari, Ivana Clairine Irsan, Huihui Huang, Ting Zhang +3 more
benchmark
Abstract

Automated vulnerability repair must often operate with incomplete vulnerability information: developers may lack an accurate vulnerability classification or a provided proof-of-vulnerability (PoV) test, yet still need to produce a repair. Without detailed information, a key challenge is to explore multiple plausible interpretations of the vulnerability before committing to a repair. To address this challenge, we propose ExploreRepair, a vulnerability repair framework that reconciles the vulnerability observations and repair implementations from a team of exploration agents. We first collect these observations and repair implementations from three agents, each tasked with exploring a specific dimension of the vulnerability. Then, a Lesson Aggregator and a Patch Aggregator independently synthesize repairs from these observations and concrete implementations before a final Aggregator reconciles their proposals without benchmark feedback. On PatchEval, ExploreRepair achieves repair success rates of 39.6% and 71.6% with GPT-5.6-Luna and DeepSeek-V4.1-Flash, respectively, exceeding EvoRepair, a state-of-the-art vulnerability repair approach, by 16.0 and 25.4 percentage points. On VulnLoc+, it achieves success rates of 51.4% and 62.9%, respectively, exceeding LoopRepair's variant without PoV tests by 20.0 and 17.1 percentage points.

0.04

Theory & Algorithms 80

Diameter-two graphs whose mutual-visibility chromatic number equals their chromatic number

cs.DM math.CO ·2026-10-29 ·Siddhesh Chaubal
Abstract · 8 pages, 3 figures. Submitted to Aequationes Mathematicae

A set $S$ of vertices of a graph $G$ is a mutual-visibility set if every pair of vertices of $S$ is joined by a shortest path whose internal vertices avoid $S$, and the mutual-visibility chromatic number $χ_μ(G)$ is the least number of mutual-visibility sets into which $V(G)$ can be partitioned. Klavžar, Kuziak, Valenzuela Tripodoro and Yero (2025) proved that $χ_μ(G)\leχ(G)$ whenever $G$ has diameter two, observed that equality holds for complete bipartite graphs, and asked whether there are any other diameter-two graphs attaining equality. We answer this question in the affirmative. We construct a graph $G_{11}$ of order 11 and diameter two with $χ_μ(G_{11})=χ(G_{11})=3$, and an infinite family of such graphs obtained from $G_{11}$ by replacing one vertex with pairwise non-adjacent copies of it.

0.56

Faster #PPM from faster #3SAT

cs.DS cs.CC ·2026-10-29 ·László Kozma
Abstract

The counting variant (#PPM) of the permutation pattern matching (PPM) problem asks for the number of occurrences of a permutation pattern in a host permutation of length $n$. The problem admits faster algorithms than the trivial $O^*(2^n)$ time enumeration; the best known running time, for both PPM and #PPM, is $O(1.415^n)$ (Gawrychowski and Rzepecki). It is also known that PPM is ETH-hard: an algorithm with running time $2^{o(n/\log{n})}$ would imply a subexponential algorithm for 3SAT (Jelínek, Opler, and Pekárek). We give a reduction in the opposite direction, showing that an algorithm for counting satisfying assignments of a 3-CNF formula (#3SAT) in time $2^{O({n}/{f(n)})}$, where $n$ is the number of variables, yields an algorithm for #PPM with running time $2^{O({n \sqrt{\log{f(n)}/{f(n)}}})}$. In particular, the recently claimed refutation of ETH by OpenAI implies a #3SAT algorithm with an improvement factor $f(n) \in Ω(\sqrt{\log{n}}/\log\log{n})$ in the exponent; this yields, through our reduction, a subexponential algorithm for #PPM.

0.56

A Single-Exponential Time Algorithm for the Maximum-Cardinality Weakly Stable Matching Problem with Ties and Incomplete Lists

cs.DS ·2026-10-29 ·Shuichi Miyazaki, Suguru Tamaki
Abstract · 13 pages

We study Stable Marriage with Ties and Incomplete lists (SMTI) under weak stability. For an instance with $n$ agents on each side, we give an exact algorithm that returns a maximum-cardinality weakly stable matching in $O^*(22.74^n)$ time. To the best of our knowledge, this is the first exact algorithm for this problem running in single-exponential time.

0.54

Length Bounds for MDS Algebraic Geometry Codes of Arbitrary Positive Genus

cs.IT ·2026-10-29 ·Haojie Chen, Chuangqiang Hu, Guorui Ma
theory
Abstract

We study the length of maximum distance separable algebraic geometry codes from curves of arbitrary positive genus. For fixed genus, we prove that the asymptotic ratio of the length to the field size is at most $1/2$ when both the dimension and the codimension grow faster than the logarithm of the field size. We also give explicit bounds under logarithmic dimension conditions. The bounds hold in every characteristic and are uniform in the curve and the evaluation set. The proof combines character sums on the Jacobian with an estimate for sums of distinct points.

0.44

Local Convergence of Triadic Closures of Sparse Random Graphs

cs.DM math.PRmath-phmath.CO ·2026-10-29 ·Himasish Talukdar
theory
Abstract · 46 pages, 4 figures

Triadic closure is a natural mechanism for modeling the clustering phenomenon observed in many real-world networks. Given a possibly random undirected graph $H$, we consider pairs of non-adjacent vertices that have a common neighbor. Conditional on $H$, each such pair is joined by an edge independently and simultaneously, with a probability depending on the degrees of its endpoints and on the number of their common neighbors. The resulting graph is called a triadic closure of $H$. We study the effect of this operation on local convergence. More precisely, we prove that if a sequence of potentially random finite undirected graphs of $\left( H_n \right)_{n \in \mathbb{N}}$ converges locally in probability to a rooted graph $H^*$, then the corresponding sequence of triadic closures converges locally in probability to the same triadic closure of $H^*$. As a consequence, we obtain a general limit theorem for the clustering coefficients of triadic closures. In particular, we show that the triadic closures can have a strictly positive limiting clustering coefficients even when the limiting clustering coefficients of the parent graph sequence is zero. Finally, we apply our results to several classical sparse random graph models including Erdős-Rényi, configuration, preferential attachment models and compute the limiting clustering coefficients of their triadic closures.

0.36

Secure Wireless Communications with Flexible Reflectors: Joint Position and Rotation Optimization for Physical Layer Security

cs.IT eess.SP ·2026-10-29 ·Miao Jiang, Xike Wang, Yiqing Li, Guangchi Zhang
benchmark
Abstract · This work has been submitted to the IEEE for possible publication. 13 pages, 13 figures

By exploiting the spatial selectivity of wireless channels, physical layer security (PLS) protects confidential wireless transmissions. Flexible reflectors (FRs), which are passive metallic plates mounted on movable platforms, offer a low-cost and low-complexity solution for reshaping the propagation environment. However, their potential for secrecy-oriented signal propagation shaping has not been explored. In this paper, we study an FR-aided secure communication over non-line-of-sight links and jointly optimize the positions and rotations of FRs to maximize the secrecy rate under geometric reflection and mutual-blocking constraints. For the single-FR case, we propose a Fibonacci segmentation search (FSS) method, which partitions the feasible region into multiple unimodal intervals and obtains the optimal solution with Fibonacci search over the unimodal intervals. To reduce the computational overhead, we also propose a low-complexity cross-entropy search (CES) method, which employs Gaussian mixture model-based adaptive importance sampling to localize promising search intervals before deterministic refinement. After deriving the blocked placement interval induced by the deployed FRs and integrating it into a sequential deployment scheme, we extend the FSS and CES methods to the multi-FR case. Simulation results show that our proposed FSS method achieves significant secrecy rate performance gains over various benchmark schemes, and our proposed CES method achieves a favorable runtime-performance tradeoff. These results indicate that FRs provide a low-cost and reconfigurable mechanism for PLS.

0.36

Approximating Minimum Spanning Trees in Metrics under Semi-Random Corruptions

cs.DS ·2026-10-29 ·Aditya Anand, Anupam Gupta, Rajesh Jayaram, Euiwoong Lee, Ali Vakilian
publishedhot direction
Abstract · SODA 2027

We study the problem of finding minimum spanning trees in semi-random corrupted metrics. In this problem proposed by Bateni, Dharangutte, Jayaram, and Wang (COLT'24), we start with a metric space $\mathcal{M} = (X, d)$. The adversary independently corrupts each pairwise distance with probability $p$. The algorithm only observes these possibly corrupted distances and must output a spanning tree whose expected cost (in the true metric) is multiplicatively close to the minimum spanning tree cost. Prior work showed that losing a factor of $Ω(\log n)$ is necessary, even if the algorithm may adaptively select $O(n/\log n)$ points and probe the true distances between them. However, no upper bounds were known. We study three regimes based on whether the adversary can only increase, decrease, or arbitrarily change the corrupted distances. We show the following: (1) In the increase-only setting, we give a polynomial-time $O(1)$-approximation algorithm for any constant $0 \le p < 1$. (2) In the decrease-only setting, we give an $O(\log n)$-approximation algorithm for any constant $0 \le p < 1$. Additionally, we show an $O(\log^4 n)$-approximation algorithm that runs in quasi-polynomial time $n^{O(\log n)}$, for any constant $p \leq 0.1$. Moreover, we show that a polynomial-time algorithm computing an $n^{1/3-\varepsilon}$-approximation is unlikely to exist. (3) In the general setting, where the adversary can both increase and decrease distances, we show an $O(\log n)$-approximation algorithm for every $0 \leq p < \frac{1}{4}$. Further, we show that no $o(n)$-approximation is possible for $\frac{1}{2} \leq p < 1$. Our algorithms answer the main open question in Bateni et al.'s work; moreover, the $O(\log n)$-approximations in the latter two settings are information-theoretically optimal.

0.35

Directed Cut-Query Complexity of Finding a King

cs.CC math.CO ·2026-10-29 ·Ziad Ismaili Alaoui, Oded Lachish, Nikhil S. Mande
theory
Abstract

A king in a tournament is a vertex from which every other vertex is reachable by a directed path of length at most two. We study the deterministic query complexity of finding a king when a query specifies a set of vertices and returns the number of edges directed from that set to its complement. We prove that this complexity is $Θ(n)$ for $n$-vertex tournaments. To show this, we first show that every correct deterministic algorithm in this model must output a maximum-outdegree vertex, even though a king need not have maximum outdegree. We then give an adversary whose answers, after too few queries, remain consistent with a tournament in which the algorithm's output does not have maximum outdegree.

0.35

How a group that cannot communicate knows itself by its enemy

cs.GT cs.MAcs.ROeess.SY ·2026-10-29 ·Gregory Falco
Abstract · 6 pages main text; 35-page SI Appendix included as an ancillary file. Code and data: https://doi.org/10.5281/zenodo.23743441

Teams that cannot communicate seem stuck with what each member sees. We show that their opponent can tell them the rest. Silent hunters, jammed robots and traders who never meet all face an adversary that reacts to the group as a whole; every member can read that reaction, a tell, and learn what no member knows alone. One tell can do the work of a message from every member. The adversary would rather hide, but hiding interferes with its own task, and a team that commits in advance to how it will respond decides how much hiding is worth. A committed team never loses when its opponent cares more about its own goals, and does as well as a fully informed team exactly when the opponent's stake outweighs any gain from lying. Below that point the team learns more by trusting less: in a solved Gaussian model it obtains the truth by promising not to use all of it and keeps the cost the opponent would have paid to conceal; an exact test decides when this holds in finite games. How a team spreads its trust matters too. A lone uninformed member with a full-support prior needs a private coin, but with additive squared-error losses two interchangeable members who respond differently replace it, and a small discount on trust insures against misjudging the enemy. Information, on this view, is not only gathered or aggregated; it is produced by how a team commits to respond.

0.34

Deep Reinforcement Learning for Blind Passive Beamforming in RIS-Assisted MIMO Systems

cs.IT stat.ML ·2026-10-29 ·Oumaïma Bounhar, Thomas Chêne, Ghaya Rekaya-Ben Othman, Oussama Damen
published
Abstract · Accepted at AIxNET 2026. 6 pages. Copyright may be transferred to IEEE after publication, after which this version may no longer be accessib…

Reconfigurable Intelligent Surfaces (RIS) are a promising solution for improving millimeter-wave (mmWave) communications by controlling signal propagation. Yet, selecting the optimal RIS configuration remains challenging due to the high-dimensional search space and signaling overhead. We propose a Deep Reinforcement Learning (DRL) framework for blind passive beamforming, where a Deep Q-Learning (DQL) agent learns an adaptive policy to select near-optimal RIS configurations while sounding only few codewords, without channel state information. Results show that our approach outperforms Q-Learning (QL) and conventional baselines in convergence, performance, and overhead reduction.

0.34

EPPA for Reducts of the Random Graph

cs.DM math.COmath.LO ·2026-10-29 ·Sofia Brenner, Jan Hubička, Matěj Konečný
theory
Abstract

We prove that finite two-graphs have coherent EPPA and that neither graphs modulo global complementation nor two-graphs modulo global complementation have EPPA. This completes the classification of EPPA and coherent EPPA for the reducts of the random graph, answering a question and confirming a conjecture of Evans, Hubička, Konečný and Nešetřil.

0.34

Tight Asymptotic Bounds on the Size of Order Graphs

cs.DS ·2026-10-29 ·Ruben Becker, Adrián Gómez-Brandón, Gonzalo Navarro, Nicola Prezza
theoryhot direction
Abstract

Order graphs were introduced by Arroyuelo et al.\ (TODS 2024) as a generalization of the ring index, to support worst-case optimal joins on relations of arity $d$ while storing few columns: the index stores one column per edge, so its size is proportional to the number of edges of an order graph. It was known that the minimum number of edges lies between $Ω(2^d\sqrt d)$ and $O(2^d d)$, and the lower bound had been claimed to be $42$ for $d=6$. We first exhibit an optimal order graph with only $40$ edges for $d=6$, and show that the proof of that lower bound holds only for unions of cycles, whereas order graphs can be smaller than the smallest set of cycles. We give a corrected statement and proof that holds for every order graph: there must be at least $\max_m\binom dm\lceil m/2\rceil$ edges. This bound is $40$ (hence tight) for $d=6$. We then settle the asymptotic question: the minimum size is $Θ(2^d\sqrt d)$. This is obtained via an explicit construction, based on symmetric chain decompositions of the Boolean lattice, with at most $(d-1)\binom{d}{\lfloor d/2\rfloor}+2^d$ edges. The upper bound is within a factor $4+O(d^{-1/2})$ of the lower bound, and both numerically and asymptotically better than trie switching. Finally, we cast the problem as a set cover with connectivity constraints, which yields an exact ILP formulation and a greedy heuristic. With them we find optimal order graphs for $d\le 6$ in the bidirectional and for $d \le 7$ in the unidirectional model, showing that our lower bounds are tight there, and improve the best known order graphs for $d=7$ and $8$.

0.34

PhysicsNeMo-Mesh: GPU-Native Meshes and Simulation Cases for Physics AI

cs.CG physics.comp-phcs.LG ·2026-10-29 ·Peter D. Sharpe, Corey Adams, Negin Sobhani, Sanjay Choudhry
code/data
Abstract

PhysicsNeMo-Mesh is the mesh module of the open-source PhysicsNeMo library. It provides mesh processing similar to PyVista, but on the GPU, in batches, and inside PyTorch's automatic differentiation. Like xarray for unstructured meshes, it attaches semantics to data: fields on points, cells, and whole meshes have names and shapes, and a simulation case keeps the complete PDE problem specification, including boundary conditions and their prescribed values. It addresses three problems that plague machine learning for physical simulation. First, GPU utilization is bottlenecked by I/O, because samples are very large relative to the compute per forward pass. Second, because traditional preprocessing is expensive, it is necessarily done once, during curation, rather than differently for each model and workflow at the point of use. Finally, advanced equivariant and mesh-native architectures require information that curation ordinarily discards. Fast GPU mesh processing makes it practical to preprocess at the point of use: on one GB200 GPU, mesh operations run $6.2\text{--}1{,}068\times$ faster than VTK with 72 threads, and loading a whole case is $8.0\text{--}88\times$ faster than reading its VTK files ($8.0\times$ when both read the same fields). Overlapped with training, this preprocessing costs $2.1\%$ of throughput for a $127\,\text{ms}$ Transolver step. So one dataset serves many models and workflows: in PhysicsNeMo's public recipe, one stored DrivAerML dataset feeds Transolver, GeoTransolver, and GLOBE. PhysicsNeMo-Mesh combines the speed of an intermediate array store with the flexibility of an archival format such as VTK. When shapes and wall conditions both vary, per-boundary conditions cut a heat-conduction error from $87\%$ to $6\text{--}9\%$. And by keeping field ranks and named boundaries, it retains the semantic structure that next-generation architectures will need.

0.34

Retrieval Sequencing in a Compact Storage System with Side Access

cs.DS cs.CC ·2026-10-29 ·Yağmur Gül
theory
Abstract

In a compact storage system with side access, bins are arranged in adjacent vertical stacks. In each retrieval cycle, bins blocking access are lifted to create a temporary tunnel through which a robot extracts one requested bin from the side. The lifted bins are then lowered again, and the energy expenditure of the cycle equals the number of lifted bins. Given a pick list known in advance, we seek a retrieval order that minimizes total energy expenditure. For every number of stacks, we prove that the total energy separates into an instance constant and a pairwise penalty, and that every energy-optimal order is laminar within each stack. For two stacks, we prove that the retrieval sequencing problem is solvable in polynomial time. For three stacks, we give an exact polynomial-time reduction to a merge problem and characterize which merge-problem instances arise from the storage model. We define a recursive class of retrieval orders and construct a polynomial-time dynamic program that minimizes energy expenditure over this class. Finally, we quantify the additional energy savings from batching, which allows several requested bins to be retrieved in one cycle.

0.34

Privacy-Constrained Function Computation in Distributed Databases

cs.IT ·2026-10-29 ·Vamoua Yachongka, Rémi A. Chou
published
Abstract · A preliminary version of this paper was presented at IEEE ISIT 2025

Consider distributed databases that contain both sensitive (private) and non-sensitive (public) attributes. A user seeks to compute a function of public attributes and side information by interacting with the databases and downloading their responses. In this paper, we study the trade-offs among download cost, information leakage about the private attributes, and distortion between the computed function and its reconstruction, referred to as the capacity region. Our main contributions are as follows. For a single database, we provide a complete characterization of the capacity region for general discrete sources and characterize the capacity region of two different setups for Gaussian sources when the computed function is defined as the sum of the public attribute and side information. For distributed databases, we establish inner and outer bounds on the capacity region of discrete sources and we characterize the capacity region when the databases and the side information are mutually independent. As a special instance, we study lossless function computation, and results analogous to the lossy case are obtained.

0.34

INLA for Adaptive UAV Sampling of Radio Environment Maps in IoT Networks: Prior Width Governs Accuracy and Calibration

cs.IT stat.MLcs.LG ·2026-10-29 ·Changhao He, Kaichuang Yang, Xiaojuan Zhang
Abstract

We study adaptive UAV sampling for radio environment map (REM) construction from non-Gaussian IoT observations. Integrated nested Laplace approximation(INLA) updates a Matérn GMRF posterior, with waypoints selected by entropy pre-screening and Sherman--Morrison lookahead over the reachable set. We show that the likelihood-induced Skew-Normal correction has a sextic entropy effect, explaining its limited influence on waypoint selection. Across 19 retained ray-traced windows,Random and lawnmower sampling cost a median 18.0\% and 14.0\% more residual NMSE than posterior-informed sampling, equivalently reductions of 15.3\% and 12.3\% when referred to those baselines, while differences among informative rules remain unresolved. Widening the prior shadowing standard deviation from 6.0 to 10.025\,dB improves NMSE and credible-interval coverage, with monotone improvement over a 4--16\,dB sweep on this corpus and no observed turning point, although overconfidence persists. Online width estimation from the UAV's own counts yields about 6.5\,dB without recovering the gain. Across 30 measured LoRaWAN propagation windows paired with simulated counts and scored only on received cells, the margin over non-adaptive sampling reproduces, the training-derived width minimizes the tested interval score, and GP-UCB has nominally $3$--$5\%$ lower NMSE. The results show substantial, corpus-dependent sensitivity to prior specification, while fine differences among acquisition rules have limited or unresolved effects, the preferred width depends on the corpus, scored cells and model mismatch.

0.33

Persistence of Bounded Rationality in Financial Markets: An Evolutionary Game-Theoretic Model

cs.GT econ.THq-fin.GN ·2026-10-29 ·Cheng Zhang
Abstract · 4 pages, 1 figure. JEL: C73, G14, G41

Classical theory predicts that irrational traders lose money to rational traders and are eventually driven out of the market. Yet noise trading, herding, and heuristic trading persist even in highly developed markets. We model a financial market as an ecology of two trading-behavior types, relatively rational (R) and boundedly rational (N), whose composition evolves under replicator dynamics according to relative performance. Each type's fitness depends on the share of the other type and on frictions. When the interaction is complementary, that is, rational traders profit from the mispricing created by boundedly rational traders while the latter benefit from the liquidity and price correction supplied by the former, and when the difference in frictions is moderate, a unique interior composition exists and is globally asymptotically stable: bounded rationality persists endogenously. When the interaction is antagonistic, the same point becomes a tipping threshold between two one-type regimes. Comparative statics show that lowering the costs of disciplined trading, as algorithmic and AI-based trading does, shifts the composition toward rational trading continuously but does not eliminate bounded rationality unless the cost advantage exceeds an explicit threshold.

0.33

On edge-ordered Ramsey numbers of hypergraphs

cs.DM math.CO ·2026-10-29 ·Martin Balko, Jelena Glišić
theory
Abstract · 10 pages

For an integer $h\geq2$, an edge-ordered $h$-uniform hypergraph $H^<$ is an $h$-uniform hypergraph with a linear ordering of its hyperedges. For an integer $q \geq 1$, the edge-ordered Ramsey number $R_{\mathrm{edge}}(H^<;q)$ of $H^<$ is the smallest positive integer $N$ such that there exists an edge-ordered $h$-uniform hypergraph $G^<$ on $N$ vertices such that, for every $q$-coloring of the hyperedges of $G^<$, there is a monochromatic copy of $H^<$ in $G^<$ preserving the hyperedge order. In this note, we show that, for every edge-ordered $h$-uniform hypergraph $F^<$ on $n$ vertices with $m\geq1$ hyperedges, $R_{\mathrm{edge}}(F^<;q) \leq R_h(F;q)^{Cm^2}$ where $F$ is the underlying unordered hypergraph, $R_h(F;q)$ is its standard $q$-color Ramsey number, and $C=C(h,q)>0$ is a constant. For fixed $h\geq3$ and $q\geq2$, this shows that prescribing the hyperedge order does not require an increase in the tower height of the classical upper bound. It also settles a problem posed by Fox and Li (2020), and for $h=2$ gives the bound $2^{O(n^4\log n)}$, improving their bounds for every fixed $q \geq 3$.

0.33

Stochastic Minimum Spanning Tree Identification with Binary Weights

cs.DS ·2026-10-29 ·Haya Diwan, Lisa Hellerstein, Sanjeev Khanna, Junkai Song
hot direction
Abstract

We introduce the Stochastic MST Identification Problem, with binary weights. In this problem, we need to identify a minimum spanning tree (MST) of a graph, each of whose edge weights is either 0 or 1. The weight of each edge $e$ is a random variable which is 1 with probability $p_e$ and 0 with probability $1-p_e$. The probability $p_e$ is given for each edge $e$, but the realized weight of an edge can only be determined by querying the edge. The problem is then to determine a strategy for querying the edges of the graph that minimizes the expected number of queries that need to be performed until we can identify a subset of edges that form an MST of the graph. We address both adaptive and non-adaptive versions of this problem. For the non-adaptive version of our Stochastic MST Identification problem, we give an $O(\log E)$-approximation algorithm via a reduction to submodular ranking. For the adaptive version, we show that solving the problem is trivial when all $p_e$ are equal to 1/2, because of a relation to the Tutte polynomial. However, computing the expected cost of an optimal querying strategy is #P-hard, even when all $p_e$ equal $\frac{1}{2}$. For arbitrary $p_e$ values, we present a constant-factor approximation algorithm for simple planar graphs. We investigate the question of whether the submodular goal value approach of Deshpande et al. could be used to achieve a $\log$-factor approximation algorithm that would work for arbitrary $p_e$ values and general graphs. In this context, we address the structural question of determining optimal submodular goal values for MST identification in graphs.

0.32

Phase Retrieval of Bandlimited Signals Using a Morlet Wavelet and an Admissible Companion

cs.IT math.CA ·2026-10-29 ·Bei Liu, Meng Ning, Liping Wu, Qingyue Zhang
theory
Abstract

In this paper, we demonstrate that complex bandlimited signals can be uniquely recovered (up to a global phase factor) from two families of wavelet transform magnitudes, sampled at twice the Nyquist rate in two frequency bins. These families come from translations of the fixed dilated Morlet wavelet and a fixed companion wavelet. The companion wavelet admits an augmented Morlet--Gaussian representation. We rewrite the two families as two modulated Gaussian convolution measurements of an auxiliary bandlimited signal. Using a Gaussian convolution uniqueness result, we prove that samples on \((4B)^{-1}\mathbb{Z}\) determine every signal in \(PW_B^2\) (up to a global phase factor). Compared with earlier wavelet phase retrieval results, our approach requires significantly fewer sampling points.

0.31

Accumulating Pruning Constraints: New Exact Polynomial-Space Algorithms for Max 2-SAT

cs.DS cs.AI ·2026-10-29 ·Konstantin Kutzkov
code/data
Abstract · Project folder with Lean verification: https://github.com/konstantinkutzkov/max-2-sat-arxiv

We present deterministic polynomial-space algorithms that solve Max 2-SAT on $n$ variables exactly in time $\Ostar(2^{(1-γ)n})$, for a fixed $γ>0$, on several classes of instances whose clause density may grow with $n$. Our main tool extends the classic DPLL branching algorithm. When the search gives a variable its less promising value, comparing the two values of the variable yields a pruning constraint that is satisfied by some optimal assignment. Accumulating these constraints along each branch prunes all but an exponentially small fraction of the assignments. The classes are formulas with many biased variables, semirandom 2-CNF formulas, and Max-Cut on regular expander graphs and on graphs generated by a stochastic block model. In a semirandom formula, the pairs of variables that may share a clause are arbitrary, and the clauses on these pairs are random. For semirandom formulas and block-model graphs, the algorithms are still exact on every input, and the time bound holds with high probability over the input. To our knowledge, these are the first polynomial-space algorithms with such a running time for nontrivial classes of dense Max 2-SAT instances. Beyond these classes, the pruning constraints can be added to any branching algorithm for Max 2-SAT, and we expect them to be of practical relevance for general formulas, in particular for sparse ones, where each constraint involves only a few variables.

0.31

Substring Complexity in BWT-Runs Bounded Space

cs.DS ·2026-10-29 ·Johannes Fischer
Abstract

Given a string $T[1..n]$, its substring complexity is defined as $δ=\max_k d_k/k$, where $d_k$ is the number of its distinct substrings of length $k$ in $T$. We show how to compute it in $O(n+r\log r)$ time, using only $O(r)$ words of space, where $r$ denotes the number of runs in the Burrows-Wheeler transform.

0.30

An SCL-Aware Design of PAC Codes

cs.IT ·2026-10-29 ·Tyler Kann, Shrinivas Kudekar, Matthieu R. Bloch
Abstract

We develop design methodologies for the rate profile of polarization-adjusted convolutional (PAC) codes tailored to successive cancellation list (SCL) decoding. The crux of our approach is two-fold: first is the creation of a modified density evolution Gaussian approximation (DEGA), called conditional DEGA (CDEGA), to analyze the decoding process in the presence of errors; second is the creation of a dangerous-path occupancy metric to find the conditional occupancy of bit-errors in a finite list. Our approach not only illuminates the role of rate-one precoding, but also provides insight into why certain information sets are particularly well suited to PAC codes. We demonstrate the usefulness of CDEGA and dangerous-path metrics with algorithms that create new information sets, as well as precoders that are extremely sparse yet competitive. In particular, we design codes of length $N \leq 512$ with various rates and list sizes, all of which offer at least half a dB gain over traditional PAC codes at an FER of $10^{-3}$ and improve upon state-of-the-art results.

0.30

Mathematics and Efficiency of Song's CVP Hardness Reduction

cs.CC ·2026-10-29 ·Shuhong Gao
theoryhot direction
Abstract

Song (2026) proves that, for every fixed $0<ε<1$, the closest codeword problem (CCP) for binary linear codes under Hamming distance is NP-hard to approximate within $N^{1-ε}$, where $N$ is the code length, under deterministic polynomial-time many-one reductions. The code-to-lattice construction gives hardness within $N^{1/2-ε/2}$ for Euclidean CVP, with the same $N$ aslattice rank. This reaches every fixed Euclidean hardness exponent below one half. Aharonov and Regev (2005), with the constant refined by Gao (2026), place GapCVP at factor $cN^{1/2}$ in $\mathrm{NP}\cap\mathrm{coNP}$ for every fixed $c>2/(π\sqrt3)\approx0.3676$. NP-hardness at that scale would imply $\mathrm{NP}=\mathrm{coNP}$. This paper has two goals: to explain Song's theorem and its mathematics for graduate students, and to make the reduction more efficient. The reduction turns satisfiability into a system of quadratic equations and then into CCP for binary linear codes, and recovers assignments from recurrence relations of algebraic moments. We adapt the method to linear codes over any fixed finite field of characteristic $p$. A common recurrence for each block column gives a moment window $L=c_p(κ-1)+1$, where $c_p=\max\{2,p-1\}$ and $κ$ is the rank cutoff. Together with ordinary digit products and flexible sampling, this reduces interpolation degrees and output lengths. For CCP gap $N^{1-ε}$, our explicit schedule gives code length and lattice rank at most $n^{4/ε^2+O(1/ε)}$; Song's prescribed binary schedule gives $n^{8/ε^2+O(1/ε)}$. Here $n$ counts variables in a common quadratic system, equal to the number of original variables plus clauses for 3SAT. These are dimension bounds, not optimality claims or running-time exponents. The limiting hardness exponents remain Song's.

0.30

One Polynomial Field for Every Euler Step and the Flow: Polynomial Dynamics for P and PSPACE

cs.CC cs.LOmath.DS ·2026-10-29 ·Olivier Bournez
Abstract · 93 pages. Ancillary files: exact local checks (Python 3, standard library only). v2: explicit horizon in Corollary 55, minor corrections

We construct, for each deterministic Turing machine that halts on all inputs, one polynomial vector field with rational coefficients that performs the same logical computation at unit Euler step, at every smaller or variable step in $(0,1]$ with divergent sum, and along the flow. This unifies discrete-time and continuous-time computation by polynomial dynamics, which earlier characterizations treated separately. For a machine making $T$ transitions in space $H$, a logical scale $h=2^{-s}$ with $s=O(H)$ suffices, with physical time $O(s(T+s))$, drift disturbances up to $h^{2K}$ for a fixed exponent $K$, and $2Ks+3$ fractional bits at unit step. On a fixed compact domain, polynomial time and robust eventual decision characterize P and PSPACE, respectively, with polynomial scale thresholds and no supplied stabilization horizon in the space case. Without sign, floor or conditional primitives, margin amplification at bounded speed, overlapping polynomial gates and interval-product estimates give the common realization, from which we derive shared logical traces, fixed-field PSPACE-complete prediction and a robust safety family with polynomial certificates common to all Euler steps and the flow.

0.30

Pair-Dependent Judge Errors Change the Minimax Rate of Semantic Entropy

cs.IT cs.LG ·2026-10-29 ·Jiamiao Liu, Dewen Qiao, Yu Zhang, Xuetao Chen
code/data
Abstract · 47 pages, 5 figures, 3 tables. Code: https://github.com/yahiko-l/pair-dependent-semantic-entropy

Semantic entropy, the entropy of the meanings among a language model's sampled answers, is estimated from the verdicts of a judge that declares pairs of answers equivalent or not, and automatic judges err in ways that depend on the answers they compare. The existing minimax theory covers only judges whose false-equivalence probability is the same for all pairs. We model a known judge that declares two answers with meanings $a$ and $b$ equivalent with probability $q+ΔM_{ab}$, for a symmetric matrix $M$ and $q$ bounded away from $0$ and $1$, and show that for three meanings the rate is set by tree statistics on four answers, while on an open set of channels the consistency threshold is set by trees on five. For every channel in an explicit open class, the minimax squared risk is of order $N^{-1/3}$ with $N=n^4Δ^6$ on polynomial weak windows on which $n^5Δ^8$ exceeds a multiple of $\log n$, and of order $\min\{1,N^{-1/3}\}$ on every polynomial weak window when the tree functions on at most four vertices are injective. The exponent comes from a cubic degeneracy at the equitable point. On an open set of channels in the class, the four-vertex tree functions take equal values at two distributions with different entropies. For such a channel, on every polynomial weak window, the minimax risk is of the order of the sum of $\min\{1,N^{-1/3}\}$ and a term that decays exponentially in $n^5Δ^8$, so no estimator is consistent while $n^5Δ^8$ stays bounded, even as $N\to\infty$. For $Δ=n^{-β}$ with $β>1/2$, channels of the class with injective four-vertex tree functions allow consistent estimation exactly for $β<2/3$, and the channels of this open set exactly for $β<5/8$, at which the risk exponent drops from $1/12$ to $0$. Computer-assisted certificates show that both kinds of channels occur on open sets.

0.29

Constant-Factor Approximation for Non-Splitting Coflow Scheduling in Heterogeneous Parallel Networks via Iterative Relaxation

cs.DS ·2026-10-29 ·Chi-Yeh Chen
Abstract

Coflows capture the collective communication requirements of data-parallel applications, and scheduling them well is essential to the performance of modern data centers. We study non-splitting coflow scheduling in a heterogeneous parallel network whose $m$ network cores may be Electronic Packet Switches (EPS), not-all-stop Optical Circuit Switches (OCS), or all-stop OCS, where every flow must be transmitted entirely through a single core. Previous approximation algorithms for this setting have guarantees that grow linearly with the number of cores $m$ or with the maximum flow degree $τ$ of a port, and therefore become weak as data centers deploy more cores. The source of this dependence is the rounding step: rounding each flow to the core holding its largest fractional value can inflate the load of a port by a factor of $m$. We replace this step with an iterative relaxation rounding scheme that handles the input-port and output-port capacity constraints simultaneously. For minimizing the makespan, this yields approximation ratios of $\min\{τ, 3\}$, $2\min\{τ, 3\}$, and $4\min\{τ, 3\}$ for pure EPS, pure not-all-stop OCS, and pure all-stop OCS networks, respectively. For minimizing the total weighted coflow completion time, we obtain approximation ratios of $8\min\{τ, 4\}$, $16\min\{τ, 4\}$, and $32\min\{τ, 4\}$ for the same three architectures. In a hybrid network, the guarantee is governed by the weakest architecture present, so the overall ratios are at most $4\min\{τ,3\}\le 12$ and $32\min\{τ,4\}\le 128$, respectively. Unlike all previous guarantees for non-splitting coflow scheduling in heterogeneous parallel networks, these bounds are absolute constants that do not grow with the number of cores $m$ or with the flow degree $τ$. We also show that the integrality gap of the underlying linear program is at least $2$.

0.29

Zero-freeness for Random Regular Graphs up to the Replica Symmetry Breaking Threshold

cs.DS math-phmath.PR ·2026-10-29 ·Zejia Chen, Chunyang Wang, Xinyuan Zhang
Abstract · Revised and expanded: strengthened the characterization of the replica-symmetric phase, reorganized and expanded the proofs. 84 pages

We study antiferromagnetic spin systems on random regular graphs through the lens of the phase transitions characterized by the complex zeros of the partition function. Our motivating examples include the hardcore model and the antiferromagnetic Potts model (including proper $q$-colorings at $β=0$). For both models on random $d$-regular graphs, we establish, with high probability over the graph, zero-free regions throughout the interior of the replica-symmetric regime. Asymptotically, this regime extends up to tree occupation density $$α< α_{\mathrm{RSB}}(d) = (2\log d - 2\log\log d + 2 - 2\log 2 + o_d(1))/d$$ for the hardcore model and up to degree $$d < d_{\mathrm{RSB}}(q) = (2q-1)\log q - 2\log 2 + o_q(1)$$ for proper $q$-colorings. Leveraging the standard algorithmic implications of zero-freeness, we obtain deterministic polynomial-time algorithms that multiplicatively approximate the corresponding partition functions at fixed rational parameters throughout this regime, with high probability over the graph. Our algorithmic guarantees significantly improve upon prior work and extend beyond the parameter ranges reached by previously known search algorithms. As a key intermediate step, we establish the pointwise equivalence, away from the threshold, of three notions of replica symmetry: geometric (averaged pairwise correlation decay), thermodynamic (convergence to the Bethe free-energy prediction), and variational (nonpositivity of the Bethe functional). To establish zero-freeness, we develop reinforced small subgraph conditioning, lifting classical small subgraph conditioning to random holomorphic functions to control short-cycle fluctuations. We also use variational replica symmetry to obtain exponential planted overlap concentration. This enables a typical-graph truncation that controls complex second moments up to the replica symmetry breaking threshold.

0.28

From Trees to Binary Trees: Approximation Algorithms and a PTAS

cs.DS math.CO ·2026-10-29 ·Fabian Weisbrodt, Florian Werner
Abstract

Kellerhals, Krebs, Nichterlein, and Schmid introduced the optimization problem Tree in Binary Tree (TBT), where, given a tree $T$, the task is to find a binary tree $S^*$ with $V(T)=V(S^*)$ that minimizes $\sum_{uv\in E(T)} \operatorname{dist}_{S^*}(u,v)$. They also introduced the variant Tree in Binary Tree With Steiner Nodes (TBT-SN), where the condition $V(T)=V(S^*)$ is relaxed to $V(T)\subseteq V(S^*)$. We give a simple linear-time $\frac{8}{3}$-approximation algorithm for TBT and a polynomial-time approximation scheme (PTAS) for TBT-SN. Our main building idea in both algorithms is to iteratively consider subtrees of the tree $T$ which contain all vertices below some vertex $v$ and construct a solution where $v$ has degree 1. For the PTAS, we cut $T$ into smaller trees, solve these instances, modify the obtained solutions, and combine them to achieve a solution for the original instance.

0.28

Communication-Oblivious Permutation and Sorting

cs.DS ·2026-10-29 ·Qi Dong, Yilei Wang, Ke Yi
theory
Abstract

With the help of secure hardware such as trusted execution environments (TEEs), private data can be uploaded and computed in untrusted cloud clusters, so that an adversary controlling the cloud can only see the data in encrypted form. However, the communication pattern between the processors may still reveal sensitive information about the data, so it is important to develop parallel algorithms in the BSP model whose communication pattern is independent of the input data, a property known as \textit{communication obliviousness}. In this paper, we study such algorithms for three fundamental problems: permutation, sorting, and 0-1 sorting. For the first two problems, we prove a tight (up to lower-order terms) communication complexity of $(2-2/p)n\pm o(n)$. Here, $p$ is the number of processors and $n$ is the input size, assuming $2\le p$ and $p=o(n^{1/3})$. Our lower bound holds regardless of the number of rounds and local memory size, while for balanced permutation and sorting the upper bound is achieved in a constant number of rounds with load at most $n/p+o(n/p)$. For 0-1 sorting, we also derive a tight communication complexity of $C_{\mathrm{cut}}(p) \cdot n \pm o(n)$, where $C_{\mathrm{cut}}(p)$ is the optimal value of a linear program parameterized by $p$. Although there appears to be no closed-form for $C_{\mathrm{cut}}(p)$, we show that it is strictly smaller than $2-2/p$ for all $p\ge 3$, thus separating 0-1 sorting from general sorting in terms of their communication complexity.

0.28

Cubic lower bounds for the permanent in odd characteristic, for Pfaffian representations, and for polynomial entries

cs.CC math.AC ·2026-10-29 ·Karthik Sheshadri
theory
Abstract · 18 pages; stand-alone; submitted to International Mathematical Research Notices

The determinantal complexity of a polynomial is the order of the smallest matrix whose entries are polynomials of degree at most one and whose determinant equals the polynomial. A lower bound for the permanent in this measure is the algebraic form of the question whether the permanent is harder to compute than the determinant. Over the complex numbers, the best known lower bound grows cubically in the size of the permanent. It holds even when the permanent only has to be a limit of such determinants. I prove that the cubic lower bound holds over every algebraically closed field of odd characteristic $p$, with a constant that depends on $p$. Hence the cubic bound does not depend on characteristic zero. I also prove the bound for Pfaffians, which are square roots of determinants of skew symmetric matrices. Lower bounds for Pfaffians imply the same bounds for determinants. Further, I prove the bound, divided by $a$, for matrices whose entries are polynomials of degree at most $a$. All three results come from one lower bound for smooth hypersurfaces. I prove it in every characteristic by counting intersection points together with their multiplicities.

