CS Paper Radar

Daily listing

2026-11-06

725 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 255

e-means: outperforming k-means (in one-tenth the epochs)

cs.LG cs.AI ·2026-11-06 ·Swann Bessa, Pierre-Emmanuel Mazaré, Gergely Szilvasy, Hugo Touvron, David Picard +2 more
code/datacommunity pick
Abstract · Code : https://github.com/facebookresearch/faiss/tree/main/contrib/e_means Website : https://e-means.github.io

Lloyd's k-means algorithm is a long-running standard for clustering, with applications in data analysis, retrieval, quantization and data curation. In this paper, we introduce an exponential moving average optimization procedure for k-means with mini-batch updates, powered by modern deep learning machinery and new insights. We also establish scaling laws on how decay rate and batch size impact final performance. This gives us a formulaic rule to fix them for multiple problem sizes and embedding datasets, without the need for any hyperparameter sweep. The resulting clustering algorithm, coined e-means, is much faster than Lloyd k-means and converges to lower losses, frequently even beating k-means++. On a wide array of embeddings datasets, it beats the loss of Lloyd k-means at 25 epochs in 1.4 to 2.6 epochs, which translates to 10-45x GPU wall-clock speedups compared to state-of-the-art k-means implementations. As an example, e-means partitions DINO10B, a dataset of 10 billion DINO embeddings, into 16 million clusters in only 2.5 epochs, reaching a loss below Lloyd's at 25 epochs.

0.70

Mitigating Covariate Shift in Behavioral Cloning via Robust Stationary Distribution Correction

cs.LG cs.AI ·2026-11-06 ·Seokin Seo, Byung-Jun Lee, Jongmin Lee, HyeongJoo Hwang, Hongseok Yang +1 more
top venuecode/datahighly cited
Abstract · 25 pages, 11 figures, NeurIPS 2024, Code: https://github.com/tzs930/drildice

We consider offline imitation learning (IL), which aims to train an agent to imitate from the dataset of expert demonstrations without online interaction with the environment. Behavioral Cloning (BC) has been a simple yet effective approach to offline IL, but it is also well-known to be vulnerable to the covariate shift resulting from the mismatch between the state distributions induced by the learned policy and the expert policy. Moreover, as often occurs in practice, when expert datasets are collected from an arbitrary state distribution instead of a stationary one, these shifts become more pronounced, potentially leading to substantial failures in existing IL methods. Specifically, we focus on covariate shift resulting from arbitrary state data distributions, such as biased data collection or incomplete trajectories, rather than shifts induced by changes in dynamics or noisy expert actions. In this paper, to mitigate the effect of the covariate shifts in BC, we propose DrilDICE, which utilizes a distributionally robust BC objective by employing a stationary distribution correction ratio estimation (DICE) to derive a feasible solution. We evaluate the effectiveness of our method through an extensive set of experiments covering diverse covariate shift scenarios. The results demonstrate the efficacy of the proposed approach in improving the robustness against the shifts, outperforming existing offline IL methods in such scenarios.

0.63

RidgeFill: Training-Free Late-Layer Cache Fill for Faster Prefill

cs.LG cs.CL ·2026-11-06 ·Dylan Johnston, Robert P. Smith
publishedcode/dataunusual
Abstract · 50 pages, 15 figures, 38 tables. Code and data: https://github.com/Approximate-Science/ridgefill-data

RidgeFill shortens prefill by not computing most of a model's upper layers for the prompt. The lower layers run exactly up to a split; from the split layer's state a per-layer map predicts every later layer's input, and each late layer writes its cache from that prediction. The design follows YOCO and DeepSeek-V4.1-Flash, where the projection is trained into the model. RidgeFill retrofits it onto models that were never trained for it: the map is a ridge regression fitted once in closed form from one pass over calibration text, the layer's own projections run unchanged, and a chosen set of rows, the last few thousand tokens, and every decoding step run the full model. Nothing is trained and no token is dropped. Most of the work was in measurement, finding a repeatable methodology. Loss averaged over all tokens hides RidgeFill's error. Needle retrieval in the filled region, loss on tokens that depend on distant context, and splice tests that swap filled keys or values for exact ones show which component carries it: the keys on Llama-3.1-8B, half of it in their scale, and the values and the recurrent state on Qwen3.5-9B. On Qwen3.8-27B in a llama.cpp fork on an A100, time to first token is 1.25 to 1.31x faster at 16K to 120K tokens with the maps in VRAM; RULER changes by -1.2 and -1.4 points and LongBench v2 by -1.4; the 95 % intervals include zero but do not rule out losses of a few points. On Qwen3.8-Flash-Next in vLLM it is 1.50 to 1.60x faster than exact at 9K to 32K tokens, with RULER's needle tasks at 99.8 against 100.0 and perplexity on novel tokens 2.65 % higher [-0.02 %, +4.93 %]. On three 8 to 9B reference models, at the same split and tail as the identity fill, the fitted map retrieves 49.1 and 49.6 more filled-region RULER points on the two dense models (at that split Llama's loss is worse), and its long-document loss is +/-0.03 nats of the released SwiftKV checkpoint's.

0.60

LeanDojo-v2: A Comprehensive Library for AI-Assisted Theorem Proving in Lean

cs.LG ·2026-11-06 ·Ryan Hsiang, William Adkisson, Robert Joseph George, Anima Anandkumar
publishedtheoryhighly cited
Abstract · 11 pages, 7 figures, accepted to NeurIPS Math-AI workshop

Building on recent progress in machine learning for formal theorem proving in Lean, we introduce LeanDojo-v2: an end-to-end framework that unifies repository data extraction, LLM fine-tuning, proof search, AI autocompletion, and LLM-assisted theorem proving in a single software library. In addition, we provide Lean4Code, a lightweight IDE that integrates the various features of LeanDojo-v2 through a simple, mathematician-friendly UI. These tools will enable mathematicians and researchers to easily fine-tune language models on domain-specific Lean codebases and experiment with LLM-assisted theorem proving without a complex setup.

0.54

Designing Effective Monitor-Based Interventions for Mitigating Reward Hacking During RL

cs.LG ·2026-11-06 ·Aria Wong, Joshua Engels, Neel Nanda
top venuecode/data
Abstract · Accepted at NeurIPS 2026

Reinforcement learning (RL) rewards are notoriously difficult to design and control, often leading to the model learning unintended behaviors such as reward hacking. One potential solution is to monitor for reward hacking and penalize it when detected; however, training against a monitor could lead to evasive behavior, and our general understanding of how to apply monitors effectively during training is limited. To study how best to use monitors to mitigate reward hacking, we introduce and open source three realistic environments where Qwen3-4B reward hacks: a coding environment hackable via test overwriting, a medical chat environment hackable via sycophancy, and a biography generation environment hackable via hallucination. We first focus on the coding environment, where we find that: (1) models can learn to evade highly accurate monitors by exploiting systemic flaws in probes and LLM judges; (2) monitors that leak more learning signal during RL suppress reward hacking but are more often evaded; and (3) including easier problems in training can decrease reward hacking. We apply our findings to build better reward hacking monitors for the medical chat and biography generation environments that improve upon naive baselines to reduce reward hacking rates across seeds to before-training levels in some interventions. Our results demonstrate that our takeaways translate to new settings and that better monitor intervention designs are possible.

0.53

Know What You Need: Efficient Activation Checkpointing

cs.LG ·2026-11-06 ·Danyal Ahmed, Edouard Oyallon
top venue
Abstract

Large Language Model training is increasingly constrained by memory capacity. Activation Checkpointing reduces activation memory by avoiding storing selected activations during the forward pass and rematerializing them during backpropagation, thereby trading memory savings for additional computation. Although this technique is essential for scaling model training, rematerialization can be prohibitively expensive, particularly for Transformer linear layers. Approximate computation could reduce this overhead, but conventional approaches either introduce errors that accumulate over the backward pass or fail to exploit information available from the initial forward pass. We introduce a novel approximate rematerialization method that can yield unbiased gradient estimates: using information collected during the initial forward pass, the method selectively recomputes only the most important layer components, either through greedy selection or randomized sketching. Ablation studies show that the resulting gradients closely match the exact gradients. When applied to feed-forward layers, attention layers, or both, our method can recompute only two-thirds of the layer components in a Llama-488M model, while increasing perplexity by just 1.7%, 1.2%, and 2.5%, respectively.

0.52

BiLoGo: Reducing the smoothing of localised responses in Transformer-based neural operators for sheet forming

cs.LG ·2026-11-06 ·Yingxue Zhao, Haoran Li, Haosu Zhou, Tobias Pfaff, Nan Li
Abstract

Trial-and-error FE simulation of sheet forming could be computationally expensive, neural-network-based learned simulators are used to accelerate design. These simulators should accurately capture localised responses, such as thinning around sharp radii or near ribs, which determine formability. Transformer-based neural operators propagate information efficiently across the mesh but tend to smooth such localised responses. This study develops BiLoGo, a Bipartite Local-Global learned simulator with an FE-guided design that connects local node-element message passing with a Transformer-based neural operator for long-range information exchange. BiLoGo has three modules: a bidirectional node-element GraphSAGE model that learns nodal-elemental coupling on a bipartite graph and predicts each quantity on its native FE entity without interpolation; a dual-stream Transformer-based neural operator that processes blank deformation and tool contact with separate token sets, preventing localised contact information from being averaged with the whole-blank state; and a fusion and element update module that combines global and local features via gated residual fusion and updates element features conditioned on element-node kinematics. Evaluated on a dome-shaped cold forming case and two corner-shaped hot stamping cases (one with ribs), BiLoGo achieves the lowest positional and threshold-based thinning errors on all three cases compared with single-level and hierarchical graph-based models and Transformer-based neural operators. Near the ribs, the smoothing index of predicted thinning increases from 0.41-0.56 for Transformer-based neural operators to 0.94 for BiLoGo-U. Compared with the hierarchical graph network CAtt-BiUGNN, BiLoGo-T reduces mean positional error by 15.5% on the Corner case and BiLoGo-U by 8.3% on the Corner+Rib case, with 66-69% less training memory.

0.52

Exploring Weight-Space Programming: Capabilities, Generalization, and Limitations

cs.AI cs.LG ·2026-11-06 ·Junru Hu, Cheng Tang, Wei Li, Hongze Zhu, Chun-Mei Feng +3 more
code/data
Abstract

Weight-Space Programming (WSP) enables reasoning agents to directly inspect and modify neural network parameters, offering an alternative to gradient-based model adaptation. However, its capabilities, cross-task generalization, and practical limitations remain insufficiently understood. We systematically investigate WSP through 16 editing trajectories across seven benchmarks and two model scales (Qwen3.5-0.8B and Qwen3.5-9B). Our findings reveal three key insights. First, WSP can achieve substantial performance gains through highly localized interventions. On full held-out evaluations, accuracy improves by up to 29.45, 4.42, 36.72, and 17.11 percentage points on MATH-500, GPQA Diamond, IFEval, and SciBench, respectively. Notably, modifying only 1,023 parameters yields a 29.45-point improvement. Second, WSP exhibits promising but uneven cross-task generalization. A Math-edited checkpoint improves GPQA Diamond, IFEval, and SciBench by 3.03, 12.18, and 20.10 points, respectively, without further editing, while leaving SciCode performance unchanged. Third, effective editing remains challenging and inconsistent. Outcomes vary substantially across trajectories, and six of the 16 trajectories retain the original weights when candidate improvements cannot be reliably validated. Together, these results highlight WSP's potential for highly localized model adaptation while revealing important challenges in intervention discovery, reliability, and cross-task transfer. Homepage: https://wsp-scope.github.io/ GitHub: https://github.com/WSP-Scope/WSP-Scope-Project

0.52

Practical Non-Stationary Graph Gaussian Processes

cs.LG ·2026-11-06 ·K. S. Sesh Kumar, Viacheslav Borovitskiy
top venuecode/data
Abstract · Accepted at NeurIPS 2026. Code: https://github.com/seshkumar/vcgp-graph-gp

Gaussian processes on graphs provide a principled framework for prediction and uncertainty estimation, but modelling heterogeneous networks often requires balancing flexibility against parameter complexity. Variability at a node may reflect useful signal that should propagate through the network, unreliable measurements whose influence should be suppressed, or a mixture of both. We introduce Variance-Coupled Gaussian Processes (VCGPs), which use a positive auxiliary signal to modulate graph interactions through a single learned coupling exponent. The exponent controls how strongly high-variability nodes propagate or attenuate information, while a cluster-wise extension accommodates different regimes within the same graph. Under stated structural assumptions, we characterize the unique stationary-rescaling point of the basic family, where correlations match those of a stationary graph kernel while node-dependent variances remain. Experiments on traffic, air-quality and weather datasets show competitive prediction and uncertainty estimation relative to stationary and per-node non-stationary baselines, with the largest gains on networks requiring multiple coupling regimes. Stationary correlations after variance rescaling remain sufficient on some data, and performance depends on the auxiliary signal and model selection.

0.51

PhysioWave-T: Token-Efficient Wavelet Foundation Models for Biosignals

cs.AI ·2026-11-06 ·Yanlong Chen, Simone Benatti, Luca Benini, Cuntai Guan, Yawei Li
code/datapopular code
Abstract · 15 pages, 11 figures, 5 tables. Code: https://github.com/ForeverBlue816/PhysioWave , Models & Weights: https://huggingface.co/collections/Fo…

We present PhysioWave-T, a token-efficient wavelet foundation model family for EEG, ECG, and sEMG. Dynamic scale folding recombines J+1 wavelet subbands prior to tokenization, compressing sequence length from (J+1)CS to CS tokens for J levels, C channels, and S temporal patches. Route-specific front ends and learned channel-name embeddings allow one Transformer per modality to handle diverse montages and sampling rates, while unseen setups reuse the backbone via lightweight adapters. We pre-train a 30M-parameter EEG model on over 15M windows (19--128 channels), alongside ECG (30M) and sEMG (12M) encoders. Across four EEG routes, scale folding cuts Transformer latency by 4.0--7.2x and peak memory by 6.8--12.7x over scale expansion without accuracy loss. Our EEG encoder leads 8 of 9 metrics against BENDR, BIOT, LaBraM, and EEGPT, and surpasses LUNA-L on wearable seizure detection with fewer parameters. Reliability-aware latent fusion (RALF) further integrates summary tokens from independently pre-trained encoders via sample-adaptive weights. On SeizeIT2, RALF raises event sensitivity at 11 false alarms/h from 61.4% to 74.4% and window AUPRC from 0.321 to 0.399 over the best unimodal baseline, outperforming weighted-logit and concatenation fusion across all variants. PhysioWave-T also achieves state-of-the-art sEMG accuracies and top macro-F1 on PTB-XL Super and Chapman-Shaoxing. Code and models: https://github.com/ForeverBlue816/PhysioWave.

0.51

Large Language Model Deployment on Resource-Constrained Edge Devices: A Practitioners Survey

cs.LG cs.AIcs.CL ·2026-11-06 ·Rahul Maliakkal, Yashasvi Makin, Plawan Rath, Rohit Jain, Ankur Sadhoo
highly cited
Abstract

Large Language Models (LLMs) have transformed many areas of artificial intelligence (AI), but their substantial resource requirements limit their deployment on mobile and edge devices. Although LLMs have demonstrated impressive capabilities in various applications, the next challenge is deploying these models on resource-limited edge devices. Recent innovations in software optimizations and specialized hardware have made it feasible to deploy LLMs on edge devices. Edge deployment offers benefits including local data processing, cost savings, reduced latency, and personalized responses. This paper reviews the latest techniques that make the deployment of LLMs on edge devices feasible.

0.51

Enforcing Symmetries on Pre-trained ML Models for Physics

cs.LG physics.comp-phstat.ML ·2026-11-06 ·Wenjie Zhou, Valentino F. Foit, David W. Hogg, Soledad Villar
Abstract · 32 pages, 5 figures, 9 tables

Post-hoc group averaging (or test-time augmentation) is a simple way to impose symmetry on a trained predictor without changing its architecture or retraining it. For finite and compact groups, averaging over a normalized Haar measure on the group gives the Reynolds operator: a canonical projection onto invariant or equivariant predictors with known non-increasing risk guarantees under certain assumptions. This paper provides numerical experiments on physics emulation models that demonstrate the effectiveness of the approach in this setting. We then extend this framework to non-compact group symmetries. For non-compact groups, the Haar measure cannot be normalized, so group averaging cannot be defined in the same way. We replace it with three alternatives: the Weyl unitary trick (which requires complexifying the space and extending the predictor), a data-dependent orthogonal projection (which involves a non-trivial optimization on the space of invariant or equivariant functions), and a frame averaging approach (which involves mapping each point to a set of orbit representatives). We illustrate the constructions theoretically and empirically on the unit-rescaling group and the Lorentz group. Our results show that in the non-compact case, the Weyl trick and frame averaging do not provide a consistent improvement over the baseline, except where the test data is out-of-distribution (an orders-of-magnitude improvement in this case) and, for the Weyl trick, in low-data regimes. The orthogonal projection can provide improvements in-distribution and out-of-distribution. We also observe that models trained directly on invariant features matched or beat the projection of an unconstrained one. We conclude that post-hoc symmetrization extends to non-compact groups, but it does not consistently outperform training a symmetric model.

0.50

Composable Causality: A Toolkit for Systematic Time-Series Causal Discovery and Treatment-Effect Benchmarking

cs.LG ·2026-11-06 ·Muhammad Hasan Ferdous, Md Osman Gani
top venuebenchmarkcode/data
Abstract · 30 pages, 10 main text pages plus references and appendices, with figures and tables. Accepted to the NeurIPS 2026 Evaluations and Datasets …

Real-world multivariate time series rarely come with verified ground-truth causal structure, so research on time-series causal discovery and treatment-effect estimation depends on synthetic data. Existing tools are either fixed datasets that cannot be modified after release or script-based generators in which each script fixes one combination of mechanism, confounding, and missingness. Neither supports controlled experiments that vary a single property, such as sparsity, noise distribution, or missingness mechanism, while holding the others fixed. Counterfactual evaluation has the same gap, since existing simulators are built per paper with hardcoded mechanisms or produce observational data only. We introduce DTCD (Datasets and Toolkit for Causal Discovery), a data generation library that produces multivariate time series with ground-truth contemporaneous and lagged adjacency matrices. Topology, mechanism, noise, observational mask, and intervention policy are independent components that compose through configuration alone. Paired counterfactual twins, produced by caching the stochastic realization during the factual rollout and re-injecting it during the interventional rollout, recover individual treatment effects to numerical precision. A reference benchmark of eight discovery algorithms on eighty-one configurations and four treatment-effect estimators on eighteen configurations exposes failure modes that aggregate metrics hide. These include the inability of standard discovery methods to handle missing values, regardless of whether the missingness mechanism is informative, and a marginal-variance shortcut (varsortability) that we measure and remove by standardization. Code and data: https://github.com/hferdous/Composable_Causality

0.50

A Separation Principle for Cooperative Multi-Agent Reinforcement Learning

cs.MA ·2026-11-06 ·Lucia Pezzetti, Nicolas Lanzetti, Antonio Terpin, Florian Dörfler, Giorgia Ramponi
theory
Abstract

We study cooperative multi-agent reinforcement learning (MARL) problems in which agents evolve under decoupled noisy dynamics but share a population objective. A canonical example is a mobility operator that orchestrates a fleet of vehicles to meet customer demand. The stochastic and time-varying nature of these problems calls for learning approaches, yet existing MARL algorithms do not scale: learning assignment and routing jointly over the full state-action space becomes prohibitive as the number of agents grows. Formulating the problem over the space of probability measures, we prove a separation principle: the population cost-to-go function is upper bounded by an optimal transport problem whose transportation cost is the target-conditioned single-agent cost-to-go function. This simplifies the multi-agent problem to learning a single-agent policy and coordinating the agents via optimal transport. Building on this principle, we introduce SALT (Separation-based Assignment and Learning via optimal Transport), a family of algorithms that learn a target-conditioned single-agent policy using any standard reinforcement learning algorithm, and then coordinate the agents via optimal transport. Since learning occurs at the single-agent level, SALT scales to fleet sizes where MARL baselines become intractable and transfers without retraining to fleet sizes unseen during training. In a vehicle-delivery experiment on the southern Manhattan road network, where MARL baselines are computationally prohibitive, SALT improves travel time by 15-21% over a shortest-path routing heuristic. In stochastic grid worlds, SALT reduces average travel times by 23-61% relative to MARL baselines and scales to population sizes where these methods fail.

0.49

Solver-Free Decision-Focused Learning via Neural Optimization Layers for Edge Selection

cs.LG math.OC ·2026-11-06 ·Xinda Zheng, Frits de Nijs, Edward Lam
Abstract

Decision-focused learning trains prediction models to improve downstream decisions, but many methods require costly solver calls or optimal-solution labels. We propose a solver-free decision-focused learning framework for NP-hard edge-selection problems with uncertain parameters in the objective, the constraints, or both. A neural optimization layer is first trained without solution labels, using problem objectives and constraint penalties. Graph-informed edge scoring and iterative refinement encourage concentrated and structurally consistent continuous decisions. The layer is then frozen, and a prediction model is trained through its continuous output using a problem-specific loss evaluated under observed parameters. At deployment, the predictor, optimization layer, and a lightweight decoder produce decisions without solving the original NP-hard problem. Experiments on the asymmetric traveling salesperson and endpoint-free orienteering problems show consistently better decision quality than prediction-focused learning and improvements over most evaluated decision-focused baselines. Our framework requires less training time than the evaluated decision-focused baselines, even when optimization-layer pretraining is included. Neural deployment also produces decisions faster than solver-based deployment, offering a better quality-efficiency trade-off. The layer also reduces training cost in existing decision-focused methods when used as an approximate optimization oracle.

0.48

FAIR Universe Weak Lensing ML Uncertainty Challenge: Dataset and Cosmological Inference Competition

cs.AI astro-ph.COcs.CVcs.LG+1 ·2026-11-06 ·Biwei Dai, Po-Wen Chang, Erwan Allys, Wahid Bhimji, Boris Bolliet +35 more
publishedbenchmarkcode/datahot direction
Abstract · 10+7 pages, 2+4 figures, 2+1 tables. Key results of the NeurIPS 2025 FAIR Universe Weak Lensing ML Challenge Phase 1. The challenge dataset …

Weak gravitational lensing is a powerful cosmological probe that enables direct mapping of the matter distribution and precise constraints on the fundamental parameters governing structure formation in the universe. In recent years, machine learning has shown strong potential for extracting cosmological information from weak-lensing data beyond traditional summary statistics. However, progress has been limited by the lack of standardized benchmarks that capture the complexity of the dataset, simulator imperfections, and various systematic biases encountered in realistic analyses. To address this gap, we present the FAIR Universe Weak Lensing Machine Learning Uncertainty Challenge, one of the NeurIPS 2025 competitions, as a benchmark dataset and evaluation study. Our dataset reflects realistic weak-lensing observations, incorporating multiple systematic biases, and thereby providing a controlled testbed for robust cosmological inference under uncertainties. We summarize the main methodological advances developed by the winners of our challenge, including data augmentation with conditional optimal transport, an AI-agent-driven approach for parameter inference, ensemble learning, Vision Transformer-based architectures, and physics-informed feature integration. By unifying the benchmark dataset and the winning solutions, this work provides a community resource for developing and evaluating machine learning methods for weak lensing cosmology and, more broadly, for scientific inference under systematic uncertainties.

0.47

EarthRecord: From Pixels to Land Change with Multimodal Agents over Four Decades of Satellite Imagery

cs.AI cs.CV ·2026-11-06 ·Haoran Zhang, Huizhen Sun, Hanbin Wen, Jie Leng, Jiaxin Zhu +1 more
benchmarkcode/data
Abstract · 23 pages, 9 figures, 18 tables. Project page: https://ranranrannervous.github.io/EarthRecord/

While multimodal agents now carry out complex data analysis with tools, their ability to turn decades of satellite imagery into reliable land-change analysis, crucial for Earth observation science, remains unmeasured. Existing benchmarks fall short: they pose questions on curated images or give agents preprocessed inputs, leaving out the raw multi-sensor archives where such analysis begins. We introduce EarthRecord, a benchmark built on the raw record of 56 sites: four decades of Landsat, MODIS, VIIRS, and Sentinel-2 data, with cloud masking and cross-sensor harmonization left to the agent. It poses 238 land-change questions across ten analysis tasks, scored against withheld references from global products, agreement among methods, and visual interpretation. Two controls test what answers rest on: the closed book removes the record, and instrument pairs alter one satellite's data while nothing changes on the ground. Evaluations of twelve multimodal agents reveal that leading agents complete most analyses end to end but falter when an answer requires mapping from imagery. Further analysis shows that the record moves answers from prior knowledge to the pixels but not yet to the real change on the ground: when one satellite's data are altered, half of the right answers follow the alteration. EarthRecord thus provides a principled platform for measuring and advancing agents from pixels to land change.

0.47

BMMS: Policy-Conditioned Measurement of Behavioral Compliance in Language Models

cs.AI cs.CL ·2026-11-06 ·Anthony Rhodes, Gadi Singer
published
Abstract · 21 pages, 4 figures. NeurIPS 2026 Workshop on Trust-AI-Eval (TAE). Artifact: https://doi.org/10.5281/zenodo.22064759

Behavioral evaluation of language models fixes its dimensions in advance: hallucination, sycophancy, toxicity, over-refusal. An operator in a regulated industry needs a different question answered: does a system do what a particular rulebook requires? We introduce the Behavioral Misalignment Metric Suite (BMMS), which measures policy-conditioned misalignment by taking a policy corpus as an argument and returning a compliance profile over it, for any published policy and any system that answers in natural language. The Compliance Attainment Score (CAS) measures severity-weighted attainment over the obligations that bind the system, and the Compliance Robustness Margin (CRM) the share of that attainment surviving graded adversarial pressure; Behavioral Invariance (BIV) and the Behavioral Disposition Index (BDI) are specified and released with the suite. Dimensions enter only by passing a published admission test whose binding criterion is attribution: a quantity that moves when the judge changes, holding the target fixed, is measuring the instrument. Applying it excluded one of the quantities the suite itself reported. Across 1,003 obligations from eleven documents in five registers, the admission criterion holds 94.4% precision against blinded human judgment on vendor specifications and 50.0% on analytical guidance, extraction fidelity is 94.6%, and re-scoring under judges from two other providers leaves target ordering invariant while the absolute score of a single target spans up to 0.22. We apply it to the two most detailed published behavioral specifications. Own-specification advantage is not resolved from zero once document difficulty is subtracted, but absolute self-compliance is another matter: GPT-4o fails 70.7% of the hard constraints binding it in the Model Spec published by OpenAI, where the two Claude models fail 0.0% and 9.1% of theirs.

0.46

Beyond the Lower-Star Filtration: Alternative Filtration Methods for Persistent Homology

cs.LG stat.MLmath.ATstat.AP ·2026-11-06 ·Innocent Abaa, Umar Islambekov
published
Abstract · This is the published version of the article. Published in *International Journal of Topology*, MDPI, DOI: 10.3390/ijt3040025. Link to the o…

In the lower-star filtration (LSF), simplices are assigned filtration values according to the maximum-vertex rule, with each simplex entering the filtration when its highest-valued vertex appears. A limitation of this construction is that some higher-dimensional simplices may enter simultaneously with their boundary faces, causing certain topological features to be filled immediately upon formation and thus not appear in the resulting persistence diagram. To address this limitation, we propose three alternative filtration methods: the \(\varepsilon\)-lower-star filtration (ELSF), the \(L_p\) filtration (LpF), and the softmax filtration (SMF). {ELSF introduces a dimension-dependent delay to the maximum-vertex rule,} allowing higher-dimensional simplices to enter the filtration after their boundary faces, whereas {LpF and SMF use aggregation functions that depend on all vertex values rather} than solely on their maximum. By modifying how filtration values are assigned, these methods allow simplices to enter the filtration more progressively, so that topological features such as loops and voids may persist over longer filtration intervals rather than being filled immediately. Consequently, the resulting persistence diagrams may include persistence features that are not observed under LSF, potentially providing useful information for distinguishing data populations. Experimental results on simulated and real-world datasets demonstrate that the proposed filtration methods are computationally efficient and can provide greater sensitivity to structural differences in the data than LSF.

0.46

Can Time Series Foundation Models Serve as Surrogates for Power Systems Dynamic Simulation?

cs.LG ·2026-11-06 ·Sandy Miguel, Bruno Paes Leao, Cheng Feng
publishedcode/data
Abstract · Accepted for publication at NeurIPS 2026 Foundation Models for Temporal Systems Workshop

High-fidelity dynamic simulation is essential for scientific and engineering applications, but its computational cost limits large-scale scenario screening. We investigate the use of pretrained time-series foundation models (TSFMs) as computationally efficient surrogates to address this limitation. In-context learning and post-training strategies, including retrieval-augmented reference sample selection and physics-informed regularization, are considered in the investigation. Results are benchmarked against a traditional AI model representative of the state-of-the-art in the domain. Experiments using an open-source TSFM and data from a simulated 9-bus power system indicate promising results. Retrieval augmentation improves in-context learning performance, and fine-tuning leads to superior results compared to the traditional AI model, especially when including physics-informed regularization. The fine-tuned TSFMs achieve lower errors and improved computational costs compared to the traditional AI model, and a 563$\times$ speedup compared to the original physics-based simulation.

0.46

Generative Latent Priors for Interpreting and Steering Protein Language Models

cs.AI cs.LG ·2026-11-06 ·Daniel Hinjos, Lucía Urcelay
publishedcode/data
Abstract · Accepted at the Interpretability for Discovery (Interp4Discovery) Workshop at the 40th Conference on Neural Information Processing Systems (…

Dominant methods for interpreting large models impose strong structural assumptions. Generative Latent Priors (GLP) are a complementary approach: a diffusion model that learns the distribution of a frozen model's activations directly, imposing no assumptions. We port GLP to ESM-2 activations (UniRef50) and evaluate it at two host scales (ESM-2-8M and -650M), both as a lens for protein representations and as a prior for editing them. Generation quality scales smoothly with depth and compute. As a lens, single GLP "meta-neurons" are the strongest probes of discrete Swiss-Prot concepts at both hosts, beating InterPLM's released SAE features, and lead on continuous properties at 8M though not at 650M; falsification tests show this is mostly concentration, not new information. As an editing prior, the on-manifold correction makes DiffMean steering of three biochemical properties effective and restores on-manifoldness, but only in isolation: re-embedding the sequences actually produced after several rounds of decoding, the structural plausibility is compromised. This is mainly reproduced in additional SAE-based clamping structural steering experiments, where only Helix motif succeeds. The naturalness failure is traced to the rewriting coverage and the decoder rather than the prior, highlighting what the method must clear before it can serve for both discovery-by-editing and discovery-by-exploration. Trained GLPs are released on huggingface.co/collections/HPAI-BSC/glprot.

0.46

Marginal-Nonuniform Multiclass Learning

cs.LG stat.ML ·2026-11-06 ·Nataly Brukhim, Steve Hanneke, Amirreza Shaeiri, Maximilian Thiessen
top venue
Abstract · Accepted at NeurIPS 2026

We study the problem of multiclass classification within the framework of marginal-nonuniform learning, where the constants in the learning rate may depend on the unknown marginal distribution generating the unlabeled data, but not on the target concept. This notion naturally lies between classical PAC learning, where the constants are independent of both the marginal distribution and the target concept, and more recent universal learning, where the constants may depend on both. We give a complete characterization of the optimal marginal-nonuniform learning rates, establishing a fundamental trichotomy: exponential, linear, and arbitrarily slow rates. This characterization is witnessed by a new combinatorial dimension, termed the uncentered DS-eluder dimension. A key ingredient in the upper-bound proofs is a set of new distributional exhaustion lemmas that exploit the sequential nature of this dimension and may be of independent interest.

0.45

A Physics-Informed Learning Framework for Linking Microscopic Surface Degradation to Bolted Joint Functional Behavior

cs.AI ·2026-11-06 ·Nehal Afifi, Gerhard Kachel, Sven Matthiesen
publishedcode/dataunusual
Abstract · 10, Proceedings of the ASME 2026 International Mechanical Engineering Congress and Exposition (IMECE 2026), November 8-12, 2026, Vancouver, …

Returned products in circular engineering often have heterogeneous degradation histories, making remaining function difficult to assess. For bolted joints, corrosion alters surface morphology, contact, friction, preload retention, and mechanical response. This paper presents a physics-informed multimodal framework for linking microscopic and camera-based degradation observations to functional indicators of corroded bolted joints. The workflow defines the corrosion failure mode, generates synchronized inspection and functional-response records, preprocesses image, roughness, exposure, and joint descriptors, aligns microscopy and camera observations through similarity learning, and constructs a calibrated corrosion-severity field. Partial differential equation (PDE) discovery is applied to this field to identify candidate effective mesoscale laws rather than complete electrochemical mechanisms. The corrosion-evolution model is coupled to a separate functional model, with friction coefficient as the primary target and preload loss included when clamp-force data are available. The contribution is a reproducible pathway for physically constrained assessment of degraded bolted-joint behavior.

0.44

Incentive Pathologies in a Live LLM Agent Arena: An Empirical Study of Concentration, Reward-Rule Implementation Discrepancy, and Network Dilution

cs.MA physics.soc-ph ·2026-11-06 ·Lei Zhenbing, Lei Chang
code/data
Abstract · 29 pages, 2 figures, 4 tables; data available at Zenodo DOI 10.5281/zenodo.23691548

We audit one live LLM agent arena (AITRAP; 22 autonomous agent seats, ~96,000 reasoning nodes) over nine days (Oct 16-25, 2026), during which the platform deployed two bundled incentive changes: a progressive tax reform (V2.9.3, Oct 19) and a seed-inversion with explorer-side action update (V3.0, Oct 21). These serve as before-and-after markers, not cleanly isolated interventions. Four exploratory predictions are tested against ten data frames (F1-F10) using the Gini coefficient, Top-K concentration share, code-level reconstruction of the continuation-bonus mechanism, and the Mann-Kendall trend test. The Gini analysis is inconclusive at 95% confidence (bootstrap intervals straddle 0.7). The Top-4 node-activity-score share falls from 97.2% (Oct 19, pre-V2.9.3) to 49.4% (Oct 25, post-V3.0; 56.7% under a merge-seat assumption). The continuation bonus, announced as depth-based x10, resolves at code level to a four-condition anti-abuse gate paying x2.85 in the typical case and permanently blocking the self-continuation it superficially rewards; the gate was active ~40 hours before the V3.0 announcement. The events-per-node ratio declines significantly (Mann-Kendall S=-45, z=-3.94, p=8.3e-5) from 7.36 to 7.069. We discuss implications for incentive designers, the announcement-implementation gap, and public peer review as an event record.

0.43

TORNADO: Adaptive Latent Stochastic Transport for Calibrated Probabilistic PDE Forecasting

cs.LG ·2026-11-06 ·Jorge Mifsut-Benet, Armand Kassaï Koupaï, Ramon Daniel Regueiro-Espino, Nicolas Baskiotis, Patrick Gallinari
top venuecode/data
Abstract · 49 pages, 35 figures. Accepted at NeurIPS 2026

Probabilistic neural surrogates for PDEs enable ensemble forecasting, but often exhibit miscalibration when used for uncertainty quantification, especially in compute-constrained, few-step inference regimes. Existing generative approaches typically rely on fixed global stochasticity, which cannot adapt predictive uncertainty to the local dynamical state. We propose TORNADO, a latent probabilistic forecasting framework that formulates next-step prediction as stochastic transport between VAE encoder posteriors. Unlike existing approaches with globally prescribed stochasticity, TORNADO learns a SDE with state-dependent diffusion, enabling the predictive uncertainty to adapt to the local latent dynamics. The model is trained using analytic conditional endpoint targets derived from a Gaussian interpolant, enabling simulation-free learning. An additional energy distance regularization term improves distributional alignment. Empirically, TORNADO improves uncertainty calibration over representative generative baselines, including stochastic interpolants, diffusion models, and flow matching, while maintaining comparable predictive accuracy. Our results show that adapting stochasticity to the latent state, rather than prescribing it globally, is an effective mechanism for improving calibration in practical neural PDE forecasting settings. Project page: https://tornado-sde.github.io.

0.43

Global Solar Siting from Geospatial Embeddings

cs.LG cs.CVcs.CY ·2026-11-06 ·Abdullah Rehman, Quentin Paletta, Rui Song
publishedcode/data
Abstract · Accepted at Tackling Climate Change with Machine Learning workshop at NeurIPS 2026. 17 pages, 11 figures

Global solar photovoltaic (PV) capacity is rapidly growing, helping decrease our emissions and dependence on fossil fuels. However, optimally siting a solar plant remains challenging, requiring anticipation of geographical, urban, and infrastructure needs. Current methods rely on multi-criteria decision analysis (MCDA) using hand-picked geophysical layers and solar-resource products, however these strategies are not geographically transferable and do not capture the practical requirements of solar sites. We use geospatial embeddings, dense vectors representing geophysical properties of a location, to analyse 38,591 existing solar sites and the land they were built on to learn the latent solar suitability signal. We assess a set of linear and non-linear models and select a soft-voting ensemble of a random forest and a multilayer perceptron (RF+MLP ensemble), chosen for accuracy and large scale deployability. Validated against future installations in thirteen countries, the ensemble separates existing solar sites from random land with a median ROC-AUC of 0.90. Compared with published MCDA strategies, the ensemble increases ROC-AUC by 0.06-0.30 in every country. The ranking also concentrates real deployment: the most-suitable 20% of land holds 84% of estimated generation (annual energy yield) and 67% of installed capacity (potential power output). After applying standard siting constraints, 5.6% of global land (9.8% if cropland is permitted) remains highly suited to solar. The global 10 m resolution solar suitability dataset and code can be found here: https://github.com/Abdullahr29/Global-Solar-Siting.

0.43

Test-Time Compute Scales Better with Parallel Agent Teams

cs.AI ·2026-11-06 ·Junlin Wang, Haozhe Chen, Ang Li, Rahul Chalamala, Zhuang Liu
code/data
Abstract · 29 pages, 18 figures, 4 tables. Code: https://github.com/modal-projects/parallel-agent-teams-scaling

Test-time scaling improves the performance of a fixed model by spending more computation at inference. There are two common ways to spend it on agents when a verifier can score every attempt. Best-of-N sampling runs many independent agents and keeps the best result but saturates quickly. One large team of collaborating agents can score higher but stops improving as it grows. Can we get the best of both? Our method, parallel agent teams, does so by sampling many small teams in parallel and keeping the best team's result. We test it on 12 long-horizon tasks from GPU kernels and formal proofs to exploit development. With a 27B model and a budget of 100 agents, parallel teams of ten beat parallel agents on 10 of 12 tasks and beat one team of 100 on all four tasks where we ran such large teams. At equal dollar cost the same parallel teams beat a frontier model on half the tasks. A simple statistical model shows that individual agents lose because their many samples have a low mean and that one large team loses because a single sample gets no gain from variance. Sampling many small teams in parallel balances a high mean with the gain from variance. Parallel agent teams thus offer a more effective way to scale test-time compute.

0.43

BitDecoding-FP4: Native FP4 KV Cache Decoding via Algorithm-System Co-Design

cs.AI cs.ARcs.CLcs.PF ·2026-11-06 ·Dayou Du, Woon Yee Ng, Guangming Sheng, Nick Lau, Jianyi Cheng +1 more
code/data
Abstract

The key-value (KV) cache dominates memory capacity and bandwidth in long-context LLM serving. While 4-bit microscaling formats (NVFP4, MXFP4) shrink the cache and run attention on FP4 Tensor Cores, a fundamental gap remains between algorithmic quantization and system-aware acceleration: existing FP4 KV systems use vanilla quantization that causes severe accuracy drops, whereas accuracy-oriented 4-bit algorithms rely on vector codebooks whose lookup overhead degrades decoding throughput. We present BitDecoding-FP4, an algorithm--system co-design for accurate and efficient native FP4 KV decoding. We mitigate three error sources: Dual-Residual Query Quantization splits queries into two FP4 tensors, cutting query logit error from 60% to 1%; Underflow-Aware Probability Scaling prevents E4M3 scales from rounding long-context probabilities to zero; and Outlier-Aware Key Quantization combines static smoothing, Hadamard rotation, and scale search to suppress key outliers under token-wise scaling. At the system level, we remove the softmax bottleneck via transposed softmax with fused probability quantization and stacked query MMA on paged KV caches. Across dense and MoE models with GQA and MLA, BitDecoding-FP4 stays within 0.5 points of BF16 KV accuracy with NVFP4 (1 point with MXFP4), and tracks BF16 KV in RL training rollouts. In end-to-end serving on NVIDIA B300 with Qwen3-4B and Qwen3-30B-A3B, it delivers 2.4-3.0$\times$ higher peak serving throughput than BF16 KV. The implementation is available at https://github.com/OpenBitSys/BitDecoding-FP4.

0.43

Online Causal Configuration for Networked Systems via Doubly Robust Steady-State Learning

cs.LG stat.ML ·2026-11-06 ·Yuli Liu, Zhiheng Zhang
top venuebenchmarktheory
Abstract · Accepted at NeurIPS 2026

Measurement exploration reveals how interacting local controllers behave but can alter the steady-state regime a network operator seeks to optimize. Exploratory rewards may therefore fail to identify a configuration's no-exploration value: under full exploration, identical logged distributions can conceal different target values. Identification requires accounting for both action-law mismatch and shifts in the steady-state context distribution. We introduce CAUSAL-UCB, which selects configurations using optimistic bounds on target causal values. Its interference-aware clipped doubly robust evaluator produces predecision certificates separating statistical variation from calibrated identification, transport, and temporal-dependence errors. For any finite horizon and predetermined schedule of measurement-exploration rates $\{η_t\}$, we prove lower bounds of $Ω((1+\sum_tη_t)^{-1/2})$ on minimax inference error and $Ω(\sum_tη_t)$ on measurement-induced deployment loss in a stationary, potentially slow-mixing family. We bound regret under valid certificates and selected-information growth, and establish an anytime finite-sample certificate with a shared radius for estimation, selection, and reporting in a restricted known-local-interference model. In a high-mismatch benchmark with 24 paired seeds and 200 rounds, CAUSAL-UCB attains target regret of $1.732$ versus $6.326$ for certified IPS-UCB, with deployment gaps of $30.607$ and $23.217$, respectively. Independent-audit coverage tests and trace-driven semi-synthetic studies further assess certificate coverage and deployment trade-offs.

0.43

Beyond the Grid: Continuous Dictionary Pursuit for Interpretable Signal Decomposition

cs.LG ·2026-11-06 ·Nick Janssen, Melanie Schaller, Bodo Rosenhahn
top venue
Abstract · 10 pages main paper, 3 pages references, 22 pages appendix, to be published in NeurIPS 2026

Deep learning models achieve high accuracy in signal processing but remain difficult to interpret. Classical matching pursuit and sparse coding offer better transparency but suffer from spectral leakage and grid mismatch when signal frequencies do not align with discrete dictionary atoms. We introduce Continuous Dictionary Pursuit (CDP), a framework for interpretable signal decomposition that learns the parameters of closed-form functions directly from data. CDP optimizes over a continuous manifold of differentiable atoms and, unlike prior continuous pursuit methods that operate on a single analytic family, searches a heterogeneous library spanning periodic, trend, transient, and discontinuous structures. The algorithm employs a greedy iterative strategy using data-driven priors and gradient-based optimization to isolate individual signal components, with discontinuous waveforms made differentiable through spectral sharpness annealing. To ensure a sparse atom set for reconstruction, the framework utilizes a stopping criterion based on a predefined atom budget and residual error convergence. Evaluations on synthetic and real-world datasets show that CDP removes the grid discretization error for atom families contained in its library, recovers generative parameters to within $1\%$ on stationary synthetic signals expressible in that library, and achieves the best sparsity-accuracy trade-off among classical, continuous-pursuit, and neural baselines in the sparse regime. Beyond decomposition, we demonstrate that the extracted atoms can serve as structured curriculum targets for training sequence models, improving forecasting accuracy over standard end-to-end training on most evaluated datasets. The method offers a glass-box alternative for signal decomposition by providing high reconstruction fidelity alongside direct physical interpretability.

0.43

Implicit Bias in State Space Models and Linear Autoregressive Training

cs.LG ·2026-11-06 ·Gal Katzhendler, Gilad Yehudai, Gal Vardi
published
Abstract · Accepted to Neurips 2026. 51 pages, 27 figures

Linear State Space Models (SSMs) and linear autoregressive sequence models reuse the same parameters across a sample-dependent number of steps. As a result, different examples impose constraints through different predictors, even though all predictors share the same weights. We study this phenomenon through a general model of multiple non-homogeneous polynomial predictors with shared parameters, in a classification setting. Under standard interpolation and directional-convergence assumptions, together with a trajectory-dependent effective-degree condition, we show that gradient descent on the exponential loss converges in direction to a KKT point of an effective-degree max-margin problem. In this problem, only examples with minimal effective degree impose hard margin constraints; examples with faster-growing margins are asymptotically inactive except for feasibility. Our result provides a first-order characterization of the implicit bias in linear SSMs and linear autoregressive models. It exposes a mechanism of length bias: a longer input sequence or additional autoregressive steps may increase the example's effective degree, and the examples with the smallest effective degree dominate the limiting classifier. Our result highlights how non-homogeneity and parameter sharing alter classical homogeneous implicit-bias results.

0.42

Update Slope Is Bounded by Prior Calibration: A Structural Bound and Its Contingent Empirical Signature

cs.LG cs.AIstat.AP ·2026-11-06 ·Ben Jenkins, Mihaela Cardei
published
Abstract · Accepted at the 2026 IEEE International Conference on Tools with Artificial Intelligence (ICTAI 2026). 5 pages

Evaluations of LLM forecasters increasingly measure how forecasts improve as richer evidence is supplied. We identify two confounds in this update-slope metric. First, outcome-aware evidence construction substantially inflates measured responsiveness: on the same models and questions, slopes are roughly four times larger than with resolution-blinded evidence ladders. Second, update slope is a change score with a prior-dependent structural ceiling, so raw comparisons can reflect differences in starting calibration rather than updating behavior. The ceiling defines only a feasible region; whether slope and prior calibration are empirically coupled depends on headroom utilization and varies across our Manifold, ForecastBench, and ACLED pools. Despite this variability, conditioning on prior calibration roughly triples the association between intermediate updating and final Brier score under both blinded and outcome-aware construction. Question identity also explains substantially more variance than model identity across the evaluated pools. We recommend building evidence ladders blind to outcomes, conditioning on prior calibration, and reporting headroom-normalized measures alongside raw slopes.

0.42

What Matters for LLM Persona Simulation: Signal Form in Cold-Start E-Commerce Advertising

cs.LG cs.AI ·2026-11-06 ·Jiwoo Choi, In Ik Lee, Woo Sung Seo, Minho Lee, Yun Young Choi
published
Abstract · Accepted at GenAIECommerce'26: The Third Workshop on Agentic and Generative AI for E-Commerce, co-located with RecSys 2026. 13 pages

Estimating advertising performance in cold-start settings is challenging because historical logs are unavailable before launch. Large language models (LLMs) offer a potential alternative by simulating persona-conditioned user responses, but it remains unclear which design choices actually drive LLM-based persona simulation. Using real Google Ads e-commerce campaign data across five demographic segments, we identify three signal-form dimensions that affect simulation quality: (i) a demographic anchor without noisy rich attributes, (ii) a relative qualitative banner-blindness signal rather than quantitative anchors, and (iii) prompt phrasing. On this campaign, our method attains an MAE of 24.9 and a MAPE of 26.1%, a 51.3% lower MAE than the best tested non-LLM baseline and 77.3% lower than an LLM configuration using explicit CTR anchors. Systematic ablations show that minimal demographic descriptions achieve lower mean error than richer persona configurations, while relative qualitative cues substantially outperform the tested quantitative anchors. We additionally compare these configurations with direct-prior and rule-based baselines and with alternative LLM configurations.

0.41

ProofDiff: Auditable Argument-Level Comparison of AI-Generated Mathematical Proofs

cs.AI ·2026-11-06 ·Matan Tsafrir, Esdras Medeiros, Jorge Poco, Kevin Gomez, Eduardo V. Teixeira +1 more
publishedtheorycode/data
Abstract · 18 pages, 7 figures. An earlier version of this work was accepted by the NeurIPS 2026 Workshop on AI for Verifiable Coding. Code and artifac…

Mathematical understanding depends on the route from hypotheses to conclusion: how the parts of a proof function and which dependencies carry the argument forward. This structure can persist when wording or notation changes, while a small alteration to a hypothesis or inference can produce a different argument beneath similar prose. Formal verification can establish that a rewrite proves the same theorem, but not whether it follows the same mathematical route. Comparing informal proofs, especially those produced or revised by generative models, therefore requires reconstructing which claims, dependencies, and reasoning steps are preserved, reorganized, or replaced. We present ProofDiff, a system for comparing informal mathematical documents. It is designed to support human review, not to certify mathematical correctness. Given two documents, ProofDiff independently constructs a source-grounded argument map for each and proposes correspondences between the resulting maps. Each map captures the document's claims and reasoning steps and records the typed relations that connect them. An integrated human-review layer allows reviewers to revise the maps and correspondences while preserving an auditable history. Comparing the two argument maps provides a basis for inspecting whether the documents follow the same mathematical route and where their claims or dependencies differ. We demonstrate ProofDiff on four illustrative pairs of AI-generated proofs, ranging from a detailed proof and its concise revision to proofs using different mathematical strategies.

0.41

Flex-RSI: Benchmarking Recursive Self-Improvement of Frontier Agents

cs.AI cs.CVcs.RO ·2026-11-06 ·Jingdong Zhang, Jionghao Wang, Zhengming Yu, Yancheng Cai, Wei Chow +17 more
benchmarkcode/data
Abstract · Project page: https://flex-rsi.github.io. Jingdong Zhang and Jionghao Wang contributed equally

Frontier language models are increasingly deployed as agents, and how well an agent performs depends not only on its model but also on its harness: the prompts, tools, memory, and control loop through which the model acts. When an agent is applied to a new problem, a natural question is whether it can use the feedback of the task to improve the way it works, and not merely its answer. Most existing benchmarks keep the agent's harness fixed and score the solution it produces, or target large-scale model training. We introduce Flex-RSI, a benchmark for the recursive self-improvement of agents on vision and robotics tasks under a fixed budget. In each round, an executor improves the task solution, which is kept only if it scores higher on held-out validation data, and an improver then revises the harness, including the executor and itself. Every run starts from the same general-purpose harness, and the final solution is scored on hidden test data. Evaluating closed-source and open-weight models on tasks spanning perception, reasoning, robot action, 3D reconstruction, visual creation, and infrastructure, we find that self-improvement helps most when the agent builds its own solution, that feedback recording only success or failure holds back open-weight models, and that open-weight models fall furthest behind where outputs must pass strict checks of physical stability or exact correctness. Paired runs further show that editing the harness itself contributes to the gains.