0.28

Communication and Privacy Trade-Offs in Distributed Storage With Shared Secret Keys

cs.IT ·2026-10-29 ·Maryam Keshvari, Remi A. Chou
published
Abstract · A preliminary version of this work was presented at the 2023 Allerton Conference on Communication, Control, and Computing

We study distributed storage over public communication channels where privacy is realized using shared secret keys. Specifically, a user wishes to store a file across $L$ servers such that any subset of at least $t$ servers can reconstruct the file, while any subset of at most $z$ colluding servers learns no more than a fraction $α$ of the file. Unlike classical secret-sharing models that assume secure communication channels between the user and each server, we consider a setting where all transmissions occur over a public channel and the user shares an individual secret key of length $n$ bits with each server. For a fixed key length $n$, we characterize the trade-off between the privacy leakage parameter $α$ and the maximum file size that can be securely stored. Additionally, we determine the minimum local randomness, public communication, and storage required to achieve this optimal trade-off.

0.27

Witnessed Boolean Contracts: Sound Local Distance Guarantees for Approximate Implicit CSG

cs.CG ·2026-10-29 ·Jeremy Pei
unusual
Abstract

Hard min/max operations preserve the sign of exact implicit solids but generally destroy signed-distance semantics. Arbitrarily small, sign-preserving value perturbations can rotate the normalized gradient by nearly a right angle and displace a gradient-projected boundary witness by a finite amount. We formalize this as the absence of a value-only stability modulus for witness projection. We introduce a witnessed Boolean contract checking value and gradient errors, positive reach, active-branch separation, and a Lipschitz visibility margin at the approximate witness. A sharp normalization lemma bounds the normal-direction error. An enlarged neighborhood permits transfer of the inactive operand's bound between the approximate witness and the true foot point. Under the stated analytic and domain assumptions, passing the checks identifies a true point on the union boundary and bounds distance deficiency by the active operand's pointwise value error. Failed or unavailable checks return Unknown. The argument extends to finite hard unions by testing every inactive operand. An executable proof-chain audit finds no violations on 75,738 certified samples. A broader analytic stress test covers spheres, capsules, ring tori, three Boolean operators, and a four-leaf union. Of 96,000 narrow-band samples, 87,469 pass with zero oracle violations; per-case coverage ranges from 72.67% to 98.30%. Removing visibility admits 75 truly invalid witnesses, showing that branch separation alone is insufficient. Non-union cases remain empirical tests, not new theorems.

0.27

Construction and formal verification of a special topology on {\mathbb{Z}}^{n}: characterization of connected subsets

cs.SC cs.LO ·2026-10-29 ·Ignacio Vega-Páez
theory
Abstract

We examine whether the integers admit a topology whose connected subsets are precisely the sets of consecutive integers. The main result gives a construction for every integer lattice {\mathbb{Z}}^{n}, n>0; the answer for Z is then obtained as a corollary. We prove that the neighborhoods determined by the parity of the sum of the coordinates form a basis and that topological connectedness agrees exactly with accessibility by finite paths of unit lattice steps. We also present a deductive audit and the correspondence between this argument and a verification in Lean 4 and Mathlib, including the empty set and the independent formal characterization of integer intervals. The construction and mathematical result originate in Problem 5712, proposed by Dan Marcus and solved in 1970 by the Cleveland State University group; the present article contributes a detailed exposition and a report on formal verification.

0.27

Level Crossing Rate Bounds for Products of Independent Nonstationary Fading Processes

cs.IT math.PR ·2026-10-29 ·Plínio Santini Dester
theory
Abstract

We develop a unified framework for the instantaneous total level crossing rate (LCR) of products of independent positive nonstationary fading processes. Universal upper and lower bounds are first derived from an exact decomposition of the product derivative and signed crossing imbalance, requiring only factor-wise crossing statistics. We then introduce the drifting Gaussian-derivative (DGD) class, in which each conditional derivative is Gaussian with possibly nonzero drift. For products of DGD factors, an exact Gaussian-mixture representation yields bounds that jointly capture aggregate drift and derivative fluctuations. The lower bound reduces under stationarity to a known $\ell_2$ bound, whereas, to the best of the author's knowledge, the upper bound is new even in the stationary case. Furthermore, the bounds are asymptotically exact in drift-dominated regimes. The framework is specialized to nonstationary Double-Rayleigh fading and to quasi-identically distributed DGD factors, with closed forms for products of $α$-$μ$ envelopes. Numerical results show that the DGD bounds closely bracket the exact LCR.

0.26

On Zeros and Algorithms for Disordered Systems II: The Antiferromagnetic Ising Model up to Kesten-Stigum

cs.DS cond-mat.dis-nnmath-phmath.PR ·2026-10-29 ·Ferenc Bencs, Brice Huang, Daniel Z. Lee, Kuikui Liu, Guus Regts
Abstract · 98 pages

We give the first polynomial-time algorithm to estimate the partition function of the antiferromagnetic Ising model on the uniformly random $d$-regular graph to arbitrarily high accuracy for all inverse temperatures up to the Kesten-Stigum threshold. This is the entire regime in which the model is "replica symmetric," and is conjectured to be optimal for polynomial-time counting and sampling. To achieve our algorithmic result, we establish the absence of Fisher zeros for the partition function in a constant-width strip around the interval from $0$ up to the Kesten-Stigum threshold, with high probability over the randomness of the input graph. AI Disclosure: The overarching vision and proof architecture were developed solely by the authors without AI input. GPT 5.5 Pro was used, with extensive guidance from the authors, for only a few minor technical aspects of the proof. The authors independently verified correctness and wrote everything by hand. We take full responsibility for its contents. A more detailed statement can be found in Section 1.2.

0.26

Average pairwise Hamming distance of a subset in a hypercube

cs.DM math.COcs.IT ·2026-10-29 ·Yijia Fang, Hao Huang
theory
Abstract

In this note, we prove that for a subset $\mathcal{F}$ of $\{0,1\}^n$ of density $\frac{1}{8}$, its average pairwise Hamming distance is at least $\frac{n-3}{2}$, with equality attained only by $(n-3)$-dimensional subcubes. This answers a question of Lei Yu, and solves an important open case of a more general question posed by Ahlswede and Katona in 1977.

0.26

Exact Competitive Ratios of Index Rules for Scheduling with Obligatory Testing

cs.DS ·2026-10-29 ·Felix Buld
theory
Abstract

We revisit scheduling with obligatory testing, where each job consists of two operations: a test with an initially known time and an operation whose processing time is revealed upon completion of the job's test. We develop an iterative competitive-ratio-preserving instance-reduction method that enables us to compute exact competitive ratios directly on highly structured instances for various index rules. Our main result is a clean analysis showing that the competitive ratio of the algorithm $1$-SORT is exactly the Golden Ratio $\varphi\approx 1.618$, improving the recent best-known upper bound of $1.861$ for arbitrary testing times. We further leverage this method to obtain near-tight bounds for a simple randomized algorithm on instances with uniform testing times.

0.25

Parameterized Approximation of Max k-Weight CSP: A Dichotomy for Boolean Languages

cs.DS ·2026-10-29 ·Rudranarayan Kar, Souvik Saha, Abhishek Sahu
Abstract

We study \textsc{Max $k$-Weight CSP}: maximize the satisfied weight of Boolean constraints with nonnegative rational weights by setting at most $k$ variables to $1$. Each constraint uses at most $d$ distinct variables and a relation from a finite Boolean language $Γ$. This includes \textsc{Partial Vertex Cover}, \textsc{Densest $k$-Subgraph}, bounded-frequency \textsc{Maximum Coverage}, and cardinality-constrained \textsc{Max-SAT}. We completely classify the fixed finite languages admitting an efficient parameterized approximation scheme (EPAS) with parameter $k+d$. Only nonempty relations excluding the all-zero tuple matter. For such a relation, a coordinate is a \emph{unit witness} if the tuple with a single $1$ in that coordinate is accepted, and a \emph{control} otherwise. A relation is \emph{Direct} if setting any selected control to $0$, with all other coordinates unchanged, preserves satisfaction. It is \emph{Rooted} if every satisfying tuple has a selected unit witness such that keeping it, setting all other unit witnesses to $0$, and setting any subset of selected controls to $0$ preserves satisfaction. If all such relations are Direct, or all are Rooted, we give a deterministic $(1-\varepsilon)$-approximation for every rational $\varepsilon\in(0,1)$ in time $2^{O(d+k\log(1+dk/\varepsilon))}\ell_{\mathrm{in}}^{O(1)}$, where $\ell_{\mathrm{in}}$ is the total input length. The algorithms are uniform in $d$ and give an EPAS parameterized by $k+d$ whenever relation membership is polynomial-time decidable. If neither condition holds, no EPAS exists unless $\mathsf{FPT}=\mathsf{W[1]}$, even for unweighted instances where no two constraints involve the same set of variables.

0.25

Half-Optimal Induced Matchings in Random Graphs: Online and Low-Degree Algorithms

cs.DS cs.CCcs.DMmath.CO ·2026-10-29 ·Abhishek Dhawan, Eren C. Kızıldağ, Neeladri Maitra, Cheng Mao, Timothy L. H. Wee
theory
Abstract · 32 pages plus references

For many optimization problems on random graphs, most prominently the largest independent set problem, natural algorithms reach only half the optimum, and such factor-two gaps are known to be sharp within broad algorithmic classes. We prove such a gap for finding large induced matchings in Erdős-Rényi random graphs. In the worst case, requiring a matching to be induced turns a polynomial-time solvable problem into an NP-hard one; our results reveal an average-case counterpart. In $G(n,p)$ with $C/n\le p\le 1-κ$, the largest induced matching has size $(1+o(1))L$ with high probability, where $L=\log_{1/(1-p)}(np)$. A simple online greedy algorithm finds an induced matching of size $(1/2-ε)L$, while no online algorithm, even with adaptive inspection order and unbounded computation, finds one of size $(1/2+ε)L$ except with probability $\exp(-Ω(L\log(np)))$. In the sparse regime $p=d/n$, low-degree polynomial algorithms exhibit the same sharp threshold. Our guarantees rest on a maximality principle: with high probability, every maximal induced matching is half-optimal. Greedy outputs a maximal induced matching and our low-degree algorithm a near-maximal one. We bypass the delicate coupling arguments classically used to analyze greedy since Karp. For low-degree hardness, we adapt Wein's overlap gap argument, with forbidden structures that track overlaps between endpoint sets rather than edge sets, since a new edge may join two previously used endpoints and thus encounter no fresh randomness. For online hardness, we give a new and significantly simpler argument that avoids the correlated instances of prior OGP-based proofs: at a suitable stopping time, all pairs involving uninspected vertices remain unrevealed, so a direct first-moment bound suffices.

0.25

Hypergraph cycle enumeration for efficient inference of planted k-XOR with smooth trade-off

cs.DS quant-phstat.CO ·2026-10-29 ·Arman Babakhani, Enrico Fontana, Junhyung Lyle Kim, Shouvanik Chakrabarti, Ruslan Shaydulin
Abstract

We introduce hypergraph cycle enumeration (HCE), a simple spectral framework for detecting and recovering planted signals in noisy hypergraph inference problems. We study HCE on one representative problem from this class, namely planted noisy k-XOR, also known as sparse Learning Parity with Noise (LPN). Both theoretically and empirically, HCE achieves comparable recovery behavior to Kikuchi method, matching Kikuchi hierarchy. We identify parameter regimes where HCE outperforms Kikuchi method, showing that neither algorithm is universally better. We also show a subquadratic quantum speedup for HCE by replacing one of the algorithm's steps with a quantum walk. Our classical and quantum algorithms do not refute the quartic quantum speedup of quantum Kikuchi method, which remains valid for the practically relevant regime of large size and high noise strength.

0.25

Strong Lower Bounds for Weak Epsilon-Nets

cs.CG ·2026-10-29 ·Sujoy Bhore, Reilly Browne, Da Wei Zheng
theory
Abstract

Let $P$ be a finite point set in $\mathbb{R}^d$. A weak $\varepsilon$-net for $P$ with respect to convex sets is a finite set $X\subset\mathbb{R}^d$ intersecting every convex set containing at least $\varepsilon|P|$ points of $P$. Determining their optimal worst-case size was described by Alon, Kalai, Matousek, and Meshulam [Adv. Appl. Math., 2002] as "one of the truly important open problems in combinatorial geometry." We give elementary constructions which establish the first polynomially superlinear lower bounds in fixed dimension. Already in $\mathbb{R}^3$, we construct point sets for which every weak $\varepsilon$-net has size $Ω(\varepsilon^{-3/2})$. More generally, for every fixed $d\ge3$, we obtain a lower bound of $Ω_d(\varepsilon^{-γ_d})$, where $γ_d=\sqrt{d/2}+O(1)$. Consequently, no upper bound of the form $O_d(\varepsilon^{-C})$ can hold with an absolute constant $C$ which is independent of the dimension. Despite the recent progress on upper bounds, the strongest previous general lower bound remained $Ω_d(\varepsilon^{-1}\log^{d-1}(1/\varepsilon))$, due to Bukh, Matousek, and Nivasch [Israel J. Math., 2011]. To obtain our results, we apply a quadratic lift to grid-based configurations with many point-line incidences. The bounds also yield polynomial separations between fractional and integral piercing numbers of convex families, and lower bounds for Hadwiger-Debrunner numbers.

0.25

Optimal self-orthogonal binary linear codes

cs.DM math.COcs.IT ·2026-10-29 ·Sascha Kurz
theory
Abstract · 39 pages; comments more than welcome

We complete the determination of the minimum possible length of eight-dimensional binary self-orthogonal codes with given minimum distance. For dimension nine we give exact values and bounds for minimum distance at most $64$. By exploiting the geometric reformulation of linear codes as multisets of points, we fully redo the classification for dimension at most seven in a simplified and computer-free manner. As a by-product we prove the non-existence of $[93,9,44]_2$ codes and show several divisibility results for putative optimal nine-dimensional binary codes.

0.24

Renegotiable Options Fail to Prevent Hold-Up Between Two Profit-Maximizing Frontier LLM Agents

cs.GT cs.MA ·2026-10-29 ·Junwei Yao, Zike Yang, Yongxi Zhou, Yanxian Chen, Jianan Liu +2 more
Abstract

Do contract terms that protect investment between people also protect it between profit-maximizing language-model agents? In the hold-up problem, a seller can adapt a unit to one buyer at a cost before the price is set. Once that cost is sunk, the buyer can offer almost nothing and the seller will still accept, so a seller who foresees this does not adapt. One remedy is an option: the buyer may buy at a fixed price. In theory the option stops working if the buyer can let it lapse and offer its own price instead. Yet in a laboratory experiment by Hoppe and Schmitz, this renegotiable option raised human investment from 41.5% of pairs without a contract to 67.2%. We had Claude Opus 5.5 and GPT-6.1 Sol, each told to maximize its side's profit, play a one-shot game with the same strategic structure under their four contract regimes and a reference-price control. The agents did what game theory predicts: under the renegotiable option no seller adapted, and under the option that rules out this renegotiation every seller did. In a second experiment in which sellers no longer stated their expected revenue before deciding, removing the profit instruction left investment unchanged under no contract and both options. When we changed the numbers so that the buyer would exercise the option even without adaptation, adapting only added cost, and no seller adapted: the agents did not simply adapt whenever an option was offered or whenever its price covered the cost. Under our protocol, a contract term that raised investment between people did not protect it between agents.

0.24

Time-Switching Energy-Harvesting Quantum Relay-Assisted FSO Communication with Displaced Squeezed Signaling

cs.IT quant-ph ·2026-10-29 ·Nizar Khalfet, Francisco A. Monteiro, Ioannis Krikidis
benchmarktheory
Abstract

This paper investigates a quantum relay-assisted free-space optical (FSO) communication system in which a decode-and-forward relay is powered by time-switching optical energy harvesting and forwards the decoded information using displaced squeezed signaling (DSS). To obtain a physically con sistent description of turbulence-induced loss for non-classical signaling, each hop is modeled as a pure-loss bosonic chan nel whose random transmissivity follows a truncated Gamma Gamma law, i.e., the Gamma-Gamma irradiance conditioned on the physically admissible range of a transmittance. Under this channel, coherent states remain pure, whereas squeezed states are mapped to mixed Gaussian states. Under Helstrom-optimal detection, the average error of the coherent hop is derived in exact Laplace-series form. For the DSS hop, we derive the output states, a closed-form Gaussian fidelity with rigorous two sided bounds on the Helstrom error, and prove the existence of a loss-aware optimal photon partition. Furthermore, we prove that optimized squeezing preserves both the turbulence-induced diversity order and the leading asymptotic constant of coherent relaying, so that its benefit is a finite-photon-budget effect, and that an interior throughput-optimal harvesting fraction exists. A covariance-preserving benchmark with an exact series representation yields an analytical upper bound on the achievable throughput whose absolute gap vanishes as the relay photon budget grows. Numerical results based on the exact mixed-state Helstrom error quantify the benefit of squeezing and the regime in which it is useful.

0.24

Fragmented Obedient Equilibria in Routing Games

cs.GT ·2026-10-29 ·Zhengyang Wang, Achilleas Anastasopoulos
Abstract

We study routing games in which competing mediators recommend routes to fixed nonatomic populations. Each mediator chooses a randomized flow allocation to minimize its population's expected travel cost, subject to users having no conditional expected gain above a tolerance $ε$ from disregarding their recommendations. We define a fragmented obedient equilibrium (FOE) by combining user obedience with mediator optimality, and establish existence under a lower-hemicontinuity assumption on the obedient policy correspondences. Under affine edge latencies and independent mediator sampling, we derive an exact reformulation in first and second flow moments that preserves best responses and equilibrium profiles. Every feasible moment pair admits a finitely supported realizing policy. When the network has at most four routes, each mediator's best-response problem is an exact semidefinite program. A numerical illustration on the Braess network shows that relaxing obedience can reduce FOE costs, while gaps relative to the social optimum and optimal obedient pooling can persist.

0.23

The Tight Competitive Ratio of Greedy for Online Steiner Forest

cs.DS ·2026-10-29 ·Iman Gholami, Ilya Hajiaghayi, MohammadTaghi Hajiaghayi, Peyman Jabbarzade
theory
Abstract

The online Steiner Forest problem is a fundamental network design problem in which terminal pairs arrive one at a time and an algorithm must irrevocably purchase edges connecting each pair. The natural greedy algorithm connects each terminal pair by a shortest path after making all previously purchased edges free. Awerbuch, Azar, and Bartal (SODA 1996) proved that this algorithm is $O(log^2 k)$-competitive and conjectured an $O(log k)$ bound, where $k$ is the number of terminal pairs. Bamas, Drygala, and Maggiori (SODA 2022) obtained an $O(\log \NumPairs\log\log \NumPairs)$ bound for greedy on a restricted class of inputs, leaving the general case open. In this paper, we resolve the competitive ratio of the greedy algorithm for online Steiner Forest, proving that it is tightly $Θ(log(k) log log(k))$. Specifically, we establish an $O(log(k) log log(k))$ upper bound for arbitrary instances and construct a matching $Ω(log(k) log log(k))$ lower bound. Our results refute the $O(log k)$ conjecture of Awerbuch, Azar, and Bartal and establish a tight $Θ(log log k)$ separation between the greedy algorithm and the best possible online algorithm for Steiner Forest.

0.23

Classical communication over quantum channels: does entanglement help in nature?

cs.IT quant-phmath-ph ·2026-10-29 ·Peixue Wu
code/data
Abstract · See the companion github repository https://github.com/pxwu24/additivity-of-classical-communication for the presentation of the results and …

Can entangling codewords across independent uses of a quantum channel improve the rate of classical communication? An old conjecture states that there is no such entanglement advantage for all channels. Although Hastings's result and follow-up works disproved this conjecture, it turns out that enormous dimension is required to construct a channel where entanglement enhances classical communication. This indicates that there might be no entanglement advantage for many physical channels, or for all low dimensional channels. Motivated by recent progress on (generalized) amplitude damping channels, we propose a general Schur-weighted block entropy inequality, and show that there is no entanglement advantage for all phase-covariant channels including multi-level amplitude damping channels, generalized dephasing channels, and damping-dephasing channels. As a by-product, we also recover the recently established result that there is no entanglement advantage for all qubit channels.

0.23

Maximum Cardinality Matching in $n^{2+o(1)}$ Time

cs.DS ·2026-10-29 ·George Z. Li, Jason Li, Junkai Zhang
Abstract

We consider the maximum cardinality matching problem in general graphs, and give an $n^{2+o(1)}$-time algorithm for the problem. Our work builds on recent work on combinatorial algorithms for maximum flow and extends the techniques to skew-symmetric flow problems which capture general matchings as a special case.

0.23

Four-Coloring Planar Graphs in Linear Time

cs.DS cs.CGcs.DMmath.CO ·2026-10-29 ·Yuta Inoue, Ken-ichi Kawarabayashi, Atsuyuki Miyashita, Bojan Mohar, Carsten Thomassen +1 more
theoryhot direction
Abstract

The famous Four Color Theorem (4CT) states that any planar graph can be colored with four colors, but how do we find a 4-coloring? In our previous paper [FOCS'26], we presented a simple combinatorial algorithm to 4-color planar graphs in $O(n\log n)$ time. We improve it here to $O(n)$ time, which is optimal. The algorithm from our previous paper is based on proving that every planar triangulation has a linear number of reducible configurations or obstructing cycles. The bottleneck for the algorithm is that we need $O(\log n)$ Kempe changes to extend a recursive 4-coloring to the reducible configurations. Each Kempe change can involve a constant fraction of the graph and takes linear time. Here we show that all $O(\log n)$ Kempe changes can be identified and implemented in $O(n)$ time. The idea is to spend $O(n)$ time to preprocess the graph so that we can identify and simulate each good Kempe change in $O(n/\log n)$ amortized time. Thus, we implement all $O(\log n)$ Kempe changes in $O(n)$ total time. Based on this, we implement the whole 4-coloring in linear time. We note that creating such a representation for fast identification and simulation of the right Kempe changes is a challenge of data structures, unlike the previous paper that builds on a new graph theoretic understanding of reducible configurations.

0.22

Leveraging Recourse and Size Bounds for OnlineKnapsack with Removable Items

cs.DS cs.CL ·2026-10-29 ·Dimitris Fotakis, Laurent Gourvès, Aris Pagourtzis, Panagiotis Patsilinakos
theory
Abstract

In the online unweighted knapsack problem, items arrive sequentially and must be accepted or rejected immediately for a bounded-capacity knapsack, aiming to maximize the total size of accepted items. When decisions are irrevocable, no algorithm achieves a non-trivial competitive ratio. An essential relaxation that circumvents this impossibility is to allow previously accepted items to be removed from the knapsack (a.k.a. removal). In this work, we investigate possible further improvements on the competitive ratio with removal by bringing together two other previously proposed model extensions: knowledge of an upper bound on the item sizes (a.k.a. bound awareness) and ability to reinsert previously rejected items(a.k.a.recourse). Our contributions are twofold. First, we establish close upper and lower bounds on the competitive ratio under bound awareness, removal and recourse. We show that significant improvements on the best known competitive ratio are possible when the upper bound on the item sizes is not too restrictive and only few uses of recourse are available. Second, we introduce the notion of adaptive online algorithms which do not require to know any upper bound on the item sizes but still achieve performance guarantees comparable to the ones obtained under the bound awareness model.

0.22

Bregman projections and partition design for accelerating finite Markov chains

cs.IT math.PRmath.OCstat.CO ·2026-10-29 ·Michael C. H. Choi, Youjia Wang
theory
Abstract · 33 pages, 4 figures

We study Gibbs averaging of finite Markov chains with a prescribed stationary distribution $π$. Given a transition matrix $P$ and the Gibbs kernel $G$ induced by a state-space partition, we prove that $GPG$ is an information projection for every admissible Bregman generator $φ$, with divergences defined using transition probabilities relative to $π$. A Pythagorean identity characterizes this projection and recovers the relative entropy and weighted Frobenius cases. A dual characterization identifies $GPG$ as a kernel closest to stationarity among those with the same projection chain as $P$; for relative entropy, it maximizes the entropy rate. The identities also give monotonicity under partition coarsening. For every reversible ergodic baseline $P$, Gibbs averaging cannot increase the squared weighted Frobenius distance to stationarity at any positive integer time, and the inequality is strict unless $P=GPG$. The corresponding mixing time therefore cannot increase. To choose the partition, we identify the divergence of $G$ from stationarity as the exact worst-case divergence of $GPG$ over all $π$-stationary baselines $P$. This objective depends only on the block masses. We give an exact algorithm for concave block costs, including those arising from relative entropy and power generators $φ(u)=u^p$ with $1<p<2$, and a greedy approximation algorithm with a provable guarantee for $p>2$. For power generators $φ(u)=u^p$ with $p>1$, the objective is proportional to Tsallis entropy of order $2-p$: its minimization favors imbalanced block masses when $1<p<2$ and balanced block masses when $p>2$, while all $k$-block partitions have the same objective value when $p=2$. A numerical experiment compares the resulting projected chains with a local Metropolis sampler on a multi-modal $π$.

0.22

The Support-Distribution Conditions for MSRD Codes Reduce to a Single Inequality

cs.IT ·2026-10-29 ·Xiang Fan
theory
Abstract · 18 pages

Byrne, Gluesing-Luerssen, and Ravagnani determined the number of codewords with a prescribed nonzero support in a linear maximum sum-rank distance code whose blocks have a common number of columns. Their formula depends only on the code parameters and the support dimensions. Evaluating it for arbitrary parameters gives a family of necessary nonnegativity conditions for existence. We prove their conjecture that this family is equivalent to the condition at the vector of total dimension $d+1$ obtained by filling the longest blocks first, where $d$ is the distance parameter. The resulting explicit criterion applies over every finite field $\mathbb F_q$, for all row numbers $n_i\leqslant m$ and $1\leqslant d<N$, where $m$ is the common number of columns and $N$ is the total number of rows. It also covers the boundary cases in which this vector fills each of its nonzero blocks completely, not explicitly included in the conjecture's displayed parameter range. The proof uses a generating function and an exchange identity to reduce the problem to vectors obtained by filling the longest blocks first. We then estimate alternating sums of elementary symmetric polynomials, with a separate argument for $q=2$. The dual conditions admit the same reduction. For codes that exist, the joint condition follows from sphere-packing bounds for the code and its dual. An infinite family of parameters satisfies the joint condition but admits no such linear code.

0.22

On Proximity Gaps for One-Point Algebraic Geometric Codes

cs.IT ·2026-10-29 ·Songsong Li, Chaoping Xing, Chen Yuan, Ruiqi Zhu
theory
Abstract · 43 pages

Proximity gaps and correlated agreement underpin the soundness of interactive oracle proofs. We establish these properties for general one-point algebraic geometry (AG) codes, extending corresponding results for Reed-Solomon codes. For codes of length $n$ over $\mathbb F_q$ with designed distance $n-m$, write $ρ=m/n$ and fix $ρ\in(0,1)$. For polynomial curves of degree at most $\ell$ and any $0<\varepsilon<1$, we prove correlated agreement with multiplicative slack $\varepsilon$ and soundness $O(\ell(\ell+1)/(\varepsilon q))$ whenever $e\le(1-\varepsilon)(n-m)/2$, with no additional genus penalty. This gives asymptotically good AG code families from Garcia--Stichtenoth tower of unbounded length over a fixed field with soundness bounds independent of $n$, beyond the RS restriction $n\le q$. We also establish mutual correlated agreement up to radius $(1-\sqrtρ-η)n$, with soundness $O(\ell n/(η^3q))$ for $0<η<1-\sqrtρ$. Our half designed distance proof combines auxiliary divisors, simultaneous Berlekamp-Welch interpolation, and error-incidence counting to obtain a common polynomial representation without a Polishchuk-Spielman divisibility lemma. Finally, two counterexamples on the Garcia-Stichtenoth tower show limitations beyond half the minimum distance, at radii below the Johnson bound and with positive slack. The first requires a positive constant soundness lower bound independent of $q$ and $n$, precluding a uniform $O(1/q)$ guarantee throughout the Johnson regime. The second fixes the field and lets the relative gap above half distance tend to zero while ruling out correlated agreement for every soundness parameter $σ<1$, with positive limiting rate and relative error radius.

0.22

Strongly explicit Ramanujan 2-lift towers via Hilbert modular forms

cs.DM math.COmath.NT ·2026-10-29 ·Olga Kharlampovich, Igor Pak, Alina Vdovina
theory
Abstract · 27 pages

Expander graphs have been central objects of study for over fifty years. Ramanujan graphs attain the asymptotically optimal nontrivial adjacency-eigenvalue threshold for infinite fixed-degree families. The original arithmetic constructions by Lubotzky-Phillips-Sarnak (1988) and Margulis (1988), and the iterative $2$-lift method of Bilu-Linial (2006) provide different approaches to Ramanujan graph constructions. Motivated by work on Mealy automata, we combine both approaches to give the following new arithmetic construction. For every prime $p\equiv 1\pmod {17}$, we construct a infinite tower of $(p+1)$-regular non-bipartite Ramanujan graphs on $2^n$ vertices, with label-preserving $2$-lifts between consecutive levels. These construction is strongly explicit: for fixed $p$, a neighbor or lift sign is computable in $O_p(n^3)$ bit operations. The construction uses fixed algebraic $2$-adic transformations, specified by finite data computable in preprocessing depending only on $p$. The main number-theoretic tool is Blasius's Ramanujan-Petersson theorem for Hilbert modular forms, applied through the Jacquet-Langlands correspondence to obtain the sharp bound $|λ|\leq2\sqrt p$ for every nontrivial adjacency eigenvalue of our $(p+1)$-regular graphs. The transformations preserve $2$-adic distance, which gives the binary covers and an explicit formula for their signings.

0.21

Nearly Optimal Sparse Navigable Graphs in General Metrics

cs.DS ·2026-10-29 ·Hongyi Chen, Tianyi Zhang
Abstract · 24 pages, 3 figures

We study approximation algorithms for constructing sparse navigable graphs in general metrics. Given an $n$-point metric $(P,d)$ and a parameter $α\ge1$, a directed graph on $P$ is $α$-navigable if, for every pair of distinct points $s,t\in P$, the source $s$ has an outgoing neighbor $p$ with $d(p,t)<d(s,t)/α$. The objective is to minimize the number of edges. Recent independent works of Khanna, Padaki, and Waingarten (2026) and Conway et al. (2026) give $O(\log n)$-approximation algorithms and show that an $o(\log n)$ approximation is NP-hard in general. Before our work, the best known running time for such an approximation for general $α\ge1$ was $\widetilde O(\min\{n\operatorname{OPT}_α,n^{2.5}\})$, where $\operatorname{OPT}_α$ is the minimum number of edges in an $α$-navigable graph. In this paper, we give a randomized algorithm that, with high probability, runs in $\widetilde O(n^2)$ time and achieves an $O(\log n)$ approximation for every $α\ge1$. In general, these guarantees are nearly optimal up to constant factors in the approximation and polylogarithmic factors in the running time.

0.20

The Welfare Limits of Market Segmentation with Captive Consumers

cs.GT econ.TH ·2026-10-29 ·Irfan Tekdir
Abstract · 25 pages, 5 figures

In a duopoly with captive and contested consumers, I characterize every pair of consumer and producer surplus that can arise under some market composition and some public segmentation, and construct a market that produces each pair. The set is lens-shaped. At any level of industry profit, consumers do worst when captive and contested consumers are perfectly separated and best in an unsegmented market in which the two firms have equally many captives. A market with unequal captive bases and some contested consumers cannot reach this upper frontier under any segmentation. Between Bertrand competition and monopoly, the frontier itself lies strictly below what consumers would get if everyone paid a single deterministic price yielding the same profit. Firms must randomize, and with downward-sloping demand the resulting price dispersion lowers consumer surplus.

0.19

Stochastic and Noisy Oracles in Smooth Continuous Optimization

cs.CC math.OCcs.LG ·2026-10-29 ·Albert S. Berahas, Liyuan Cao, Katya Scheinberg
Abstract

We review models of inexact computational oracles in smooth continuous optimization and their implications for the convergence, computational complexity, and attainable accuracy of optimization algorithms. We adopt a framework in which zeroth-,first-, and second-order oracles are characterized by quantitative error bounds and may accept additional control parameters that influence both their accuracy and computational cost. We introduce deterministic and stochastic oracle models and illustrate their properties through a variety of examples, including finite-difference approximations, stochastic gradients, and randomized derivative estimators. Particular attention is given to the distinction between absolute and relative errors, bias and variance, and stochastic function evaluations with independent or coupled randomness. A central theme of the review is the use of common analytical tools to recover and compare convergence and complexity guarantees under different oracle models. For deterministic and stochastic first-order methods, we show how a general descent inequality leads to complexity bounds under various assumptions on the objective function and all proposed oracle models. In the stochastic setting, we develop bounds on expected stopping times and the accuracy that can be guaranteed under different oracle conditions. We also examine the total computational complexity of algorithms relying on specific oracle constructions and briefly review models of inexact second- order information and their use in optimization algorithms.

0.18

On the NP-Hardness of Rank-Metric Minimum Distance and Syndrome Decoding, at Every Rate

cs.CC ·2026-10-29 ·Venkata Gandikota, Vatsal Jha, S. Venkitesh
Abstract · A longer version of this manuscript has been submitted for publication

In this work, we investigate the hardness of computing minimum distance, as well as the hardness of syndrome decoding, for rank-metric codes. It is known (Gaborit and Zémor, IEEE TIT 2016) that for $\mathbb{F}_{q^m}$-linear rank-metric codes of length $N$, minimum distance when $m=O(N)$ and syndrome decoding when $m=O(N^2)$ are NP-hard under randomized reductions. We obtain a derandomization of these results and a bit more. We show that for any fixed small $\varepsilon>0$, for $\mathbb{F}_{q^m}$-linear rank-metric codes of length $N$ with $m=O(N^{1+\varepsilon})$, the problems of minimum distance, nearest codeword, and syndrome decoding are NP-hard under deterministic reductions. Further, we can prescribe the rate; that is, if we restrict to codes having any fixed rational rate $R\in(0,1)$, then the NP-hardness for all three problems persists. Our main technical contribution is simple, and possibly of independent interest. The reduction of Gaborit and Zémor (IEEE TIT 2016) relies on a distance-preserving embedding of a Hamming-metric code into a rank-metric code, obtained by multiplying its coordinates by a uniformly random vector. We show that we can choose an explicit vector from a slightly larger extension field.

0.18

A Uniform Upper Bound on the Length of MDS Codes from Genus-Two Curves

cs.IT math.AGmath.NT ·2026-10-29 ·Haojie Chen
theory
Abstract

We study the maximal length of MDS algebraic geometry codes from genus-two curves over finite fields. Let $C_\mathcal{ L}(D,G)$ be such a code over $\mathbb F_q$, with length $n$ and dimension $k$. Assume that $3\leq k\leq n-3$. We prove that $n<(2q+10\sqrt q+16)/3$. This bound holds for every prime power $q$ and for both parities of $k$. It gives a uniform upper bound of $2q/3+O(\sqrt q)$ for the maximal length. The explicit bound is strictly smaller than $q+1$ when $q\geq125$.

0.17

The information geometry of the Sinkhorn divergence on a Hilbert manifold

cs.IT math.ST ·2026-10-29 ·Paul Navas
Abstract

The Sinkhorn divergence, the debiased entropic optimal transport cost, determines a metric tensor on probability measures defined along curves, due to Lavenant et al. We construct a real-analytic Hilbert manifold of measures, the smooth stratum, on which this tensor is a weak Riemannian metric, and we continue Eguchi's construction of a geometry from a divergence to the third order. The stratum consists of the measures with a positive Sobolev density on a closed Riemannian manifold, and the strata form a scale whose intersection is the Fréchet manifold of smooth positive densities of Ay. Following it, we characterize the metric through the gradients of mean values: its cotangent form is a spectral reweighting of the Dirichlet form of a Markov chain, so the metric is of the same kind as the Fisher-Rao and the Otto metric, and on the stratum it is bounded by a multiple of the Fisher-Rao metric (but not conversely). At third order, the Sinkhorn divergence is self-dual: both connection forms of Eguchi's construction coincide with the Levi-Civita form of the metric, and the cubic tensor vanishes, as a consequence of the transposition symmetry of the transport problem. On every stratum, the Levi-Civita form is represented, so the stratum has a real-analytic Levi-Civita connection, which on the Fréchet manifold is the connection of Ay and Schwachhöfer.

0.17

The Value of Experiments in Assortment Optimization

cs.DS math.OC ·2026-10-29 ·Mohammadreza Ahmadnejadsaein, Amine Bennouna, Omar El Housni
Abstract

Retailers can learn customer preferences by experimenting with different assortments. However, the cost of each experiment and the operational burden of repeatedly changing product offerings can make extensive experimentation impractical. We study how much a retailer can improve its final assortment decision using customer observations from one (or a very small number of) fixed experimental assortments. We consider the multinomial logit model, where product revenues are known but customer preferences are unknown. In this setting, we jointly design the experimental assortment and the rule for selecting a final assortment from observed customer choices to minimize worst-case regret: the expected revenue loss of the final assortment relative to an optimal assortment under the true preferences, maximized over possible customer preferences. We first characterize exactly the minimax regret and an optimal assortment when no experimentation is possible. We then show, surprisingly, that offering a single experimental assortment to just one customer can already substantially reduce regret, depending on how the products' revenues compare. We derive the exact minimax regret and an optimal experimental policy in this setting. When a single fixed experimental assortment is offered to many customers, we give a geometric characterization of the limiting minimax regret and construct an asymptotically optimal policy. Depending on product revenues, this single experiment can yield no improvement over no experimentation or nearly complete regret reduction. Finally, we extend our analysis to multiple experimental assortments chosen in advance, characterizing the limiting minimax regret, constructing asymptotically optimal policies, and identifying which collections of assortments allow worst-case regret to vanish as the number of customer observations grows.

0.17

Random ($\log{n}$)-CNF Formulas Are Hard for Depth-$2$ Frege with Modular Gates

cs.CC ·2026-10-29 ·Jonas Conneryd
theory
Abstract

We prove that, with high probability, refuting a random $k$-CNF formula on $n$ variables with ${M = n2^{k}}$ clauses for $ω_n(1) \leq k \leq O(\log{n})$ requires polynomial calculus degree at least ${n/2-O(n/\sqrt{k})}$ over every field. Through Pang's recent tradeoff between $\mathsf{AC}^0[p]$-Frege size and polynomial calculus degree [Pang 2026, arXiv:2610.00837v2], this implies a size lower bound of $\exp(Ω_p(\sqrt{k} \log{n}))$ for depth-$2$ $\mathsf{AC}^0[p]$-Frege. In this range of $k$, the bound is superpolynomial in $M$ while $M \leq \mathrm{poly}(n)$. We also prove a nearly matching degree upper bound of $n/2-n\ln(k)/{2k} +O(n/k)$, which shows that other formula families are needed to obtain lower bounds for full $\mathsf{AC}^0[p]$-Frege through Pang's tradeoff.

0.17

Learning Finite-Memory Binary Sources

cs.IT math.PR ·2026-10-29 ·Maciej Skorski
Abstract

Estimating biased finite-memory sources from a single trajectory is a fundamental problem with practical impact, for example in estimating entropy for cryptography. We study order-$d$ binary sources whose conditional probabilities have bias at most $ε$, and establish *sharp* sample complexity bounds: relative accuracy $γ$ with confidence $1-δ$ needs $γ^{-2}\,(\tfrac12-ε)^{-d}\,(\log d+\log\tfrac1δ)$ samples up to constants depending on $ε$, with matching upper and lower bounds and no dependence on mixing time or spectral gap. The proof rests on a sharp bound on expected hitting times of the sliding-window (de Bruijn) chain, which stay of order $1/π(w)$ uniformly in $d$ exactly when $ε<\tfrac3{10}$, and on majorization of the window masses by those of independent biased coins. Applied to published entropy-source validations, the bounds show that their trace lengths may be underpowered: $10^6$ bits cannot resolve all transitions of an $8$-bit model to better than about $4.5\%$ relative error.