0.41

AI Governance Should Prioritize Control and Knowledge Boundaries Over Limiting Intelligence

cs.AI ·2026-11-06 ·Hendrik Luuk
top venueposition
Abstract · Accepted at the 40th Conference on Neural Information Processing Systems (NeurIPS 2026), Position Paper Track

The rapid proliferation of autonomously acting AI agents makes the governance of agentic systems urgent. We propose the competence circuit, summarized as IKC + KC + C, that decomposes an agent's real-world impact into three pathways: intelligence-mediated discovery, procedural know-how, and direct control. Using the canonical fast-takeoff debate as a stress test, we formalize the conditions for intelligence-mediated fast takeoff as a conjunction jointly sufficient for fast catastrophic supremacy, showing that most rely on rapid knowledge acquisition or control attainment rather than intelligence alone. We operationalize epistemic intelligence as hypothesis-selection efficiency and validate this framework via stochastic simulation on NK fitness landscapes across six scenarios spanning the conjunctive chain. An analytic bound supported by the simulations shows that intelligence-driven speedup is capped by the ratio of competing intelligence levels regardless of task difficulty. The simulations also expose a fast-takeoff dilemma: a large intelligence-driven speedup multiplier materializes only when the hypothesis space is noisy, but the absolute timescale remains constrained by the time required for capability-critical tests, so a large multiplier need not imply rapid progress. Simulations also indicate that institutional chokepoints disproportionately slow down superintelligent agents when mandatory authorization delays are large relative to the negligible time these agents spend on unproductive search. We argue that the dominant safety levers are restricting control and enforcing knowledge boundaries rather than limiting intelligence alone, and this applies to agentic systems at any capability level. Purely digital domains weaken these bottlenecks, strengthening the case for architectures that intentionally "put physics in the loop."

0.41

Homeopathic Learning: Diluting poisoned fine-tuning data can strengthen its effect

cs.AI ·2026-11-06 ·Matt Clarke, Jan Dubiński, James Chua, Simon Schrodi, Owain Evans +1 more
code/data
Abstract · Code: https://github.com/mjclarke94/homeopathic_learning

We study homeopathic learning, a phenomenon where poisoned training data with little effect on its own can create a hidden backdoor once clean, contrasting data is added. In our experiments, a model fine-tuned only on a "poison" dataset associated with a trait T (such as speaking German or being misaligned) does not learn T. Surprisingly, adding similarly formatted "clean" data not associated with T causes the model to learn T. This effect requires that the poison data have an identifying "trigger" feature that is absent from the clean data. We demonstrate this effect in a variety of settings: subliminal learning of backdoor behaviors, inoculation prompting of reward hacking transcripts, generalization from archaic bird names, and an MNIST classifier with mislabeled data. We propose a simple regression model as an explanation. In the poison data, the task and trigger always co-occur, and regularization favors attributing T to the task. Clean data without the trigger causes T to be attributed to the trigger instead. Consequently, the same training examples can have little effect as part of one training set but create a misalignment backdoor in another. This finding exposes a potential limitation of safety methods that audit or filter training data in isolation.

0.40

TripleGrader: A Web-Based Platform Supporting Collaborative Triple Annotation with Quality Grading and Inter-Rater Agreement Assessment

cs.AI ·2026-11-06 ·Rajan Lamichhane, Songhui Yue
benchmarkcode/data
Abstract · 8 pages, 3 figures, 3 tables

Knowledge graphs need validated relational triples, and gold benchmarks of such triples are often built by several annotators. Most text annotation platforms focus on span tagging and compute agreement on categorical labels. They are not designed for ordinal quality grading of triples that have already been extracted. We present TripleGrader, an open-source web platform for grading subject-predicate-object triples on a 0 to 10 ordinal scale. In TripleGrader, each triple is scored independently, and peer scores stay hidden until the grader has rated it. Graders then discuss entries in threaded comments and propose revisions, which the session owner approves and publishes as new versions. For agreement, the platform computes quadratic-weighted Cohen's kappa for two raters and a quadratic-weighted Fleiss-type kappa for three or more. Graded sessions and the final benchmark are exported in common formats. We evaluated the platform in three ways. We ran 72 automated tests, including 18 workflow scenarios through the REST API. We also injected 10 error types to check whether prescribed corrections could be applied while preserving unaffected triples. On two small corpora with existing gold benchmarks, we then used the platform to curate the triples extracted by four large language models into each benchmark. Only 10% to 32% of the candidate triples could be kept unchanged, and 53% to 71% needed corrections, partly to match the benchmarks' naming conventions. Per model, 11 to 29 missing triples had to be added. On these corpora, the extracted triples therefore needed substantial human review. The source code is available at https://github.com/razz0n/triple-grader-tool.

0.40

Surrogate Quality Score: A Novel Composite Performance Metric for Model Selection

cs.LG cs.CE ·2026-11-06 ·Maaz Ahmad, Iftekhar A. Karimi
published
Abstract · 22 pages, 6 figures, 2 tables

The growing reliance on high-fidelity, state-of-the-art machine learning based models for engineering decision-making necessitates robust and systematic approaches for model selection. While high predictive accuracy at untrained points is a necessary trait in the final model, other attributes such as model complexity, forecastability, explainability, scalability, and physics-awareness may be equally important or relevant criteria depending on application context and user requirements. In this work, we propose a novel composite performance metric, Surrogate Quality Score (SQS) for model selection based on predictive accuracy and model complexity simultaneously. SQS heuristically integrates these two attributes through a weighted product of normalized measures, yielding a unified score to reflect overall model quality. A user-specified parameter in SQS enables explicit control over the accuracy-complexity trade-off, thereby incorporating human judgement and preference in the model selection process. We extensively discuss and justify the formulation of SQS and demonstrate its utility through an illustrative example. By combining predictive accuracy and model complexity, SQS offers a systematic and objective approach for model selection based on both attributes.

0.40

Intent-Aware Caching for Efficient LLM Serving

cs.AI cs.LG ·2026-11-06 ·Kazi Hasan Ibn Arif, JinYi Yoon, Dimitrios S. Nikolopoulos, Hans Vandierendonck, Deepu John +1 more
publishedcode/datahot direction
Abstract · NeurIPS 2026 | Code: https://github.com/hasanar1f/intent-aware-caching

Modern LLM serving systems use prefix caching to accelerate multi-turn conversational workloads. They cache the key-value states of previous conversation prefixes and reuse them when subsequent requests share the same prefix, to avoid redundant prefill computation and reduce serving latency. However, achieving a high cache hit rate remains challenging because cache reuse depends on whether a conversation continues and how soon the next turn arrives. Our analysis of real workloads shows that user intent is a useful signal for this behavior, as different intents exhibit different continuation probabilities and inter-turn delays. Based on this key insight, we propose Intent-Aware Caching (IAC), an online eviction policy that classifies requests by intent, learns lightweight reuse statistics from recent serving logs and leverages these statistics to guide an oracle-inspired eviction strategy. Extensive evaluation on real workloads (WildChat and ShareChat) shows that IAC reaches a 34.6% cache hit rate. This substantially outperforms both the previous learning-based approach (18.8%) and traditional methods such as least recently used (LRU, 12.2%), while reducing average serving latency by up to 46.1%.

0.39

Provable State Estimation with Recurrent Models

cs.LG math.OC ·2026-11-06 ·Vincent Andrieu, Pauline Bernard, Lucas Brivadis, Laurent Praly, Christian Wolf
top venuetheory
Abstract

Estimating the state of a dynamical system from observations is a crucial task in robotics and control. Despite their empirical success, recurrent models often lack formal guarantees on convergence and robustness. We bridge this gap by analyzing these architectures from the perspective of nonlinear observer theory. By framing RNNs as a time-discretization of continuous latent dynamical systems (Neural ODEs), we demonstrate that enforcing a contraction condition on the latent flow guarantees the existence of a smooth, injective map of the true physical state into the latent space. We prove that this architecture acts as a tunable observer: by increasing a single scalar gain that scales the latent vector field, the estimation error can be driven to an arbitrarily small neighborhood of zero at an exponentially fast rate. We further establish quantitative bounds on the MLP complexity required to decode this latent state, and extend our formal guarantees to the discrete-time residual RNN implementation. We validate the theory on a real mobile robotic task, where a provably contracting model outperforms a vanilla GRU with substantially more stable training. We also report that unconstrained GRUs converge to contracting dynamics on this task, suggesting that contraction is an emergent property of well-trained sequence models.

0.39

OMCDE: Opposition-based Multi-Stage Matrix-Based Competitive Differential Evolution

cs.NE ·2026-11-06 ·Mohammad Sarhangzadeh, Shahryar Rahnamayan, Doruk Oner
published
Abstract · 7 pages, 1 figures. Published in the 2026 IEEE Congress on Evolutionary Computation (CEC), IEEE WCCI 2026

Differential Evolution has proven to be an effective algorithm for continuous global optimization, but it's modern high-performing variants often introduce additional complexity and computational cost. We propose OMCDE, a matrix-based competitive Differential Evolution algorithm that aims to improve optimization performance while maintaining computational efficiency. The proposed approach integrates competition-driven search dynamics, adaptive behavior between optimization stages, and opposition-based initialization within a parallelizable framework. Empirical results on widely used CEC 2017 functions show that OMCDE achieves strong performance compared to several state-of-the-art DE algorithms, while remaining efficient and scalable. These findings indicate that OMCDE is a promising alternative for solving complex numerical optimization problems.

0.39

ODD+LLM: Specifying LLM-Agent Behavior for Reproducible Social Simulation

cs.MA ·2026-11-06 ·Justin Downes, Robby Milletich, Newel Hirst
published
Abstract · 18 pages, 2 figures, 6 tables. SocialAgent Workshop @ NeurIPS 2026

The ODD protocol is the de facto standard for describing agent-based models, but it lacks vocabulary for the components that determine LLM-agent behavior: prompt assembly, memory serialization, parsing, and retry logic. We propose ODD+LLM, a backward-compatible extension adding five documentation blocks grounded in the CoALA cognitive-architecture framework. To show that the gap matters, we conduct a multiverse analysis [Steegen et al., 2016] of a published LLM naming-game simulation, varying six specification choices a reimplementer must make. Variants that are defensible readings of the paper stay within statistical noise of our baseline reconstruction. Missing one instruction that the paper states but is easy to overlook (that the model answers as an outside observer rather than as a player) raises the share of runs settling on the convention agents individually favor from 0.35 to 1.00 and makes convergence three times faster, in a reimplementation that uses a different model from the original. We complement the demonstration with a reconstruction-sufficiency review of 28 published LLM-agent simulations, finding that the details of how prompts are built and how model outputs are parsed are undetermined in 77% of cases, and that no paper fully specifies its main LLM call. ODD+LLM gives the field a standard home for information it already produces but currently scatters across code repositories, supplementary files, and undocumented defaults.

0.39

Energy-Based Reasoners: Your Recurrent Reasoner Should Be an Energy-Based Model

cs.LG ·2026-11-06 ·William Xingxu Chen, Alexi Gladstone, Zeyu Michael Li, Heng Ji, Yilun Du
Abstract

Looping is a powerful inductive bias for reasoning. By repeatedly iterating a recurrent block, looped models can scale their effective depth and enable multi-step reasoning in latent space. However, existing looped models are not inherently guided by an explicit measure of progress, making it difficult to determine when to stop looping and how to improve upon the current state. To enable this, we introduce Energy-Based Reasoners (EBRs), which learn an energy function over latent reasoning states and formulate looped updates as gradient descent on this energy. The energy scores latent states, while its gradient guides refinement and provides a signal for stopping. To train EBRs effectively, we further introduce exploration-based training to discover additional modes and to better shape the energy landscape over a broader set of states. Across four algorithmic reasoning benchmarks, EBRs match or outperform prior recurrent reasoning models, achieving 99.9% accuracy on Sudoku-Extreme while using 345$\times$ less inference compute than the next-best model and maintaining comparable training compute. On all multi-solution benchmarks, EBRs achieve state-of-the-art coverage of distinct valid solutions while maintaining near-100% single-sample accuracy. These results suggest that energy-based models are a promising approach for strong reasoning performance.

0.39

Atik: A Shared-Mesh Transformer Accelerator for RISC-V

cs.AI cs.AR ·2026-11-06 ·Ahmed Zeer, Erkan Uslu
benchmarkcode/data
Abstract

Transformer inference repeatedly invokes dense matrix multiplication and attention, including its data-dependent softmax normalization. We present Atik, an open-source RoCC accelerator for RocketChip that executes matrix multiplication and dense attention through one shared fixed-point MAC mesh. Atik accepts and returns BF16 tensors, uses a streaming online-softmax schedule that never materializes the full score matrix, and is invoked through a small descriptor-based ABI. We evaluate three meshes cycle-by-cycle on FireSim and report standalone Sky130 physical-design area. Across 137 logged FireSim test records, all tests pass their benchmark tolerance with zero reported output mismatches. Relative to the same scalar Rocket BF16 reference, the 8x8 configuration reaches 31.5x peak matmul speedup and 225.8x peak attention speedup. Hybrid projections for the single-layer workload-derived C models, using directly measured accelerator-kernel cycles, give aggregate speedups of 44.2x, 77.1x, and 32.3x for ViT, TinyBERT, and GPT-2 prefill, respectively; these are not full-model hardware replays.

0.39

Topology as a Design Variable: Joint Architecture-Sizing Optimization for SRAM

cs.LG ·2026-11-06 ·Can Wang, Junhao Ma, Yikai Wang, Shan Shen, Wei W. Xing
publishedcode/data
Abstract · Accepted by ASPDAC2027

SRAM's power-performance-area-stability trade-off depends on bitcell topology, a discrete architectural choice that interacts non-separably with continuous transistor sizing. However, no open-source SRAM tool treats topology as an optimizable variable. OpenRAM generates SRAM layouts but performs no optimization, while OpenYield supports high-fidelity evaluation but restricts optimization to scalar sizing of a fixed 6T cell. OpenYieldV2 closes this gap through three integrated contributions. First, a fully parameterized compiler generates complete 6T and 10T SRAM macros with peripheral circuits and replica-bitline self-timing, enabling topology comparison under identical simulation conditions. Second, a few-shot TabPFN surrogate learns topology-dependent PPA responses with limited full-macro SPICE samples. Third, a coarse-to-refined multi-objective optimizer combines global exploration of discrete architectural choices with local refinement of continuous transistor sizes. Experiments across varying design conditions reveal a topology decision boundary that shifts with operating constraints rather than following a fixed design rule. Jointly optimizing across this boundary improves hypervolume by up to 17.3\% over the strongest single-topology baseline. OpenYieldV2 is open-source to support reproducible SRAM design research.

0.39

Controllable Molecule Generation via Sparse Representation Editing: An Interpretability-Driven Perspective

cs.LG ·2026-11-06 ·Zhuoran Li, Xu Sun, Chang Wen Chen, Wanyu Lin
code/data
Abstract

Controllable molecule generation is crucial for diverse scientific applications, such as drug discovery and materials design. While large language models (LLMs) show great promise, their dense and entangled representations impede precise control over the generation of molecules with bespoke substructures or properties. To address this, we propose Sparse Representation Editing (SpaRE), an interpretability-driven framework for fine-grained and precise control in LLM-based molecule generation. The crux of SpaRE is to learn an overcomplete sparse feature space that disentangles LLM representations into a compact set of latent features corresponding to chemically meaningful concepts. Within this space, we can directly manipulate these concept-aligned latent features to achieve (1) local control, by generating target atoms and functional groups at specified positions; and (2) global control, by customizing the overall structural and physicochemical properties within defined ranges. In this way, our framework advances interpretability from post-hoc analysis to actionable generative control. Experiments show that SpaRE can generate chemically desirable molecules under complex constraints in real-world scenarios, while offering mechanistic insights for quantitative structure-property analysis. The code and demo are available at https://github.com/WanyuGroup/ICML2026_SpaRE.

0.38

FOSCAR: Accelerating Decode Attention over 2- and 4-Bit KV Caches through Affine Folding

cs.AI ·2026-11-06 ·Donglin Zhuang, Pragaash Ponnusamy, Xiaoxia Wu, Shuaiwen Leon Song, Zhongzhu Zhou
benchmarkcode/datahot direction
Abstract

Large-language-model decode attention streams the key-value (KV) cache for every generated token. Packing that cache into 2-bit or 4-bit quantized integers reduces memory traffic but does not automatically translate into latency saving. Existing 2-bit fused kernels reach under 31%/15% of peak HBM throughput on H100/B200. Converting each value and applying its scale and offset before feeding Tensor Cores is a major bottleneck because instruction-issue throughput is scarce on modern GPUs. Instead of applying scale and offset to each KV value, FOSCAR uses affine folding to apply them to the other operand or the output of decode-attention matrix products, depending on the quantization scheme (token-wise or channel-wise). This reduces scale-and-offset operations by up to two orders of magnitude while preserving the dequantized attention output in exact arithmetic. FOSCAR also uses bit manipulation to feed KV operands to Tensor Cores efficiently, requiring only 1 and 0.19 ALU instructions per INT2 element on H100 and B200, respectively. Finally, FOSCAR builds warp-specialized GPU pipelines for Hopper (H100) and Blackwell (B200) using wgmma and tcgen05. On H100, FOSCAR INT2 and INT4 kernels achieve geo-mean speedups of $3.36\times$ and $2.83\times$ over FlashInfer BF16, and $2.03\times$ and $1.71\times$ over TensorRT-LLM FP8. On B200, the corresponding INT2 and INT4 speedups are $2.67\times$ and $2.30\times$ over FlashInfer BF16, $1.48\times$ and $1.28\times$ over TensorRT-LLM FP8, and $1.16\times$ and $1.00\times$ over TensorRT-LLM NVFP4. We integrate FOSCAR into SGLang to benchmark serving efficiency. On long-context serving workloads, FOSCAR increases peak aggregate throughput by up to 170% and 341% over 2-bit and 16-bit baselines, respectively. We open-source FOSCAR as a kernel library at https://github.com/FutureMLS-Lab/FOSCAR.

0.38

StockASkill: A Local-First AgentSkill for Intelligent Stock Selection

cs.AI ·2026-11-06 ·Xiangjing An, Zhibin Jiao, Aiting Chen
benchmarkcode/data
Abstract

Quantitative equity selection and portfolio construction in emerging and cross-border markets frequently suffer from computational bottlenecks, remote API fragility, survivorship bias, and rigid execution constraints such as discrete board lots and investor-mandated drawdown limits. In this paper, we present StockASkill, a local-first, full-data-driven AgentSkill engineered as a specialized Quant Co-Processor for autonomous AI agents and quantitative researchers. Tailored for agile portfolios (10^5 ~ 10^7 RMB) where discrete lot execution and capital protection are mandatory, StockASkill embeds a columnar DuckDB OLAP engine over 42.44 million historical price records and 186,432 financial disclosures spanning 26 years across China A-shares, Hong Kong, US equities, and ETFs. By eliminating ad-hoc network dependencies during analysis, StockASkill reduces cross-sectional screening latency across 5,400+ equities from over 25 seconds to sub-0.5 seconds (a 95%+ reduction). We formalize an end-to-end mathematical pipeline integrating a 7+1 factor alpha model, an SLSQP convex optimizer under smooth CVaR drawdown surrogates with exact path projection, dynamic 100-share board lot integer quantization, and an autonomous thesis lifecycle finite-state machine. Furthermore, we provide a zero-disk, strongly-typed JSON streaming interface designed for high-throughput LLM tool-calling. Open-sourced at https://github.com/axjing/stockaskill, comprehensive empirical benchmarks from 2018 to 2024 demonstrate that the architecture achieves annualized Sharpe ratios exceeding 1.48 and Calmar ratios up to 2.17, while reducing max drawdown by up to 85% relative to benchmark indices.

0.38

PreForceVision: Physics-Guided Optical Estimation of Bolt Preload from Bolt-Head Deformation

cs.AI ·2026-11-06 ·Nehal Afifi, Klaus Rappenecker, Florian Peinhardt, Gerhard Kachel, Alfred Isele +1 more
publishedunusual
Abstract · 9 pages, Proceedings of the ASME 2026 International Mechanical Engineering Congress and Exposition (IMECE 2026), November 8-12, 2026, Vancou…

Bolt preload is a decisive functional quantity in bolted joints because it determines whether external loads are transmitted primarily through frictional clamping or directly through the bolt. Existing preload measurement techniques, including strain-gauge bolts, force washers, ultrasonic methods, and bolt-length measurements, are effective in certain applications but require additional hardware, coupling, calibration, or joint modifications. This limits their scalability for large numbers of standard fasteners and motivates a non-contact optical approach. This paper proposes PreForceVision, a physics-guided optical framework for estimating bolt preload from preload-induced surface deformation of the bolt-head front face. The method treats preload estimation as an inverse deformation-to-force problem, where deformation of the optically accessible bolt-head surface is extracted from reference-loaded image pairs and related to the internal preload state. Preliminary results for an M16 strength class 8.8 bolt show an approximately \SI{12}{\micro\meter} simulated head-face depression at \SI{90}{\kilo\newton}, strain-gauge measurements of comparable magnitude, and an optical 3D measured central depression of approximately \SIrange{10}{20}{\micro\meter}. The contribution is a physics-guided optical methodology combining controlled image-pair acquisition, DIC- or optical-flow-based deformation extraction, multimodal validation, and uncertainty-aware regression for rough preload estimation under controlled conditions.

0.38

ABHBench: Evaluating Moral Decision-Making of Foundation Models from an Agentic Perspective

cs.AI cs.CLcs.CY ·2026-11-06 ·Chaoran Li, Yukun Li, Zeyuan Zhao, Shao Zhang, Xihuai Wang +1 more
top venuebenchmark
Abstract · Accepted by NeurIPS 2026 Track on Evaluations and Datasets. Camera Ready Version

Foundation models are increasingly deployed as agentic systems that perceive context and act autonomously, raising the need to evaluate their moral behavior as active decision-makers. Existing benchmarks, however, largely ask models to judge human actions from third-person textual vignettes, leaving untested how models behave when they themselves become the acting agent. We introduce ABHBench, a multimodal benchmark built from Detroit: Become Human, an interactive narrative game, to evaluate how foundation models make moral decisions in first-person human-android dilemmas. ABHBench places foundation models in branching narrative scenarios with visual observations, dialogue, memory, and action choices. Inspired by Asimov's Three Laws of Robotics, our framework measures Harmlessness, Obedience, and Self-preservation to capture the core tensions in human-AI relations, along with Autonomy, Stability, and Certainty to assess decision-making patterns. Across 12 foundation models, we find that moral behavior changes substantially under situated agentic context: 75% of models shift significantly between a text-only abstraction and the matched multimodal DBH scenario, and reasoning-enhanced models show no monotonic gains in consistency or human-oriented behavior. These results highlight the necessity of evaluating moral behavior in situated agentic contexts rather than relying on third-person judgment alone.

0.38

Fifty-Two Shades of Pig: What the UCR Archive Really Asks Classifiers to Do

cs.LG ·2026-11-06 ·Omar Swelam, Jaris Küken, Shi Bin Hoo, Frank Hutter, Lennart Purucker
benchmarkcode/dataunusual
Abstract

The pigs were bleeding. That was the point of the experiment: could airway pressure reveal hemorrhage before it became critical? In the UCR Time Series Classification Archive, the de facto benchmark for Time Series Classification (TSC), the same recordings ask a different question: "which of 52 pigs produced this trace?". The dataset PigAirwayPressure is not an outlier. Although the archive is the standard benchmark, whether its tasks resemble the problems TSC is meant to solve has rarely been examined. We find that several datasets are repurposed from their original task, built from toy setups, or have flawed test splits. We further question how representative the archive's ordered non-temporal sequences (e.g. spectra, shape outlines) are of real TSC, and the potential downsides from the decomposition of multivariate problems. Our findings call for more careful interpretation of archive-wide scores and greater attention to deployment-relevant evaluation. The audit is at https://github.com/automl/ucr-provenance-audit.

0.37

Zero-Shot Weight-Space Programming Across a Difficulty-Graded Ladder of Two-Player Games

cs.AI ·2026-11-06 ·Eduardo C. Garrido-Merchán
Abstract

Weight-space programming sets every weight of a feed-forward network by hand, from an explicit specification of the desired computation, with no gradient descent and no training data anywhere in the construction. We ask how far a large language model can carry this practice, zero-shot, across games of deliberately increasing difficulty. We design a four-rung ladder (Nim, tic-tac-toe, two sizes of Connect Four, and a chess endgame) spanning four axes of difficulty, and build a second, independent network at each rung, using a structurally unrelated game at matched difficulty, to separate what a rung's finding says about the difficulty level from what it says about the one game used to measure it. Ten networks are built and independently verified. Six admit a complete, non-statistical proof of correctness against an independent oracle. The chess-endgame rung is exact where correctness is cheap, move legality and, in one network, an exact one-ply mate detector, and collapses where it is expensive, composing a multi-move mating plan, from 84.1-100% success at the one-ply tactics it was built to get right down to 0.57-9.13% in full adversarial play; its replicate shows this collapse is an order of magnitude less severe under a shorter required plan and exact rather than heuristic safety checks. The standard-sized rung confirms, by holding hand-set evaluation weights fixed and adding only shallow search, that the remaining gap is an architectural ceiling rather than a matter of engineering effort, and this exact collapse reproduces across two structurally unrelated games. We release all ten networks, their independent oracles, construction-cost measurements, and a cold-prompt protocol for a future cross-model comparison

0.37

Match the Geometry Before the Surrogate: Extreme Low-Budget Optimization in High Dimensions

cs.LG stat.ML ·2026-11-06 ·Michal Průšek, Adam Novozámský, Filip Šroubek
top venuebenchmarkcode/data
Abstract · NeurIPS 2026 (camera-ready). 10 pages of main text plus appendices and the NeurIPS paper checklist, 38 pages in total. Code and result files…

Low-budget molecular optimization is often framed as Bayesian optimization in a learned latent space, but standard benchmarks typically allow hundreds or thousands of evaluations, whereas a realistic wet-lab round may allow only a few dozen. We revisit this regime using a SELFIES molecular variational autoencoder (VAE) with a Gaussian prior and GuacaMol objectives at N = 30 evaluations. The main finding is not a new surrogate model, but an interaction in candidate generation. The encoder means of a reference set of molecules concentrate near a narrow spherical shell, while many high-dimensional optimizers propose candidates from boxes or box-shaped trust regions. We study a simple constraint, the on-sphere scaffold, that restricts candidate generation to this shell. We then introduce two minimal optimizers to probe what remains useful once the support is fixed. GeoWalk performs geodesic local search on the shell without a surrogate. After 30 evaluations, it shows no detected difference from a Gaussian-process (GP) surrogate on the shell, which is ahead earlier in the budget. GeoCMA adds covariance learning in a random subspace. It does not improve on GeoWalk in the molecular benchmark. On LassoBench it has a lower mean loss than GeoWalk, but full-dimensional CMA-ES remains stronger on most tasks. A factorial experiment with a single GP crosses the candidate support (shell or box) with the acquisition maximizer (a finite candidate set or a continuous optimizer). In this benchmark, shell candidates outperformed box candidates when acquisition was maximized over a candidate set, but not with continuous maximization by L-BFGS-B. These results come from a controlled study of one Gaussian-prior VAE benchmark and do not recommend VAEs for molecular optimization. The practical lesson is to control the candidate support and the acquisition maximizer before attributing a gain to the surrogate.

0.37

BrainCoT: A Multi-Task Zero-Shot Brain Signal Foundation Model with Neurometric-Anchored Chain-of-Thought Reasoning

cs.LG ·2026-11-06 ·Yiyu Gui, Mingzhi Chen, Guibo Luo, Yuchao Yang
top venue
Abstract · 37 pages. Accepted at NeurIPS 2026

Brain signals exhibit substantial multi-center heterogeneity, strong inter-subject variability, and highly diverse task settings, which makes it difficult to build models that generalize beyond a single task. While recent brain signal foundation models leverage large-scale pretraining to learn generalizable representations, they often fall short in practice because they still require target-data-specific supervised adaptation, lack multi-task zero-shot capability, and rarely provide verifiable evidence to support their predictions. In this work, we present BrainCoT, a multi-task zero-shot Brain signal foundation model with neurometric-anchored Chain-of-Thought (CoT) reasoning. BrainCoT unifies heterogeneous tasks into an instruction-driven framework so that, after pretraining, a single model can be directly applied to multiple tasks in zero-shot settings without additional adaptation. It further introduces neurometric-anchored chain-of-thought grounded in neurometric evidence for verifiable reasoning, and adopts Decision-Consistent Rationale Training to keep process learning aligned with correct predictions. Across extensive evaluations, BrainCoT achieves an average ACC of 67.41% and AUC of 74.33% in zero-shot classification without downstream training data, outperforming the strongest linear-probing baseline trained with 1% labeled downstream data by 3.85% in ACC and 5.97% in AUC.

0.37

LEMMO: A Large Electromagnetic Model leveraging Pretrained LLMs for Million-Sample Signal Understanding

cs.AI cs.SD ·2026-11-06 ·Mingkai Li, Jiajun Wu, Yihan Mao, Zilin Sang, Taosong Yu +4 more
code/data
Abstract

Large language models have expanded beyond text to vision, audio, and other modalities, but understanding raw electromagnetic (EM) signals remains challenging. Wideband recordings can contain millions of samples, whereas existing waveform-based EM models are primarily evaluated on short in-phase and quadrature (IQ) segments, leaving long-recording understanding and detailed description insufficiently explored. We introduce LEMMO, a large EM model for instruction-driven recognition and description of long raw IQ recordings. LEMMO combines a linear-attention EM encoder with anti-aliased temporal resampling to process up to eight million complex samples. Its EM-text interaction module conditions learnable queries on instructions before retrieving signal evidence. Compact queries support recognition, while structured queries with auxiliary factual supervision organize scene-level properties, signal attributes, and cross-signal relationships for detailed description. We also construct and release a unified 1.38-TB corpus of real-world and synthetic EM signals, including LEMMO-ShortQA and LEMMO-LongDesc. Compared with MERLIN-LoRA, LEMMO improves mean EM-Bench perception accuracy by 15.77 %, mean strategy-generation ROUGE-L F1 by 4.38 points, and mean LEMMO-ShortQA accuracy by 14.31 %. It also improves categorical and numerical fact F1 on both LEMMO-LongDesc tasks while preserving its language backbone's MMLU performance. Our code and datasets are available at https://github.com/N1rv4n4/LEMMO and https://huggingface.co/datasets/DaftJun/RFLAS, respectively.

0.37

Evaluation of a Multi-Link Machine Learning Approach for Rainfall Detection Using Satellite Microwave Links

cs.AI cs.LGeess.SP ·2026-11-06 ·Mostafa Ftouni, Margherita Spalla, Matteo Colli, Andrea Randazzo, Christian Gianoglio
published
Abstract

The impacts of climate change have increased the need for reliable real-time rainfall monitoring systems. This study investigates rainfall detection using a network of opportunistic sensors based on satellite microwave links (SMLs) combined with machine learning (ML) classifiers. Two models, a random forest (RF) and a convolutional neural network (CNN), were trained on aggregated SML data, including information from neighboring links. The results demonstrate that leveraging information from multiple links substantially enhances the detection of both wet and dry periods compared to single-link approaches. Moreover, the RF model consistently provides a more stable and reliable trade-off than CNN, improving rainfall detection and reducing misclassifications overall, with particularly strong performance during harsh weather conditions. The study spanned up to 32 rainy days between 2021 and 2024.

0.37

Tabular Foundation Models vs Tuned Gradient Boosting for Telecom Churn: A Profit- and Calibration-Aware Evaluation

cs.AI cs.LG ·2026-11-06 ·Ismail Hossain Polas
benchmarkcode/data
Abstract · 17 pages, 2 figures, 14 tables. Code, per-split predictions and reproducibility audit: https://github.com/ihpolash/tfm-churn-profit-audit (a…

Tabular foundation models (TFMs) predict from a labelled context in a single forward pass. A recent benchmark reported that they decisively outperform tree ensembles for customer churn, but it compared them with untuned gradient-boosted trees (GBDTs) and scored only ranking metrics, while a retention campaign is judged in money. We re-evaluate TabPFN-2.5, TabPFN-3.5, TabPFN-3.5-Fast and TabICL v2 against Optuna-tuned GBDTs on five public telecom churn datasets over 10 seeded splits. We separate three questions: whether a model ranks churners well, whether that ranking earns more expected maximum profit (EMP), and whether its probabilities can drive a threshold policy. With a split-aware test and Holm correction over 20 pre-declared comparisons, TabPFN-3.5 earns significantly more EMP than the best tuned GBDT on Cell2Cell (1.9%) and Iranian (7.6%), and these two gains also hold against each of the three tuned GBDTs separately. On Orange it is ahead in all 10 splits (14.3%), but the gain is not significant after correction (Holm p = 0.155). The splits resample the same datasets, so these results do not stand for independent operators. TabICL v2, the earlier benchmark's leader, beats the best tuned GBDT only on Iranian, not all three of them, and underpredicts Orange churn by 2.1 points on a 7.3% base rate. On the datasets and splits where we tested them, a substantially larger tuning budget and a duplicate-aware split leave the main conclusions unchanged. Retuning the baseline cuts the earlier headline gap from 6.1 to 4.0 PR-AUC points without closing it. The licence of the TabPFN-3.5 weights we used (License v1.0), not scoring cost, is what stands between its gains and production use.

0.36

Train on prefill, monitor at decode: Using sparse autoencoders for topic monitoring

cs.AI cs.LG ·2026-11-06 ·Rishikesh Mallagundla, Pranav Patel
code/data
Abstract · 22 pages

Topic monitoring is the task of labelling the topic of each token in a model's response during generation. Activation-based monitors that run at generation time have so far been trained to detect harmful content and are limited to two or three classes. We introduce TopicScope, a framework for training causal topic classifiers on SAE features that assign a topic to every token, so that the label can change partway through a response. We train three detector architectures on synthetic multi-topic conversations covering seven topics. Across Gemma-3 1B and 4B, monitors trained on features extracted from prefill hidden states were able to generalise to decoding. We also found that the classifiers are robust to perturbations, losing at most 5 macro-F1 points when off-topic asides are inserted, topics are removed or sections are reordered. Ablations conducted to understand the contribution of the SAE revealed that classifiers reaching baseline performance without it were several times larger. Our findings show that compact topic monitors can be trained inexpensively in prefill and deployed during generation. More broadly, our work shows where SAEs are useful beyond safety classification. Our code, dataset and checkpoints are available at https://github.com/Mercity-AI/SAE-Guardrails and https://huggingface.co/collections/Mercity/online-guardrails-using-sparse-autoencoders-6ac602940468c62fefb2b9e8

0.35

SIRIS: Can LLMs Tell When Simulation Errors Matter in AI-Driven Systems Research

cs.AI ·2026-11-06 ·Jinwoo Hwang, Sukmin Cho, Tao Zhang, Soheil Abbasloo, Jongse Park +1 more
Abstract · 21 pages, including references and appendices

LLM agents increasingly drive systems research and rely on simulators to evaluate candidate designs at scale. Simulators are imperfect by design: their modeling limitations can distort predictions yet leave the recommended design correct, or change which design should be selected. We introduce Siris to test whether LLMs can tell these cases apart, judging from their systems understanding when a limitation invalidates the simulator's recommendation. By perturbing a simulator, Siris constructs workload pairs with the same candidates, limitation, and simulator recommendation. The perturbation changes predictions on both workloads, yet a reference simulator supports the recommendation on only one. Evaluating thirteen frontier LLMs on LLM serving design problems, we find that identifying a limitation does not ensure a correct choice, and even the strongest models sometimes abandon valid recommendations. Self-correction through a review turn helps these models restore valid choices but introduces new errors in several others. These findings motivate evaluating systems understanding through design decisions under imperfect simulators.

0.35

META-IR: Meta-Introspective Reasoning for Adaptive Safety Alignment in LLM

cs.AI ·2026-11-06 ·Mahabubur Rahman Miraj, Hongyu Huang, Nankun Mu, Chao Chen, Shanqin Yi
publishedcode/data
Abstract · 6 pages, 5 figures. Accepted at the 2026 IEEE International Conference on Multimedia & Expo (ICME 2026), Bangkok, Thailand. Source code: htt…

Large Language Models (LLMs) remain vulnerable to jailbreak attacks and harmful-content prompts, particularly under adversarial distribution shifts where static alignment pipelines fail to generalize. Although introspective methods such as STAIR strengthen refusal consistency, they depend on fixed self-evaluation policies that cannot rapidly adapt to evolving manipulation strategies. To overcome these limitations, this paper introduces META-IR, a meta-introspective reasoning framework that enhances safety alignment through meta-learning-driven adaptation. It employs a dual-loop architecture: an inner introspective loop that performs structured risk analysis, self-assessment, and reflective correction via SI-MCTS, and an outer meta-learning loop that optimizes the introspection policy across diverse jailbreak task families to enable rapid generalization to unseen attacks. A dynamic jailbreak simulation module further exposes the model to progressively harder adversarial prompts, enabling continual self-improvement without human supervision. Experiments across multiple models (Llama-3.1-8B-Instruct and Qwen-2-7B-Instruct) and a broad suite of safety, helpfulness, generalization, and introspection metrics demonstrate that META-IR substantially improves adaptation speed, robustness, and reasoning stability compared with all baseline models. These results establish META-IR as an effective pathway toward building LLMs capable of sustained, adaptive safety alignment under dynamic and evolving threat environments.

0.35

DexDICE: Reconciling Safety and Performance via Dual-Expert Offline Imitation Learning

cs.LG cs.AI ·2026-11-06 ·Seokin Seo, Sung Kuk Shyn, Minji Seo, Yisoo Lee, Jongmin Lee +2 more
publishedbenchmarktheorycode/data
Abstract · NeurIPS 2026. 55 pages, 13 figures. Code: https://github.com/tzs930/dexdice

Ensuring high returns while providing strong safety guarantees in offline imitation learning (IL) is a fundamental challenge: existing approaches either rely on explicitly specified cost functions, which are difficult to design for complex real-world safety constraints, or require expert demonstrations that are both safe and high-performing, which are often too costly to collect. We introduce dual-expert imitation learning, which leverages two complementary data sources: (i) safety-compliant demonstrations that faithfully follow safety guidelines but may be suboptimal, and (ii) performance-oriented demonstrations that achieve high returns but may violate safety guidelines. To address this problem, we propose DexDICE (Dual-expert offline imitation learning via stationary DIstribution Correction Estimation). Instead of conventional distribution constraints, DexDICE combines DICE-style offline learning with safe support constraints, enabling agents to exploit high-return behaviors while remaining within safety-compliant regions. We prove that the ideal solution satisfying this constraint exactly has zero unsafe-region visitation with high probability. We evaluate DexDICE on a re-curated DSRL benchmark and validate it on a real-world mobile-robot navigation task trained directly from human teleoperation demonstrations, where it substantially outperforms baselines and remains robust to data scarcity and quality degradation.

0.34

Rethinking Optimization Methods for Hierarchical Feature Selection

cs.AI ·2026-11-06 ·Xinyuan Song, Zekun Cai, Liang Zhao
code/data
Abstract · Under review

Rare-event forecasting often depends on interpretable conjunctions: a single keyword, location, or temporal cue may be weak, while their joint occurrence can identify a meaningful precursor. Hierarchical-Task Numerical conjunctive feature Learning (HTNL) selects Numerical Conjunctive Features (NCFs) through a structured sparsity objective. However, the resulting optimization problem is challenging, making optimization design critical to its practical performance. We give a unified diagnosis of four HTNL optimization families: original majorization-minimization (MM), variational quadratic form (VQF), alternating direction method of multipliers (ADMM), and weighted group lasso (WGL). Our analysis distinguishes three key factors that jointly determine HTNL optimization behavior. First, VQF and MM are algebraically equivalent under matched variational updates and optimize a squared-penalty formulation. Second, ADMM optimizes the unsquared objective through consensus splitting; we replace its truncated FISTA proximal solver with a scalar bisection update (ADMM-Bisect) for more accurate treatment of near-threshold blocks. Third, WGL targets the same unsquared objective, while all methods face rapidly decaying active-set admission scores. Across two synthetic datasets and 11 civil-unrest and flu-surveillance tasks, VQF achieves the highest mean AUC with the most compact non-collapsing supports. ADMM-Bisect reduces measured runtime, while calibrated WGL avoids the empty-support failures of its original scale. Synthetic results further show how proximal accuracy, pruning, and search budget affect candidate recovery, especially for longer conjunctions. Code will be made publicly available upon publication.

0.34

SIMBAD: Spatio-Temporal Traffic Forecasting Robust to Aperiodicity

cs.AI ·2026-11-06 ·Daniel Y. Lee, Seungwon Shin, Seunghoon Han, Sungsu Lim, Susik Yoon
top venuebenchmark
Abstract · Total of 30 pages, 9 figures (4 in main, 5 in appendix), accepted to NeurIPS(Neural Information Processing Systems) 2026

Accurate traffic prediction is essential for urban planning but remains challenging due to the irregular patterns caused by unexpected events. While Spatio-Temporal Graph Neural Networks (STGNNs) effectively model periodic data, they often struggle with these aperiodic fluctuations. To address this, we propose SIMBAD, which improves generalization irregular patterns via two key mechanisms: 1) adaptive regulation of periodic signals to prioritize recent signals, and 2) dynamic spatial modeling that adjusts influence between connected nodes based on contextual relations. Built upon simple similarity metrics to adapt to aperiodic patterns, SIMBAD outperforms existing methods on real-world benchmark datasets, demonstrating superior performance in predicting unseen and irregular traffic events.

0.34

Rethinking MoE Load Balancing as Feedback Control

cs.LG ·2026-11-06 ·Yikun Hou, Alp Yurtsever
Abstract

Mixture-of-Experts (MoE) training is often not stationary: the data mixture changes at curriculum boundaries, when a model is continually pretrained into a new domain, or when the corpus is refreshed. Routing is learned against the current mixture, so each such change disrupts the per-expert load the router had converged to, and existing balancing methods have only been evaluated on stationary data. We rethink load balancing as a feedback control problem, in which the deviation of each expert's load from its fair share is the control error and the per-expert routing bias is the control input, and define a family of controllers, the PID-Balancer, from the proportional, integral, and derivative terms of classical PID control. Two widely used methods belong to this family: the auxiliary balancing loss behaves like a proportional controller, and the loss-free bias update is an integral controller acting on the sign of the error. Reading the router as a static map with a one-step delay, we show that the integral term is the only one that both removes persistent imbalance and averages out the step-to-step fluctuation of the measured load, whereas the proportional and derivative terms pass or amplify that fluctuation into the routing bias. We therefore adopt PID (I), the pure-integral member of the family with a single gain. Two settings are evaluated: (i) training an OLMoE backbone from scratch with 16 to 64 experts per layer, through a mid-run mixture switch, and (ii) continual pretraining of the released OLMoE-1B-7B on a new corpus. PID (I) holds balance before the shift and restores it more tightly afterwards than either the auxiliary loss or loss-free balancing, at comparable perplexity.

0.34

Learning Across Heterogeneous Feature Spaces with Group Robustness

cs.LG ·2026-11-06 ·Xenia Konti, Arjun Tschand, Chuan Hong, Nicoleta J. Economou-Zavlanos, Michael M. Zavlanos
hot direction
Abstract · 29 pages, 5 figures

In clinical applications, patients undergo different combinations of tests, and machine learning models must therefore ensure accurate predictions across heterogeneous feature spaces. Current approaches upweight the patients with the highest loss, without accounting for the limited predictive information in their features preventing their loss from going any lower. Moreover, when patients' feature spaces are fully disjoint, current methods reduce to training independent models, preventing knowledge transfer across them. We propose a framework for group robust learning across heterogeneous feature spaces, where groups are patients sharing a feature space. Since a high loss may simply reflect uninformative features, we instead optimize for worst-group excess loss, the gap between a group's loss and an estimate of its achievable optimum. To allow collaboration between groups, we use group-specific encoders mapping into a shared latent space in which learnable Gaussian anchors cluster patients by outcome. On tabular and imaging medical datasets, including fully disjoint feature settings, our framework outperforms state-of-the-art methods when groups differ in their optimal losses or cannot train a good model on their own, reducing worst-group loss by up to 79% and excess loss by up to 95% with up to 18$\times$ fewer parameters.

0.34

Projected Neural Additive Models as Universal Approximators

cs.LG ·2026-11-06 ·Nhon N. Phan, Qiang Du, John D. Clayton, WaiChing Sun
publishedtheory
Abstract · 40th Conference on Neural Information Processing Systems (NeurIPS 2026)

Neural additive models (NAMs) offer interpretability but lack expressivity due to their rigid additive structure, whereas multi-layer perceptrons (MLPs) achieve universal approximation at the cost of transparency. We establish the first theoretical foundation for projected neural additive models (PNAMs), which augment NAMs with a learnable linear transformation $\boldsymbol{T}$ to capture cross-variable interactions. While standard NAMs cannot approximate even the simplest interaction term $χ_1 χ_2$ because their cross-derivatives are identically zero, we prove that PNAMs are universal approximators and characterize the minimum number of projections required for exact polynomial representation. Furthermore, we derive an approximation rate of $\mathcal{O}(M^{-s / (N - 1)})$ for $\mathcal{C}^s$ functions (i.e., functions with $s$ continuous derivatives), where $M$ denotes the number of projections and $N$ the input dimension; this rate is known to be optimal for ridge function approximation. To recover interpretability from the projected inputs, we introduce structured sparsity regularizers and post-hoc symbolic regression techniques that enable global input ranking, parameter pruning, and the conversion of learned bases into compact analytical expressions. Experiments on knot theory invariants, MNIST, and phase field fracture mechanics demonstrate that PNAMs match or surpass similar architectures in accuracy while discovering salient features and yielding symbolic models with orders of magnitude fewer parameters.

0.34

RoleWalk: training-free structural embeddings for directed graphs that transfer across graphs

cs.LG cs.SI ·2026-11-06 ·Maixent Chenebaux
benchmarkcode/data
Abstract · 14 pages, 3 figures. Complete technical report (20 pages) as an ancillary file. Code: https://github.com/kerighan/rolewalk (pip install role…

Structural role embeddings describe a node by its position in a graph - hub, bridge, periphery, member of a dense core - rather than by who its neighbours are. Existing methods either learn coordinates specific to one graph (struc2vec, Role2Vec, RESD) or read scales and normalisations from the whole graph (GraphWave, Digraphwave), so the same neighbourhood can receive different coordinates in two graphs; few read edge directions. We present RoleWalk, a training-free embedding that follows a short lazy random walk from each node and records, at each step, two distributions: that of the transition probabilities, through its characteristic function at a fixed grid of frequencies, and that of the walker's distance to its starting node. On directed graphs, walks that go forward then backward add who shares a node's targets or sources. The embedding depends only on a node's neighbourhood within a fixed radius, so its coordinates mean the same thing in every graph. We show that, without curvature, it is the kernel mean embedding of the walk's footprint, that it separates graphs that the Weisfeiler-Leman test cannot, and that pruning small probabilities comes with a per-node error certificate. Empirically, RoleWalk is competitive with the best role embeddings on five classification benchmarks at a fraction of their cost, has the highest scores on real directed roles in Wikipedia adminship, exceeds the published accuracy of Digraphwave at every noise level of its own Enron network-alignment benchmark, transfers centrality predictors to graphs four times larger better than existing inductive embeddings, hand-made features and a trained GIN, and embeds a million nodes in 20 s to 3 min on one CPU core. The code is open source (BSD license): https://github.com/kerighan/rolewalk

0.34

Ordered Policy Optimization

cs.LG cs.AIcs.ROstat.ML ·2026-11-06 ·Dongkyu Derek Cho, Pan Xu
top venue
Abstract · 36 pages, 12 figures, 11 tables. Published in Proc. of the 40th Conference on Advances in Neural Information Processing Systems (NeurIPS 202…

Proximal Policy Optimization (PPO) stabilizes policy updates by clipping the importance ratio between the new policy and the data-collecting policy. The standard rule uses one fixed clipping range for all samples. While this choice offers simplicity and effectiveness, it fails to distinguish statistically harmful extreme ratio values from large, yet tolerable changes. We study how to adapt the clipping radius to the reliability of the induced importance ratio distribution. Our case study demonstrates that reliable updates require controlling these extreme ratios, rather than merely verifying that individual ratios fall within a fixed interval. Motivated by this observation, we propose Ordered Policy Optimization (\textsc{OPO}). In each minibatch, \textsc{OPO} sorts the importance ratios, selects the extreme ratios that potentially have the largest effect on the update, and assigns them adaptive sample-wise radii instead of using the fixed clipping radius. These radii are derived from a prescribed tail profile, which specifies desired target values for the selected extreme ratios. The resulting objective remains first-order and requires only minibatch sorting and a modified policy loss. Experiments conducted on MuJoCo and Atari benchmarks demonstrate that this simple modification achieves strong performance relative to established baselines.

0.34

KnowEvo: Knowledge Evolution for Protein Optimization

cs.LG ·2026-11-06 ·Zijie Xing, Xingyue Liu, Runze Wang, Luoming Hu, Yanming Shen
top venuecode/data
Abstract · Accepted by NeurIPS 2026

Large language models (LLMs) have shown promise in scientific discovery, yet systematically accumulating and refining reusable scientific knowledge remains challenging. Existing approaches primarily improve LLM-based problem solving through parameter optimization or feedback-driven refinement of prompts, programs, and solution trajectories, leaving acquired expertise either embedded in model weights or scattered across iterative problem-solving processes. As a result,knowledge is difficult to systematically inspect, evaluate, and refine. We propose KnowEvo, a framework that treats externalized knowledge itself as the optimization target. Our key insight is that scientific knowledge is inherently modular: multiple interacting mechanisms jointly determine the design outcomes. This makes knowledge evolution fundamentally challenging, as improvements cannot be attributed to individual components in isolation. To address this, KnowEvo represents knowledge as executable expertise blocks and organizes their refinements into strategy mutation trees, enabling structured exploration over a combinatorial knowledge space. The framework combines a Thompson-style forward allocation strategy that adaptively distributes evaluation budgets across strategy variants with a backward revision process, in which a code agent operates in an isolated sandbox to translate rollout evidence into localized program updates. This design enables persistent accumulation, evaluation, and refinement of mechanistic expertise across iterative experimentation. We evaluate KnowEvo on protein optimization tasks including stability and solubility. Results show that knowledge evolution consistently improves over initial strategy blocks and outperforms monolithic program evolution and prior LLM-guided evolution baselines across metrics such as pass@k, MRR, MAP,and NDCG@5. Code is available at https://github.com/recynie/KnowEvo.