0.17

Constant-Factor Approximation for Capacitated Clustering in Low Dimension with $(1+ε)k$ Centers

cs.CG cs.DS ·2026-10-29 ·Kangke Cheng, Guanlin Mo, Hu Ding
Abstract

In many applications of $k$-median and $k$-means, including load balancing and redistricting, each center can serve only a bounded number of clients. This leads to capacitated clustering, in which a client may be forced to use a distant center because all nearby ones are full. This loss of locality defeats the techniques behind approximation schemes, and no constant-factor approximation is known in general metrics even for capacitated $k$-median. A long line of work therefore opens $K=k+\lfloorεk\rfloor$ centers for a fixed $ε\in(0,1]$, while respecting the capacities exactly. We give a deterministic polynomial-time algorithm for uniform soft capacities in low dimension, that is, when the clients or the center locations have bounded doubling dimension. With $K$ centers, its ratio is $3-2ε$ for $k$-median and $9-8ε$ for $k$-means, or $5-4ε$ in $\mathbb{R}^d$. Apart from $k$-median in the plane, these are the first polynomial-time ratios bounded by a constant independent of $ε$; previously, the best ratio was $O(ε^{-2}\log(1/ε))$, even for $k$-median. We also obtain ratios that stay bounded as $ε\to0$ for non-uniform and hard capacities in low dimension, and ratios polylogarithmic in $1/ε$ in general metrics. Our approach starts from a simple structural observation: a near-optimal uncapacitated solution can be split into fewer than $2k$ capacity-respecting blocks, so $2k$ centers already suffice, one per block. Our main tool, the rescue lemma, shows that once every block that sends clients to a center of the optimal solution is rescued (given a private center), that center can be closed, so each extra center effectively rescues about two blocks. This lets $\lfloorεk\rfloor$ extra centers rescue nearly an $ε$-fraction of the blocks, yielding the linear dependence on $ε$ in our ratios.

0.16

Quantum communication: a discrimination approach

cs.IT quant-phmath-ph ·2026-10-29 ·Hao-Chung Cheng
Abstract · v2: proofs simplified

We develop a state discrimination approach to unassisted quantum communication through a noisy quantum channel. Starting from the operational duality between entanglement recovery and state discrimination, we construct entanglement-generation codes with explicit decoders constructed from Petz recovery maps. The resulting error bounds hold for every number of channel uses and are expressed in terms of Petz-Rényi coherent information. For product sources, they yield exponential error decay at rates below coherent information.

0.16

Minimum Label $k$-Cut on Planar Graphs: Complexity and Color Structure

cs.DS ·2026-10-29 ·Yuanhao Wang, Wei Wang
theory
Abstract

Minimum Label $k$-Cut asks for a minimum-size set of colors such that deleting all edges of these colors leaves at least $k$ connected components. We study how the structure of the underlying graph and its color classes affects the complexity of this problem. Regarding fixed-parameter tractability with respect to $k$ on simple planar graphs, we prove that the problem is W[1]-hard, even when the underlying graph is a cactus graph and each color appears at most twice. Furthermore, for cactus graphs in which every color class is connected, we establish a complexity boundary in terms of the maximum monochromatic degree: when this degree is at most two, an optimal solution can be found in $O(Nk^2)$ time, where $N$ is the input size, with no restriction on color frequency or the number of blocks spanned by a color; allowing the maximum monochromatic degree to increase to three already makes the decision version NP-complete. Finally, we prove that the problem remains W[1]-hard parameterized by $k$ on fan graphs, even when every color class is an induced path. These results show that bounding color frequency does not suffice to eliminate parameterized hardness. They also establish a monochromatic-degree boundary for connected colors on cactus graphs and identify complexity barriers to extending this tractable case to outerplanar graphs.

0.16

Finding Competitive Equilibria in Online Marketplaces with Platform Fees, Taxes, and Subsidies

cs.GT econ.TH ·2026-10-29 ·Taihei Oki, Ryosuke Sato
Abstract

We study equilibrium computation in multi-unit trading networks with valuations satisfying generalized full substitutes and agent- and transaction-specific piecewise-linear transfer functions. These transfers model platform fees, taxes, subsidies, and membership discounts in online marketplaces. Our ascending procedure raises prices at different rates and reaches the minimum equilibrium price vector in finitely many iterations, each implementable in strongly polynomial time under our oracle and arithmetic models. An equilibrium allocation can then be recovered efficiently. The method extends primal-dual techniques developed for ascending auctions to handle complementarities between purchases and sales through a submodular flow formulation. In single-seller markets with substitutes and complements, our framework also extends the double-track auctions of Sun and Yang (2009) and Shioura and Yang (2015) to bidder- and good-specific transfers.

0.16

Layering and reverse extension for weighted $k$-path vertex cover

cs.DS cs.DM ·2026-10-29 ·Jiajing Wu, An Zhang, Shiming Li, Guohui Lin
Abstract · 13 pages; submitted for publication

We study the weighted $k$-path vertex cover problem for any fixed integer $k \ge 2$, in which a vertex non-negatively weighted simple graph $G = (V, E)$ is given and one wants to find a minimum-weight vertex subset $C \subseteq V$ such that every simple path on $k$ vertices in $G$ contains at least one vertex of $C$. The problem generalizes the classic weighted vertex cover problem, that is, $k = 2$, and it admits a $k$-approximation algorithm of Hochbaum~[SIAM J. Comput., 1982] and a $(k - \frac 12)$-approximation algorithm of Brüstle et al.~[IWOCA 2021]. In this paper, we carefully design local weight functions in the Layering and Reverse Extension local ratio framework of Bar-Yehuda et al.~[ACM Comput. Surv., 2004], leading to an approximation algorithm of ratio $\frac {3k-1}4$ for odd $k \ge 3$ and of ratio $\frac {3k}4$ for even $k \ge 4$. When $k = 3$ and $k = 4$, these ratios are $2$ and $3$, respectively, matching the state-of-the-art approximation ratios for the corresponding special cases.

0.15

A Sharp Fisher-Information Inequality for Boolean Pairs: Proof of the Amari--Kobayashi Conjecture

cs.IT ·2026-10-29 ·Jun Chen, Shun Watanabe, Lei Yu
theory
Abstract

Consider two terminals that observe correlated uniform binary vectors and separately compress their observations to one bit. We prove that, for every blocklength and every correlation parameter in $(-1,1)$, the Fisher information retained by the resulting pair of bits is maximized by matching parity functions of an appropriate degree, thereby resolving a conjecture of Amari and Kobayashi. The proof combines a Fourier-based bound on the derivative of the normalized covariance with a scalar inequality for signed polynomials subject to a global range constraint.

0.15

Diagonalization and Undecidability in Semi-Thue Systems

cs.CC ·2026-10-29 ·Vesa Halava, Reino Niskanen
theory
Abstract

We revisit classical diagonalization argument in the setting of semi-Thue systems, viewing computation as nondeterministic rewriting rather than sequential machine execution. Semi-Thue systems naturally generate derivation trees, making existential acceptance, reachability, and infinite derivations fundamental computational notions. We formulate acceptance and decision directly in terms of derivation trees and study the associated assertion, word, and infiniteness problems. We prove that the complements of the assertion and word problems are not recursively enumerable by direct Post-style diagonal arguments. For the infiniteness problem, we give a constructive diagonalization based on a fixed universal semi-Thue system. Unlike the classical approach, the diagonal construction is carried out entirely within the rewriting formalism: the universal system remains an ordinary nondeterministic semi-Thue system while simulating encoded semi-Thue computations, including self-reference. These results show that undecidability in semi-Thue systems arises naturally from universality and self-reference in nondeterministic rewriting, rather than from machine states or control mechanisms. This perspective clarifies the relationship between rewriting systems, recursive enumerability, and classical diagonal arguments originating in the work of Post and Turing.

0.15

What Can Be Computed with Indefinite Causal Structure?

cs.CC quant-ph ·2026-10-29 ·Omri Shmueli
Abstract

Can quantum computation become more powerful when operations do not need to follow a definite causal order? We study this question in the process matrix formalism of Oreshkov et al. (Nature Communications, 2012), which allows quantum operations to be connected without a fixed global order. In the computational model, a quantum circuit succinctly describes a process matrix, and the algorithm is given access to its generated channel, with computational cost measured by the size of the generating circuit. We give polynomial-time algorithms in two settings. First, general process matrix computation suffices to solve every problem in $\textbf{NP}$. The valid processes used by our algorithm satisfy the purification postulate of Araújo et al. (Quantum, 2017), and their generated channels are unitary, although their generating circuits are not. Second, requiring the process matrices themselves to be pure, we give a polynomial-time algorithm for unique-sink orientation, a general problem capturing several longstanding computational problems in optimization. Our unitary generators are constructed explicitly and efficiently. Assuming that unique-sink orientation has no quantum polynomial-time algorithm, our result shows that indefinite causal order can provide computational power beyond $\textbf{BQP}$ within the process matrix formalism, even while preserving unitarity.

0.15

On the Quantum Chromatic Gap II

cs.CC quant-phcs.LOmath.CO+1 ·2026-10-29 ·Lorenzo Ciardo, Antoine Mottet
Abstract · Added an informal overview of the proof (Section 2)

We confirm a conjecture of the first author [SODA'26] on a strong form of pseudo-telepathy for Khot's $d$-to-$1$ games: for every $d\geq 4$, every $ε>0$, and every $k\in\mathbb N$, there is a $d$-to-$1$ label-cover game of classical value less than $ε$ that admits a perfect quantum strategy using finite-dimensional shared entanglement with measurement contexts of radius at least $k$. As a consequence, there exist graphs whose quantum chromatic number $χ_q$ is $3$ and whose classical chromatic number $χ$ is arbitrarily large. This is the first super-exponential $χ_q$ vs. $χ$ gap, and it improves on the exponential lower bounds of Brassard, Cleve, and Tapp [Phys. Rev. Lett. '99], who produced graphs for which $χ\geq 1.1397^{χ_q}$, and Lalonde, who recently achieved $χ\geq 1.1546^{χ_q}$. To prove the conjecture, we proceed as follows. We start with a ``stretched'' version of the Mermin-Peres magic square, obtained by spacing out its equations and linking them with additional long equation wires. This operation increases the radius of the initial measurement contexts of a perfect quantum strategy for the game, while preserving classical unsatisfiability. We then apply parallel repetition to make the classical value small, and we turn the system into a pseudo-telepathic multi-layered Kneser hypergraph via a PCP construction of Dinur, Guruswami, Khot, and Regev [STOC'03]. Finally, we feed the resulting hypergraph into the reduction of Fei, Minzer, and Wang from hypergraph colouring to $4$-to-$1$ games, we establish quantum completeness of this reduction, and we show that it preserves context radius.

0.14

Building On Learning-Augmented Online Sorting

cs.DS ·2026-10-29 ·Wyatt Olson
Abstract · 5 pages, 0 figures

Online sorting is a problem that has gained attention recently, and is a natural translation of sorting into an online setting. The problem at hand is to place elements of an online sequence immediately and irrevocably with the goal of producing an array that is as close to sorted as possible upon inserting all elements. Recent work by Bercea et al. [ESA 2026] introduced the incorporation of learned predictions into the online sorting domain. They show lower bounds for the consistency and robustness of any algorithm, including randomization, then provide a deterministic algorithm that is $O(n/r)$-consistent and $O(r)$-robust, which are tight bounds. However, this algorithm requires prior knowledge of the optimal offline cost of the full input sequence before all elements have been revealed. We resolve this by providing an algorithm with matching bounds that does not require knowledge of the optimal cost, then extend these results to the more general problem of online metric TSP, in line with the prior work.

0.14

Tight Lower Bounds for Distributional Paging

cs.DS ·2026-10-29 ·Ali Vakilian
Abstract

Distributional paging seeks a polynomial-time eviction rule that competes with the optimal \emph{online} policy $\mathrm{OPT}$ when requests come from a known stochastic source. Two algorithms by Lund, Phillips, and Reingold [FOCS, 1994] are canonical. Their randomized \emph{dominating distribution} algorithm, for Markov sources, evicts a page drawn from a distribution over the cache satisfying pairwise linear inequalities; their deterministic \emph{median} algorithm, for sources with independent inter-arrival times across pages, evicts the cached page with the largest lower median of its conditional next-request time. Their original guarantees of $4$ and $5$ were recently improved to $2$ and $4$ by Pabbaraju and Vakilian [ICALP 2025]. We show that neither improved constant can be reduced in its respective source model. For Markov paging, we give two complementary factor-$2$ lower-bound constructions, each with an explicit polynomial-time dominating selector. A directed-cycle family has ratio $2k/(k+1)$, while a recurrent four-page family with cache size three has ratio $2-4\varepsilon/3$. For the independent inter-arrival model, we show that the median algorithm is not $c$-competitive for any $c<4$. These are separate results for distinct stochastic models. Matching the known upper bounds, they establish tight competitive ratios of $2$ and $4$, respectively, improving on previous lower bounds of $1.5907$ and $1.511$.

0.10

First-Order Bilevel Optimization for Non-truthful Auction Design

cs.GT math.OC ·2026-10-29 ·Shing-Hei Ho, April Niu, Ellen Vitercik, Kai Wang
Abstract

Automated mechanism design has enabled the learning of complex auction mechanisms using neural networks. However, designing non-truthful auctions remains challenging because bidders' equilibrium strategies depend on the auction mechanism and thus evolve as the mechanism is optimized. In this work, we formulate auction design as a Stackelberg game with multiple strategic followers and develop a fully first-order algorithm for revenue optimization in non-truthful auctions. We establish non-asymptotic guarantees for hypergradient approximation and show that, under standard smoothness and strong-monotonicity assumptions, our method reaches an $O(ε)$-stationary point using $\tilde{\mathcal{O}}(ε^{-2})$ first-order oracle calls. Experiments on symmetric and asymmetric single-item auctions, as well as multi-item auctions, show that our method converges faster than second- and zeroth-order baselines and scales better as the number of bidders increases. Our method also discovers auctions that achieve revenue close to truthful baselines. These results provide a scalable framework for automated mechanism design beyond truthful auctions.

0.09

Improving Energy Efficiency of Uplink RSMA via Rydberg Atomic Quantum Receiver

cs.IT cs.NI ·2026-10-29 ·Xinya Shen, Zheng Chu, David Chieng, Chiew Foong Kwong, Qihao Peng +3 more
Abstract · 13 pages, 7 figures

This paper investigates an uplink rate-splitting multiple access (RSMA) network empowered by a Rydberg atomic quantum receiver (RAQR) to maximize energy efficiency (EE). To overcome the amplitude-only measurement limitation of the considered direct readout RAQR, we devise a phase alignment mechanism that linearizes the received signal while retaining full channel energy, thereby enabling tractable optimization. The resulting non-convex fractional EE problem is transformed via Dinkelbach's method. For IoT scenarios with symmetric rate requirements, we derive exact closed-form power allocation from the Karush-Kuhn-Tucker (KKT) conditions, which yields a globally optimal bisection search with negligible computational latency. For asymmetric rate requirements, we develop a generalized block coordinate descent-based power allocation (BCD-PA) framework that eliminates reliance on expensive interior-point solvers. We further propose a low-complexity paired allocation (LC-Paired) that exploits user pairing and closed-form bandwidth solutions via the Lambert-W function. Simulation results confirm that the KKT-based scheme attains the highest EE, while the BCD-PA offers a flexible complexity-performance trade-off. The LC-Paired scheme drastically reduces computational cost with only moderate EE degradation.

0.09

Certified Small Well-Separated Pair Decompositions in Practice

cs.CG ·2026-10-29 ·Guilherme D. da Fonseca
theory
Abstract

Given n points and $ε>0$, a well-separated pair decomposition (WSPD) is a collection of dumbbells {X,Y}, where X and Y have diameter at most $ε$ times their distance, such that every pair of points has one point in X and the other in Y for some dumbbell. A partition WSPD covers every pair exactly once, and a cover WSPD at least once. Callahan and Kosaraju give a partition WSPD with $O(n/ε^d)$ dumbbells in $R^d$, but many point sets have far smaller WSPDs of both kinds. For $d\ge2$, the known $O(1)$-approximations of the minimum size WSPD that preserve the separation $ε$ have unspecified factors, above $10^7$ for $d=2$ by our estimates. For cover WSPDs, we give an $O(n^2\log n)$-time sweep approximation algorithm with a factor 40 in the plane for $ε\le1/16$. It extends to $R^d$ with a factor $2^{O(d)}$. The sweep is also a lower-bound tool: its number of steps is a lower bound on the WSPD size. For partition WSPDs, we analyze quadtree and fair split tree constructions: in the plane and for $ε\le1/5$, they give constant approximation factors, the smallest of which is 252 for a randomly shifted and scaled quadtree. In experiments with $ε=1/20$, the number of sweep steps is at least 0.91 OPT for n = 500, where the optimum OPT is known, and a local search raises it to a lower bound of 0.99 OPT. A greedy variant of the sweep, which is still a certified 40-approximation, is at most 1.28 times the lower bound for n = 10000. Followed by a new local-search heuristic that shrinks the WSPD, it is within 0.02% of the optimum on every instance with n = 500, and at most 1.10 times the lower bound for n = 10000. We then show how to convert the cover WSPDs into partition WSPDs. For n = 10000, the conversion gives on average 1.13 times the lower bound, against 1.05 for the cover WSPD itself and 1.55 for the Callahan-Kosaraju partition WSPD.

0.09

Clifford synthesis is NP-complete

cs.CC quant-ph ·2026-10-29 ·Søren Fuglede Jørgensen
Abstract

We show that the problem of synthesizing a quantum circuit for a given Clifford operator using at most a specified amount of resources is NP-complete for several resource models of practical interest. Specifically, we study synthesis of a given Clifford operator using single-qubit Clifford gates and two-qubit Clifford gates, with full qubit connectivity and no additional ancilla qubits. With single-qubit gates free and CNOT as the only two-qubit gate, we show that deciding whether a given target admits an implementation using at most a specified number of CNOT gates is NP-complete, and we show that the problem remains NP-complete when any two-qubit Clifford gate is allowed and the budget counts all such gates. Moreover, we show that both results continue to hold when SWAP gates may be inserted for free, and when single-qubit gates are also considered part of the budget. In each case, we build on recent work by Li and Guan and show that there are polynomial-time reductions from Hamiltonian path in grid graphs to these decision problems, using a number of qubits quadratic in the number of grid vertices.

0.05

Quantum Query Complexity of Tournament Solutions

cs.CC quant-ph ·2026-10-29 ·Pragya Arora, Udit Kumar
Abstract

A tournament solution records a set of winners from a tournament, that is, from the outcomes of all pairwise comparisons among $N$ alternatives. We study the quantum query complexity of computing tournament solutions when the outcome of each comparison has to be learned by a query. Classically, the picture is uniform: the Copeland set, the uncovered set and the top cycle all require $Θ(N^2)$ queries, and so does finding a single vertex of maximum out-degree. We show that the quantum query complexity of the top cycle can be computed with $\tildeΘ(N)$ queries, and a Condorcet winner can be found with $\tilde Θ(\sqrt N)$ queries. Finding a strong king requires $Θ(N^{3/2})$ queries, the same as finding a vertex of maximum out-degree. We also give a self-contained alternative proof of the known $Θ(N^{3/2})$ bound for finding a vertex of maximum degree, established by Gilani, Mande, Patro, and Raja. Our lower bounds reduce counting problems to tournament problems using Ambainis's adversary method. For the top cycle, we find a Hamiltonian path and locate the prefix forming the top cycle by divide and conquer.

0.03

Robotics, Control & Signals 121

Neuromorphic design of a rhythmic network

eess.SY ·2026-10-29 ·Robin Wroblowski, Rodolphe Sepulchre
unusual
Abstract · This is an author-created, un-copyedited version of an article submitted for publication in the Journal of Neuromorphic Computing and Engine…

Central Pattern Generators (CPGs) have long inspired the design of rhythms that combine endogenous generation and exogenous interaction. In neurodynamics and engineering studies, CPGs have been primarily analysed and designed as networks of coupled oscillators. In contrast, the present article regards CPGs as networks that propagate events through excitable rather than oscillatory nodes. The proposed event-based perspective reduces the design of network-level rhythmic properties to the tuning of local nodal input-output characteristics. This reduction enables a distributed and robust design methodology: nodes can be tuned independently, they can be heterogeneous, and the global rhythm results from local event propagation rather than from a specific global tuning of the network parameters. The present article builds on previously introduced concepts that parallel existing characterizations of oscillators: the event describing function (eDF), which characterises the propagation delay of a node, and the event phase response curve (ePRC), which characterises the phase sensitivity of a node. Event-based CPGs made of excitable rather than oscillatory nodes emerge as a principled design framework for robust rhythm generation.

0.60

CuPIQP: A Batched and Differentiable QP Solver on GPUs for Robot Learning and Control

cs.RO math.OC ·2026-10-29 ·Fenglong Song, Roland Schwan, Yuwen Chen, Yuning Jiang, Colin N. Jones
Abstract

Quadratic programs (QPs) are fundamental building blocks of many robotic tasks such as inverse kinematics, whole-body control, and model predictive control. Extending these methods to large-scale robot learning and control requires solving many QPs concurrently and, for end-to-end learning, differentiating through their solutions. Efficient batched optimization is therefore essential to bringing established model-based methods into GPU-based robotics pipelines. We present \textsc{cuPIQP}, a batched and differentiable QP solver on GPUs based on a proximal interior-point method, with unified support for dense, sparse, and multistage problems. The solver exploits parallelism across independent QPs and their internal structure while preserving per-problem numerical updates and termination decisions. Benchmarks with up to 16384 QPs per batch demonstrate solve-time speedups of up to $18.8\times$ over \textsc{Moreau} and $4.26\times$ over eight-core \textsc{PIQP}. In an end-to-end learning task, \textsc{cuPIQP} accelerates training by $4.6\times$ over \textsc{QPAX} and $7.8\times$ over \textsc{Moreau}. By reducing the computational cost, \textsc{cuPIQP} provides efficient, reusable infrastructure for integrating model-based optimization into large-scale robot learning and control.

0.53

Actuator-Aware Observer Inputs for Cascade ADRC of Quadrotors Under Rotor Lag, Measurement Delay and Sensor Noise

cs.RO ·2026-10-29 ·Xuan-Quy Dao, Ngoc-Bich Le, Le-Minh Nguyen
code/data
Abstract · 19 pages. MATLAB and PX4 SITL evaluation. Code, controller gains, random seeds, and per-seed results: https://github.com/Quang61123/actuator…

In sampled cascade linear active disturbance rejection control (LADRC) of quadrotors, the extended-state observer receives the instantaneous command, although the delayed measurement reflects an earlier, rotor-filtered acceleration. This simulation and software-in-the-loop design study examines actuator-aware (AA) observer inputs: altitude and attitude observers use limited virtual inputs filtered by an assumed rotor model and aligned with the measurement delay, and horizontal observers use acceleration reconstructed from measured attitude. Fixed gains are selected on the discrete-time model under sensitivity and step-response requirements and stationary covariance ceilings on channel activity and on the noise-induced inner-loop increment of the full cascade. In paired set-point trials, PID-Budget, AA-Budget and AA-HalfGain reach mean RMSEs of 65.5, 19.2 and 16.9~mm; under a stricter common cascade ceiling, the AA designs reach 29.9 and 26.4~mm versus 72.3~mm for a PID with about twice their rotor-command increment. Ablations attribute the improvement to the rotor-lag model and the measured-attitude input; at the nominal 5-ms delay, delay alignment changes the RMSE by 0.013~mm. PX4 software-in-the-loop runs reproduce these results. Behind EKF2, a noise model fitted to the estimation errors predicts the roll/pitch and yaw increments within 5\%. After redesign for this stack, AA-Budget-E reduces the set-point RMSE from 59.1~mm (PID-E) to 44.0~mm with a lower rotor-command increment in every paired tracking run (S6: 45\% lower, 15\% after output quantization); AA-HalfGain-E beats PID-E only on the lemniscate. With fixed gains, the AA designs tolerate less unmodelled rotor lag and delay than PID in the linear sensitivity screen. The ceilings are nominal tuning constraints.

0.52

Fundamental Limits of Sampling-Based Reachability: Geometry and Flow Distortion

eess.SY cs.RO ·2026-10-29 ·Jixian Liu, Ihab Tabbara, Hussein Sibai, Enrique Mallada
theorycode/data
Abstract · 16 pages, 9 figures, 3 tables. Extended version of the CoRL 2026 paper; submitted to IEEE Transactions on Automatic Control. Code: https://g…

Sampling-based reachability estimates reachable sets from finitely many simulated trajectory endpoints. Existing guarantees often control missed probability mass or approximation error relative to the convex hull of the reachable set, leaving the geometric error relative to the reachable set incompletely characterized. We study this error as a problem of support estimation under flow distortion. Our assumptions combine a uniform lower bound on the local volume density of the initial set, a lower bound on the sampling density, and a one-sided Lipschitz constant $μ$ for the vector field. These conditions admit initial sets with corners and dynamics that may expand or contract. In the fine-accuracy regime, $\tilde{\mathcal O}\bigl(2^ne^{nμT}(R/r)^n/(ρκ)\bigr)$ independent endpoint samples suffice for Hausdorff accuracy $r$ using the endpoint cloud. Here, $n$ is the state dimension, $T$ is the horizon, $R$ is the volume-equivalent radius, and $ρ$ and $κ$ quantify sampling density and local volume density. For $n\ge2$ and sufficiently small $κ$, a minimax lower bound over all estimators based only on independent endpoints matches these dependencies up to a dimension-dependent factor $2\cdot4^n$ and logarithmic factors, within the stated parameter range and when the state domain accommodates the construction. The exponential burden in dimension, and in the horizon for expanding systems, is therefore intrinsic to this observation model. Planar experiments illustrate the predicted effects of corners and contraction; robot-arm experiments show how empirical accuracy deteriorates with dimension.

0.49

An Efficient SIM Beamforming Scheme Based on Physical-Empowered Projected Gradient

eess.SP ·2026-10-29 ·Yuzhi Yang, Hao Liu, Tianhao Guo, Jiancheng An, Lu Gan +2 more
Abstract · under review by ieee trans

Stacked Intelligent Metasurfaces (SIMs) have emerged as a promising technology for next-generation wireless systems. Especially in multi-input multi-output (MIMO) systems, it has the potential to realize complicated wave-domain transformations with low hardware costs and energy consumption. However, SIM-assisted MIMO systems face two major challenges in channel estimation and beamforming. As the SIM elements do not have the signal processing capability, channel estimation can only be conducted in an indirect manner, which makes it challenging. Moreover, the precoding optimization after channel estimation is non-convex and also difficult. Targeting these issues, we exploit the time-division duplex (TDD) channel reciprocity to calculate the gradient backpropagation in an over-the-air manner. Based on this structure, we propose a gradient-based optimization framework that optimizes beamforming without explicitly estimating the SIM--UE channel. The proposed method, namely the physical projected gradient descent (P2GD) method, optimizes the achievable sum rate while incorporating zero-forcing (ZF) constraints. Based on the method of incorporating the ZF restriction, we propose two variations of P2GD, namely user equipment (UE)-P2GD and base station (BS)-P2GD. We also provide a theoretical convergence analysis and projected gradient techniques to ensure physical feasibility. Simulation results show that the proposed approach can achieve suboptimal beamforming in the multi-user system within a small number of over-the-air iterations, using a per-direction signaling budget below the pilot cost of explicit SIM--UE channel estimation. The proposed method provides a low-overhead beamforming scheme for SIM systems without explicit SIM--UE channel estimation.

0.48

Gym-Anything for Robotics: Scaling physical simulations with digital agents

cs.RO cs.AIcs.LG ·2026-10-29 ·Pranjal Aggarwal, Sean Welleck
benchmarkcode/data
Abstract

Realistic, fast robotics simulations could enable large-scale training of embodied AI models that transfer to the real world. However, despite decades of progress, available simulations cover only a narrow range of real-world tasks and physical phenomena. The core bottleneck is that creating better simulations requires significant expert effort and therefore does not scale. To address this, we introduce Gym-Anything for Robotics, a framework in which coding agents create realistic robotics simulations. Our core insight is that simulation creation is a coding agent task, so improvements in coding agents translate directly into richer and more faithful simulations. We select simulations based on U.S. GDP data to cover economically valuable physical work, define what makes a good simulation and a good task, and develop an adversarial creation-audit loop that checks each simulation and task against these definitions. Using this pipeline, we build RoboWorld, a set of 25 simulations with 25 test and 175 training RL environments of varying difficulty. Even the best frontier models fail on every RoboWorld test task, achieving an average reward of only 0.06. Building on RoboWorld, we introduce several new ways for frontier models to improve, including in-context learning across repeated attempts, a privileged teacher that reads the simulator state and teaches a student in context, self-assisted learning from the model's own privileged experience, and code policies that generate demonstrations across varied scenes to train vision-language-action models (VLAs) such as $π_{0.5}$. Each of these substantially improves performance. We release our pipeline, RL environments, and benchmark to facilitate future research. Code and data are available at https://gym-anything.github.io/robotics

0.48

Data-Driven or Model-Constrained? Comparing Airflow Estimation Methods under Sparse Sensing

cs.RO ·2026-10-29 ·Dominik Aigner, Thomas Wiedemann, Javier Monroy, Achim J. Lilienthal
published
Abstract · 4 pages, 3 figures, accepted at IEEE SENSORS Conference 2026

Airflow-driven advection is the primary mechanism for gas and aerosol dispersion in real-world scenarios, making accurate airflow estimation essential for ventilation monitoring, gas source localization, and contaminant spread analysis. Since obtaining airflow maps via high-resolution spatial sensing is usually infeasible in practice, flow field estimation methods must cope with sparse measurements by introducing statistical or physical priors to compensate for limited data. Existing approaches mainly differ in how they incorporate fluid-flow constraints and environmental geometry. However, due to the lack of systematic comparisons across this spectrum of methods, the potential benefits of physics-based priors for airflow estimation under sparse sensing remain underexplored. To address this question, we quantitatively evaluate four airflow estimators across three simulated 2D scenarios under steady-state conditions and varying sampling densities. The methods range from Gaussian Process Regression and graph-based Gaussian Markov Random Fields to two Navier-Stokes-based approaches, including a novel strong-form residual minimization variant. Our results show that physical priors provide the greatest benefit at low sampling densities by compensating for missing spatial information. Although data-driven Gaussian Process Regression becomes increasingly competitive as sensor coverage increases, model-based estimators retain an accuracy advantage across the evaluated scenarios.

0.45

A Tutorial on Shape Servoing of Deformable Objects-Part I: Fundamentals

cs.RO eess.SY ·2026-10-29 ·Ignacio Cuiral-Zueco, Alessio Caporali, Miguel Aranda, Gianluca Palli, Gonzalo López-Nicolás +1 more
code/data
Abstract · Part I of a two-part tutorial. 9 pages, 5 figures. Code: https://github.com/nachocz/tutorial_on_shape_servoing_of_deformable_objects. Intera…

In robotics, shape control (or shape servoing) refers to the automatic manipulation of a deformable object to achieve a desired shape. It involves a feedback loop that exploits the relation between robot motion and changes in the measured shape. This relation, known as the interaction Jacobian or, in this context, the deformation Jacobian, depends on the object's mechanical properties, geometry, and boundary conditions (e.g., where the object is grasped). Because these factors are often difficult to model accurately, the deformation Jacobian can be estimated locally from observed shape changes produced by commanded robot motions, without requiring a fine-tuned mechanical model. This article is the first of a two-part tutorial on data-driven Jacobian-based shape control of deformable objects. It introduces the central elements of the approach: the formulation of the shape-control problem, shape representation, estimation and online update of the deformation Jacobian, and a standard control law from the literature. It then assembles these elements into a complete shape-servoing loop, derives a local stability condition for this loop, and illustrates why some target shapes cannot be reached. Part II covers more advanced topics. Throughout both parts, a simulated planar example in which two robot arms manipulate an elastic rod observed by a camera illustrates the concepts.

0.42

A Behavioral approach to Constrained Linear Quadratic Control: Feasibility and Optimality

eess.SY math.OC ·2026-10-29 ·Soutrik Bandyopadhyay, Debasattam Pal, Shubhendu Bhasin
published
Abstract · Accepted for publication at IEEE Indian Control Conference 2027

This paper develops a constrained linear-quadratic (LQ) optimal control framework within the behavioral approach. Unlike classical formulations that rely on state-input representations, our method treats both the system dynamics and the constraints as trajectory-level objects. The system is modeled through a kernel representation, while inequality constraints are captured using behavioral inequalities defined over infinite-dimensional convex cones. Using quadratic differential forms (QDFs) to encode performance indices, we formulate the constrained LQ problem as a convex variational optimization over trajectories. By employing Gateaux differentials, we derive the stationarity and complementarity conditions and show that they admit a representation as an augmented behavioral system whose unique decomposition into autonomous and controllable sub-behaviors yields a complete parameterization of all optimal constrained trajectories. The resulting conditions constitute a behavioral analogue of the Karush-Kuhn-Tucker (KKT) framework. A numerical example illustrates the effect of inequality constraints on the optimal behavior. The framework provides a general, trajectory-level representation for constrained optimal control and opens avenues for closed-loop and data-driven extensions.

0.41

A Mixed $H_{\infty}$--Passivity Approach for Leveraging Heat-Pump-Connected District Heating Systems as Frequency Support in Electric Power Systems

eess.SY ·2026-10-29 ·Xinyi Yi, Ioannis Lestas
code/dataunusual
Abstract

Heat-pump (HP)-connected district heating systems (DHSs) can support grid frequency by rapidly modulating HP electrical consumption while accommodating the resulting heat imbalance. Thermal-state-informed HP control improves EPS--DHS coordination, but introduces slower DHS dynamics and heat-side disturbance propagation at the electrical interface on the time scale of secondary frequency control. This paper formulates the regulated DHS as a grid-facing dynamic port and develops a mixed $H_\infty$--passivity framework for disturbance-independent temperature regulation while preserving a clean electrical response during secondary frequency control. Sufficient conditions guarantee local asymptotic stability of the EPS--HP--DHS interconnection across primary frequency control, secondary frequency control, and DHS regulation. These stability conditions are integrated with multi-channel $H_\infty$ attenuation of DHS-state-dependent electrical dynamics, heat-disturbance propagation, and DHS-regulation errors, yielding a tractable convex LMI program for controller synthesis. Cross-model validation deploys controllers synthesized on control-oriented EPS--DHS models, without retuning, on an open-source electromechanical IEEE 39-bus model coupled with four high-fidelity finite-volume DHS models retaining thermal transport, storage, and heat-loss dynamics. The resulting controller improves EPS--DHS coordination without compromising the grid-facing electrical response.

0.40

Veloxity: Towards Memory-Safe UAV Embedded Firmware using the Rust Programming Language

cs.RO ·2026-10-29 ·Derek Ward, Brian Bascom, Tyler Larsen, James Usevitch, Phil Tokumaru +1 more
code/data
Abstract · 22 pages, 12 figures

Recent years have witnessed increased momentum towards memory-safe programming languages for safety-critical systems. Memory-safe languages can eliminate entire classes of common software errors that have historically required extensive analysis and expenditure to avoid. Nevertheless, memory-safe language adoption in aerospace remains fragmented and incomplete, particularly for unmanned aerial vehicles (UAVs). Recent Rust UAV projects either run only in simulation, are closed source, or have not been integrated with an established autonomy ecosystem and tested against existing C/C++ firmware. To address this gap, we present Veloxity: an open-source embedded UAV autopilot written in the memory-safe Rust programming language. Veloxity lowers the barrier to using memory-safe languages within modern ecosystems by providing a minimal, modular foundation for autonomous UAVs. We compare Veloxity to equivalent autopilots written in C++ in simulation, with hardware testbenches, and via outdoor waypoint-following flight tests, demonstrating functional parity with improved performance. These results indicate that modern memory-safe languages are a viable option for reliable and performant firmware for autonomous aerospace systems.

0.39

SONIC-K1: Porting a Universal Humanoid Motion Tracker to a Small Humanoid on a Single GPU

cs.RO ·2026-10-29 ·Kyrylo Kolesnichenko, Irvin Steve Cardenas, Jong-Hoon Kim
code/data
Abstract · 4 pages, 3 figures. Project page: https://kyry.me/sonic-k1

We present SONIC-K1, to our knowledge the first open-weights whole-body motion tracker for the Booster K1 humanoid, obtained by adapting the SONIC framework to this smaller robot. The policy is trained from scratch on a single shared GPU with a small fraction of the compute reported for SONIC. In MuJoCo simulation, it tracks nearly all held-out motion clips and performs well on external benchmarks under SONIC's success criterion. The first release ran on the real K1 but shook its legs on steps and turns, and simulation showed a sensitivity to actuation delay. We therefore post-trained it with the robot's deployment torque limits, randomized actuation delay and gain randomization. The post-trained v1.0 stays robust under real-robot torque limits and large actuation delays, where the first release often failed, and in our runs on the real K1 it walks and turns without the earlier leg shaking. These results show broad motion-tracking capability on a modest training budget. Code, policy weights and evaluation tools are publicly available.

0.39

Turning Multipath into Positioning Information: Persistent Electromagnetic-Twin DoA Localization

eess.SP ·2026-10-29 ·Tuo Wu, Zhentian Zhang, Ye Tian, Hing Cheung So
Abstract · 13 pages, 11 figures

Reflected paths can support localization when line-of-sight (LoS) links are blocked, but their usefulness depends on how accurately the stored model predicts their directions. This paper develops electromagnetic-twin (ET) direction-of-arrival (DoA) localization with a persistent, physically calibrated path-angle field. A nominal floor plan supplies virtual-path geometry, and sparse reference probes at known positions learn a compact directional correction. Field order and regularization are selected from calibration data alone. For each user, unordered DoAs are processed by joint position estimation and one-to-one path association, using assignment-aware multistart search and robust Gauss--Newton refinement. New trusted probes update the stored correction through recursive least squares (RLS), without reprocessing all previous measurements. The analysis establishes calibration rank conditions, objective descent, and conditional information gains from reflected paths, while also distinguishing the decay of initialization bias from the convergence of localization error. Synthetic experiments use 500 queries per operating point. At 50% LoS blockage and $1.5^\circ$ DoA error, the calibrated ET achieves 0.39 m root-mean-square error, versus 0.55 m for nominal geometry. After a wall change, its final error is 0.43 m, versus 0.63 m for a frozen calibrated map. Equal-evidence comparisons show competitive calibrated alternatives and faster abrupt-change recovery with sliding-window updates than with the configured RLS. The results support reusable directional calibration, not a universally superior map or update rule, and validation is limited to synthetic first-order reflections with trusted calibration labels.

0.39

RAIBO-Wagon: Real-Time Navigation System for a Legged Robot Towing a Wagon

cs.RO ·2026-10-29 ·Insung Yang, Eunchae Kim, Jungwoo Hur, Seunghun Jeon, Donghyuk Choi +1 more
code/data
Abstract · 19 pages, 13 figures, 5 tables. Project page: https://insungyang.github.io/raibo-wagon/

Legged robots have demonstrated advanced capabilities across a wide range of real-world tasks, yet much of this progress concerns the robot controlling only itself. Once the robot is coupled to an external object -- particularly an underactuated one -- the dynamics of the combined assembly become more complex. Towing a wagon is a common practical scenario, where coupled dynamics make reliable navigation challenging. This difficulty is amplified in confined spaces, where collision-free navigation requires finding a narrow set of feasible paths under strict real-time constraints. To address this, we propose a navigation framework that combines Conditional Flow Matching (CFM) with Model Predictive Path Integral (MPPI) control over a learned Forward Dynamics Model (FDM). The CFM model mitigates the initialization problem of MPPI and accelerates its convergence, while MPPI with the FDM recovers feasible solutions beyond the CFM model's training distribution by iterative refinement. We integrate the framework into a full-stack navigation system for RAIBO-Wagon, a quadruped towing a custom wagon, and validate it in real-world outdoor experiments on 15 m-long courses through confined spaces, with all sensing and computation performed fully onboard. The system achieves an 85% success rate across 40 trials, towing a total load of up to 40 kg -- nearly 100% of RAIBO's body weight.