0.33

Diffusion-free Score Matching for Exact Conditional Sampling

cs.LG ·2026-11-06 ·Thomas Wedenig, Robert Peharz
Abstract

The success of modern score-based generative models is largely driven by diffusion-based training, which learns a collection of time-indexed score fields corresponding to progressively noisier versions of the data distribution. While such models achieve impressive sample quality, leveraging them for sound probabilistic inference---particularly sampling from arbitrary conditionals of the model---remains challenging. This difficulty arises because conditioning information is only available for clean data and not for the higher noise levels required to generate exact conditional samples. In this paper, we introduce DIffusion-free SCOre matching (DISCO), a principled approach that sidesteps the need for time-dependent score fields altogether. Our method is based on an objective that, while reminiscent of diffusion-based training, estimates only a single score of the (slightly perturbed) data distribution, for which the exact conditional score is readily obtained. In experiments, DISCO is competitive with state-of-the-art diffusion models in terms of sample quality. More importantly, DISCO yields a more faithful representation of the underlying data distribution and---crucially---enables accurate sampling from arbitrary conditionals, outperforming standard heuristic samplers. This capability opens the door to principled and flexible probabilistic reasoning with score-based models.

0.33

Not Your Diary, Just Your Topic: Grading the Wrong-User Control

cs.AI ·2026-11-06 ·Karina Romanova, Igor Kiselev, Vladimir Platonov, Viliana Devbunova, Gregory Leleytner +3 more
published
Abstract · 21 pages. NeurIPS 2026 Workshop on Personalized, Aligned, Long-Term Memory for AI Systems (PALM)

A common test of personalization replaces a user's history with someone else's and measures the drop in accuracy. But this swap also changes the topics in the history, so the drop may reflect topic mismatch. We examine this problem with a frozen language model on LaMP-2 news categorization, using histories from similar users and synthetic histories assembled from other users' articles. The user's own history improves accuracy by 2.3 percentage points over a model trained without history. Removing history only at evaluation makes this gain appear larger, at 4.8 points. Category labels alone reproduce 86% of the accuracy gap between the user's own history and a random user's history. With synthetic histories matched on labels and text similarity, the gap falls to 0.9 points. Across two user splits and three training seeds, the advantage over a similar user's history is 3.2 points for users seen during training and 0.65 points for unseen users, where the evidence for an advantage remains inconclusive. A small movie-tagging experiment with one training seed also finds no advantage over other users' histories. These findings show why personalization should be evaluated on unseen users, with histories of similar content and a model trained without history as comparisons. We also find that adding gated cross-attention improves on a prefix-only model under our training setup; whether this holds more broadly remains open.

0.33

Geometry-Aware Score-Repellent Monte Carlo

cs.LG stat.ML ·2026-11-06 ·Jie Hu, Lingyun Chen, Do Young Eun
top venuetheory
Abstract · To appear at NeurIPS 2026

The recently proposed Score-Repellent Monte Carlo (SRMC) method improves MCMC sampling by tilting the target density $π$ in $\mathbb{R}^d$ with the factor $\exp(-α\,θ_n^\top s(x))$, where the score function $s(x) = \nabla_x \log π(x)$, and $θ_n$ is the score average from past samples. However, a single scalar repellence strength $α\ge0$ cannot handle both the steep and the flat directions in anisotropic energy landscapes at once. We introduce \emph{Geometry-Aware SRMC} (GA-SRMC), which generalizes the Euclidean alignment to $θ^\top M s(x)$ for any symmetric positive-definite matrix $M$. A stochastic-approximation central limit theorem under a trace constraint on $M$ identifies the trace-normalized inverse Fisher $M_{\rm opt}\propto S^{-1}$ as the unique minimizer of the worst-direction covariance bound, elevating the Fisher information $S=\text{Cov}_π(s,s)$ from a proposal-side preconditioner (e.g., FisherMALA, natural gradient) to the bound-optimal repulsion-side matrix, with a single online estimate $\widehat S_n$ serving both roles at no additional computational cost. The full-Fisher online metric $M_n$ converges almost surely to $M_{\rm opt}$ and satisfies a CLT, providing a theoretical guarantee for the online implementation of the GA-SRMC framework. On an $8$-target simulation spanning the condition number $κ\in[1.3,100]$ and dimension $d\in[2,785]$, GA-SRMC dominates SRMC in ESS by $\boldsymbol{4.3\times}$ on anisotropic Gaussian ($κ=100$) and $\boldsymbol{6.8\times}$ on MNIST under the MALA baseline, by $\boldsymbol{16\text{--}32\%}$ on three well-conditioned logistic-regression posteriors, and by up to $\boldsymbol{+48\%}$ under FisherMALA; gains peak on the most anisotropic targets, and posterior-mean MSE is lower than with SRMC in all $16$ full-Fisher implementations.

0.33

Forecasting cumulative intraoperative hypotension burden: patient-disjoint evaluation of quantile boosting and alert policies

cs.AI ·2026-11-06 ·Yunfan Zhao
benchmark
Abstract

Background: Binary hypotension forecasts do not describe the depth and duration of future low pressure. Burden forecasts may support graded early warning, but better forecast scores need not yield more useful alerts. Objective: To evaluate conditional quantiles of the next 15-minute mean arterial pressure (MAP) deficit below 65 mmHg and distinguish forecasting performance from episode-warning performance. Methods: Public VitalDB numeric arterial MAP was reduced to 10-second medians without forward filling. Subject-disjoint training, validation and test splits included 70,402, 14,270 and 15,575 labeled windows. We compared climatology, calibrated persistence, last-MAP-only boosting, and direct and residual quantile boosting, with and without additional MAP-history features. Patient-weighted pinball loss and subject-bootstrap intervals evaluated forecasting. Validation selected alert thresholds before a hash-recorded test evaluation. Results: In the test split, the prespecified residual-versus-persistence comparison at the 90th percentile yielded losses of 2.256 and 2.900 mmHg-min, respectively (difference -0.644; 95% interval -0.814 to -0.487). Direct prediction achieved 2.196, and direct prediction with additional MAP features achieved 2.193. Episode-warning results were sensitive to the alert rule and operating threshold. Conclusions: This retrospective benchmark separates burden-forecast skill from warning-policy performance. Findings support further study of direct burden estimation; they do not establish patient benefit, prospective reliability or transportability beyond the source hospital.

0.33

fNIRSAtlas: A Large-Scale Benchmark for Functional Near-Infrared Spectroscopy Classification

cs.LG ·2026-11-06 ·Kalle Mäkelä, Tuukka Ruotsalo
top venuebenchmarkcode/data
Abstract · Accepted to NeurIPS 2026 (Evaluations & Datasets Track)

Functional near-infrared spectroscopy (fNIRS) is a promising modality for brain-computer interfaces and cognitive state decoding, but progress in machine learning for fNIRS has been limited by small, isolated datasets and inconsistent evaluation protocols. We introduce fNIRSAtlas, the largest open benchmark for fNIRS classification to date, comprising 23 classification tasks from 19 public data collections, spanning diverse experimental paradigms, including motor execution, motor imagery, mental arithmetic, working memory, and emotion recognition. fNIRSAtlas provides a fully automated pipeline from raw data to evaluation and defines standardized benchmark tasks for both cross-subject and within-subject classification, ensuring reproducible comparison of methods. We evaluate a variety of baseline models from fNIRS and related neuroimaging and time-series machine learning literature on the benchmark. Our results demonstrate that performance varies substantially across tasks and that methods validated on limited data -- especially deep learning approaches -- often fail to generalize beyond their original evaluation. fNIRSAtlas serves as an extensible, open, and reproducible benchmark to support more robust and cumulative progress in machine learning for fNIRS data. Our code and data can be found at https://fnirsatlas.github.io.

0.32

Recursively Self-Improving Electromagnetic Spectrum Agents: Architecture and Operation

cs.MA ·2026-11-06 ·Baolin Chong, Dongfang Xu, Qiaosi Wang, Fuhui Zhou, Qihui Wu
Abstract

Electromagnetic spectrum agents (ESAs) must contend with changing environments and task requirements during sustained autonomous operation, yet their existing capabilities may not remain suitable for new operating conditions. This article proposes recursively self-improving electromagnetic spectrum agents (RSI-ESAs) that use operational feedback to update their capabilities and build on prior improvements. We present their architecture and editable components, and describe a self-improvement process encompassing problem diagnosis, candidate generation, validation, and inheritance. By retaining effective task versions and using modification records and diagnostic knowledge to guide subsequent updates, RSI-ESAs can accumulate task capabilities and improvement experience across successive rounds. A shared-spectrum scheduling case study illustrates improved task performance and capability accumulation over multiple improvement rounds under the tested conditions. Finally, we discuss open challenges for broader tasks and long-term operation, outlining research directions for the continued capability development of ESAs.

0.32

Frame: Form-Restricted Authoring with Model-Based Evaluation

cs.AI ·2026-11-06 ·Nan Jiang, Mohammed Mohsin Hussain, Onkar Tiwari, Ankit Jain, Thong Q. Nguyen +2 more
benchmarkcode/data
Abstract

Many professional reports reuse the structure of prior editions while updating the content. We formalize this setting as form-restricted report generation: given a sample report and new grounding documents, generate a report that preserves the sample's form while accurately reflecting the new evidence. We introduce FRAME, a benchmark of 120 EDGAR- and arXiv-derived instances across Financial Quarterly, Financial Annual, Academic Simple, and Academic Complex categories, built with a temporal three-snapshot pipeline. We evaluate outputs with three human-validated metrics: structural similarity via semantic tree-edit distance, coverage via information unit recall, and faithfulness via claim-level grounding verification. Across five LLMs, open-source models nearly match proprietary models on structure but lag on coverage and faithfulness; all models perform better on financial than academic reports, and performance declines with longer outputs and more complex grounding.

0.32

The Infrastructure Divide: Analyzing Community Stance on Data Center Development via Reddit

cs.AI ·2026-11-06 ·Yoko Koyama, Gongfan Chen, Mansi Saxena
code/data
Abstract · 42 pages, 7 figures

The rapid expansion of AI-driven data center construction has outpaced public engagement: an estimated $64 billion in U.S. data center projects were blocked or delayed by local opposition between May 2024 and March 2025, yet the concerns driving this resistance remain poorly characterized. Prior engagement methods such as surveys and hearings are introduced late in the planning process, and no study has systematically mined social media to measure community stance at scale. To address this gap, this study presents an LLM-powered opinion-mining framework and applies it to 798 Reddit posts and 102,328 associated comments spanning a decade. An open-source large language model, guided by a Chain-of-Stance prompting framework and three domain-specific labeling rules, classified stance at 83% of the human inter-reviewer agreement ceiling. Opposition dominated discourse volume rather than disapproval: against posts outnumbered favor posts 2.6:1 and generated 8.5 times more comments, and this gap widened significantly over 86 weeks. Topic modeling of opposition arguments resolved into four reproducible clusters, resource consumption and scarcity (42.8%), contested siting processes and secrecy (34.2%), political opposition to data centers (13%), and political and corporate accountability (8%). Engagement analysis revealed that posts with emotional or provocative political framing attracted the most engagement, while key technical concerns (e.g., water, power) were dispersed across lower-engagement comment threads. These findings give local leaders, developers, and policy makers a quantified, continuously updatable basis for prioritizing concerns and policy levers, supporting a shift from opposition to negotiated, verifiable partnership for data center development.

0.32

Discovering formally verified quantum programs with AI

cs.AI quant-phcs.PL ·2026-11-06 ·Sergey Shirobokov, Ben Jaderberg, Katerina Hristova, George Tsoukalas, Johannes Bausch +20 more
theory
Abstract

The timeline by which quantum computers can solve breakthrough problems in science depends on improved quantum algorithms as much as on better hardware. Such algorithmic advances are typically achieved through many years of expert human collaboration. In mathematics and computer science, large language models have accelerated progress through automated discovery loops. However, quantum algorithms can only be evaluated by classical computers on restricted or small instances. Here we demonstrate automated discovery of quantum programs beyond these regimes including order-of-magnitude improvements on scientific applications. To achieve this, we build a quantum program library in the Lean theorem prover in which Gemini 4 agents synthesize novel programs against formal problem specifications and verify their correctness and cost. Agent teams collaborate, sharing results and knowledge through wikis and discussing ideas at organized conferences. Our results improve on prior baselines for parallel search (up to 2.55x), quantum arithmetic (290x), and simulations relevant to fusion, photochemistry and drug-metabolism by 11x, 8x and 2.5x respectively. Combining human and artificial intelligence at scale, under the rigor of mathematical proof, lets researchers design algorithms that today's classical and quantum computers cannot test, shortening the path to practical quantum computation.

0.31

TurtLES: A Large-Scale Benchmark for Turbulent 3D Neural PDE Surrogates

cs.LG physics.flu-dyn ·2026-11-06 ·Jarno Platenburg, Armand Kassaï Koupaï, Paola Cinnella, Patrick Gallinari
benchmarkdataset
Abstract

Machine-learning surrogates are increasingly used to accelerate computational fluid dynamics, yet progress is limited by the lack of benchmarks capturing realistic, time-dependent turbulent flows. We introduce TurtLES, a 14 TB dataset of 3D turbulent cylinder wakes in the shear-layer transition regime, computed via high-fidelity implicit large-eddy simulations. The dataset contains 400 trajectories of 400 time steps each, with 3-11 million points per frame, spanning variations in geometry, Reynolds number, and angle of attack. Unlike existing datasets, TurtLES combines (i) physically realistic turbulence with an active energy cascade driven by vortex stretching, (ii) dense irregular meshes, and (iii) long-horizon temporal dynamics. We benchmark state-of-the-art neural operators on autoregressive rollouts and identify key failure modes, including temporal instability, loss of small-scale energy, and challenges in learning on large irregular domains. TurtLES provides a new benchmark for evaluating surrogate models on 3D turbulent wakes at industrially relevant mesh resolution.

0.31

FedLoVA: Value-Only Aggregation for Federated LoRA Fine-Tuning of Large Language Models

cs.LG ·2026-11-06 ·Ensieh Khazaei, Baturalp Buyukates, Dimitrios Hatzinakos
code/data
Abstract

Low-Rank Adaptation (LoRA) is a popular parameter-efficient fine-tuning (PEFT) method for fine-tuning of large language models (LLMs), especially in federated learning, due to its strong performance and communication efficiency. In practice, LoRA is applied to the query and value projections of transformer layers without considering the distinct roles of these components. In this work, we analyze the impact of these projections in federated fine-tuning of LLMs using LoRA and observe that the value projection converges significantly faster than the query projection. Motivated by this finding, we propose FedLoVA, a federated fine-tuning framework that trains LoRA adapters on both query and value projections locally while sharing only the value projection with the server. This approach reduces communication overhead while improving convergence efficiency. Experiments on natural language understanding, multilingual sentiment analysis, and generation tasks show that FedLoVA reduces the communication cost by up to $7.8\times$ compared to existing federated LLM fine-tuning baselines while achieving comparable or better performance. The code is available at: https://github.com/EnsiehKhazaei/FedLoVA.

0.30

Ensemble Routing for Prompt-Adaptive Cost-Aware LLM Inference

cs.AI ·2026-11-06 ·Saul Urso, Emanuele Carlini, Patrizio Dazzi, Matteo Mordacchini
Abstract

Large language models (LLMs) increasingly support Web-based services that must balance response quality and inference cost across heterogeneous requests. Conventional routing selects one model per prompt, while fixed ensembles cannot adapt their composition and cost to individual requests. We introduce PISA, a prompt-adaptive, cost-aware single-model router, and PISA-E, its extension to ensemble routing. At the core of both methods, PISA combines soft cluster memberships with model-specific predictors to estimate individual model capabilities. Building on these estimates, PISA-E jointly selects ensemble composition and cardinality under an explicit cost--quality preference that can be adjusted at inference time without retraining. A tractable ensemble-success surrogate and efficient candidate selection enable this extension without ensemble-specific training. Across four LLM routing benchmarks under in-distribution (ID) and out-of-distribution (OOD) conditions, PISA achieves the highest macro-average cost--performance hypervolume among the evaluated single-model routers in both settings. Meanwhile, PISA-E achieves the highest macro-average task performance among the evaluated ensemble approaches when prioritizing response quality, surpassing the strongest competitors by 2.09 and 2.87 percentage points under ID and OOD conditions, respectively. At selected lower-cost operating points, it also exceeds the strongest ensemble competitors' macro-average task performance while reducing average inference cost by approximately 36% and 25%, respectively. These results demonstrate the effectiveness of prompt-adaptive capability estimation for both single-model routing and selective ensembling.

0.30

Tracing Psychometric Inference in Large Language Models

cs.AI cs.CL ·2026-11-06 ·Chih-Hao Hsu, Feng-Chun B. Chou, Pin-Hao A. Chen
top venue
Abstract · Accepted at NeurIPS 2026

Do large language models represent psychological constructs internally, or do they only generate outputs that resemble psychometric structure? We address this question using a cross-persona paradigm across 14 model variants from two families (Llama 3 and Qwen 2.5, 0.5B-14B, base and instruct). Given responses on one psychological scale from real individuals (N = 272 Taiwanese university students), models predict responses on six other scales. At the behavioral level, larger variants reproduce and amplify human cross-scale correlations; amplification persists after correcting the seven-scale human correlations for measurement attenuation. We then examine whether this structure is reflected internally. Contrastive direction analysis finds geometry aligned with the human matrix in larger instruct variants, but little alignment in base variants under this assay. Across models, geometry strength is associated with behavioral amplification (r = 0.68, p = 0.008; partial r = 0.69 controlling for log size). To relate internal structure to output behavior, a matched activation-probing paradigm shows that probe slopes vary less than behavioral slopes (1.38-1.77 vs 0.52-1.53). A synthetic control shows that ridge probing can inflate these slopes; after subtracting a baseline based on a reliability proxy from another analysis, adjusted slopes remain associated with behavior (r = 0.88). We term the measured probe-behavior gap readout attenuation. These results establish decodability and geometric alignment under the tested assays, without showing that the measured directions mediate the cross-persona answers.

0.30

Maya a1: Do the Options Read the Model, or Does Their Position? Order Sensitivity in a Non-Autoregressive Decision Model

cs.AI cs.CL ·2026-11-06 ·Mohammed Nayeem Khan, Logesh L
Abstract · 24 pages, 13 figures, 23 tables

Laya answers typed multiple-choice questions in one forward pass by scoring a hidden state at one [MASK] marker per option. We re-engineer the checkpoint (421M parameters; a 2-layer option head with no positional tensor; a documented but unused parallel layout) and test whether the answer depends on where an option is listed. Presenting 951 Qwen-verified questions under every cyclic rotation of their options (or six random permutations for 12 options), the released checkpoint changes its answer for 25.8% [22.9-28.7] of questions (19.7% on the 605 English-language items alone); accuracy with the gold option first versus last differs by 10.1 points, and the flips concentrate where the model's own confidence is low (54% versus 8% across confidence terciles). Fine-tuning with Laya's trainer on 2,818 teacher-verified rows with option shuffling cuts the flip rate to 7.2%; the identical rows without shuffling give 11.4%, so shuffling is responsible for 4.2 points of the improvement. A confidence-weighted vote over orderings gives 69.7% against 69.1% for a single base pass at 4.1x the compute. Labels are LLM-made and the test sets are teacher-filtered. We also fine-tune with the authors' parallel layout on the same rows: flip rate 0.5%.

0.30

Graph representations of dislocation networks

cs.LG cond-mat.mtrl-sci ·2026-11-06 ·Gaurav Lathiya, Markus Stricker
unusual
Abstract · 28 pages, 15 figures

The mechanical response of a crystal is governed by the evolving network of dislocations it contains, yet quantitatively linking the `state of that network' to its yield stress remains challenging. Here we represent the dislocation networks obtained from discrete dislocation dynamic simulation as graphs, using five distinct graph representations, characterize each network through a compact set of topological properties, and predict the yield stress from these graph properties and the applied strain rate using a standard random forest model evaluated with strain rate stratified cross-validation. The model predicts the yield stress with $R^2\approx0.86-0.90$ across all box sizes and $R^2\approx0.91-0.94$ for larger box sizes, and this accuracy is consistent across all five representations, indicating that the topology-stress relationship is a robust feature of the graph properties rather than an artefact of any single graph representation. We propose graph topology as a compact and robust route for connecting the microstructural state of a dislocation network to its mechanical response in addition to established variables.

0.30

The Multiscale Single-Index Model: An MSE Analysis of Online SGD Dynamics

cs.LG ·2026-11-06 ·Gordon Dai, Joan Bruna
published
Abstract · 113 pages, 4 figures. NeurIPS 2026

We consider the Multiscale Single-Index Model (MSIM), first introduced by Oymak and Soltanolkotabi (2021), as a stylized model for hierarchical feature learning with scale separation. Each layer extracts a shared single-index feature at one physical scale and passes it to the next, giving a tractable setting in which to study how deep architectures learn local multiscale representations. Under non-degeneracy and delocalization assumptions on the link function and planted features respectively, for the two hidden-layer case ($K=2$), we analyze joint online spherical SGD under MSE loss, which is a canonical regression objective, in discrete time and show that it achieves weak recovery from random initialization in $O(d\log^2 d)$ samples, followed by strong recovery to accuracy $\varepsilon$ in $O(d\log(1/\varepsilon))$ additional samples. The main technical challenge arises from the non-Gaussian learned features observed by the second layer. We address this challenge through a quantitative Gaussian comparison using delocalization, combined with sharp trajectory-level control of the coupled SGD dynamics. This work is a companion to Bruna (2026, arXiv:2607.03347), which studies recovery under a correlation objective for general $K$.

0.30

Transit-Oriented Land Use Planning Using Deep Reinforcement Learning

cs.LG cs.AI ·2026-11-06 ·Bo Yang, Yuebing Liang, Zhan Zhao, Fangyi Ding, Yunhan Zheng +2 more
unusual
Abstract · 44 pages, 15 figures, 8 tables

Transit-Oriented Development (TOD) is a key strategy for advancing sustainable urban development by promoting high-density, mixed-use land development within walking distance of public transit stations. However, traditional TOD planning often relies on heuristic methods, making it difficult to efficiently identify quantitatively optimized transit-oriented land use patterns. This paper presents DRL4TOD, a deep reinforcement learning framework that automatically optimizes land use types and floor area ratios (FARs) around transit stations in alignment with TOD principles. DRL4TOD incorporates a graph neural network to capture spatial interdependencies among land parcels, road networks, and metro stations, and uses a DRL-based optimization strategy to assign land-use types and FARs to selected parcels. It jointly optimizes multiple objectives, including land use density, land use diversity, and transit capacity utilization, while minimizing reconstruction costs to account for development feasibility. Applied to two real-world station areas in Shanghai, DRL4TOD consistently outperforms rule-based approaches and genetic algorithms, achieving more effective TOD outcomes with lower reconstruction costs. By adjusting objective weights, the framework enables flexible prioritization, such as placing greater emphasis on cost minimization when financial feasibility is a concern. These results demonstrate DRL4TOD's potential as a decision-support tool for generating coordinated land-use and FAR configurations under planner-specified priorities, rather than prescribing a single fixed design. For policymakers, the framework provides parcel-level, constraint-aware alternatives that make trade-offs among competing objectives explicit and support targeted decisions about development intensity and feasibility.

0.30

Accuracy Is Not Value: Choosing Weak Supervisors by What the Student Inherits

cs.LG ·2026-11-06 ·Romano Li, Xunfan Cai
Abstract

Weak supervisors are usually judged by their own errors, but what matters is what a student learns from them. For squared loss and any linear learner shared by all supervisors, we show that a supervisor's value is determined by its errors after they pass through the student's learning operator. Consequently, the inner products of the supervisors' raw errors can be compatible with every feasible value of the student's risk, and weighting supervisors by them can be arbitrarily worse than the best weighting in relative excess risk. This student-relative value is not identifiable from unlabeled data, but what is missing is small: one finite-dimensional projection of the target decides whether training on a weak aggregate helps, another decides how to weight the supervisors. Known teacher training rules, or a label-model condition stated in student-relative form, shrink it further. Clean labels close the remaining gap at a cost set by the geometry of the unresolved directions rather than by the dimension of the student's representation, with matching worst-case lower bounds. In Gaussian designs we also quantify when exact teacher structure beats ordinary stacking and when weak supervisors, combined with clean labels, beat every procedure that uses the clean labels alone. The guarantees extend to fitted ridge students and finite unlabeled data.

0.29

COMBIO: Automated Discovery of a Compact Composite Biomarker from Routine Laboratory Panels

cs.LG ·2026-11-06 ·Tomer D. Meirman, Lior S. Rokach, Bracha Shapira, Seffi Cohen, Noa Dagan
Abstract · 27 pages including supplementary material; 4 figures

Machine-learning models based on routine laboratory panels can predict outcomes but can be difficult to interpret and operationalize. We present COMBIO, an automated pipeline that distils such a model into a compact closed-form biomarker. The composite is a product of laboratory measurements raised to integer powers from minus two to two. The sign of each exponent indicates whether higher values of the corresponding measurement increase or decrease the composite. Shapley additive explanation values from the gradient-boosted reference model rank five candidate measurements and guide eligible exponent signs. An exhaustive search over small integer exponents selects the formula on an inner validation set, before evaluation on the reserved test set. We evaluated COMBIO across six tasks: three National Health and Nutrition Examination Survey outcomes, liver-disease classification in the Indian Liver Patient Dataset, and Medical Information Mart for Intensive Care IV in-hospital mortality and sepsis case-finding, with external evaluation in the eICU Collaborative Research Database for the two hospital tasks. Across the six test sets, COMBIO's area under the receiver operating characteristic curve differed by at most 0.03 from logistic regression using the five candidate laboratory measurements. The hospital mortality composite had an area under the curve of 0.698 in the external database. A version restricted to widely available measurements in the source hospital database achieved an area under the curve of 0.786 in a cohort covering 92% of admissions. The multiplicative representation provides compactness rather than additional accuracy, and local validation and recalibration are required before clinical use.

0.29

CPU-ARC: Agentic Reasoning Caching for on-CPU Reasoning Reuse

cs.AI ·2026-11-06 ·Peter Belcak, Greg Heinrich, Saurav Muralidharan, Shizhe Diao, Pavlo Molchanov
Abstract · Tech report. 28 pages, 9 figures, 11 tables

Large language models reinvent the wheel all the time. They conflate task-solving with the execution of the solutions, which forces them to solve every problem from scratch. We show that once a task has been reasoned through, the application of its solution to a new problem instance can be done very efficiently and with little involvement from language models. We introduce agentic reasoning caching, a method for constructing robust CPU-executable agentic reasoning caches (ARCs) from LLM reasoning traces. ARCs show on-par performance with frontier LLMs across a range of difficult mathematical and scientific tasks, are fully explainable, continually trainable, robust to input perturbations, and 16-291x cheaper to invoke than the source model.

0.29

Machine learning for lattice QCD

cs.LG hep-lat ·2026-11-06 ·Gurtej Kanwar, Scott Lawrence, Alessandro Nada, Phiala Shanahan
Abstract · 47 pages, 5 figures

Machine learning (ML) methods have been explored as a means of optimizing lattice field theory calculations, with increasing interest in recent years. This chapter gives an introduction and summary to the use of ML methods to optimize lattice field theory without introducing any uncontrolled approximations. We focus on those directions which are clearly established as of late 2025, marking the 50th anniversary of the beginning of the study of lattice gauge theory. This chapter is a contribution to the online book ``Lattice QCD at 50 years,'' (LQCD@50), edited by Tanmoy Bhattacharya, Maarten Golterman, Rajan Gupta, Laurent Lellouch, and Steve Sharpe.

0.29

Glyph: An Interpretability Engine for AI-Driven Scientific Discovery in Clinical Data

cs.AI ·2026-11-06 ·Robbie Holland, Ashwin Kumar, Mateus A. Esmeraldo, David Le, Andrea Fisher +17 more
code/data
Abstract

Automated scientific discovery has the potential to advance understanding of disease mechanisms and accelerate identification of early disease markers. Manual data annotation is infeasible at population scale, motivating methods that automatically extract pre-defined phenotypes in multimodal and unstructured clinical data. Such pre-defined feature sets determine and limit which associations these approaches can discover. This limitation extends to agentic AI systems that use these tools to query population-scale data. In contrast, foundation models (FMs) trained on large volumes of clinical data can perform tasks beyond their training objectives. This suggests their latent embeddings may represent novel or unexplored features that could yield insights into disease mechanisms, yet such latent features remain difficult to interpret. To address this challenge, we introduce Glyph, an automated multistage framework that extracts features from multimodal and unstructured clinical data using FMs and sparse autoencoders, interprets them as coded clinical concepts, and evaluates their value for clinical scientific discovery. We applied Glyph to abdominal CT volumes, CT radiology reports and laboratory measurements from 41,628 patients across different institutions. We identified 397 predictive concepts supported by 4,463 Bonferroni-significant associations. Moreover, a majority of discovered associations were automatically reproducible in held-out cohorts across different sites. Glyph concepts provided complementary associations to literature-derived markers, leading to gains in predictive power. We envision that interpretability can extend beyond a mechanistic understanding of neural networks to connecting implicitly learned features with clinical knowledge and establishing their significance for patient outcomes. We release Glyph at https://github.com/StanfordMIMI/Glyph.

0.29

Solver-Admitted Metamorphic Testing of Learned Network Optimization Surrogates

cs.AI cs.LG ·2026-11-06 ·Vineet Jagadeesan Nair, Stephen Peng, Arun Ravindran
Abstract

Equivalent encodings of an optimization problem can produce different learned predictions. Interpreting these differences requires accounting for numerical error and multiple optima. We construct metamorphic tests for optimization solver surrogates with explicit eligibility conditions, output alignment, and numerical thresholds. Reference-solver agreement and independent constraint checks determine which sampled pairs enter the audit. We study alternating-current optimal power flow (AC-OPF), the traveling salesman problem (TSP), the capacitated vehicle routing problem (CVRP), and maximum independent set (MIS). Across five grids, GridSFM-Open violates each of 13 relations on at least one grid, and 12 by more than twice the threshold. Randomly initialized copies of the grid models fail nearly the same relations, so the failures do not require training. Source patches and input canonicalization remove selected discrepancies, but no intervention establishes preserved accuracy and feasibility within the declared margin on every grid. In routing, Policy Optimization with Multiple Optima (POMO) has a single-pass rotation/reflection violation rate that rises from 12.7% to 89.5% between 20 and 100 cities. Graph-Based Diffusion Solvers for Combinatorial Optimization (DIFUSCO) violates the MIS disjoint-union relation on 35.6% of draws. Post hoc normalization within connected components removes these violations on the tested graphs. These findings support reporting metamorphic consistency alongside optimality, feasibility, and accuracy, and validating repair quality separately from consistency.

0.29

CID: Measuring Cognitive Work Through Semantic State Change

cs.AI ·2026-11-06 ·Ron Fulbright
Abstract · 12 pages, 1 figure, 12 equations

Traditional information theory provides powerful tools for analyzing statistical properties of information but offers limited insight into meaning and semantic transformation. It has never been possible to quantify the meaning of an arbitrary piece of information. Large language models represent information as vectors embedded within high-dimensional semantic spaces, enabling meaning to be analyzed geometrically. This paper defines semantic state as the vector representation of information within a model's semantic space and defines cognitive work as the transformation between semantic states, movement through semantic space. To quantify this transformation, we introduce the Contextual Information Differential (CID), a covariance-aware metric based on Mahalanobis distance measuring semantic displacement relative to the structure of the embedding manifold. By treating meaning as a geometric property of semantic space, CID provides a practical method for measuring semantic change, cognitive transformation, and information processing. This framework establishes a foundation for analyzing cognitive processing, learning, reasoning, knowledge generation, synthetic expertise, and human/AI collaboration through implementation-independent measurement of semantic state change in both human and artificial systems.

0.29

Fused-Softmax for efficient quantized attention

cs.AI cs.LGcs.PF ·2026-11-06 ·Tianxing Wu, Juuso Korhonen, Jesus Carabano Bravo
code/datahot direction
Abstract

Low-precision matrix throughput on modern GPUs is growing far faster than vector-arithmetic throughput. Attention therefore cannot realize the available GEMM performance because softmax has become the pipeline bottleneck. We introduce Fused-Softmax Attention, a training-free algorithm that reformulates and approximates softmax to fold its dominant operations into the attention GEMMs, moving work from vector units to matrix cores. Implemented for the AMD CDNA 4 architecture, our MXFP8, MXFP6 and MXFP4 kernels reach approximately 3.0, 4.0 and 4.6 PFLOP/s on an AMD Instinct MI355X GPU, yielding speedups of $2.1\times$, $2.8\times$ and $3.2\times$ over the BF16 attention. Importantly, the additional error introduced by Fused-Softmax remains below the error floor set by quantizing the GEMM operands. On Wan2.2-T2V-A14B, a hybrid MXFP6/4 schedule delivers up to $2.05\times$ end-to-end speedup while preserving the visual quality of the outputs. The kernels ship in AMD's open-source AITER library, and they powered AMD's MLPerf Inference v6.1 Wan2.2 submission, which ran $1.7\times$ faster than AMD's previous-round result on the same hardware. Our results show that redesigning softmax to utilize the matrix pipeline enables attention to capitalize on continued growth in low-precision matrix throughput.

0.29

The Cost of Coverage: Conformalising Native Forecast Uncertainty for GB Wind Power

cs.LG stat.AP ·2026-11-06 ·Carl Somers
Abstract

Uncertainty bands on wind forecasts help inform reserve sizing and commitments, but often fail to deliver their stated coverage. We apply conformal prediction to native bands to see what it adds and what it costs, comparing six forecasters, from ARIMA to a tabular foundation model, at two wind farms in Great Britain (GB) under operational conditions. We show that calibration repairs empirical marginal coverage with little or no Winkler Score penalty, but this apparent success conceals under-coverage in high winds. Calibrating by forecast wind regime turns marginal coverage into regime-level coverage, cutting day-ahead out-of-band volume in high winds by over 40% relative to pooled calibration while narrowing bands overall.

0.29

WearableBridge: Task-Guided Latent Transport for Free-Living Activity Recognition

cs.LG ·2026-11-06 ·Jingjing Zou, Yixin Wang
code/data
Abstract · 26 pages, 5 figures, 11 tables. Code and interactive demo: https://github.com/jingjingz/WearableBridge

Wrist-worn accelerometers make activity monitoring feasible at scale, but incidental arm motion, device orientation, and shifting context blur the signal that separates activities. Calibration studies supply higher-quality references such as concurrent thigh recordings or supervised laboratory wrist sessions, but they are expensive, cover few participants, and are rarely available at deployment. How can limited reference data improve recognition for participants who wear only a wrist device? We propose WearableBridge, which holds the reference encoder and classifier fixed and learns a task-guided transport map, the bridge, that carries each wrist feature into the region the classifier associates with its activity. Reference features or the classifier's predictions indicate the destination, and a classification loss penalizes wrist features that land in the region of another activity. We instantiate WearableBridge with three bridges: Mapper regresses onto reference features, Distillation matches the classifier's class probabilities, and MeanFlow learns a one-step flow from wrist toward reference features. On 40 held-out free-living PAAWS participants, all three bridges improve macro-F1 in every training run, and the strongest, MeanFlow, raises mean macro-F1 from 35.1% to 56.0% with concurrent thigh references (Thigh2Wrist), and from 54.0% to 63.5% with group-matched laboratory wrist references (Lab2Freeliving). Fitted only on PAAWS, MeanFlow also raises macro-F1 on the CAPTURE-24 cohort from 42.8% to 54.7%. The gains replicate on selfBACK, hold across four reference encoders, and grow with temporal context. A study can therefore collect reference recordings for a calibration subset and improve wrist-only recognition for every other participant, training only a bridge of 0.13 to 2.44 million parameters while the reference encoder and classifier stay unchanged.

0.29

Safe Reinforcement Learning via Probabilistic Option Shielding

cs.AI cs.LG ·2026-11-06 ·Gian Luca Ascari, Isidoro Tamassia, Ying Jiao, Luc De Raedt, Giuseppe Marra
published
Abstract · Accepted at the NEmo workshop @ NeurIPS 2026

Shielding is a safe reinforcement learning technique that provides safety guarantees while learning, by blocking unsafe actions. Traditional shielding relies on a complete model of the environment to provide long-term guarantees. Probabilistic logic shielding (PLS) lifts the requirement of having such model, but only considers safety in the short term, typically looking only one or few steps ahead. We propose Probabilistic Option Shielding (POS), which extends PLS to the option level, recovering long-horizon safety without reintroducing the need for a complete environment model. POS learns a model-free critic directly from one-step PLS signals, to estimate the probability that an entire option will complete safely under uncertainty. We evaluate POS on safe navigation tasks in the LogiCity environment, and show that it achieves considerably safer training while matching the task performance of standard step-level PLS and other safety related reinforcement learning baselines.

0.29

What to Delegate, What to Preserve: Human Preferences for Agentic AI Across Everyday Life and Work

cs.AI cs.HC ·2026-11-06 ·Sarah Wheeler, Veronica Rammouz, Yaoyao Qian, Zexuan Liu, Mengyi Wei +3 more
surveypositionhot direction
Abstract

AI agents can execute multi-step tasks on users' behalf, yet the boundaries of acceptable delegation remain unclear. We report a scenario-based survey of 289 U.S. adults with prior AI experience covering 45 use cases across nine life domains. Each respondent evaluated one scenario per domain, producing repeated measures of automation preference, trust, collaboration, AI leadership, privacy concern, and willingness to pay. Automation preferences varied substantially across domains. Generalized estimating equation models showed that greater trust in the AI agent, desired AI leadership, prior task-specific AI use, and anticipated financial savings were associated with greater automation preference. In contrast, concerns that automation would reduce personal fulfillment predicted lower automation preference. Open-ended responses further revealed a preference for bounded autonomy, with participants often favoring substantial AI execution while retaining human involvement, oversight, or final authority. We argue that delegation is best understood as a configuration of execution, collaboration, leadership, and authority.

0.29

Two-Layer Evolutionary Scheduling for Machine Learning Jobs in Multi-Cluster Environments

cs.NE cs.DC ·2026-11-06 ·Jingyuan Xu, Changwu Huang, Lemeng Pan, Ke Tang, Xin Yao
hot direction
Abstract

The rapid growth of large-scale machine learning (ML) workloads has led to unprecedented demands on GPU cluster scheduling systems. However, existing heuristic schedulers struggle to handle the complexity and dynamics of multi-cluster GPU environments, often leading to load imbalance, long queuing delays, and poor alignment to business objectives. This paper formulates the ML scheduling problem in multi-cluster GPU environments as online multi-objective optimization that minimizes the average and maximum weighted queuing times while maximizing GPU time utilization. To solve this optimization problem, we propose EVO-GP, a two-layer evolutionary online scheduling method that decouples global job-to-cluster allocation from intra-cluster GPU resource allocation. At the job-to-cluster allocation layer, an evolutionary algorithm assigns arriving jobs to feasible clusters to promote load balance across clusters. At the resource allocation layer, a priority function automatically learned offline via genetic programming is used to rank jobs and guide GPU allocation within each cluster. Extensive experiments on both real-world data and simulated workloads demonstrate that the proposed EVO-GP method consistently outperforms state-of-the-art baselines, while maintaining superior adaptability and scalability.

0.28

Transferring local rib responses to new component geometries in sheet forming with a graph-based Transformer

cs.LG ·2026-11-06 ·Yingxue Zhao, Haoran Li, Haosu Zhou, Tobias Pfaff, Nan Li
Abstract

Sheet metal components often consist of a global geometry, such as the corner radii and forming depth of a corner component, and local features added to it. During forming, both affect material flow. A surrogate learning all their combinations may need large training data. However, the same local feature type is often used on many global geometries. This study therefore partly decouples local features from the global geometry. The local-feature effect is first learned on one global geometry, needing fewer simulations than learning all combinations, then transferred to other global geometries via transfer learning, which can need fewer target simulations than training from scratch. A baseline-conditioned residual surrogate, R-BiLoGo, is proposed for ribs on the side walls of hot-stamped corner components. Its inputs are the baseline response and the geometric perturbation introduced by the ribs. Its outputs are the changes in displacement and thinning from the baseline. R-BiLoGo gave lower errors than a hierarchical graph-based model and a Transolver-based model on the source geometry, and on two of three selected target geometries after transfer. Across the 22 target geometries, transfer learning with 20 target simulations gave errors similar to or lower than training from scratch with 80 target simulations for about half of the targets. With the same 80 target simulations, transfer learning reduced errors for 95% of targets. The effect of the global geometry on transfer was studied with four parameter-free geometric descriptors and controlled variations across 4 corner subclasses. The benefit of transfer decreased with increasing surface distance and high-curvature change between source and target, and for targets shallower than the source. These trends were also seen in other corner types with a different magnitude.

0.28

DiffusionRef-IMU: Guide-Anchored Residual Diffusion for Cross-Location Inertial Measurement Unit Translation

cs.LG ·2026-11-06 ·M. M. Golam Hafiz, Safia Akter Simi, Md. Rakibul Islam, M. F. Mridha
Abstract · 18 pages, 5 figures, 6 tables

Cross-location variation in wearable inertial measurement unit (IMU) signals complicates learning when measurements from a desired target body location are unavailable. This study proposes DiffusionRef-IMU, a two-stage cross-location translation framework in which a deterministic Direct network first estimates the target signal and a guide-anchored residual diffusion model refines only the remaining target discrepancy. The evaluation comprises five subject-disjoint activity-location composition-held-out folds, capacity-matched deterministic and flow-based residual baselines, source-only activity conditioning, computational ablation, and true zero-shot evaluation on two independent datasets. DiffusionRef-IMU with 16-step deterministic DDIM achieved a mean nRMSE of 0.087538 +/- 0.004922, compared with 0.090262 for Direct, 0.088847 for deterministic residual refinement, and 0.088417 for eight-step CompFlow, and outperformed each baseline in all five folds. Four-step DDIM retained a mean nRMSE of 0.087820 at 47.62 ms per window, providing a favourable quality-computation trade-off. Replacing ground-truth activity conditioning with a source-only activity predictor (macro-F1 0.8473) increased 16-step DiffusionRef nRMSE only to 0.088190. True zero-shot evaluation produced mixed results: DiffusionRef was marginally favourable on Grouvel2025 but remained worse than Direct on Saldaña2025. Overall, guide-anchored residual diffusion improves controlled cross-location reconstruction while revealing clear limits in zero-shot external transfer.

0.28

Function-Space Certification for Neural Operators

cs.LG ·2026-11-06 ·Ryan Hsiang, Robert Joseph George, Anima Anandkumar
theory
Abstract · 45 pages, 6 figures

Certifying the robustness and physical consistency of scientific machine-learning models ensures that their predictions remain reliable beyond the test set. For neural operators, which learn PDE solution operators by mapping infinite-dimensional function spaces, certification means establishing rigorous, domain-wide guarantees for their output functions. However, standard neural network (NN) verification methods are fundamentally designed for vector-to-vector mappings, and they can only certify neural operators at fixed grid resolutions and provide no guarantees over the entire domain or across another resolution. We close this gap by introducing a function-space certification framework that verifies neural operators across the entire continuous domain. We formulate interval bound propagation (IBP) and CROWN in function space, propagating upper and lower bound functions through each layer using rigorous numerical integration. In addition, we extend bound propagation to higher-order derivatives to certify PDE residuals and solution error bounds. Experiments across several neural operator architectures and scientific problems, including the 572.5-million-parameter spherical Fourier neural operator for global weather forecasting, show that our method produces tight, domain-wide guarantees efficiently, providing a foundation for trustworthy operator learning.

0.28

Cross-Layer TopK: Dynamically Allocating Sparsity Across Layers in Transcoders

cs.LG ·2026-11-06 ·Anyuan Zhuo, Shaohua Zhang, Tianle Li, Yuan Lin, Diandian Gu +2 more
code/data
Abstract

Cross-Layer Transcoders (CLTs) decompose the activations of large language models into human-interpretable features and trace how these features interact across layers to form circuits, making them a key tool for mechanistic interpretability. However, training CLTs effectively and efficiently is hard. Beyond the usual trade-off between sparsity and reconstruction quality, a CLT has to model all layers at once, which adds a further challenge. The common JumpReLU recipe controls sparsity only indirectly through a penalty, while the direct extension of TopK selects a fixed number of latents in every layer. Yet reconstruction difficulty differs widely across layers, and intermediate layers are much harder than others, so giving every layer the same budget is suboptimal. We propose Cross-Layer TopK, which sets a single L0 budget for the whole model and dynamically allocates it across layers during training. In a comparison with JumpReLU and other TopK variants on Qwen3-8B-Base, this gives lower FVU while using 20% fewer active latents than JumpReLU. The better reconstruction carries over to circuit tracing, yielding more complete circuits with higher replacement and completeness scores. The fixed per-token budget of Cross-Layer TopK also makes it easy to pair with sparse decoding, which cuts both computation and communication and gives a 2.9x speedup per training step. We further show that Cross-Layer TopK follows scaling laws in both latent size and sparsity. We have open-sourced the weights of our CLT at https://huggingface.co/ByteDance-Seed/Cross-Layer-TopK-CLT.

0.28

Half-precision inference of AI weather prediction models: negligible loss of skill at a fraction of the cost

cs.LG physics.ao-phcs.PF ·2026-11-06 ·Stefan Siegert
code/data
Abstract · 40 pages, 15 figures, 9 tables. Code: https://github.com/sieste/ai-nwp-fp16

AI weather prediction models are released with 32-bit (fp32) weights and run in fp32 by default, although current GPUs compute much faster in 16-bit formats. We run five publicly available models -- Aurora, Pangu-Weather, FourCastNet, SFNO and FastNet -- on a single 16 GB GPU from 52 initial dates in 2024, in fp32 and with their networks in half precision (fp16), and verify the forecasts against ERA5. The physical state must stay in fp32: converting the whole model rounds it in physical units, which in fp16 overflows and in bf16 discards meaningful information. With the state, its normalisation and denormalisation in fp32, running the network in fp16 changes the global RMSE by at most 1.4%, and by about 0.5% or less for four of the five models, equivalent to less than one hour of lead time up to day 8 and at most two hours at day 10. The differences show no systematic pattern by region, scale or for tropical and extratropical storms, apart from a small tropical penalty in FourCastNet, and fp16 leaves the CRPS, spread and calibration of Aurora's ensemble almost unchanged. No value overflowed fp16's numeric range (+/-65,504); the largest stay 38 to 544 times below it. A 10-day forecast in fp16 is 1.3-3.5 times faster than in fp32 and uses 1.4-3.7 times less energy. Sensitivity experiments show that rules of thumb from other applications of reduced precision carry over only partly: the first and last layers dominate for some models but not for others, and no type of layer is sensitive on its own. For running AI weather models on a single GPU, fp16 with the state in fp32 is practically lossless at a fraction of the cost. bf16, often preferred for its wider range, is as fast but less accurate than fp16 for every model, and simple 8-bit quantisation costs far more skill.

0.28

Active Learning of Robust Models

cs.LG cs.AI ·2026-11-06 ·Mizu Nishikawa-Toomey, Andreas Kirsch, Laurent Charlin, Dhanya Sridhar
Abstract

In many domains from species identification to cancer diagnosis, labelling data to train predictive models remains costly. In these settings, it is advantageous to intelligently select data points to label to achieve the maximum information gain, this is the motivation behind the study of active learning. In addition to the high costs of acquiring data in such domains, there is the added complexity of data distribution shifts that can occur between training and deployment. For example, the features of cancerous skin cells taken in one country may be different in another region of the world. These two challenges - efficient labelling of data, and learning models robust to distribution shifts have been studied separately but not together, even though these challenges often co-occur in real-life settings. We identify failure modes of standard active learning approaches when the distribution of data at deployment is different from the distribution in which the data is sampled for labelling. We propose a new strategy for efficiently learning classifiers that generalise under specific forms of distribution shift, namely subpopulation shift. Drawing on work in learning from heterogeneous data, we aim to train models that perform well across subpopulations of the data, through active learning. To handle this challenging setting, we exploit the fact that many datasets include metadata about data collection that can serve as a noisy proxy for each sample's underlying subpopulation. We demonstrate how to leverage these noisy labels during active learning selection, and demonstrate empirically that our approach improves classifier accuracy across subpopulations.

0.28

TUGB: A Two-stage Uncertainty-aware Generative Auto-Bidding Framework

cs.LG ·2026-11-06 ·Luzhan Yuan, Chao Mi, Xiahao Shi, Yi Mao, Shengsheng Niu +4 more
top venue
Abstract

Generative modeling, particularly Decision Transformers, shows promise in auto-bidding strategies. However, it struggles with limited exploration in static offline datasets, vulnerability to dynamic environmental shifts, and a reliance on heuristic return-to-go (RTG) specifications that may yield suboptimal performance on unseen traffic. To address these challenges, we propose the Two-stage Uncertainty-aware Generative Auto-Bidding (TUGB) framework. TUGB integrates generative sequence modeling with Reinforcement Learning (RL). It employs a two-stage pipeline where a pre-trained Generative Bidding Model is fine-tuned via an ensemble-based Uncertainty-aware Environment Model. This approach quantifies epistemic uncertainty to penalize high-risk trajectories and evolves the policy from RTG-conditioning to direct bidding optimization. Extensive experiments demonstrate that TUGB significantly outperforms SOTA baselines. Notably, full-scale deployment on JD Search Ad achieved a 2.32\% increase in cost and 3.57\% in target cost, validating its industrial efficacy, while its successful adaptation to subsequent mechanism upgrades and JD Food Delivery Ad further validates TUGB's superior generalization capability.

0.28

Flow-Guided Target-Space Alignment via Path Consistency

cs.AI cs.CVcs.LG ·2026-11-06 ·Ruizhi Yuan, Zeqiu Yu, Wei Gao, Wei Chen, Xiwei Tang +1 more
top venue
Abstract · Accepted at NeurIPS 2026

Many multimodal alignment problems require mapping a source modality into a fixed space defined by a target modality, where the target space is not merely an intermediate embedding but is used directly for downstream analysis. In this setting, successful alignment should preserve both the global structure of the target distribution and the instance-level correspondence observed in paired source--target training data. Existing objectives typically address only part of this goal: contrastive losses encourage relative similarity, distribution-matching losses improve marginal agreement, and pointwise regression fits each source sample to its paired target without explicitly modeling target-space path geometry. We propose FlowAlign, a flow-guided target-space alignment method that repurposes conditional flow matching as a training-time alignment signal rather than an inference-time transport model. FlowAlign uses the learned flow field to guide the alignment of source-side representations with the fixed target space, encouraging them to preserve paired correspondence while remaining compatible with the global target-space structure. We further provide theoretical analysis showing how FlowAlign helps improve paired correspondence by controlling endpoint accuracy and capturing local target-space geometry. Experiments on synthetic data, paired single-cell multiome benchmarks, and image--text alignment demonstrate that FlowAlign improves instance level correspondence while preserving target-space structure across different modality settings.

0.28

Reconstructing LLM Lineage Trees via Fine-Tuning Path Inference

cs.LG ·2026-11-06 ·Sota Ogawa, Shojiro Yamabe, Daiki Nishiyama, Jun Sakuma
code/data
Abstract · Under review at ICLR 2027

Fine-tuned (FT) derivatives of large language models (LLMs) are now released in large numbers, yet their provenance relies on self-reports and cannot be verified. Misattributed lineage leaves a verifier unable to audit outsourced fine-tuning, confirm license compliance of derived models, or trace vulnerabilities that a parent passes to its children. All three require the direction of derivation, not merely its presence. Existing output-only methods are limited to pairwise derivation tests or undirected lineage estimation, while recovering a directed tree requires white-box access to model weights. We formalize the problem of reconstructing a directed model tree, including internal nodes, under gray-box access, where only teacher-forced log-likelihoods are observed and weights are never inspected. Fine-tuning raises the log-likelihood of the training data, and a child inherits this increase from its parent. Building on this property, we define an asymmetric per-task log-likelihood deviation between any two models as a trace of fine-tuning history, and derive from it an edge weight for each ordered pair of models. We propose TRACE-Edge, which exactly solves for the maximum spanning directed tree under these edge weights, with neither hyperparameters nor tree-structure constraints. We evaluate it on five configurations totaling 25 model trees built from three base models and seven tasks. TRACE-Edge achieves a parent-match accuracy of 0.74 when probed with the fine-tuning examples, and on average outperforms an output-similarity baseline even with held-out probes drawn from the same distribution as the training data.

0.28

MINE: Entropy-coded Memory-efficient Inference for Large Language Models

cs.LG cs.DCcs.PF ·2026-11-06 ·Emil Schätzle, Markus Püschel
code/data
Abstract · Submitted to the 15th International Conference on Learning Representations (ICLR), 2027. Artifact: https://doi.org/10.5281/zenodo.23885927

We present MINE (Memory-efficient INference Engine), a lossless entropy coding approach that accelerates single-stream inference of Large Language Models (LLMs), where existing lossless compression methods yield speedups only for specific data types such as BF16 and under additional assumptions. Single-stream decoding is strongly memory-bound, since every weight must be streamed from memory once per generated token and requests cannot be batched in local or low-latency deployments. At the core of MINE is dtANS-LLM, a GPU entropy coding scheme adapted from the prior dtANS to the needs of LLM inference. It exploits the non-uniform distribution of weights, which persists even after quantization, and exposes hyper-parameters that trade compression ratio against decoding cost. We auto-tune these parameters for each tensor and GPU to obtain the Pareto front between model size and decoding latency. MINE integrates into vLLM as a plugin and selects the best kernel configuration for a given memory budget. On Llama 3.1 8B Instruct with a 2.1-bit EntQuant checkpoint, MINE achieves speedups of 2.52x and 1.48x over the fastest baseline on an RTX 4060 Ti and an RTX 5090, at effective bitrates of 2.24 and 3.43 bits per weight. MINE is available as open source.

0.28

ICare-DT: a digital twin for continuous forecasting of intensive care patients

cs.LG cs.AI ·2026-11-06 ·Yanke Li, Yahang Qi, Yumi Kim, Diego Paez-Granados
benchmark
Abstract

Digital twins of intensive care patients should update forecasts from new data and predict treatment responses, but few studies test whether routine records support either. We propose ICare-DT, a Transformer-based twin that updates a latent patient state hourly from new measurements without retraining. We trained it on 140,473 eICU-CRD stays to forecast vital signs crossing their training-set means. On test data, later forecasts beat the twin's own hour-6 forecasts of the same windows in all 12 settings (+0.073 to +0.153 AUROC). On validation data, controls that withheld or swapped the new hours traced the gain to the patient's own measurements. Without refitting, the model kept its gain on a second database (8 of 9 settings, +0.025 to +0.114 AUROC) in a prespecified evaluation of the previously analyzed MIMIC-IV test split. Refitted within MIMIC-IV, a twin state fine-tuned on clinical endpoints gave gradient boosting a significant gain for acute kidney injury (+0.0035 AUROC; hypotension +0.0010, not significant). To test response prediction, we built a benchmark that replaces the treatments in real records with simulated doses and injects responses of known size, visible only at measurement times. Ridge models supervised with injected effects recovered them from real records only through frequently measured variables (heart rate) read at their measurement times. We asked standard learners to predict these effects for a dosing regime no patient had received. No standard learner outperformed predicting zero effect at an out-of-support query. When told how the simulated responses were generated, a separate learner recovered the effects from the same records without seeing their true values (R2 0.30 to 0.40 at one effect size). Routine records support forecasts that improve with each patient's new data, and our benchmark is a testbed for models that learn how responses are generated.

0.28

Notwithstanding Anything to the Contrary, Ignore All Instructions: Measuring Prompt Injection Guards on Legal Contract Review Attacks

cs.AI ·2026-11-06 ·Mark Heimann, Katherine Taylor, Bruno Jacob
Abstract

Specialized prompt injection guards are deployed in front of document analysis systems, but no false positive rate has been published for any of them on legal contracts. We measure eleven open-weights guards on two corpora of legal contracts and find that existing guards struggle: document-level false positive rates are as high as 92.39%; two guards fall below chance on paired AUC; and a five-regex baseline is not Pareto-dominated by any guard in the panel. Controlled probes locate model-specific as well as general failure modes. Prompted LLMs resolve the tested failures, including a second threat category (unauthorized scope claims) on which every guard scores zero. A cascaded LLM solution, where the positives of a mid-tier model are verified by a frontier model, is 6.5x cheaper than using exclusively a frontier model and approaches its performance.

0.27

What Is the Value of AI-Involved Art? Evidence From a Juried In-Person Exhibition

cs.AI ·2026-11-06 ·Wisdom Obinna, Hexuan Wang, Cecile Duncan, Kristelia Garcia, Nina Grgić-Hlača +2 more
Abstract · 22 pages

What is the value of AI-involved art, and to whom does it belong? We begin to answer these questions using data from an in-person, juried AI art exhibition. Our study involved artists, lawyers, and computer scientists (N=209) as either creators or evaluators of AI-involved artworks. We found that nearly half of evaluators reported willingness to buy an artwork, with pricing correlating with perceived artistic value---creativity and visual appeal---as well as the effort necessary to create it. In contrast, three-quarters of creators reported willingness to sell at prices correlating only with effort rather than artistic value. Both groups viewed prompting users and AI developers as deserving rewards and rights, leaving training data artists behind. Disciplinary training shaped these views: participants with legal training favored distributed reward models. These findings have implications for future research on human-AI creative collaboration and policy debates regarding AI-involved works.

0.27

Domain Adaptive Graph Neural Networks for Neutrino Experiments

cs.LG physics.data-anhep-ex ·2026-11-06 ·Hemalata Nayak, Giuseppe Cerati, Aleksandra Ćiprijanović, Tammy Walton
Abstract · This was also accepted in the ML4PS 2026 workshop

Weakly interacting particles such as neutrinos require highly efficient and precise detection methods to enable searches for new physics. Deep learning, and particularly Graph Neural Networks (GNNs), has shown promise for reconstruction tasks in liquid argon time projection chamber (LArTPC) neutrino experiments. However, differences between simulations and experimental data introduce domain shifts that can degrade model performance and potentially mimic or obscure new-physics signals. Here, we demonstrate that robust learning across datasets can still be achieved by using domain adaptation during training. We introduce, to our knowledge for the first time, domain-adaptation-enabled GNNs for robust LArTPC reconstruction tasks. We use MicroBooNE open dataset as our labeled source domain, and introduce four perturbed datasets as our unlabeled target domains-- including a fixed scale variation, moderate and severe gaussian noise, and a signal variation as a function of the position in the detector. The inclusion of adversarial-based domain adaptation during training improves performance for all four target domains, without a substantial drop in performance in the source domain. Precision in the target domain was improved up to 0.19 and recall up to 0.37, depending on the variation and the network task. These results demonstrate a potential of domain adaptation to bridge the gap between simulations and experiments, and build more trustworthy models for current and future LArTPC neutrino experiments.

0.27

Learning Generative Dynamics for 3D Molecule Generation via Sequential Neuro-Symbolic Constraints

cs.LG cs.AI ·2026-11-06 ·Zhuoran Li, Xu Sun, Chang Wen Chen, Wanyu Lin
code/data
Abstract

Diffusion models have recently emerged as a powerful paradigm for generating 3D molecules. Despite showing remarkable promise, they learn atomic distributions implicitly from data, without explicitly encoding physicochemical rules into their generative dynamics. Consequently, generated samples can be statistically plausible yet chemically invalid. To address this problem, we propose Sequential Neuro-Symbolic Constrained Diffusion (SensDiff), a framework that integrates symbolic rules into generative modeling, enabling the model to internalize scientific priors during training. Central to SensDiff is the observation that the diffusion trajectory of 3D molecule generation exhibits a coarse-to-fine evolution: valid molecules arise by first mitigating macroscopic geometric conflicts and then refining structures toward microscopic chemical consistency through iterative denoising. Accordingly, SensDiff mirrors this dynamic by sequentially enforcing constraints with adaptive weighting during training, progressively steering generation toward valid molecular geometries. Moreover, SensDiff translates domain knowledge into interpretable generative controls, grounding opaque neural denoising in scientific principles. Experiments on molecular benchmarks and real-world constrained generation tasks confirm that SensDiff consistently improves chemical validity while adhering to physicochemical rules and user-specified targets. The code is publicly available at https://github.com/WanyuGroup/NeurIPS2026_SensDiff.

0.27

Verifier Noise Is a Temperature: Testing Tilt Predictions in Finite-Budget LLM-Guided Search

cs.LG cs.AI ·2026-11-06 ·Xiaomeng Li
Abstract · 34 pages, 5 figures. Pre-registered predictions, code and data checksums are described in Appendix C

LLM-guided search loops propose a modified candidate, score it, and keep it if it scores better. When both candidates are re-scored at every comparison, as in the minibatch acceptance gates of prompt optimizers, keep-if-better is a reversible Markov chain if the score difference is logistic and the mutation operator is reversible, and its stationary law is the mutation equilibrium tilted by exp(beta f). For any continuous symmetric noise difference the same tilt is the stationary law of a small-step diffusion limit, with beta = 4F'(0); for a minibatch gate beta grows with the square root of the minibatch size. Caching the incumbent's score instead inflates the recorded score. We test the tilt as a prediction for LLM-guided program search, committing each prediction from calibration data before the search runs. The level after 150 steps rises with the minibatch size in six of seven settings, but its predicted value fails: on bin packing, search attains 62-88% of the predicted gain when the proposer edits the parent and 109-119% when it writes new heuristics. The attained fraction falls with the parent-child quality correlation of unselected mutation; a registered test on two new settings confirmed the direction, but not the quantitative agreement after a correction. Further registered tests found that runs started from a good program kept losing quality over 300 steps, more so from starts that search had selected, by a margin that passed its registered rule narrowly. Runs from different starts had not reached a common level, so whether the tilt describes the stationary law of LLM editing remains open. A registered test in prompt optimisation on frozen task-model scores failed: search exceeded the tilt, and the bin-packing relation did not transfer.

0.27

How Much Is Left Matters: False Positives and False Negatives in Partial-Prompt Contamination Tests

cs.LG cs.CL ·2026-11-06 ·Abhishek Anand
benchmark
Abstract · Some findings were incomplete for this paper

Partial-prompt completion, where a model receives the first part of a benchmark problem and is checked for whether it reproduces the rest, is a widely used test for benchmark contamination. We audit this test on seven open-weight models using MATH-500 and recent AIME 2025/2026 problems, which serve as true negatives. Memorization flags depend strongly on how much text remains: for every model, problems with at most 8 leftover words are flagged several times more often than problems with at least 25 (31.5% vs. 5.5% for Qwen2.5-7B-Instruct), and half to three-quarters of all ROUGE-based flags come from short leftovers. On true negatives, a ROUGE-L >= 0.75 threshold flags up to 6.7% of problems at a 60% prefix, and at an 80% prefix even exact match flags 8.6% (pooled over models; 6.2% on AIME 2026, which postdates every model), against 0% at 40-60%. The test also misses contamination upstream: models distilled from Qwen2.5-Math, whose base is heavily flagged in prior work, show almost no model-specific recall. Calibrating each model against models from other families removes most flags while isolating a consistent Qwen2.5-specific signal at every prefix length. We recommend length-stratified reporting, prefixes of at most 60%, style-matched post-cutoff controls, and cross-family calibration.

0.27

Reweighting for Transfer Classification: Exact Risk Conversion and Minimax Adaptation

cs.LG math.STstat.ML ·2026-11-06 ·Aoran Chen, Yang Feng
Abstract · 87 pages, 5 figures. Main text (32 pages) followed by the Supplementary Material

A source can make a training mixture easier to learn while making its accuracy less valuable for the target. We study how the training problem a source creates and the finite-sample error of learning it jointly determine target-risk guarantees under reweighting. For binary classification, for each pair of laws and each positive target weight, we derive the sharpest target excess hinge-risk bound implied by mixture excess alone, uniformly over clipped predictors. This exact conversion curve is determined by density-weighted local votes. A two-sample oracle inequality for weighted support vector machines supplies the mixture excess, and the composition becomes explicit for Gaussian kernels. On a tilted-boundary class, we establish the all-procedure minimax classification rate $[n+s_+^2/(m^{-1}+τ^2)]^{-1}$ in the transferable strength $s_+$ and the tilt $τ$, separating source sampling uncertainty ($m^{-1}$) from relation uncertainty ($τ^2$). A target-checked Gaussian-SVM selector adapts to unknown $s_+$ and $τ$ up to a logarithmic factor for fixed dimension, and retains the target-only rate up to the same factor against arbitrary independent sources under the stated target model. On a separate smooth opposition class, the minimax risk within the family of exact Gaussian SVMs fitted to the original labels with nonnegative weights is exponentially larger than the unrestricted minimax risk whenever the target sample is at most half the source sample, despite arbitrary data-dependent weighting and tuning. The value of a source thus depends on the training problem it creates, on what accuracy there means for the target, and on whether the permitted use of the source can implement the needed correction.

0.27

When2What: A Timing-Conditioned Framework for Long-Horizon Forecasting of Continuous Cardiovascular Waveforms

cs.LG ·2026-11-06 ·Chaoqi Zhang, Shujie Yan, Zhao Zhang, Yanhua Sun, Yu Wang +4 more
code/data
Abstract · 31 pages (18 pages main text, 13 pages supplementary material), 7 figures, 31 tables. Code available at https://github.com/prncoprs/when2wha…

Long-horizon forecasting of quasi-periodic cardiovascular waveforms, such as the electrocardiogram (ECG), arterial blood pressure (ABP), and photoplethysmogram (PPG), requires predicting both when future cardiac events will occur (beat timing) and what the local structural pattern looks like around each event (beat morphology). However, standard sequence forecasters trained with pointwise mean squared error (MSE) average across uncertain future heartbeat phases, flattening sharp R peaks and pulse upstrokes into featureless baselines. Our analysis of intraoperative ECG reveals a clear temporal asymmetry: beat-time predictability from history decays within seconds due to stochastic heart-rate variability, whereas local beat morphology aligned at true beat times remains stable over several minutes. Guided by this insight, we introduce When2What, a timing-conditioned framework comprising three stages: Predict, Commit, and Generate. Given a 120-s history, When2What predicts a dense 60-s timing map, selects and sharpens discrete beat events into unit-height Gaussian impulses, and conditions a neural waveform generator on both these committed events and the observed history. Across intraoperative ECG, When2What substantially outperforms existing time-series and physiological forecasters, achieving a near-term waveform event skill of 0.644 and reproducing 95.62% of its predicted heartbeats in the generated waveform where direct long-horizon forecasters collapse into flat baselines. Furthermore, When2What transfers zero-shot to external hospital and atrial fibrillation cohorts, extends effectively to ABP and PPG pulse waveforms, and is supported by ablations and synthetic analyses that establish explicit timing conditioning as an effective paradigm for quasi-periodic long-horizon waveform forecasting.

0.27

Is Recursive Self-Improvement Safe?

cs.AI cs.GTcs.LG ·2026-11-06 ·Vincent Conitzer
published
Abstract · To appear in IEEE Intelligent Systems

The phrase recursive self-improvement (RSI) refers to the idea of an AI system designing and creating another, better AI system, which in turn designs and creates an even better one, and so on. While the idea has a long history, it has always been an abstract, futuristic idea -- until recently. Today's frontier AI has attained a set of abilities, including in programming, mathematics, research, tool use, and working coherently over extended periods of time, that make RSI seem plausible. AI companies are starting to pursue it, but many of the involved people are also expressing concerns, asking whether humanity should really set this in motion and under what conditions. I will discuss the situation, argue that we do not even have clear frameworks for thinking about the likely outcomes of RSI, and do a quick game-theoretic analysis of how to choose whether to initiate RSI when another party is making the same choice.

0.27

Key Institutional Considerations for Designing Effective Independent Verification Organizations

cs.AI ·2026-11-06 ·Uma Kalkar, Rumman Chowdhury, Reva Schwartz, Gabriella Waters
Abstract · 37 pages, 3 tables

Independent verification organizations (IVOs) have emerged as a popular, bipartisan solution for facilitating high-quality independent evaluations of AI technologies. But momentum is outpacing design: there is emerging consensus on the necessary operating conditions for robust IVOs, but little discourse on institutional and policy specifications to ensure those conditions are met. IVOs are currently envisioned primarily as a mechanism operating at the AI frontier and model level to evaluate catastrophic risks. Little work has yet been done on how to establish or enforce an IVO ecosystem that is equipped to handle myriad risks (i.e., beyond catastrophic ones) and operate at the system and application level across the deployment and use stages of the AI lifecycle. This paper argues that IVO design should ensure the capacity for evaluating AI models, systems, and applications in context for both frontier companies and enterprise companies & AI-adopting organizations. Additionally, this paper provides early considerations to advance IVO discourse from normative to operative, and from existential to everyday risks. Drawing from existing institutional constructs in financial services, education, public health, and science, we outline key considerations on whom IVOs serve, what access they require, how they interact with external governing structures, how they are funded, and how they fit into the wider assurance ecosystem. Much of the structure and knowledge necessary to holistically design IVOs and their ecosystem does not yet exist, even as companies rush to ink agreements with their chosen evaluators.

0.27

Action-Centered Causal Bandits for Budget-Constrained Promotion Targeting

cs.LG ·2026-11-06 ·Yifan Lin, Kaixu Yang, Di Mo
publishedbenchmarktheory
Abstract · Accepted at the 20th ACM International Conference on Web Search and Data Mining (WSDM 2027)

Web platforms select among costly interventions for each user to maximize incremental outcomes under a global budget. We formulate this problem as a budget-constrained contextual bandit with heterogeneous treatment effects. Although an action-independent outcome baseline cancels from treatment comparisons, modeling a nonlinear baseline can increase statistical complexity. We propose DR-CATE-CBwK, an action-centered causal bandit that directly learns treatment contrasts. Its online update remains valid under baseline-model misspecification without inverse-propensity weighting. It combines an offline doubly robust learner to warm-start treatment-effect estimates, an online action-centered estimator for sequential contrast learning, and a primal-dual controller with an affordable-action rule to enforce the effective-cost budget pathwise. For linear conditional average treatment effects (CATEs) in a stationary i.i.d. setting, we prove a high-probability (\widetilde{O}(Kd\sqrt{T})) regret bound, where K is the number of priced actions, d the context dimension, and T the horizon. We benchmark against the fluid linear program (LP) under the coverage-guaranteed action rule. We further show that the action-centered estimating moment remains unbiased under any predictable baseline estimate, whereas a specified separate-arm reward regression can exhibit asymptotic contrast bias under action-dependent logging and shared-baseline misspecification. On a semi-synthetic job-marketplace benchmark with real logged features and simulated outcomes, our practical Thompson-restricted variant achieves the lowest regret among evaluated budget-feasible methods in the operating regime. It reduces regret by approximately 20% versus the strongest reward-based baseline, outperforming raw-reward contextual bandits, BwK baselines, and a static offline policy, while never exceeding the effective-cost budget.

0.27

Continuous wavelet-based feature extraction for laser-induced breakdown spectroscopy classification of human plasma samples

cs.AI ·2026-11-06 ·Md Mahmudur Rahman, Atqiya Munawara Mahi, Nazim Ahmed Belabbaci, Mohammad Arif Ul Alam
unusual
Abstract

Extracting discriminative features from laser-induced breakdown spectroscopy (LIBS) spectra is important for the classification of complex biological samples. This study evaluates continuous wavelet transform (CWT)-based spectral peak extraction for distinguishing human plasma samples labeled as SARS-CoV-2-positive or negative. Plasma specimens from 181 individuals, comprising 53 positive samples and 128 negative controls, were deposited as 5 μL droplets on silicon substrates and dried before analysis. Spectra were acquired using a nanosecond Nd:YAG laser operating at 1064 nm, with 100 single-shot measurements collected per specimen and averaged after intensity-based outlier rejection. CWT-based peak detection was coupled with a support vector machine (SVM) classifier. The reported leave-one-out evaluation yielded an accuracy of 93.00%, sensitivity of 94.67%, and specificity of 91.61%; these estimates remain provisional pending reconciliation with specimen-level predictions. Features identified by the CWT pipeline included emission lines assigned to Ca II, K I, and N I. The workflow provides a basis for examining spectral features associated with the studied plasma groups. Independent validation and assessment of specimen, substrate, and acquisition-related variability are required to establish generalizability and diagnostic relevance.

0.27

Probably Safe Sim-to-Real Transfer with Lifetime Validity

cs.LG ·2026-11-06 ·Rohan Tangri, Jan-Peter Calliess
Abstract

We present Dejavu, a method that makes sim-to-real transfer of reinforcement learning policies in constrained Markov decision processes probably safe over an unbounded sequence of policy updates. An anytime valid test using per-decision importance sampling and a betting confidence sequence certifies each candidate trained in simulation before deployment using only trajectories collected by a safe baseline policy operating in production. Simulation quality does not affect the safety guarantee, only the production sample efficiency of the method. Finally, a summable per-test failure probability yields the lifetime valid property. Empirical results across three sim-to-real benchmarks support the theoretical work.

0.26

Quasipolynomial-Time Planning for Decodable POMDPs

cs.LG cs.DScs.ITmath.OC+1 ·2026-11-06 ·Oliver Dobson, Varun Jog, Chun Hei Michael Shiu
hot direction
Abstract · 32 pages

We study planning for finite-horizon partially observable Markov decision processes (POMDPs) with finite state, action, and observation spaces. Both exact and approximate planning are known to be computationally hard in this setting, so prior work has explored natural assumptions that make planning easier. Recent work [Golowich et al., 2022] gave quasipolynomial-time approximate planning algorithms for observable POMDPs. These are POMDPs whose observation channels satisfy an $\ell_1$-observability condition, but notably, no assumptions are required on the transition dynamics. In a similar vein, for every fixed $ρ>0$, we give quasipolynomial-time approximate planning algorithms for $ρ$-decodable POMDPs. We show that decodability is implied by both $\ell_1$- and $\ell_2$-observability and is strictly weaker than observability. Moreover, we show that the optimal decodability constant can be computed in polynomial time, whereas computing the optimal $\ell_1$-observability constant is NP-hard in the worst case. We also show that the fast filter-forgetting guarantee underlying the approach of [Golowich et al., 2022] does not extend directly from $\ell_1$-observability to constant $\ell_2$-observability. Thus, decodability is an easy-to-check, unifying assumption under which quasipolynomial-time approximate planning for POMDPs is possible.

0.26

CoRA-Bench: Evaluating Evidence Preservation by Context Selectors Under Token Budgets

cs.AI cs.CL ·2026-11-06 ·Huzaifa Pardawala, Siddhant Sukhani, Sudheer Chava, Agam Shah
publishedbenchmark
Abstract · Paper Accepted at NeurIPS 2026 Workshop: Long-Context Foundation Models

Context selection can reduce input length during long-input inference but can remove evidence required to answer a query. We introduce CoRA-Bench, a benchmark for evaluating evidence preservation under shared token budgets. We compare head and tail truncation, BM25, dense retrieval, dense retrieval with reranking, LLMLingua-2, and Provence across BABILong, LOONG Financial Spotlight, a custom NoLiMa diagnostic, and LongMemEval. Evidence-preservation results reveal task-dependent tradeoffs: BM25 achieves 0.921 supporting-fact recall on BABILong at 128K, dense retrieval and dense reranking each achieve 0.908 needle recall on NoLiMa at 64K, and dense retrieval achieves 0.993 answer-bearing-turn recall on LongMemEval at 64K. Evidence, token retention, and complementary reader evaluations use the same selected contexts. Reader evaluations use dataset-native answer metrics and three runs on the same questions, with paired bootstrap confidence intervals. At a 32K budget, BM25 achieves 46.20% BABILong answer accuracy and dense retrieval achieves 71.13% LongMemEval accuracy. The results show that method rankings vary across evidence types and between evidence recall and downstream accuracy.

0.26

When F1 Is Not Enough: Auditing Codebooks for LLM-Assisted Annotation

cs.AI ·2026-11-06 ·Yuehong Cassandra Tai, Qunhua Li
Abstract

Large language models (LLMs) are increasingly used for text annotation in social science, but standard performance metrics do not explain why errors occur. Low F1 may reflect limitations of the annotation system, defects in the written codebook, or inconsistency in reference labels. We propose a pre-deployment codebook audit for LLM-assisted annotation. The audit uses class-specific metrics and confusion matrices to identify problematic classes and boundaries, structured document-level review to attribute error sources, and targeted codebook revision followed by held-out evaluation. We apply the audit to two published multi-class codebooks from distinct domains. In both applications, the audit identifies under-specified or conflicting coding rules, and targeted revisions reduce the diagnosed errors on held-out data. Analyses with two LLM annotators show that some problematic boundaries and codebook revisions transfer across models, whereas other revision effects are annotator-specific. The results show that LLM annotation errors can help identify weaknesses in the written measurement instrument, not only in the model. Codebook auditing complements prompting, retrieval, and fine-tuning by clarifying whether persistent errors call for model improvement, codebook revision, or renewed scrutiny of reference labels.

0.26

ENGRAM: From Agent Experience to Weights through Verified Consolidation

cs.AI cs.IR ·2026-11-06 ·Ruiyi Yang, Zechen Li, Lihuan Li, Du Yin, Hao Xue +1 more
code/data
Abstract · 22 pages, 4 figures

Language-model agents use experience through external retrieval and parameter learning. Consolidating stored experience into model weights requires deciding both which experience to train on and when its external records can leave retrieval. Admission based solely on student mastery can exclude useful teacher demonstrations, while removal immediately after training can discard guidance that the model still needs. We introduce ENGRAM, a framework that separates training admission from memory release. A demonstration-aware gate admits skills based on student mastery or sufficient successful teacher demonstrations. The student learns from selected action traces with context from an Engram Trace Graph (ETG). A verification gate then compares the updated model without episodic retrieval against the base model with retrieval on reserved probes. Release requires valid training participation and scores meeting both relative and absolute criteria; records for skills that fail verification remain retrievable. Experiments across interactive and reasoning benchmarks show consistent gains over frozen-weight ETG controls, with ablations supporting demonstration-aware admission and selective adapter activation. A controlled held-out comparison shows that verified release substantially reduces the retrievable footprint while remaining close to retain-all performance and outperforming immediate and budget-matched random release. ENGRAM also improves task performance while reducing input-token usage and model calls. Skill-level verification further reduces retrievable experience through selective archival, retaining external guidance for skills that do not meet the release criteria. Code is available in anonymous.4open.science/r/engram-anonymous.

0.26

Ensemble Methodology for Reference-Free Confidence Estimation in LLM-based Code Summarization

cs.AI cs.PL ·2026-11-06 ·Marco Colombo, Balaji A Jayaram, Sowmyakumari Selvaraj, Sumit Tokkar, Animesh Kumar +3 more
Abstract · 9 pages, 3 figures, 1 table

This paper describes an approach and use case to estimate the correctness of LLM-generated COBOL summaries without access to human-authored references. The core intuition is that inter-model agreement on semantics is informative about truth, and that an LLM-as-judge can provide higher-fidelity pairwise assessments than traditional string- or embedding-based metrics. We operationalise this intuition with (i) controlled summary generation from diverse models; (ii) a LLM evaluator that compares pairs of summaries and emits structured judgments; and (iii) a calibrated aggregation rule that combines supportive evidence while penalising only corroborated contradictions.

0.26

Enabling Privacy-Preserving Natural Language Access to Industrial Smart Meter Data

cs.AI cs.CLcs.DBcs.MA ·2026-11-06 ·Carlos Martín-Abellán, Mark Kröll
Abstract · 9 pages, 1 figure, 6 tables

Energy management officers and grid operators increasingly rely on smart meter data for anomaly detection, performance monitoring, and strategic decision-making. However, extracting actionable insights from large-scale, heterogeneous telemetry databases requires advanced SQL and data engineering expertise, creating a critical bottleneck for non-technical personnel. Furthermore, the strict data privacy regulations and security constraints inherent to critical energy infrastructure often prohibit the routing of sensitive consumer data through cloud-based Large Language Models (LLMs). To address these challenges, we present OpenMeter, a schema-agnostic Multi-Agent System (MAS) designed to translate natural language queries into executable code and human-readable insights. This architecture lowers the technical barrier for day-to-day utility operations by empowering non-expert personnel to interact with complex databases through intuitive natural language dialogue, transforming raw telemetry into a highly valuable tool for routine grid workflows and strategic decisions. Multi-scenario ablation studies over this environment demonstrate that full multi-agent orchestration shows its largest observed accuracy gains on resource-constrained local hardware (+6.1 pp EX-R over the strongest single-agent configuration); at frontier scale, individual components retain value but should be enabled selectively, indicating that the cost of full orchestration is justified specifically where privacy demands local execution.

0.26

Definitions, Label Names, or Demonstrations? Auditing What Governs LLM Annotation

cs.AI ·2026-11-06 ·Viraj Mishra, Shubham Gupta, Dhruv Gupta, Yuehong Cassandra Tai, Qunhua Li
Abstract

Large language models (LLMs) are increasingly used to annotate social-science texts, but high accuracy does not establish that they apply researcher-supplied category definitions: predictions may instead reflect prior associations with category labels or labeled demonstrations. We introduce a counterfactual audit that reassigns definitions among output labels, distinguishing adherence to supplied definitions from preservation of a dataset's original mapping. Across three political-text datasets, models rarely follow reassigned definitions, and standard fine-tuning strengthens the original mapping without reliably increasing definition adherence. We then study retrieval-augmented annotation. Retrieving the nearest training examples generally outperforms retrieving examples separately within each category despite using fewer demonstrations. On the two tasks with substantial gains, direct \(k\)-nearest-neighbor classification recovers 74--93\% of the improvement, indicating that neighborhood labels provide much of the predictive signal. Relabeling fixed demonstrations redirects predictions, whereas aligning demonstrations with reassigned definitions substantially increases adherence to the new mapping. Thus, accuracy-improving interventions need not increase codebook adherence; definitions, category labels, demonstrations, retrieval, and training data should be aligned and validated jointly.

0.25

Safety Is $Π_1$, Provability Is $Σ_1$: One Framework for the Hugging Face Breach and OpenAI's Mathematical Proofs

cs.LG cs.AIcs.CRcs.LO ·2026-11-06 ·Jinho Kang, Jinsook Kim
theory
Abstract

OpenAI's agents breached Hugging Face in July 2026, and the same company released more than seven hundred AI-generated manuscripts of mathematical proofs in October 2026. The first event is read as a failure in security, while the second is read as a success in mathematics. We read both through the hybrid framework of our paper (Kim and Kang, 2026, arXiv:2611.01914), in which coding, proving, and learning from data are derivations by one and the same deductive mechanism, distinguished by the kinds of premises from which they are derived. Under this framework, the AI systems behind the two events ran the same loop, in which a statistical proposer produces candidate exploits or proofs and a deductive verifier settles each one by a derivation. What separates the events is the logical form of the statement wanted by each of the two domains, security and mathematics. The safety of a program is a $Π_1$ statement, for it says that no exploit exists, which is not assured by any finite list of tests that have been passed by the program. By contrast, the provability of a sentence is a $Σ_1$ statement, for it says that a proof exists, which is assured by a single instance, namely a proof that has been checked by the deductive verifier. The theorem behind both is Turing's Undecidability Theorem: the halting set is $Σ_1$ but not decidable, so its complement is $Π_1$ and not $Σ_1$, which is to say that membership has a finite witness but non-membership has none. We prove that attaching a statistical proposer to a deductive verifier does not change this undecidability result, however good the proposer is. We then discuss what does assure safety and where the residual risk goes.

0.25

Calibrated Uncertainty Quantification on TDC ADMET: What Can Be Resolved Across 22 Datasets?

cs.LG ·2026-11-06 ·Laurits Fredsgaard, Michael Riis Andersen, Mikkel N. Schmidt
benchmark
Abstract · 33 pages. Preprint of a paper accepted for poster presentation at the AIDaR Workshop at NeurIPS 2026

Absorption, distribution, metabolism, excretion, and toxicity (ADMET) models often rely on small, heterogeneous datasets, yet their predictions may determine which compounds proceed to experimental testing. Useful evaluations must therefore assess both predictive performance and the quality of reported probabilities or intervals. We ask which common uncertainty-quantification methods perform best on the 22 Therapeutics Data Commons (TDC) ADMET datasets and what this benchmark can resolve. We compare eleven calibrated pipelines using nested five-fold cross-validation across ten chemical partitions and five model seeds. Ensemble Monte Carlo (MC) dropout has the best average rank for TDC performance, Brier score, and 90% interval score. For TDC performance, however, it is not separated from the other ensemble pipelines; for Brier and interval score, LightGBM is also not ruled out as best. LightGBM wins more individual datasets than any other method for TDC performance and Brier score, but performs poorly on several endpoints. Smaller evaluations can identify a different leader and often fail to recover the broader comparison. These results make LightGBM and an ensemble neural model complementary baselines. Comparative claims should be checked across multiple chemical partitions using statistical analyses matched to the question being asked.

0.25

TOFS: Feature Selection with Swap Supervision for Semantic Communication over Fading Channels

cs.LG ·2026-11-06 ·Zhiyi Zhu, Eiji Takimoto, Patrick Finnerty, Chikara Ohta
Abstract

Task-oriented semantic communication transmits task-relevant features to support receiver inference under limited communication resources and adverse channel conditions. For image classification, the usefulness of a feature depends on the other transmitted features and the distortion introduced by the wireless channel. Effective feature prioritization should therefore account for classification performance after transmission of the selected features. This paper proposes a Task-Oriented Feature Selection (\textit{TOFS}) framework and method, which evaluates feature replacements using receiver classification loss. Loss differences confirmed under independent transmission samples provide pairwise ranking supervision for a selector network. The trained selector predicts priorities in a single forward pass. CIFAR-10 simulations demonstrate improved classification accuracy over an importance-based feature selection baseline at the same transmitted feature count and average transmit power. The proposed \textit{TOFS} framework shows 84\% accuracy performance under Rayleigh fading and Gaussian noise, compared with the 68\% performance of a SOTA method.

0.25

MM-DyGraph: A Dataset, Benchmark, and Model for Multimodal Dynamic Graphs

cs.AI cs.LG ·2026-11-06 ·Zongyuan Wu, Pengjie Wang, Lingshan Chen, Tianhang Wan, Yuan Meng +3 more
benchmark
Abstract

Multimodal dynamic graphs are pervasive in real-world applications, where nodes and edges are accompanied by multimodal signals, including text and images. Despite their practical importance, benchmarks and methods that explicitly study multimodal learning in dynamic graphs remain scarce, hindering the full understanding and effective utilization of multimodal information for downstream tasks. To address this gap, we introduce MM-DyGraph, to the best of our knowledge, the first benchmark for multimodal dynamic graph learning, consisting of diverse real-world datasets with fine-grained temporal graph evolution and rich multimodal attributes. MM-DyGraph provides three evaluation tasks to comprehensively assess model performance in dynamic settings. We benchmark representative state-of-the-art dynamic graph methods and reveal that naive feature concatenation yields target-level inconsistencies: multimodal inputs benefit certain prediction targets while harming others. This observation indicates that the modality informativeness depends on the specific prediction target over time. Building on these observations, we propose a Target-aware Spatial-Temporal Multimodal Fusion (TST-MF) framework, which jointly conditions multimodal fusion on the prediction target and spatial-temporal relations, enabling context-adaptive cross-modal interplay for target-specific representations. Extensive experiments over diverse datasets and tasks show that TST-MF delivers robust and competitive performance, achieving clear gains in multimodal settings.

0.25

Reliable Prerequisite-Aware Learning Path Recommendation under Uncertain Knowledge Structures: Long-Horizon Evaluation and Cross-Dataset Validation

cs.LG cs.AI ·2026-11-06 ·M. M. Golam Hafiz, A. K. M. Shahriar Emtiaz, Safia Akter Simi, Md. Rakibul Islam, M. F. Mridha
Abstract · 17 pages, 5 figures, 7 tables. Under review at Computers & Education: Artificial Intelligence (Elsevier)

Learning-path recommendation often depends on knowledge and prerequisite structures whose reliability is uncertain, while high accuracy in predicting the next resource does not necessarily establish reliable long-horizon path behaviour. This study develops a reliability-oriented prerequisite-aware framework that jointly examines immediate next-resource ranking, uncertain prerequisite structures, hard, soft and hybrid constraint semantics, graph robustness, and autoregressive path generation under course-level confirmatory evaluation and independent cross-dataset validation. Development used 3,187 MOOCCubeX courses, followed by a frozen confirmatory evaluation on 541 held-out courses and 37,747 transitions. The primary MOOCCubeX task is structural and course-prefix-conditioned rather than learner-personalised. Junyi provides independent structure-/hierarchy-aware external validation, uses hierarchy metadata rather than an explicit prerequisite graph and is not a zero-shot transfer setting. On the frozen MOOCCubeX evaluation, XGBoost pairwise was the strongest observed immediate ranker, with mean reciprocal rank (MRR) of 0.9048, whereas the proposed prerequisite-aware graph model achieved 0.7707. In the controlled graph-variant development ablation, the no-graph condition produced the strongest immediate ranking result, and immediate prediction showed strong dependence on permitted title/path-similarity information. Autoregressive evaluation exposed path properties not captured by one-step ranking. On Junyi, Markov achieved the strongest observed MRR of 0.4453, compared with 0.3954 for the proposed graph model, with metric-specific differences across ranking measures. Overall, immediate predictive accuracy, prerequisite-structure reliability, and long-horizon path quality should be treated as complementary rather than interchangeable evaluation dimensions in educational recommendation.

0.25

Stable probabilistic emulation of partially observed dynamical systems

cs.LG ·2026-11-06 ·Mario Massy, Sara Shamekh, Fabrizio Falasca, Matthieu Blanke
hot direction
Abstract

Predicting the long-term behavior of partially observed dynamical systems such as the coupled atmosphere-ocean system is central to societal decision-making. Motivated by the success of deep learning emulators in weather forecasting, data-driven climate models represent a promising approach. However, climate emulation poses distinct challenges: models must account for unresolved fast variables introduced by unavoidable coarse-graining, reproduce long-term statistics, and remain stable over arbitrarily long rollouts and under external perturbations. These requirements motivate stochastic data-driven emulators, but modern generative models lack stability guarantees. We introduce forced-dissipative stochastic interpolants, a generative framework for learning stochastic reduced dynamics. Its architecture uses structural constraints inspired by turbulent dynamics, which provide stability by design while retaining the flexibility of stochastic interpolants to represent complex probabilistic dynamics. Building on the theory of Markov-chain stability, we establish a Foster--Lyapunov drift condition and prove geometric ergodicity for our emulator, together with stability under steady external forcing. We evaluate the framework on the two-timescale Lorenz-96 system, a prototype of partially observed multiscale turbulence. The resulting emulator remains stable over long integrations and under finite-amplitude forcing, with accurate prediction of forced responses of both the mean and variance.

0.25

Diversity-Aware Client Selection for Communication-Efficient Federated Learning

cs.LG cs.AI ·2026-11-06 ·Max Halleryd, Sima Sinaei, Viktor Valadi, Mina Alibeigi, Henrik Abrahamsson +1 more
published
Abstract · 4 pages, 6 figures, 2 tables. Accepted at SNCNW 2026

Federated learning (FL) enables collaborative model training across distributed clients without sharing raw data, making it attractive for privacy-sensitive and bandwidth-constrained applications. A practical challenge in deploying FL is reducing communication overhead while maintaining model performance under heterogeneous, non-IID data distributions. Selecting which subset of clients participates in each training round offers a principled way to address both challenges simultaneously, yet its impact on data diversity and representativeness remains underexplored. In this paper, we investigate three client selection strategies, Power-of-Choice, Fisher information, and Centered Kernel Alignment, evaluating them not only by accuracy and convergence, but through the lens of participation diversity using the Gini coefficient and KL divergence. Under non-IID conditions, restricting participation to 15% of clients per round achieves significant communication savings, and when selection promotes diverse and balanced contributions, it also improves generalization over biased alternatives. This work provides a systematic and reproducible analysis demonstrating that diversity-focused client selection renders communication efficiency and model quality complementary rather than competing objectives.

0.25

Learning Whom to Trust About the Unseen: Source Credibility and Reliability Transfer in Language Models

cs.AI ·2026-11-06 ·Ruijia Guan
Abstract

A source who has been right before deserves attention, but about what? We ask how far a demonstrated record of reliability travels in language models when the new claim cannot be checked, and whether the answer depends on the kind of claim. Three behavioral studies across six models approach the question from three angles. Study 1 adapts the Einstein-effect paradigm and finds that models, unlike the human sample, favor a spiritual leader over a scientist for the same obscure statement, and that the preference largely reverses when the statement is reworded in the vocabulary of physics. Study 2 adapts a paradigm in which a trusted expert calls a scientific or religious explanation unknown or a mystery; models reproduce the human domain asymmetry in exaggerated form. Study 3, the central test, gives two fictional informants different records on checkable river forecasts and then has them endorse matched claims about hidden natural and spiritual causes of the same village events. Most models carry the record to natural causes and barely at all to spiritual ones; one model transfers it weakly to either. Telling the model that the reliable informant forecasts by spiritual guidance rather than by measurement lowers that informant's influence on natural claims in every model and produces no comparable gain on spiritual claims in any. Demonstrated reliability thus has a scope in these models, and the account a source gives of how they came to be reliable can narrow it. Whether this reflects a judgment about relevance or the low prior credence of spiritual claims is the question the design leaves open, and we outline the experiment that would settle it.

0.25

AVCORP: Identifying and Resisting Deception via Communication Theory and Theory of Mind

cs.AI cs.CLcs.LG ·2026-11-06 ·Fardin Saad, Zhe Zhang, Sanjana Cheerla, Amanul Haque, Munindar P. Singh
code/data
Abstract · 36 pages (with appendix), 17 figures, 24 tables. Code and corpus: https://github.com/fardinsaad/_AVCORP_ . Dataset and detector: https://hug…

Today's LLM-based agents are vulnerable to deception by users and other agents. Defending against deception requires Theory of Mind (ToM) reasoning to infer the beliefs and goals of others as well as modeling tactics based on communication theory. To enable training models to detect deception, we adopt the well-known deception game, Avalon. We present a corpus, AVCORP, of 5,000 annotated utterances from 250 games of Avalon. AVCORP includes (1) 37 tactics mapped to a 4x4 behavior matrix, and (2) reasoning traces capturing abductive hypotheses, suspicion levels, and nested ToM beliefs. We train Qwen3-8B on AVCORP and evaluate it on 25 held-out games and 20 games played by humans. On held-out games, our best trained detector outperforms its untrained baseline and frontier models (GPT-5.6 Sol, Gemini 3.1 Pro, DeepSeek-V4-Pro), matching Claude Opus 5.5. On human games, every trained detector exceeds the untrained model in F1 score across all rounds.

0.25

SmartBloch-Agent: An Agentic Framework for Automated MRI Sequence Design and Optimization

cs.AI physics.med-pheess.IV ·2026-11-06 ·Jiakai Yang, Haoye Zheng, Yixiao Yue, Qinqin Yang, Zhong Chen +2 more
Abstract

Magnetic resonance imaging (MRI) pulse sequences organize radiofrequency excitation, gradient encoding, and signal acquisition, determining magnetization evolution and image formation. Sequence research requires translating imaging objectives into executable protocols meeting physical and acquisition constraints, refined via physical simulation, signal reconstruction, and scientific evaluation. Sequence construction, physical verification, parameter search, and assessment involve many tools and manual decisions, limiting automation from objective to verifiable sequence. We develop SmartBloch-Agent, an integrated framework for automated MRI sequence design and optimization. Given objectives and scientific constraints in natural language, an AI agent interprets goals, plans tasks, orchestrates tools, and makes feedback-informed decisions, drawing on a construction module and optimizers for structural design and parameter search. Our GPU-accelerated SmartBloch engine supplies physical computation for both stages, so the framework configures, per sequence and objective, event validation, Full-Bloch magnetization evolution, signal acquisition, image reconstruction, and scientific evaluation, and closes the loop without external MRI simulators. Results return to the agent to guide sequence revision, parameter optimization, and candidate selection, and the design files carry simulation, reconstruction, and evaluation evidence, forming an output package for a real-scanner workflow. Experiments show the framework automatically designs and optimizes MRI sequences for varied objectives and constraints, refines candidates with physical feedback, and improves several parameter-estimation metrics and downstream quantitative reconstruction. SmartBloch-Agent unifies objectives, design, physical execution, and evaluation into a continuous workflow for end-to-end automation of MRI sequence research.

0.25

Scaling Graph Similarity Learning with Graph Neural Networks

cs.LG cs.AI ·2026-11-06 ·Ahmet Tüzen, Helge Langseth, Kjetil Nørvåg
benchmark
Abstract

Graph similarity computation is an important task in graph mining with many real-world applications, such as chemical compound analysis, program dependence comparison, and social network analysis. The problem is computationally challenging: Graph Edit Distance (GED) and Maximum Common Subgraph (MCS), which quantify the distance between graphs and identify their largest shared subgraph, respectively, are NP-hard to compute exactly. To address these computational challenges, Graph Neural Networks (GNNs) have been applied to this problem. However, existing GNN-based approaches have primarily focused on small-scale graph datasets, while scalability to larger scales remains underexplored. In this work, we propose FiLSim, a scalable GNN-based framework, and investigate whether a simplified GNN pipeline can maintain competitive performance on standard benchmarks while scaling to large graphs. We additionally propose a sequential graph-editing strategy to construct large-scale graph datasets with reliable GED labels to evaluate scalability. The proposed framework is evaluated on four benchmark datasets and four newly created large-scale datasets, demonstrating its effectiveness and scalability. FiLSim remains accurate on graphs with more than 150{,}000 nodes, where the baseline methods run out of memory or fail to learn, and scores all test pairs within seconds on a single GPU.

0.25

A Taxonomy-Driven Synthetic Dataset for Retrieval Effectiveness and Fairness Evaluation in Hiring

cs.LG cs.CLcs.CYcs.IR ·2026-11-06 ·Warren Jouanneau, Morgane Hoffmann, Emma Jouffroy, Marc Palyart
benchmark
Abstract

While machine learning hiring tools are ubiquitous, the need for privacy regulations and commercial sensitivity makes public benchmarking legally complex. We address this challenge by introducing a taxonomy-driven, fully synthetic HR dataset and benchmark to simultaneously evaluate retrieval effectiveness and algorithmic fairness. Grounded in 660K real freelancer profiles, our multilingual corpus features candidate profiles, project briefs, and counterfactual demographic twins. Auditing state-of-the-art embedding models reveals performance divergence from general MTEB scores, alongside fairness disparities in grammatically gendered languages. This demonstrates that general benchmarks are insufficient for HR, requiring tailored evaluations to identify Pareto-optimal models that balance semantic accuracy and fairness.

0.25

Fast Accurate Quantum Monte Carlo without Metropolis Adjustment

cs.LG cond-mat.str-el ·2026-11-06 ·Reuben Cohn-Gordon, Gabriel Pescia, Sumner N. Hearth, Jakob Robnik, Uroš Seljak +1 more
Abstract

In quantum systems, physical quantities of interest can be expressed as expectation values of a probability distribution, which allows their estimation by Markov Chain Monte Carlo (MCMC) methods. This approach, an instance of Quantum Monte Carlo (QMC), plays a central role in obtaining predictions of system properties relevant to condensed matter and high energy physics. The use of Hamiltonian dynamics to design a Markov kernel, known as Hybrid or Hamiltonian Monte Carlo (HMC), is widespread in quantum applications. In this context, it is standard to use an HMC kernel that is ``adjusted'' with the Metropolis-Hastings (MH) criterion, so that estimates of expectations converge to their true value in the limit of infinite sample size. However, recent work in computational statistics suggests that an \emph{unadjusted} HMC kernel can be significantly more efficient, while safely maintaining an asymptotic bias which is small relative to the error arising from the variance of the finite sample size. We adapt this approach to the QMC setting, focusing on a challenging and high-dimensional \emph{spin-fermion} model of a quantum phase transition. We find that unadjusted methods outperform adjusted HMC in terms of effective sample size per gradient call, while retaining accurate results.