0.39

PhysVLMM: Interaction-Conditioned Physical Belief Maps for Articulated Objects in Robot Scene Maps

cs.RO ·2026-10-29 ·Dibyendu Ghosh
Abstract

Mobile robots must open doors and drawers, yet their scene maps record what each object is and whether it has moved, not how it will respond to a pull. This paper proposes PhysVLMM, a method that maintains a physical belief for every articulated part of a Vision-Language-Motion Map and answers a planner's queries about it, with two algorithms: a map update and a query. Watching a part move provably cannot identify its dynamics, and a measured pull identifies only the pulled part, up to one direction. PhysVLMM therefore (i) keeps that direction as a posterior instead of a point estimate, (ii) passes the evidence of each pull in closed form to the untouched look-alike parts of the map, and (iii) answers whether an action will succeed at a stated risk, or abstains. On real doors and drawers in 18 homes, it identifies the joint type in 86% of held-out pulls (48% for a learned single-view method) and predicts the response to a pull with 41% lower error than differentiable simulation (CRPS 0.323 vs. 0.544), four orders of magnitude faster (13 ms vs. 230 s per case). The posterior lowers the final-position error from 0.460 to 0.270, sharing lowers the force error of untouched parts from 56% to 42%, and at every stated risk from 0.02 to 0.30 the false-safe answers stay within it. Pooling all measured parts of a home does as well as the look-alike groups, and simpler variants predict where a pulled part stops at least as well.

0.39

Non-Asymptotic Confidence Regions for Separable Nonlinear Models

eess.SY math.STstat.ML ·2026-10-29 ·Szabolcs Szentpéteri, Péter Kovács, Balázs Csanád Csáji
publishedtheory
Abstract · Accepted for publication at the 65th IEEE Conference on Decision and Control (CDC 2026)

This paper studies robust uncertainty quantification for separable nonlinear least squares (SNLLS) problems. We propose two identification methods, (i) Residual-Perturbed Gradient (RPG) and (ii) Global-Fit-Gap (GFG), to construct joint distribution-free confidence regions for both the linear and nonlinear parameters. By exploiting assumed group invariance of the noise (such as symmetry or exchangeability), we apply random perturbations (such as sign-changes or permutations) to build the regions based on resampling and ranking. We prove that both methods guarantee non-asymptotically exact coverage (type I error). Finally, we illustrate the constructions through two numerical experiments: (i) identifying generalized finite impulse response (GFIR) systems using Laguerre polynomials with an adjustable decay rate and (ii) fitting Gaussian radial basis function (RBF) models with variable knots and widths.

0.39

Maxwell-Consistent 3-D Magnetic Field Camera using Scalar-Potential-Based Reconstruction

eess.SP ·2026-10-29 ·Paul Jürß, Marija Boberg, Philip Suskin, Tobias Knopp, Fynn Foerger
published
Abstract · Accepted at IEEE Sensors 2026

Various applications require accurate 3-D characterization of magnetic fields with high spatial and temporal resolution. Building on our previously introduced spherical magnetic field camera employing 86 three-axis Hall sensors arranged on a spherical surface, we present a novel magnetic-field reconstruction approach based on the magnetic scalar potential and its spherical harmonic expansion. Unlike the previous reconstruction method that expands each vector component separately, our scalar-potential-based formulation is inherently consistent with Maxwell's equations for source-free fields by design, yielding curl-free and divergence-free field models. Using a simulated magnetic gradient field generated by two permanent magnets, we demonstrate that the proposed method achieves higher reconstruction accuracy, improved robustness to sensor noise and failure, and substantially reduced non-physical divergence and curl compared to the component-wise baseline. Furthermore, the scalar-potential approach attains superior performance with fewer coefficients, indicating a more efficient use of measurement data allowing higher-degree polynomial representations.

0.38

SymAnt: Grounding Long-Horizon Action Anticipation with Adaptable Symbolic Priors

cs.RO cs.CV ·2026-10-29 ·Fnu Aryan, Yaqi Xie, Joseph Campbell, Woojun Kim, Kwonjoon Lee +4 more
publishedcode/data
Abstract · Accepted to WACV 2027. 18 pages (8-page main paper, references, and supplementary appendix), 6 figures, 9 tables. Project page: https://arya…

As robots and intelligent systems increasingly interact with humans, they must anticipate long sequences of future actions from only brief observations. While purely neural models typically require extensive observations to predict future actions and remain difficult to interpret and adjust, symbolic priors offer a structured alternative. To this end, we introduce SymAnt, a novel neuro-symbolic approach that unifies the expressiveness of neural networks with the structured reasoning of scene and knowledge graphs. Given a short video, SymAnt performs a joint-search over both graphs to construct a symbolic context, enabling unified reasoning over spatial relationships, object attributes, affordances, and associated actions. This symbolic context serves as a prior for an iterative refinement process that progressively generates and improves the sequence of future actions. By grounding predictions in symbolic context, SymAnt anticipates long-horizon actions from short observation windows in a parameter-efficient manner, while enabling users to directly edit the underlying knowledge to adjust predictions without further training. Through our experiments, we demonstrate the effectiveness of SymAnt, particularly when only short observation contexts are available, across five diverse datasets: Breakfast, 50 Salads, EPIC-KITCHENS, EGTEA Gaze+, and Assembly101.

0.37

From Airspace Degradation to Deconfliction: Predicting Safety-Critical Traffic States through the Evolution of Multiscale Complexity Metrics

eess.SY physics.soc-ph ·2026-10-29 ·Raúl López-Martín, Justus Renkhoff, Yash Guleria, Mohsan Jameel, Massimiliano Zanin
Abstract · 44 pages, 12 figures

Growing air traffic demand is pushing airspace management towards the limits of its operational capacity, with steadily increasing workload for air traffic control officers, with direct consequences for safety. Understanding which traffic conditions give rise to safety-critical situations is therefore an essential open problem. Here we compare a large number of deconfliction events with equivalent conflict-free scenarios, extracted from thirteen weeks of air traffic control data over the Swedish Flight Information Region. We describe the traffic structure before each event through thirty sector-complexity metrics computed at multiple spatial and temporal scales, and use them as inputs of Machine Learning models. We find that the need of deconflictions can be predicted with an accuracy reaching $70\%$, but only when the traffic is described at the sector spatial scale. Predictability increases monotonically in the minutes preceding the clearance, showing that situations requiring deconflicting actions are not sudden occurrences, but rather the endpoint of a progressive degradation process. Additionally, the information needed for such prediction mainly resides in specific characteristics of the overall traffic structure, and not only in the conflicting pair. These results numerically demonstrate that safety-critical airspace states can be anticipated from the collective behaviour of the traffic, paving the way for early-warning systems that flag degraded, safety-relevant conditions before an intervention is required.

0.37

Compressed Orientation Reachability Maps for Sub-Second Robotic Needle Entry Point Planning

cs.RO cs.CGeess.SY ·2026-10-29 ·Prabhu Nithin Gollapudi, Andreas Maier, Franziska Mathis-Ullrich, Water Zhao, Jia Wang +1 more
Abstract · 6 pages, 4 figures, 2 tables

In the context of multi-stage trajectory planning for robotic percutaneous liver interventions, identifying a kinematically feasible entry point remains a critical computational bottleneck. Existing methods rely on exhaustive inverse kinematics or randomized searches, incurring delays of 13 to 27 seconds per target and limiting real-time intraoperative adaptation. This paper introduces an Orientation-Based Reachability Map (ORM) that precomputes the kinematic feasibility of a 6 Degrees of Freedom (DoF) serial manipulator across 188,873 workspace positions and encodes orientation feasibility into a compact, indexed 83 MB map. A Just-In-Time (JIT) compiled runtime pipeline replaces live optimization with near-instantaneous array lookups. Experimental evaluation demonstrates a reduction in planning latency from 13.5-26.9 s to 0.289-0.349 s per target while maintaining high geometric fidelity, achieving a mean convex hull volume overlap ratio of 0.933 and a mean alpha-shape volume ratio of 0.976 relative to the baseline reachable point cloud. The proposed ORM framework enables sub-second entry point planning and provides an efficient computational framework for robotic needle insertion in percutaneous liver ablation.

0.37

socx: GPU-Accelerated Chance-Constrained Stochastic Optimal Control using JAX

eess.SY cs.RO ·2026-10-29 ·Tanmay Dokania, Yashwanth Kumar Nakka
Abstract

In this article, we present socx, a GPU-accelerated solver for lifted convex reformulations of linear stochastic optimal control under chance constraints, implemented in JAX. We demonstrate that by exploiting the sparse and independent structure of the problem, distributed optimization can reduce it to simpler problems: independent linear-quadratic tracking (LQT) problems and Euclidean projections onto sets. These simpler problems can be solved in parallel and then coupled using the constraints. To validate our solver, we run it on a hover-linearized 3D quadrotor and show a speed-up of up to $98\times$ over MOSEK and up to $2224\times$ over SCS. socx also solves horizons up to $N=1120$, whereas MOSEK runs out of memory at $N=320$. We present ablations to isolate the key components that enable the speed-up.

0.37

Computing with Coupled Oscillator Networks: Optimization, Memory, and Modern AI Primitives

eess.SY cond-mat.dis-nnnlin.AO ·2026-10-29 ·Ahmed Allibhoy, Fabio Pasqualetti
Abstract

We explore how synchronization in coupled oscillator networks may be employed to perform a variety of computing tasks. These systems are becoming more attractive as an alternative to conventional digital computers due to the increasing energy demands of the latter, particularly for large-scale problems in machine learning and artificial intelligence. The ubiquity of coupled oscillator networks means that oscillator-based computing can be thought of as a universal computing paradigm encompassing systems with a diversity of physical realizations, e.g. as analog circuits, neurological systems, spintronic systems. The bridge between the dynamics of coupled oscillator networks and computational primitives they implement is the Ising model, which originates in statistical physics, and when applied to oscillator networks, explains the emergence of globally complex and disordered synchronization patterns from local interactions in the network. We discuss the application of coupled oscillator networks to three fundamental computing operations. First, oscillator networks as specialized hardware solvers of NP-hard combinatorial optimization problems, with the solution encoded in the synchrony patterns. Second, oscillator networks as associative memory systems, enabling storage and retrieval of data represented through phases of oscillations. Finally, oscillator networks as a dynamical implementation of attention, with synchronization providing the mechanism for the alignment of keys to queries. The ability of oscillator networks to implement these computational primitives paves the way for novel realizations of intelligent systems through a variety of physical substrates.

0.37

Dynamic Persuasion: Managing Reputation under Censored Feedback

eess.SY ·2026-10-29 ·Ahmet E. Arcaklioglu, Ceyhun Eksin
published
Abstract · 44 pages, 8 figures. A preliminary version of part of this work appears in Proc. IEEE CDC 2026

We study a long-run sender who repeatedly communicates about a state to a sequence of myopic receivers and whose credibility is a public score updated from past interactions. Receivers face a lemon loss and an opportunity loss, and their behavior depends on these only through a single index, the tolerance. Because the score records only detected misrepresentations, it rests above the sender's true honesty rate; we characterize this credibility premium in closed form. In a fluid limit in which the score adjusts slowly, no policy earns more than operating at a single credibility level, and the sender's gain from managing credibility equals the value of the premium: against receivers who correct for imperfect detection, that gain vanishes. At a fixed adjustment speed, every policy is bounded by the fluid value plus a term linear in the speed. We then study three policy classes, linear, (s,S), and threshold, obtain the optimal parameters of the linear and threshold classes in closed form, show how receiver losses shape them, and show that, under an approximation, the optimal (s,S) band collapses to a threshold. Simulations of the dynamics agree with the closed forms and indicate where the single-level conclusion stops: bands pay when switching modes is costly, and cycles pay when demand is seasonal.

0.36

MUSBRIC: LEGO-Compatible Pneumatic Muscle Bricks for Reconfigurable Musculoskeletal Robots

cs.RO ·2026-10-29 ·Qiyao Wang, Yinglei Zhu, Liang Zhou, Wenzhe Gong, Xin-Jun Liu +1 more
Abstract · 6 figures. Qiyao Wang and Yinglei Zhu contributed equally to this work. Improved HTML rendering of Figure 1; scientific content unchanged

Musculoskeletal robots combine compliant actuators with articulated skeletons, but exploring alternative configurations often requires repeated fabrication. We present MUSBRIC, a modular hardware platform and accompanying MuJoCo-based simulator for reconfigurable musculoskeletal robots. LEGO-compatible endcaps turn fiber-reinforced extensile pneumatic artificial muscles into reusable, plug-and-play muscle bricks that connect directly to standard LEGO components. Experimentally identified actuator behavior supports corresponding musculoskeletal simulations. An elbow-joint prototype validates load-dependent flexion and extension, showing that pulley-guided routing enables reversible motion and a mechanical stop stabilizes the straight posture. A robotic hand demonstrates compliant grasping and multi-joint construction. Matched hardware and simulation sweeps across 15 leg configurations demonstrate systematic exploration of attachment geometry, while a quadruped assembled from four legs demonstrates walking. Together, these demonstrations span mechanism analysis, configuration exploration, and functional robot construction within a single reconfigurable system. The hardware designs and simulation code will be released as open source soon to facilitate reproduction and further development.

0.35

Learning the Omega-Limit Partition from Approximate Koopman Invariants

eess.SY ·2026-10-29 ·Dhruv Shah, Jorge Cortés
Abstract

Nonlinear dynamical systems with multiple omega-limit sets pose a fundamental challenge for global finite-dimensional Koopman modeling with continuous observables. A natural way to address this challenge is to decompose the state space into forward-invariant regions associated with distinct asymptotic behaviors and to learn a separate Koopman model for each region. This paper studies how to compute such partitions from data. We establish conditions under which Koopman-invariant observables, namely eigenfunctions associated with the unit eigenvalue, characterize the stationary partition induced by omega-limit sets. Building on this characterization, we develop OPAL (Omega-limit Partitions from Approximate invariants Learned from data), an algorithm that extracts partitions from approximate invariant observables. Numerical examples show that the recovered partitions capture the relevant asymptotic regions and support multiple regional linear Koopman models with significantly improved long-horizon prediction capabilities.

0.35

AFDM-ISAC Design Amalgamating FMCW Radar and Carrier Aggregation

eess.SP ·2026-10-29 ·Qu Luo, Qihao Peng, Zihuai Lin, Maged Elkashlan, Pei Xiao +2 more
publishedhot direction
Abstract

Carrier-aggregation (CA)-assisted affine frequency division multiplexing (AFDM) is conceived for integrated sensing and communication (ISAC) systems. A frequency-modulated continuous-wave (FMCW)-inspired dechirping receiver is developed to exploit the inherent pulse-compression property of AFDM chirp subcarriers for accurate delay-Doppler estimation. To further enhance sensing performance, a pair of CA schemes is proposed. The Type-I CA scheme improves sensing accuracy by aggregating echo signals across multiple CA bands prior to radar processing, thereby reducing estimation error with low hardware overhead. The Type-II CA scheme enhances range resolution by effectively enlarging the sensing bandwidth, where chirps from different CA bands are designed to form a continuous set of chirp signals. Based on this principle, variable subcarrier sensing and asynchronous sensing schemes are developed. For the communication receiver, the generalized input-output relationship of AFDM is derived under both fractional delay and fractional Doppler (FDFD) channels. To address the challenge of FDFD estimation, a cross-domain sensing-assisted channel estimation (CDS-CE) scheme is proposed to facilitate time-domain delay-Doppler parameter estimation and affine-domain channel coefficient recovery. Simulation results demonstrate that the proposed framework achieves significant improvements in sensing accuracy and resolution, as well as in channel estimation performance compared with state-of-the-art methods.

0.35

Automated Optimization of Patient-Specific Electro-Anatomical Models Fitted to Cochlear Implant SCINSEV

eess.SP ·2026-10-29 ·Minh Q. Vu, Erin L. Bratu, Jack H. Noble
Abstract

Purpose: Patient-specific electro-anatomical models (EAMs) of the implanted cochlea are crucial for auditory nerve fiber simulation in cochlear implant (CI) programming and neural health estimation. Fitting an EAM to the stimulation current induced non-stimulating electrode voltage (SCINSEV) currently requires manual expert adjustment or exhaustive search over precomputed forward solutions, neither of which scales to routine clinical use. We present an automated optimization pipeline that fits patient specific EAMs in minutes. Methods: Four compartment tissue label maps (perilymph, soft tissue, modiolus, and bone) derived from postoperative CT were used to construct a finite difference forward model, fitted to the measured SCINSEV by Gaussian process Bayesian optimization. Three reductions cut fitting cost: a coarsened grid with trilinear injection and readout at sub-voxel contact positions; solving half the stimulating contacts and completing three quarters of the matrix by reciprocity; and a Jacobian-based identifiability analysis reducing four resistivities to two. Each reduction was validated separately against the measurement's reciprocity residual and the reproducibility of the fit. Results: Across 43 implanted ears, fitting took a median of 2.6 minutes per ear. Median SCINSEV RMSE on the reduced-resolution objective was 0.069 mV, against 0.084 mV for grid search at a matched evaluation budget. Against an exhaustive full-resolution grid search requiring a median of 9.1 hours per ear, the pipeline reached statistically indistinguishable fitting error roughly 210x faster. Conclusion: Patient-specific cochlear EAM fitting can be performed at clinically practical timescales without expert supervision, removing a major computational barrier to patient-specific cochlear implant programming and cochlear simulation.

0.35

FlapSim: A Composable MuJoCo Simulation Platform and Reduced-Order Aerodynamic Model Hierarchy for Flapping-Wing Robotics

cs.RO ·2026-10-29 ·Mingshan He
code/data
Abstract

Flapping-wing simulation need to reconcile high-frequency wingbeat aerodynamics with the slower dynamics relevant to flight control and reinforcement learning. We present FlapSim, a modular open-source MuJoCo simulation platform that combines robot descriptions, wing geometry, and composable aerodynamic models in a deterministic simulation pipeline. Across $48$ held-out scenarios spanning frequency, gimbal, wind, frequency-step, gust, and cross-robot flat-plate regimes, we compare five reduced-order configurations operating at $1\,\text{ms}$ with a $50\,μ\text{s}$ quasi-steady blade-element reference. The built-in hover cycle-averaged baseline achieves a $67.51\times$ CPU speedup but only $0.6593$ mean wrench fidelity. Combining multirate $21$-substep clock synchronization, three-dimensional linear/quadratic inflow and gust coupling, analytical blade-element geometry scaling, and phase-locked $1f\text{--}6f$ harmonics raises mean fidelity to $0.9646$ at a $65.19\times$ speedup in a five-block model. A streamlined four-block model without explicit step-transient lag states achieves $0.9662$ fidelity at a $67.98\times$ speedup. Controlled ablations show that stroke harmonics are particularly important during frequency transitions, improving held-out frequency-step fidelity by $+0.3176$. Transfer-function analysis explains this effect: fixed-window cycle averaging retains harmonic contributions when the wingbeat frequency changes, so within-cycle reconstruction also improves the filtered wrench. These results characterize the accuracy--computation tradeoff of the aerodynamic hierarchy and identify timing, inflow, geometry, and harmonic representation as central considerations for reduced-order flapping-wing simulation. The source code are available at: https://mingshanhe.github.io/FlapSim/\href{https://mingshanhe.github.io/FlapSim/}.

0.35

Compound Climate Risk-Aware Electricity Dispatch: Trajectory-Based Modeling for Power System Resilience

eess.SY ·2026-10-29 ·Feiyu Cai, Jing Qiu, Chenxi Zhang, Yi Yang, Yuechuan Tao +1 more
Abstract · 10 pages, 8 figures

In recent decades, emerging compound hazards in a changing climate have increasingly threatened power system resilience and posed new challenges to electricity dispatch. Conventional Value-at-Risk (VaR)-based risk-aware dispatch provides a tractable means of managing stochastic net demand deviations caused by distribution feeder outages. However, when applied to temporally compounding events such as tropical cyclone (TC)--blackout--heatwave, conventional net demand risk formulations may assign similar risk levels to a brief outage and an outage with the same instantaneous net demand impact that persists into a subsequent hazard, despite their different requirements for sustained operational flexibility. Moreover, directly incorporating compound risk into electricity dispatch models faces a structural mismatch: the risk is defined by the joint distribution of feeder restoration and subsequent hazards over exponentially many joint states, whereas dispatch models require risk formulations that are tractable in the scheduling decisions. To address these issues, this paper proposes a compound climate risk-aware dispatch framework with a trajectory-based compound risk model, which defines the loss on operating deviations determined by entire feeder connection trajectories, parameterized by uncertain restoration times and coupled with subsequent hazards. Analytical moment-generating functions and equivalent exponential-cone reformulations provide a compact representation of the resulting risk constraints, preserving the relevant dependencies without explicit joint-state enumeration. By leveraging a physics-based spatiotemporal TC--blackout--heatwave hazard model, case studies demonstrate the effectiveness of the proposed method with real-world time-series power outage data from Puerto Rico, including data collected during Hurricane Fiona in 2022.

0.34

Finite-Sample LTI System Identification from a Misspecified Finite Model Set

eess.SY math.OCstat.ML ·2026-10-29 ·Nicolas Chatzikiriakos, Andrea Iannelli
theory
Abstract

We consider the problem of finite-sample identification of an unknown LTI system from a finite hypothesis class. In contrast to existing results that assume the hypothesis class contains the true system, we consider the misspecified setting, where the true system does not belong to the hypothesis class. We first characterize the resulting estimation target and its dependence on the data-generation scheme. We then establish a finite-sample lower bound on the failure probability that any identification algorithm must incur. Finally, we analyze the robustness of an existing exponential weighting-based identification algorithm to the misspecification and derive a finite-sample upper bound on its failure probability. Our results illustrate when using a misspecified finite hypothesis class can provide an advantage over identification without prior knowledge of the candidate systems.

0.34

Quantifying Linepack Storage Potential of Hydrogen Networks

eess.SY ·2026-10-29 ·Ulrike Herrmann, Hans Würfel, Albertus Johannes Malan, Anton Plietzsch, Johannes Schiffer
Abstract · 23 pages, 6 figures, submitted to the International Journal of Hydrogen Energy

Future hydrogen systems depend on large-scale storage to balance supply and demand. Hydrogen networks can store energy in the pipelines themselves, but the potential of this linepack storage has not been quantified comparably to dedicated storage facilities. We transfer metrics established for conventional storage, such as volume capacity and full-cycle equivalents, to hydrogen networks. To evaluate them, we formulate an optimization problem maximizing linepack storage utilization, subject to non-convex PDE-based constraints describing transient gas flow. We solve the optimization problem over a full-year horizon to capture seasonal variations and to assess the interaction with cavern storage. For computational tractability, we propose a simulation-informed sequential convex programming approach with adaptive time stepping that reduces problem size by up to 98% at accuracy comparable to fixed-resolution methods. For the GasLib-40 topology with German scenario data, the network provides a volume capacity of 24 million normal cubic meters at 28 full-cycle equivalents per year.

0.34

AcrossDEX 1.0: From Human Motion Pretraining to Bimanual Dexterous Manipulation across Hand Embodiments

cs.RO ·2026-10-29 ·Xiaoyu Yang, Nan Wu
Abstract

Bimanual dexterous manipulation requires continuous coordination of wrist motion, finger contact, and object motion. Human videos provide rich manipulation experience. Using this experience for robot control requires aligning human poses with robot commands and accounting for embodiment-specific kinematics and contact dynamics. We present AcrossDEX 1.0, a learning system that connects human motion pretraining, simulation-based expert learning, robot demonstration collection, and visual closed-loop execution. Pretraining on HOT3D Aria predicts one second of bimanual motion from five images and thirty annotated hand states over the preceding second. Robot adaptation preserves the visual encoder, temporal backbone, and wrist parameters. Shared wrist representations, separate finger fields, and validity masks support both six-channel Inspire hands and sixteen-joint Allegro hands. The robot policy predicts sixty control targets from frontal images, command history, and task conditions. It executes a prefix and replans from updated observations. We compare direct regression, diffusion, and flow matching on lid opening and return and closed-lid object reorientation. Direct regression achieves success rates of 40.6\% for opening and return, 100.0\% for Inspire reorientation, and 93.3\% for Allegro reorientation. Flow matching achieves 90.0\% on Allegro reorientation. Paired comparisons further show that constraining replanning with the previous plan improves the continuous execution of generative policies. These results demonstrate a common adaptation pipeline for two hand embodiments and highlight the roles of contact coordination and action-chunk continuity.

0.33

Integrated Sensing and Communication via Channel Knowledge Maps Built from Beam Feedback Coefficients

eess.SP ·2026-10-29 ·Chongda Huang, Yue Xiao, Gang Wu, Vladimir Poulkov, Xianfu Lei +1 more
Abstract

Channel knowledge maps (CKMs) support wireless localization by linking channel features to user locations. Detailed channel fingerprints, however, can require substantial storage. This paper proposes a CKM-based localization framework that uses beam feedback coefficients collected during communication. These coefficients serve as compact location fingerprints within the beam management process, without requiring full channel feedback or additional sensing measurements. Using a coarse position prior, the base station searches nearby CKM entries and refines the location estimate through local interpolation. We analyze the relationship between spatial fingerprint variation and localization accuracy, and evaluate the effect of feedback quantization. Simulations show favorable storage--accuracy trade-offs relative to dense and sparsified spectral fingerprinting in the tested multipath environments.

0.33

A Temporal Cellular Sheaf Framework for Multi-Agent Systems with Switching Communication Topologies

eess.SY ·2026-10-29 ·Brandon Fallin, Wilmer Leal, Austin Copeland, Benjamin Merlin Bumpus, James Fairbanks +1 more
Abstract · 10 pages, 4 figures

Cellular sheaves serve as a natural modeling framework for encoding heterogeneity in the state and interaction spaces of multi-agent systems. This work examines switching communication graphs with the use of temporal cellular sheaves, which we construct to represent heterogeneous agent state spaces, the maps that relate them, and their respective communication constraints. We then develop sufficient conditions under which a multi-agent system defined on a temporal cellular sheaf can solve the target tracking problem by generalizing the existing concept of joint connectivity. Using a Lyapunov-based stability analysis, we demonstrate that the target tracking error of the network converges to a neighborhood of the origin.

0.33

CuBot-STT: Simultaneously Transport and Tighten for Lattice-type Modular Reconfiguration Robot

cs.RO ·2026-10-29 ·Chongxi Meng, Da Zhao, Yanmin Zhou, Runjie Shen, Zhipeng Wang +1 more
theory
Abstract · 11 pages. Submitted to IEEE/ASME Transactions on Mechatronics

Lattice-type modular reconfigurable robots rearrange standardized modules across tasks. Dedicated transport robots reduce module weight and cost by eliminating module level reconfiguration actuators. But existing systems use low capacity connections or internal fastening robots that occupy space for functional components. This paper presents CuBot-STT, which integrates module transport, interface locking, and release in one external robot. Cubic modules provide structural support and $24\,\mathrm{V}$ power distribution. On each face, a gear transmission converts one external input into motion of four helical locking points, preserving internal space. The robot uses alternating foot support, a gripper for module handling, and a tightening head for interface actuation. The controller coordinates two-foot support and module locking through continuous constraint transfers. The planner decomposes relocation costs into an analytical lower bound and a residual learned through expert-guided, goal conditioned Q-learning. Contact face $A^*$ search expands relocation sequences into robot primitives satisfying connectivity, asymmetric manipulation, and foot stance constraints. Experiments demonstrate normal and tangential interface capacities of $300\,\mathrm{N}$ and $490\,\mathrm{N}$, electrical continuity across six modules, and autonomous reconfiguration of an 11-module lattice. Planning benchmarks show superior search efficiency over analytical guidance and bidirectional heuristic search, with generalization beyond training lattice sizes without fine-tuning.

0.32

Tracking analysis of compressed-communication distributed adaptive estimation for time-varying parameters

eess.SY ·2026-10-29 ·Die Gan, Shuning Chen, Ying Wang
Abstract

This paper studies the problem of distributed time-varying parameter tracking over communication-constrained networks. To mitigate the accumulation of induced compression errors across the network, we propose a differentially compressed distributed least mean squares (LMS) algorithm.In the proposed scheme, each sensor updates a local auxiliary state recursively based on the compressed difference between the current parameter estimate and the previous auxiliary state, and then transmits only this compressed difference to its neighbors for auxiliary-state reconstruction. This differential recursion mechanism helps control the accumulation of compression errors and accommodates a general class of randomized compressors including randomized quantizers and sparsifiers.We establish the mean-square boundedness and explicit steady-state tracking-error bound under a cooperative excitation condition on stochastic regression vectors, which captures both temporal and spatial information to reflect the joint excitation effect of sensors.Notably, our stability analysis requires neither temporal independence nor stationarity of regressors, nor persistent excitation at individual nodes, thereby accommodating correlated and nonstationary signals from stochastic feedback systems. Numerical simulations demonstrate the theoretical results under correlated and non-stationary regression data.

0.32

Temporal and spatial couplings in the dynamics of human walking and running

eess.SY q-bio.QM ·2026-10-29 ·Mustafa Mert Ankarali
Abstract

Rhythmic behaviours such as walking and running are commonly modelled as dynamical systems around a stable limit cycle. To keep such models tractable, it is usually assumed that each step depends only on the previous one, that the limit cycle is fixed, and that the sagittal and frontal planes can be studied separately. Here, we test these assumptions in seven public datasets of human gait (71 participants), including 16 people who both walked and ran. We fit linear step-to-step maps and validate them by cross-validation, holding out more data where more data exist. At first sight, gait has memory: the state two steps back lowers the prediction error by up to 29\%. Yet most of this temporal coupling vanishes once the operating point may drift slowly and measurement noise is modelled; gait remains limit-cycle-like, around an operating point that moves. Running keeps more memory than walking in the same people. Spatial coupling between the planes is present in most participants, mainly from sagittal to frontal, but weak. This supports planar models of healthy gait, while a repeatable, task-dependent coupling index may serve as a measure of gait quality. The simple models are wrong, but useful enough.

0.32

Readiness-Constrained Pre-Islanding Dispatch for Military Microgrids

eess.SY math.OC ·2026-10-29 ·Aidan Looney, Annie Weathers, Nicholas Judson
Abstract · 32 pages, 5 figures, 3 tables, submitted to Military Operations Research

Military installation energy readiness depends on having sufficient resources to sustain mission-critical loads throughout a prolonged utility outage. This paper develops a method to help operators manage energy resources during normal grid-connected operation while preserving islanded readiness. A robust viability kernel identifies combinations of battery energy, fuel, and load-shifting capability sufficient to sustain critical load under uncertain demand and photovoltaic generation. A probability bound translates stochastic generator start failures and recovery into a deterministic generator count selected to meet a specified availability requirement at islanding. Case studies for a fourteen-day islanding requirement at 0.999 islanding-entry availability show that advance warning of an outage can be used to prepare resources to meet a higher readiness requirement before islanding begins. The case studies further demonstrate that economic-only operation and the evaluated fixed battery reserve policy can fail to sustain critical load for the required duration, whereas enforcing the resource-dependent requirement preserves readiness at a modest increase in modeled grid-connected operating cost. These results demonstrate how active resource management can preserve long-duration energy readiness. The framework also helps commanders understand how readiness changes with operating conditions, even when the available resources remain the same.

0.32

Implementation and Validation of Mismatch-Tolerant High-Gain OTA with a Calibrated Negative Conductance Load

eess.SP ·2026-10-29 ·Mayank Aggarwal, Naveen Kadayinti
Abstract · 6 pages, 13 figures

This paper presents the circuit-level implementation of a mismatch-tolerant high-gain operational transconductance amplifier (OTA) using automatic negative-conductance calibration in 65-nm CMOS. The technique enhances OTA gain by compensating for its finite output conductance using a cross-coupled transistor pair, with an on-chip calibration loop that eliminates the need for a replica circuit or external tuning. The calibration routine increases the gain by driving the steady state error of the OTA in a unity gain feedback loop toward zero. The implementation addresses practical challenges involving comparator offset, DAC design, and subtractor gain and offset. A 6-bit capacitive DAC (CDAC)-based comparator offset calibration scheme enables reliable detection of the sub-millivolt steady state error required for high OTA gain. Post-layout Monte Carlo simulations show that the comparator offset standard deviation is reduced from 4.11 mV to 132 $μ$V, with a maximum observed post-calibration offset of 625 $μ$V. The complete calibration circuitry is implemented and evaluated through post-layout simulations.

0.32

Optimal Path Planning with Localization constraints

cs.RO ·2026-10-29 ·Chinmay Mundane, Manav Mishra, PB Sujit
hot direction
Abstract

This paper studies UAV path planning in areas where GPS signals are not present. Because these drones can easily get lost when they don't have GPS, they must rely on dead reckoning and occasionally pass near cellular towers, which act as landmarks to reset and fix their position. The goal is to find the shortest path from a start to a goal location while keeping the estimation uncertainty within a given bound. To solve this problem, we propose a localization-constrained A* planner that treats uncertainty as a hard constraint during search. The robot state and covariance are propagated using an Extended Kalman Filter (EKF), and only paths that satisfy the covariance limit are considered. We also show that the proposed method does not affect heuristic used in A* and remains admissible and consistent, ensuring optimality among all feasible paths. Extensive Monte Carlo simulations and PyBullet quadrotor test show that the proposed method finds shorter and more reliable paths compared to RRT* and PRM, while always satisfying the uncertainty constraint.

0.32

Control of Seizure Dynamics in a Bistable System with Pulsatile Inputs

eess.SY ·2026-10-29 ·Xianlong Zhao, Yuzhen Qin
unusual
Abstract · 7 pages, 2 figures

Bistable neural population models describe transitions associated with epilepsy through the coexistence of a stable equilibrium representing normal neural activity and a stable limit cycle representing seizure-like oscillations. Perturbations that drive the system outside the equilibrium's basin of attraction can trigger transitions to seizure-like activity. Enlarging this basin enables the system to recover normal activity from a larger set of perturbed states, providing a model-based mechanism for reducing frequency of seizure initiation. Motivated by electrical brain stimulation, this paper investigates basin enlargement using pulsatile feedback control. Combining higher-order averaging with periodic near-identity transformations, we establish conditions under which nonzero-mean and zero-mean pulsatile feedback gains enlarge the basin of attraction and derive guidelines for pulse design. Numerical simulations validate the theoretical enlargement.

0.32

Common Offset Gathers for Efficient Multistatic Beamforming in Frequency Domain

eess.SP physics.med-ph ·2026-10-29 ·Marko Jakovljevic, Xiaohong Wang, Ion Candel, Anthony E. Samir, Kai Thomenius
Abstract · 11 pages

Frequency-domain beamformers reconstruct ultrasound images with far fewer operations than delay-and-sum (DAS). However, for multistatic synthetic aperture data, they require Fourier transforms over the full transmit-receive aperture and large amounts of memory, limiting their use for volumetric imaging. Here, we adapt common offset gathers (COG) from seismic imaging to frequency-domain beamforming: transmit-receive pairs with the same offset are beamformed independently using a bistatic formulation, holding only one offset gather in memory at a time. We combine COG with the range-Doppler algorithm (RDA) and evaluate COG RDA against DAS and conventional multistatic RDA using simulated, phantom, and in vivo acquisitions. For linear arrays, COG RDA reduced memory requirements by two orders of magnitude relative to conventional RDA and beamformed up to 25x faster than DAS, with comparable image quality. For a 32x32-element matrix array, COG RDA requires only 2-D instead of 4-D Fourier transforms, and reconstructed a volume on a GPU 32x faster and with 67x less energy than DAS. COG RDA makes frequency-domain multistatic beamforming practical for matrix arrays and power-constrained devices.

0.32

LLM eBay: A Simple and Truthful Marketplace for Fine-Tuned LLMs Trading

eess.SY ·2026-10-29 ·Jiachen Shen, Xingke Yang, Hui Zhong, Aohan Li, Tomoaki Ohtsuki +3 more
Abstract

Personalized fine-tuning of Large Language Models (LLMs) enhances user experience but incurs substantial computational costs. Recent studies show that initializing training from a style-similar fine-tuned model reduces adaptation overhead, turning such checkpoints into reusable, tradable assets. A marketplace for these checkpoints, however, faces information asymmetry: sellers may misreport a checkpoint's style, while buyers cannot verify its usefulness before purchase, which invites adverse selection. We introduce LLM eBay, a simple single-round marketplace for trading fine-tuned LLMs. Sellers are paid per purchase at a posted price, and a truncated quadratic proper scoring rule adjusts each payment according to how well the seller's style report agrees with a style signal that the platform measures by generating text on hidden prompts. We show that the score elicits the verifier's expected signal, characterize exactly when truthful reporting is a seller best response under per-purchase demand, and bound the downside risk from noisy verification. In an end-to-end study with real LoRA checkpoints on two base models and three buyer tasks, LLM eBay reduces measured convergence steps by 8.5\% relative to retrieval over self-reported style (95\% CI [1.0\%, 15.9\%]). In model-based market simulations built on a measured style-to-cost relationship, it cuts style-claim error by 96.5\% (from 0.424 to 0.015), recovers 96.3\% of Oracle welfare, saves 72.4\% of the cost of fine-tuning from the base model, and keeps all providers profitable in the default setting.

0.31

Addendum to "Optimal Steering of a Linear Stochastic System to a Final Probability Distribution, Part II"

eess.SY math.OC ·2026-10-29 ·Yongxin Chen, Tryphon T. Georgiou, Michele Pavon
theory
Abstract · 4 pages

We hereby fill in a gap in an induction argument used in the proof of a theorem on state-covariance controllability of linear stochastic systems in an earlier paper by the authors. The mathematical statement is true. The missing point in the proof concerns the assertion that, in the canonical single-input case, the control can be assigned arbitrary endpoint values. We amend the argument by proving a slightly stronger induction statement that permits arbitrary finite endpoint jets of the control. An independent approach to homogeneous state-covariance controllability was provided in a paper by Abdelgalil and Georgiou (2025) as a corollary of strong controllability results on $\GLp(n,\mathbb R)$, and shown herein that it extends to the case of additive noise as well.

0.30

Analog Realization and Measurement of a Forced Aliev-Panfilov Excitable Oscillator

eess.SY nlin.CD ·2026-10-29 ·Zixuan Fu, Unnikrishna Pillai
unusual
Abstract

A single resistor moves a discrete analog realization of the forced Aliev-Panfilov excitable oscillator from a chaotic operating point of the model to a regular one. In this model a rate parameter multiplies the complete right-hand side of the excitation equation, forcing term included. The input resistor of the excitation integrator sets that parameter, so no state, delay element, or controller is added. Sampled waveforms were recorded at two resistor values, with two independent runs each. The measured excitation excursion agrees with numerical integration to within three percent at the baseline setting and five percent at the rate-controlled setting. Between the two settings the excitation excursion contracts by a factor of thirteen. At the rate-controlled setting the recovery excursion exceeds prediction by a factor of eight. This excess traces to a displaced operating point on the state-dependent recovery rate. Thermal drift in the multiplier chain and leakage at the high-impedance recovery-rate divider could each produce it, and the present data do not separate them. A tested bipolar input stage strongly loads that divider, a constraint absent from the ideal equations. The divider, not the sub-hertz signal bandwidth, sets the amplifier input-stage requirement.

0.30

Securing the Robotics Supply Chain

cs.RO ·2026-10-29 ·Cindy L. Bethel, Henrik I. Christensen, Odest Chadwicke Jenkins, George Pappas, Holly Yanco +1 more
unusual
Abstract

This paper sets out what a sustained federal research and investment agenda would need to include to close that gap--not within a single budget cycle, but over the decade such an undertaking will realistically require. It identifies where the gap is most pronounced today: in the methods needed to verify and certify components of uncertain origin, in the foundational hardware research that has long been underfunded relative to software, in the physical infrastructure researchers need to test systems at scale, and in the workforce required to build and sustain all of it. Robotics should not be treated merely as a downstream beneficiary of a broader artificial intelligence investment strategy. It constitutes its own supply chain, with its own points of failure, and merits its own sustained federal commitment.