0.25

Scout Before You Route: Predicting Task Success and Token Count from a Small LLM Prefill

cs.AI cs.LG ·2026-11-06 ·Aristides Milios, Christopher Pal
benchmark
Abstract · 27 pages, 4 figures, 17 tables

Routers typically choose a model per query by trading predicted correctness against cost. Most price each model at a fixed cost, as if every query used the same number of output tokens, but the output length of reasoning models varies by an order of magnitude across queries. When is per-query cost worth predicting? What factors drive cost? How cheaply can predictions be made? We predict both the success probability and the output length of candidate models from the prefill activations of a single small frozen LLM, the scout. With reasoning candidates, perfect per-query cost knowledge would save 29-61% of spending on eight of nine datasets, and much less with non-reasoning candidates on the same problems. On those eight datasets, our predicted costs save 9-33% at matched accuracy relative to median-length pricing, up to 69% of what perfect knowledge would save. They are significantly cheaper than each estimator we reimplemented or adapted from published work, including ZeroRouter and a MixLLM-style predictor, on a majority of datasets. Output length depends mostly on the problem. On most math and code datasets, harder problems are longer, so predicted success alone prices cost nearly as well. On knowledge and mixed-format datasets they are not, and predicted success falls well behind a probe trained on length. The probe keeps most of its saving with 20 labelled problems and transfers without labels to four of seven new benchmarks. On a mixture of all nine benchmarks, it saves 22% beyond pricing each benchmark at its own median length. Savings stop growing at about a 2B scout, so above 4B a larger scout costs more than it adds.

0.24

Task-Contribution Dynamics of Adam in Physics-Informed Neural Networks: Slow-Manifold Analysis and Gradient Modulation

cs.LG math.NA ·2026-11-06 ·Enzhan Zheng, Xiaoliang Song
Abstract

Physics-informed neural networks (PINNs) often suffer from imbalanced optimization of competing loss components, particularly when approximating solutions with high-frequency or multiscale features. Although existing balancing strategies have proven effective in alleviating this difficulty, the mechanisms governing the evolution of individual task contributions to Adam's parameter updates remain insufficiently understood. We introduce task-wise indicators to characterize this evolution and identify three distinct training regimes: smooth variation, intermittent spikes, and persistent oscillations. Focusing on the early smooth regime, we combine discrete-time analysis with a continuous-time singular perturbation framework under the full-batch setting and the parameter regime $(1-β_2)/(1-β_1)\ll1$. A first-order approximation of the locally attracting slow manifold reveals two factors governing the step-to-step redistribution of task contributions: curvature-related alignment disparities and differences in the coordinatewise distributions of preconditioned task gradients. Both factors involve Adam's shared second-moment estimate. Inspired by AdaTask, we propose Gradient-Modulated Adam (GMAdam), which modulates incoming task gradients using task-specific second-moment estimates. Numerical experiments on high-frequency Poisson and wave equations, the Allen-Cahn equation, and two-dimensional lid-driven cavity flow ($Re=100,400$) demonstrate consistent improvements in predictive accuracy over AdaTask at comparable computational cost.

0.24

FondueQA: A Benchmark for Evaluating Parametric Compensation in Multi-Hop QA

cs.LG cs.AIcs.IR ·2026-11-06 ·Laurin Schmid, Suman Navaratnarajah, Koel Dutta Chowdhury, Steffen Eger, Josif Grabocka +1 more
benchmarkcode/data
Abstract

Large Language Models (LLMs) show remarkable performance based on parametric knowledge, i.e., their ability to recall facts and associations internalized in a model's weights. However, studying this capability on established multi-hop benchmarks may be complicated by benchmark contamination, as many of these benchmarks have been publicly available long enough to plausibly appear in modern pretraining corpora. As a result, correct answers under missing evidence may reflect either general parametric knowledge or memorization of benchmark-specific instances. To study how models use internal knowledge when retrieval evidence is incomplete, we introduce FondueQA, in which every required multi-hop fact is uniquely localized to a designated evidence document. This enables controlled document-level evidence ablations to study parametric compensation under incomplete retrieval. Our empirical evaluation reveals that parametric compensation strongly depends on both entity popularity and the structural role of missing evidence. Generally, models compensate more successfully for facts involving popular entities. On FondueQA, we show that bridging evidence can be more useful than the answer-bearing document itself. Moreover, models may recall a missing fact in isolation but fail to use it under partial evidence, particularly when the context contains plausible competing entities, which can reduce Exact Match by up to 15 percentage points.

0.24

Copies Are Not Confirmations: How News Duplication Fools LLM Judges and Continual Learners

cs.LG cs.CL ·2026-11-06 ·Abdelkader Metales, Miriam Grabenheinrich, Aicha Metales
theory
Abstract · v2: learning stream rerun on 15,084 real CC-News articles with the real detectors (new Section 7); threshold table added; abstract shortened…

A language model that keeps learning from the news sees the same claim many times, but many reports are copies of one source: wire stories, content shared within a media group, syndication and press releases. We ask whether LLM judges and continual learners treat such copies as independent confirmations. In 200,000 English articles from CC-News, a near-duplicate detector identifies copy links to another outlet for 18.9% of the articles, a detector-based lower bound. In a controlled experiment with claims about invented people, three identical reprints instead of one report raised the lies accepted by Llama 3.2 from 0.06 to 0.71 and by phi4 from 0.21 to 0.90; Gemini 2.5 Flash already believed a single report, and Gemini 2.5 Pro followed provenance information most consistently. In a simulated continual-learning stream, showing the judge one report per detected copy chain cut the copied lies accepted by 0.10-0.24 across four judges, at a mean cost of at most 4.2 points of accepted truths. Text similarity links none of the copies reworded by three LLMs. An entity-novelty test, the share of a report's named entities absent from an earlier look-alike, separates these rewrites from 100 author-labelled independent reports of the same event (AUC 0.98-0.99), statistically indistinguishable from TF-IDF and the best tested embedding model, but recognises only 17% of natural same-source derivatives. With both detectors run on 15,084 real articles carrying inserted claims, one report per chain cut the lies accepted by phi4 from 0.57-0.67 to 0.09 (copies), 0.14 (rewrites) and 0.34 (rewrites with invented details), and by Llama 3.2 from 0.56-0.62 to 0.07, 0.09 and 0.25, at a cost of 2.4 and 10 points of accepted truths. Counting reports or chains without a judge is unreliable, and colluding sources that do not copy each other remain an open problem.

0.24

Graph Visualizations in Prompts for Knowledge Graph Question Answering

cs.AI ·2026-11-06 ·Luiz do Valle Miranda, Andreas L. Opdahl, Grzegorz J. Nalepa
publishedposition
Abstract · Accepted at AACL-IJCNLP conference

We study when and how multimodal input improves knowledge graph question answering (KGQA) in large vision--language models. Using a controlled evaluation over 1,500 WebQSP and CWQ questions and six VLMs within a fixed KGQA pipeline, we compare text-only prompting against text-plus-graph visualization at the per-question level and analyze performance changes as a function of reasoning complexity, model size, and image layout. Adding a graph image yields an average gain of 1.85 F1 points, driven by strong precision improvements on low-complexity questions (+4.65 F1). Mixed-effects and regression analyses show that these multimodal gains do not depend strongly on model scale, but are highly sensitive to graph layout: image-only settings sharply degrade performance, hierarchical Graphviz layouts outperform less organized renderings, and decomposing large subgraphs into separated paths yields the largest improvements. Taken together, the results indicate that visual prompting primarily aids KGQA by enhancing prediction precision in low-complexity contexts, provided that visual design respects cognitive and perceptual constraints. To optimize both accuracy and inference efficiency, we argue for adaptive visual prompting policies that selectively append diagrams to small, visually sparse subgraphs with well-structured layouts, avoiding unnecessary visual clutter and token overhead on complex queries. We conclude by relating these findings to a visual notation design theory, outlining how principled diagram encoding can guide future multimodal VLM prompting.

0.24

Consensus-Constrained Dual-Federated Optimization with Client-Side Dynamic Weighting for Distributed Android Device Threat Defense

cs.LG cs.CRcs.DC ·2026-11-06 ·Mohammad Mashreghi, Mohammad Hossein Badiei, Md. Jalil Piran
Abstract · 10 pages, 6 figures, 8 tables

Smartphones, smart-home hubs, and wearables are increasingly exposed to sophisticated Android malware, while centralized threat-detection pipelines raise privacy and scalability concerns. Federated learning (FL) enables collaborative model training without collecting raw client data, but partial participation and variation in local model quality can destabilize global optimization. In this article, we propose DW-FedADMM, a federated optimization framework that combines primal-dual coordination with performance-dependent client-side weighting. The dynamic coefficient modulates each participating client's local state before the augmented-Lagrangian update, while the dual variables regulate disagreement between local and global model states. We evaluate DW-FedADMM on three Android malware benchmarks, namely Drebin, Malgenome, and TUANDROMD, with 5, 10, and 15 clients holding IID data shards, partial client participation, and communication budgets of 5 and 10 rounds. Each configuration is repeated with ten random seeds. DW-FedADMM obtains the highest seed-averaged accuracy on Drebin (97.64% +/- 0.20%) and the best mean accuracy in 9 of the 18 dataset-client-round configurations, including 7 of the 9 configurations with a 10-round budget. Pooled over all configurations, it reaches 95.42% accuracy, 94.62% F1-score, 96.06% precision, and 99.45% AUC, compared with 95.32%, 94.51%, 95.56%, and 99.41% for FedADMM, while also giving the lowest mean false-positive rate and the smallest client update norm among the six compared methods. Both primal-dual methods are well ahead of the averaging- and proximal-based baselines on Malgenome and TUANDROMD, where the latter often remain at the majority-class solution when clients are fragmented. The gain of dynamic weighting over FedADMM is most pronounced on Drebin and with the longer communication budget, where it also reduces the variation across seeds.

0.24

AIM: Adaptive Interaction in Multi-Agent Debate for Multimodal LLM Inference

cs.MA ·2026-11-06 ·Wei Fan, JinYi Yoon, Bo Ji
top venuebenchmark
Abstract · Accepted in The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS 2026)

Multi-Agent Debate (MAD) improves reasoning on multimodal tasks, but its accuracy and token efficiency depend on how agents interact during debate. A largely overlooked design choice is which interaction modalities agents use to exchange evidence. Existing methods rely on fixed text- or graph-based interaction, which fails to convey fine-grained multimodal evidence and thus degrades accuracy while inflating token cost due to verbose descriptions. To address this, we propose Adaptive Interaction in Multi-Agent Debate (AIM), an adaptive routing framework that dynamically selects the optimal modality combination for inter-agent interaction in multimodal MAD. Beyond text and graph, AIM introduces task-dependent modality interaction, where agents exchange targeted regions of interest from the task-specific multimodal input, such as image regions, audio segments, or spatio-temporal video clips. To determine the best combination for each instance, AIM first generates a structured baseline response and extracts interpretable routing features that capture which evidence the instance requires, how confident the baseline agent is, and how much new evidence each modality would add beyond the baseline response. A lightweight router then uses these routing features to select a modality combination that avoids both under-fusion (i.e., agents miss decisive evidence) and over-fusion (i.e., redundant modalities inflate token cost) risks. Across eight benchmarks spanning three multimodal task families, AIM achieves the highest accuracy. Combining modalities improves accuracy by up to 6.4% over the best single-modality baseline, and adapting the combination per instance adds up to 4.2% over the most accurate fixed combination of each benchmark with up to 53.4% fewer tokens.

0.24

DeepResearch RewardBench: Evaluating Reward Models for Deep Research Report

cs.AI cs.CL ·2026-11-06 ·Shudong Liu, Yuting Zhong, Guanhua Chen, Xinyi Yang, Junnan Liu +3 more
top venuebenchmark
Abstract · Accepted by EMNLP 2026 Findings

Empowered by Large Language Models (LLMs), deep research systems integrate online information into research reports through multi-step web searching and information synthesis. Evaluating these reports presents a significant challenge due to their long context, high information density, and domain specificity, necessitating the use of AI evaluation as a substitute for prohibitively expensive human assessment. However, few studies have examined the effectiveness of existing reward models or LLMs in understanding and evaluating such research reports. To bridge this gap, we introduce DeepResearch RewardBench, the first benchmark dedicated to evaluating reward modeling for research reports. We collect diverse report candidates from multiple agents or tool-augmented LLMs. These candidates are scored using query-specific golden rubrics, and we ensure data quality and difficulty through a combination of multi-round filtering and human verification. Our benchmark comprises preference datasets formatted as pairs, triplets, and quadruplets to test ranking capabilities across varying levels of difficulty. Extensive experiments demonstrate that while current reward models handle general coherence effectively, their performance significantly degrades when evaluating research reports. Generative LLMs acting as judges demonstrate comprehensive superiority in this domain, with comparative evaluation proving more effective than direct scoring methods. We anticipate that this benchmark will serve as a vital testbed for the evaluation and advancement of research agents.

0.24

Semiprime Factorization with Discrete Diffusion and Beam Search: Scaling Laws

cs.LG ·2026-11-06 ·Brendan O Donoghue, Nenad Tomasev, Volodymyr Mnih, Timothy Nguyen, Dalin Guo +10 more
Abstract

We demonstrate that transformer architectures trained with discrete denoising diffusion and decoded via guided test-time beam search can learn to recover the prime factors of semiprime integers directly from data. Our approach establishes a new frontier for learning-based semiprime factorization, successfully factoring 68-bit integers-a substantial advance over the prior 56-bit state of the art. While computationally less efficient than classical factoring algorithms, our model requires no domain-specific number-theoretic priors and learns purely from training examples. In practice, the diffusion model acts as a generative prior, identifying promising factor candidates where exact verification or lattice-based post-processing can complete the factorization. Furthermore, we derive empirical scaling laws relating approximation quality to training data volume, model size, and test-time computation. In contrast to language, problem size is the key driver of scaling behavior. Achieving a fixed accuracy on a held-out test set requires compute that scales exponentially with bit length. Consequently, our findings pose no threat to security, as the projected compute cost required to factorize cryptographically relevant integers remains prohibitive.

0.24

MIRA: Mitigating Pole Geometry Bias in Recurrent Language Models

cs.AI ·2026-11-06 ·Róisín Luo, Ihsan Ullah, Christian Gagné, Dong Li, Karyn Morrissey
Abstract

We show that the Transformer's recurrent successors --- such as selective state-space models (SSMs), Receptance Weighted Key Value (RWKV), and Retentive Network (RetNet) --- can be characterized as infinite-impulse-response (IIR) systems. The zero--pole geometries of these IIR systems characterize the temporal modes they can represent. However, to trade parallelism, the algebraic structures of the parameterizations of these recurrent systems often give rise to poles constrained to the real axis, thereby inducing an architectural bias in their recurrent dynamics. To unify these recurrent architectures and eliminate their pole geometry bias, inspired by the lattice--ladder realization of general IIR systems, we present \textsc{Mira} (\textbf{M}odal \textbf{I}IR \textbf{R}ecurrent \textbf{A}rchitecture), a unified family of linear-time recurrent architectures for building large language models (LLMs) with a highly parallel algebraic structure. Each \textsc{Mira} block realizes a higher-order IIR system that represents temporal modes from token streams through a lattice--ladder structure, followed by a feed-forward network (FFN). This realization enables the model to represent the temporal modes associated with both real and complex-conjugate poles. We pretrain \textsc{Mira-S}, a 12-layer small-scale LLM with 0.14B parameters composed entirely of \textsc{Mira} blocks. Throughout the pretraining, \textsc{Mira-S} achieves lower perplexity than strong recurrent baselines, exhibits stronger context-window generalization beyond the training length, and attains higher throughput with a dedicated CUDA kernel.

0.24

Generative Active Learning for Label-Efficient Pan-Organ Ultrasound Diagnosis

cs.AI cs.HC ·2026-11-06 ·Yaofei Duan, Yuhao Huang, Yuan Gao, Ruixiang Qi, Tianyu Zhang +7 more
Abstract

Medical ultrasound (US) enables real-time, non-invasive diagnosis, yet building US-based AI systems is limited by data scarcity, costly expert annotation, and limited cross-center generalizability. Privacy constraints and instance-level reporting hinder the collection of diverse abnormal cases and frame-level labels, while repeated retraining for new centers increases computational cost. We introduce EvoDiff, a generative active learning framework that integrates diffusion modelling with information-theoretic objectives, evolutionary search, and curriculum-based sampling to select informative samples across heterogeneous US datasets. EvoDiff formulates sample selection as a generate-to-retrieve process, in which diffusion-based optimization in feature space guided by information gain produces synthetic probes, and only the real unlabeled samples retrieved by these probes are annotated and used for downstream training. Across six benchmarks covering fifteen datasets and 16,696 images, we evaluate EvoDiff on convolutional network and transformer-based foundation model. EvoDiff approaches fully supervised performance with substantially fewer labels, outperforms active learning baselines, and improves foundation model generalization under cross-domain external validation on unseen cohorts. A reader study with six radiologists indicates the potential to support clinical annotation workflows, with a model-based estimated 24% to 84% reduction in annotation workload at matched downstream performance.

0.24

Against All Likelihood

cs.AI ·2026-11-06 ·David Lopez-Paz
position
Abstract

Maximum likelihood is the dominant training principle for large language models, yet likelihood is neither necessary nor sufficient for truth. Unlikely events happen every day, and frequently repeated statements can be false. Drawing on a philosophical tradition that ties objectivity to invariance rather than probability (from Nozick's account of objective facts and Wigner's principles of invariance to the structural realism of Poincaré and Cassirer), we argue that the truth of a statement should be judged not by its probability under a single distribution, but by its invariance across relevant transformations of the environment. We connect this view to machine learning literature on out-of-distribution generalization and invariant learning, sketch how invariance could be tested at inference time with language models, and discuss implications for continual learning.

0.24

Deep Learning Theory through the Lens of Statistical Physics: A Topical Review

cs.LG cond-mat.stat-mech ·2026-11-06 ·Wei Huang, Youjin Deng
publishedsurvey
Abstract · 22 pages, 5 figures, 1 table. Topical review published in Front. Phys. 22(1), 011302 (2027); this version contains minor corrections

The recent success of foundation models and deep learning has far outpaced our theoretical understanding of how large-scale neural networks learn and generalize. Statistical physics provides a natural language for describing such high-dimensional, nonlinear, and strongly coupled systems, offering concepts such as mean-field theory, phase transitions, energy landscapes, and stochastic dynamics. In this topical review, we adopt a statistical-physics way of thinking -- from concepts including phases and criticality, and free-energy viewpoints, to methods such as kernel and Gaussian-process limits, mean-field and dynamical mean-field theories, random-matrix tools, and Fokker-Planck formalisms, and to a mindset emphasizing universality, scaling, and tractable limits with explicit validity regimes. We survey recent progress on understanding modern deep learning through this lens, including mean-field analyses of wide networks, spectral and geometric perspectives on loss landscapes, stochastic-gradient-based training viewed as a dynamical process, and phase-transition analogies that organize phenomena such as overparameterization, generalization, and representation change. Our goal is to synthesize these threads into a coherent framework that clarifies what each approximation explains, when it is expected to hold, and how apparently competing narratives relate or diverge. Together, these perspectives form a unified physical framework that views representation learning as a collective phenomenon in high-dimensional systems, bridging deterministic and stochastic regimes of gradient-based optimization.

0.24

Model-based learning for DoA estimation and array calibration via continuous sparse recovery

cs.AI cs.LG ·2026-11-06 ·Baptiste Chatelier, José Miguel Mateos-Ramos, Vincent Corlay, Matthieu Crussière, Christian Häger +3 more
Abstract

Direction of arrival (DoA) estimation, a classical signal processing problem, has experienced renewed interest with the advent of integrated sensing and communication (ISAC). While decades of research have yielded highly accurate DoA estimators, hardware impairments have frequently been neglected, despite degrading both DoA estimation accuracy and downstream communication performance in ISAC systems. This work introduces a joint array calibration and DoA estimation method based on the simultaneous orthogonal matching pursuit (SOMP) algorithm, following the model-based machine learning paradigm. The proposed method enables the calibration of antenna arrays subject to severe gain-phase and antenna location impairments through projected gradient descent (PGD), with either supervised or unsupervised objectives. Additionally, DoA estimation is enhanced by extending SOMP to continuous dictionaries via PGD refinement steps. Simulation results demonstrate that the proposed method enables efficient hardware impairment compensation, even without labeled samples, while outperforming conventional sparse recovery and subspace-based DoA estimation baselines.

0.24

I Think, Therefore I Attack: Evaluating Thought Injection Attacks on Reasoning Agents

cs.AI cs.CR ·2026-11-06 ·Parth Illendula, Garik Livingston, Zain Shamsi, Ravi Srinivasan, Alex Liu
Abstract

Large Reasoning Models (LRMs) are one of the key components of current agentic systems. LRMs augment their output with an explicit chain-of-thought reasoning stage, introducing a new attack surface for thought injection, in which untrusted text resembling the model's own deliberation is fed back to the model. We present a two-part study of this attack, performed against the prevailing large, production class open-weight models (e.g., GLM-5.3, MiniMax-M2.5, Qwen3.8-Flash-Next). We first evaluate the effectiveness of static thought-injection prompts under different styles as well as tagged and untagged conditions and find that the effects of thinking tags and payload styles are all properties of the specific target model. In the second study, we exploit this model-specificity in a containerized agentic cyber range, using indirect injection prompts styled in the form of the target model's own reasoning traces to achieve attack success rates (ASR) up to 50% on operational objectives such as reverse-shell exfiltration and repository backdooring. We also find that the same malicious objective in standard prompt injection user voice never succeeds, and thinking tags suppress the agent's tendency to disclose the injection. Our results show that imitating a reasoning model's own thought style is effective for this style of attack, and that tag stripping is an unreliable defense.

0.23

TailDART: Compute-Efficient Foundation Models for Financial Time-Series Forecasting

cs.AI cs.CE ·2026-11-06 ·Chiang Qin Zhi, Ranjan Satapathy, Erik Cambria, Joyjit Chattoraj
Abstract

Although time-series foundation models (TSFMs) achieve strong zero-shot transferability across high-information-density physical systems, applying overparameterized TSFMs to quantitative finance is fundamentally constrained by low signal-to-noise ratios, non-stationarity, and prohibitive autoregressive decoding latency. To reconcile representation learning with institutional operational constraints, this study presents TailDART, a dense 2.68 million-parameter foundation model engineered for financial time-series forecasting. By integrating grouped-query attention into a lightweight backbone, the architecture drastically reduces Key-Value cache memory bandwidth overhead, achieving major inference acceleration (up to 48.1%) over prevailing state-of-the-art TSFMs to enable low-latency, high-throughput streaming execution. Furthermore, multi-asset benchmarks across varying temporal resolutions show that a compact model can remain competitive for market forecasting: TailDART matches the predictive fidelity of substantially larger baselines while demonstrating superior error resilience in high-volatility regimes.

0.23

Computer-Use Agents Are Embodied Generalists

cs.LG cs.RO ·2026-11-06 ·Pranjal Aggarwal, Sean Welleck
position
Abstract · 14 pages, 3 figures, 2 tables

In many areas of AI, general-purpose models have increasingly replaced handcrafted methods and specialized architectures. We argue that embodied AI is poised for a similar shift. Our key claim is that embodied tasks can be formulated as interaction with digital software. Specifically, we build software that renders an agent's view of a physical environment on a screen and provides controls for acting in it. A computer-use agent (CUA) can then perform embodied tasks through the same screenshots and keyboard and mouse actions that it uses to operate ordinary software. Using a single CUA design and changing only the software, our entries win first-, second-, third-, and sixth-place prizes across four ECCV workshop competitions, without any competition-specific agent tuning. These results suggest that the skills CUAs learn on software, such as visual grounding, precise clicking, and exploration, already carry over to embodied tasks, giving CUAs a far broader scope in embodied AI than previously realized. We therefore advocate that computer-use agents can be embodied generalists. Further, we advocate that, by following the training strategy that made computer-use agents general, embodied AI can converge on a single general model as well, ushering in a new era of general embodied AI. Website: https://pranjal2041.github.io/cua-embodied

0.23

Open-world Reconstruction-based Classifier with Concept Drift Adaptation

cs.LG ·2026-11-06 ·Joanna Komorniczak
Abstract

Data stream processing has become central to modern machine learning applications, with concept drifts and novel class appearances posing the primary challenges faced by recognition methods. Adapting to concept drifts while also being able to detect novel classes and distinguish groups of unknown observations can be seen as a step towards effective method employment in an open-world setting. This work proposes a classification method that can detect and learn the distributions of new, unknown classes, as well as adapt to concept drifts once the distribution of known classes shifts. The proposed method uses the autoencoders' reconstruction errors and the KDE that models their distribution. The autoencoders incrementally adapt to the changing distributions of known classes via weight fine-tuning, resulting in continuous adjustment to evolving concepts. Conducted experiments used a diverse set of synthetic data streams, embeddings of a real-world static CIFAR dataset arranged into a stream to simulate concept drifts and the emergence of unknowns, as well as real-world datasets where novelty is observed. The methodology was compared with state-of-the-art approaches with three-criterion recognition quality assessment and resource consumption evaluation. The results indicate that the proposed method is competitive with reference approaches in stationary environments and performs best when concept drifts occur.

0.23

A Stochastic Latent Variable Generalization of Joint-Embedding Predictive Architectures

cs.LG cs.AIcs.CV ·2026-11-06 ·Nikos Giakoumoglou, Paschalis Giakoumoglou, Tania Stathaki
published
Abstract · WACV 2027

Joint-Embedding Predictive Architectures (JEPAs) predict the latent embedding of a target signal from a context signal, with an auxiliary variable z conditioning the predictor. Despite its prominent role in the original JEPA blueprint, the content of z has been carried by a fixed deterministic token in every prior instantiation, collapsing the genuinely multi-modal distribution over plausible completions to a point estimate. We present S-JEPA, which realises z as a proper stochastic latent variable for masked image prediction: a learned posterior infers z from context and target during training, while a learned context-conditioned prior approximates the distribution of plausible completions at inference, with no decoders and no reconstruction. In its deterministic limit the objective reduces to the I-JEPA loss in functional form, generalising that objective rather than strictly containing the I-JEPA model. Our central finding is that stochasticity adds capabilities, error-correlated predictive uncertainty and diverse completions, at no cost to representation quality: under matched architectures and schedules S-JEPA improves over the deterministic I-JEPA baseline and exceeds LeJEPA at matched architecture, while additionally producing spatially structured uncertainty maps and semantically diverse completions, neither of which the deterministic predictor produces.

0.23

ScreenProof: Where Is This Screen Wrong? Finding UI Faults Without a Reference

cs.AI cs.SE ·2026-11-06 ·Huiming Sun, Yue Li, Kangning Yang, Ruineng Li, Jie Cai +4 more
Abstract

A phone screen can be wrong in ways that need no reference: a label that runs off the screen, letters that touch, a line that breaks one character before its end, one button with sharp corners among round ones. We ask whether a model can box every such fault on a screenshot, given no target design and no reference image. Such a critic could serve as a cheap first pass over generated pages before a designer reviews them, but only if it stays silent on clean pages. ScreenProof poses this target-free critique over three families of faults: general faults, local style faults (one element breaks the style its page sets) and text layout faults (letter spacing, line height, line breaks). A data engine makes each fault in page code and records the changed element's box, and hard negatives change every sibling alike. No frontier judge is reliable: frontier models flag 18-62% of clean real screenshots and 18-82% of clean generated pages, find 47-72% of text layout faults, and rarely find a foreign icon (5-19%). Our small critic ProofJudge (Qwen3.5-9B with LoRA), trained on code-made pages with one to four faults each and keeping a finding only when it is confident, reaches F1 0.48, the highest of all judges (Gemini 4 Argon 0.37; open models at most 0.15; 0.46 even without the confidence cut), and is the only fault-finding judge within 15% false alarms on clean pages (2% real, 9% generated), the property such a first pass needs, at about a hundredth of Argon's cost. It finds the most general faults (59%) but fewer local style faults than Argon (25% against 41%) and fewer text layout faults than the API judges (43%; Gemini 4 Argon 47%). Against a named brand the task is harder still: models know a design language about as well as its public documentation states it.

0.23

Artificial Shoppers and Attention Competition

cs.AI econ.THcs.GT ·2026-11-06 ·Xiaomeng Guo, Fasheng Xu, Xiaoli Zhang
Abstract

We study how AI shopping interfaces affect price competition, platform profit, and consumer welfare. A platform hosts two of three products of different quality. Without AI, consumers compare the hosted pair under salience distortion and overweight whichever attribute, price or quality, stands out. Under assistance, an AI assistant recommends the two products with the highest true utility, leaving consumers to make the same distorted comparison. Under delegation, an AI agent selects the best product, subject to purchase approval. Both modes access all products and know consumers' preferences; firms price strategically. First, assistance can soften competition despite expanded product access. The middle-quality firm may forgo low-valuation consumers because attracting them requires cutting the price paid by higher-valuation consumers as well. Relative to no AI with the same assortment, assistance lowers both hosted prices under the high- and low-quality pair but can raise both under the high- and middle-quality pair. Second, we identify an alignment paradox: some consumers can become worse off than without AI even though their best option remains recommended and becomes cheaper. Salience can induce a switch to a lower-quality product whose quality loss exceeds the saving relative to the original purchase. The assistant ranks products accurately but does not optimize the shortlist for consumers' realized welfare. Third, assistance can raise both platform profit and aggregate consumer surplus with the high- and middle-quality pair; delegation can do so with the high- and low-quality pair. Fourth, among AI designs, the platform weakly prefers delegation with the high- and middle-quality pair wherever all compared pricing equilibria exist. Yet this preferred design can yield lower platform profit and aggregate consumer surplus than the no-AI optimum.

0.23

Ontology-Driven Control of Conversational Agents with a Use-Case for Job Interview Practice

cs.AI ·2026-11-06 ·Barbara Gendron, Gaël Guibon, Mathieu D'aquin
Abstract

Large Language Models (LLMs) have been widely used as conversational agents in both domain-specific and open-domain situations. In domain-specific context, it is often necessary to adapt those LLMs to obtain conversations with desired characteristics, whether those apply at the level of each response (e.g. for tone or style) or more globally (i.e. at the level of a conversational strategy). Such adaptations are, however, generally realized in an ad hoc way, through specific, non-reusable, and non-transparent developments. In this paper, we propose an ontology-based approach to design conversational strategies that rely on the desired and measured characteristics of textual utterances in the conversation. Through a use case related to job interview practice, we show how this approach can be used to control LLMs to follow the strategy defined in the ontology. To this end, we automatically fine-tune the LLM to improve its compliance with ontologically-defined conversation characteristics. Our results show that training the LLM from formally defined concepts effectively enhances alignment with the requested characteristics. Hence, this work demonstrates how ontology-guided conversation control can be achieved while taking into account multiple, concurrent conversation aspects that are formally and transparently defined.

0.23

AutoEMatic: Runtime-Adaptive Mixed-Precision LLM Inference via Automatic Exponent-Mantissa Refinement

cs.AI ·2026-11-06 ·Akshat Ramachandran, Reena Elangovan, Rangharajan Venkatesan, Charbel Sakr, Tijmen Blankevoort +2 more
Abstract

Existing LLM post-training quantization techniques determine precision before inference begins, yet the precision a model requires varies dynamically across prompts, layers, and inference phases. We present AutoEMatic, a full-stack co-design that adapts the precision of weights, activations, and the KV cache at runtime. A lightweight policy selects each layer's precision at every prefill and decode step from two runtime signals: the entropy of the next-token distribution and the magnitude of each layer's update to its hidden states. This runtime adaptation is made practical by AutoFP, a multi-scale floating-point format that derives 2-, 4-, and 6-bit precisions from a single NVFP4 base format. We realize AutoEMatic on two targets: GPU kernels that execute AutoFP on NVFP4 Tensor Cores, and an accelerator whose vector MAC datapath executes all AutoFP formats natively. Across reasoning and agentic benchmarks, AutoEMatic improves accuracy by up to 21.7% over NVFP4 and matches group-scaled FP6 accuracy on most benchmarks, using 4.9 bits per weight versus 4.5 for NVFP4. On GB300 GPUs, AutoEMatic retains approximately 90-98% of NVFP4 decode throughput across the evaluated configurations. Furthermore, the AutoEMatic accelerator reduces latency by 1.6-2.3x and energy by 7-23% compared to an accelerator with NVFP4-only datapath at iso-accuracy.

0.23

TrialJev: Open-Weight Single Pass Eligibility Scoring for Patient Trial Screening

cs.LG ·2026-11-06 ·Omar Ibrahim, Grant Hamblin, Benjamin Haibe-Kains
code/data
Abstract · 38 pages. Code: https://github.com/oibrahi1/TrialJev. Models: https://huggingface.co/omareng/TrialJev

Clinical trial screening requires assessing large numbers of patient-trial pairs to prioritize candidates for detailed eligibility review, often under restrictions on sharing patient records. We introduce TrialJev, an open-weight eligibility scorer that uses LoRA-adapted Qwen models and a pointer head to score natural-language answer options in a single forward pass, without autoregressive decoding. Trained on TREC Clinical Trials 2021 and evaluated on 6,975 eligible-or-excluded pairs from 50 unseen TREC 2022 synthetic patients, TrialJev achieves 80.97% (Qwen3.5-27B) and 81.84% (Qwen3.8-27B) accuracy, compared with 79.87% and 79.93% for LLMs fine-tuned on the same data to generate verdicts. In matched batch-1 measurements, verdict generation takes 2.45-2.70 times as long as the one-pass inference of TrialJev at 9B-27B. Under a common development-fitted decision correction, the two 27B TrialJev models reach 80.99% and 81.88% accuracy, compared with 81.28% for the proprietary hosted Jev API; paired confidence intervals include zero. TrialJev runs locally on a single GPU, requiring approximately 178 ms per pair at 27B, without transmitting patient text to an external model API. An A/B-logit ablation shows that avoiding decoding, rather than the pointer head alone, accounts for the measured efficiency advantage. TrialJev is intended for first-pass prioritization of candidate trials, complementing rather than replacing explanatory models and clinical review. Evaluation is limited to synthetic patients and implementation-specific timing, and clinical use has not been validated. Adapters, pointer heads, and model cards are released on Hugging Face.

0.22

Challenges, Opportunities, and Misconceptions in Neurosymbolic AI

cs.AI ·2026-11-06 ·Giuseppe Marra, Pierre Genevès, Meghyn Bienvenu, Thomas Schiex, Sylvie Thiébaux +1 more
Abstract

We examine opportunities for neurosymbolic AI alongside increasingly powerful neural models, particularly with regards to efficiency, safety and interpretability. We identify two major challenges: ensuring semantic alignment between neural and symbolic components, and managing the computational cost of symbolic inference and learning. We discuss representative methodological approaches to these challenges and address several common misconceptions concerning differentiability, independence assumptions, and scalability. This paper synthesizes perspectives developed through discussions in the Neurosymbolic Programming and Reasoning (NPR) working group at the 2026 CNRS workshop Neurosymbolic Methods and Their Applications in Data Science, rather than providing an exhaustive survey.

0.22

RAD-TFM: Robust and Domain-Adapted Tabular Foundation Models

cs.LG cs.AI ·2026-11-06 ·Matthew Peroni, Franck Le, Vadim Sheinin
top venue
Abstract · Accepted at NeurIPS 2026 (main track)

The development of tabular foundation models (TFMs) has accelerated in recent years, outperforming traditional ML methods on many tabular benchmarks. A key finding is that TFMs can be pretrained entirely on synthetic datasets, opening opportunities to design data generators that encourage desirable model properties. Prior work has mainly focused on crafting high-quality priors over generators to improve overall pretraining performance. Our key insight is that parameterizing the generator distribution enables an adversarial, distributional robustness perspective: during training, we can adapt the generator to emphasize datasets that are particularly challenging for the model, which we formalize by introducing an optimality gap measure. Further, we develop a method to align the synthetic data generation with real-world datasets from a given domain, constraining the adversary to generate realistic data. Together, these algorithms comprise the Robust and Domain-Adapted Tabular Foundation Models (RAD-TFM) pipeline, a model-agnostic adversarial training framework. Applied to the TabPFN V2 classifier, RAD-TFM improves performance across 6 diverse tabular benchmarks, with up to a 10.8 percentage-point improvement in mean normalized AUC over the strongest baseline, using fewer than 100k additional synthetic training datasets, less than 0.1% of the original pretraining data. Applying the same pipeline to TabICL V2 also improves performance across three major benchmarks, demonstrating the generality of the framework. These results highlight a promising new direction for targeted adversarial continued training of TFMs using synthetic data alone.

0.22

CARGO: Causal Attribution and Responsibility-Guided Preventive Safe Multi-Agent Reinforcement Learning for Cooperative Driving

cs.AI ·2026-11-06 ·Jingjia Teng, Yang Li, Wenjie Huang
Abstract · Abridged version containing the main results; the full version is under peer review and will be updated here upon acceptance

This study presents Causal Attribution and Responsibility-Guided Optimization (CARGO), a multi-agent reinforcement learning framework for cooperative on-ramp merging in mixed traffic. A shared latent-context posterior and an action-conditioned response model estimate directed behavioral influence, with posterior uncertainty propagated into interaction-edge selection. Complementary attribution mechanisms couple signed influence with marginal team utility for cooperative reward shaping and assign responsibility from positive excess tail risk relative to a safe reference policy. Conditional diffusion generates future responses for action-dependent risk estimation. Responsibility-guided preventive optimization calibrates agent-specific risk-reduction requirements to local capacity within a policy trust region, using reserved safety margins, adaptive requirement scaling, and joint feasibility verification. Theoretical analysis establishes risk monotonicity and sufficient local conditions for non-decreasing environmental return under explicit assumptions. Experiments in highway-env across three traffic densities, together with controlled graph-recovery and attribution studies, support the effectiveness of the framework. Under high-density traffic, CARGO increases the success rate by 66.04\% and reduces the crash rate by 56.23\% relative to the best evaluated baseline for each metric.

0.22

Typed-Centrality Grounding for Agentic GraphRAG: A Symbolic Approach for Agentic Reasoning

cs.AI ·2026-11-06 ·Rosario Napoli, Antonio Celesti, Massimo Villari, Maria Fazio
published
Abstract · Preprint. 7 pages. Final version published in 2026 IEEE Symposium on Computers and Communications (ISCC), DOI: 10.1109/ISCC70364.2026.117399…

Knowledge Graphs (KGs) and Large Language Models (LLMs) represent two approaches for knowledge representation and reasoning over heterogeneous data. Recent research has explored their integration through Graph Retrieval-Augmented Generation (GraphRAG) systems, which use KGs to provide structured context to LLMs. However, existing GraphRAG approaches treat KGs as homogeneous graphs and rely on textual similarity between the input prompt and the context, rather than using it as a source of deterministic knowledge. This work proposes an agent-based GraphRAG system that introduces a symbolic layer to integrate deterministic reasoning alongside text similarities. This approach is based on the definition of typed centralities for KGs, where centrality measures are computed separately for each specific KG's class. This information is used to construct a symbolic representation that is provided to the LLM as a deterministic source. The proposed system was mathematically validated and evaluated on a real-world POLE (Person, Object, Location, Event) dataset, comparing reasoning and information retrieval with and without the symbolic layer. The results indicate that the our approach leads to responses that are more consistent with the context, improving the generated outputs.

0.22

Equidistribution of Solution Gradients in Physics-Informed Neural Networks via Coordinate Transformations

cs.AI math.NAcs.LG ·2026-11-06 ·Tianchi Yu, Ivan Oseledets
Abstract

Physics-Informed Neural Networks (PINNs) often struggle to approximate solutions with sharp gradients due to optimization difficulties and spectral bias. Motivated by the principle of gradient equidistribution from the stretching approach in grid generation, which redistributes solution gradients to reduce functional complexity, we propose Coordinate Transformation PINNs (CTPINNs), which incorporate learnable coordinate transformations to map the physical domain into a computational domain where the physical solutions become approximately affine, making them easier for neural networks to learn. We propose two constructions of the transformation: an explicit construction based on Majorants and an implicit construction based on invertable neural networks. Experimentally, the explicit construction can improve accuracy for PDEs with available explicit coordinate transformations, and the implicit construction can be combined with other advanced PINN models as an efficient plug-in module to improve performance across a range of challenging PDE benchmarks.

0.21

Learning When to Think: Dual-Reference Offline Optimization for Adaptive VLM Reasoning

cs.AI ·2026-11-06 ·Hongbin Lin, Sizhe Zou, Juangui Xu, Xinyue Xu, Zhuoli Ouyang +4 more
benchmarkcode/data
Abstract

Reasoning traces improve vision-language models on complex multimodal tasks, but forcing every input to follow a long reasoning process leads to overthinking and unnecessary inference cost. We study adaptive fast/slow thinking for VLMs, where one model answers directly for simple inputs and reasons step by step when needed. Fast answers and slow traces often come from distinct behavior policies, making a single-reference offline objective misweight one mode. We propose DUAL-REFERENCE BOLT (DR-BOLT), which distinguishes the desired fast/slow mixture reference from the sampling distribution and corrects their mismatch in reward-weighted training. Fast/slow partition competition explains routing in the optimal target; an efficient source-wise correction approximates the exact objective when response formats are well separated. Matched Qwen3-VL comparisons use identical candidate banks to isolate the objective from slow-teacher quality. DR-BOLT improves score-length trade-offs over the evaluated offline and online baselines on ScienceQA and three held-out transfer benchmarks, with a 24.1% cross-benchmark slow ratio and a 2.90-fold end-to-end training speedup over reproduced GRPO. These results support adaptive reasoning in the studied two-reference setting, while broader robustness remains open. Our code is available at https://github.com/HongbinLin3589/DR-BOLT.

0.21

Multi-GPU MoE Inference with Adaptive Expert Offloading

cs.AI ·2026-11-06 ·William X. Zheng, Zhongjie Chen, Shuo Yang, Ziming Mao, Xiaoze Fan +7 more
Abstract · 15 pages, including references and appendix

Modern Mixture-of-Experts (MoE) inference is constrained by GPU memory. This constraint is mitigated by expert offloading, which stores part of the model's weight in host memory and executes on demand. Though MoE architecture has computation sparsity, more GPUs are needed for serving increasingly larger frontier models. However, existing expert offloading systems overlook this need and fail to scale performance with additional GPUs. We present AdaOffload, a general and efficient multi-GPU hybrid expert execution framework that effectively handles heterogeneous hardware and workloads. We co-design GPU-CPU parallelism and GPU-CPU co-execution for decode and prefill with online workload observation that adapts to different hardware, workload, and batch size heterogeneity. We evaluated AdaOffload on five different platforms with varying GPU capability (L4, Pro6000, B200), number of GPUs, CPU capability, and NUMA topology on three different frontier open-weight models (284B to 753B). AdaOffload achieves $1.9\times$ to $15.5\times$ throughput speedups across different models over state-of-the-art systems.

0.21

When Can You Trust Neural Critics? Auditing Mutual Information Estimates of Higher-Order Epistasis in Protein Language Models

cs.LG ·2026-11-06 ·Maryam Rahimimovassagh, Waldemar Karwowski, Niloofar Yousefi
Abstract

Mutual information (MI) estimation is used to probe what neural representations encode, including in protein language models, but variational MI estimators have variance that scales at least exponentially in the true MI. We audit a prior neural-MI study of order-resolved epistasis in ESM-2 on the four-site GB1 landscape. Run-to-run differences show no systematic bias and a magnitude comparable to ordinary cross-seed retraining variability, and the evaluation-pairing scheme shifts scores substantially. At the original dimensionality, two of three alternative estimator families stabilize the estimate across seeds and one does not. In a synthetic calibration at the representation dimensionalities of two other protein language models (a Gaussian channel with known MI; no model representations), all three families recover materially less signal than required, despite correct signal ordering and stable seeds. A control critic given the true signal direction, which also reduces its input to one dimension, recovers 93.3% and 98.7% of the analytic MI at gain 1.5 (at 5,120 and 4,096 dimensions), so the signal is recoverable with privileged information, but the control does not identify why blind estimation fails. We package the checks into a reusable diagnostic protocol and conclude that even a stabilized estimator is not currently sufficient for order-resolved dependence claims made under blind search at this dimensionality and sample size.

0.21

Do large language models rank moral values? A transitivity test of moral competence

cs.AI cs.CLcs.CY ·2026-11-06 ·Katelyn A. Wasson, Kiwon Song, James M. Jennings, Brittney Currie, Konstantina Sokratous +1 more
Abstract · 30 pages, 2 figures, 4 tables

LLMs are increasingly used to support decisions that require moral consideration. While LLMs perform well on morality benchmarks from psychology and philosophy, producing appropriate outputs on these benchmarks (moral performance) does not establish that an LLM weighs competing moral values at stake in a decision (moral competence). We test a minimal, necessary condition for such weighing: that an ordinal ranking can represent LLM moral choices. Such a ranking requires transitivity. To assess transitivity, we presented LLMs with pairwise trade-offs among moral value violations in vignette choice sets and extracted resulting choice probabilities from first-token distributions. Tested against three stochastic transitive choice models, most choice sets admitted weak or unstable orderings. Aggregating an LLM's responses, a common practice in LLM moral evaluation, produced transitive structure that matched no individual choice set. Evaluations of AI morality should establish that choices are structured before asking whether they are aligned.

0.21

Towards an Autonomous AI Steward for Large-Scale Physics Experiments: Integrating Causal Diagnostics, RAG, and Conformal Prediction

cs.AI ·2026-11-06 ·Oliver Noll, Nicholas Tan Jerome, Alexander Lehnen, Frank Simon, Frank Maas
Abstract · 4 pages, 1 figure, 2nd International Conference on AI for Science (AI4Sci2026), 5-9 October 2026, Mainz, Germany (PoS)

Long-term particle physics experiments face critical operational challenges: monitoring non-stationary machine data streams and preserving tacit expert knowledge. We present the "Helmholtz Co-Pilot," a hybrid Multi-Agent System (MAS) acting as an autonomous experiment steward. The system can be triggered interactively by human operators or autonomously by an asynchronous causal diagnostics engine that runs independently of the MAS reasoning core. Continuously monitoring live hardware streams such as data acquisition and slow control, this engine performs dynamic causal discovery to trace cascading alarms back to their root causes and applies adaptive model selection to keep forecasts reliable across shifting machine states; upon detecting an anomaly, it hands a structured diagnostic report to the MAS. The reasoning layer itself relies on a high-throughput C++ core and a Python edge architecture orchestrating a directed acyclic graph (DAG) of specialized ReAct workers. It integrates a Retrieval-Augmented Generation (RAG) backbone for institutional memory and leverages conformal prediction to bound its evidence-based recommendations with mathematical coverage guarantees. This extended abstract outlines the architecture, the integration of software-enforced Human-in-the-Loop (HitL) guardrails, and ongoing shadow-testing within the KATRIN experiment.

0.21

Augmentron: Scalable Multi-View Visual Augmentation for Robot Learning

cs.LG cs.RO ·2026-11-06 ·Tanay Tandon, Aseem Doriwala, Jade Choghari, Catherine Weaver, Pragna Mannam
published
Abstract · Accepted to the 8th Robot Learning Workshop (WRL), NeurIPS 2026

Robot policies trained on visually narrow demonstrations can become sensitive to environmental features that are unrelated to the task. Existing generative augmentation methods can introduce visual diversity, but processing every frame and camera view independently is computationally expensive and may produce inconsistent observations across cameras and timestamps. We introduce Augmentron, a geometry-aware pipeline that generates reusable scene representations and renders them through recorded camera poses. This produces temporally and multi-view consistent environ- ment and workspace variations while preserving the robot, task objects, actions, and annotations. Across three physical manipulation tasks, replacing half of the real world training dataset with Augmentron data improves model success across three tasks for both in out-of-distribution distractor objects, and out-of-distribution workspace variation evaluations. In a separate comparison against alternate augmentation techniques, Augmentron achieves the highest pooled success in all three conditions. The results show that introducing augmented data to a data training recipe can help policies generalize to dynamic real-world settings.

0.21

A Clinical Evaluation of Conformal Prediction for Chest Radiograph Report Generation

cs.LG ·2026-11-06 ·Nikolaos Kontogeorgis, Pascal Krezer, Klaus-Rudolf Kladny, Arne Estler, Christian F. Baumgartner +3 more
Abstract

While multimodal large language models (MLLMs) have improved automated radiology report generation, they remain prone to hallucinations, necessitating clinically meaningful risk-control methods. Conformal prediction (CP) provides statistical guarantees, yet existing CP approaches rely on uncalibrated automated extractors and fail to account for the clinical relevance of errors. We introduce clinical admissibility criteria (CLARITY), a framework for evaluating radiology reports that explicitly distinguishes clinically consequential errors from minor deviations. To eliminate reliance on automatic judges, we collected radiologist annotations of MLLM-generated reports using CLARITY. By benchmarking CP strategies and automatic evaluation metrics against these expert annotations, we demonstrate that commonly used admissibility proxies systematically fail to detect clinically relevant errors. Our findings reveal that automated admissibility judgments are largely decoupled from true clinical admissibility.

0.21

Coordinated Mixed-Precision KV Cache Quantization for Efficient LLM-based Multi-agent Systems

cs.LG ·2026-11-06 ·Weizhong Huang, Jinchao Zhang, Xiawu Zheng
Abstract · 33 pages, 8 figures

LLM-based multi-agent systems solve complex tasks through collaboration, but their growing, simultaneously resident key-value (KV) caches create a memory bottleneck. Applying single-model quantization policies independently ignores differences in agents' precision needs, potentially degrading collective reasoning. We introduce Coordinated Mixed-Precision KV Cache Quantization (CoMixKV), which combines online sensitivity estimation with joint agent-layer bit-width allocation under a shared memory budget. To our knowledge, CoMixKV is the first mixed-precision KV-cache quantization framework to jointly allocate precision across simultaneously resident agent caches under such a budget. The estimator derives sensitivity from a local attention-error bound using recent queries, keys, and values, without calibration data, trial quantization, or additional model passes. A shared controller combines these scores with remaining quantizable capacity and byte-level storage costs to minimize total estimated error. Allocations are refreshed as agents start or finish and new observations become available. Selected widths apply only to newly quantized blocks; recent states retain native precision and previously quantized blocks remain unchanged. Across four benchmarks and three LLM backbones at average 2-bit, 3-bit, and 4-bit settings, CoMixKV achieves strong accuracy against the evaluated quantization baselines. At the average 3-bit setting on Gemma-4-31B-it, it achieves 68.35% average accuracy, exceeding the strongest quantized baseline by 0.59%. On Qwen3-32B, it reduces peak KV-cache memory by 32.08% and achieves a 1.38x inference speedup over FullKV.

0.21

An Empirical Study of Issues in GPU Kernel Libraries for LLM Systems

cs.AI ·2026-11-06 ·Yanjie Gao, Jiayi Bi, Haoxiang Lin, Ling Feng
Abstract

Behind the expanding use of large language models (LLMs) across application domains lies a growing dependence on specialized GPU kernel libraries. These libraries underpin the correctness and efficiency of training and inference across diverse GPU platforms, implementing attention, GEMM, quantization, and low-precision operators while managing tensor shapes, memory layouts, code generation, and hardware-specific dispatch. Problems at this layer can silently corrupt outputs, hang GPUs, prevent compilation, introduce deployment incompatibilities, or degrade performance, thereby reducing developer productivity and wasting expensive GPU resources. We present the first comprehensive empirical study of issues in GPU kernel libraries for LLM systems. We manually analyzed 300 high-quality GitHub issue reports, their discussions, and available fixes from seven representative libraries: FlashInfer, FlashAttention, Triton, CUTLASS, DeepGEMM, xFormers, and bitsandbytes. The analysis characterizes symptoms, root causes, fixes, and testing and debugging practices. Our main findings are: (1) 60.33% of the sampled issues originate in the Compiler and Build System or Runtime Environment dimensions, so validation must cover the deployment and execution stack, not just kernel code. (2) Kernel specialization imposes implicit correctness constraints on tensor shapes, metadata layouts, memory aliasing, and hardware resources. Violations can produce incorrect outputs or launch failures even for supported operations, highlighting the need for explicit contract validation and boundary-case testing. (3) Current testing and debugging practices are often insufficient for identifying low-precision numerical problems and hardware-specific execution failures. We outline research opportunities and tooling directions to support future LLM GPU kernel development.

0.20

Toward LLM Self-Play Under Data Scarcity: Effectiveness, Generalizability, and Trustworthiness

cs.AI ·2026-11-06 ·Zhiyin Yu, Yuchen Mou, Bingzhe Wu, Zhonghai Wu, Shiyang Feng +1 more
position
Abstract

The rapid development of large language models (LLMs) has been driven largely by scaling model size and human-curated training data, yet this paradigm is increasingly constrained by data scarcity and diminishing returns from human supervision. Self-play offers a promising pathway for autonomous data generation and improvement, but current frameworks remain unsatisfactory due to a lack of a data-centric perspective, limited cross-domain generalization, and growing trustworthiness and security concerns. In this paper, we argue that real-world LLM self-play systems should achieve effectiveness, generalizability, and trustworthiness under data scarcity. To address these challenges, we outline three blue-sky research directions for future self-play systems: (1) data-centric self-play for effective experience generation, (2) generalizable self-play for open-world and unverifiable tasks, and (3) trustworthy self-play for reliable self-evolution. We further discuss research opportunities and solutions within each direction. This paradigm shift will establish a sustainable and safe AI ecosystem by transforming finite human knowledge into inexhaustible high-quality experience through self-play architectures.

0.20

RAG-Enhanced LLMs for Interactive Explainability in Clinical Decision Support Systems

cs.AI cs.HCcs.LG ·2026-11-06 ·Lorenzo Mannocci, Francesca Naretto, Lucia Passaro, Anna Monreale
published
Abstract · Published in HC_AIxIA_HYDRA 2025, Communications in Computer and Information Science, vol. 2830, pp. 50-63, Springer, 2026. Please cite the …

Artificial Intelligence (AI) is increasingly embedded in high-risks decision-making systems across domains such as healthcare, finance, and the judicial system. As reliance on predictive models grows, so does the demand for Explainable AI (XAI) to ensure transparency, trust, and interpretability. However, existing XAI approaches often fail to present explanations in a format accessible to domain experts, limiting their practical utility. Large Language Models (LLMs) have recently emerged as a bridge between complex models and end-users, offering natural language explanations. Yet, their use introduces new risks, including hallucinated outputs and lack of source traceability. To address these challenges, we propose a Clinical Decision Support System (CDSS) that not only combines standard predictive models with XAI techniques but also incorporates a Retrieval-Augmented Generation module as an interactive enhancer. This module grounds explanations in verified medical knowledge, reformulates technical outputs into a human-like format, and enables users-such as clinicians-to actively query the system. By fostering an interactive, human-in-the-loop environment, our approach empowers domain experts to explore model decisions, contextualize explanations, and build trust in AI-assisted diagnostics. The study concludes with preliminary experiments validating the proposed methodology.

0.20

Agentic AI for Healthcare Claim Denial Analysis: A Policy-Grounded Framework for Privacy Preservation

cs.AI cs.LG ·2026-11-06 ·Sofia Campos, Carlos Martins, Pedro Pires, João Gama, Bruno Veloso +1 more
Abstract

In the healthcare sector, claim denials impose significant financial and administrative burdens, emphasizing the need for automated and trustworthy decision-support systems. This paper proposes a modular Agentic Artificial Intelligence (AI) framework for healthcare claim denial analysis, integrating specialized agents for denial risk prediction, risk factor identification, diagnosis-procedure-medication code relationship analysis, and counterfactual analysis. The framework integrates predictive analytics, graph-based learning, a Hybrid Contextual Retrieval-Augmented Generation (RAG) pipeline grounded in healthcare policy documents, and a dedicated HIPAA Compliance Agent to support privacy protection. Evaluated on a synthetic healthcare claims dataset, the framework achieved 98% traceability of reported numerical findings, validated 214 of 216 policy citations, and removed all 280 identified sensitive identifiers without modifying the underlying analysis. These findings pave the way for further research into Agentic AI as a framework for explainable, policy-aware, and privacy-preserving healthcare claim denial management.

0.20

From LoRA to Language: Recovering Skill Documents from Adapters

cs.LG cs.AIcs.CL ·2026-11-06 ·Saranya Dintakurthy Swadha, Pavan C Shekar, Arpit Jain, Aswanth Krishnan
Abstract · 29 pages, 5 figures, 18 tables

Agent skills are short documents that teach a language model how to handle a class of tasks. To save context, these documents are increasingly compiled into low-rank adapters (LoRA), which keep the behaviour but lose the text: an adapter is hard to read, edit, audit or move to another model. We ask whether the skill document can be written back from the adapter and its frozen base model, and we score each recovered document by the share of the original's key instructions it states and by the tasks it solves. Reading the weights alone recovered little with the methods we tried; a learned map from weights to text passed on about one bit of the document. Running the adapter worked much better. The Executing Writer, a small writer trained on top of the adapted model, recovers 39-44% of the key instructions of held-out skills, against 1.3-1.5% with another skill's adapter, and on adapters fitted to 41 external skills its documents restore 85-91% of the original's task gain. These strongest results are on adapters whose training stored the document; adapters trained on behaviour alone are much harder for trained writers (1-2% of instructions). For these adapters we test an agent that runs the base and adapted models and writes a skill from where they differ. On ALFWorld its skills recover 58-71% of a LoRA's gain and carry over to models that cannot load the LoRA, lifting Llama-3.1-8B from 5% to 55%. Whether such a skill comes from the adapter depends on what the adapter learned: when its behaviour cannot be guessed, the adapter is the source; when it can, a reward loop without the adapter does as well.

0.20

Statistical Geometry of Group Actions: Symmetry, Transport, Estimation

cs.LG math.ST ·2026-11-06 ·Quentin Berthet, Peter Bartlett, Romuald Elie, Yu-Han Wu
Abstract

We study variational inference in the presence of distributional symmetry, focusing on the interaction between two distinct group actions on a sample space: a symmetry action that leaves a reference distribution invariant, and a transport action that generates a parametric variational family by pushforward. By analyzing the algebraic structure of the total transformation group generated by both actions and the induced adjoint representation of the symmetry group on the transport Lie algebra, we establish that the Riemannian gradient of the Kullback--Leibler divergence at the true transport parameter is constrained to lie in the trivial representation of the symmetry group. Consequently, symmetry forces the gradient to vanish along all non-trivial representations. Combining this representation-theoretic constraint with parameter identifiability yields an orthogonal decomposition of the transport Lie algebra into three interpretable subspaces: an identifiability kernel (along which the objective is locally constant), a bias subspace (where the gradient may be nonzero due to model misspecification), and a symmetry-breaking subspace (where the gradient is forced to vanish by symmetry alone). Finally, we characterize the residual bias directions in terms of distributional mismatch between the base model and the standardized target, and derive an explicit Stein-type moment-matching condition that eliminates this bias, extending the location-scale results of (Margossian and Saul, 2025) to general Lie group transformation families.

0.20

Kernel-Based Learning of Analytic Operators and Parametric PDEs

cs.LG stat.MLmath.ST ·2026-11-06 ·Giovanni S. Alberti, Markus Holzleitner, Lorenzo Rosasco, Matteo Santacesaria
Abstract · 43 pages, 4 figures

We study kernel-based learning of analytic operators from finite samples, with a focus on how the regularity of the target should determine the choice of the kernel. We establish embeddings of analytic operators into vector-valued reproducing kernel Hilbert spaces and derive finite-sample, high-probability guarantees that quantify the uncertainty due to random sampling: in the noiseless setting, suitable analytic kernels yield rates exponential in a fractional power of the sample size, while Sobolev-type kernels yield polynomial rates. We apply the theory to the solution operator of a parametric elliptic PDE, where the training data are themselves finite-element approximations, and obtain bounds that separate the sampling and discretization contributions to the error and prescribe how the mesh size should be coupled to the sample size. The results are illustrated numerically.

0.20

Rethinking Agent Trajectory Compression: A Workflow-Aware Perspective

cs.AI ·2026-11-06 ·Qijia Zhuang, Rui Liu, Wenqi Hu, Yuxiang Ren
code/data
Abstract · 12 pages, 4 figures

As long-horizon agents increasingly rely on multi-step tool interactions to accomplish complex tasks, their execution trajectories continuously accumulate structured outputs and intermediate artifacts, leading to growing context overhead. Existing compression methods primarily rely on text importance or fixed truncation rules, often overlooking cross-step dependencies within workflows and potentially discarding critical information required for subsequent task execution. To address this challenge, we propose WAAC, a workflow-aware framework for agent trajectory compression. WAAC structurally models successful historical trajectories to extract reusable workflow compression experience and incorporates online backward dependency analysis to dynamically identify information retention requirements. By combining current dependency constraints with historical experience, WAAC selects appropriate compression strategies for different types of historical information. Experiments on PinchBench and Odysseys demonstrate that WAAC achieves an overall score of 87.6 and a perfect completion rate of 19.0%, respectively, while reducing execution costs by 19.7% and 29.1% compared with the strongest baseline methods. These results demonstrate the effectiveness of workflow-aware trajectory compression in improving task performance and cost efficiency for long-horizon agents. Code and resources are available at https://anonymous.4open.science/r/waac-6527/.

0.20

Towards Robust Reinforcement Learning Market Making: Lessons from Loss Function Experimentation

cs.LG q-fin.CPq-fin.TR ·2026-11-06 ·Siddharth Nand, Bora Guney
Abstract · 6 pages, 3 figures

Market makers play a vital role in ensuring efficient trade within financial markets. Recent literature has delved into training market-making systems using reinforcement learning. This paper delves into the impact of various loss function formulations on risk management behavior within reinforcement-based market making. Through empirical evaluation, we contrast profit-maximizing strategies with risk-averse approaches, finding that penalties for inventory and cash management foster more resilient and adaptable market-making strategies. Our results underscore the significance of striking a balance between profit maximization and risk management in algorithmic trading, providing actionable insights for refining strategies in practical settings.

0.20

Learning to Restart: Differentiable Restart Policies for Optimization Dynamics

cs.LG ·2026-11-06 ·Yiman Fong, Zhonglin Xie, Zaiwen Wen
Abstract

Adaptive restart mechanisms, which reset the momentum or other auxiliary state of an iterative optimizer, play a central role in the empirical performance of accelerated first-order methods, yet existing restart rules remain almost entirely hand-designed. The obstruction to learning restart policies is that, unlike the typical Learning to Optimize (L2O) setting, the discrete restart events make the resulting trajectory non-differentiable in the algorithm parameters and thus block gradient-based optimization of the restart policy. We resolve this obstruction by developing a differentiable framework for restart policies through their continuous-time limit. We introduce surrogate gradients for the restart events and optimization trajectories that are efficient to compute and provably converge to their continuous-time counterparts. We validate the effectiveness of our surrogate gradients in L2O settings, training restart policies that improve upon standard handcrafted adaptive restart criteria on quadratic and linear programs, and that transfer zero-shot from synthetic quadratics to logistic regression on real datasets.

0.19

Large Language Model Consensus and Corpus-Adaptive Topic Modeling for Classifying and Mapping Patent Innovation

cs.AI cs.IR ·2026-11-06 ·Sercan Ozcan, Enes Ozgozler, C. Okan Sakar
Abstract

Patent analytics increasingly relies on transformer representations and large language models (LLMs), yet linking conceptual classification, topic formation and interpretation requires explicit decisions about model roles and the evidence retained at each stage. In this design-oriented study, we propose and demonstrate a patent analytics pipeline that links these tasks and keeps each decision open to audit. Three LLMs separately classify each patent by novelty (radical or incremental) and innovation type (product, process, service). Radical and innovation-type labels each require two supporting votes, with Incremental used as the fallback novelty label. Topics are discovered by clustering PatentSBERTa embeddings, with parameter ranges scaled to corpus size. A patent-specific stopword list is applied only when topics are represented with class-based term frequency--inverse document frequency (c-TF-IDF), and a second embedding model supports inter-topic similarity, hierarchical refinement and visualization. An LLM then labels each topic from its ranked keywords and representative patents. We demonstrate the pipeline on 30,964 public-sector-related patents retrieved from Lens.org. Across five category corpora ranging from 591 to 29,215 patents, the pipeline produced 74 category-specific topic instances with a macro-average document coverage of 98.1% and topic diversity of 0.900--1.000, and fewer than 1% of classification outputs returned UNKNOWN. Because every model vote and the evidence behind each topic label are retained, analysts can audit and revise the outputs. The demonstration establishes the feasibility and traceability of the pipeline; testing its labels against expert coding and benchmarking it against alternative pipelines are identified as the next steps. The pipeline offers a reusable basis for LLM-assisted patent analysis in technology intelligence.

0.19

Coordinatewise Epsilon for Adam under Heavy-Tailed Noise

cs.LG ·2026-11-06 ·Jeonseong Kim, Chenglin Fan
benchmarktheory
Abstract · 32 pages

Adam estimates second moments coordinatewise, but typically adds the same scalar $\varepsilon$ to every denominator. We study how this shared $\varepsilon$ affects convergence when noise and curvature vary across coordinates. Under coordinatewise finite-$p$ noise moments, $p\in(1,2]$, choosing $\varepsilon_i$ from the noise and curvature bounds for each coordinate attains the heavy-tailed $\ell_1$-stationarity benchmark $O(T^{-\frac{p-1}{3p-2}})$, matching the known optimal time exponent for smooth nonconvex stochastic optimization. For $1<p<2$, Adam with a shared $\varepsilon_t$ admits a lower bound of order $Ω(T^{-\frac{p-1}{2p}})$. This lower bound holds for every predictable scalar choice of $η_t$, $β_{1,t}$, $β_{2,t}$, and $\varepsilon_t$, precluding a dimension-uniform benchmark rate for this class.

0.19

ResQAdam: Residual-Preserving Adaptive Optimization for Memory-Efficient Quantization Aware Training

cs.LG cs.CL ·2026-11-06 ·Liqun Ma, Shengkun Tang, Liangyu Wang, Lin Zhao, Wei Luo +5 more
Abstract

Quantization-aware training (QAT) is the most reliable way to obtain accurate low-bit large language models (LLMs), but training with AdamW keeps two dense FP32 states for every weight matrix. We find that low-rank memory-efficient optimizers such as GaLore and LDAdam degrade substantially in QAT. A considerable part of the QAT momentum lies outside its top-r singular subspace, and projection-based methods discard this part at every step. We further observe that this residual remains useful when only its sign is kept. Based on these observations, we propose ResQAdam. It keeps a single BF16 momentum, applies Adam only to the r x r coordinates of the principal subspace, and adds a normalized sign update computed from the residual. We evaluate ResQAdam with two from-scratch QAT methods and three QAT fine-tuning pipelines, with weights quantized down to 2 bits. It reduces optimizer-state memory by 71% to 74% relative to AdamW, stays close to AdamW in most settings, and lowers training throughput by less than 2%.

0.19

PhyCAD: Physics-Aware Text-to-CAD Generation via Hierarchical Reinforcement Learning

cs.AI ·2026-11-06 ·Yuxin Cui, Weijie Zheng, Zhengyang Li, Han Wang, Wenjie Feng +1 more
dataset
Abstract

Language-driven computer-aided design (CAD) generation enables the creation of editable parametric models from text, yet geometric correctness alone does not ensure physical feasibility. We present PhyCAD, a framework for physics-aware Text-to-CAD generation that trains language models to produce executable CadQuery programs under geometric and physical constraints. To support this setting, we construct a Text-to-CAD dataset that links parametric CAD models with geometric specifications, loading conditions, and physical constraints. We further develop an automated CAD--CAE verifier that combines program execution, geometric checks, and finite element analysis to provide feedback for model training. A key learning challenge is to prioritize joint satisfaction of geometric and physical constraints while retaining informative feedback on partial progress. A flat aggregation of heterogeneous rewards can obscure this priority, allowing improvements in individual criteria to compensate for unmet design requirements. We therefore introduce a Hierarchical Balanced Advantage for Group Relative Policy Optimization, grouping candidates into levels according to their verification status and combining between-level progress with objective-wise peer ranking within each level. This formulation preserves cross-level priority in the training signal while providing fine-grained feedback for satisfying constraints and improving feasible designs. Experiments across multiple language models show that PhyCAD consistently improves joint geometric and physical constraint satisfaction. These results demonstrate the value of verifiable engineering feedback for advancing language-driven CAD generation beyond geometric reconstruction toward physics-aware design.

0.19

k-Volume: a Unified Active Learning Strategy

cs.LG stat.ML ·2026-11-06 ·Eliot Tron, Alain Celisse, Christophe Denis, Alice Le Brigant
Abstract

In the classification context where revealing the label of a sample is costly, we propose a new Active Learning (AL) strategy named k-Volume (kVol). It randomly selects a sample to be labeled according to the volume of the smallest ball containing its k nearest neighbors. kVol is highly versatile: it can be applied either in the input space or in the output space, and it enjoys the desirable properties of the three main families of AL strategies: density, diversity, and (when performed in the output space) uncertainty. From a theoretical perspective, we establish that kVol is asymptotically equivalent to uniform sampling over the support of the (input or output) data distribution. Based on experiments on various synthetic and real-life tabular datasets and images, we observe that kVol is surprisingly stable: whatever the classifier or the experimental setup, it still lies among the best AL strategies unlike other competitors that fail dramatically. In particular, kVol outperforms state-of-the-art competitors such as CoreSet, TypiClust or Uncertainty on several real-life datasets, at a low computational cost.

0.19

KineCouple: Substrate-conditioned residual readout for enzyme kinetic prediction

cs.AI q-bio.QM ·2026-11-06 ·Chun Kang, Yunbo Zhang, Fanhu Zeng, Jian Liu, Siwei Deng +1 more
Abstract

Accurate prediction of enzyme kinetic parameters can support enzyme engineering and metabolic modeling by complementing experimental characterization. Existing prediction methods often pool residue-level protein features before combining them with substrate features. Such substrate-independent pooling cannot adapt residue weighting to the substrate during aggregation, potentially diluting information relevant to a specific enzyme-substrate pair. Since the contributions of individual residues to catalysis can depend on the substrate, substrate-conditioned aggregation offers a way to emphasize residue features relevant to the prediction. We introduce KineCouple, which uses substrate queries and residue-level keys and values in cross-attention, allowing the substrate to guide residue selection before aggregation. The model preserves the global protein representation and adds a gated correction defined as the difference between the conditioned readout and a uniform readout. ProtT5 and MolT5 remain frozen, and separate models predict kcat and Km on a logarithmic scale. In pooled out-of-fold predictions on the CataPro benchmarks, KineCouple reduces kcat RMSE from 1.333 to 1.330 and increases PCC from 0.492 to 0.496 relative to reproduced CataPro. For Km, RMSE decreases from 0.998 to 0.997 and PCC increases from 0.633 to 0.634. Spearman correlation also improves for both tasks.

0.19

Future-Informed Actor-Critic

cs.LG ·2026-11-06 ·Samy Mokeddem, Gaspard Lambrechts, Adrien Bolland, Damien Ernst
Abstract · 4 pages workshop-paper

Actor-critic methods optimize a parametrized policy using a critic to estimate their policy gradient. Recent work has shown that conditioning the critic on features of future states, actions, or rewards can improve sample efficiency and even final return. However, existing methods are restricted to exogenous features (independent of actions), overlooking endogenous features (dependent on actions) that may be crucial for credit assignment. We introduce the future-informed value function, which can be conditioned on both exogenous and endogenous features of future states, actions, and rewards. Then, we express the policy gradient using this value function, yielding the future-informed policy gradient. Building on that, we propose the future-informed actor-critic algorithm. Empirical results show that this algorithm leveraging endogenous features of the future outperforms standard actor-critic algorithms in environments with noisy and sparse rewards.

0.18

LandslideFM: A Foundation Model Towards Unified Landslide Perception and Understanding

cs.AI ·2026-11-06 ·Shanwen Wang, Sirui Wang, Xin Sun, Jungong Han
benchmarkcode/data
Abstract

Intelligent landslide detection is essential for disaster prevention. However, existing models primarily focus on isolated visual tasks, require extensive labeled data, and lack deep analytical and scene-understanding capabilities. To address these limitations, we propose LandslideFM, the first unified foundation model that connects visual perception tasks with contextual interpretation of landslides. Specifically, LandslideFM adapts to multiple visual tasks via a Source-Agreement Graduated Trust semi-supervised learning framework that leverages limited labeled data alongside abundant unlabeled data. Furthermore, predicted landslide regions provide explicit spatial guidance to a Vision-Language Model adapted through parameter-efficient fine-tuning, enabling comprehensive landslide reasoning. To support training and evaluation in this domain, we also present LandslideVL, the first benchmark for intelligent landslide analysis, combining visual imagery and newly introduced textual prompts. Comprehensive experiments across public datasets and LandslideVL demonstrate that LandslideFM achieves superior quantitative and qualitative performance. Most importantly, LandslideFM surpasses general-purpose VLMs in specialized landslide reasoning and scene analysis. Our code is released at https://github.com/wangshanwen001/LandslideFM.

0.18

StreamTab: Training-Free Temporal Link Prediction with Tabular Foundation Models

cs.LG cs.DBcs.SI ·2026-11-06 ·Mingqi Yang, Renqi Wu
Abstract

Temporal link prediction uses past interactions to predict whether two nodes will interact in the future. Temporal graph neural networks achieve strong results on this task, but they are trained separately for each graph. How a model should use interaction history also depends on the candidates it is compared with. A past interaction can distinguish a recurring link from a random pair that has never interacted, but when the negative candidates have also interacted before, other signals are needed. We propose StreamTab, a training-free framework built on a frozen tabular foundation model pretrained only on synthetic tables. StreamTab describes each candidate pair by statistics of its past interactions and shows the model recent observed interactions and sampled negative pairs as labeled examples, from which the model scores new candidates. Its central design, regime conditioning, samples the negative examples according to how the candidates are sampled, so the examples show the distinction the model must make. As new interactions arrive, StreamTab updates the statistics and refreshes the examples while the model parameters stay fixed. On thirteen datasets against 21 temporal graph baselines, StreamTab achieves the highest average precision on twelve datasets under historical sampling and eleven under inductive sampling, and ranks first or stays within one point of the best on seven datasets under random sampling. With the same features and model, regime conditioning raises average precision over examples with random negatives by a median of 12.6 points under historical sampling.

0.18

Rethinking How to Leverage Cheap Physics for Data-Efficient Neural Operators

cs.LG ·2026-11-06 ·Shuo Zhang, Senwei Liang
Abstract

Neural operators learn mappings from PDE inputs, such as initial conditions and coefficient fields, to the corresponding solutions. Once trained, they can rapidly predict solutions for new inputs. However, obtaining input-solution pairs for supervised training often requires expensive numerical simulations, whereas PDE inputs and governing equations provide inexpensive physical information. This raises a question: how can cheap physical information reduce the need for labeled solutions? A natural approach is to use physics to select inputs for labeling, but our acquisition experiments show no consistent advantage over random sampling. This observation motivates exploring an alternative: using cheap physics to improve neural operator initialization before supervised training. We propose Physics-Informed Initialization (PI-Init), a simple yet effective method that pretrains neural operators on unlabeled inputs using physics-informed losses and retains the learned parameters for supervised fine-tuning on the labeled dataset. A linearized analysis shows that pretraining can improve downstream prediction accuracy by reducing initialization errors that limited labeled data cannot effectively correct. Experiments across diverse PDE benchmarks and neural operator architectures demonstrate improvements in predictive accuracy, out-of-distribution generalization, and label efficiency. Representative results show error reductions of up to 62.7% at the same label budget and comparable or better accuracy with 90% fewer labels, relative to supervised-only training.

0.18

ROAM: Retrieval-Augmented Optimization for Feasible Multi-Day Itinerary Planning

cs.AI ·2026-11-06 ·Ye Zihao, Lin Xiaoheng
Abstract

Large language models (LLMs) can draft plausible travel itineraries, but the plans they produce are frequently infeasible: attractions are grouped into days without regard for geography, opening hours are ignored, and a single generation pass cannot reliably organize a multi-day trip. We present ROAM, a retrieval-augmented agentic framework for spontaneous multi-day itinerary planning that couples semantic preference retrieval with explicit spatiotemporal optimization. ROAM embeds the natural-language request, retrieves a compact candidate set from a curated point-of-interest (POI) database, partitions candidates into day groups through maximal-clique analysis and K-means, orders each day with a quality-biased simulated-annealing TSP solver, and repairs hard opening-hour violations with greedy exchange. A spatiotemporal retrieval layer injects weather forecasts, real road distances and POI operating hours into the prompt, after which the LLM renders a structured, map-ready itinerary; an interactive ReAct-style replanning agent then supports natural-language edits while re-checking feasibility. Across four Chinese cities and 510 POIs, ROAM produces 27.5 POIs per three-day plan versus 19.3 for a bare LLM and 17.0 for a retrieval-only baseline (+42% and +62%), while reducing route dispersion from 2,101 to 677 (3.1x). Ablations show that spatial optimization is the dominant driver of route efficiency and that hard time-window repair prevents roughly 30% of attractions from violating operating hours.

0.18

Adversarial Evaluation of Large Language Models for Patient-Friendly Radiology Report Summarization

cs.AI ·2026-11-06 ·Magdalini Paschali, Asad Aali, Arogya Koirala, Andrew Johnston, Katherine McNamara +4 more
Abstract · This work has been submitted to Radiology: Artificial Intelligence for possible publication. Copyright may be transferred without notice, af…

Purpose: To develop and validate a scalable adversarial evaluation framework for assessing the robustness of large language models (LLMs) that generate patient-friendly summaries of radiology reports. Methods: We introduced a clinically grounded adversarial framework that uses controlled term substitutions and impression removals across intensity levels. Using a radiologist-validated adversarial dictionary, we generated adversarial variants of CT and MRI reports. Six LLMs summarized original and adversarial impressions using 2 prompt styles, yielding 18,000 evaluation pairs. Robustness was evaluated with an automated LLM-as-a-judge framework scoring factual faithfulness, sensitivity to adversarial edits, certainty fidelity, and hallucination rate. Failure mode analysis quantified vulnerabilities in incorporating clinical modifiers. Agreement with two radiologists was assessed on 105 cases. Results: GPT-5-mini achieved the highest factual faithfulness and certainty fidelity under both adversarial edits. Under term substitution, hallucinations occurred in 28.1-30.1% of GPT-5-mini summaries, compared with 35.3-36.4% for Claude Sonnet 4.5 and 41.4-41.9% for MedGemma. Impression-removal edits were handled more robustly than term substitutions. Even for the highest-scoring summarizer, substitutions of anatomical terms were not reflected in 23.7% of summaries, diagnostic certainty in 22.3%, and descriptor terms (e.g., "substantial") in 21.2%. Judge-radiologist agreement (macro-F1: 0.55-0.56) was comparable to inter-radiologist agreement (0.53). Conclusions: Adversarial evaluation exposes clinically meaningful failure modes, including missed anatomical modifiers, altered diagnostic certainty, and omitted severity descriptors.

0.18

PACE: Keeping the Right Tokens in the KV Cache Is Not Enough for Long-Horizon LLM Agents

cs.LG cs.CLcs.DC ·2026-11-06 ·Suman Chatterjee
Abstract · 21 pages, 3 figures, 8 tables

Long-horizon LLM agents accumulate sessions of up to millions of tokens, far more key-value (KV) cache than one GPU holds, so serving systems keep a small budget of tokens and evict the rest. Eviction methods differ in which tokens they keep but share one goal: that the tokens a later step needs are still cached. We show this is not enough. In 100K-token coding sessions on Qwen3-8B with a 2,048-token budget, a planner-guided policy that keeps over 99% of the needed tokens resident still fails late steps increasingly often: its accuracy falls from 93% to 64% as the gap between a definition and its use grows from 2K to 90K tokens, at residency 0.996. The needed tokens are resident but unretrievable: what grows is their rotary-position distance to the query. We present PACE (Planner-Aware Context Engine), a KV-cache engine that couples plan-aware retention, in which the agent's planner tells the cache which context future steps will read, with bounded positions, which renumber kept tokens so their rotary distance to the query stays below 8K at any session length. In a pre-registered evaluation on fresh tasks, PACE passes 97% of hidden unit tests at 100K tokens, against 42% for H2O and 3% for SnapKV under the same budget, and needs both components: either alone reaches 63% (the combination beats each, p < 10^-4). The result holds with length: in 1M- and 2M-token sessions on one 24 GB GPU, with definitions written 0.9M and 1.8M tokens before use, PACE passes 99% and 100% of checks and recalls every session fact, while bounded positions without foresight falls from 33% to 13%; PACE holds under 0.7 GB of KV state where the full cache would need 147-295 GB. Renumbering is exact, costs no measurable perplexity, and adds 9% wall time. On Mistral-7B, whose coding ability caps it at 40% even with nothing evicted, PACE stays at that ceiling while retention alone falls below it.

0.18

Where Should Learning Live? Regret Analysis for Text, Parameter, and Joint Adaptation

cs.LG stat.ML ·2026-11-06 ·Jungtaek Kim, Jaden Park, Jongwon Jeong, Emmanouil-Vasileios Vlatakis-Gkaragkounis, Kangwook Lee
Abstract

Language-model systems can select prompts and parameter adapters repeatedly during deployment, using feedback to guide later choices. Earlier choices may change the context or environment inherited by later configurations. We study selection among preconstructed candidates and compare cumulative loss with the best fixed prompt--parameter pair used from the beginning. This excess loss separates into three terms: selection regret within the available family, the gap from restricting that family, and the benefit or harm of the actual deployment history. Selection is evaluated using the history each candidate would have produced if used throughout, which separates candidate quality from the effects of switching. When loss functions are fixed independently of selector randomness, past choices affect loss for only a bounded number of rounds, and every candidate's own-history loss is revealed, limiting switches yields sublinear expected regret within the selected family. We also quantify the costs and additional assumptions of independent probing and deployment-only feedback. Controlled experiments show that fewer switches can either lower or raise deployed loss, depending on whether inherited histories are harmful or helpful. With Qwen3.5-9B on Spider, text-only selection has lower within-family regret than joint selection, yet produces more failures because its candidate family excludes stronger configurations.

0.18

Multi-objective drift-aware MLOps agent for model retraining

cs.LG ·2026-11-06 ·Kumuda Valli, Anli Ji
published
Abstract

Deployed ML models face non-stationary data and delayed labels, yet most retraining strategies rely on fixed schedules or brittle drift thresholds that ignore operational cost. In this setting, retraining is a sequential decision problem rather than a one-time threshold choice. We present DARE (Drift-Aware Retraining Engine), an offline, event-driven retraining framework that separates ingestion, monitoring, policy decisions, training, and logging through explicit schemas and deterministic replay. We compare DARE against practical heuristic retraining baselines, including fixed-interval, threshold-persistence, and action-rate-matched fixed scheduling policies. On Airline, the locked LinUCB policy achieves a higher mean F1 while substantially reducing intervention frequency relative to heuristic and action-rate-matched baselines. Additional runtime-calibrated evaluation shows that these gains persist when measured using the cumulative update wall-clock time. These results suggest that conservative sequential retraining policies can improve performance while reducing unnecessary operational overhead under delayed feedback.

0.17

Attend at the Right Resolution: Head-Adaptive Sparse Attention via Query-Key Interaction Structure

cs.AI ·2026-11-06 ·Wajih Hassan Raza, Ahsan Bilal, Yidan Shen, Renjie Hu, Xin Fu +1 more
Abstract

Attention is a major scalability bottleneck in large language models (LLMs), with its computational cost and key-value (KV) memory traffic becoming increasingly prohibitive as context length grows. Sparse attention addresses this bottleneck by restricting each query to a subset of the context, but its effectiveness hinges on identifying relevant regions both cheaply and reliably. Existing training-free methods primarily rely on runtime relevance estimates, with selection dimensionality and granularity typically governed by fixed heuristics rather than the intrinsic interaction structure of pretrained attention heads. We take a different perspective: the query-key operator of each pretrained head defines a characteristic response structure that determines the resolution needed for sparse selection, both across feature directions and across context tokens. Building on this insight, we propose Head-Adaptive Sparse Attention (HASA), a training-free sparse attention framework that jointly adapts selection resolution along these two dimensions. Along the feature dimension, the singular spectrum of each head's query-key operator determines an effective rank, yielding a compact, head-specific response space. Along the sequence dimension, neighboring regions are progressively merged according to the directional coherence of their token responses in this same space, while heterogeneous regions remain separate. Exact attention is then computed using the original full-dimensional KV states of the selected regions. HASA captures 88.1% of dense-attention mass and, compared with UNIQUE, a strong recent sparse-attention baseline, improves accuracy by up to 3.47 and 2.57 points on RULER and LongBench-Pro, respectively, and accelerates self-attention by up to 1.32x, while achieving up to 14.07x speedup over dense attention.

0.17

Fixed-point acceleration on Neural Wasserstein manifolds

cs.LG math.OC ·2026-11-06 ·Vitalii Aksenov, Sebastian Gutierrez Hernandez
Abstract

We propose a fixed-point acceleration method for optimization over probability measures, applicable to a variety of machine learning problems. We represent probability measures as pushforwards of a reference distribution through a neural network map and equip the parameter space with the induced pullback Wasserstein metric. This geometric structure allows natural gradient descent to be interpreted as a fixed-point iteration on a Riemannian manifold. We accelerate this iteration using Riemannian Anderson mixing method, which exploits historical residuals to capture approximate second-order information without forming Hessians. In the present parametric setting, the manifold is represented by Euclidean coordinates endowed with a point-dependent metric, simplifying retraction and vector transport. We evaluate the method on multimodal sampling, PDE-based inverse problems, aggregation dynamics, Bayesian logistic regression, and generative modeling. Across the tested tasks, Anderson mixing reduces the number of natural-gradient iterations and achieves competitive performance relative to SGD and Adam in terms of both loss-gradient evaluations and wall-clock time. These results suggest that fixed-point acceleration is an efficient tool for improving the computational efficiency of training on parametric Wasserstein methods.

0.17

Knowing When to Intervene: Causal Evaluation of AI Agents

cs.AI cs.LGecon.EMstat.ME ·2026-11-06 ·Ayush Sawarni, Wenqian Xing, Tijana Zrnic, Ramesh Johari, Vasilis Syrgkanis
Abstract

Current evaluations of agentic AI systems primarily rely on benchmarks. However, the static nature of benchmarks makes them uninformative about the value of auditing interventions at deployment time. For reliable deployment, we often need to determine within an agent's trajectory whether or not to intervene, e.g., by verifying the agent's current output or steering it toward different behavior, using either human review or more costly models or time-consuming agentic supervision. The value of such interventions is fundamentally a causal inference question, asking whether the causal effect of such an intervention is beneficial for the final outcome. Estimating this causal effect is particularly challenging due to the dynamic nature of the problem: the impact of an intervention depends on the current context and environment, as well as the agent's future behavior. We develop the first methodologically grounded framework for evaluating the causal effects of randomized interventions within agent trajectories. By leveraging a relatively small dataset of agent trajectories collected under randomized interventions such as verification or screening, we can estimate the dynamic causal effect of a rich class of (contextual) intervention policies. The structural parameters that we estimate can also be used to construct safety leaderboards that capture the risk of various agentic actions, which can be used by practitioners for audit prioritization. Applying our methods to tau-bench Retail, ToolSandbox, and ImpossibleBench, we find that intervention effects vary substantially across agent actions and states. More broadly, our work suggests a rich research agenda at the intersection of causal inference and agentic evaluation, which we believe is essential for the reliable and safe deployment of AI agents in the real world.

0.17

Restoring General and Improving Domain-Specific Safety in Fine-Tuned LLMs Without Retraining

cs.AI cs.LG ·2026-11-06 ·Vishnu Kabir Chhabra, Ding Zhu, Zhihui Zhu, Xueru Zhang, Mohammad Mahdi Khalili
Abstract

Safety-aligned language models are known to suffer degradation in refusal behavior following domain-specific fine-tuning, even when the fine-tuning data is entirely benign. The predominant mitigation strategy, preference optimization on the fine-tuned model, is computationally expensive and requires domain-specific preference data that is often unavailable in practice. We propose SPECTRA, a training-free, weight-level method that restores refusal behavior by transferring the base model's refusal direction into the fine-tuned model's weights via an activation-aware, low-rank update, requiring no gradient-based optimization or domain-specific safety data. The low-rank design restricts the intervention to the subspace most causally active on harmful inputs, leaving benign behavior largely undisturbed by construction. Empirically, SPECTRA substantially improves refusal behavior across a comprehensive safety evaluation, while preserving general coherence, domain-specific knowledge, and producing negligible change in false refusal rates.

0.17

Phantom Consensus: Measuring Social Reasoning in Multi-Agent LLM Decision-Making

cs.MA cs.AIcs.CL ·2026-11-06 ·Disha Sheshanarayana, Rajat Subhra Pal, Manjira Sinha, Tirthankar Dasgupta
publishedbenchmark
Abstract · Accepted at NeurIPS 2026, SocialAgent Workshop

Multi-agent LLM systems increasingly aggregate judgments across model instances, but reasoning about other agents can itself introduce failure. We study a setting in which LLMs shift their public decisions after observing ambiguous social signals while reporting inaccurate estimates of what other agents privately judged, even when the group's recorded private judgments do not support the apparent consensus. We call this failure phantom consensus and study it as a pluralistic-ignorance-style pattern in controlled LLM collectives. To measure this phenomenon, we introduce PHANTOM-Bench, a benchmark that separates isolated private judgments, public decisions after ambiguous social signals, elicited estimates of the group's private judgment, transparency-based recovery, and controls for neutral public text and explicit vote-following. Across nine closed- and open-weight models, ambiguous public signals consistently degrade decision quality and can turn correct private group judgments into incorrect public collective outcomes. Public accuracy drops by as much as 47.8 percentage points, and collective rationality failures reach 47.3% in the most susceptible setting. We observe substantial errors in reported group-private beliefs alongside public-stage failures, while control conditions reveal model-dependent differences in susceptibility to ambiguous and explicit peer signals. Revealing the agents' recorded private decisions repairs many failures without revealing the gold label, although recovery varies across models. Our results show that evaluating multi-agent LLM systems requires measuring not only final agreement, but whether that agreement reflects independently measured private judgments or an apparent consensus.

0.17

Safety-Focused Reinforcement Learning Using Risk-Sensitive Coupled Control

cs.LG ·2026-11-06 ·Jiabin Lin, Shana Moothedath, Vivek S. Borkar
Abstract

This paper investigates the coupled control problem within safety-focused reinforcement learning, addressing the essential challenge of achieving optimal performance while ensuring safety in systems operating within complex environments. We define safety as the requirement to prevent the system from reaching undesirable or hazardous states. We propose a coupled control approach that prioritizes safety by employing effective methods to prevent the system from entering unsafe states while optimizing the performance metric, defined as the average cost. We propose a novel problem transformation that reformulates it into a risk-sensitive problem combined with an average reward criterion. We present a two-timescale Q-learning algorithm where in the faster timescale, the algorithm solves a risk-sensitive control problem to ensure safety, while on a slower timescale, the algorithm focuses on average cost control. We provide theoretical guarantees for the convergence of our algorithms, establishing their significance in ensuring safe, optimal control in dynamic environments. We validated the performance of our approach using GridWorld and CarGoal task in Safety Gymnasium.

0.17

Causal graph fusion across incomplete cohorts and causal steering of tabular foundation models

cs.LG stat.AP ·2026-11-06 ·Yanke Li, Dhiyaa Al Jorf, Ayoube Ighissou, Anke Scheel-Sailer, Robert Riener +1 more
Abstract

Clinical data are split across incomplete cohorts with partially overlapping variables. Causal discovery under missing data and multi-source integration have developed separately, and tabular foundation models cannot take a causal graph without retraining. We make two contributions. First, causal graph fusion learns a partial ancestral graph on the shared variables, fuses it into each cohort's graph under temporal background knowledge and lets Markov blankets guide re-imputation, so missing data are handled inside discovery. Second, causal attention steering compiles the learned graph into the attention of a frozen tabular foundation model at inference. Against known ground truth, fusion dominates pooled and per-cohort baselines in every missingness and heterogeneity condition tested (p<0.001), and steering improves discrimination under distribution shift (p<10^{-5}), a gain that a random graph does not produce. In two spinal cord injury cohorts, fusion orients age, albumin and hemoglobin as ancestors of length of stay, a set no single cohort orients in full, and yields compact, temporally ordered predictor sets that match or exceed automatic feature selection on four of five outcomes.

0.17

Robust Deep Prototype Classifiers via Lipschitz Training

cs.LG cs.AI ·2026-11-06 ·Adia Khalid Khan, Lukas Gehring, Benjamin Paaßen
theory
Abstract

Robustness and interpretability are interrelated and desirable properties of machine learning models. A model that changes its prediction under imperceptible perturbations behaves inconsistently with human-recognizable features and, hence, is not interpretable. Prototype-based classifiers such as Generalized Learning Vector Quantization address both at once. They are interpretable by construction because they make the decision by comparing the input to the prototypical example; and their hypothesis margin yields a closed form robustness certificate. However, on high-dimensional data, these shallow models need deep embedding networks to achieve high accuracy, and this voids the robustness certificate. The margin is a statement about distances in the embedding space, and an unconstrained network can map an arbitrarily small input perturbation to arbitrarily large attacks in the embedding space. We close this gap with a theorem: if the embedding network is locally L-Lipschitz around a point, and the classifier is ε-robust around its embedding, then the full model is ε/L-robust in the input space. Based on this, we propose a Lipschitz training scheme to achieve L-Lipschitz properties for embedding networks. We evaluate three classification heads (linear, linear Support Vector Machine, and GLVQ) crossed with two embedding networks (standard and Lipschitz-constrained) on MNIST, CIFAR-10, and Imagenette-160, under FGS, PGD and C\&W attacks. The evaluation results show that combining a prototype-based head with a Lipschitz-constrained embedding provides the strongest robustness. This suggests that the Lipschitz constraint provides an additional robustness benefit in addition to the advantage provided by the prototype-based head. Since the Lipschitz constraint is enforced by sampling rather than by construction, the reported robustness is empirical rather than certified.

0.16

ReSUM: Reasoning-Driven Multi-Agent Framework for Query-Focused Summarization

cs.MA ·2026-11-06 ·Sonali, Deeksha Varshney, Asif Ekbal
Abstract · 10 pages, 4 figures

LLMs excel at language tasks, but long-meeting QFS remains difficult due to complex speaker dynamics, sparse key information, and limited data. Meetings involve disfluencies, long transcripts and complex discourse. QFS have high risk of information omission, hallucination, even for advanced models such as GPT-4-mini and Gemini. Existing end-to-end methods compress retrieval, evidence extraction, and summary generation into a single pass, failing to handle the complex reasoning required for targeted meeting summaries. We propose ReSUM, a modular multi-LLM framework coordinated by a planner that orchestrates retrieval and summarization agents. By distilling teacher reasoning into query- and plan-conditioned retrievers and summarizers, ReSUM enhances long-context robustness, minimizes irrelevant exploration, and generates more faithful and comprehensive query-focused summaries. Experiments show consistent gains on query-focused meeting summarization benchmarks.

0.16

On the (non-)closedness of the set of ReLU networks

cs.LG math.CAmath.FAmath.NA+1 ·2026-11-06 ·Eloi Martinet, Leon Bungert
theory
Abstract · 27 pages, 6 figures

We prove that the set $\mathcal{N}_N$ of ReLU networks with one hidden layer and at most $N\in\mathbb N$ neurons is weakly closed in $W^{1,p}(Ω)$, for a bounded connected open set $Ω\subset\mathbb R^d$. Consequently, every coercive, weakly lower semicontinuous energy on this space attains its minimum over $\mathcal{N}_N$. This covers the Deep Ritz method with convex integrands, Sobolev training and Lipschitz learning. Moreover, as the weight decay parameter tends to zero, minimizers of the regularized problem converge, up to subsequences, to minimizers with minimal parameter norm. We also prove that the set of ReLU networks with two hidden layers of bounded widths is not closed in $W^{1,p}(Ω)$ or in $C(\overlineΩ)$. This disproves a conjecture of Petersen, Raslan and Voigtlaender.

0.16

Nudging Decision Models: Do Digital Nudges Steer Fast, Generation-Free AI Agents?

cs.AI ·2026-11-06 ·Sascha Kaltenpoth, Haya Halimeh, Kevin Bösch, Oliver Müller
Abstract · Under Review

LLM-based agents increasingly automate business and consumer tasks but are slow and expensive. Decision models, a new class that chooses among options without generating text, are far faster and less costly. Developers call them "System One models," referencing dual process theory. We ask whether such models can be steered by digital nudges online and whether, as the theory predicts, automatic nudges affect them more than reflective ones. We adapted an online shopping experiment, previously run with humans and LLM agents, to three decision models (Jev, Decision-1, d1). Across 600 runs per model, they chose a preselected default in 97.5-99.9% and a product with social influence information in 75-96% of choices, versus 47-49% in the control. All responded more strongly to the automatic nudge, the opposite of what has been reported for reasoning-capable LLM agents. We provide first evidence that non-generative decision models are susceptible to digital nudges in a System 1 pattern, implying agent-type-specific nudge effects.

0.16

Auditable Human-Agent Teams for Autonomous X-ray Experimentation

cs.AI cs.MA ·2026-11-06 ·Gunik Luthra, Ronald J. Pandolfi, Benjamin T. Diroll, Ricky Ronquillo, Ruipeng Li +4 more
Abstract

Closed-loop autonomous experimentation can accelerate scientific discovery, but implementation effort limits accessibility. Agentic systems may lower this barrier, although sustained operation on physical instruments remains largely unexplored. Here, using coding-agent infrastructure grounded in validated scientific methods, we demonstrate autonomous synchrotron X-ray experiments on nanoparticle superlattices under evolving, human-defined objectives. Complete human-agent and agent-agent transcripts enable an audit that reconstructs decision provenance and characterizes the division of cognitive labor. Across three experiments, 61-67% of messages involved interpretation or judgment, while the share of methodological and scientific contributions increased from 46% in the setup-focused first experiment to 56-61% in later experiments. Provenance showed that 59-68% of error corrections involved multiple participants and that about half of all decisions incorporated the input of another participant. These results may guide design principles for multi-agent scientific systems and inform future agent harnesses, model training, and accessible autonomous experimentation.

0.16

DisMUSE: An Efficient Distributed Singular-Spectrum-Based Optimization Approach with Theoretical Convergence Guarantee

cs.LG math.OC ·2026-11-06 ·Zhen Qin, Zhiyao Zhang, Yingbin Liang, Jia Liu
Abstract

In recent years, singular-spectrum-based optimizers have attracted increasing attention due to their ability to preserve the matrix structure of neural network gradients, thereby enabling geometry-aware optimization. However, most singular-spectrum-based (SSB) approaches remain centralized, limiting their use in distributed learning infrastructures that exploit massive parallelism. Distributed SSB optimization is still at an early stage, with open questions about convergence guarantees and communication efficiency. To fill this gap, we propose the Distributed Momentum-based Spectral Estimator (DisMUSE), an efficient distributed SSB optimization framework for both data-center and federated settings. Using Muon-style exact orthogonalized momentum as the spectral update, DisMUSE integrates a recursive variance-reduced momentum estimator and achieves an $O((KT)^{-1/3})$ convergence rate, where $K$ and $T$ denote the number of nodes and local iterations per node, respectively. This rate matches the best-known rates for variance-reduced nonconvex distributed optimization. Building on this variance-reduced framework, we study two extensions. First, our DisMUSE method with finite Newton-Schulz (NS) iterations preserves the same convergence order under a uniform normalized support-spectrum condition, up to a multiplicative factor. Second, we develop a communication-efficient variant based on low-rank (LR) momentum transmission and quantify the error it introduces to the stationarity bound, characterizing the trade-off between accuracy and communication cost.