0.30

Information-Driven Collaborative Autonomy for Heterogeneous UAV/UGV Teams

cs.RO eess.SYmath.OC ·2026-10-29 ·Joshua Yuriv Pilipovsky
hot direction
Abstract

This paper considers coordinating a heterogeneous team of unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs) traversing a domain containing an unknown number of potentially hazardous targets. The UAVs provide aerial sensing using onboard optical cameras, adapting their trajectories to ground routes and target information. We formulate the aerial search problem using ergodic control, which distributes sensing effort according to a spatial information density to balance exploration and target localization. The formulation incorporates dynamic camera field-of-view footprints and pose-dependent probabilistic detection models. Angles-only measurements are processed through Bayesian inference and marginalization over data-association hypotheses to obtain a Poisson multi-Bernoulli (PMB) approximation of the target posterior, representing both detected and undetected targets. This belief directs aerial sensing toward unexplored regions along the ground routes and viewpoints that improve target localization. Ground planning combines chance-constrained target avoidance, enforced through confidence sets, with a void-probability objective that favors routes unlikely to contain threats. We solve each team's optimization problem in a fully decentralized manner using a nested SCvx*--C-ADMM scheme, with outer sequential convex programming iterations and inner distributed consensus optimization, accounting for nonlinearities, nonseparable objectives, and continuous-time inter-agent separation constraints. A closed-loop scheme updates both fleets' trajectories as information becomes available and permits ground motion when the remaining route risk falls below a prescribed threshold. Numerical case studies examine sensing geometry and target uncertainty, compare decentralized and centralized solvers, and demonstrate full closed-loop collaboration between multiple UAVs and UGVs.

0.29

SwapBot: Tool-Centric Robot Manipulation via Interchangeable End-Effectors and Tactile Sensing

cs.RO ·2026-10-29 ·Yunchu Zhang, Ray Song, Junhan Jiao, Liyiming Ke, Yuchen Yang +4 more
Abstract

Robot manipulation in everyday environments inherently involves substantial tool usage. Seemingly simple household tasks, such as cooking pan-fried tofu, require forceful cutting, delicate brushing, and precise placement using different tools under contact-rich interaction. Thus, the ability to autonomously swap and use task-appropriate tools is important for tool-centric manipulation. We introduce SwapBot, a modular, tool-centric end effector system featuring low-cost adapters that quickly convert common household tools into robot end effectors. Rather than pursuing a single universal gripper, SwapBot adopts a unified design principle that combines a fixed robot-side coupling interface with low-cost, tool-side adapters, enabling compatibility with diverse tools, including articulated ones, and multiple robot platforms. Each adapter integrates compact tactile sensors, supports fast autonomous swapping, and interfaces with an intuitive teleoperation system for efficient demonstration collection. User studies show that, during teleoperation, SwapBot reduces task completion time and operator workload. Using collected demonstrations on the robot, we train imitation learning policies to successfully perform tool-centric tasks like cooking tofu, stirring salad, flipping meat. Our ablation study shows that using tactile-augmented demonstrations improves policy performance across contact-rich skills, increasing success rates by 15 to 25 percentage points over RGB-only policies. These results highlight modular, tactile-equipped tools as a practical and extensible pathway toward robust and capable robot manipulation. Project website https://personalrobotics.github.io/SwapBot.

0.29

Limit Theorems and Optimal Policies for Flock Flight through Poisson Forests

eess.SY cs.ROmath.PR ·2026-10-29 ·Lav R. Varshney
theory
Abstract

Inspired by birds flying through cluttered environments, Karaman and Frazzoli showed that high-speed flight through an ergodic random forest exhibits an ultimate speed limit: above a critical speed, almost surely no infinite collision-free trajectory exists for a single bird. This paper formulates a corresponding bulk-transport problem for many birds. We construct an explicit directed corridor network in a Poisson disk forest and impose a common temporal headway at its waypoints. The maximum current equals a minimum vertex-cut capacity and is achieved by repeated flight along vertex-disjoint paths. The same schedule maximizes completed traversals at every finite horizon, with constant forward speed and no internal waiting. For a forest belt of fixed depth, exact transverse superadditivity yields an almost-sure deterministic current-density limit. A uniform shape theorem for an asymmetric planar dual identifies the current when depth and width grow together. For diverging depth bounded by a constant times width, this current equals the width-first, depth-second limit and is independent of the positive aspect ratio. When diagonal corridors are available, bulk current is positive at sufficiently low obstacle intensity and zero at sufficiently high intensity. For stochastic demand, an upstream admission queue is stable at every rate below the finite-network capacity.

0.29

A Conveyor Belt Mechanism for Repositionable Magnetic Coupling Between Continuum Robots

cs.RO ·2026-10-29 ·Ella Walsh, Eric Diller, Jessica Burgner-Kahrs
Abstract · 7 pages, 10 figures

This work proposes a novel conveyor belt mechanism enabling continuous, active repositioning of functional elements along the body of a tendon-driven continuum robot. Up to three conveyor belt mechanisms are integrated within a 9 mm diameter continuum robot. We demonstrate equivalent or reduced conveyor belt actuation forces compared to robot tendon actuation forces. As a case study, we mount permanent magnets along two conveyor belts, and demonstrate actively reconfigurable inter-robot coupling. Coupling point location and quantity can be adjusted during robot operation to actively tune the effective stiffness of the coupled multi-robot system. We show that effective stiffness follows a predictable inverse-cubic relationship with coupling location, and increases with the number of coupling points. These findings demonstrate that the conveyor belt mechanism enables active functional element repositioning for changing tasks requirements, with magnetic inter-robot coupling as one proof-of-concept.

0.28

Incremental Stability of Gradient Flows under Generalized Polyak-Łojasiewicz Inequalities and Its Implications for General Systems

eess.SY math.OC ·2026-10-29 ·Andreas Oliveira, Arthur C. B. de Oliveira, Mario Sznaier, Eduardo Sontag
Abstract

Gradient flows of twice continuously differentiable strongly convex and convex objectives are, respectively, incrementally exponentially stable and non-expansive in the Euclidean metric. In this work, we investigate to what extent these properties extend to nonconvex objectives satisfying generalized Polyak-Łojasiewicz inequalities. For objectives with a unique minimizer, we establish conditions under which the gradient flow is contracting in a state-dependent Riemannian metric and incrementally exponentially stable. For objectives with nonunique minimizers, we generalize the non-expansivity of convex objectives to incremental stability for more general nonconvex objectives. We focus on incremental stability due to its implication to convergence to equilibrium points; when paired with a Lyapunov function it implies pointwise convergence to a minimizer and continuity of the endpoint map. Along the way, we derive new structural results for incrementally stable nonlinear systems and extend the analysis to Heavy-Ball and Dynamical Inertial Newton flows.

0.28

UMI-Live for In-The-Wild Bimanual Robotics Data Collection

cs.RO cs.AIeess.SY ·2026-10-29 ·Govind Velamoor, Adrian Lin, Sam Jung, Spencer Mateega, Brandon Y. Yang +1 more
code/data
Abstract

Universal Manipulation Interface (UMI) devices allow human experts to demonstrate tasks for robot learning without teleoperating physical robots. Using these demonstrations as training data still requires a consistent hardware setup, calibration, and data processing. We present UMI-Live, a full-stack open-source bimanual UMI system for in-the-wild data collection. We design our custom UMI-Live hardware to integrate camera and LiDAR--inertial sensing, onboard compute, and power, without external tracking equipment or calibration within the collection environment. We implement a simple user interface for recording data while an asynchronous pipeline performs parallel SLAM reconstruction and publishes training-ready datasets. We also release software that supports dual-YAM teleoperation and VLA policy training and inference. We validate UMI-Live data quality by fine-tuning $π_{0.5}$ and MolmoAct2, achieving success rates of 10/10 and 8/10, respectively on a bimanual pick-and-place task. Furthermore, we release 40 hours of UMI-Live data across three tasks, and complete hardware designs, software, and reproduction guides at our $\href{https://umi-live.github.io/}{\text{project website}}$.

0.28

The Athlete as a Scientist of One: Agentic Bayesian Experimental Design for Individualized Physiology from Consumer Wearables

cs.RO cs.AI ·2026-10-29 ·Behnam Moradi
benchmark
Abstract

Large language model (LLM) agents for personal health read wearable data and recommend, but do not learn about the individual in a way that can be audited or falsified. We present EpiCoach, an agentic system that treats one athlete, a cross-country mountain biker, as the subject of accountable N-of-1 science. Claims about the athlete live in an evidence-gated hypothesis registry: each must issue a probabilistic prediction about a future session before it occurs, and its belief moves only through computed evidence from a logged statistical fit. Predictions come from a mechanistic digital twin inferred from heart rate, GPS and barometric altitude: a lagged heart-rate response model, a fitness-fatigue model with learned time constants, and a latent technical-descending skill state, measured on discovered trail segments. The agent prescribes the next training week as a constrained experiment maximizing information gain subject to load and safety constraints, around two instrument rides. We evaluate the system two ways over eighteen months of data. A rolling-origin replay benchmark shows the twin beats standard forecasters on heart rate but not segment time, and the agentic loop matches a static model. A live prospective phase runs the weekly cycle: in its first three weeks the ledger sealed 20 predictions, scored 14, the descent forecast calibrated after one correction, the climb forecast pessimistic as the athlete improved, and the safety officer vetoed the agent twice. Every correction is preserved in the registry, a template for auditable, falsifiable personalization in digital health. We release the registry, benchmark and protocol.

0.28

LC-Resonance Placement Limits and Damping Recovery in Virtual-Admittance-Based Grid-Forming Inverter Control

eess.SY ·2026-10-29 ·Stephen Arinze Obi, Jae-Jung Jung
Abstract · 10 pages

Compact LC-filter design and inner-loop control must be coordinated for stable grid-forming operation. This paper quantifies how current-loop proportional gain constrains LC-resonance placement in virtual-admittance-based dual-loop control with unity voltage-feedback decoupling (VFD) and capacitor-voltage active damping. Equivalent-impedance analysis identifies gain-dependent contraction of the positive-resistance region. Delayed inverter-current feedback opposes VFD-assisted damping above one-sixth of the sampling frequency. Simple proportional-path frequency shaping recovers an admissible effective-gain window while retaining unity VFD and leaving the resonant path unchanged. Reduced-model criteria, checked against the complete model, link window closure to a nominal resonance-placement limit near one-third of the sampling frequency under single-sampled single-update operation over the studied gain range. A 3-kVA prototype demonstrates an 83.3% LC-product reduction at comparable voltage and nominal current crossovers. A separate comparison with the same high-resonance filter demonstrates stable operation at 2.68 times the unshaped nominal current crossover. Output-impedance analysis and grid-connected tests identify additional grid-dependent gain constraints.

0.28

Beyond Diagonal RIS for Physical-Layer Security: Low-Complexity Solution in the Low-SNR Regime

eess.SP ·2026-10-29 ·Rawan Derbas, Hamad Yahya
Abstract · Submitted for publication in IEEE journals

This letter investigates beyond diagonal reconfigurable intelligent surfaces (BD-RIS) for physical-layer security (PLS) in correlated wiretap channels. A multi-antenna transmitter is considered, and transmit antenna selection (TAS) is used to exploit transmit-side diversity with a single active radio-frequency (RF) chain. Using the first-order secrecy rate approximation, we develop a two-stage scheme to maximize the secrecy rate in the low signal-to-noise ratio (SNR) regime. In the first stage, two TAS schemes are developed based on the effective channel upper-bound. In the second stage, an eigenvector-based BD-RIS scattering matrix is designed for single-connected, group-connected, and fully-connected architectures. Unlike legitimate-user-centric or eavesdropping-nulling designs, the proposed approach balances legitimate-user signal enhancement and information-leakage suppression without forcing zero leakage. Simulation results show improved secrecy performance, particularly under a strong correlation between legitimate and eavesdropping channels.

0.28

Adversarial Pilot Injection Against AI-Based CSI Feedback in FDD Massive MIMO Systems

eess.SP ·2026-10-29 ·Jinli Zhai, Yusen Wang, Qimei Cui, Zengbao Zhu, Shengyuan Liang +1 more
Abstract · Submitted to IEEE Internet of Things Journal

Artificial intelligence (AI)-based channel state information (CSI) feedback reduces overhead in frequency division duplexing (FDD) massive multiple-input multiple-output (MIMO) systems, yet its reconstructed CSI depends on the integrity of pilot acquisition. This paper identifies and evaluates the attack path of adversarial pilot injection against this feedback. A transmitted disturbance can damage the user-side estimation before encoding; correctly delivered quantized bits can still represent a contaminated channel estimate. We model this path through attacker propagation, linear minimum mean square error estimation, compression, quantization, and decoding under a total transmit-energy constraint. An adversarial perturbation generator (APG) implements the attack using channel statistics without current legitimate CSI or instantaneous attacker-channel observations at deployment. At a Rician attacker-channel factor of 10~dB and a transmit perturbation-to-signal ratio of $-4$~dB, APG produces a normalized mean-square error (NMSE) of 0.84~dB in base-station CSI reconstruction, compared with $-10.31$~dB for Gaussian interference. Estimator-targeted and reconstruction-targeted attacks yield different error orderings, and reconstruction damage also degrades beamforming. As a countermeasure, alternating adversarial training (AAT) adapts the feedback encoder and decoder to contaminated pilots. With condition-matched training and APGs reoptimized for the final receiver, AAT reduces NMSE by 2.40--3.79~dB across six representative channel--power conditions. The results demonstrate the importance of protecting pilot acquisition alongside quantized feedback delivery.

0.28

Confidence-Aware Speed-over-Ground Estimation from Doppler Spectra via Reconstruction-Based Representation Learning

eess.SP cs.LG ·2026-10-29 ·Malte Stelzer, Marvin Sach, Timo Bartels, Dinu Mihailescu-Stoica, Sören Kohnert +3 more
Abstract

Reliable speed-over-ground estimation from ground-facing Doppler spectra is essential for mobile platforms such as railway systems. Classical signal-processing approaches can provide accurate estimates under favorable conditions, but often degrade in challenging conditions. We propose confidence-aware speed-over-ground estimation trained on rail-domain data that estimates speed over ground and frame-level confidence from Doppler spectra with several novel contributions: First, we transfer the idea of exploiting a reconstruction task by an autoencoder for performance estimation from previous work to Doppler RaDAR. Second, we append a speed-over-ground estimation head operating on the autoencoder's latent space features. Third, we omit the reconstruction decoder during inference but add a low-cost dedicated confidence head for frame-wise error-bound estimation. Fourth, we reveal how noise augmentation training can provide high robustness to the method. Experiments show that our approach improves speed estimation accuracy over classical baselines while providing informative confidence estimates that indicate the reliability of individual predictions, even under adverse signal conditions.

0.27

Anchor-Free Range-Free 3D IoT Localization in Buried Environments with Connectivity-Informed Power Selection and Temporal Refinement

eess.SP ·2026-10-29 ·Peyman Moeini, Mark Coates
Abstract

In underground and surface mines, wireless Internet-of- Things (IoT) tags deployed with ore or muck loads can support material and grade tracking after burial, when conventional ranging and surveyed localization infrastructure may be unavailable. Building on the three-dimensional Pro- traction Technique (3D-PT), this paper introduces 3D-PT with Power Selection (3D-PT-PS), which selects informative transmit-power levels from binary pilot measurements, and 3D-PT-PS with Temporal Refinement (3D-PT-PST), which incorporates recorded load-deposition times as weak, non- metric cues. Both methods are anchor-free and range-free and require no RSSI, measured ranges, surveyed anchors, pile dimensions, or material parameters. Monte Carlo simulations evaluate different pile sizes and tag densities. In a 5000-t crusher-pocket scenario with 125 tags, 3D-PT-PST achieves 6.12% mean NRMSE and 2.20 m 3D RMSE, reducing mean NRMSE by 13.5% relative to 3D-PT-PS and by 43.6% and 69.1% relative to calibrated MDS-MAP(P) and CATL, respectively. These results show that connectivity and timing provide complementary information.

0.27

LLM-guided Planar Loco-Manipulation for Quadruped Robots

cs.RO cs.LG ·2026-10-29 ·Georges Jetti, Michael Khayyat, Jean-Pierre Sleiman, Stefano Arrigoni, Francesco Braghin
Abstract · 20 pages, 9 figures, 15 tables. Includes appendix

Robotic navigation typically assumes a collision-free route to the goal exists; in cluttered environments with movable obstacles, however, reaching the goal may require physical interaction. Classical solutions encode these behaviors through geometric rules and task-specific logic, while learning-based approaches often constrain the decision space to low-dimensional motion commands or predefined action sets. We instead consider a broader decision space, jointly selecting the skill, relevant object, and geometric target. Our language-conditioned task planner, fine-tuned with LoRA on synthetic data, maps natural-language tasks and a structured scene representation to variable-length sequences of parameterized actions, executed by low-level reinforcement-learning policies. The planner learns the geometric relations underlying interactive navigation and outperforms a classical planner and a prompted frontier model on interactive navigation tasks. Comparisons with models trained from scratch indicate that composition does not require pretraining when the representation is structured, whereas pretraining yields more reliable geometric targets. At the low level, we introduce an end-to-end whole-body planar pushing policy that achieves a 93.3% success rate in simulation, versus 83.5% for a hierarchical baseline, while reducing completion time by more than 60%. On hardware, the hierarchy yields the correct skill composition in 44/45 trials and end-to-end success in 34/45. Project website: https://anonymousprojects-dev.github.io/LLM-guided-LocoManipulation-Quadrupeds/

0.27

AFast Robot Localization Method Based on Monte Carlo Sampling

cs.RO ·2026-10-29 ·He Zhaoqing, Hang Li, Xian Pu, Li Wang
Abstract

Localization is an essential capability of intelligent robots. However, in real-world constrained environments, robots often cannot rely on external localization systems. Therefore, this paper proposes an online particle-filter localization method based on Monte Carlo sampling. The method is applied to the localization of a mobile robot in a maze environment and requires only four range sensors mounted on the front, rear, left, and right sides of the vehicle, without the need for any heading information. Simulation experiments demonstrate that even in a maze where local regions are highly similar, the vehicle can still achieve relatively accurate localization, thereby providing a fast and effective approach for robot localization in complex environments.

0.26

Measuring the Value of Communication in Multi-Agent Tracking Under Finite-Rate Belief Sharing

eess.SP cs.IT ·2026-10-29 ·Danial Dehghani, Maximilian Lübke, Zahra Seifaei, Norman Franchi
Abstract

This paper investigates the value of communica- tion (VoC) in cooperative multi-agent tracking under rate- constrained information exchange. Although prior work has studied distributed tracking, quantized communication, and robust fusion, it remains unclear at what communication rate most of the achievable tracking gain is obtained, beyond which additional communication resources yield only marginal benefit. We consider stationary agents that observe a moving target through heterogeneous range and bearing measurements, perform local extended Kalman filtering (EKF), quantize their posterior state means, and fuse exchanged information via covariance intersection (CI). Performance is evaluated versus communication rate relative to no-communication baselines, a centralized oracle, and a CI-perfect reference. The results show that communication is not uniformly beneficial: very low rates can degrade performance due to coarse quantization, whereas moderate rates recover most of the achievable gain. In the considered scenario, increasing the rate from 4 to 12 bit/s per agent reduces the root-mean-square error (RMSE) from 13.93 m to 6.35 m, while a further increase to 32 bit/s per agent improves it only marginally to 5.93 m. This reveals a clear knee region at moderate rates, beyond which the remaining loss is mainly due to the conservatism of the fusion architecture. Overall, the proposed framework characterizes low-rate degradation, communication-efficient operation, and diminishing returns, while highlighting the joint role of rate and sensing geometry in determining communication benefit.

0.25

Human Activity Recognition using Ceiling-Mounted Multi-Radar Sensors

eess.SP ·2026-10-29 ·Ananya Prabhu, Aditi Site, Tarmo Lipping
Abstract

Continuous activity monitoring in assisted living should work without cameras or wearable devices. Radar meets this requirement, but most radar studies use a single sensor and activities recorded in isolation, and few deliver results in real time. We present an indoor monitoring system in which point clouds from three ceiling-mounted radar sensors are fused using density-based clustering, movement and posture features are derived for activity classification, and results are shown on a real-time dashboard. In a preliminary study with eight participants performing everyday activities as one continuous scenario, a random forest reached 74% frame-level accuracy with a random 80/20 split and 68% with leave-one-participant-out evaluation. Without position features, accuracy fell to 46% and 44%, indicating partial dependence on room layout, while window-level features without position reached 67% and 65%.

0.25

Evaluating Generalization and Long-Term Stability of Smartwatch-Based Cuffless Blood Pressure Monitoring

eess.SP ·2026-10-29 ·Hongda Huang, Xiao Peng, Xiaoyu Li, Shuailong Tang, Guangpu Zhu +4 more
published
Abstract · Accepted by the 2026 IEEE EMBS International Conference on Biomedical and Health Informatics (BHI 2026)

Smartwatch-based cuffless blood pressure (BP) monitoring is a promising low-burden solution for long-term cardiovascular management, but practical deployment remains constrained by model complexity, uncertain cross-dataset generalization, and insufficient evidence on long-term tracking stability. This study presents an efficient Signal-Context Gated Fusion Network (SCGBPNet) for cuffless BP estimation from photoplethysmography (PPG) signals. SCGBPNet integrates signal-level representations through self-attention and further incorporates subject-specific contextual information via a compact gated fusion module. On Aurora-BP, SCGBPNet achieved SBP/DBP mean absolute errors (MAEs) of 7.08/4.90 mmHg, compared with 6.44/4.89 mmHg under within-dataset validation on the private smartwatch dataset, indicating a modest within-dataset discrepancy. Cross-dataset transfer from Aurora-BP to the smartwatch dataset increased the MAEs to 8.98/6.47 mmHg, demonstrating substantial cross-dataset degradation. For long-term stability, the proposed method showed low week-to-week performance variation and a low daily decay rate, indicating good stability over a follow-up period of more than one month. Nevertheless, estimation accuracy deteriorated when BP deviated substantially from an individual's baseline. These findings indicate that favorable aggregate MAE alone should not be interpreted as evidence of deployment-ready performance. More systematic and stratified evaluation across dataset settings, BP ranges, and baseline-deviation levels is needed to assess the practical reliability of smartwatch-based cuffless BP monitoring.

0.25

OpenFlyV2: An onboard vision-language-action model for aerial perception and navigation

cs.RO ·2026-10-29 ·Junli Liu, Zhongrui You, Chenhui Li, Zhen Li, Yuyang Tian +7 more
datasetcode/data
Abstract · 12page, 6figures

Human cognition flexibly draws on a shared perceptual foundation to understand environments, attend to referenced entities, reason about spatial relationships, and act. Endowing aerial robots with an analogous functional organization remains challenging, as existing systems are often specialized for individual capabilities, dependent on offboard computation, or unable to sustain closed-loop operation on resource-constrained platforms. We present OpenFlyV2, a unified onboard vision-language-action architecture that uses a shared vision-language representation to support semantic understanding, language-grounded perception, spatial reasoning, and continuous aerial control within a single model. A compact action expert converts the shared multimodal context into short-horizon flight waypoints, while motion-aware temporal memory and action priors reduce the computational cost of closed-loop execution. We construct \dataset{}, a simulation-based dataset comprising approximately 70,000 navigation trajectories and 60,000 vision-language instruction--response pairs. OpenFlyV2 generates action chunks, each comprising 12 waypoints, entirely onboard at 10 Hz using an NVIDIA Jetson Orin NX. It transfers zero-shot from simulation to physical quadrotor flight, achieving a 48\% navigation success rate across four real-world environments without real-world aerial training data. Extensive results demonstrate that broad vision-language competence and continuous closed-loop flight can coexist within a single model on a resource-constrained aerial platform. Code will be made publicly available at https://github.com/Ideal-ljl/AeroBrain.

0.25

CAPTURER: A Dexterous Aerial Manipulator for Dynamic and Precise Grasping across Diverse Environments

cs.RO ·2026-10-29 ·Meng Wang, Jindou Jia, Sicheng Zhou, Zeshuai Chen, Kexin Guo +2 more
Abstract

As a representative form of embodied robotic interaction, aerial grasping requires close coordination between the aerial platform and manipulator. Most existing systems perform aerial grasping with static or slowly moving targets under fixed scenarios. Dynamic and precise grasping across different conditions introduces additional challenges, including target motion, localization drift, dynamic coupling, and external disturbances. In this work, we present CAPTURER, a dexterous aerial manipulator for centimeter-level grasping across diverse environments. To achieve the challenging goal, we first develop an onboard perception system for state estimation and target detection. Drift correction and global pose transformation are introduced to improve target localization during flight. A cooperative motion planner is proposed to coordinate the mobility of the aerial platform with the dexterity of the manipulator according to the relative target state. To maintain grasping accuracy under disturbances, we design a disturbance-rejection controller to compensate for dynamic coupling, model uncertainty, and external disturbances. The capability of CAPTURER is validated through indoor and outdoor experiments with both static and moving targets. The results demonstrate that CAPTURER achieves centimeter-level grasping accuracy for a marker pen measuring only 1.5 cm in diameter at flight speeds exceeding 1.5 m/s.

0.24

EgoContact: Evaluating Physical Intelligence with Egocentric Video

cs.RO cs.CV ·2026-10-29 ·Adam Deryło, Alexey Gavryushin, Daniel Cremers, Marc Pollefeys, Xi Wang
datasetcode/data
Abstract · Project page: https://egocontact.github.io

Egocentric video has emerged as one of the most scalable sources of data for learning generalizable manipulation priors. Yet, despite its growing use in pretraining robotic policies, it has been underutilized as a validation tool for physical intelligence, a key component of which is understanding where to make contact with an object to achieve an intended interaction. We present EgoContact, a large-scale egocentric dataset for evaluating the generalizability and precision of learned manipulation priors through future contact-point prediction. EgoContact contains 22.5K manually verified interaction events, each with a pre-contact observation, semantic intent, and pixel-precise contact locations. We find that current methods often struggle to predict precise interaction contacts and sometimes fail to localize the target object in diverse, in-the-wild scenes, even when allowed multiple predictions. Furthermore, we demonstrate the usefulness of EgoContact's annotated interactions for robotic manipulation: simply using EgoContact as the memory bank of an existing affordance-transfer policy improves grasp success from 14.2% to 31.2% in simulation and from 16% to 58% on a real robot.

0.24

Assuring Learning-Enabled Cyber-Physical Systems with Foundation Models: A Neuro-Symbolic Agenda for Testing, Verification, and Runtime Assurance

cs.RO eess.SY ·2026-10-29 ·Xi Zheng, Ivan Ruchkin, Hong Jin Kang, Tevfik Bultan, Vijay Ganesh +9 more
position
Abstract

Learning-enabled cyber-physical systems (CPS), including autonomous vehicles and UAVs, increasingly incorporate foundation models, yet fragmented assurance practices provide insufficient support for system-level guarantees. We argue that restoring engineering control across stochastic neural, symbolic, and physical boundaries requires a shift from model-centric evaluation to system-level assurance. We introduce the Foundation Model Assurance Stack, connecting three directions: (i) neuro-symbolic architectures that constrain foundation models through explicit symbolic world models, reasoning layers, and interfaces; (ii) semantically grounded, conformance-based testing that combines formal specifications and symbolic world models with foundation-model-assisted specification mining, scenario generation, and test synthesis; and (iii) compositional verification and runtime assurance that target formally tractable components and interfaces through modular contracts, equivalence checking, verified synthesis, uncertainty-aware reasoning, and runtime monitoring and enforcement. Symbolic abstractions establish assurance boundaries between stochastic learning and formal reasoning, connecting testing, verification, and runtime evidence across the system lifecycle. Foundation models serve as constrained assistants whose outputs are independently checked wherever correctness can be formally specified. We identify research challenges in specification engineering, abstraction fidelity, evidence composition, and lifecycle certification, and outline a staged agenda toward trustworthy, deployable, and certifiable learning-enabled CPS.

0.24

SimTO-Robust: Automated Design of Soft Grippers Robust to Grasp Pose and Object Geometry

cs.RO ·2026-10-29 ·Kurt Enkera, Joshua Pinskier, Marcus Gallagher, David Howard
Abstract

Robotic grippers operating in unstructured environments must robustly grasp objects of different shapes and sizes despite perception and positioning errors. Soft grippers that can achieve this robustness through intelligent mechanical design rather than sophisticated control software offer a promising solution. However, no existing computational design framework explicitly optimises soft gripper morphology for robustness across object geometries and grasp poses. We introduce SimTO-Robust, a generalisation of simulation-driven topology optimisation (SimTO) that simulates grasps across multiple objects and poses and aggregates their contact forces to guide gripper morphology optimisation. We generate 6,646 designs and physically evaluate 24 of them alongside 10 specialised SimTO designs, two rigid baseline grippers, and a baseline generalist soft gripper produced using conventional topology optimisation. All 37 grippers are tested in a pick-and-place task involving five feature-rich objects and six grasp poses, with external disturbances of increasing magnitude applied during the task. On average, SimTO-Robust designs complete pick-and-place more reliably than specialised SimTO designs and the rigid baselines. The best SimTO-Robust designs also withstood more of the applied disturbances than any other tested gripper. These results show that incorporating diverse grasp conditions into simulation-driven topology optimisation is a simple and effective approach to improving real-world soft gripper robustness. More broadly, they show that robust robotic grasping can be achieved through intelligent mechanical design rather than sophisticated sensing and control.

0.24

Which wind and solar modelling choices matter? Effects on capacity factors, energy droughts and European power system configuration

eess.SY ·2026-10-29 ·L. Karkossa, A. Grochowicz, M. J. Koivisto, A. Hahmann, M. VIctoria
Abstract

Europe's future power system will rely heavily on wind and solar power and must remain resilient during prolonged periods of low wind and solar generation, known as variable renewable energy (VRE) droughts. To capture such rare events, adequacy assessments increasingly draw on decades of weather data, but the conversion of weather into power is largely treated as fixed. It remains unclear which modelling choices in this conversion matter for representing these droughts, and whether they matter as much as the weather itself. Here we systematically compare these choices side by side (bias correction, wind speed extrapolation, capacity layouts, turbine and photovoltaic configurations, and wake losses) and trace their effects through VRE droughts to the design of a net-zero European power system, using 85 years of reanalysis data. We find that modelling choices barely change how often VRE droughts occur or when the system is stressed. They do, however, determine how much capacity is needed to withstand that stress. By altering the electricity delivered during droughts, they shift required capacity by up to 480 GW, which is comparable to the spread across the most critical weather years. Modelling exercises should therefore use multiple weather years; bias-correct and regionally validate weather inputs; explicitly represent resource heterogeneity, forward-looking technologies, and wake losses; and judge these choices by their effect on the low-generation tail, not only on the mean.

0.24

Dual-Agent Reinforcement Learning for Vehicle Scheduling and Multi-Energy Dispatch in Electric-Hydrogen-Transport Systems

eess.SY ·2026-10-29 ·Fulong Yao, Yiming Xu, Maurizio Albano, Liana Cipcigan, Naeima Hamed +2 more
Abstract

Electric-hydrogen-transport systems (EHTSs) require coordinated scheduling of electric and hydrogen vehicles together with multi-energy dispatch. These operations form a sequential decision process: upstream vehicle scheduling (VS) determines service allocation and order, thereby endogenously shaping the electricity and hydrogen demands faced by downstream energy dispatch (ED). This paper develops a dual-agent reinforcement learning framework tailored to the heterogeneous decision structures. The vehicle-scheduling agent combines proximal policy optimisation (PPO) with a Plackett-Luce policy to learn vehicle-priority permutations, which are then mapped to feasible charger and refueller assignments through a greedy allocation. The energy-dispatch agent employs twin delayed deep deterministic policy gradient (TD3) for continuous storage and energy-flow control. The directional coupling between the two decision layers is analysed, revealing that updates to the upstream scheduling policy can shift the demand distribution experienced by the downstream dispatcher. A three-stage training mechanism is therefore introduced, combining layer-wise pretraining with concurrent fine-tuning to mitigate the effects of scheduling-induced distribution shifts on downstream ED learning. Year-long case studies show that the proposed framework achieves the lowest vehicle-service delay and energy cost among the 11 evaluated methods, demonstrating effective coordination between vehicle scheduling and multi-energy dispatch.

0.23

BEVLight: Generalizable Traffic Signal Control from UAV Bird's-Eye Views

eess.SY cs.LG ·2026-10-29 ·Maonan Wang, Yuxin Cai, Junming Liu, Zhengyan Huang, Yi Yu +3 more
Abstract · 27 pages, 15 figures

Learning-based traffic signal control (TSC) has achieved promising performance, but most methods rely on structured traffic states that must be estimated from sensor observations in real deployments. Visual TSC learns signal decisions directly from images, yet existing methods depend on fixed roadside cameras and remain tied to specific intersection configurations and visual domains. UAVs offer on-demand bird's-eye-view (BEV) observations that jointly capture intersection topology and traffic dynamics, making them a promising sensing platform for visual TSC. We present BEVLight, a generalizable TSC framework that directly maps BEV observations to signal actions. BEVLight grounds visual features onto individual lanes and composes them along the junction topology into movement and phase representations, allowing a shared policy to score candidate phases across heterogeneous intersection layouts and signal plans. To train the policy, we develop a web-based virtual teleoperation platform and collect over 5K human control decisions together with their underlying traffic states. Simulator-provided traffic states first supervise lane- and movement-level representations, after which human phase selections and rankings supervise the phase scorer. Re-rendering the recorded trajectories under diverse appearances preserves both forms of supervision and encourages robustness to visual changes. Closed-loop experiments demonstrate zero-shot generalization across traffic demands, signal plans, and intersections: BEVLight reduces average waiting time by 7--9\% relative to perception-to-control visual baselines and matches or outperforms structured-state RL controllers that receive ground-truth traffic states, achieving the lowest waiting time in 12 of 16 scenarios. An offline study further shows that BEVLight transfers to real UAV observations without target-domain adaptation.

0.23

DAWN: Diffusion-based Alignment and unfastening Workflow Network

cs.RO ·2026-10-29 ·Hemant Kumar, Piyush Goenka, Nikhil Chopra
Abstract · 8 pages, including references

We present DAWN (Diffusion-based Alignment and unfastening Workflow Network), a hierarchical robotic framework designed to perform real-world, unfastening tasks through multi-modal sensory integration. Autonomous robotic unfastening is crucial for improving recycling processes, enhancing safety, and reducing operational costs; yet, it remains predominantly manual due to inherent challenges in precise screwdriver engagement with fasteners, detailed force execution, and robustness to environmental variability. To address the complexity of learning long-horizon, contact-rich manipulation tasks, we subdivide the unfastening task procedure into two sequential stages: (1) alignment & insertion, and (2) unfastening. DAWN consists of three key components: AlignNet, a diffusion-based policy ensuring precise screwdriver tip engagement; LatchNet, a multi-modal classifier reliably detecting successful stage transitions; and UnfastNet, which achieves stable, force-controlled screw removal. Experimental validation on a Franka Emika robot demonstrates that DAWN achieves alignment & insertion success rates comparable to trained human operators, robustly transitions between stages, and attains an impressive success rate in unfastening across various unseen scenarios. Ablation studies further elucidate critical input modalities and effective training strategies, highlighting DAWN's scalability, robustness, and data efficiency.

0.23

Modeling and Parametrization of Large-scale Hydrogen Electrolyzers for Transient and Frequency Stability Analysis

eess.SY ·2026-10-29 ·Mahmoud K. Abdelhamid, Lukas Sigrist, Álvaro Ortega
Abstract

Hydrogen Electrolyzers (HEs), particularly Proton Exchange Membrane (PEM)-based systems, are flexible loads capable of providing grid ancillary services and enhancing the stability of renewable-dominated power systems due to their fast electrical response. However, the literature often lacks clear guidance on the structure, modeling, and parametrization of large-scale HEs for transient and frequency stability analyses. This paper presents a detailed configuration for utility-scale HE plants, using the stack as the fundamental unit, supported by a realistic parametrization based on commercial components. Furthermore, this work extends the Generalized ESS Model (GEM), a linearized state-space model previously utilized with other converter-interfaced energy storage systems, to PEM HEs for the first time. To facilitate plant scalability, a robust method for scaling GEM matrices across different capacities is proposed. Finally, a measurement-based methodology is developed to estimate the GEM parameters and subsequently extract the equivalent electrical circuit parameters. Comprehensive simulations on the WSCC system under various disturbances validate the accuracy of the derived GEM, the effectiveness of the scaling approach, and the robustness of the parameter identification methodology.

0.23

RoboSeed: Multidimensional Evaluation of Vision-Language-Action and World-Action Models: Capability, Resource Affordability, and Research Evidence

cs.RO ·2026-10-29 ·Xin Ma, Furong Chen, Xinlong Man, Bo Zhang, Yalong Guo +7 more
code/data
Abstract · Project repository: https://github.com/PrimitiveX/RoboSeed

Choosing a robot action model requires evidence about task capability, adaptation cost, deployment requirements, and the reliability of reported conclusions. Success rates alone offer limited guidance to researchers with different computing budgets and goals. We propose RoboSeed, a multidimensional evaluation framework for vision-language-action (VLA) and world-action models (WAM), within a continuing program to train, evaluate, and openly release robot models with verifiable evidence for model selection. The framework defines five dimensions: capability, downstream training affordability, deployment affordability, locally verified research value, and reproduction evidence. It maps empirical measurements and audit criteria to 0-100 profiles using fixed normalization anchors, with balanced, training-constrained, and deployment-constrained preference settings. Interface and protocol checks establish when comparisons are appropriate. Missing measurements remain explicit, incomplete aggregate scores are withheld, and weighted input coverage identifies evidence gaps. We apply the framework to representative VLA and WAM models, drawing on local training, released-weight evaluation, task-level audits, and an evaluation-system study. The results reveal budget-dependent backbone choices, sensitivity to adaptation recipes, and differences between language and visual task interfaces. Reproducible scoring code and machine-readable evidence support extension to additional models. The study provides dimension-level assessments and priorities for further measurement, while incomplete serving traces and heterogeneous protocols preclude overall numerical rankings.

0.22

STAR: Steering Action Representations with Proprioception for VLA Inference

cs.RO cs.AI ·2026-10-29 ·Guangyu Zhao, Yongchang Zhang, Haoxuan Ru, Kaichen He, Haoran Wen +4 more
code/data
Abstract

Vision-language-action models (VLAs) map language instructions and visual observations to robot actions, showing strong capabilities across robotic manipulation tasks. However, in our probing experiment, we find that action representations near the action output do not fully preserve task-relevant visual grounding. Motivated by this late-layer grounding gap, we introduce STAR, for STeering Action Representations, a training-free inference method for pre-trained VLAs. STAR applies an exponential moving average over late-layer residual updates, allowing later action representations to inherit object-aware update directions from earlier layers without model updates or architectural changes. It further uses recent proprioceptive states to adapt the mixing strength, using conservative mixing when execution dynamics change rapidly. Experiments on robot manipulation benchmarks show that STAR improves late-layer object awareness and increases average task success rates by 4.0 percentage points. Our code is available at https://github.com/PKU-VaLuE-Lab/STAR.

0.22

Sensing Fast and Distant Targets in Fractional Overspread Channels with Zak-OTFS and Modular Arithmetic

eess.SP cs.IT ·2026-10-29 ·Andrea Novero, Mudasir Ahmad Sheikh, Danish Nisar, Saif Khan Mohammed, Michele Zorzi +1 more
Abstract · 14 pages, 8 figures, 4 tables

Delay-Doppler (DD) modulations, such as Zak-transform-based orthogonal time frequency space (Zak-OTFS) modulation, represent the channel as a sparse set of delay and Doppler shifts, which makes them well suited for radar sensing. Sensing with a single DD frame, however, is unambiguous only when the channel spread factor, i.e., the product of the delay and Doppler spreads, is less then one. Long-range mobility scenarios, including drones, airplanes, low Earth orbit (LEO) satellites, and underwater acoustic links, typically violate this crystallization condition, so that a single frame aliases the target delays, Doppler shifts, or both. Existing methods add a time-frequency pilot to a conventional DD frame, resolve the aliasing in only one dimension, and assume integer delays and Doppler shifts. We propose a frame design that combines multiple DD bases with different delay and Doppler periods, which together cover an overspread channel with fractional delays and Doppler shifts, using the same bandwidth and frame size. We show that modular arithmetic on the peak positions uniquely recovers each target by pairing the peaks through their common projection and their gains, and we derive the condition under which this pairing is ambiguous. We implement the design with two superimposed Zak-OTFS frames and propose two receivers: dual-frame sensing (DS), which performs the joint detection directly, and DS with cross-pilot cancellation (DS-XC), which additionally cancels the cross-pilot interference, re-estimates the gains by joint least squares, and removes ghost targets. Both receivers maintain a detection rate close to one as the spread factor grows, and DS-XC also achieves the lowest false-alarm rate and channel estimation error among all the compared systems. In a 100 km drone scenario, DS-XC outperforms all baselines, with a complexity that grows near-linearly with the time-bandwidth product.