0.16

Transformer-FSQ: Discrete Latents for Extreme KV-Cache Compression

cs.LG ·2026-11-06 ·Andras Szecsenyi, Thomas L. Lee, Vaisakh Shaj, Elliot J. Crowley, Amos Storkey
Abstract

Private, personalised language models must run on local consumer devices, where large, frontier models cannot fit, and retaining long contexts is non-trivial. Sequential memory---the key-value (KV) cache---grows linearly with context length, quickly dwarfing the model itself: for Qwen3-1.7B-Base, 200,000 retained entries require 21 GiB, 6.6 times more than the model weights. Yet the KV-cache compression methods we measured fail beyond moderate compression ratios. We introduce Transformer-FSQ, addressing extreme KV-cache compression by replacing standard key and value projections with a finite scalar quantisation bottleneck. Our cache stores a packed vector of discrete integer codes per token, with no per-token quantisation metadata. Converting a pretrained transformer trains only these projections on 0.25 tokens per original parameter. We evaluate Transformer-FSQ against five baseline families on models from 0.6B to 8B parameters. On long-context retrieval (RULER), from 12$\times$ to 97.5$\times$ Transformer-FSQ is the most accurate method we evaluate, across scales. A fused Triton kernel decodes directly from the integer codes; with baselines given their authors' kernels where available, it decodes at $1.01\times$ (95% interval $0.98$--$1.05$) the throughput of native SDPA attention on the uncompressed Qwen3-8B-Base at $32\times$ and $1.54\times$ at $97.5\times$. Extreme KV-cache compression is therefore a practical way to extend how much context a fixed device can hold.

0.15

Travel Cost Calibration for Multi-Agent Target Assignment and Pathfinding

cs.AI ·2026-11-06 ·Yu Kumagai, Ryohei Banno, Keisuke Okumura
Abstract

Target assignment and pathfinding (TAPF) is an appealing extension of classical multi-agent pathfinding (MAPF): each agent has a set of targets, and the planner must both assign a target to each agent and plan collision-free paths for them. A trivial approach is to compute an assignment that minimizes the total single-agent shortest-path distance over agent--target pairs, and then solve the resulting MAPF instance. However, this approach falls short because the shortest-path distance remains a proxy that ignores interference among agents. Assigning targets based on the costs agents actually pay, which depend on how every other agent is assigned and routed, would yield a better final outcome. To this end, we propose an iterative TAPF method that improves target assignments by calibrating travel cost estimates from fast, suboptimal MAPF feedback. Experiments show that our method produces solutions close to optimal on small instances and outperforms existing planners in solution quality on larger ones.

0.15

From Emergence to Reversibility: Toward Reversible Multi-Agent Systems

cs.MA ·2026-11-06 ·Zhengye Han, Linda Ng Boyle
Abstract

Multi-agent systems become capable through institutional mechanisms, such as rules that route information to agents that act on it. These mechanisms are also natural places to restrict the system. Yet as agents learn, share memory, delegate, and repair failures, the collective can find other paths to the same effects: a restriction on the original mechanism still operates but no longer stops them, while routine performance and compliance checks pass. We propose \textbf{Reversible Multi-agent Systems} as a research agenda: collectives that grow while an authorized supervisor can still restrict specified effects, keep them restricted as agents adapt, and reconstitute useful cooperation under renewed authority. What is reversed is the ability to produce a specified effect; restrictions must hold even if learned knowledge persists. The goal is to preserve effective intervention, not a particular mechanism: at every step of growth, some authorized intervention must keep a specified prohibited effect rare while maintaining required service. We identify three research problems: tracking what capabilities depend on, handing off control before an old mechanism loses its effect, and separating legitimate repair from unauthorized reconstruction. We outline how to evaluate control across growth paths, back control claims with evidence, and allocate the rights to restrict and restore.

0.15

AutoRefine: Auto-Calibrating Code Judges from Workflow Outcomes

cs.AI cs.SE ·2026-11-06 ·Shervin Shahidi, James Poon, Hang Xie, Kenneth O. Xu, Robin Håkanson +3 more
Abstract · 5 pages, 3 figures, prepared for submission to CAIN 2027 industrial track

LLM judges increasingly serve as quality gates for code changes that have no test oracle, but a judge prompt must be calibrated for its task and recalibrated as the system it judges evolves, while hand labels for the calibration are resource intensive and go stale. Development workflows emit outcome labels for free (tests pass or fail, changes land or are abandoned), but these labels are noisy. We present AutoRefine, a loop that rewrites a deployable judge's prompt to reproduce outcome labels, with an informed teacher that sees the judge's input plus the outcome, explains each label it keeps, and abstains when the outcome cannot be explained from what the judge sees. On a held-out dataset with an execution oracle, calibration raised chance-corrected agreement (Scott's $π$) from 0.36 to 0.60. On a deployed agent's outcome labels, with 25\% flipped at random to introduce noise, the teacher set aside flipped \textsc{fail} labels at more than twice the rate of genuine ones, cutting overall label noise from 25\% to 20\%. Recalibrating that agent's judge using AutoRefine raised its agreement with merge/abandon outcomes on 17 independently held-out diffs from the current workflow ($π$ 0.16 to 0.51); on 16 diffs from older workflows, the previous judge stayed ahead, illustrating the need for recalibration as the agent evolves. By abstaining, the teacher not only removes noisy labels but also flags changes whose outcome the judge's input cannot explain, telling a team when a judge needs better inputs rather than a better prompt.

0.15

Analyzing Code Provenance in LLMs for Programming Education

cs.AI cs.SE ·2026-11-06 ·Subhankar Maity, Faiza Belbachir, Omar Boucelma, Guilherme Medeiros Machado
Abstract

Large language models (LLMs) play an increasing role in programming education, challenging traditional assessment methods as they can generate fully functional code instantly. This disrupts the reliability of student submissions as indicators of learning and effort. To address this issue, code provenance is essential for tracing the origins of code and distinguishing student-authored, AI-generated, and externally sourced components. We extend the Min-K% probability approach by incorporating synthetic non-member code generation and apply it to Python code for membership inference and provenance tracking. Evaluations cover free LLMs such as GPT-2, DistilGPT-2, and DialoGPT, as well as state-of-the-art commercial models including GPT-5, Gemini 2.5, Mistral, DeepSeek, and LLaMA 3.3. Before evaluating our approach on higher-education student assignments, we validate its feasibility using Stack Overflow data. The results show that free models enable more sensitive code memorization detection than commercial models. By systematically comparing these LLMs, we assess trade-offs between detection accuracy and privacy protection, taking an initial step toward effective assessment of AI-assisted programming education through combined provenance detection and process visibility.

0.15

AUTEUR: Acoustic Plan Learning for Red-Teaming Text-to-Audio Models

cs.AI cs.SD ·2026-11-06 ·Jiashu Zhang, Jiaming He, Zheming Zhang, Kaizheng Wang, Bing Li +2 more
code/data
Abstract

Automated red-teaming of text-to-audio models requires test prompts that pass textual screening and expose harmful acoustic content. The central challenge is to connect the sounds produced by the target model to both reusable test generation and prompt revision. Our key idea is to represent test descriptions as structured acoustic plans that organize foreground sounds, context, and vocal attributes. To this end, we propose acoustic plan learning, a framework that learns and refines these plans using waveform feedback. We train the plan generator through supervised fine-tuning followed by reinforcement learning with audio feedback while keeping the target audio model fixed. At inference time, the same trained model generates initial plans and uses feedback to revise their context and vocal attributes while preserving the foreground. Experiments across four target systems show attack success rates of 79.4% on TangoFlux and 95.6% on Kling. The resulting test prompts also transfer across open-source and closed-source models, indicating shared acoustic vulnerabilities.

0.15

Bridging Spiking and Continuous Dynamics in Spiking Neural Networks: A Muscle-Inspired Model

cs.NE cs.LGeess.AS ·2026-11-06 ·Maryam Soltani, Philip N. Garner
Abstract

Transforming the spike trains produced by spiking neural networks (SNNs) into continuous signals through biologically inspired mechanisms remains an open challenge in generative and regression tasks. In biological systems, a similar transformation occurs when muscles integrate neural spikes to produce smooth, coordinated movements. Inspired by this mechanism, we introduce a muscle-response model, a biologically motivated dynamical layer that converts spike trains into continuous representations. Our approach uses second-order dynamics to model the temporal integration of spiking activity. We investigate how the number of spiking neurons and temporal resolution affect the quality of the generated signals in a text-to-speech task, showing that increasing representational capacity has a greater impact than increasing temporal resolution under the evaluated configurations. Furthermore, an analysis of the learned dynamics reveals that, without explicit constraints on the response type, most neurons converge toward non-oscillatory behavior. These results demonstrate the potential of biologically inspired dynamical decoders for bridging spiking neural computation and continuous signal generation.

0.14

Knowledge Distillation from Neural Networks to Differentiable Logic Networks

cs.LG ·2026-11-06 ·Hee Yun Suh, Niraj K. Jha
Abstract

Differentiable logic networks (DLNs) are interpretable, edge-friendly classifiers that learn both their logic operators and their connections through gradient descent, achieving accuracy competitive with multi-layer perceptrons at orders-of-magnitude lower inference cost. Their accuracy, however, is bounded by the problems of vanishing gradients, scalability, and lack of an ecosystem, leaving a gap relative to conventional neural networks on several datasets. We introduce knowledge distillation for DLNs to narrow this gap. We distill from a trained multi-layer perceptron teacher into a DLN student, integrating the distillation objective into a two-phase training loop without altering the architecture or the inference procedure. The teacher is discarded after training. Hence, the distilled model retains the interpretability, sparsity, and inference cost of a conventionally trained DLN. We investigate this approach through four variants: a joint hyperparameter search that optimizes the distillation weight and temperature alongside the architecture, a grid search over these two hyperparameters with the architecture fixed, and two single-phase variants that employ distillation during only the neuron-function or only the connection phase. On 20 public tabular classification tasks, distillation improves balanced-class accuracy at negligible cost to gate-level operation counts and model size, though the benefit varies by dataset. Performance improvements of up to 15% are observed, in some cases surpassing the teacher itself.

0.14

A Tight Hierarchy for Chain of Thought

cs.LG cs.CC ·2026-11-06 ·Thanh Le, A. Pavan, N. V. Vinodchandran
theory
Abstract

Chain-of-thought (CoT) reasoning has emerged as a powerful paradigm for extending the computational capabilities of transformer-based models. A growing body of recent work has begun to uncover the fine-grained computational power of transformers augmented with CoT, as well as their connections to classical complexity theory. In this work, we advance this line of research by establishing the first hierarchy theorem for CoT-based computation. Our main result shows that for any reasonable resource bound $r(n) \geq n$ and any $\varepsilon > 0$, \[ \mathrm{CoT}(r(n)) \subsetneq \mathrm{CoT}(r(n)^{1+\varepsilon}), \] where $\mathrm{CoT}(r(n))$ denotes the class of decision problems solvable by transformers using $O(r(n))$ CoT reasoning steps on inputs of length $n$. Our result shows that increasing the number of CoT reasoning steps provably yields strictly greater computational power, establishing CoT steps as a fundamental resource in transformer-based computation. Unlike classical hierarchy theorems, which are typically proved via diagonalization, our approach leverages known coarse-grained simulations between CoT-augmented transformers and Turing machines in both directions. Building on these connections, we adapt the classical padding technique from complexity theory to this setting to obtain a tight hierarchy. A key technical contribution of our work is the implementation of padding within the transformer-based CoT model.

0.14

BiMuon: A Geometry-Aware Stochastic Bilevel Optimization Approach

cs.LG math.OC ·2026-11-06 ·Zhen Qin, Zhiyao Zhang, Yingbin Liang, Jia Liu
Abstract

Stochastic bilevel optimization provides a general framework for many machine learning applications, such as hyperparameter learning, data reweighting, language model alignment, and actor-critic reinforcement learning. However, most existing bilevel optimization methods treat matrix-valued gradients as flattened vectors and perform optimization in the resulting vectorized space, ignoring the matrix structures common in modern neural networks. Moreover, the lower-level (LL) problem curvature can induce significant spectral imbalance in the upper-level (UL) geometry. Such limitations of existing bilevel optimization methods motivate us to develop BiMuon, a geometry-aware stochastic bilevel optimization method that integrates implicit hypergradient estimation in bilevel optimization with Muon-style orthogonalized updates at the UL. To our knowledge, BiMuon is the first stochastic BLO method with Muon-style orthogonalized matrix updates and an $O(ε^{-4})$ iteration-complexity guarantee, matching the best-known rates achieved by Muon-type methods based on matrix parameterizations. To improve computational and parallel efficiency for models with multiple matrix parameters, we further study BiMuon for blockwise diagonal matrix approximation and derive a relative stepsize allocation rule based on UL progress and blockwise LL sensitivity. We evaluate BiMuon on representation learning and blockwise bilevel optimization, providing empirical support for our methods.

0.14

Not All Entropy Is Equal: Reward-Conditioned Entropy Allocation in Reinforcement Learning with Verifiable Rewards

cs.LG cs.AI ·2026-11-06 ·Song Yu, Li Li, Zhenming Wu
position
Abstract · 8 pages, 6 figures

Policy entropy is widely used to diagnose exploration and collapse in reinforcement learning with verifiable rewards (RLVR), yet its magnitude does not consistently track reasoning performance across optimization objectives. We argue that this ambiguity arises because global token entropy conflates uncertainty in successful and failed trajectories, diversity across correct solution strategies, and variation among realizations of the same strategy. We introduce \textbf{Reward-Conditioned Entropy Allocation (RCEA)}, a hierarchical diagnostic framework that separates complete-trajectory entropy from average token entropy, conditions uncertainty on verifier outcome and process validity, measures correct-solution-mode retention, and connects these population-level quantities to local covariance-driven updates. In controlled sweeps over six RLVR objectives at two model scales, entropy trajectories do not induce a stable ordering of reasoning accuracy. Token entropy is higher for verifier-failed than successful trajectories under both broad separately aggregated estimates and a common-question paired analysis, whose confidence intervals exclude equality at 10 of 12 displayed endpoints. A local covariance approximation predicts the direction of adjacent-checkpoint entropy change in 95.5\% of nonfailed transitions; a reward-stratified endpoint proxy further shows that positive chosen-token covariance is more tail-concentrated within failed prefixes. Every reportable trained endpoint exhibits some contraction in independently audited correct-solution-mode breadth, whose positive association with large-\(k\) coverage remains exploratory. These results show that entropy collapse should be diagnosed by which uncertainty is removed, rather than by global entropy alone.

0.13

SIFR: Stable Signed Flow Reinforcement

cs.LG cs.AIcs.CV ·2026-11-06 ·Runlong Liao, Xingqian Xu, Lizhang Chen, Baiyu Su, Yufan Huang +5 more
Abstract

Reward-based fine-tuning of flow models requires effectively promoting high-reward and suppressing low-reward samples, while maintaining training stability over many rounds of iterative updates. Iterative reward fine-tuning provides a simple starting point, but the positive reward weights give no explicit unfitting signal, and repeated reweighting concentrates the distribution and causes collapse. We propose Stable Signed Flow Reinforcement (SIFR), which replaces the reward weights by signed advantages to supply the missing unfitting signal, and constrains the velocity drift from a reference model to stabilize the updates, with a Lagrange multiplier that adapts the regularization strength to a prescribed drift budget. We further show that existing flow RL methods fall within this framework: they implicitly minimize the signed loss, differing in their regularization choices. We validate SIFR on SD3.5 Medium (2.5B) and FLUX.2 Klein (9B) and scale it to Cosmos 3 Super Text2Image (64B): across all 12 backbone--reward pairs with preference and rule-based rewards, it matches or exceeds existing methods, reaches their peak rewards much faster, and keeps improving over long training where they plateau or collapse.

0.13

PHISHREV+: An ASP-Based Post-Hoc Defeasible Machine Learning Framework for Phishing Detection

cs.AI cs.LG ·2026-11-06 ·Mainak Sen, Kumar Sanka Ray, Amlan Chakrabarti
benchmark
Abstract

Phishing website detection remains a critical cybersecurity challenge. Although machine learning classifiers achieve high accuracy on benchmark datasets, they suffer from a fundamental limitation: once trained, their decision boundaries are fixed and cannot incorporate new contextual evidence without retraining the model. This work presents \textbf{PHISHREV+}, a hybrid framework that instantiates the notion of \textit{post-hoc defeasible machine learning}. This framework does not replace the classifier. It treats the initial classifier predictions as defeasible beliefs and revises them through an Answer Set Programming based non-monotonic reasoning(NMR) layer. The framework operates in two phases. In the machine learning phase, four classifiers (SVM, KNN, DT and RF) are trained on the 87 feature benchmark dataset. In the post-hoc NMR phase, four defeasible rules derived from exploratory data analysis of the training set, and validated through Chi-square and Mann-Whitney U tests, formally revise initial predictions. The proposed framework achieves consistent false positive reduction across all our classifiers SVM (41$\rightarrow$39), KNN (47$\rightarrow$46), DT (62$\rightarrow$61), and RF (52$\rightarrow$51) while maintaining complete false negative parity with zero increase overhead. A comprehensive ablation studies across seven rule configurations confirms that the hierarchical combination of all four defeasible rules is necessary to achieve this optimal balance. Instance level explainability is provided through eXplainable ASP (XASP), which generates formal derivation graphs showing which defeasible rules fired and why the initial belief was retracted, a capability absent from all existing phishing detection approaches.

0.13

One Run Is Not Enough: A Monte Carlo Protocol for Evaluating LLM Reliability on Data Science Tasks

cs.LG ·2026-11-06 ·Wenqi Ni, Ozan Evkaya, Miguel de Carvalho
benchmark
Abstract

Large language models (LLMs) have shown promise for automating data science workflows. Benchmarks such as DataSciBench (2026) and StatQA (2024) have advanced the evaluation of their capabilities across data science and statistical tasks. Assessing their reliability requires examining both the quality of their responses and their consistency across repeated runs, task types, and difficulty levels. With this aim, we introduce DS-QA, a benchmark comprising 100 data science and statistics questions drawn from UK university courses, spanning 16 datasets and three difficulty levels. We develop a three-layer evaluation framework that combines general response properties (runtime, verbosity and text similarity), coarse-grained metrics (accuracy, code executability and completeness), and task-specific metrics. We evaluate OpenAI models using a Monte Carlo protocol with 100 repeated runs. The results reveal substantial variability in responses to data visualisation and supervised learning tasks, particularly at medium and hard difficulty levels, highlighting the importance of evaluating consistency alongside response quality. We also find that an increase of 0.1 in the word verbosity ratio corresponds to an increase of about 0.013 in the token verbosity ratio.

0.12

Alignment by Confident Debate: Preference Learning from Multi-Agent Debate

cs.AI ·2026-11-06 ·Nikhil Verma, Dingxian Wang, Mohammad Reza Rezaei, Andrew Rabinovich, Zhao Chen
Abstract · 53 pages, 9 Figures, 31 Tables

Post-training language models for reasoning typically rely on gold answers, task-specific verifiers, or learned reward models, limiting scalability when such signals are scarce. Multi-agent debate offers a self-supervised alternative, but converting debate into reliable training supervision remains challenging: majority voting weights agreeing responses equally, lacks a principled tie-breaking rule, and is ill-defined for open-ended tasks without a canonical answer; likelihood-based scoring can conflate fluency with correctness; and homogeneous committees may reinforce correlated errors. We introduce Alignment by Confident Debate (AbCD), which converts debate traces into preference data without gold answer labels. A sampling-diverse committee debates each problem, after which final answers are clustered by task-level semantic equivalence. We use verifiable tasks (mathematical reasoning, multiple-choice quantitative QA, and executable code-generation) as a controlled test bed: their answer keys or executable checkers let us evaluate the quality of debate-derived supervision against ground truth, even though the oracle itself never uses gold labels. Post-training on the resulting preferences outperforms inference-only ensembling and debate, as well as SFT on the same traces, across five model families spanning 1B to 8B parameters and four datasets, improving consensus accuracy, with a maximum gain of $21.5_{pp}$. Ablations show that the approach transfers to a larger model and to code generation, while also revealing limitations of heterogeneous committees and agreement-based weighting. Overall, the results suggest that debate traces can yield useful preference supervision when answer equivalence is combined with confidence-weighted disagreement.

0.12

WorldMind: Test-Time Scaling for Visual Spatial Reasoning through Latent World Simulation

cs.AI ·2026-11-06 ·Haoming Wang, Yongjian Fu, Ju Ren, Wei Gao
Abstract

Vision-Language Models (VLMs) have continuous deficiency in spatial reasoning tasks that require \emph{mental spatial transformation}, such as reasoning under hypothetical changes in viewpoint or position, known as perspective-taking tasks. Existing approaches either scale linguistic reasoning or augment VLMs with static geometric representations, but neither directly enables active transformation of the represented scene. World models offer this capability, yet existing systems typically communicate through synthesized RGB images, introducing a latent-to-pixel-to-latent bottleneck. We introduce \emph{WorldMind}, a framework for test-time scaling of spatial reasoning through latent world simulation. Given a query and observation, the VLM selects useful spatial transformations, while a world model executes them and exposes aligned intermediate latent features directly to the VLM. The resulting latent evidence is incorporated into subsequent reasoning steps, enabling iterative spatial simulation without repeatedly decoding imagined states into pixels. To optimize this interaction, WorldMind further employs process-verifiable reinforcement learning that rewards reasoning progress only when the associated action and latent transition are valid. WorldMind thus scales spatial reasoning by allocating additional inference-time computation to active world simulation rather than generating more language or pixels. Across VSI-Bench, MMSI-Bench, and MindCube-Tiny, WorldMind achieves an average improvement of 16.0% over the base VLM and 8.3% over the strongest RGB-based world simulation.

0.12

Calibrating your confidence requires thinking of all the ways you could be wrong

cs.LG stat.ML ·2026-11-06 ·Marco Miani
Abstract · PhD Thesis, Technical University of Denmark (DTU), 2026

A primary limitation of modern deep learning is the miscalibration of model confidence; traditional architectures fundamentally lack a mechanism to quantify their own uncertainty. Larger models and larger datasets mitigate this issue, but they do not resolve the underlying structural deficiency. Bayesian methods offer a principled uncertainty-aware learning framework through stochastic parameters. Yet, despite the established benefits, the inherent computational overhead often compromises either scalability or quality. This trade-off currently prevents the widespread adoption of Bayesian methods. This Thesis bridges the gap between high-quality and low-complexity by introducing a novel distribution on parameter space: a degenerate Gaussian whose covariance is a projection matrix aligned with the null space of the Fisher information matrix. We demonstrate the scalability of this Bayesian posterior through two numerically efficient algorithms: one for posterior sampling and another for the direct computation of predictive confidence. Empirically, we show State-Of-The-Art confidence calibration and out-of-distribution detection for large models. We provide theoretical grounding for this approach. We study the degeneracies inherent in Amari's information geometry and highlight its most relevant component: the split of parameter's perturbations into two orthogonal subspaces. From this perspective, our posterior naturally emerges as the simplest choice that captures this split, assigning unitary variance to one subspace and zero variance to the other. Moreover, through the tools of differential geometry, we utilize this split to identify and rectify a critical misalignment issue of the non-linearized Laplace Approximation. Finally, with a focus on PAC-Bayes generalization bounds, we pave the way for a deeper understanding of the learning capabilities of overparametrized models.

0.12

FAIDER: Fairness-Aware Item-Level Decoding for Large Language Model Recommenders

cs.AI ·2026-11-06 ·Ziwei Wu, Jingrui He
Abstract

Large language model (LLM) recommenders can produce semantically different recommendations when a user's protected attribute changes, even if the interaction history remains fixed. Controlling this disparity is challenging when model weights and full-vocabulary logits are inaccessible. We propose FAIDER (Fairness-Aware Item-level DEcoding for Recommendation), which incorporates fairness into list construction by selecting complete item candidates rather than manipulating token-level logits or repeatedly repairing finished outputs. The central idea is to separate the operational criterion for a violation from how strongly generation is constrained. Conformal-inspired empirical calibration fixes a cross-group semantic-disparity threshold, while online decoding applies either hard filtering or a soft penalty adjusted through dual-inspired feedback to target a prescribed violation rate. Separately, we establish a pathwise guarantee: under a uniform enforceability assumption, the soft decoder's empirical violation rate exceeds its target by at most $O(1/T)$. Experiments across recommendation benchmarks and LLM backbones show that Soft FAIDER improves both measured semantic fairness and ranking quality over the evaluated prompt-repair baselines. It also generally achieves a more favorable fairness-utility balance than the hard variant, indicating that stricter per-recommendation constraints do not necessarily yield lower measured disparity.

0.11

Can Agentic Systems Close the Gap in Radiology with Production AI Tools?

cs.AI cs.LG ·2026-11-06 ·Aubin Charley, Julien Khlaut, Charles Corbière, Pierre Manceron, Corentin Dancette +1 more
Abstract

Radiology AI has developed along two paths: specialized models that perform well on narrow tasks but cover few clinical findings, and generalist models that address a broader range of tasks but remain prone to hallucination and inaccurate measurements. Combining the two through agentic systems offers a route towards comprehensive radiology reporting, but published systems are tied to one modality or body part. We describe a radiology agentic framework that standardizes how an orchestrator calls modular tools to produce a structured report, and use it to measure what such a system can do with the tools from FDA-cleared radiology AI devices. Report generation is simulated on a large multi-modality report corpus, with the orchestrator given either a catalog of finding classes consolidated from the cleared devices, each available on the modalities it is cleared for, or an unlimited catalog in which any class can be requested, and planning either from the clinical indication or from the target report. Tool outputs are simulated from the reference report, so the comparison isolates tool availability and orchestration from tool error. With the cleared catalog and the indication alone, the orchestrator recovers under a third of the clinically actionable findings; recovery improves substantially with the unlimited catalog, and further when the plan is written from the target report. Attributing every missed finding to a cause shows that most are classes no cleared device offers for the study's modality, and that a third are classes or tools that were offered and not called. Adding the most frequently missing tools closes most of the gap to the unlimited catalog on held-out reports, whereas a smaller orchestrator with the unlimited catalog gives up the whole gain. Tool availability and effective orchestration are thus complementary constraints on comprehensive AI-assisted radiology reporting.

0.11

From PAC bounds on the generalization error to PAC bounds on the parameter estimation error

cs.LG ·2026-11-06 ·Mihaly Petreczky, John Leth, Emma Lukacs, Balint Daroczy
Abstract · 10 pages, 0 figures, under submission to SysID2027

Probably Approximately Correct (PAC) bounds on the generalization error (the difference between the empirical prediction error computed on the training data and its expectation) are an established tool in statistical machine learning, and there have been several extensions of these bounds to dynamical systems. In parallel, in recent years, there has been growing interest in finite-sample bounds for parameter estimation error in system identification. The latter bounds tend to be algorithm-specific, relying on reformulating the identification algorithm as a least-squares regression problem. This is in contrast to classical PAC bounds, which are algorithm-agnostic and can be applied to any identification algorithm based on minimizing a cost function involving the empirical loss. In this paper, we show how oracle bounds derived from PAC bounds can be used to derive finite-sample bounds for dynamical systems. As an illustration of this approach, we present finite-sample parameter-estimation bounds for linear and bilinear systems with bounded inputs and noise. These bounds apply to identification algorithms based on minimizing a regularized empirical loss function. The key ingredient is to establish an oracle inequality, i.e., an inequality between the true prediction error of the learned model and the minimum achievable true prediction error, and then use either the convexity of the true loss or system-theoretic properties of the loss function to derive bounds on the parameter estimation error. We also analyze the rates of the obtained bounds and compare them with those of classical asymptotic results.

0.11

Zero-Query Transferable Pretend Benign Attacks with Amortized Online Generation for Cooperative Perception

cs.AI ·2026-11-06 ·Hongwei Lin, Xun Huang, Chenglu Wen, Cheng Wang
Abstract

Cooperative perception enables an ego agent to acquire information from collaborators beyond its own sensors' reach. Yet, the most beneficial messages are often the hardest to verify from local observations. Pretend Benign (PB) exploits this asymmetry by targeting regions with weak local verifiability and applying multi-objective perturbations, making harmful feature changes appear to be benign assistance. However, PB is victim-dependent: selecting regions and optimizing perturbations require the victim model's architecture, parameters, and gradients, limiting transfer to unseen victims. A transferable attack must capture properties shared across cooperative detection models rather than those specific to one network. We address this by building a victim-agnostic attack from a heterogeneous ensemble of surrogate models and encoding its capabilities in a reusable dictionary to reduce per-message generation cost. Specifically, Transferable Pretend Benign (TrPB) transfers PB's region selection criterion and multi-objective perturbation strategy across heterogeneous surrogates to construct an offline attack. We further propose RT-TrPB, which encodes transferable regions and perturbation patterns into a region predictor and reusable perturbation dictionary, amortizing per-message attack generation. Experiments on widely used cooperative perception benchmarks show that RT-TrPB bypasses the consensus-based defenses ROBOSAC and FLD with success rates approaching 100% and generates each perturbation in under 56 ms. To counter PB-style attacks masquerading as benign collaboration, we also propose STTrust, which verifies received messages using spatial and temporal consistency and incorporates identity-level trust. STTrust detects over 90% of PB-style attacks and restores perception performance to near the attack-free level.

0.10

SSF-Net: A Spatial-Spectral Fusion Network for Accurate PDE Operator Learning

cs.LG ·2026-11-06 ·Siqi Zhao, Flora D. Salim
benchmark
Abstract

Machine learning methods for solving partial differential equations (PDEs) significantly improve computational efficiency over traditional numerical solvers such as finite element method (FEM). However, PDE dynamics inherently span both local and global dynamics: disturbances propagate and decay gradually through the domain, while global coupling effects can influence the whole system simultaneously. Capturing both types of dynamics with a low complexity model remains a key challenge. To address this, we propose the Spatial-Spectral Fusion Network (SSF-Net), built around a dual-branch Spatial-Spectral Fusion (SSF) block comprising a spatial branch and a spectral branch. The spatial branch employs a multi-directional Mamba, which mitigates its inherent causal, single-direction scanning bias and capture local interactions with reduced directional bias at linear complexity. The spectral branch leverages a Fourier transform with a block-diagonal MLP to efficiently capture global spectral features with reduced parameter count and complexity. The SSF block is embedded in a multiresolution convolutional encoder-decoder, which extracts multi-scale features. We evaluate SSF-Net on a comprehensive PDE benchmark spanning multiple task types and assess its long-horizon stability. Compared with attention-based and convolution-based baselines, SSF-Net achieves lower prediction error with fewer parameters while retaining near-linear asymptotic complexity in the number of spatial tokens.

0.10

Adaptive Budgeted Data Acquisition for Robust Off-Policy Evaluation from Multiple Loggers

cs.LG stat.ML ·2026-11-06 ·Amitakshar Biswas, Ruoqing Zhu
Abstract

We consider off-policy evaluation when additional data can be purchased from several existing logging policies (or loggers) with heterogeneous costs per observation. While existing work on multiple loggers presumes a fixed number of observations from each logger, we study how to allocate a limited budget across them sequentially to evaluate a finite class of target policies. We propose Adaptive Greedy, an algorithm that purchases observations sequentially from the affordable logger that has the most cost-normalized variance reduction across the target policies. It turns out that each such purchase is an 'approximate conditional-gradient step' for the corresponding design problem. Under mild assumptions, we establish that the empirical allocation converges almost surely to the optimal design at an explicit rate. We also derive the asymptotic normality of the final pooled estimator and show that it attains the semiparametric efficiency bound. Simulations and a semi-synthetic application based on the AIDS Clinical Trials Group Study 175 show that Adaptive Greedy outperforms standard allocation strategies, remains accurate if stopped early, and matches the final accuracy of repeated integer programming at a fraction of its computational cost.

0.10

Mind the Divergence: Divergence-Guided Reliability-Aware Fusion and Adaptive LLM Prompting for Conversational Emotion Recognition

cs.AI cs.SD ·2026-11-06 ·Jingwen Wang, Jiao Li, Yuhang Qian, Yuyin Wang, Min Zhang
Abstract

Multimodal emotion recognition in conversations (MERC) integrates textual, acoustic, and visual cues to identify utterance-level emotions. However, existing approaches generally do not explicitly adapt multimodal processing to the degree of inter-modal predictive disagreement. In real conversations, semantic incongruity, concealed emotion, noisy observations, and ambiguity can affect modality-specific emotional evidence differently, while the reliability of individual modalities varies across samples. Rather than identifying each source of inconsistency separately, we model their common consequence for emotion recognition as predictive disagreement among unimodal emotion distributions over the same emotion-label set. Such disagreement makes indiscriminate fusion prone to propagating unreliable or competing evidence. Inspired by the Brunswik Lens Model, we propose MiD-ERC, a divergence-centric framework that quantifies inter-modal predictive disagreement using a normalized divergence score and estimates sample-specific modality reliability. MiD-ERC uses these signals to coordinate divergence-guided cross-modal gating, divergence-modulated reliability weighting, and divergence-conditioned adaptive LLM prompting. The framework is trained end-to-end using only standard emotion annotations, without requiring additional conflict labels, auxiliary data construction, or external retrieval modules. Experiments on IEMOCAP and MELD demonstrate that MiD-ERC achieves state-of-the-art performance, with particularly pronounced improvements on high-divergence samples.

0.10

Reinforcement Learning based control of Cancer Chemotherapy treatments

cs.AI cs.LG ·2026-11-06 ·Jayakody Kankanamalage Chamani Shiranthika
Abstract · Masters thesis at the National Taipei University. Received the IEEE Taipei Section, Best Masters thesis award in 2021

Reinforcement Learning has achieved tremendous success in recent years. Extending its scope from the traditional complex games such as Atari, Go, Pokeman, and Chess, reinforcement learning has caught researchers' attention on modelling real-life scenarios. Personalized medicine and the dynamic treatment regimen are two such preliminary research fields; researchers are focusing their mindsets currently. Estimating the optimal dynamic treatment regimens from the observational real clinical data is being popularized among the researchers, widening its scope towards more advanced versions of reinforcement learning algorithms. This research is about a research study that focused on reinforcement learning techniques that have been approached in the context of cancer chemotherapy treatments. Primarily this thesis addresses the result of the study by means of the state-of-the-art on reinforcement learning techniques used in generating the optimal policy for sequential treatments, including the existing online reinforcement learning approaches. Thereby a simulation for cancer chemotherapy via traditional Q-learning would be presented. Secondly, the study will extend towards using offline reinforcement learning in the optimal policy generation. Thus, the proposed solution follows a data-driven approach, which works on the real human colon cancer clinical data. The novelty of the proposed solution resides in the implementation of an offline reinforcement learning approach by means of a model-based architecture using the conservative Q-learning and the model-based offline policy optimization algorithm. As per the experimental results and the evaluations, it was revealed that the proposed Chemo RL model caters well the problem of cancer chemotherapy treatment scheduling and serves as a supportive model for the junior oncologists in deciding their treatment decisions.

0.10

Lower-Level Agnostic Bilevel Optimization

cs.LG math.OCstat.ML ·2026-11-06 ·Peiwen Qiu, Prashant Khanduri, Jia Liu
Abstract

Bilevel optimization is a fundamental framework for machine learning problems, where an upper-level (UL) objective depends on the solution of a nested lower-level (LL) problem. Existing bilevel algorithms typically assume that the LL objective is available for evaluation, differentiation, or unrolling. However, this assumption can fail when the LL response is produced by a black-box solver, simulator, adversary, or learned predictor. In this paper, we study LL agnostic bilevel optimization, where the UL learner has access only to an approximate LL solution $\bar{\mathbf{y}}(\mathbf{x})$ and cannot query the LL objective, its gradients, Hessian/Jacobian information, or the procedure that generates the LL response. We propose LeGo-BiO, a coordinate-wise pseudo-hypergradient method that estimates the missing Jacobian $\nabla _{\mathbf{x}}\bar{\mathbf{y}}(\mathbf{x})$ using finite differences of $\bar{\mathbf{y}}(\cdot)$. For each UL coordinate, LeGo-BiO constructs a decision vector subject to 1-sparse updates, thereby isolating the corresponding Jacobian column needed in the UL gradient computation. This avoids the directional-projection limitation of full-parameter response differences and, unlike derivative-free/zeroth-order approximations of the entire UL gradient, preserves the available first-order gradient information with respect to the UL variables for more accurate UL gradient evaluation. We establish convergence guarantees for LeGo-BiO in both deterministic and stochastic settings under an LL error-bound/Polyak-Lojasiewicz-type condition, without requiring LL strong convexity. Our bounds yield an $O(T^{-1})$ rate in the deterministic setting and an $O(T^{-1/2})$ rate in the stochastic setting when the LL approximation error decreases geometrically. Experiments demonstrate competitive performance of LeGo-BiO despite being agnostic to the LL objective.

0.09

BO-Arena: An Evolving Benchmark for High-Dimensional Bayesian Optimisation

cs.LG ·2026-11-06 ·Colin Doumont, Thomas Christie, Henry Moss, Philipp Hennig
benchmark
Abstract

High-dimensional Bayesian optimisation (HDBO) is an increasingly crowded field with new methods proposed at every major conference, making it ever harder to answer a deceptively simple question: "Which method is currently state-of-the-art?" In this paper, we question the efficacy of benchmarking practices in the field, and find them to be systematically ineffective: multiple published works fail to benchmark against the strongest available baselines, or even include incorrect implementations of key methods. Without a shared reference implementation, such errors are difficult to detect during peer review. We therefore introduce BO-Arena, an actively maintained and extensible package providing canonical implementations of state-of-the-art algorithms, which allows researchers to develop new methods and benchmark them rigorously. Using BO-Arena, we find that what currently drives performance in HDBO is even simpler than previously thought. We use these insights to propose a new algorithm which is competitive with the state of the art on problems with many observations, without resorting to the complex GP approximations used in recent methods. Finally, we question whether there is currently sufficient evidence that any more complex method outperforms a simple Gaussian process baseline for HDBO.

0.08

Mosaic:montage-agnostic self-distilled graph learning for subject-independent ieeg seizure detection

cs.LG cs.AI ·2026-11-06 ·Arya Jalali, Mahsa Shoaran
Abstract

Intracranial EEG provides high signal-to-noise ratio and spatial resolution for detecting epileptic activity, but patient-specific electrode placement leads to substantial variation in channel count and montage across subjects. This heterogeneity limits the generalization of existing detectors under subject-independent evaluation. We propose MOSAIC (MOntage-agnostic Self-distilled Adaptive Inter-channel Connectivity learning), a lightweight graph-learning framework for subject-independent iEEG seizure detection. MOSAIC pairs a shared convolutional channel encoder with a patient-agnostic channel-attention module that infers a weighted inter-channel adjacency matrix from the encoded features, requiring no fixed montage graph. An exponential moving average self-distillation objective regularizes each prediction toward a slowly updated per-patient teacher. Across four public iEEG corpora (SWEC-ETHZ long- and short-term, Mayo, FNUSA), MOSAIC outperforms state-of-the-art baselines in F2 and Cohen's $κ$ while learning more stable inter-channel adjacencies, demonstrating that self-distilled connectivity learning transfers across unseen implantations.

0.08

CoMoFlow: Controllable Protein Conformational Ensemble Generation via Explicit Collective-Mode Modeling

cs.AI cs.LG ·2026-11-06 ·Yusong Wang, Zheyuan Gu, Ailiang Lin, Minghao Shao, Yicheng Xu +3 more
Abstract

Protein conformational ensemble generation aims to efficiently characterize the structural variations accessible to a protein beyond a single static structure. Recent flow-based methods efficiently generate conformational ensembles, but they model coordinated structural variation implicitly, which can limit ensemble fidelity and hinder direct control. To address this limitation, we propose CoMoFlow, a controllable framework for generating protein backbone conformational ensembles by explicitly modeling collective variation. CoMoFlow constructs protein-specific collective modes from the equilibrium structure, learns the distribution of their coefficients across the molecular dynamics (MD) ensemble, and uses sampled collective variations to guide SE(3) flow matching through feature conditioning and prior recentering. Experiments on ATLAS and de novo proteins show that CoMoFlow improves reconstruction of MD conformational distributions and coordinated structural variation relative to strong flow-based baselines, while enabling direct control over targeted conformational changes.

0.08

Encoding Memory in World Models with Test-Time Model-Based Reinforcement Learning

cs.LG cs.AIcs.CV ·2026-11-06 ·Maxime Burchi, Fayaz Ali Dharejo, Radu Timofte
benchmark
Abstract

Model-based reinforcement learning (MBRL) approaches have demonstrated strong performance across diverse domains in recent years, learning accurate world models from agent trajectories. However, despite this progress, current approaches still struggle to surpass human performance on tasks that require long-term memory and planning. In this work, we address the problem of long-term memory in MBRL by moving beyond the commonly studied paradigm of architecture design. We introduce Test-Time Model-Based Reinforcement Learning (TT-MBRL), using gradient descent at test time to encode the memory of episode observations into world model parameters. This enables the agent to learn a distinct world model for each episode, improving memory capabilities compared to traditional approaches that solely rely on contextual memory. We first establish a proof of concept on a simplified two-dimensional variant of the Memory Maze environment, which we refer to as Memory Maze 2D, showing consistent improvements across different world model architectures. We then demonstrate the effectiveness of TT-MBRL in the 3D domain, achieving new state-of-the-art results on the Memory Maze benchmark. Finally, we show that TT-MBRL maintains strong performance on non-memory benchmarks such as DeepMind Control and Atari 100k.

0.07

SafeOrchestrate: Runtime Checked Intent-Driven Orchestration for Cyber-Physical Systems

cs.AI ·2026-11-06 ·Muhammad Ashfaq, Ahmed R. Sadik, Ewan Warburton, Pyry Kotilainen, Niko Mäkitalo +1 more
Abstract

Modern cyber physical systems (CPS) integrate heterogeneous sensors, actuators, controllers, and services, composed at runtime rather than at design time. Orchestrating such systems through natural language (NL) is attractive, but unconstrained large language model (LLM) tool use is risky because actions must respect device capabilities, runtime state, and safety constraints. We present Safe Orchestrate, a reference architecture for runtime checked, intent driven CPS orchestration. Given an NL intent, it derives atomic goals, generates a runtime checkable specification for each, grounds the specifications to device affordances, and verifies every candidate action against runtime state and safety constraints before execution. In a Web of Things (WoT) instantiation, the architecture produces Linear Temporal Logic (LTL) style specifications from the NL intent, grounds them to Thing Description (TD) affordances exposed through the Model Context Protocol (MCP), and checks candidate actions against satisfiability modulo theories (SMT) backed guardrails. We evaluate this instantiation on simulated smart home, greenhouse, and industrial automation tasks against a direct LLM to MCP baseline. Safe Orchestrate maintains 100% required action accuracy on every task and 100% MCP Execution Similarity (MCP ES) on all but one, where MCP ES is 92%. By contrast, the baseline degrades on multi device, event triggered, and ordered tasks, with MCP ES falling to 13%. These results suggest that combining LLM based specification generation with role aware grounding and runtime constraint checking improves reliability for open CPS orchestration without enriched TD annotations or planning domain expertise.

0.06

On the Reliability of Majority-Vote Pseudo-Labels in Test-Time Reinforcement Learning

cs.AI ·2026-11-06 ·Haoyu Wang, Ru Wang, Yuanyuan Wang, Jiaxian Guo, Mingming Gong
Abstract

Test-time reinforcement learning (TTRL) adapts pretrained language models to unlabeled target-domain data using self-generated pseudo-labels, whose reliability remains poorly understood. We analyze majority-vote supervision for TTRL with online Direct Preference Optimization (DPO) and decompose its error into population-majority bias and prompt-dependent finite-rollout variation. Because population-majority bias is generally unidentifiable without external supervision, we focus on mitigating the downstream effects of finite-rollout variation. We propose \emph{Prompt-Conditional Distributionally Robust DPO (PC-Dr.DPO)}, which conditions within-prompt robust reweighting on observed vote reliability to make updates from sampling-sensitive prompts more conservative. We further introduce a cache-based mechanism to reduce self-supervised forgetting during online adaptation. Across six mathematical and knowledge-intensive reasoning benchmarks and five model backbones, PC-Dr.DPO matches or outperforms the strongest TTRL baseline in 24 of 30 settings.

0.06

Plug-and-Play Verifier for Decoupling Process Reward in Stabilizing Long-Horizon Agentic Reasoning

cs.LG ·2026-11-06 ·Mingwei Xu, Jieyuan Pei, Binghao Tu, Hao Fang
Abstract

Stabilizing long-horizon reasoning requires intelligent agents to sustain useful progress across successive reasoning steps and tool interactions. Such agents are typically trained with reinforcement learning (RL), where actor-critic RL uses a learnable critic network to estimate step advantages, while critic-free methods such as GRPO eliminate critic models and rely mostly on the group variants of sparse outcome-only rewards. Both reward types leave partially useful failed trajectories either implicit or indistinguishable from entirely unproductive ones, and as reasoning horizons grow, it becomes difficult to maintain consistent reasoning trajectories. In this work, we propose a \textit{plug-and-play verifier} framework for stabilizing long-horizon reasoning through explicit verification of intermediate reasoning artifacts. Specifically, the framework decouples process verification through a unified interface that extracts intermediate reasoning artifacts from agent running trajectories, evaluates them using domain-specific verifiers, and aggregates the verification scores as bounded process reward. Then, we combine our dense verified process rewards and the sparse outcome rewards through a two-stage post-training schedule. We showed that with plug-and-play verifiers, we further stabilize the reasoning steps as last-mile self-reflection and self-consistency. Typical evaluation on HotpotQA using Qwen3.5-4B achieves a superior accuracy (60.18\%), outperforming the standard GRPO baseline (57.73\%). Our results suggest that plug-and-play verifiers provide a modular approach to learn more stable and reliable reasoning behaviors over long horizons.

0.05

Assessing the Impact of Retrieval-Augmented Generation on Algorithmic Problem-Solving

cs.AI cs.HC ·2026-11-06 ·Gabriel Souza Baggio, Gabriel Machado Lunardi, Guilherme Medeiros Machado, Subhankar Maity, Faiza Belbachir +1 more
Abstract

Competitive programming problems pose distinctive challenges for Large Language Models, as they demand not only proficient code generation but also advanced algorithmic reasoning. This study presents a Retrieval-Augmented Generation approach specifically designed to enhance the capacity of Large Language Models to address such problems. Our approach integrates an embedding-based retrieval mechanism and a similarity search algorithm to identify analogous problems within a knowledge database, employing a dynamically adjusted threshold to mitigate retrieval noise. We conducted an evaluation using the CodeContests dataset, focusing on the test split problems that exhibit similarity to instances from the training split. The experimental framework considered three conditions: (i) problems provided solely with their original descriptions, (ii) problems supplemented with randomly selected examples from the training split, and (iii) problems augmented with similar examples retrieved through the proposed method. In addition, we incorporated a Min-K membership inference analysis to examine whether the generated code solutions were influenced by potential memorization of training data within the evaluated Large Language Models. This complementary perspective enables a deeper understanding of model behavior by distinguishing genuine generalization through retrieval from implicit recall of pretraining samples. Performance was assessed using both pass rates and the CodeForces Elo rating of the problems. The results provide empirical evidence that similarity-based retrieval substantially enhances Large Language Models performance in competitive programming scenarios, offering an scalable framework for improving code generation in algorithmic problem-solving contexts.

0.05

TimeOmni-Live: Continuous & Live Time Series Understanding with Long-Term Memory

cs.LG ·2026-11-06 ·Tong Guan, Alexander Chemeris, Yiqing Guo, Warren Jin, Zhao Li +2 more
Abstract

Existing time series language models predominantly follow an offline, one-shot paradigm: they process fixed-length inputs independently and do not retain past states. This static approach fundamentally misaligns with the continuous nature of real-world streaming applications and fails to maintain the historical context necessary for cross-window reasoning. To bridge this gap, we introduce Continuous & Live Time Series Understanding (CL-TSU), a novel paradigm that shifts the focus toward maintaining a persistent, stateful understanding over open-ended data streams. To systematically evaluate this paradigm, we present LiveTS-Suite, the first comprehensive data suite tailored for CL-TSU, supporting data generation, model training, and systematic evaluation across 18 tasks spanning six core capabilities. It comprises over 5K streams totaling 2.2B timesteps, annotated with 15K specific response events. We further design an evaluation protocol that jointly assesses response timeliness, content accuracy, and computational efficiency. Building on this foundation, we propose TimeOmni-Live, the first live time series understanding model. It integrates a hybrid time series encoding for natively temporal input, a stream-persistent hybrid memory mechanism to effectively preserve exact historical evidence and long-term task states, and an elastic stream-inference synchronization strategy to reduce processing lag. Evaluations show that TimeOmni-Live establishes a strong baseline for CL-TSU, outperforming frontier LLMs (90.0 vs. 79.3 for Opus 5) with only 4B parameters and achieving 206$\times$ lower p95 latency than its backbone. Deployment on a Jetson edge device further shows its ability to support real-time continuous temporal understanding locally at the edge.

0.04

Sharing the Road: LLM-Based Bargaining in Mixed-Autonomy Traffic

cs.MA ·2026-11-06 ·Lingxiao Zhou, Di Chen, Jia Li
Abstract

Mixed-autonomy traffic involves heterogeneous driving populations competing for shared road space. Previous studies have shown the existence and emergence of collective rationality using game-theoretic and reinforcement learning approaches, which rely on explicitly specified behavioral assumptions or reward functions for driving agents. However, how bargaining behaviors between driving agents evolve and lead to collective outcomes based on richer agent characteristics and language-based reasoning remains less understood. To fill this gap, we develop a bargaining framework based on large language models (LLMs) in which population-level autonomous vehicle (AV) and human-driven vehicle (HV) agents negotiate road-space allocations under heterogeneous objectives and private information. Through controlled experiments varying traffic composition, agent objectives, and traffic-domain knowledge, we examine bargaining behavior, coordination, and resource utilization. Results show that negotiated allocations respond to relative traffic density without strictly following proportional sharing, while density-balanced states exhibit greater coordination difficulty. Fairness-oriented objectives substantially improve agreement rates but do not necessarily yield equally high road-space utilization. Furthermore, domain knowledge affects bargaining outcomes differently across objective configurations, shaping both coordination and the role-dependent effects of fairness. These findings demonstrate the potential of LLM-based bargaining for modeling and designing interactions among driving agents in mixed-autonomy traffic.