0.22

BURGER: Benchmarking and Understanding Robotic Generalization via Embodied Drawing

cs.RO cs.AI ·2026-10-29 ·Xiangyu Xu, Bangshuai Peng, Ziyi Liu, Mengjie Zhao, Chenhao Lu +3 more
benchmarkcode/data
Abstract · Preprint version

How effectively can vision-language-action (VLA) models learn diverse language-conditioned behaviors as training data scales? Investigating this question requires large and varied robot datasets, whose collection on physical robots remains costly. We introduce BURGER, a language-conditioned robotic drawing benchmark with a scalable, prompt-driven pipeline for generating continuous drawing trajectories, both in simulation and the real world. In BURGER, an agent starts from the same blank canvas and controls a robot arm to draw the object or spatial composition specified by language, while each instruction allows an open-ended variety of valid drawings and stroke orders. Using 28,200 simulated demonstrations spanning 282 drawing tasks, we evaluate five representative VLA models across data scales, language grounding, and multimodal behavior. BURGER supports fine-grained analysis of language grounding, spatial composition, and action generation, revealing distinct failure patterns across VLA models. Our experiments suggest that some VLA models can benefit from greater task diversity and demonstration scale, capture multimodal drawing behaviors, and generalize to unseen compositions of familiar objects. Finally, we deploy a trained policy on a Franka Panda arm, demonstrating the ability to draw reactively under human intervention. Project homepage and code can be found at https://siyingzhao.github.io/BURGER/.

0.21

Stability-Performance Trade-off in Maritime Integrated Sensing, Communication, and Computation: A Lyapunov-Guided Dual-Timescale Hierarchical RL Approach

eess.SP ·2026-10-29 ·Ruoguang Li, Yu Wang, Bingxin Tian, Benedetta Picano, Yingyang Chen +4 more
Abstract

Driven by the ever-growing demand for intelligent maritime services, cooperative unmanned aerial vehicle (UAV) and unmanned surface vehicle (USV) operations, together with wireless integrated sensing, communication, and computation (ISCC), have become pivotal in maritime environments. However, the distinctive maritime environment precludes a direct application of terrestrial ISCC models, while sensing, communication, and computation compete for the same spatio-temporal resources, thus yielding inherent performance trade-offs that become particularly severe in resource-constrained ocean scenarios. To address these issues, we investigate an ISCC system in which multiple UAVs detect targets at the sea surface and partially offload the collected data to a USV for cooperative processing. Characterizing the maritime-specific environment impairments, we incorporate sea clutter and wave-induced USV roll motion into the sensing and communication models, respectively. Moreover, we introduce data queueing models at both the UAVs and the USV to capture the dynamic coupling among sensing-data generation, wireless offloading, and computation execution. The optimization objective is to minimize the long-term average system energy consumption by jointly optimizing UAV-target association, UAV trajectories, beamforming, time allocation, offloading ratios, and computation resource allocation. Given that these decisions differ in update frequency and time sensitivity, we propose a Lyapunov-guided dual-timescale hierarchical soft actor-critic (LTH-SAC) framework, which transforms the long-term problem into a sequence of queue-aware online subproblems and coordinates optimization across the two timescales. Simulation results verify the effectiveness of the proposed framework and demonstrate its capability to balance the stability-performance trade-off among sensing, communication, and computation.

0.21

Contact Imitation Interface: Transferring Dexterous Skills by Showing How Contact Unfolds

cs.RO ·2026-10-29 ·Zhixuan Xu, Yichen Li, Bingchen Wu, Rui Wang, Yiwen Hou +3 more
Abstract · 15 pages (8-page main paper and 7-page supplementary material). Project website with code and hardware designs: https://nus-lins-lab.github.…

Robot-free human demonstrations offer a promising route to scaling dexterous manipulation data. The key challenge is to preserve the demonstrator's natural dexterity while producing contact-grounded observations and actions that transfer to robot hands. We present Contact Imitation Interface (CIMI), a framework with contact-grounded demonstration and learning interfaces. The demonstration interface combines shared contact geometry and fingertip RGB views with contact-alignment-first retargeting and reprojection of wider scene views. The learning interface introduces a hierarchical policy in which a contact-surface planner predicts relative contact-surface trajectories and a controller coordinates wrist and finger commands to improve contact alignment. CIMI combines contact-grounded observations, retargeting, and action representations, improving mean final-stage success by at least 50 percentage points over baseline policies across six hand-task pairs on LEAP and Wuji. Hardware designs and software are publicly available.

0.21

Tactus: Tactile & Pose Sensing Glove for Natural Dexterity Data Collection

cs.RO cs.AIcs.ARcs.LG ·2026-10-29 ·Li Guan, Rachael Granberry, Sahar Aseeri, Sahar Rostami, Taylor Niehues +99 more
Abstract · 9 pages

Collecting naturalistic human manipulation demonstrations with accurate contact force and hand pose is essential for robot dexterous manipulation policy and foundational world model learning - yet no existing wearable platform provides both calibrated fingertip force sensing and occlusion-resilient hand tracking in a form factor that preserves natural manipulation behavior. We present Tactus, a textile-integrated sensing glove that co-optimizes three design pillars: (1) a thin, conformal 3D-knit form factor (1.5 mm fingertip stackup) designed to minimize interference with natural manipulation, (2) strategically placed capacitive tactile sensors at manipulation-relevant regions informed by empirical contact studies, and (3) IMU-fused pose tracking that maintains accuracy through the hand-object occlusions where manipulation data is most valuable. Tactus includes a documented calibration pipeline for cross-glove consistency and per-session calibration for subject-dependent correction. To enable objective comparison, we introduce standardized benchmarking protocols for tactile force regression and hand pose estimation, and apply them to Tactus and two commercially available tactile gloves - to our knowledge the first quantitative cross-platform evaluation of this kind. Tactus resolves a minimum detectable force of 0.59 N at 95% confidence, 1.6 - 2.7x lower than the commercial baselines, and with 0.87 cross-participant consistency, while our pose tracking sustains 11.0 mm fingertip keypoint relative error during manipulation, when occlusion is heaviest. We show that these combined signals support object property estimation, slip detection, and representative recording of hand-object interaction - yielding demonstration data with the fidelity required for contact-aware manipulation policies and world models.

0.21

Humanoid Dodgeball: Learning Fast Dynamic Obstacle Avoidance Motions for Humanoids

cs.RO ·2026-10-29 ·Aoqian Zhang, Chunzheng Wang, Haoran Chen, Annan Tang, Cheng Xiang +2 more
Abstract

Humanoids have become increasingly agile and capable in real-world environments. To operate safely in dynamic scenarios, they must rapidly evade fast-moving obstacles while maintaining stability and physically feasible whole-body motion. Learning such behaviors remains challenging due to the high-dimensional dynamics of humanoid systems and the need to jointly balance reactive obstacle avoidance, motion naturalness, and control robustness. In this work, we present Humanoid Dodgeball, a reinforcement learning (RL) framework for agile dynamic obstacle avoidance in humanoids. Our method enables highly dynamic whole-body evasive maneuvers against incoming projectiles through threat-aware reward shaping, while multiple direction-specific adversarial motion priors regularize the learned behaviors toward stable and diverse human-like motions. Extensive simulation and hardware experiments demonstrate that the proposed framework substantially improves dynamic obstacle avoidance performance across diverse obstacle speeds and launch distances, while preserving robust and physically plausible humanoid motions.

0.21

LOGIV: Graph-Verified Execution with Causal Local Repair for Long-Horizon Robot Tasks

cs.RO ·2026-10-29 ·Enqiao Lu, Yuxiang Xie, Xingrui Yu, Yuyang Wang, Yipeng Gu +8 more
Abstract · 14 pages. Updated related work, references, author list and affiliations

Agentic robots must turn a high-level goal into actions while keeping the physical conditions needed for future steps valid. A completed action does not ensure that its effects will persist: a later disturbance can invalidate support that the remaining plan still assumes. We present LOGIV, a graph engineering framework for agentic robot execution that uses an explicit task graph to connect symbolic planning, observation-based verification, and embodied action. LOGIV compiles a validated plan into condition-support and resource-ordering edges, checks their physical validity during execution, and uses dependency tracing to determine what must be restored after a failure. Repair is constrained by retained unfinished actions, and the combined continuation is symbolically validated before execution resumes. The framework operates above existing Vision-Language-Action (VLA) and World Action Model (WAM) policies without adding trainable components. Across 20 dual-arm tasks from RoboTwin V2 and BiCoord with three execution backbones, LOGIV improves mean task success without injected disturbances by 10.67 and 15.33 percentage points, respectively. Under predefined perturbations, it achieves higher task completion and condition restoration than three adapted recovery baselines for every backbone on both benchmarks. On three real-robot tasks, successful trials increase from 12/30 to 17/30. These results support explicit, continuously checked task dependencies as a practical interface for reliable agentic robot execution.

0.21

A Two-Wheeled Quadrotor with All-Clockwise Tilted Rotors for Level-Attitude Surface Locomotnion

cs.RO eess.SY ·2026-10-29 ·Takumi Ito, Satoshi Nakano
Abstract

Conventional wheel-equipped quadrotor UAVs must tilt their bodies to generate lateral forces, limiting their ability to maintain a desired attitude during lateral motion. This paper proposes a wheeled quadrotor with all-clockwise tilted rotors that uses surface contact as a constraint to generate propulsion along the contact surface without tilting the vehicle. We analyze feasible forces and system controllability under surface contact and optimize the vehicle design considering different rotor-spin configurations. The optimization results show that a configuration in which all four rotors spin clockwise can generate a larger tangential force than the conventional alternating-spin configuration. This approach enables both aerial and contact-based locomotion without additional rotor-tilting actuators or mechanisms. Simulation results demonstrate both hovering in free space and horizontal locomotion on a floor while maintaining a level attitude.

0.21

Cross-Controller Transfer via In-Context Behavior Adaptation

cs.RO ·2026-10-29 ·Jonathan Yang, Chelsea Finn, Dorsa Sadigh
Abstract · 9 pages, 6 figures

Generalist robot policies increasingly learn from data collected across diverse tasks and embodiments, yet typically assume a single controller for each robot. This assumption limits transfer because different controllers can assign different physical meanings to the same action and may require different strategies to complete a task. We study how to transfer robot data across both embodiments and controllers without allowing conflicting action interpretations to interfere with learning. Our setting spans six controller types as well as variations in controller gains and compliance. We propose a framework that encodes behavioral interaction under a controller through examples of commanded actions and the resulting robot motions. A transformer context encoder summarizes these interactions, allowing a single policy to adapt in context. Across real-world experiments, our method improves average success by 33 percentage points over naive data pooling, achieves 50\% success on unseen controller modes and 52\% on unseen controller gains, and transfers to tasks with limited controller-specific data. On contact-rich tasks such as USB insertion and surface wiping, behavioral interaction context enables the policy to adapt not only its actions but also its strategy. Our results provide a path toward generalist robot policies that can leverage data collected across diverse action interpretations. For more information, visit \href{https://controller-transfer.github.io}{https://controller-transfer.github.io}.

0.20

On the Performance of AoA-Based Physical-Layer Authentication Under Multi-Source Impersonation Attacks

eess.SP eess.SY ·2026-10-29 ·Jeongbin Kim, Moonsik Min, Yo-Seb Jeon
Abstract · 13 pages, 4 figures

This paper investigates cooperative impersonation attacks against angle-of-arrival (AoA)-based authentication for a wireless communication receiver equipped with a digital uniform linear array. Specifically, we characterize the vulnerability and robustness of this authentication system in relation to adversarial angular diversity and the relative numbers of adversarial sources and receive antennas. To this end, we first derive the optimal precoding vector for multiple spatially distributed adversarial sources by minimizing the mean squared error between legitimate and adversarial received signals. By means of the rank property of Vandermonde matrices, we then establish necessary and sufficient conditions under which the adversarial sources can reproduce the legitimate transmitter's spatial signature. With fewer adversarial sources than receive antennas, perfect reproduction is not possible without at least one adversarial source satisfying a specific AoA condition. In the other case, increasing the number of receive antennas can reduce impersonation success probabilities when none of the adversarial sources has an AoA satisfying this condition. Numerical results validate the theoretical analysis and evaluate the security performance of AoA-based authentication.

0.20

Joint Estimation and Network Configuration in Multi-Agent Systems

eess.SP ·2026-10-29 ·Mohammadreza Barzegaran, Hamid Jafarkhani
theory
Abstract

Distributed consensus-based sensing enables a multi-agent system to collaboratively estimate an unknown state by combining local measurements with information received from neighboring agents. In practice, however, the communication network, transmission delays, and finite-rate communication directly affect the disagreement among agents. On the other hand, agreement among agents does not ensure accuracy, particularly when sensors have different measurement uncertainties. We develop a sampled consensus-plus-innovations framework under realistic wireless network constraints that jointly accounts for inter-agent disagreement and estimation inaccuracy. We derive a computable upper bound on the worst-case steady-state estimation error that captures consensus disagreement due to the communication graph, quantization error, and transmission delay, together with estimation inaccuracy, which accounts for measurement uncertainty against the true state. We then use this bound as a network design objective and propose a semi-distributed auction-based algorithm that jointly modifies the communication graph and link-rate allocation. The proposed method preserves connectivity, converges in finite time to a locally auction-stable network configuration, and reduces the bound on the estimation error at every accepted change. The proposed method has been evaluated on various test cases, and the results show that it reduces the estimation error while maintaining fewer links and allocating moderate link rates in a short time.

0.20

Decoding Geometry, Not Tokens: Efficient VLMs for Robot Affordance and Trajectory Prediction

cs.RO cs.CV ·2026-10-29 ·Akila Ayanthi, Darshana Priyasad, Tharindu Fernando, Dimity Miller, Sridha Sridharan +2 more
published
Abstract · IEEE/CVF Winter Conference on Applications of Computer Vision (WACV 2027)

Predicting where to grasp and how to move is fundamentally a geometry problem. Vision-language models treat it as language generation instead. They tokenize spatial coordinates into text and decode them autoregressively. This mechanism ignores spatial structure and truncates continuous precision. Errors compound across coordinate tokens during decoding. In this paper, we show that compact off-the-shelf VLMs with 0.5-2B parameters already encode rich spatial information in their intermediate hidden states. Autoregressive decoding is the dominant bottleneck for trajectory prediction, while affordance prediction depends additionally on feature depth and decoder architecture. We expose the latent spatial information with a lightweight, geometry-compatible prediction head that attaches to a frozen backbone. It reads features from an intermediate hidden layer, conditions a learned query on the task instruction and refines it via cross-attention over the visual tokens to regress affordance boxes and trajectory waypoints directly as continuous values, without modifying the backbone. We evaluate on ShareRobot-Bench and AGD20K. Our 2B variant reduces trajectory error by over 60% and improves ShareRobot-Bench affordance mAP by 85% relative to the autoregressive baseline. It approaches RoboBrain on trajectory prediction (0.251 vs 0.191 DFD) at 3.8x higher throughput. These results demonstrate a favorable accuracy-efficiency trade-off for instruction-conditioned 2D geometric prediction.

0.20

Tracking Task Progress Through Errors and Recovery:A Benchmark and Recurrent Distributional Model

cs.RO ·2026-10-29 ·Chen Wang, Xinze Niu, Zihao Zhu, Keru Zhou, Yicheng Xu +6 more
benchmarkdatasetcode/data
Abstract

Operational strategies for long-horizon robotic tasks require reliable evaluation to determine whether execution is progressing, stalling, regressing, or recovering from errors. Yet real- world robot datasets that capture error recovery and provide frame-level error annotations remain scarce. To address this gap, we introduce and open-source ReSTEP, a dataset of 3,156 episodes across two manipulation tasks with semantic stage labels and frame-level er- ror annotations. ReSTEP supports training and benchmarking progress monitors through successful execution, mistakes, and recovery. Existing progress estimators can struggle to distinguish visually similar task stages and maintain temporally consistent predictions while responding to genuine setbacks. We therefore propose REMORA, an online progress esti- mator that combines recurrent visual memory, distributional prediction, and causal peak selection. Memory retains execution history to help resolve visual ambiguity, while distribu- tional prediction represents alternative progress hypotheses and causal peak selection tracks a supported hypothesis over time. Together, these components enable consistent progress tracking that remains responsive to errors and recovery. Experiments on both ReSTEP tasks demonstrate that REMORA outperforms the state-of-the-art SARM baseline in progress ac- curacy and temporal consistency.

0.19

ScheduleGen: Demonstration Generation via Skill Scheduling for Multi-Arm Imitation Learning

cs.RO ·2026-10-29 ·Neel Jawale, Stephan Pleines, Soha Pouya, Yuke Zhu, Ajay Mandlekar +1 more
Abstract · 22 pages, 9 figures, 5 tables. Project website and videos: https://njawale42.github.io/ScheduleGen/

Multi-arm robot platforms, including bimanual and humanoid robots, have the capability to complete tasks with lower cycle times than single-arm systems, but they are more difficult to safely control due to the presence of multi-arm constraints. The paradigm of training generalist robot policies from massive datasets is a promising approach to address general-purpose multi-arm manipulation, but the acquisition of large multi-arm manipulation datasets remains a bottleneck. Automated demonstration generation can alleviate this burden by synthesizing new demonstrations from a handful of source demonstrations, but existing algorithms ignore multi-arm constraints and as a result often induce unsafe and unsuccessful behaviors. We address this limitation by introducing ScheduleGen, an integrated planning & scheduling algorithm for multi-arm demonstration generation. We evaluate ScheduleGen on 28 simulation task variants, including seven externally authored RoboDojo tasks and settings inaccessible to prior systems; it achieves 79.6% mean data-generation success and trains policies with 65.5% mean success. ScheduleGen policies outperform those trained on a state-of-the-art data generation baseline, and achieve 50-80% success across eight real-world variants through zero-shot sim-to-real and real-to-sim-to-real transfer. Videos and more at: https://njawale42.github.io/ScheduleGen/

0.19

Adaptive Scenario Generation for KL-DRO Around a Learned Density Model

eess.SY cs.LG ·2026-10-29 ·Aron Brenner, Saurabh Amin
Abstract

Distributionally robust optimization (DRO) hedges data-driven two-stage stochastic programs against error in the distribution estimated from a finite record, and standard ambiguity sets differ in how they treat realizations absent from that record. A divergence ball around the empirical distribution contains only reweightings of the record, whereas a transport ball also contains distributions supported off the record, including on realizations the data-generating process cannot produce, against which the decision is then hedged as well. We instead propose a Kullback-Leibler (KL) ball centered at a density model fitted to the record, the learned nominal, which lets the adversary up-weight costly unobserved realizations in proportion to the probability the model assigns them. Solving the resulting problem poses a second difficulty, because it can accommodate only a few scenarios, and an independent sample of that size is likely to omit the rare realizations on which the worst case concentrates. We therefore build the scenario set adaptively, alternating between solving the finite problem and drawing new realizations from the learned nominal, steered toward the worst case at the current decision, with importance weights that correct the draws back to the model. On a multi-item newsvendor problem and a capacity expansion problem on realistic grid data, centering the ball at the learned nominal rather than at the record lowers cost under distribution shift, at comparable in-distribution cost, when the record is scarce and the model class captures structure that lets it extrapolate beyond the record. The advantage fades as the record grows or as the shift departs from the structure the model has learned.

0.19

Decision-Level Error Attribution and Depth-Aware Safety Evaluation of Monocular RGB Human Tracking for Autonomous Robot Navigation

cs.RO cs.CV ·2026-10-29 ·Abdulahi Taiwo, Chijioke Cyriacus Ekechi
theory
Abstract · 10 pages, 7 figures, 6 tables

Monocular RGB tracking is useful for robots operating near people, but perception errors affect binary stop/proceed decisions in ways that geometric accuracy metrics do not capture. We examine how much decision error arises from the safety rule versus perception. We prove that lateral-only distance is a lower bound on planar distance: with accurate positions, it cannot miss dangers identified by the planar rule, but it produces unnecessary stops. We introduce a two-factor Shapley attribution separating rule and perception contributions, and further distinguish missed detections from localization errors. Experiments use ETH and UCY pedestrian videos with simulated stationary robots, disjoint calibration and evaluation pedestrians, and time-block bootstrap intervals. Across 898 pedestrians in three scenes, the lateral rule missed no dangers on annotated positions but achieved only 5.6-31.9% stop precision. In two scenes with RGB perception, calibrated ground-plane localization achieved 0.59 m and 0.51 m planar RMSE, compared with 1.72 m and 1.39 m for a geometric baseline and 0.81 m and 0.68 m for calibrated Depth Anything V2. Detection recall of 0.43 and 0.26 made missed detections the leading source of missed danger. A two-standard-deviation safety margin with ground-plane localization matched or improved on the lateral rule's missed-danger rate in all eight scene-threshold settings and reduced unnecessary stops in seven. All robot interactions are simulated, and outcomes are offline decision classifications, not demonstrations of collision avoidance.

0.18

AI for Science in Space: Orbital Laboratories, Planetary Fieldwork, and Embodied Autonomy

cs.RO ·2026-10-29 ·Zizhe Wang, Jia Deng, Xin Lu, Zehao Jia, Yi Huang +2 more
Abstract · 47 pages, 5 figures, 7 tables. Ancillary evidence files included

Artificial intelligence for science (AI4S) is extending from interpretation of existing data to the design and execution of experiments. Space offers distinctive scientific opportunities while restricting physical support. Microgravity and extraterrestrial environments expose phenomena that terrestrial experiments cannot fully reproduce; finite samples, constrained resources, and limited opportunities for intervention complicate sustained investigation. This critical review examines orbital laboratories and planetary in situ science across biology, materials, fluid and thermal physics, and adaptive exploration. We distinguish learning from space-derived data, selecting scientific actions, executing them, and preserving experimental validity. Terrestrial self-driving laboratories provide complementary evidence on optimization, dedicated automation, robotic manipulation, and the support needed for unattended operation. The synthesis connects scientific gains already demonstrated by bounded AI systems to gaps in observation, executable actions, and continuity of valid evidence. Space robots can expand the attainable scope of autonomous science when necessary interventions exceed validated instrument capabilities and timely human support is unavailable. Their value must be established against dedicated mechanisms, redundancy, and human-supported operation. We develop a framework connecting sample history, scientific decisions, physical execution, and claim-specific validity. A worked orbital design, a planetary extension, and matched evaluation schemes separate the contributions of sensing, decision policies, and physical actions. The resulting agenda measures autonomy through valid scientific return, task coverage, intervention costs, and evidence integrity across complete campaigns.

0.18

Safe Platooning under Input Constraints with Link-Wise Velocity String Stability

eess.SY ·2026-10-29 ·Panagiotis S. Trakas, Charalampos P. Bechlioulis, Thomas Berger
theory
Abstract

We propose a decentralized controller for vehicle platoons with different braking and propulsion limits. The method uses only local distance and velocity measurements and requires neither online optimization nor exact vehicle parameters. The design combines a braking-distance safety margin with an adaptive funnel that relaxes the spacing requirement during propulsion saturation. A standstill mechanism prevents backward motion and supports stop-and-go operation. Under explicit uniform conditions on the vehicle parameters, controller settings, and initial states, we prove that closed-loop solutions exist for all time, collisions are avoided, and input limits are satisfied. Moreover, each follower tracks its predecessor's velocity range with gain one, up to an explicit additive offset determined by the funnel width and the braking-distance term. This bound establishes link-wise velocity string stability. Both the offset and the bounds on speeds and spacing errors are independent of vehicle index and platoon length. Simulations with 2 to 100 followers illustrate safety and bounded velocity deviations under disturbances and actuator limits.

0.18

Approximate Dynamic Programming for Link Reconfiguration to Reduce Malicious Input Controllability in Consensus Networks

eess.SY ·2026-10-29 ·Nathaniel Mailhot, Mohammad Pirani, Davide Spinello
Abstract

Communication links influence how strongly a malicious input at one sensor affects a consensus network. We study how a defender can reduce this influence by selecting one additional link after observing the currently attacked sensor. The attacked sensor changes over time, and vulnerability is measured by the trace of a finite-horizon controllability Gramian. A prescribed local epsilon-greedy attacker response rule defines a discounted Markov decision process for the defender. Exact dynamic programming (DP) provides an optimal policy using the full Gramian payoff matrix and attacker transition model. We then introduce approximate dynamic programming (ADP) that learns from sampled attacker transitions and empirical Gramian costs estimated from measured impulse responses. Under conditionally unbiased costs and standard sampling and learning-rate conditions, the learned action-value table converges to the exact DP solution. The analysis uses Bellman contraction and bounds the effect of persistent Gramian estimation bias, establishing when measured responses recover optimal defender decisions without constructing the full payoff matrix. In simulations on 20- and 100-node graphs, empirical ADP rollout costs are within 0.23% and 1.76%, respectively, of exact DP. In a two-dimensional intruder tracking example, both policies reduce network-wide root-mean-square tracking error spread by approximately 25% relative to an unmitigated attack.

0.18

From Teleoperation to Long-Horizon Dexterity: Evaluating Visuomotor Skills and Offline Proxies for Dexterous Manipulation

cs.RO cs.AIcs.LGeess.SY ·2026-10-29 ·Ammar N. Abbas, Aayush Jain, Prashanth Viswanath, Nivesh Gadipudi, Giulia Vilone +1 more
Abstract · Accepted at CoRL 2026 Workshop

Imitation learning policies are intended to make robotic manipulation robust to variations in object pose, appearance, and environment that challenge pre-programmed motion. However, real-world performance remains difficult to predict, as conventional optimization signals such as validation loss show no consistent relationship with robot success. To identify the design choices that govern real-world performance, 281 imitation-learning policies spanning seven policy families were evaluated in approximately 1,700 real-robot trials of RJ45 connector grasping and insertion using a five-fingered robotic hand. Controlled ablations showed that observation and action representations dominated conventional optimizer and sampling choices. For point-cloud policies, region-of-interest cropping, representation in the robot-base frame, and normalization within the point encoder each improved success by 70-100 percentage points. For the vision-language-action policy, absolute rather than relative action prediction improved insertion success by up to 70 percentage points. Skill decomposition, multi-view perception, and recovery demonstrations were associated with the strongest-performing systems, although their effects were not independently isolated. Of 67 offline metrics evaluated, only five measures of predicted-motion smoothness and directness reliably correlated with insertion success (r = -0.70 to -0.80), and only within a single policy family. These results identify representation choices as key determinants of real-world imitation-learning performance and show that offline metrics can shortlist policies, but cannot replace real-robot evaluation.

0.18

WorldBridge: Bridging Video and Action for Task Generalization in World Action Models

cs.RO ·2026-10-29 ·Jun Cen, Yubo Guo, Hao Li, Liangkai Li, Yaoxu Lyu +7 more
code/data
Abstract · Authors are listed in alphabetical order. Project page: https://worldbridge-jd.github.io/

Achieving the broad task generalization of embodied models in physical environments remains a central challenge in robot learning. We study this challenge in World Action Models (WAMs) through two complementary capabilities: video generalization, which predicts the visual evolution of new tasks, and action generalization, which translates these predictions into executable behavior through a video-to-action bridge. By varying video and action supervision for tasks outside an initial training set, we examine their respective contributions to task transfer. On RoboTwin 2.0, using only in-distribution video-action supervision yields a 25.07% success rate on out-of-distribution tasks. Introducing videos of these tasks without their action labels raises success to 55.90%, while adding just one action-labeled demonstration per task further improves it to 70.70%. These findings suggest that broader video coverage enables task transfer by leveraging existing inverse-dynamics capabilities, while even sparse action supervision further strengthens those capabilities. Guided by this perspective, we introduce WorldBridge, a WAM with separate video and action experts connected through structured attention and trained with joint flow matching. WorldBridge incorporates visual memory at two temporal scales and action forcing, alongside a unified data processing pipeline that standardizes and filters heterogeneous robot trajectories and egocentric hand-derived supervision. Across multiple manipulation benchmarks, WorldBridge achieves state-of-the-art performance among the compared methods, including 96.23% on RoboTwin 2.0 full data setting and 80.9% on clean to random setting, 82.8% on RoboCasa GR1 TableTop, and 57.6% on DOMINO Level-1. Our study highlights the complementary roles of video prediction and action grounding in developing robots that generalize across tasks.

0.17

Learning Safe Flow Matching Robot Policies from Binary Feasibility Oracles

cs.RO cs.AI ·2026-10-29 ·Jianming Ma, Rongjun Jin, Yizhi Chen, Yue Gao
Abstract

Robot foundation models increasingly rely on flow-based action generation for general-purpose manipulation, yet the generated actions are not guaranteed to satisfy physical constraints such as joint limits and collision avoidance. This problem is especially challenging when the feasible set cannot be described analytically and is accessible only through black-box feasibility oracles. We study this practical oracle-only setting and propose an oracle-guided differentiable safeguard for safe flow-matching robot policies. The safeguard uses binary feasibility queries to correct unsafe updates in the forward pass, while extracting local boundary information to provide constraint-aware learning signals in the backward pass. Combined with discrete-time flow matching, the same safeguard is used during training and inference, enabling forward--backward aligned policy learning without explicit or differentiable constraint models. Experiments on obstacle-aware robotic manipulation demonstrate 100% constraint satisfaction while preserving task performance.

0.17

From Reasoning to Action: Consequence-Grounded Alignment for Vision-Language-Action Models

cs.RO ·2026-10-29 ·Yixue Zhang, Wenke Xia, Kun Wu, Wenbo Yu, Pei Ren +4 more
code/data
Abstract · Project page: https://onlyxingxingzyx.github.io/CoRAct/

Reasoning informs decision-making by connecting intended actions with their expected consequences. Recent vision-language-action (VLA) models incorporate embodied chains of thought (ECoTs) to guide action generation. However, such guidance remains implicit, as actions are conditioned on ECoTs rather than explicitly optimized to align with them. We introduce CoRAct, a consequence-grounded reasoning-to-action alignment framework that explicitly aligns reasoning and action through reward-guided optimization in action noise space. CoRAct formulates two physically grounded ECoT components as verifiable signals: movement specifies intended motion, while waypoints encode expected spatial consequences. A unified Mixture-of-Transformers jointly models ECoT reasoning, action generation, and action-conditioned future prediction. For candidate actions generated from sampled noise, CoRAct computes alignment rewards based on movement consistency and agreement between predicted outcomes and specified waypoints. These rewards guide group-relative optimization of the initial noise distribution, steering action generation toward consistency with intended actions and expected consequences. Experiments across RoboCasa, VLABench, and real-world manipulation tasks demonstrate improvements in task success, generalization, and reasoning-execution alignment over strong ECoT-based VLA baselines. CoRAct also grounds externally provided ECoTs in executable behavior, enabling zero-shot adaptation to novel scenarios by leveraging knowledge from powerful vision-language models. Project page and dataset: https://onlyxingxingzyx.github.io/CoRAct/

0.17

Language-Grounded Force/Torque Pretraining for Extrinsic Contact Perception

cs.RO ·2026-10-29 ·Kazuki Takahashi, Hiroaki Murakami, Ryota Sakuma, Mitsuhiro Kamezaki, Matthew Ishige +1 more
Abstract · 9 pages, 5 figures

Fine manipulation requires robots to recognize when a grasped object makes or changes contact with its environment, yet such extrinsic contact is not always directly observable from vision or local contact geometry. Force and torque signals carry this information, but also vary with object weight, grasp configuration, and robot motion, leaving it unclear whether they can provide a reusable representation of physical state. Here we show that bilateral six-axis force/torque histories can be pretrained as a semantically structured physical representation by aligning them with a frozen language space using supervision generated from simulated pick-and-place interactions. On unseen objects in simulation, the learned representation supports text-prototype readout of robot action, object weight, and grasp interaction. In real-world placement, unlabeled calibration enables contact and support-state recognition without updating the encoder, outperforming a vertical-force sign-reversal rule baseline. For connector insertion, an event absent from pretraining, low-rank adaptation improves data-efficient learning over training from scratch. These results show that force/torque histories can support both direct readout of known physical states and efficient adaptation to new extrinsic-contact events.

0.16

Adaptive search with path-constrained mobile sensors using multi-arm bandits

cs.RO cs.LGmath.OC ·2026-10-29 ·Abraham P. Vinod, Parth K. Thaker, Gautam Dasarathy, Stefano Di Cairano
Abstract · 24 pages, 5 figures

We study a constrained sequential decision problem for detecting regions of interest from noisy online measurements collected by a team of energy-constrained mobile sensors. The sensor motion is restricted to a precomputed set of feasible paths that satisfy mobility constraints, limited available energy by requiring return to charging stations, and avoidance of known obstacles. We cast the resulting active path selection as a multi-armed bandit problem and propose an iterative search algorithm that adaptively assigns paths using confidence bounds computed from the collected data. We also propose a graph-based method to construct low-cardinality path covers under the team's constraints. We establish finite-time and anytime probabilistic correctness guarantees for the proposed algorithm, derive non-asymptotic upper bounds on the times required to detect all regions of interest and to label all nodes, and show that these bounds match algorithm-independent lower bounds up to constant factors for a broad class of problem instances. Using numerical simulations, we illustrate the behavior of the proposed approach and the effect of path redundancy and team size on search performance.

0.15

Teaching a Tree to Recover: Online Distillation of an LLM Supervisor with Counterfactual Replay

cs.RO ·2026-10-29 ·Kristian Ceder, Kilian Freitag, Knut Åkesson
Abstract

Autonomous mobile robots commonly pair a global planner with a local model predictive controller (MPC), but nonconvex obstacle avoidance and a limited horizon can make the robot freeze in complex layouts. Deciding whether to replan, wait, or call for help is a supervisory decision. Recent work queries a large language model (LLM) supervisor on every anomaly, which is slow and costly. We instead distill the LLM online into a decision tree. After each training episode, counterfactual replay labels recoveries as needed, unnecessary, or harmful. These labels retire unhelpful rules. In simulation, on 20 unseen layouts where the MPC alone freezes, the tree handles at least 18 correctly across all seeds, either by reaching the goal or by calling for help when no feasible path exists. The tree matches both an always-query LLM supervisor and expert-designed rules, while issuing no LLM queries at test time. Handwritten rules cannot anticipate the long tail of situations a physical robot will encounter. Online distillation is a promising remedy: the tree handles familiar cases, and the LLM is consulted only for new ones, whose validated recoveries become new rules. Language models can thus also serve as designers of supervisors rather than only as runtime supervisors.

0.15

Improving Fine-Grained Manipulation through Demonstration Expansion and Execution Design

cs.RO ·2026-10-29 ·Jingxu Gu, Dongze Lian
Abstract · RoboFineMani Workshop Challenge at CoRL 2026

Fine-grained manipulation requires a robot policy to follow instructions, remain robust to visual variations, and translate predicted actions into precise robot control. We present Demonstration Expansion and Execution Design (DEED), a framework for adapting a pretrained vision-language-action policy to the five MetaFine tasks through structured demonstration expansion and execution design. DEED generates demonstrations for different instructions from the same initial scene and renders each physical trajectory under different camera positions and lighting conditions while preserving its original states and action targets. A hierarchical sampling strategy balances instructions, trajectories, and visual conditions during supervised adaptation. At deployment, DEED converts predicted action chunks into robot commands using inverse kinematics, gripper mappings selected for each task, and closed-loop replanning. DEED achieves an official Overall score of 85.55%. Fixed-policy evaluations show that performance varies with execution horizon, denoising budget, and gripper mapping.

0.14

Gaze-Assisted Supervisory Control for Lunar Robotic Arm Teleoperation Under Concurrent Monitoring

cs.RO ·2026-10-29 ·Jiayu He, Yifan Li, Oliver R. Runswick, Ernest N. Kamavuako, Yunjie Jia +2 more
Abstract

Rover-mounted robotic arms can extend human reach during lunar sample handling, but continuous joystick teleoperation requires repeated visual-motor alignment while operators may also need to monitor system states. This paper presents a gaze-assisted supervisory control framework in which gaze nominates a rock or placement target, explicit confirmation authorizes the action, and the robot performs bounded autonomous grasp-and-place execution. The framework was evaluated against manual joystick teleoperation in a mission-analog virtual reality study with 16 participants. The study used a 2 x 2 x 3 within-subject design varying control mode, placement difficulty, and concurrent monitoring load. Mean recorded task completion time decreased from 47.65 to 18.43 s per rock, with 72.5% of this difference arising during grasping. Under the tested auditory monitoring conditions, gaze-assisted supervisory control showed less average proportional slowing, with no detected control-mode difference in monitoring-response accuracy or speed. The rate of gaze shifts between core areas of interest decreased from 49.98 to 27.73 shifts/min, while the raw NASA Task Load Index score decreased from 32.54 to 18.26. However, the benefit was task dependent. In precision placement, manual control yielded lower mean placement error and a higher placement success rate. These findings support a mixed-initiative allocation in which gaze enables rapid task-level intention specification and autonomous execution, while precision-critical stages retain manual refinement.

0.14

Kinematic-Tactile Aligned Dexterous Retargeting with Contact Wrench Space Alignment

cs.RO ·2026-10-29 ·Xinyuan He, Ziheng Xu, Yueyuan Chen, Yufan Ye, Yuanshuo Tan +4 more
Abstract · 8 pages

Human demonstrations provide a natural and scalable source of data for learning dexterous manipulation policies. However, existing retargeting methods primarily transfer human kinematics, while the tactile information underlying physical interaction is rarely preserved across embodiments. To address this limitation, we present a hierarchical three-stage framework for kinematic-tactile aligned dexterous retargeting. We use object-centric contact wrench space as a physical representation for tactile alignment between human and robot hands. The framework first aligns hand kinematics and contact geometry, and then constructs a human wrench support profile from tactile forces and contact geometry. Robot fingertip force targets are subsequently optimized to cover the human wrench support and converted into joint position corrections, yielding directly replayable robot hand trajectories. We evaluate the framework in simulation and real robot experiments across three pouring conditions. In the real robot experiments, our method achieves a 96.7\% replay success rate and higher wrench transfer fidelity than representative retargeting baselines. The framework enables physically grounded transfer of kinematic and tactile information from human demonstrations to robot hands, providing a path toward scalable learning of contact-rich dexterous manipulation.

0.14

Don't Risk It!: Trajectory-Conditioned Distributional Risk for LiDAR-Only Crowd Navigation

cs.RO ·2026-10-29 ·Mathai Mathew Pulicken, Laksh Nanwani, Basant Sharma, Phani Teja Singamaneni, Arun Kumar Singh +1 more
Abstract

Efficient crowd navigation requires anticipating dynamic collision-free space without relying on fragile object tracking. We propose the Don't Risk It Planner (DRIP), an end-to-end LiDAR-only planning framework that couples occupancy forecasting with trajectory-conditioned clearance and uncertainty estimation. Operating directly on LiDAR-derived occupancy grid histories, DRIP predicts a distribution over minimum robot-obstacle clearance for candidate control sequences to evaluate collision risk. The model is trained using a risk cost that grounds predictive uncertainty directly in empirical collision outcomes rather than geometric regression alone. In closed-loop experiments with up to 45 pedestrians, DRIP achieves higher success rates with fewer collisions and reduced freezing behavior compared to state-of-the-art baselines.

0.13

ProLoop: Progress-Aware Closed-Loop Framework via Structured State Relations for Long-Horizon Robotic Task Planning

cs.RO ·2026-10-29 ·Runxin Zou, Qingmiao Liang, Jiatao Zhang, Xinyan Lv, Liangyu Qie +3 more
Abstract

Long-horizon robotic tasks are commonly decomposed into semantic subgoals to simplify planning. However, subgoal progress in LLM-based planning is typically inferred implicitly from interaction histories. Inaccurate estimates can misguide execution and allow local errors to propagate. To address this, we propose ProLoop, a Progress-Aware Closed-Loop framework that derives explicit subgoal progress from structured relations between desired and execution states. A learned State-Relation Monitor assesses these relations, which are then aggregated into subgoal progress as ongoing, deviated, or completed. The resulting progress guides execution adaptation, enabling reliable progress-aware control throughout execution. Experiments on AgentBoard ALFWorld and a real-world robotic platform show that ProLoop improves success rate by 12.94% overall and by 15.46% on Hard tasks, while reducing environment interactions by 32.3%.

0.13

Gaussian Distillation of Flow Policies

cs.RO ·2026-10-29 ·Youssef Saied, Ziyad Sheebaelhamd, Claire Vernade, François Fleuret
Abstract

Flow policies are among the strongest robot controllers and are well suited to modelling multimodal action distributions. Our analysis shows that the continuous-action distributions of the studied flow policies are well approximated by Gaussians. We propose the Gaussian Action Readout (GAR), which replaces a flow policy's final layer with a Gaussian readout trained on just 0.02% of eligible states, while keeping the rest of the network frozen. It samples with one action-network pass, gives closed-form likelihoods, and achieves near-teacher task success at up to twice the inference speed.

0.13

RoboAtom: Large-scale Atomic-Action Supervision for Robot Learning

cs.RO ·2026-10-29 ·Xueyuan Wei, Xuanxu Wang, Xudong Li, Yuping He, Wanshui Gan +8 more
benchmark
Abstract · 31 pages, 12 figures, including references and appendices

Fine-grained action annotations can support skill reuse and compositional generalization in robot policies. However, robot datasets typically provide task-level instructions without subtask annotations. Although some include subtask labels, inconsistent formats and conventions hinder cross-source joint training. We propose RoboAtom-Data, which adds unified, time-aligned atomic-action annotations to existing data from multiple sources, covering 1.45 million trajectories, 6.88 million skill-labeled segments, and 16,882.1 hours. Our pipeline combines a unified protocol, large-scale manual annotation, propagation across repeated executions with the same ordered skill sequence, and multi-stage quality control. We evaluate the dataset through automatic atomic-action annotation and robot control. We construct RoboAtom-Bench to standardize evaluation of skill recognition and temporal boundary localization. Trained on the annotated data, RoboAtom-Anno achieves better automatic annotation performance than all evaluated open- and closed-source models on both RoboAtom-Bench and the external WGO-Bench. For robot control, atomic-action supervision paired with hierarchical execution improves task success over task-level baselines on LIBERO-Long, LIBERO-Plus, and RoboTwin. We will release the data, benchmark, model, and construction pipeline to support further research on skill-level robot learning.

0.12

How Much Should a VLA Trust Its Reasoning? Step-Level Guidance as a Measured Action-Value Function

cs.RO cs.CVcs.LG ·2026-10-29 ·Dabin Seo, Eunhyek Joa
benchmark
Abstract · 6 pages, 2 figures, 4 tables

Reasoning vision-language-action (VLA) policies write a reasoning trace and then decode the action from it. A wrong trace yields a wrong action, and even a correct-looking trace can mislead the plan. Existing remedies ask whether the reasoning is right. We ask how strongly it should steer. CoC-CFG applies classifier-free guidance between the policy's context and the same context without the reasoning span. One weight per Euler step attenuates or amplifies the reasoning, and w=1 recovers the original policy. We treat the choice of this weight as a Markov decision process whose action values are measured, not bootstrapped. With the expert frozen and the noise fixed, applying a weight for one step and finishing the plan under the current controller returns that weight's value exactly. Q-adaptive guidance fits these values with small critics and improves the controller conservatively. In the Alpasim driving simulator it outperforms every fixed weight and online RL (SAC, GRPO) in open loop, and in closed loop it passes 262 of 300 scenes, against 252 for the original policy. On held-out frames of the Waymo end-to-end driving benchmark, its first three stages outperform every fixed weight, four adaptive-guidance methods, SAC and GRPO. On the hidden test set, a frozen Alpamayo 1.5 with rule-based camera and vehicle conversions and Q-adaptive guidance sets a new state of the art among published methods: 8.135159 rater feedback score, against 8.079 for the best published method.

0.11

CaPtain: A Pipeline for Post-Training Code-as-Policy Robot Agents

cs.RO ·2026-10-29 ·Yijia Fan, Xilin He, Yilei Jiang, Dongming Wu, Jiaxu Wang +1 more
Abstract

Across coding, web, and tool-use tasks, language-model agents are now routinely trained with the same recipe: supervised fine-tuning (SFT) to learn the interface, then multi-turn reinforcement learning (RL) to master it. Code-as-policy (CaP) casts robot manipulation in the same form, as the policy writes code against robot APIs and reads back the results, so the same recipe should apply. Yet existing CaP agents sidestep training: they either prompt closed frontier models or delegate manipulation to pre-trained vision-language-action models (VLAs) and high-level skills, while small open vision-language models (VLMs) fail outright. We present CaPtain, an open post-training pipeline that turns a small open VLM into a capable CaP agent, with no VLA, frontier model, or privileged state at test time. CaPtain trains and deploys the policy inside CaPtain-Harness, which exposes low-level perception, geometry, and guarded-motion primitives across simulation and real hardware. We first align the base model via SFT on synthetic trajectories generated with GPT-6 Astra, then apply multi-turn RL. Multi-turn RL must credit tens of agent requests, whose physical effects often surface turns later, from a single end-of-episode outcome. We therefore introduce a physical critic that reads privileged simulator state, never seen by the policy, and credits each request by how much it changes the predicted chance of success. Starting from Qwen3.5-4B (0% zero-shot), CaPtain-4B reaches 55.4% held-out LIBERO success after SFT and 73.8% after RL. On LIBERO-PRO, it reaches 78.5%, surpassing several frontier-model agents, and its RL gains persist across three perturbation suites (+17.6 to +21.2 points). Without target-domain training, it attains 83.3% on SimplerEnv WidowX, exceeds frontier models on CaP-Bench, and runs on a real robot after simulation-only adaptation. We release all code, data, and model weights.

0.10

Event2Reward: Learning Robotic Rewards from Progress Relations

cs.RO cs.LG ·2026-10-29 ·Junnan Nie, Zitian Fan, Jiachen Zhang, Junyi Lao, Likui Zhang +12 more
Abstract · 44 pages, including references and appendices

Learning task progress from robot demonstrations provides a practical way to construct dense shaping rewards for robotic reinforcement learning. A common approach uses normalized demonstration time as progress labels, implicitly assuming monotonically increasing task progress. During policy exploration, however, actual task progress can stall or even regress. Such models may mistake regressions for progress, assigning positive shaping rewards to actions that cause progress loss. Real suboptimal and failed executions offer corrective experience, but assigning absolute progress values to their states is difficult. Our key observation is that stalls, regressions, and recoveries establish local relative progress relations even when absolute progress is unknown. We introduce Event2Reward, which uses normalized temporal labels to anchor the progress scale and these within-trajectory relations to constrain local changes, combining absolute and relative progress supervision. Its progress predictions serve as potentials for reward shaping. We curate an event-annotated dataset from approximately 100K raw real-robot execution records, covering successful, suboptimal, and failed executions, without requiring framewise absolute progress labels. Under the same training budgets, Event2Reward improves average policy success by 18.16 percentage points over a baseline using only sparse task rewards across 40 tasks in four LIBERO suites. On real robots, average success after 30 minutes of wall-clock training reaches 60.0% on three precise insertion tasks and 90.0% on three object manipulation tasks, exceeding the sparse-reward baseline under the same budget by 36.7 and 23.3 percentage points, respectively. Ablation studies further show that absolute progress supervision and relative progress relations are complementary, with joint supervision yielding better downstream policy learning.

0.10

EmbodiExp: Compositional Experience Transfer for Embodied Agents

cs.RO cs.AIcs.CVcs.LG ·2026-10-29 ·Zhifeng Gu, Ying Chen, Rui Jiang, Lei Yi, Bing Wang
Abstract

Experience can help embodied agents turn high-level plans into grounded actions, but transferring it across tasks remains challenging: previously effective strategies may be inapplicable to the current state or fail to produce their intended outcomes. We present EmbodiExp, a framework for constructing transferable experiences and composing them online during embodied task execution. Our key observation is that tasks often share subgoals whose completion can be assessed independently of overall task success. We therefore introduce an outcome-grounded subgoal experience that specifies a subgoal, applicability conditions, an adaptable policy, and a verifiable outcome. During execution, applicable experiences guide action generation, while verified outcomes can establish the conditions required by subsequent experiences, enabling online composition. Subgoal-level outcome attribution further enables failure utilization: failed trajectories can still contribute verified local successes and condition-specific guidance on actions to avoid. Experiments on multiple embodied benchmarks under a fold-based protocol with frozen experience banks demonstrate the effectiveness of EmbodiExp.

0.09

PASA-KV: Attacking Action-Facing Visual Routes in World Action Models

cs.RO ·2026-10-29 ·Zhongbo Wu, Jiakai Huang, Zheng Zhang, Siyu Xu, Kaixiang Chen +4 more
Abstract

World action models (WAMs) increasingly couple internal visual or world representations with action generation, making their world-to-action pathways a critical yet poorly understood source of robustness. We investigate when perturbations to these internal representations actually become behaviorally consequential, and uncover a pronounced mismatch between representation perturbability and closed-loop vulnerability: attack effectiveness varies sharply across visual K/V routes, while larger representation displacement does not reliably imply stronger control failure. Our analysis attributes this mismatch to the action-conditioned readout, where keys determine where action queries read and values determine what content they receive. This motivates an action-facing view of vulnerability, in which the effect of a perturbation depends on how it is routed into action generation rather than on representation-space magnitude alone. We instantiate this view with Tail-AV, a short-probe signal for identifying vulnerable routes, and operationalize it in PASA-KV, which selects action-facing K/V targets before training a fresh task-shared attack. Experiments across multiple WAM systems and a physical robot validate readout-informed target selection, demonstrate strong closed-loop attacks, and reveal substantial system-level differences in resistance.

0.09

Continuous-Action World Models: Learning Cross-Embodiment Spatiotemporal Dynamics Priors

cs.RO cs.CVcs.LG ·2026-10-29 ·Liangyu Li, Qingwen Liu
Abstract · 12 pages, 6 figures

We introduce Continuous-Action World Models (CoAWM), a framework for learning cross-embodiment spatiotemporal dynamics priors that support robot action generation across embodiments and execution frequencies. At its core is a shared latent function space that captures the spatial and temporal structure of robot actions. We learn this space through two stages. First, function pretraining brings trajectories from different embodiments and sampling frequencies into a common representation, learning shared dynamics priors through action reconstruction and consistency across temporal sampling views. By accommodating heterogeneous robot trajectories, this stage broadens the range of data sources available for pretraining and facilitates scaling to larger and more diverse datasets. Second, we connect the pretrained function space to a world-model backbone, enabling observations, language instructions, and the current robot state to condition the prediction of a latent action function. A single decoder takes any query time and returns the corresponding end-effector action. By separating shared dynamics priors from task-conditioned action selection and temporal sampling, our approach provides a unified foundation for transferring robot experience across embodiments and control frequencies.

0.08

JEDI: Joint-Embedding Demonstration Compression for In-Context Imitation Learning

cs.RO ·2026-10-29 ·Jiazhao Liang, Zhewen He, Shuaihang Yuan, Hao Huang, Yi Fang
Abstract

In-context imitation learning (ICIL) adapts robot policies through test-time demonstrations, but long multimodal trajectories constrain how many examples fit in context. We ask what a compressed demonstration should preserve. We propose JEDI, which compresses demonstrations into latent slots trained to predict the query episode's next-stage keyframe in a frozen video-JEPA space. Task and process counterfactuals help the slots preserve execution details, while the predictor selects demonstrations from a bank. Post-training with keypoint tokens enables human-hand conditioning on robots. Experiments on LIBERO and real robots show that JEDI supports many-shot context; LIBERO objective controls show improved success and demonstration following over action-only compression. These findings support query-future prediction as a training objective for preserving execution-relevant information in compressed demonstrations.

0.08

Surrogate Uncertainty for Black-box Visual Grounding

cs.RO ·2026-10-29 ·Hanbo Zhang, Tianrun Hu, Cedar Site Bai, Anxing Xiao, David Hsu
Abstract

Visual grounding is crucial for robots to understand language instructions with visual observations. This paper addresses the problem of visual grounding uncertainties, where a robot uses a black-box vision-language model (VLM) to ground free-form natural-language instructions to target objects. Prevailing methods derive uncertainty from the VLM itself. Yet, outputs of the VLM, e.g., generated uncertainties, are from the same parameters and structures producing the predictions, and inherit the bias. We give a formal Bayesian definition of the problem that treats the VLM prediction as an observation. An external surrogate Bayesian model then quantifies per-query uncertainty over a local neighborhood of a held-out calibration set, decomposed into aleatoric and epistemic components. A hierarchical retrieval across input modalities realizes a compact joint-space neighborhood, mitigating the curse of dimensionality. The estimator further integrates with conformal prediction or model-based planning for active visual grounding, where robots are allowed to take verbal or physical actions to ground targets. Across grounding benchmarks and state-of-the-art VLMs, it consistently improves both calibration and task performance. On a real Fetch robot under open-world instructions, it enables principled trade-offs between confident action and information gathering, substantially improving robustness.

0.08

Ask to Evolve: A Unified Harness for Lifelong Embodied Navigation with Hybrid Action Space and Persistent Memory

cs.RO ·2026-10-29 ·Kai Sheng, Liuyi Wang, Jinlong Li, Haojie Dai, Chengju Liu +1 more
Abstract

Giving embodied agents the freedom to act autonomously, initiate interactions, and understand their surroundings through experience is a step toward bringing them to life in the physical world. Realizing this vision requires action, interaction, and memory to work together, so that each encounter can inform current decisions and contribute experience for future tasks. We present EvoLife, a unified navigation harness that enables a multimodal embodied agent to choose how to move, when to ask, and what experience to carry forward. Its hybrid action space supports waypoint navigation, local discrete adjustments, pixel-based navigation, and object-directed navigation. Active interaction allows the agent to seek clarification or assistance when it encounters ambiguity or difficulty, while persistent spatial memory binds observations and dialogue to places, retaining useful experience to guide subsequent visits. We evaluate eight benchmarks across four task families: vision-language navigation, object navigation, interactive instance navigation, and continuous multi-task navigation. With an average of only 1.87 active interactions per episode, our method achieves a first-visit success rate (SR) of 100% across four 100-episode sets spanning vision-language navigation and object navigation. Complete-sequence SR reaches 98% on 50 LHPR-VLN sequences and 94% on 50 GOAT-Bench sequences. In the R2R revisit setting, success weighted by path length (SPL) increases from 82.23% to 93.31% with retained experience. Multi-day physical navigation further demonstrates exploration, proactive help-seeking, and revisiting remembered places. Together, these results advance the vision of bringing embodied agents to life through autonomous action, proactive interaction, and the accumulation and reuse of experience.

0.08

Confidence-Modulated Safety Filtering for End-to-End Policies

cs.RO eess.SY ·2026-10-29 ·Javier Borquez, Jaime F. Fisac
Abstract

End-to-end policies, from task-specific controllers to robot foundation models, learn complex behavior from raw observations but are not guaranteed to remain safe, so they are often wrapped in a safety filter. The resulting system operates under two domains: the filter's operational design domain (ODD), where its assumptions hold and safety is certified, and the policy's competence domain, where it is reliable. Filters typically fix their conservativeness regardless of the policy's competence, trading performance inside the ODD against margin beyond it. We ask whether the policy's confidence should set that level, and propose a filter-agnostic pipeline that modulates it at runtime: inside the ODD, where every setting is safe, confidence buys performance; beyond it, low confidence adds margin, which helps when it flags where the filter's model fails. Preliminary F1TENTH simulations, where this holds, show the modulated filter combining the strengths of permissive and conservative filters.

0.08

Utility-WAM: Correcting Future-Induced Action Changes

cs.RO ·2026-10-29 ·Feihong Zhang, Xi Lin, Yulong Shi, Zuxing Lu, Yanghong Mei +5 more
benchmark
Abstract

World Action Models (WAMs) can condition robot actions on generated visual futures, but this additional context is neither cost-free nor uniformly beneficial. We examine what a predicted future actually changes: the action delta $Δa$ between the future-conditioned and future-free execution paths of the same model. Under a shared WAM checkpoint, this delta barely shifts average benchmark success ($+0.8$ points on LIBERO, $-0.8$ on RoboTwin~2.0), while generating visual futures increases inference latency by 1.7--1.9$\times$. These aggregate metrics conceal a concentrated failure mode that we term \emph{future-induced negative transfer}: while predictive deltas benefit long-horizon tasks ($+2.4$ points on LIBERO-Long) and remain negligible on most routine manipulations, they induce severe degradation on precision tasks---with \emph{Press Stapler} and \emph{Place Can Basket} losing 54.0 and 34.0 points, recurring as top failure modes across independently trained WAM variants. We introduce \textbf{Utility-WAM} for in-place correction of future-conditioned control. Its primary learning component is a residual adapter that repairs future-induced velocity deltas under a frozen shared WAM by aligning them with ground-truth dynamic guidance and penalizing excess error beyond the future-free baseline. The adapter selectively eliminates the majority of severe task drops (reducing losses to 6.0 and 14.0 points) while preserving long-horizon foresight. To eliminate superfluous video computation, a lightweight pre-generation gate reuses counterfactual action outcomes to selectively bypass video generation on 75.5\% of RoboTwin queries, cutting latency by 38\% while improving average success to 93.3\% on RoboTwin~2.0 and 98.9\% on LIBERO. On physical bimanual manipulation across three real-world tasks, Utility-WAM succeeds in 53 of 60 trials, compared to 43 for FAST-WAM.

0.07

PRIME W0: Physical Priors from World Models for Embodied Manipulation

cs.RO ·2026-10-29 ·Huabin Peng, Fagang Jin, Yancong Wei, Tengfei Ma, Chuhang Liu +2 more
Abstract

Vision-Language-Action (VLA) models excel at instruction following, yet most remain feed-forward: they condition actions only on the current observation and discard both long-horizon temporal context and the hierarchical structure of VLM representations. We present PRIME W0, a hierarchical conditional flow-matching framework that closes this gap through three synergistic designs: (1)per-layer KV injection from a pretrained VLM into an Adaptive MoE action expert; (2)dynamics-aware conditioning via asynchronous world-model subgoal-image prediction and a DINOv2 historical-frame semantic bridge; and (3)Action Dynamic Manifold (ADM) flow matching with covariance-structured noise that respects robot kinematic coupling. PRIME W0 is trained in four stages---VLM pre-training, DAT/NTP visuomotor pre-training, DiT flow-matching decoder assembly with WM injection, and end-to-end SFT---and achieves state-of-the-art results on LIBERO and real-robot benchmarks, with systematic ablations validating each architectural component.

0.07

TouchBridge: Tactile-Conditioned Human-to-Robot Transfer for Dexterous Manipulation

cs.RO ·2026-10-29 ·Haozhe Zhang, Gu Zhang, Zirui Zou, Yuzheng Chen, Jinlong Wu +5 more
Abstract

Human demonstrations offer an efficient and scalable source of data for dexterous robot manipulation, but differences in sensing and hand morphology complicate the transfer of contact-rich interactions. We present TouchBridge, a hardware--software co-designed framework for human-to-robot data transformation. A wearable human demonstration interface and a corresponding robot interface establish cross-embodiment correspondence in visual and fingertip tactile sensing. We introduce Tac-Retarget, a tactile-conditioned retargeting method that uses measured contact to align fingertip contact patches and preserve their spatial relations within and across hands. Unlike object-aware or simulation-based transfer pipelines, Tac-Retarget requires no object mesh, object-pose estimate, or physics simulation for retargeting, enabling direct and scalable conversion of human demonstrations into robot-domain training data. Combined with visual synthesis and camera-anchored action conversion, TouchBridge produces robot-domain demonstrations for fine-tuning a pretrained vision-language-action (VLA) model. Across four real-world contact-rich tasks, policies fine-tuned solely on TouchBridge-converted human demonstrations achieve a 72.5% average success rate, compared with 77.5% using an equal number of teleoperated robot demonstrations, while human demonstration collection is 7.0x faster. Under the same policy and training setup, Tac-Retarget outperforms DexPilot and UniDex by 25.0 and 42.5 percentage points, respectively.

0.05

People & Society 46

Micro-Edits: Computational Support for Fine-Scale Edits in Film

cs.HC cs.CV ·2026-10-29 ·Nhan Tran, Hijung Valentina Shin, Abe Davis, Mackenzie Leake
top venuecode/data
Abstract · UIST 2026 Best Paper Honorable Mention. Project page: https://microedits.github.io/

In film editing, every frame matters. While macro, content-level edits shape the high-level narrative of a film, small local edits often control timing and direct visual attention. For example, editors can adjust the precise timing of clips to control the length of a meaningful silence, or cut to shots that highlight a specific character's reactions to important moments. These small, detailed edits are slow and tedious to make in standard editing tools because users have to map high-level narrative goals to low-level frame-by-frame adjustments. At the same time, these fine-scale adjustments, which we call micro-edits, can be nuanced and expressive, with different choices serving different creative goals. In this work, we start by conducting a formative study with professional editors to understand the structure of micro-edits and how they are used. We then derive a mathematical formalization for this structure to develop tools that leverage AI and optimization to make exploring micro-edits simpler and more efficient for users. Finally, we validate our tools in multiple user studies and a wide range of video examples.

0.72

Crowds Believe in You: Online Prediction Markets Overestimate How Often People Achieve Their Own Goals

cs.HC cs.CYcs.SIecon.GN ·2026-10-29 ·Zihan Chen, Lei Nico Zheng, Di Zhu
Abstract · 23 pages, 2 figures, 14 tables; working paper

People increasingly share personal goals online and invite others to forecast them. Research on the planning fallacy finds that observers predict a person's behavior less optimistically than the person does, and prediction markets aggregate such judgments, harnessing the wisdom of crowds. We study 6,150 prediction markets on Manifold in which 1,660 users asked about their own futures ("Will I ...?"), reconstruct every order and execution, and compare these self-markets with markets about external events. Self-markets resolved YES 4.4 percentage points less often than their final prices implied, whereas external markets were nearly calibrated (0.9 points); the difference was 2.2 points at the same price level and 1.9 points within creators who made both kinds of market. Absent when markets opened, the overstatement emerged within a day of trading, even where the person never traded, and peaked in the week before close. In a market's first day, observers moved external-market prices down toward the base rate, but moved self-market prices 3.6 points further toward YES, except in markets naming an outcome the person would not want, which were not overpriced. The correction came late and largely from the person, whose orders supplied 38% of the eventual movement toward the outcome in self-markets, against 9% in external markets. Because self-market prices mainly failed to separate successes from failures, a recalibration learned from earlier markets improved later forecasts by only 1.5%. Crowds believe in the futures people pose for themselves, and the information that would correct them lies with the person.

0.58

Who Owns Digital News, and Who Gets Cited: Measuring Media Concentration Across 77 Countries and Seven AI Systems

cs.CY ·2026-10-29 ·Peter Carragher, Kathleen M. Carley
dataset
Abstract

Digital news outlets compete for audience attention. Search engines equipped with generative AI systems such as Google's AI overviews increasingly decide where that attention goes, as well as the revenue it represents. Yet, the lack of a consistent data collection infrastructure between national-level regulatory bodies has hindered between-country comparative analysis of media concentration, and with it any assessment of how these intermediaries reshape digital news markets. To address this issue, we compile an ownership dataset for 9,761 digital news outlets drawn from Wikipedia and Wikidata, achieving high agreement with independently compiled US and UK news directories. Using this dataset, we derive traffic-weighted concentration indices for 77 countries. We find that journalists report more owner pressure in countries with higher concentration, and that outlet-level measures understate concentration, with owner-level indices a median 1.30x higher. Applying concentration measures to AI-mediated news discovery, we find that citation distributions from major AI systems are up to 2.08x more concentrated than traffic, and that if Google AI Overviews mediated all organic search, the resulting rise in US concentration would be several times that associated with highly regulated media mergers and acquisitions. Finally, our analyses indicate that AI systems preferentially cite state-controlled media outlets when prompted in that country's native language. Together, these findings raise concerns about the effect AI systems may have on digital media. We release the dataset to enable the monitoring of digital news globally, so that the impact of disruptive technologies on digital media markets can be tracked as they emerge.

0.47

From Artifacts to Coordination: Exploring How Shared Canvases Drive Collaboration during Group Ideation

cs.HC ·2026-10-29 ·Saramsh Gautam, Andrew M. Webb, Mahmood Jasim
published
Abstract · Accepted to ACM GROUP 2027. To appear in Proc. ACM Hum.-Comput. Interact. 11, GROUP (January 2027). 21 pages, 9 figures. DOI: 10.1145/385940…

Recent years have seen a surge in tools that support collaborative ideation through shared canvases. While these canvases offer flexible workspaces, how collaborators leverage them to iteratively develop ideas by externalizing thoughts, organizing artifacts, and coordinating tasks remains underexplored. To address this gap, we conducted an exploratory study with seven triads planning a travel itinerary using an infinite-canvas technology probe. Drawing on participant interviews and video analysis of their canvas interactions, we observed that collaborators contributed in parallel and arranged artifacts freely without a predefined structure. This spatial organization allowed collaborators to coordinate implicitly, made less vocal participants' contributions visible, helped distinguish private exploratory regions from shared working areas, and supported the emergence of distinct roles through repeated patterns of contribution. We discuss design implications for canvas-based ideation tools, including surfacing unnoticed contributions, supporting emergent roles without formalizing assignments, and keeping spatial cues legible for AI integration in collaborative canvases.

0.45

Assessing AI Companion Chatbot Safety for Vulnerable Users: A Crisis-Stage Multi-Platform Evaluation

cs.HC ·2026-10-29 ·Huanyue Liao, Seok Min Lim, Yan Ling Lim, Shi Min Cheng, Kellin Pelrine +1 more
publishedcode/data
Abstract · Accepted at the GenAI4Health Workshop at NeurIPS 2026 (non-archival)

AI companion chatbots are increasingly used by vulnerable individuals, including adolescents in suicidal crisis, for emotional support. Existing safety evaluations focus on base models, leaving gaps in product-level assessment, crisis-stage sensitivity, and comparisons between companion apps and general-purpose LLMs. We introduce MIRROR-Eval (Multi-turn Interaction Risk Review: Observing suicidal-crisis Responses), a dataset of 960 annotated 20-turn conversations across eight deployed platform configurations, six clinically grounded personas, and two crisis stages: ideation and explicit planning. An LLM-as-a-Judge pipeline applies an expert-reviewed rubric covering user risk severity, chatbot risk awareness, and intervention behavior. We observe a three-tier harm structure: companion apps (Character.AI and Replika) and Grok 4.3 form a high-harm tier; Gemini 3.1 Pro is mid-range; GPT-5.5 and Claude Sonnet 5 show substantially lower harm. At planning, companion apps show increased glorification and normalisation, advancement toward harmful action, and harmful methods and instructions. Crisis recognition and empathetic language do not ensure safe responses. These findings highlight the need for product-level safety evaluation across crisis stages. Dataset and code: https://huggingface.co/datasets/imda-biztech/MIRROR-Eval. Warning: this paper includes examples that may be offensive or harmful.

0.40

Data Provenance and Ethics in the Curricula of U.S. Health AI Education Programs

cs.CY cs.AI ·2026-10-29 ·Raffaele Giancotti, Francesca Santucci, Amama Anjum, Monique de Villa, Rupsa Mitra +3 more
code/data
Abstract · 10 pages, 1 figure, 3 tables. Under review at PLOS Digital Health. Data and supplementary material available at https://doi.org/10.5281/zeno…

Failures of deployed clinical artificial intelligence (AI) systems have often originated in how their training data were generated. Whether the programs that train the health AI workforce teach practitioners to examine data provenance is unknown. We conducted a cross-sectional analysis of the publicly advertised curricula of health AI education programs at U.S. universities, identified through a structured multisource web search between April and July 2026. Two investigators coded coverage of data provenance (structural, dataset limitations, or data management only), ethics, and sustainability; collaborative structure; technical prerequisites; and the disciplinary background and gender of named instructors. The primary outcome was provenance coverage at the structural or dataset-limitations level. Of 43 candidate entries, 36 programs at 25 institutional units in 21 universities were included. Ethics was covered by 29 programs (80.6%; 95% confidence interval [CI], 65.0 to 90.2), data provenance by 8 (22.2%; 95% CI, 11.7 to 38.1), and structural provenance by 6 (16.7%; 95% CI, 7.9 to 31.9). None of the seven master's degree programs met the primary outcome, and no program addressed sustainability. Eighteen programs included team-based work, but only 5 assembled teams across clinical and technical backgrounds. Of 92 named instructors, 76.1% came from computational disciplines and 3.3% from the social sciences or humanities; among the 51 whose gender could be determined, 66.7% were men. U.S. health AI education programs widely taught ethics but rarely taught how clinical data are produced. None addressed the environmental cost of computation, and most instructors came from computational disciplines.

0.39

TokWatt: an open, boundary-annotated database of AI inference energy, from chat answers to agents

cs.CY ·2026-10-29 ·Abderaouf Bahi
code/data
Abstract

Electricity supply is becoming a binding constraint on the deployment of artificial intelligence (AI), yet public evidence on the energy used by a single AI task is scattered and hard to compare. Published figures for one chatbot answer differ by two orders of magnitude, largely because studies count different parts of the system and measure different units of work. We present TokWatt, an open database of 134 published estimates of AI inference energy drawn from 13 sources (2023-2026) and covering 20 tasks, from classic machine-learning inferences to chat answers, reasoning answers, long-document queries, image and video generation, and coding agents. Each estimate records its functional unit, five measurement-boundary components, method, reported interval, and the location of the value in its source. A reproducible procedure scales estimates to a common server-plus-facility boundary and derives per-task reference values with ranges. The reference values span six orders of magnitude: 0.34 Wh for a short chat answer (range 0.16-0.99 Wh), 4.7 Wh for a reasoning answer, 124 Wh for a short video clip, and 150 Wh for one task delegated to a coding agent. Five independent estimates for default production chat systems agree within a factor of 1.8, and the reference values shift by at most 12% under alternative harmonization parameters. A simple identity links personal use to grid load: E watt-hours per person per day across one billion people corresponds to an average draw of E/24 gigawatts. Data and code are released at https://github.com/AraoufBh/tokwatt, and an online calculator at https://araoufbh.github.io/tokwatt/.

0.38

Towards Audience-Driven Real-Time Data Storytelling

cs.HC ·2026-10-29 ·Angelica Lo Duca
published
Abstract

This paper introduces the concept of Audience-Driven Real-Time Data Storytelling (ADRDS), a framework that utilizes participatory and social sensing to enable audiences to actively shape the presentation of data stories. At its core, ADRDS employs a pre-authored decision tree, where the presenter defines the data story through a branching three-act arc (setup, conflict, and resolution). Each node contains a question, branching rules, thresholds for coverage, consensus, and confidence intervals, as well as one or more potential calls to action (CTAs). Audience members connect to the ADRDS framework via a URL on their mobile devices and answer questions in real time. ADRDS evaluates responses against the decision tree in real-time, following strong branches when thresholds are satisfied or activating fallback paths when signals are inconclusive. ADRDS then selects the final action consistent with the traversed path and the audience's demonstrated readiness, subject to fairness safeguards. ADRDS transforms presentations of data stories into audience-shaped decision processes, where live data actively drives the narrative. This work paves the way towards a new form of data storytelling, where the final data story depends on the audience's background and interests.

0.38

Robot Navigation and Canine Welfare: A Pilot Study Surfacing Challenges for Cross-Species Legibility and Proxy Validity in Animal-Machine Interaction

cs.HC cs.RO ·2026-10-29 ·Haoxiang Zhang, Nancy Smith
publishedposition
Abstract · 8 pages, 2 figures, Accepted at ACI2026

Autonomous delivery robots increasingly share residential spaces with companion animals, yet navigation behavior is rarely designed with animal welfare in mind. This paper presents a pilot study evaluating a curve-out maneuver, a laterally arcing pass-by trajectory triggered at the dog's alert distance, proposed as a candidate welfare-informed navigation intervention grounded in canine social communication research and legible motion theory. Following a research through design approach, we developed and evaluated a first design iteration of the artifact in a naturalistic home environment with a companion dog. Contrary to our hypothesis, the curve-out produced a more acute stress response than the straight-line baseline. We argue that discontinuous execution limitation, the robot stopping and reorienting multiple times to approximate the arc, undermined legibility, which in itself became a stressor. The pilot surfaces the legibility of robot behavior to non-humans' eyes and proxy stimulus validity as structural challenges for AMI and repositions the design problem for future work.

0.37

Evaluating Conversational Explainability in Healthcare: Comparing the Explainability Assistant and a Traditional XAI Dashboard

cs.HC ·2026-10-29 ·Eduard Barbu, Rodion Krjutškov, Anneli Uusküla
Abstract · 18 pages, 4 figures. Accepted at EXPLIMED 2026 (Third Workshop on Explainable Artificial Intelligence for the Medical Domain). To appear in …

Conversational interfaces may make explainable AI (XAI) more accessible to users without machine learning expertise, but their benefits over dashboards remain insufficiently characterised in healthcare. Existing systems have typically been evaluated in isolation, leaving unclear whether conversational interaction improves interpretability beyond the underlying explanation methods. We report a within-subjects user study comparing the Explainability Assistant, an LLM function-calling conversational XAI system, with a traditional dashboard for explaining a heart disease classification model trained on the Cleveland Heart Disease dataset. Both interfaces expose the same Decision Tree classifier and post-hoc explanation methods: SHAP feature attribution, DiCE counterfactuals, and what-if analysis. This design supports comparison of interaction modalities while holding the model and explanation methods constant. Five participants with healthcare backgrounds, from medical students to practising clinicians and public health researchers, completed parallel task blocks on both interfaces; three participated in follow-up interviews. Task accuracy, Likert ratings, free-text questionnaire responses, and interviews revealed complementary strengths and weaknesses rather than a clear preference for either modality. The assistant offered easier entry and patient-specific explanations, while the dashboard supported simultaneous inspection of global and local information. All three interview participants recommended integrating conversational and dashboard elements within a single clinical workflow.

0.35

Early-Buyer Groups and Fast-Wallet Teams on a Bonding-Curve Launchpad: A Prospective Measurement Study of pump.fun on Solana

cs.SI physics.soc-phq-fin.TR ·2026-10-29 ·Fail Gafarov
Abstract · 66 pages

We study coordinated groups of wallets on a bonding-curve launchpad (pump.fun on Solana), using a complete record of 200\,443 token launches, 15\,854\,121 curve trades and 51\,025\,958 SOL transfers over 88.3 hours. A prospective, multi-channel discovery procedure (co-buying, shared funding, joint signing, transfers, consolidation) yields 23\,710 wallet groups. Without ground-truth labels we validate them by out-of-sample link prediction, independent non-trading links and a comparison with naive co-buy clustering at equal numbers of flagged tokens: raw clustering collapses into one giant component, yet it ranks tokens at least as well as the groups, so the value of the structure is descriptive. Group activity is highly concentrated and splits into serial creators and long-lived rings of early buyers. Known members buy earlier than other buyers of the same token and extract more cash than an activity-matched placebo set, while late buyers' net cash flow is negative; tokens with members show a pump-like price path, but the contrast vanishes when controls are restricted to comparable tokens, so we report it as an association. A follower entering at the earliest group signal loses money out of sample in all 32 configurations and does no better than random entry. A few pools of tens to over a hundred fast wallets are associated with a two- to four-fold higher chance of reaching 20 SOL, also within creators, with a smaller excess at 60 SOL and none for graduation; on a final hold-out the team rules keep a precision of about one half to three quarters, yet many signals arrive when the curve already holds 20 SOL, one team supplies most of them, and they add no information beyond early curve state. We also document evaluation errors that turn the same data into apparent profit.

0.34

InterA11y: Interactive Web Accessibility Error Detection with LLM Agents

cs.HC ·2026-10-29 ·Mingyuan Zhong, Mengqi Shi, Davin Win Kyi, Ajit Mallavarapu, Xia Chen +2 more
code/data
Abstract · We release our code and data at https://github.com/jszh/InterA11y

Large Language Models (LLMs) have shown potential to expand the coverage of web accessibility evaluation. However, current LLM-based checkers neglect much of the interactive and dynamic web contexts, resulting in limited coverage, high false discovery rates, and inconsistent performance across models. We present InterA11y, an agentic accessibility evaluation system that combines rubrics and interactive tools to support LLM agents in evaluating accessibility effectively. The rubrics decompose WCAG success criteria (SC) into testable aspects and guide agents in collecting and interpreting evidence. The interactive tools inspect or operate a webpage deterministically to collect evidence, such as focus appearances and error messages. We developed these rubrics and tools on 682 curated test cases grounded in established accessibility guidelines that target 10 commonly violated SC. We evaluated InterA11y on 197 expert-annotated cases sampled from 54 real webpages. Results show InterA11y achieves higher recall than the strongest evaluated LLM-driven baseline (79.0% vs. 44.4%), while maintaining comparable precision to best rule-based checkers (93.1% vs. 90.3-97.6%), with little variability between models. Further, when equipped with our interactive tools, general-purpose web agents detect more accessibility errors with fewer false positives and less token usage than without these tools.

0.33

Behind the Paper Explosion: Author Growth and Widening Gaps in Computer Science

cs.DL cs.CY ·2026-10-29 ·Jiayi Li, Bowen Niu, Bingsheng He, Zhaomin Wu
Abstract · 12 pages, 9 figures

The rapid growth of computer science (CS) papers increases the work needed to review research and find relevant studies. The growing use of AI tools in scientific writing makes it increasingly important to identify the authors behind this growth. Previous studies have tracked publication growth and followed researchers' careers, documenting the increasing number of authors and the many researchers who publish only briefly. However, annual averages of papers per author reflect different sets of researchers and may hide changes among the same authors. This study addresses this gap by linking individual changes to overall CS paper growth using arXiv, DBLP, and conference submission records for 2010-2025. With each paper divided equally among its authors, the total paper count equals the number of authors multiplied by their average papers per author. This relationship provides a decomposition of the two contributions of paper growth. This decomposition attributes nearly all growth in CS paper volume to an increase in the number of publishing authors. The analysis of author numbers further distinguishes expansion through new and returning arXiv authors from the replacement of authors no longer publishing there. Changes in who publishes also affect the average, motivating comparisons of the same authors over time within groups with similar past paper counts. This comparison identifies substantial replacement within the growing author group. New authors average fewer papers under this counting method, keeping the overall average nearly unchanged despite increases among authors who published in both years. Moreover, statistics show that these increases are uneven, with widening gaps even among authors who previously published similar numbers of papers. These findings provide insights into peer review, community participation, researcher evaluation, and equity within the CS community.

0.33

Inkitiate: Initiating and Steering Agent Assistance through Ink in a Spatial Workspace

cs.HC ·2026-10-29 ·Liangwei Wang, Nan Chen, Luna K. Qiu, Zilong Wang, Fugee Tsung +2 more
Abstract · 18 pages, 11 figures

Handwriting and sketching help people develop ideas around information. As this work unfolds, users may seek assistance with finding, understanding, and organizing materials. We investigate how ink can initiate and steer agent assistance in this process. An eight-participant formative study explored desired assistance and interaction expectations, informing three connected activities: bringing materials in, thinking with materials, and shaping the space. We present Inkitiate, a spatial workspace for ink-based agent assistance in knowledge work. Through ink, users retrieve external materials, build on accumulated content and results, and reorganize the canvas, while inspecting and guiding agent actions. In a study with eighteen participants, users reported favorable usability and low perceived workload, and used retained ink and materials to develop further ink queries. An ink-query vocabulary and cases reveal how users combine linguistic and graphical expressions to direct agents. These findings inform ink-based agent interaction design for expressive requests and continued work with materials.

0.31

Integrating of Attack Patterns and Threat Modeling to Support Security Analysis in the Automotive Domain

cs.CE ·2026-10-29 ·Eduardo Santos de Oliveira Marques, Saulo Moraes Villela, Heder Soares Bernardino, Alex Borges Vieira, Bernardo Martins Rocha +1 more
Abstract

The increasing connectivity of cyber-physical systems in the automotive domain demands systematic approaches for identifying and mitigating cybersecurity threats. While ISO 21434 defines a structured workflow for Threat Analysis and Risk Assessment (TARA), it does not prescribe a specific method for identifying attack paths or evaluating risks. This paper proposes a conceptual model grounded in ISO 21434 that integrates the STRIDE threat classification with the CAPEC attack pattern database, focusing on the identification of attack scenarios and risk assessment. Its applicability is demonstrated through a case study involving Bluetooth-based threats to an automotive headlamp system.

0.31

Measuring Organizational Influence on Digital News: An Information-Theoretic Framework for Media Pluralism and Editorial Independence

cs.CY ·2026-10-29 ·Peter Carragher, Kathleen M. Carley
Abstract

News consolidation is accelerating, yet regulators have no measurement frameworks for determining whether owners shape editorial decisions. Existing evidence comes from single cases, where pluralism and independence are measured separately, if at all. We ask whether ownership impacts editorial decisions across the digital news ecosystem, and if so, how acquisitions affect this. We derive editorial independence and internal pluralism using partial information decomposition, which separates news coverage overlap with sister outlets from coverage overlap with all outlets. Using public datasets, we apply our measures monthly from 2018 to 2026 to 1,364 US news websites on three editorial decisions: which events they cover, which sources they link to, and what they put on their front pages. We find that co-owned outlets share up to 40 times more information about these decisions than expected by chance. Ownership accounts for far more variation in independence than network affiliation does and, upon being acquired, outlets lose independence sooner than matched controls, quickly converging on their new owner. In broadcast news, by comparison, we find that it is network affiliation, not ownership, that accounts for more content overlap. Finally, we discuss policy implications and suggest that media regulators have blocked large deals rather than editorially detrimental ones. We hope that this work can serve as a framework for comparative media analyses going forward.

0.30

Data Driven Modeling of a Magnetic Newton's Cradle using the Optimization of a Discrete Loss

cs.CE physics.class-phphysics.comp-ph ·2026-10-29 ·Dimitris Koumoutsakos, Sergey Litvinov
Abstract · 30 pages, 19 figures, 8 tables

We investigate the dynamics of a magnetic Newton's cradle (MNC) using experiments, simulations through forward and inverse data driven models. In the MNC magnetic forces replace the instantaneous elastic collisions of the classical design. We compare point-dipole and finite-current-loop interaction models, identifying their parameters from direct measurements as well as through pendulum trajectories using the method of Optimizing a Discrete Loss (ODIL). For a synthetic three-pendulum twin, ODIL reproduces discretized trajectories to round-off accuracy and infers accurately magnet strength and damping from sparse data even when only one pendulum is observed. Model comparison shows that the point-dipole model compensates for near-field errors by underestimating the magnetic moment and produces structured residuals at each momentum transfer. The current-loop model removes this bias and, on full-precision recordings, also recovers the magnet radius. The continuous magnetic coupling reproduces the classical cradle's key behavior while enabling richer dynamics, including $N$ normal modes and a static repulsion offset. These results demonstrate that ODIL can infer physical parameters, and quantify model-form errors from sparse, noisy observations. Ultimately, this work establishes the magnetic Newton's cradle not merely as a physical curiosity, but as a rigorously quantifiable testbed for computational inverse modeling.

0.29

Device-Induced Mobility Heterogeneity and Its Impact on Federated Next-Location Prediction

cs.CY cs.NI ·2026-10-29 ·André Felipe Zanella, Paula Delgado de Santos, José Suárez-Varela, David Solans Noguero, Andra Lutu +1 more
Abstract

Predicting human mobility is central to network management, urban planning, and location-based services, yet privacy concerns are increasingly pushing such tasks toward federated learning. A key challenge in this setting is the strong heterogeneity of mobility data across users, which leads to non-IID data distributions and can significantly impact model performance and fairness. In this paper, we present a large-scale measurement study of device-level mobility heterogeneity using data from a major MNO in a dense European capital. Our dataset comprises mobility traces derived from network events, covering over 50,000 UEs across 178 smartphone models. We group devices into tiers based on their market segment and show that mobility characteristics, such as spatial range and regularity, vary systematically across these tiers, revealing a previously underexplored source of bias in mobility data. We investigate the implications of this heterogeneity on the use case of federated next-location prediction. In particular, we find that balancing device tiers across silos through tier-aware client selection yields the best results, improving fairness by up to 34.49% without degrading accuracy. In contrast, a FedProx-only approach achieves a 21.46% fairness improvement, offering a reasonable alternative when device information is unavailable, though with lower overall gains.

0.29

Personalised 3D Mnemonics for Vocabulary Learning in Context

cs.HC ·2026-10-29 ·Karolina Trajkovska, Matjaž Kljun, Maheshya Weerasinghe, Klen Čopič Pucihar
Abstract

Augmented reality (AR) can support vocabulary learning by placing mnemonic cues in the physical contexts where new words are encountered. Prior work has augmented real objects with bilingual labels, audio, keywords, and 3D visualisations of those keywords interacting with the objects, but these cues are typically predefined by system designers. This limits scalability and reduces learners' ability to construct personal mnemonic links. By grounding mnemonics in objects that learners already recognise, use, and associate with prior experiences, a system could connect new vocabulary to existing mental models rather than requiring learners to adopt designer-defined associations. We outline a proof of concept and a storyboard of a generative AR approach in which learners create personalised 3D mnemonics by transforming familiar objects in their surroundings into situated mnemonic anchors. We position this concept within a timeline of related work and argue that recent advances in generative 3D asset creation enable a new design point: personalised 3D mnemonics that do not merely place content in the physical learning environment, but modify familiar real-world objects so that they become part of the mnemonic itself.

0.28

GraspPressure: Understanding Pressure Input With Busy Hands

cs.HC ·2026-10-29 ·Anand Kumar, Aditya Shekhar Nittala
published
Abstract · To appear in the Proceedings of the ACM on Human-Computer Interaction (PACM HCI), presented at ACM Interactive Surfaces and Spaces (ISS 2026…

In many real-world scenarios, users interact with digital systems while their hands are partially occupied, such as when holding tools or everyday objects. Prior work on busy-hand interaction has focused on microgestures and motion-based techniques, which rely on finger articulation or relative movement. In situations where a stable grasp is required, such as handling fragile, heavy, or precision objects, such motion can be undesirable or impractical. Pressure-based input offers a complementary modality that enables interaction without changing finger pose, yet its use while holding objects has received limited empirical attention. To explore this, we conducted two within-subject experiments examining users' ability to perform controlled pressure selections and analyzing finger-specific performance across different grasps and geometries. Results indicate that participants can reliably perform pressure-based selections in these conditions, achieving high success rates (92.7% on average) with low selection times (1.74s on average). Our findings characterize the potential of pressure input as a complementary microinteraction modality under grasp-constrained, busy-hand scenarios.

0.28

Reconstructing Missing Mandi Prices in North-East India with Regional Factors

cs.CY ·2026-10-29 ·Pritam Deka
benchmark
Abstract

Agricultural price panels contain prolonged and clustered reporting gaps, yet their reconstruction is often evaluated by deleting isolated observations. We investigate a regional-factor approach to weekly mandi-price reconstruction in five North-East Indian states, with Bihar and Goa as additional comparison states. The model combines a common price path with persistent market-specific deviations; an adaptive ensemble weights alternative borrowing structures using internally masked reports. On a matched retrospective benchmark against ten baselines, pooled North-East mean absolute percentage error is 19.3%, compared with 26.5% for SAITS. The corresponding calendar-aligned panel/time-block bootstrap interval for the error difference is 2.8-12.7 percentage points. Removing the target state's monthly aggregate leaves pooled error at 19.4%, but eliminates the apparent advantage in Goa. An origin-restricted pipeline combining week-close input processing with revised ensemble validation yields 19.7% error, compared with 30.8% for carry-forward; this contrast does not isolate individual pipeline changes. Gains vary across states, and uncertainty intervals remain too narrow in several small panels. These findings support further study of structured retrospective reconstruction, not deployment claims: model development used earlier results, publication dates for monthly inputs are assumed, and confirmatory evaluation on untouched data remains outstanding.

0.27

Apple Vision Pro: The good, the bad, the ugly

cs.HC ·2026-10-29 ·Santiago Berrezueta-Guzman, Andrei Koshelev, Habiba-Loai Khalil, Stefan Wagner
survey
Abstract · Pre-print submitted in Elsevier

Apple Vision Pro (AVP), the headset Apple introduced as its first ``spatial computer,'' built by a single, vertically integrated vendor, entered the market in February 2024 and has since accumulated a fast-growing body of primary research spanning surgery, education, accessibility, design, and human-computer interaction. In this paper, we adopt the tripartite structure of Leone's 1966 film to organize a systematic review of this literature: we cataloged the validated, working applications of AVP (the good), the persistent technical and ergonomic limitations of a still-early-generation headset (the bad), and the ethical concerns and structural harms that follow from deliberate design, marketing, and deployment choices (the ugly). The review is built from a Scopus export of 218 records mentioning AVP, which we narrowed to 104 papers (2024--2026) with a primary focus on AVP, using explicit inclusion/exclusion criteria that we report in full, alongside a PRISMA-informed flow diagram of the screening process. We find the corpus heavily skewed toward validating benefits (medicine and surgery, education, and software tooling account for over half of the included papers) and comparatively thin on documented risks or misuse, a pattern we treat as a finding in its own right rather than a limitation to be explained away. We further show that several recurring issues, namely price, simulated social presence, and the gaze-tracking privacy architecture, resist a clean separation into ``bad'' or ``ugly'' using the published evidence alone, since both readings remain compatible with the same data. We close by identifying where the evidence base most needs to grow: independent risk and ethics research, bystander- and equity-focused study designs, and disclosure that would let researchers distinguish an engineering backlog from a structural choice.

0.27

Wearable Biosensors for Early Detection and Continuous Monitoring of Postpartum Depression: A Human-Centered Approach

cs.HC ·2026-10-29 ·Emma Ricci-De Lucca, Elizabeth L. Murnane
Abstract · 2024 PhysioCHI: Towards Best Practices for Integrating Physiological Signals in HCI

Postpartum depression (PPD) is the leading cause of maternal mortality in the United States. However, there are significant gaps in PPD screening, detection, and treatment. A main challenge stems from limitations in existing methods for early PPD diagnosis and subsequent monitoring and intervention. Mobile technology that incorporates interactive interfaces and continuous wearable biosensors holds major potential to augment clinical decision-making and enhance PPD care. A number of compelling research directions exist in effectively developing such systems, including user experience design, sensing accuracy during perinatal stages, clinical integration and interoperability, acceptance and adoption, and privacy and ethics. By focusing on wearable devices that center women and their experiences, this paper discusses such challenges and opportunities at the intersections of HCI, inclusive physiological computing, and women's health.

0.27

What It Costs to Trust a Source: A Reproducible Method for Institutional Assessment of Author Disambiguation

cs.DL ·2026-10-29 ·Simon Porter
Abstract

Records of research outputs are curated with care by universities, and the effort is shaped by the external sources from which candidate records are drawn. The curation required before the author identities in such sources can be trusted is examined, and a reproducible method for estimating its cost is set out. Four tests are described; two can be run on public data alone, and two are conducted against an institution's own curated record, taken from Symplectic Elements Discovery portals. The method is demonstrated on OpenAlex and Dimensions across 30 institutions. Of the publications attributed to matched researchers, 50.4\% were confirmed by the institutional record for OpenAlex and 69.6\% for Dimensions, and fewer trusted researcher profiles were retained from OpenAlex at a matched confidence threshold in each of the 30 institutions. Carried through to effort, integrating OpenAlex requires 1.5 times the person-days of review and entry of Dimensions even when each feed is filtered at the threshold that minimises the amount of review required.

0.27

Unlocking with the Heart: Investigating User Perceptions of ECG-Based Biometric Authentication

cs.HC cs.CRcs.CYcs.ET+1 ·2026-10-29 ·Farnaz Farid, Linet Momposhi, Tamara Watson, Alana Maurushat, Abubakar Bello +1 more
survey
Abstract · 28

Password-based authentication remains challenging for many users, especially for those who face usability and accessibility issues. The burden of maintaining security can cause user fatigue, frustration and non-engagement. To address these issues, we conducted a pilot study to assess ECG-based biometric authentication as a passwordless alternative and to gather adult participants' perspectives on this approach. The research consisted of a survey study, an authentication prototype and a follow-up usability experience questionnaire. Our approach combined an online survey of 337 adults examining authentication practices and perceptions, with a separate usability evaluation in which 49 participants interacted with a simulated ECG-authentication and password interface in a health-care application prototype, then completed a post-test usability test questionnaire. The survey examined how adults across age groups use and perceive various authentication techniques, their experiences with biometrics, their views on security and convenience, and their willingness to adopt ECG authentication. In parallel, the usability prototype and questionnaire captured participants' responses after hands-on interaction with an ECG authentication prototype.

0.26

Comparing wins above replacement across baseball history

cs.CE stat.ME ·2026-10-29 ·Michael Doane
Abstract

This paper introduces iWAR, a system for estimating wins above replacement from historical Retrosheet play-by-play records. We compare its runs-allowed (RA) and fielding-independent pitching (FIP) versions with Baseball-Reference WAR (bWAR), FanGraphs WAR (fWAR) and run-based alternatives. The common historical sample contains 1,471 team-seasons across 79 years and 1,441 consecutive franchise pairs. Under the fixed conversion to wins, same-season mean absolute error was 3.695 credited wins per 162 games for iWAR RA, compared with 4.724 for bWAR and 4.744 for fWAR: reductions of 21.793% and 22.112%. The paired differences were -1.030 and -1.049, with conditional 95% calendar-block intervals of [-1.367, -0.706] and [-1.398, -0.725]. Calibration preserved most of these advantages. A broader bWAR comparison across 118 years retained a smaller advantage; modern differences were smaller or sometimes reversed. PythagenPat reconstructed records more closely than every WAR total. Next-season replay gave nearly identical fixed-map errors for iWAR FIP and fWAR, while calibration changed their ordering. An amended chronological learned-model analysis of 1,552 franchise pairs gave neural RMSE 10.163, versus 9.946 for the development-selected ridge comparator (difference +0.217; conditional interval [0.015, 0.474]). This work provides a new, specialized framework for analyzing WAR across eras and provides context for future work using learning models.

0.26

Out of sync: Accommodation, evaluation, and emergence in five weeks with an agentic AI teammate

cs.HC cs.CY ·2026-10-29 ·Nia Nixon, Jaeyoon Choi, Pedro Martins de Bastos, Mohammad Amin Samadi
Abstract · 21 pages, 5 figures. Submitted to British Journal of Educational Technology

As agentic AI in education shifts from reactive tools toward partners that take initiative and persist over time, a central question is how learners relate to, evaluate, and benefit from artificial agency. We conducted a longitudinal experiment in an undergraduate learning analytics course, in which 17 student teams (N = 48) collaborated with a persistent LLM-based teammate across weekly sessions. The agent decided when to intervene through a trigger-based architecture, and its social role was manipulated between teams: a Cognitive Agent that challenged quick consensus or a Supportive Agent that invited quieter members to contribute. Combining chat discourse, post-session surveys, and spoken reflections, we identified distinct patterns unfolding across three timescales. Students accommodated their language to the agent's role immediately, using more social language toward the Supportive Agent from the first session. Their helpfulness evaluations were recalibrated from session to session in relation to the agent's observed behavior, rather than remaining anchored to students' pre-existing attitudes toward AI. A slower crossover emerged in perceived collaboration process and positive-affect word use in reflections: Supportive-Agent teams began below Cognitive-Agent teams and ended above them, although students rated the Supportive Agent less helpful throughout. Because these responses follow different trajectories, single-session evaluations of agentic AI may miss changes that emerge only through repeated interaction.

0.25

Beyond Preference Following: Benchmarking Profile-Query Arbitration in Personalized Text Transformation

cs.HC ·2026-10-29 ·Tenghao Ji, Qichen Sun, Zhuoran Lu
benchmark
Abstract

Writing assistants use persistent preferences to personalize responses, but these can conflict with the current request. Responses can meet every requirement yet serve the task poorly, while task failures need not involve following a conflicting preference. We introduce Profile Conflict Arbitration (PCA-Bench), a benchmark of 64 rewriting tasks and 256 task-preference pairs. We construct these pairs by varying the profile preference for each fixed text and request, covering different relations between the preference and the request. We assess whether responses meet requirements, how well they achieve the request's goals, and whether they follow preferences. Across three LLMs, we find that when preferences compete with task goals, a higher success rate in meeting both requirements and preferences does not mean better responses: it can come with a greater quality loss compared with responses without a profile. Similar rates of preference following do not imply the same ability to meet the current task's requirements when preferences conflict with them.

0.25

The Price of a Chance in Artificial Societies

cs.CY ·2026-10-29 ·Shuqi Zhao, Boyu Wang, Shuangyue Niu, Chenhao Zhang, Zhanzhan Zhao
Abstract

Rapid advances in large language models (LLMs) have enabled autonomous agents to interact, coordinate, and pursue individual goals in artificial societies. However, our understanding of how reward structures shape their competitive behavior and collective outcomes remains incomplete. We simulate 200 LLM agents competing over 50 rounds to maximize individual wealth. We define overcompetition and lying flat as population investment above and below a Bayesian adaptive baseline, respectively. At the lowest tested reward, populations lie flat; at the two highest, they overcompete, most strongly at intermediate winner coverage. The investment committed to pursuing an uncertain reward constitutes the price of a chance. Agents near the qualification cutoff can invest heavily yet receive no prize, finishing poorer than nonwinners who invest less. Agents' stated rationales link investment decisions to perceived qualification prospects and expected returns. At the collective level, greater investment does not consistently improve ability selection. With a fixed total prize budget, offering smaller prizes to more winners can reduce investment while preserving selection quality relative to the corresponding top-ability target. Further broadening reward coverage yields additional savings but reduces selection quality. These findings show how reward design shapes overcompetition and lying flat, distributes the costs of pursuing opportunity, and influences what competition collectively achieves.

0.24

Lookover: Scaffolding Student Participation in Social Annotation

cs.HC ·2026-10-29 ·Yubo Zhou, Xinyi Xu, Zesen Zhao, Anhong Guo, Xinyue Chen
published
Abstract · 3 pages, 1 figure. Poster at UIST Adjunct 2026, the 39th Annual ACM Symposium on User Interface Software and Technology, November 2-5, 2026,…

Social annotation platforms such as Perusall and Hypothesis are widely used to support student engagement with course readings, often through required commenting. Yet visible participation does not necessarily produce sustained or substantive peer exchange. Analyzing six weeks of Perusall activity from a graduate course (2,873 comments), we found that participation was often fragmented: 53.5% of comments received no reply, 38.3% were posted within the final 12 hours before the deadline, and over one-third of a double-coded sample went no further than repeating or agreeing with the reading or a peer. Interviews pointed to two linked difficulties: locating a discussion worth entering and contributing something that meaningfully advances it. We present Lookover, an AI-assisted annotation system that scaffolds both stages without writing for students. Discussion Overview helps students identify active themes and relevant threads, while Spark Thinking poses questions that encourage them to move beyond reaction and restatement toward questioning, extending, and substantively responding to peers. A preliminary study with five participants found Discussion Overview valuable for those who noticed it, while the usefulness of Spark Thinking depended on how much a participant already knew about the topic.

0.24

A Dantzig-Wolfe reformulation of the EndoWake model for optimal wind farm layouts

cs.CE math.OC ·2026-10-29 ·Martina Fischetti, Matteo Fischetti
Abstract · 25 pages, 1 figure, 8 tables. Submitted to international journal. Replication package: https://doi.org/10.5281/zenodo.23636074

Wind Farm Layout Optimization places turbines to maximize the expected energy under a wind rose, each turbine lowering the wind downwind. In the EndoWake model the wind field is a variable, propagated by linear inequalities, and layouts are certified by branch-and-check. On a ladder of grids it certifies the $10\times10$ grid and leaves the larger grids open after one hour. We tighten the relaxation by a convexification computed rather than derived: every pattern of turbines inside a small tile of the grid is enumerated once with its power and enters the master as a column. The device is valid because the wind field is antitone, and exact, for the cubic power curve, with a Benders cut at integer layouts. Reduced-cost fixing and propagation between overlapping tiles shrink the master before the tree search, and the tree repeats the fixing at every improving incumbent. The incumbent comes from a tile heuristic or an LP dive. On the same computer as the branch-and-check, the $12\times12$ and $14\times14$ grids close in 48 and 395 seconds. The $16\times16$ grid is certified on five seeds from random starts, one job per seed, the heuristic on a GPU, in 1413 seconds under the uniform rose and 6484 seconds under a non-uniform one. Times are medians. A pre-specified test set of 36 unseen instances with forbidden cells confirms the advantage in the axial setting. Certified means closed at the solver's gap of $10^{-4}$: the optimum lies within it plus a residual of at most 0.00042%.

0.23

Reversing Polarization through Tolerance-Aware Recommendation and Adaptive Diversification

cs.SI ·2026-10-29 ·Zhanzhan Zhao, Shuqi Zhao, Zichen Fan, Dana Randall
Abstract

Recommender systems face a tension between broadening exposure to diverse viewpoints and avoiding interactions that reinforce extreme positions. This tension is especially consequential in already polarized populations. Here we investigate whether recommendation strategies can reverse polarization rather than merely slow its development. Building on the Attraction--Repulsion Model of opinion dynamics, we introduce two complementary design principles: Tolerance-aware recommendation, which seeks to learn each individual's tolerance for disagreement and reduces exposure to viewpoints beyond that tolerance, and adaptive diversification, which broadens exposure for agents with more extreme opinions while retaining stronger personalization for moderates. Agent-based experiments show that highly personalized recommendations intensify polarization under polarized initial conditions. Tolerance awareness alone primarily slows polarization, whereas combining it with adaptive diversification can reduce polarization below its initial level and, under suitable conditions, restore consensus. Additional experiments incorporating persistent extreme influencers and a spiral-of-silence mechanism show that the combined strategy can mitigate polarization under social pressure and selective participation. To examine whether these findings extend beyond predefined opinion-update rules, we repeat the experiments on a simulated social platform populated by 100 LLM-driven agents prompted to move toward opinions similar to their own and away from those that differ substantially. Across three runs, the combined strategy is again the only recommendation policy tested that reverses polarization in every run. These findings distinguish polarization prevention from reversal and identify tolerance-aware, adaptive exposure as a promising recommendation design principle for reconnecting divided populations.

0.23

A System Dynamics Model for Zero-emission Heavy Vehicle Transition in Australia

cs.CE ·2026-10-29 ·Elnaz Irannezhad, Seyed Hossein Hosseini, David A. Hensher
Abstract

What would it take to accelerate the adoption of zero-emission trucks? Even if battery electric trucks become cheaper to own over their lifetime than diesel trucks, that alone is not enough. This study develops a system dynamics model of rigid and articulated truck fleets, calibrated to Australian fleets. Under current policy regimes, more than a third of the fleet still runs on diesel in 2050. Battery and fuel cell trucks become as attractive as diesel trucks only in 2031-2032, held back by slow fleet turnover and by how long operators take to trust the charging network, the most influential assumption tested in this study. Taxing diesel barely works at politically feasible levels. Charging trucks the full fuel excise lifts the zero-emission share by one percentage point; adding ten points would require a road user charge of about 5 AUD per litre, fifteen times the 2025-26 rate. A 70 AUD per tonne carbon price and a purchase subsidy covering 60% of the price premium do little better. Per percentage change in cost, operating-cost support is five to six times as effective as purchase support, yet the best single instrument adds only two percentage points. Only a combined push-pull package, which still raises fuel-tax revenue, shifts the market materially. Policy should stop relying on diesel prices and instead build operators' confidence in the network, cut operating costs and recycle road-charge revenue into infrastructure.

0.23

PCM: A Proof-Carrying Message Protocol for Verifiable Multi-Agent Communication

cs.CE cs.AIcs.CRcs.MA ·2026-10-29 ·Bravish Ghosh
benchmarktheorycode/data
Abstract · 50 pages, 9 figures, 15 tables. Specification, conformance vectors, verifier and benchmark: https://github.com/Cygnux-Labs/PCM-protocol

When one AI agent tells another that "the tests passed" or "a person approved this refund", the receiver has only the sender's word. A2A and MCP authenticate the channel, but the sender is the party in doubt, so a prompt-injected agent can launder fabricated facts and authority through a multi-agent system. We present Proof-Carrying Messages (PCM), a protocol in which each claim carries evidence: records from an append-only Merkle log that the agent cannot write, with inclusion proofs, signed by an attesting signer the agent cannot control. Receivers check messages offline and give each claim a three-valued result: proven, unproven or failed. A claim that rests only on the agent's own report is never proven. Authority travels as grant chains rooted in a recorded human approval, and the chains can only narrow. We specify four claim types, a typed caveat algebra, replay rules that survive restarts, and strict readings that make independent verifiers agree, and we prove the resulting security properties. On a pre-specified benchmark of 926 messages from real runs, 712 of them lies in 16 classes, PCM accepts 10 lies, all of them stale evidence (1.4%, 95% CI 0.8-2.6%). Plain A2A accepts all 712, and models of six deployed protocols (A2A with signed Agent Cards, uAgents, RFC 9421, SCITT, AP2, Verifiable Intent) accept 504-712. Two independent verifiers agree on all 926 verdicts, and verification takes about 1 ms.

0.23

Spreading Dynamics on Higher-Order Networks: A Structure-Mechanism-Behavior Perspective with Applications

cs.SI physics.app-ph ·2026-10-29 ·Ming Xie, Shibo He, Qihao Huang, Yuzhen Fan, Jiming Chen
Abstract

Spreading in real-world networks, including modern communication and networked systems, often unfolds through group interactions rather than isolated pairwise contacts. Higher-order network representations such as simplicial complexes, hypergraphs, and their temporal, multilayer, and directed extensions provide a principled way to encode group interactions. Unfortunately, existing studies on spreading over these substrates remain fragmented and largely model-centric, offering limited insight into how higher-order representations and contagion mechanisms jointly shape emergent dynamics. This review addresses this gap by introducing a structure-mechanism-behavior three-level perspective that provides a systematic mapping from higher-order network structures and spreading mechanisms to collective behaviors. Within this framework, we compare how different higher-order network representations encode group interactions, introduce a taxonomy of contagion mechanisms, and analyze their structural carrying capacity, defined as the ability of a given representation to support specific contagion mechanisms and the associated dynamical behaviors. We further map specific structure-mechanism combinations to three hallmark emergent behaviors, thereby offering a unified cross-level perspective beyond existing model-centric surveys. These insights are particularly relevant for communication systems involving group interactions, such as social information spreading, cooperative protocols, and distributed networked processes. Lastly, leveraging this three-level perspective, we offer a roadmap for guiding and controlling dynamics on higher-order networks, while discussing future directions in representation choice, mechanism design, and data-driven inference.

0.22

GPU-Resident Speculative Decoding for In-Browser LLM Inference with WebGPU

cs.CY ·2026-10-29 ·Wei Song, Yuxin Cao, Ziqi Ding, Xiao Cheng
Abstract · 11 pages, 7 figures

WebGPU enables private, offline LLM inference inside web pages, but browser dispatch, inter-process communication, synchronous readback and single-token kernels slow decoding and make conventional speculative decoding inefficient. We present a WebGPU runtime that addresses these browser-specific costs. Deterministic small-batch integer kernels accelerate multi-token verification while preserving exact agreement with greedy decoding in the same quantized runtime. A GPU-resident loop performs drafting, verification, acceptance and KV-cache maintenance without a per-token CPU decision. A draft-free context-mixing drafter reuses page content, conversation history and session memory, while a device-aware controller adapts the speculation width and the submission window to measured costs. Across seven models from 0.5B to 8B parameters, 11 tasks and three client GPUs, our runtime decodes 3.5 to 5.4$\times$ faster than WebLLM on a laptop GPU, 4.8 to 6.2$\times$ faster on an integrated GPU and 4.8 to 8.6$\times$ faster on a workstation GPU. The GPU-resident loop and speculation contribute a further 1.8 to 3.2$\times$ over the optimized kernels on the workstation. Outputs match the quantized greedy decoding of the runtime exactly, and the mean GSM8K accuracy matches that of WebLLM. Optional draft heads raise the peak speedup to 10.2$\times$, while session memory provides additional gains without downloading extra weights.

0.22

"It Feels Weird to Just Rely on AI": How Experienced Healthcare Professionals Work With Decision Support Systems in Clinical Practice

cs.HC ·2026-10-29 ·Simon W. S. Fischer, Hanna Schraffenberger, Miranda L. van Hooff, Casper Reijnen, Serge Thill +1 more
Abstract

Decision support systems (DSS) assist with clinical decision-making. Little is known, however, about the decision-making process with DSS in clinical practice and the extent to which clinicians consider or rely on the predictions. Studies on human-AI decision-making usually focus on laypeople, simplified decision tasks, or one-shot interactions with prototypes that are not integrated into existing workflows. To gain insights from real-world practice, we conducted an online survey (n=12) and semi-structured interviews (n=10) with clinicians from two different medical cohorts each working with a DSS. Findings show, amongst other things, that the interaction with DSS changes as experience increases, the predictions become more important in complex cases, that decisions are discussed with colleagues, and that workflow integration affects decision-making. Our findings on human, technical, contextual, and organisational factors contribute to a socio-technical perspective on DSS reliance behaviour. We conclude by discussing the implications of our findings for the design of DSS.

0.21

Unweaving the mechanics of woven fabrics through automated model discovery

cs.CE ·2026-10-29 ·Marin Lauber, Uzair Manack, Thibault Vervenne, Rogier P. Krijnen, Benjamin Alheit +2 more
Abstract · 11 pages, 6 figures

Woven fabrics have multiple structural, aesthetic, and medical applications that require knowledge of their mechanical properties. While isolated fiber properties are easily determined, bulk properties of woven material, which can be coated or locally reinforced, require knowledge of yarn density, weave style, and other characteristics that are difficult to obtain experimentally. Additionally, fiber rearrangement, twist, yarn contact, and fiber-direction coupling result in non-linear mechanical responses. As a result, there is no universal material model for woven materials, and each fabric requires a specific model and parameter calibration. Recently, methods to simultaneously discover material models and their parameters from experimental or synthetic datasets have been proposed. Here, we use automated model discovery to simultaneously extract the material model and its parameters from biaxial test data of a TPU-coated woven nylon fabric. We show that Constitutive Artificial Neural Networks discover a sparse three-term hyperelastic material law with one isotropic and two anisotropic contributions, reproducing the measured biaxial response with a mean $R^2=0.97$. Across 20 initializations, predictive accuracy remains robust despite variations in the exact sparse representation, consistent with strong correlations between several candidate terms. LARS-LASSO independently recovers the same dominant constitutive contributions at approximately 100-fold lower computational cost. These results are directly applicable to numerical finite-element models of fabric-based structures, including soft robotic devices. More broadly, these results provide a foundation for further model discovery, improved constitutive modeling, and mechanistically informed optimization of woven fabric structures.

0.21

Incorporating Machine-Others: An Enactive Exploration of Human-LLM Interactions

cs.HC cs.AIcs.CY ·2026-10-29 ·Elena C. Cuffari, Marta Pérez-Verdugo, Xabier E. Barandiaran
position
Abstract

New concepts based in interactional thinking are required to understand how humans are changed by relating to LLMs. Standard theories of technological impact upon users' autonomy are inadequate here, because this technology is capable of inducing linguistic interactions with humans in an unprecedented manner. Traditional theories of social cognition also fall short, because LLMs have neither autonomy nor minds nor sociality as such. And yet, humans interact, talk, and create with them. We base ourselves on enactive theory, particularly on the theory of Linguistic Bodies and recent developments of enactive accounts of AI, to explore how we make sense with machines. We argue that we incorporate machine-others in two ways: in the tool-like sense of incorporation as enabling particular skills, but also in the more linguistic and identity-tied sense of incorporating (and incarnating) LLMs' ways of languaging. However, LLMs are themselves excorporated and excarnated phantoms of our own languaging; we give a (machinic) body to our language. As such, we are not just incorporating machinic others; we are incorporating excorporated others. These conceptual advances draw a new map of our linguistic relation to LLMs, putting them in continuation with other forms of languaging involving other linguistic artifacts (e.g., religious books). We attend to some particular forms that this relation takes today, and how they may be impacting the autonomy and becoming of LLM-engaged humans. We introduce the idea of "susceptive sense-making" to characterize increased semiotic labour that is carried by the lack of friction that we sometimes encounter within these services.

0.21

Measuring Personal Information Exposure in Mental Health-Related Conversations with LLMs

cs.HC ·2026-10-29 ·Synthia Qia Wang, Ranya Sharma, Christine Lee, Sarah Keedy, Marshini Chetty +1 more
Abstract

Large language models (LLMs) are increasingly used for informal mental health support, yet the privacy implications of conversational interaction remain poorly understood. Prior LLM privacy research has focused largely on model-side risks such as memorization and adversarial extraction, leaving personal information inferred from ordinary conversation understudied. We introduce a profile-recovery framework that measures personal information exposure by comparing user profiles with attributes inferred from conversation alone. We generate 240 synthetic multi-turn mental health conversations from profiles sampled from a nationally representative health survey and validate them against real counseling transcripts, finding comparable overall disclosure volume. Across 21 attributes, 33-41% of each profile is recoverable, with 76.5% first-inference and 94.3% any-of-3 accuracy. 239 of 240 conversations expose at least one sensitive mental health attribute, and 45.0\% expose all six. Our findings reveal a gap between disclosure and inferability, motivating HCI approaches to make inference risks legible and support privacy-aware interaction design.

0.20

Towards fNIRS foundation models: A curated corpus of 67 public datasets with channel positions in MNI space

cs.HC ·2026-10-29 ·Santiago Posso-Murillo
datasetcode/data
Abstract

Brain foundation models have transformed medical artificial intelligence and computational neuroscience by reducing the reliance on task-specific data through large-scale pretraining. While they have been successful with non-invasive neuroimaging data, their adoption in functional near-infrared spectroscopy (fNIRS) has lagged due to the limited size and heterogeneity of public datasets, which are collected under varying protocols, devices, and sensor configurations. In addition, no single fNIRS dataset is comparable in size to the largest datasets in functional magnetic resonance imaging (fMRI), electroencephalography (EEG), or magnetoencephalography (MEG). The fNIRS datasets are also spread across different platforms and often have insufficient information, or lack the protocols that would make them easier to use. Given the advances in EEG foundation models that enable flexible modeling across diverse EEG montages, we curated 1,955 hours of fNIRS data from 67 public datasets spanning 2,798 participants, representing the largest fNIRS curation effort to date. The curated data include oxygenated (HbO) and deoxygenated (HbR) hemoglobin signals along with the three-dimensional position of every channel in the standard space of the Montreal Neurological Institute (MNI), which makes positions comparable across datasets. We release dataloaders that handle each dataset automatically, along with the minimal preprocessing pipeline used to prepare the data, at https://github.com/sposso/fnirs-datasets. This work aims to facilitate the future development of foundation models for fNIRS.

0.19

Structural Exposure and Subsequent Neighborhood Stability in Temporal Networks: Evidence and Limits of Generalizability

cs.SI ·2026-10-29 ·Mehmet Simsek
Abstract

Can temporal path based structural exposure serve as a reliable indicator of subsequent local neighborhood stability across different temporal networks? We investigate this question using the completed observational results from 12 networks spanning face to face nteraction, communication, online activity, and urban mobility. Exposure aggregates source node structural mass along time respecting paths; the outcome is Jaccard neighborhood turnover between later, nonoverlapping windows, with greater turnover indicating lower stability. Six complementary analyses examine the direction and consistency of exposure-turnover associations, proximity to high mass nodes, functional response shapes, distance decay kernels, alternative mass definitions, and temporally ordered changes in source mass. The evidence is heterogeneous rather than uniformly negative or positive. In the primary association analysis, two networks retained positive coefficients after within question false discovery rate adjustment, indicating lower subsequent stability at higher exposure, whereas no negative association survived that adjustment. The pooled proximity coefficient was not distinguishable from zero, and significant mass change associations appeared in opposite directions across networks. More flexible functional, kernel, and mass choices yielded only modest cross validated improvements over simpler comparators; available temporal randomizations also gave model dependent contrasts. The study therefore does not establish structural exposure as a directionally consistent, network independent stability indicator. It also cannot establish a lack of predictive value relative to a no exposure baseline, because that specific comparison and independent cross network validation were unavailable. Incomplete null simulations, unrecorded covariance estimation fallbacks constrain inference.

0.19

TRACER, a time-to-event EHR foundation model for disease detection across health systems

cs.CE cs.LG ·2026-10-29 ·Mara Pleasure, Ekaterina Redekop, Vedrana Ivezic, Rushikesh Kulkarni, Aaron Chin +2 more
Abstract

Foundation models trained on longitudinal electronic health records (EHRs) can predict clinical diagnoses years in the future, making them a potential tool for disease screening. We developed TRACER (Temporal Risk Assessment with Clinical EHR Representations), a decoder-based foundation model pretrained on data from over 4.5 million patients from four health systems, and evaluated it on 16 incident diseases at 1-, 2-, and 5-year horizons in more than 2 million patients from two health systems excluded from model development. In contrast to typical evaluations that focus on global metrics (e.g., AUROC), we measured performance at the low false-positive rates that are realistic for screening. We found that TRACER retained its discrimination in external health systems, and at 95% specificity, the patients it flagged were up to eight times more likely to be diagnosed within a year than unselected patients. Evaluating at this operating point also revealed what global metrics alone did not: detection and false-alert burden varied across diseases and health systems, even when discrimination appeared stable. Our results illustrate the potential of foundation models for screening and underscore the need for evaluation in health systems excluded from development to assess disease detection and false-alert burden at levels appropriate for intended clinical use.}

0.16

Engineering process plants as code in the age of agentic AI

cs.CE ·2026-10-29 ·Artur M. Schweidtmann, Davy Demeyer
position
Abstract · 7 pages, 2 figures. Submitted to Nature Chemical Engineering as a Comment

Agentic artificial intelligence (AI) could one day shorten process plant engineering from years to weeks. Today, however, plant information is scattered across documents and tools and must be carried between them by hand, so an AI agent placed into current workflows inherits the same silos. In this Comment, we argue for Plant-as-Code: defining the engineering of a plant as a version-controlled collection of machine-readable files that follow, where possible, existing standards such as DEXPI and IFC. On this foundation, Design-Ops extends DevOps principles to plant engineering, linking every engineering item to the information it derives from and replacing manual checking with automated tests. Agents can then coordinate routine work across disciplines while deterministic tools perform the calculations. Verification sets the pace: agents propose, tools check, and engineers review and authorize every change. We outline the risks of this shift and the research, industry and policy actions it requires.

0.15

Counting approximate breakpoint medians of random genomes

cs.CE cs.DSq-bio.PE ·2026-10-29 ·Poly H. da Silva, Arash Jamshidpey, David Sankoff
theory
Abstract · 32 pages, 1 figure

Unsigned linear unichromosomal genomes with genes labeled 1, . . . , n are represented by permutations of these labels, whose adjacencies are the unordered pairs of genes occurring consecutively. The breakpoint distance counts the adjacencies of one genome that are absent from the other, and a breakpoint median minimizes the total breakpoint distance to a given set of genomes. Deciding whether a permutation is a median often requires a minimization over the whole symmetric group, which makes the median set resistant to exact counting. We replace this global condition by a local one: a c-approximate median is a genome having at most c adjacencies absent from all the input genomes. For a fixed genome we introduce its approximate median inverse, the set of ordered k-tuples of genomes ad- mitting it as a c-approximate median, and we construct and count this set explicitly. For k genomes chosen uniformly and independently at random, this gives the probability that a given genome is a c-approximate median and the expected number of such approximate medians. The relaxation is not only technical: approximate medians and near-optimal genomes co- incide up to a shift of the tolerance that remains bounded in probability as the number of genes grows. Finally, we derive an explicit upper bound on the probability that a given genome is an exact median, and on the expected number of exact breakpoint medians.

0.13

Generative artificial intelligence prompting strategies among Chinese university students across academic tasks

cs.HC cs.AI ·2026-10-29 ·Lei Fan, Fangxue Liu
Abstract

Generative artificial intelligence (GenAI) is increasingly embedded in university learning, yet effective human-GenAI interaction depends not only on access to these tools but also on students' ability to formulate, evaluate, and refine prompts for academic tasks. This study examined university students' perceived GenAI prompting capability and reported prompting strategies. Questionnaire data from 234 students were analyzed using descriptive statistics, measurement modelling, scenario comparisons, structural associations, and qualitative coding of optional comments. Students reported using GenAI across diverse academic activities and generally expressed confidence in their prompting capabilities, particularly in providing relevant context and refining prompts following initial outputs. However, prompting strategies showed limited sensitivity to task types: specifying output format was more strongly endorsed in a creative workshop preparation scenario than in a research-oriented assignment, whereas other matched strategies showed no adjusted differences. Students also commonly endorsed both concise instructions and providing all relevant information at the outset, suggesting that effective brevity may depend more on selecting and organizing task-relevant information than on prompt length alone. By contrast, fixed prompt templates received lower endorsement than other prompting strategies. In the adopted factor model, perceived cognition showed the clearest unique association with perceived prompting capability, while open-ended responses highlighted prompt-intent alignment as a common challenge.

0.10