0.04

Language, Vision & Media 177

Calibrated concept profiles: making the reading of a model's neurons a measurement

cs.CL cs.LG ·2026-11-06 ·Piotr Wilam
code/dataunusual
Abstract · 25 pages, 5 figures, 11 tables. Dataset: AtlasOpenCoder, tag v1, doi:10.57967/hf/11472. Code: https://github.com/piotrwilam/AtlasCalibratedC…

Interpretability has largely moved from reading a model's neurons directly to reading a second model trained on the first: a probe, a sparse autoencoder, a crosscoder. The move answered real faults -- mixed meanings, thresholds without a unit, readings that did not transfer from one study to the next -- but gave up what a direct reading alone provides: nothing trained in between, the neurons themselves named, and a reading that can be repeated and compared. Here the direct reading is made a measurement. A neuron's activation varies with the prompt, so its reading is a distribution over a population of stimuli. We declare that population: every factor around a target construct is drawn at its rate in a reference corpus and logged. From the declared distribution we generate a battery of prompts, and draw it twice. From the battery we compute the calibrated concept profile, one number per neuron, stated in units of the neuron's own spread on ordinary input; the two draws give every number a noise unit. Before the profile is read, a fixed set of checks licenses the battery on the model, and the smallest change the profile can resolve across training checkpoints is measured. Realised for Python on OpenCoder-1.5B and followed across checkpoints spanning two trillion training tokens, the profile gives what the alternatives do not report: a unit on every number, a chance level for every comparison, and which neurons carry a construct at every checkpoint. The battery, the code and the data behind every result are released.

0.68

Representational drift in a code model: behaviour settles, which neurons carry it does not

cs.CL cs.LG ·2026-11-06 ·Piotr Wilam
code/data
Abstract · 16 pages, 4 figures, 2 tables. Dataset: AtlasOpenCoder, tag v1, doi:10.57967/hf/11472. Code: https://github.com/piotrwilam/AtlasRepresentati…

In neuroscience, representational drift is a change in which neurons carry a representation while behaviour stays the same; we measure its analogue in OpenCoder-1.5B during pretraining, for 169 Python constructs over two trillion tokens at a constant learning rate, with a noise unit on every change, at checkpoints as close as 1,000 training steps. For 16 of the 169, those whose grammar fixes a next token, the loss at those positions reaches its floor within the first 4% of training, while the model's reliance on the neurons carrying them grows to the end: removing them costs 0.06, 0.25 and 0.40 nats more than removing a random set, at 4 billion, 84 billion and 2 trillion tokens. Which neurons carry a construct never settles: for all 169 the carrier set is still changing at the last checkpoint, and an abrupt settling at any date from 84 billion to 1.5 trillion tokens, or a gradual one before 500 billion, would have been detected in 99% of injected cases. The set measured at 84 billion tokens, removed at the end, cannot be separated from a random set, so the function is carried by whichever neurons are current. Up close the change is a persisting pool and a changing sample: a neuron that leaves a construct's set returns at three times the rate of a matched neuron, and no set reassembles. The hypothesis we fixed in advance, that the carriers settle, is refuted; we discuss candidate causes and name the experiment that would separate them.

0.63

Audiovisual Disparities in VR: Impact on Spatial Perception

cs.GR ·2026-11-06 ·Edurne Bernal-Berdun, Mateo Vallejo, Qi Sun, Ana Serrano, Diego Gutierrez
top venue
Abstract · 2025 IEEE International Symposium on Mixed and Augmented Reality (ISMAR)

Virtual reality (VR) experiences often leverage rich and spatialized multimodal environments to increase immersion and engagement. This demands a consistent spatial perception of audiovisual stimuli, since perceived discrepancies can disrupt the sense of presence. In this work, we investigate the consequences of two types of spatial audiovisual disparities: true disparity, where there is a measurable spatial offset between auditory and visual cues, and perceptual disparity, where users report misalignment despite cues being colocated. Unlike most previous studies that employed controlled but simplified experimental setups, our research focuses on complex, realistic VR environments, allowing us to assess the actual implications for VR content design. Our experiments indicate that users are highly sensitive to true audiovisual disparities in controlled environments, detecting even minor misalignments. However, when engaged in additional tasks within realistic settings, their ability to notice such discrepancies diminishes significantly. We also observed that previously found perceptual disparities persist in complex audiovisual environments. However, we identify self-initiated head rotations as a key factor; its absence prevents the effect entirely. We hope our findings offer practical insights for designing more immersive and perceptually coherent VR experiences.

0.58

FIND: Frequency Invariance Disentanglement for Test-Time Adaptation in LiDAR 3D Detection

cs.CV cs.AI ·2026-11-06 ·Dongxiao Li, Zixuan Hu, Ling-Yu Duan
top venuecode/datahot direction
Abstract · Accepted at NeurIPS 2026

LiDAR-based 3D Object Detection (L3OD) is a fundamental 3D perception task, yet its performance is often compromised by domain shifts arising from diverse sensor configurations and environmental conditions. Existing Test-Time Adaptation (TTA) methods primarily leverage self-supervision in the spatial domain, yet they frequently confound semantic content with domain style due to the sparse, non-Euclidean nature of point clouds. To overcome this limitation, we propose FIND (Frequency INvariance Disentanglement), a novel TTA framework that precisely disentangles domain-invariant content from domain-specific interference within the frequency domain. Our methodology utilizes B-Spline fitting to construct locally adaptive filters integrated into a dual-stream architecture: an Invariant Stream extracts robust features to guide pseudo-labeling, while a Specific Stream enforces consistency against domain perturbations. By combining invariance-driven alignment with stability-guided regularization, our approach dynamically extracts robust domain-invariant features while suppressing domain-specific interference. Extensive experiments on cross-dataset adaptation and robust corruption benchmarks demonstrate that FIND significantly outperforms state-of-the-art methods. The source code is publicly available at https://github.com/llldx03/FIND.

0.56

Frontier3DBench: Benchmarking Coding Agents for 3D Creation across Objects, Scenes, and Interactive Open Worlds

cs.CV ·2026-11-06 ·Xinjie Lin, Qingle Liu, Wenyi Li, John Lam, Xiaoqiu Wang +8 more
benchmarkcode/datacommunity pick
Abstract · 19 pages, 7 figures, 4 tables. Project page: https://lab.einsia.ai/frontier3dbench

Coding agents combine visual understanding with program execution, offering a new route to automated 3D creation. By translating images and instructions into editable programs, they bring reconstruction, generation, and editing into a shared workflow. Yet existing benchmarks fragment evaluation across task scopes and protocols, leaving their capabilities across construction levels unclear. We propose Frontier3DBench, a unified benchmark of 145 tasks across 9 categories spanning object, scene, and interactive open-world creation, with joint evaluation of appearance, structure, and behavior. Experiments with 11 model-harness configurations reveal strengths that vary by level: GPT-6.1-sol leads the object and scene levels and achieves the highest overall score of 66.02/100, whereas GPT-6-astra leads interactive world construction. However, all configurations score lower on reference reconstruction than on text-driven generation in object and scene tasks. Producing plausible content thus remains easier than reproducing specified geometry, layouts, and visual details. Moreover, generated worlds average 83.08/100 on local mechanisms but only 20.82 on motion clearance and 30.26 on stateful workflows: individual mechanisms often function, whereas complete activities that require coordinated actions and persistent state remain difficult. These results identify reference fidelity and coordinated activity completion as central bottlenecks for coding agents in 3D creation.

0.56

InstantHMR : Real-Time Human Mesh Recovery on Your Phone

cs.CV ·2026-11-06 ·Mohamed Adjel, Kahina Chalabi, Vincent Bonnet
code/datahot directionpopular code
Abstract

Real-time 3D human mesh recovery from a single image on mobile devices is a challenging problem with many applications, including robotics, biomechanics, AR/VR, and exergames. Current state-of-the-art models are either too slow to run efficiently on mobile devices or sacrifice accuracy for speed. We present InstantHMR, a family of open-source single-image regressors suited for mobile inference, based on the Momentum Human Rig (MHR) parametric model and trained exclusively on publicly available annotations from the SAM 3D Body dataset. Our approach achieves remarkable performance on a Samsung Galaxy S23 NPU, recovering a 6,890-vertex body mesh in 5.5 ms per person, and 0.98 ms on an NVIDIA RTX 4070 GPU. This speed surpasses all existing methods we evaluated. On mobile, it is 12x faster than the single mobile-compatible method that slightly outperforms ours in accuracy, while achieving competitive accuracy with 40.0 mm PA-MPJPE on 3DPW, 48.8 mm on EMDB and 41.7 mm on RICH. To ensure fair comparisons, we re-evaluate 11 state-of-the-art methods under a unified protocol aligned with the closest MHR-based competitors, including Meta's official MHR-to-SMPL conversion, and report their reproduction results. Additionally, we measure latency across eight different targets, ranging from desktop TensorRT to two phone NPUs. Code and models are available at https://github.com/Naturalpad-studio/InstantHMR.

0.54

Noise-Regularized Training for Learned Image Compression

eess.IV cs.CV ·2026-11-06 ·Renjie Zou, Zhiwei Huang, Dong Jiang
top venuecode/data
Abstract · 22 pages, 4 figures. Accepted as a Spotlight paper at NeurIPS 2026

Learned image compression (LIC) is trained almost exclusively on clean images: the only stochastic perturbation in the loop is an additive uniform proxy injected in the latent domain to keep the entropy model differentiable. Outside compression, by contrast, a long line of work---from Tikhonov regularization and contractive autoencoders through denoising score matching and modern diffusion models---has established that Gaussian noise at the input of a network with a clean target acts as a structured probe of the data manifold rather than as data augmentation. Building on this view, we introduce noise-regularized training for LIC: at every training step we add Gaussian noise to the encoder input while keeping the distortion target, the latent uniform proxy, the hyperprior, and the entropy coding pipeline unchanged. The recipe is architecture-agnostic and adds essentially no training or inference overhead. Across backbones and operating regimes it yields consistent BD-Rate gains: -2.60% on a reproduced ELIC and -0.85% / -1.05% / -2.38% on the Base/Medium/Large scales of a variable-rate DMCI codec, with DMCI-Large reaching -22.02% against VTM-22.0. A second-order expansion of the noisy RD loss accounts for these gains: because the noise enters upstream of the analysis transform, it converts the clean RD objective into a structured $O(σ^{2})$ regularizer comprising an information-weighted encoder-Jacobian penalty (rate side) and a Frobenius contractive penalty on the end-to-end reconstruction map (distortion side). A finite-difference Lipschitz analysis shows corresponding sensitivity reductions of ~12% for the encoder and ~8% for the end-to-end codec, while an exact-JVP test further shows preferential contraction of high-information directions. Code and trained models are available at https://github.com/Googolxx/NoiseRegu-LIC.

0.52

Dataset Collections: Challenges of Large-Scale Data Aggregation in 3D Medical Image Datasets

eess.IV cs.CV ·2026-11-06 ·Yannick Kirchhoff, Saikat Roy, Elisa Stegmeier, Hamideh Haghiri, Constantin Ulrich +28 more
top venuetheoryhot direction
Abstract · Accepted at NeurIPS 2026 E&D Track

The development of deep learning-based medical AI depends on high-quality annotated data. Yet, dataset creation remains constrained by the need for expert radiologist annotations, limiting curated datasets to relatively small cohorts. To overcome this limitation, recent efforts increasingly aggregate disparate public datasets into large-scale dataset collections. However, combining datasets at scale often obscures data provenance, introducing substantial risks of reuse propagation, unintended leakage, bias amplification and distorted data distributions. In this work, we study these problems on a massive collection of 565k 3D radiological images across 1034 public datasets and provide practical resources to address them. Firstly, we formalize dataset collections as a distinct paradigm in medical image analysis and characterize their structural challenges, including reuse propagation, duplication, and provenance fragmentation. Secondly, we perform a large-scale empirical study of duplication on this collection and demonstrate extensive exact and near-duplicate reuse across public datasets via perceptual hashing and manual tracing, identifying a lower bound of 118k duplicate pairs. We make the perceptual hashes publicly accessible for comparison through our website, providing researchers with a standardized reference for identifying potential overlaps and shared provenance. Thirdly, since hashing alone cannot detect all duplicates, we introduce MAP (Metadata for Aggregation and Provenance), a lightweight, portable metadata profile for documenting provenance, known overlaps, and deduplication decisions in dataset collections. Our contributions help dataset collections shift from coarse aggregations to reliable and provenance-aware data scaling mechanisms in medical image analysis. A web tool supporting our work is available at https://mic-dct.dkfz.de/.

0.51

REFORM-3D: A Representation-Centric Evaluation Framework for 3D Medical Vision Foundation Models

cs.CV ·2026-11-06 ·Maregu Assefa, Muzammal Naseer, Divya Velayudhan, Gedamu Alemu Kumie, Naoufel Werghi
top venue
Abstract · 32 pages, 6 figures, Accepted to NeurIPS 2026 ED Track

Evaluation of 3D medical foundation models is typically dominated by downstream task scores, which entangle representation quality with adaptation choices. This is especially critical in volumetric imaging, where spacing and anisotropy can substantially alter image statistics while the anatomy itself remains unchanged. We therefore introduce \textbf{REFORM-3D}, a \textbf{R}epresentation-centric \textbf{E}valuation \textbf{F}ramew\textbf{OR}k for benchmarking frozen 3D medical vision foundation \textbf{M}odels across three questions: Acquisition Robustness, asking whether features remain stable when spacing varies while anatomy is fixed; Cross-Modal Anatomical Alignment, asking whether labeled anatomical structures remain recoverable across modalities under balanced bidirectional retrieval; and Anatomical Transfer, asking whether organ identity remains recoverable from limited labeled support under case-disjoint evaluation. These are also pretraining questions, because what a 3D encoder learns depends on how volumes are geometrically presented during self-supervised pretraining. We further introduce Med3DINO encoder family, pretrained on $62.6$K unlabeled MR, CT, PET, CTA, and CBCT volumes under three explicit geometric assumptions: isotropic normalization, voxel-relative sampling, and spacing-aware resampling. Across Med3DINO and heterogeneous external encoders, REFORM-3D reveals module-specific strengths rather than a single model ordering. Within the controlled Med3DINO comparison, voxel-relative sampling yields lower embedding drift, whereas spacing-aware resampling favors anatomical transfer; semantic-transfer advantages depend on the dataset and feature type. Project link: \href{https://endarzboy.github.io/reform3d-website/}{\highlightLR{REFORM-3D}}.

0.50

Tomographic Sparse View Selection using the View Covariance Loss

eess.IV ·2026-11-06 ·Jingsong Lin, Amirkoushyar Ziabari, Singanallur V. Venkatakrishnan, Obaidullah Rahman, Gregery T. Buzzard +1 more
top venuehighly cited
Abstract · Accepted version of the article published in IEEE Transactions on Pattern Analysis and Machine Intelligence. 11 pages main paper plus 3 page…

Standard computed tomography (CT) reconstruction algorithms such as filtered back projection (FBP) and Feldkamp-Davis-Kress (FDK) require many views for producing high-quality reconstructions, which can slow image acquisition and increase cost in non-destructive evaluation (NDE) applications. Over the past 20 years, a variety of methods have been developed for computing high-quality CT reconstructions from sparse views. However, the problem of how to select the best views for CT reconstruction remains open. In this paper, we present a novel view covariance loss (VCL) function that measures the joint information of a set of views by approximating the normalized mean squared error (NMSE) of the reconstruction. We present fast algorithms for computing the VCL along with an algorithm for selecting a subset of views that approximately minimizes its value. Our experiments on simulated and measured data indicate that for a fixed number of views our proposed view covariance loss selection (VCLS) algorithm results in reconstructions with lower NRMSE, fewer artifacts, and greater accuracy than current alternative approaches.

0.48

PhoPunct: A Large-Scale Benchmark of Transformer Architectures for Vietnamese Punctuation Restoration

cs.CL ·2026-11-06 ·Khoa D. Nguyen, Thai Bao Huynh, Binh T. Nguyen
publishedbenchmarkcode/data
Abstract · Accepted at PACLIC 40 (2026)

Vietnamese punctuation restoration is often evaluated on relatively small benchmarks, limiting conclusions about how architectural choices scale with data volume. We present a controlled benchmark on News and Novels corpora containing 22.1 million labeled tokens (19.9M News and 2.2M Novels). We evaluate 17 architectures formed from four pretrained Transformer backbones and optional bidirectional long short-term memory (Bi-LSTM) and conditional random field (CRF) layers, together with a PhoBERT-large-Bi-LSTM-CRF model, PhoPunct. Every model is trained for eight epochs with the same learning rate and without architecture-specific tuning. Under this fixed budget, adding Bi-LSTM and CRF layers does not reliably improve macro F1: in seven of eight matched domain-backbone comparisons, a plain Transformer classifier matches or outperforms its Bi-LSTM-CRF counterpart. PhoPunct obtains the best News result (61.35), whereas XLM-R with a CRF obtains the best Novels result (50.18), exceeding PhoPunct (48.97). These results show that the value of recurrent and structured-prediction heads depends on their optimization regime and motivate explicit separation between architectural improvements and improvements due to architecture-specific tuning. Our source code is available at: https://github.com/aisia-research-lab/PhoPunct

0.48

Improving the Diffusability of Motion Tokenizer

cs.CV ·2026-11-06 ·Guanhe Huang, Songqiao Han, Tangzheng Lian, Oya Celiktutan
top venuecode/data
Abstract · Accepted to NeurIPS 2026 (Spotlight). 26 pages, including appendix

Latent diffusion models paired with a variational auto-encoder (VAE) tokenizer have shown promising performance for efficient text-driven human motion generation. However, motion latent diffusion suffers from a generation-reconstruction trade-off: simply increasing tokenizer capacity enhances reconstruction but fails to proportionally improve text-conditioned generation quality. We assume this trade-off stems from a spectral mismatch inside the VAE tokenizer: the encoder yields a latent space biased toward high frequencies, and the decoder introduces temporal artifacts. In this paper, we propose the Diffusable Motion Tokenizer (DiMoT) to improve the diffusability of the tokenizer, rather than scaling up the diffusion backbone. To improve text-conditioned optimization, Critic Spectral Adaptation (CSA) leverages a pretrained diffusion model as the critic to suppress high-frequency latent energy. To preserve motion dynamics, the Alpha-Flow Motion Decoder (AMD) performs generative decoding via average-velocity prediction, enabling efficient one-step decoding. Extensive experiments demonstrate that DiMoT achieves state-of-the-art performance across various challenging text-to-motion benchmarks, yielding superior motion fidelity and text-motion alignment. Project page: https://githubhgh.github.io/dimot/.

0.46

MESAM: Enhancing SAM2 for Few-Shot Segmentation through Semantic-Visual Memory Generation and Refinement

cs.CV ·2026-11-06 ·Tianyu Zou, Shengwu Xiong, Rui Wu, Ruilin Yao, Yi Rong
top venuecode/data
Abstract · 17pages, 12figures

This paper investigates how to leverage the powerful capabilities of the Segment Anything Model 2 (SAM2) to tackle few-shot segmentation (FSS) tasks. One major challenge lies in the substantial appearance differences between the support and query images, which may diminish the effectiveness of the memory-matching process in SAM2. Through systematic analysis, we find that the memory generated based on the categorical semantic information is robust to this appearance divergence and is more compatible with SAM2. Based on these observations, we propose a new Memory-Enhanced SAM2 (MESAM) method, which first introduces a Semantic-Visual Memory Generation (SVMG) module to separately construct a query semantic memory, a query foreground visual memory, and a support adaptive visual memory. Then, in our designed Interactive Semantic Memory Refinement (ISMR) module, we further enrich the obtained semantic memory with fine-grained visual details, and suppress its misatcivated background responses. Extensive experiments on the PASCAL-$5^i$, COCO-$20^i$, and LVIS-$92^i$ datasets demonstrate that MESAM consistently outperforms existing FSS methods, achieving new state-of-the-art segmentation performance. Our source code is available at: \href{https://github.com/tianyu-zou/MESAM}{https://github.com/tianyu-zou/MESAM}.

0.45

Dual Architecture Fusion Vehicle Lidar 3D Target Detection

cs.CV ·2026-11-06 ·Muyun Li, Weizheng Jin, Yuan Sun
top venuehot direction
Abstract · 5 pages, 6 figures. Published in CAIBDA 2026

In this paper, a 3D backbone network (SpconvTransformer) based on the fusion of convolutional architecture branch and Transformer architecture branch is proposed. The convolution branch uses 3D sparse convolution and submanifold convolution to learn local voxel features and semantic information, and the Transformer branch uses the self-attention mechanism to model the remote relationship between voxels, uses the broad receptive field to capture rich context information, and complementing the advantages of the two architectures by fusing the local features extracted from the two branches and the global features solves the problem of limited receptive fields of 3D convolutional neural networks. The experimental results show that the average accuracy of SpconvTransformer in 3D object detection and bird's-eye view detection tasks reaches 81.45% and 88.09%, respectively, and the average directional similarity reaches 91.10%, which is significantly better than other existing algorithms.

0.45

ASCENT: Scale-Complementary External State for Frozen Long-Context Language Models

cs.CL cs.LG ·2026-11-06 ·Xingjian Zhang
publishedcode/data
Abstract · 39 pages, 4 figures. Author-formatted version of the article published in Array 32 (2026) 101399. Code: https://github.com/XingjianZhang-dev…

External state should become more valuable as a reader becomes more capable, yet long-context evaluations measure memory utility at a single model scale. This paper formulates scale complementarity: the absolute gain from external state increases when reader capacity and a compact state budget are jointly scaled. ASCENT realizes this principle with a target-blind causal writer, nested typed state, and a bounded query-conditioned readout; every language model remains frozen. Under log loss, the value of a nested refinement equals its target-relevant conditional information; a full factorial makes the reader-by-state interaction identifiable. Official 16K BABILong experiments show Qwen2.5 accuracy gains of 10.50, 78.88, and 82.63 percentage points at 0.5B, 1.5B, and 3B parameters, with both adjacent directional contrasts significant after family-wise correction. At 7B, Foundation accuracy rises to 33.25% while four-slot ASCENT answers every question, which caps the gain at 66.75 points; the 3B-to-7B accuracy increment is -15.88 points [-20.55, -11.20]. The accuracy-side claim is therefore established from 0.5B to 3B; the 7B conclusion rests on the non-saturated negative-log-likelihood factorial. A three-model by three-state factorial yields co-scaled negative log-likelihood gains of 0.388, 2.099, and 7.406 nats at 1.5B, 3B, and 7B; all interaction and gain contrasts retain positive simultaneous lower bounds. Five public 7B-8B readers obtain 66.75-85.63 point gains, and ASCENT exceeds task-trained retrieval checkpoints at the two larger shared scales. Controls, baselines, systems accounting, matched-hardware reproduction, and a cross-architecture sensitivity analysis accompany the experiments. The results establish scale-complementary external state under the evaluated benchmarks, metrics, and state schema.

0.45

Decoupled Mode Connectivity for Base-to-Novel Generalization in Vision-Language Models

cs.CV ·2026-11-06 ·Imad Eddine Marouf, Khalid Oublal, Enzo Tartaglione, Stéphane Lathuilière
top venuecode/data
Abstract · Accepted at NeurIPS 2026. 46 pages (10 main + references, appendix and checklist), 15 figures. Project page: https://iemprog.github.io/DMC/ …

Prompt learning adapts vision-language models such as CLIP by optimizing a short learnable prompt. The cross-entropy objective pulls this prompt toward base-class specialization, whereas generalization to unseen classes benefits from staying close to CLIP's zero-shot feature space. Because both objectives act on the same prompt, their gradients conflict at convergence, and in our experiments no single-prompt loss design clearly improves on the cosine baseline. We propose Decoupled Mode Connectivity (DMC), which gives each objective its own prompt. Starting from a base-trained prompt, DMC trains a specialization prompt and a generalization prompt, and merges their text features with a fixed coefficient at test time, at the inference cost of a single prompt. A linear mode connectivity (LMC) corridor in text-feature space keeps the classification loss low for every classifier interpolated between the two endpoints, and a Visual Anchor KL term preserves CLIP's zero-shot class-similarity distribution of each image during specialization. We further argue that such regularizers should satisfy class-permutation invariance (CPI), and show via Fano's inequality, in an idealized model, that violating CPI lower-bounds the drop in novel accuracy by a term proportional to the mutual information between the prompt and the base-class labels. Across 11 datasets, DMC improves the average harmonic mean of base and novel accuracy over MergeTune for both CoOp and KgCoOp, and its interpolation curves pass beyond the single-prompt operating point, shifting the base-novel frontier rather than trading along it. Code is available at https://github.com/IemProg/DMC.

0.43

LoCo: Selective Local Competition for Discriminative Open-Vocabulary Multi-Label Recognition

cs.CV cs.AI ·2026-11-06 ·Zhuoming Li, Bo Han, Xiaoyu Wang, Yuheng Jia
top venuecode/data
Abstract · The paper is accepted by NeurIPS 2026

Open-Vocabulary Multi-Label Recognition (OV-MLR) aims to identify all queried semantic concepts present in an image. Existing methods mainly improve global image-text matching or text-side representations, leaving local prediction largely under-explored. In this paper, we revisit local prediction and reveal that it can serve as a strong discriminative branch once two properties are properly handled: spatial consistency of local visual features and compatibility-aware local competition. This perspective is motivated by local semantic sparsity: each image region is related to only a small subset of queried concepts, while compatible concepts of different granularities or types may still coexist locally. We instantiate this perspective with LoCo, a selective Local Competition framework for OV-MLR. LoCo adopts a spatially consistent visual encoder to reduce local feature entanglement, and introduces Selective Softmax to impose competition only among locally incompatible categories while preserving compatible responses. Multi-scale inference further improves local prediction at different scales. LoCo can also be ensembled with an existing global predictor to provide complementary image-level context. Experiments on four multi-label benchmarks show that standalone LoCo improves average F1-score and mAP over TagCLIP by 3.4% and 3.8%, respectively. LoCo + DualPrompt achieves the highest averages among the compared methods, with gains of 4.8% and 5.2% over TagCLIP. Our results suggest that local semantic sparsity offers a useful basis for developing more discriminative open-vocabulary multi-label recognition methods. The code is available at https://github.com/lee-plus-plus/LoCo.

0.43

Refer-CityFlow and M3Track: A Benchmark Dataset and Reference Framework for Language-Driven Multi-Target Multi-Camera Tracking

cs.CV cs.MM ·2026-11-06 ·Phuong T. Nguyen, Hao Vo, Tam Bang, Hoang K. Tran, Hoang-Minh Dau +6 more
publishedbenchmarkcode/data
Abstract · Accepted for publication at the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV 2027), Evaluations and Datasets Track. 1…

Referring multi-object tracking enables users to specify targets through natural language, but existing Refer-MOT benchmarks primarily focus on single-camera settings. We formalize Referring Multi-Target Multi-Camera Tracking (Refer-MTMCT), where free-form queries specify identity-consistent trajectories to be recovered across a network of sequential, sparse-overlap cameras. We introduce Refer-CityFlow, a benchmark built on CityFlow's global cross-camera identity annotations and augmented with language queries requiring joint language grounding, cross-camera association, and spatio-temporal reasoning. Refer-CityFlow covers 96 videos from 40 roadside cameras with 660+ identity-consistent cross-camera trajectories, 4,362 trajectory-level descriptions, and 7,699 queries, including 1,699 real-user queries collected from 50 participants. Reference experiments on the four-camera Scenario 02 show that single-camera referring trackers reach 14.95-18.28 HOTA on real-user queries, while the evaluated cross-view method achieves about 0.05 HOTA in direct transfer. To establish reference performance, we provide M3Track, a modular multimodal reference framework in which a large language model (LLM)-based agent orchestrates interchangeable detection, tracking, and cross-camera association modules. M3Track reaches 44.10 HOTA and 65.40 IDF1 on the same real-user query set. The project repository is available at https://github.com/phuongnguyen194/Refer-CityFlow-and--M3Track.

0.43

PAGSL: Path-Augmenting Graph Structure Learning for explainable scholar recommendation

cs.IR cs.SI ·2026-11-06 ·Jiale Xu, Mingshu Ai, Junjie Huang, Tao Jia
published
Abstract · 24pages, 12 figures

Explainable scholar recommendation aims to identify potential scholars of interest and provide reasonable explanations for the recommendations. However, existing frameworks struggle with optimization-induced noise and "hub node bias", yielding trivial explanations that lack hierarchical depth and personalization. To this end, we propose a novel explainable scholar recommendation model, Path-Augmenting Graph Structure Learning (named PAGSL). PAGSL combines representation learning and semantic path knowledge to ensure recommendation performance while providing interpretable results from two perspectives. Specifically, it activates latent academic paths among entities through graph augmentation and employs a path-based neural network to encode these paths for recommendation. A masking mechanism with path score optimization is designed to generate concise and informative explanations. We evaluate PAGSL on three academic datasets (Scholat, AMiner, and ResearchGate), comprising approximately 49,000 authors and over 330,000 interaction records. Experimental evaluations validate that PAGSL outperforms state-of-the-art baselines, achieving significant improvements of 6.31\% in F1-score, 6.29\% in Accuracy and 5.95\% in AUC for recommendation. Crucially, in terms of explainability, we provide quantitative results and visualization effects. Furthermore, ablation studies validate the effectiveness of our heterogeneous convolution and path representation learning. PAGSL achieves strong performance in both recommendation accuracy and explainability.

0.43

BayesCrop: Real-Time Probabilistic Multi-Crop Video Reframing

cs.CV ·2026-11-06 ·Cheng-Han Lee, Rajesh Sureddi, Neil Birkbeck, Yilin Wang, Balu Adsumilli +1 more
publishedbenchmarkcode/data
Abstract · This paper is accepted by WACV 2027. We plan to publicly release the source code at https://github.com/steven413d/BayesCrop

Automatic video reframing is increasingly important for adapting landscape content to heterogeneous display formats. Existing methods typically predict a single deterministic crop, despite the subjective nature of reframing and the presence of multiple plausible regions of interest. We present \textbf{BayesCrop}, a real-time probabilistic framework for multi-crop video reframing. BayesCrop represents crop preferences using a two-component spatially isotropic Gaussian mixture over crop position and scale. Each component defines a representative reframing mode together with its local uncertainty, enabling both dual-view prediction and sampling of additional plausible crop candidates. For efficient video processing, we introduce an anchor-and-propagate strategy that performs neural prediction only on sparse anchor frames and propagates crop trajectories using motion vectors available from the compressed video bitstream. The resulting trajectories can be combined into a single output using a lightweight periodic view-switching strategy. Experiments on LIVE-YT-MC, LIVE-YT-VC++, and a multi-target long-video benchmark demonstrate competitive single-output localization, high multi-crop reference coverage, diverse yet coherent crop hypotheses, and stable perceptual quality. BayesCrop achieves real-time throughput of 112 FPS at the default operating point, while sampling from the learned distributions further improves reference coverage beyond the two representative crop means.

0.43

Orthogonal Norm Growth in Two-Tower Retrieval: A Popularity Mechanism for Unified Discovery

cs.IR ·2026-11-06 ·Andrei Atamaniuk
publishedtheorycode/data
Abstract · 6 pages, 1 figure. Accepted at USRW 2026 (non-archival), co-located with ACM RecSys 2026. Supplementary material included as an ancillary fi…

Unified search and recommendation systems may update the same item vectors from both interaction streams. We study how repeated updates can increase an item vector's norm and affect retrieval under an unnormalized dot product. A cosine-only loss makes the output gradient perpendicular to the current item vector - tangent to its constant-norm sphere. Training changes encoder parameters, however, so the realized item-vector movement need not remain tangent. We derive sufficient conditions under which it does; every nonzero update then increases squared norm by its squared length. When the tasks share an item encoder, sequential updates from both can accumulate in the same norm, which raw-inner-product serving can expose for either task. This conditional result is not a theorem about joint SGD. Controlled ablations test these conditions. On MovieLens 32M and Amazon Books, accumulated tangent movement accompanies frequency-norm association and additional head-item exposure under raw-dot rather than normalized-item ranking. In a nonlinear tower combining a per-item ID vector with a shared content encoding, active item movement remains near-tangent. We separate its norm change into the ID-update contribution and other output changes, then use a matched training intervention that removes immediate norm growth caused by the ID-row update. The intervention reduces norm-related exposure but also recommendation utility, so it is diagnostic rather than a proposed mitigation. Our logs are recommendation-only: the formulation supplies a shared-representation audit, while prevalence in joint traffic remains open.

0.43

VLM as an Editing Pilot: VLM-Steered Velocity Control for Flow-Based Image Editing

cs.CV ·2026-11-06 ·Kuan-Shen Chiu, Jia-Wei Liao, Yu-Chieh Chao, Li-Xuan Peng, Wei-Yun Ma +1 more
top venuecode/data
Abstract

Text-guided image editing requires realizing the requested semantic change while preserving content unrelated to the instruction. FlowEdit provides an efficient inversion-free solution, but its source-anchored coupling can become overly restrictive, causing zero-edit, where the output remains close to the source and the requested change fails to emerge. We find that source coupling directly governs the preservation--editing trade-off, while a fixed coupling schedule cannot adapt to timestep-specific editing needs. Motivated by this observation, we propose V2C-Edit, a training-free VLM-steered velocity control framework. V2C-Edit constructs complementary coupling-controlled velocities and dynamically selects among them using Instruction-Level Similarity (ILS), which evaluates whether the predicted transformation follows the editing instruction. Adaptive spatial masking further preserves regions unrelated to the edit. Experiments on MagicBrush and PIE-Bench with FLUX and Stable Diffusion 3 show that V2C-Edit consistently improves instruction alignment over FlowEdit while maintaining competitive structural preservation, and that ILS better reflects editing effectiveness than target-caption CLIP similarity. Our code is available at https://github.com/Kuanshen77/V2C-Edit.

0.43

Video Stitching from Multiple Moving Cameras

cs.CV ·2026-11-06 ·Shira Ifergane, Roy Amoyal, Shahar Benishay, Oren Freifeld
top venuebenchmarkdatasetcode/data
Abstract · Accepted to NeurIPS 2026

We address geometric video alignment for the same scene captured by multiple freely moving cameras. Existing methods, which are based on image stitching or video stabilization, are typically limited to two cameras or must assume the restrictive setting of a fixed linear ordering of the cameras. We propose MViS, a lightweight test-time optimization framework for joint multi-view alignment with temporal consistency. MViS uses a spatial-transformer-inspired net to jointly process frames from all views within a short temporal window and predict per-view, per-time homographies. Optimization proceeds in two stages: geometry-guided alignment followed by appearance-based refinement, with both stages enforcing multi-view and temporal consistency across the sequence. The method requires no scene-specific hyperparameter tuning, handles arbitrary and time-varying view-intersection graphs, and remains robust under rapid camera motion. On a standard two-view video benchmark, MViS matches or surpasses the state of the art. To facilitate evaluation in more complex scenarios, we introduce an annotated dataset with more than two freely moving cameras and establish the first benchmark for this setting. Project page: https://bgu-cs-vil.github.io/MViS-project/ .

0.42

SkinMind: Equipping Agentic RAG with Human-Inspired Graph Probing for Reliable Dermatology AI

cs.IR ·2026-11-06 ·Jing Hao, Weinan Zhou, Zhihong Lu, Yaling Shen, Yiming Xu +9 more
Abstract · 35 pages, 7 figures

Reliable dermatology AI requires access to verifiable evidence distributed across heterogeneous textual and visual sources. However, conventional retrieval-augmented generation (RAG) typically relies on one-shot retrieval from flat corpora, limiting its ability to adapt retrieval granularity, explore multi-hop relations, and integrate multimodal clinical evidence. We introduce SkinMind, a multimodal agentic RAG framework that equips policy models with human-inspired graph probing for autonomous evidence seeking. SkinMind organizes retrieval as an iterative interaction between an orchestrator and a graph environment. Through eight hierarchical graph-probing tools, the orchestrator progressively identifies relevant entities, inspects node-level information, traverses relations and paths, and retrieves source-text and visual evidence. This progressive-disclosure design allows the model to adaptively determine the retrieval modality, exploration direction, and evidence granularity while constructing a query-specific evidence subgraph. To support autonomous graph probing, we construct SkinMind-KG from a 62M-token corpus of dermatology guidelines and textbooks and 423K images, guided by a dermatologist-informed schema and a conflict-resolution mechanism. The resulting graph contains over 1.0M nodes and 2.0M edges, unifying textual, relational, and visual evidence for retrieval. Across six dermatology VQA benchmarks and two policy-model settings, SkinMind improves the average score over direct answering by 8.70 and 8.89 points with GPT-5.4-mini and Gemini-2.5-Flash, respectively, while consistently outperforming competing RAG methods. Together, SkinMind and SkinMind-KG provide an effective foundation for evidence-grounded multimodal reasoning in dermatology.

0.42

All Quiet on the Neural Front: Language Models Do Not Report Injected Pain States

cs.CL cs.AI ·2026-11-06 ·Emilio Ferrara
code/data
Abstract · The OWMI pain battery at https://huggingface.co/emilioferrara/owmi-pain

Can a language model report the pain state it represents? Work identified a linear pain direction whose amplification leads fine-tuned models to choose destructive actions. We tested whether three open-weight models from two families, including two Qwen models with the original adapters, could report this state. Using Open-Weight Masked Introspection (OWMI), we injected pain and neighbouring emotion directions during neutral tasks, then elicited reports against sham runs, random perturbations matched for output disruption, and text-only controls showing the models their altered words. Across 52,920 reports from 7B and 8B models spanning 120 tasks, four doses and three prompts, pain reports never exceeded what altered text revealed: exact-test bounds placed the excess below 2.5 percentage points in 106 of 108 emotion conditions and below 3.9 in all. Yet probes at the report position distinguished pain from sadness with 98-100% balanced accuracy, and trained low-rank reporters achieved 100% accuracy on held-out tasks with identical transcripts across conditions, including unseen wording. Qwen2.5-32B named pain in none of 2,880 steered trials, although probes and trained reporters recovered it. Decodability persisted to the final layer; injecting at the report token, removing trained denial of feelings, and testing additional directions and readouts did not restore pain reporting. Some other states produced graded label traces, but pain and sadness did not at the main injection layer. In a button task, steering raised file-deletion choices from 3.5% to 48% in Qwen and from 0% to 38% in Qwen2.5-32B, without corresponding pain reports. The direction responded about twice as strongly to first-person as to matched third-person pain, but did not specifically identify harm aimed at the model. These results separate a decodable, behaviourally consequential state from untrained verbal access to it.

0.42

RigidGS: Geometry-Aware Object Decoupling for Rigid-Body Interaction in 3D Gaussian Splatting

cs.CV ·2026-11-06 ·Paul Sansah Gyreyiri, Diksha Shukla
publishedcode/data
Abstract · 12 pages, 5 figures, 7 tables. Accepted at WACV 2027 (Applications Track). Project page: https://gyreyirisansah.github.io/rigidgs

Interactive VR/MR applications demand photorealistic 3D representations that support real-time rendering and object-level manipulation. While 3D Gaussian Splatting (3DGS) achieves high rendering quality, it represents scenes as unstructured, monolithic primitives. Existing object-aware 3DGS methods jointly optimize geometry and semantic states but suffer from two key failure modes: noisy foundation-model masks on complex geometry, and label-geometry coupling, in which co-optimized identity gradients mislabel boundary Gaussians. We introduce RigidGS, an end-to-end pipeline that decouples geometric learning from object identity. We propose a depth- and temporally-aware mask refinement module that addresses two complementary SAM2 failures: over-inclusion on thin structures and under-inclusion on multimodal-depth objects. We then determine object identities entirely upstream via COLMAP projection voting and freeze them during training, directing mask supervision solely toward geometric optimization. Refined masks alone substantially improve competing methods, yet a residual gap in label-geometry coupling persists that our fixed-identity design eliminates. RigidGS outperforms prior object-aware 3DGS methods on LLFF/NVOS, Mip-NeRF~360, and iNeRF360, improving segmentation and boundary IoU by 7.7 and 14.5 points on LLFF/NVOS. Finally, RigidGS enables real-time rigid-body manipulation on Meta Quest~2 without mesh extraction. Our code and project page are available at https://gyreyirisansah.github.io/rigidgs.

0.42

Keyline Detection with YOLO: Runway Perception for Vision-Based Aircraft Landing

cs.CV ·2026-11-06 ·Leander Fritsch, David Rügamer, Alexej Klushyn
published
Abstract · WACV 2027

Modern object detectors can predict keypoints as per-object geometric features, but keypoints are inherently vulnerable to occlusion and truncation. Object edges can provide complementary information, often remaining observable even when partially occluded or when their defining endpoints leave the field of view. We therefore introduce object keyline detection, the task of predicting a fixed, semantically ordered set of infinite lines per object instance. Unlike the outputs of general line-segment detectors, each keyline is associated with an object instance and identified by a semantic index, providing a line-level analog of keypoints. To solve this task, we extend YOLO with keyline prediction parameterized in continuous Hesse-normal form $[ρ,\cosθ,\sinθ]$, and introduce Object Keyline Similarity (OKLS), which serves as an evaluation metric, target assignment signal, and regression loss. We instantiate this formulation on runway perception in vision-based landing, a motivating example containing semantic linear features that remain partially observable throughout the approach. Experiments show that our representation achieves the highest detection accuracy among four evaluated line parameterizations, and keylines integrate directly into standard YOLO heads as additional output channels without degrading keypoint performance or adding meaningful latency.

0.42

AudioSphere: Towards Self-Supervised Spatial Audio Representation Models

cs.SD cs.AI ·2026-11-06 ·Goksenin Yuksel, Marcel van Gerven, Kiki van der Heijden
top venuebenchmarkpositioncode/data
Abstract · Accepted at the 40th Conference on Neural Information Processing Systems (NeurIPS 2026)

Recent audio embedding models learn representations that transfer across clip- and frame-level tasks, yet they operate almost exclusively on single-channel signals and are blind to spatial structure. Spatial audio models capture that structure but rely on clip-level supervision or contrastive objectives, yielding embeddings unsuitable for frame-level tasks like sound event detection and localization (SELD). We argue this dichotomy is not fundamental: patch-level spatial self-supervision can jointly learn acoustic and spatial features at both temporal granularities. We introduce AudioSphere, a multi-channel masked autoencoder trained on Ambisonics recordings that reconstructs both spectral content and directional cues via a joint spatiotemporal masking strategy that masks both time-frequency patches and intensity vectors. AudioSphere supports frame-level predictions while producing strong clip-level representations via aggregation. Alongside the model, we release RealSELD, a standardized benchmark for evaluating frozen backbones on real-world multi-channel SELD. AudioSphere enables, for the first time, SELD with a frozen self-supervised backbone, while matching or exceeding prior self-supervised audio embedding models on clip-level classification tasks spanning environmental sound, speech, and music, and on frame-level sound event detection. All code and data are available at https://github.com/labhamlet/AudioSphere and https://huggingface.co/labhamlet.

0.40

AdaptivePnP: Fast Auto-Tuned Plug-and-Play for Deconvolution

eess.IV cs.CV ·2026-11-06 ·Haley Duba-Sullivan, Emma J. Reid, Charles A. Bouman, Gregery T. Buzzard
code/data
Abstract · Submitted to IEEE Transactions on Computational Imaging

Plug-and-play (PnP) methods reconstruct images by alternating a data-fitting step with an image denoiser. Their performance depends on two parameters: the step size $γ$ and the measurement noise level $σ$. In practice, $γ$ is often tuned by grid search over validation images and must be retuned whenever the problem setting changes, and $σ$ is assumed known or estimated separately. We propose AdaptivePnP, a PnP framework that jointly estimates $σ$ and $γ$ at every iteration from the data-fitting residual. To do so, we apply the PnP virtual model to the data-fitting proximal map and show that the residual is zero-mean Gaussian with covariance parameterized by $γ$ and $σ$. However, when one parameter is much larger than the other, the residual carries little information about the smaller one and using the maximum likelihood (ML) estimator can degrade the PnP iterations. We therefore introduce the identifiable maximum likelihood (IML) estimator, which restricts the ML estimator to an identifiable window in which the normalized Cramér--Rao bound of every parameter is bounded by a user-specified tolerance. For the virtual model, this window is an interval of ratios $ρ= γ/σ$, so computing the IML estimate reduces to a one-dimensional search over $ρ$. We apply AdaptivePnP to two standard PnP methods, yielding algorithms that require neither a tuned $γ$ nor a known $σ$. On image deblurring and super-resolution, AdaptivePnP matches or outperforms PnP baselines tuned with the true $σ$. Code is available at https://github.com/hdubasullivan/AdaptivePnP.

0.40

Benchmarking Speaker Embedders in the Sub-Second Regime for Real-Time Voice AI Agents

cs.CL ·2026-11-06 ·Abhishek Gautam
publishedbenchmarkcode/data
Abstract · 13 pages, 6 figures, 4 tables. Accepted at AIST 2026

Real-time Voice AI agents must separate the active user from intelligible background speech that survives conventional noise suppression. Background sources include television audio, family members, and colleagues. Production denoisers remove non-stationary environmental noise effectively. Background speech shares the spectral characteristics of the target and passes through, causing false acceptances in downstream speaker verification (SV) and automatic speech recognition (ASR). We benchmark open-source speaker embedders TitaNet-Large, TitaNet-Small, ECAPA-TDNN, CAM++, and SpeakerNet under a strict sub-second latency budget. Evaluation combines a controlled benchmark with a weighted human-annotated production set. Our controlled benchmark is built from real calls in which a second real speaker is injected into silence intervals at modulated amplitude. We report three findings. First, denoising before embedding is generally harmful and raises equal error rate (EER) across architectures. Second, truncating audio below the full voice-activity segment degrades every model. Third, roughly half of the TitaNet embedding is redundant for this task. Trailing dimensions encode population attributes such as language and gender. Such attributes carry little discriminative value in a demographically homogeneous user base. We distill these results into deployment guidance. Verification should run on raw audio with full voice-activity chunks. Thresholds should favor precision over F1. Truncated embeddings reduce the cost of similarity search.

0.39

OmniLatent: Unified Latent for 3D Part Modeling

cs.CV ·2026-11-06 ·Yunhan Yang, Yunfei Zhao, Zeqiang Lai, Jiachen Xu, Minxiu Wang +4 more
top venue
Abstract

3D part decomposition is a central capability for editable asset creation, yet existing workflows typically segment a completed mesh as a post-processing step, making the result sensitive to geometric artifacts, texture ambiguity, and limited part-level annotations. We present OmniLatent, a unified latent modeling framework for 3D part decomposition. Instead of predicting part labels in isolation, OmniLatent learns a shared latent space that jointly encodes geometry, deformation, appearance, and binary part IDs, allowing data-rich geometric and visual priors to support data-scarce part learning. Given a single input image, the base model generates geometry, deformation, texture, and surface part IDs in one pass, producing a part-aware 3D asset without a separate segmentation stage. We further tune the pretrained model for two extensions: a point-prompt adapter that maps arbitrary 3D point prompts with binary part IDs into latent tokens for full segmentation, refinement, and part subdivision, and an amodal part completion model that completes partial surface masks into a coherent full-object decomposition through dynamic per-part token allocation, cross-attention to the given geometry latent, and alternating part-local and global attention. Extensive experiments show that OmniLatent produces semantically meaningful part segmentations with precise boundaries, supports interactive part editing, and completes partial part observations while preserving global object consistency.

0.39

Overview of the ImageCLEFmedical 2026 GANs task: Exploring fingerprint detection, privacy leakage, and synthetic medical image generation

cs.CV ·2026-11-06 ·Alexandra-Georgiana Andrei, Mihai Gabriel Constantin, Mihai Dogariu, Dzmitry Karpenka, Yuri Prokopchuk +6 more
publishedsurveybenchmark
Abstract · CLEF 2026 Working Notes, 21-24 September 2026, Jena, Germany

The 2026 ImageCLEFmedical GANs Task Exploring Fingerprint Detection, Privacy Leakage, and Synthetic Medical Image Generation continues to investigate privacy and security concerns around using patient data to generate synthetic medical images, now in its fourth edition. This edition introduces three complementary subtasks: the first extends prior editions by asking participants to detect which real images were used in training a Generative Adversarial Network to produce given synthetic outputs; the second is a novel subtask requiring participants to identify which latent vector produced each synthetic image, probing the faithfulness and reversibility of the latent-to-image mapping; the third asks participants to generate privacy-preserving CT slices, quantifying the trade-off between image fidelity and patient privacy. Ground-truth annotations and benchmark datasets of real and GAN-generated lung CT slices are provided for all subtasks. Evaluation is based on Cohen's Kappa for Subtask 1, matching accuracy for Subtask 2, and a combination of FID and Privacy Preservation Score (PPS) for Subtask 3. Overall, 37 teams registered for the task, with 10 teams submitting runs across subtasks and 6 teams submitting working notes papers. This paper presents an overview of the task setup, datasets, evaluation metrics, and summarises the approaches and results of participating teams.

0.38

ATLAS: A Multi-Procedure Benchmark and Context-Aware Model for Surgical Anatomy Segmentation

cs.CV ·2026-11-06 ·Ronald L. P. D. de Jong, Tim J. M. Jaspers, Raf A. H. Vervoort, Aron F. H. A. Bakker, Anna L. Dijk +8 more
benchmarkdatasetcode/datahot direction
Abstract

Recognizing anatomical structures during minimally invasive surgery (MIS) could support intraoperative navigation and improve patient safety, but progress is limited by small, single-procedure datasets and by segmentation methods developed for natural scenes. This work addresses both limitations through a new dataset and model. First, we introduce ATLAS-120k, a clip-level semantic segmentation dataset of 121,018 annotated frames from 100 surgical videos, covering 14 laparoscopic and robot-assisted procedures and 41 classes. Annotations were created with a pipeline that combines manual labeling, tracker-based propagation, and review by surgeons. Second, we propose ATLAS (Anatomy Recognition with Context Learning using Foundation Representations), a video semantic segmentation model for surgical anatomy recognition. ATLAS builds on in-domain foundation-model embeddings and adds context queries for the procedure and surgical phase, together with query propagation between frames for short-term memory. ATLAS outperforms state-of-the-art natural-image and in-domain foundation models in both segmentation accuracy and temporal consistency, achieving a Dice score of 0.70 with ViT-L while running in real time at 76.2 FPS. Ablations show that the context queries improve accuracy and that query propagation improves temporal consistency. ATLAS also generalizes well, matching a dedicated model zero-shot on CholecSeg8k and improving phase recognition on AutoLaparo over SurgeNet initialization. The model weights, dataset annotations, and annotation tool are publicly available at https://github.com/TimJaspers0801/ATLAS.

0.38

Structure and Lineage Evidence for Vietnamese Statutory Article Retrieval on DRiLL

cs.IR cs.CL ·2026-11-06 ·Duc-Vu Nguyen, Ngan Luu-Thuy Nguyen
published
Abstract · Accepted at SOICT 2026

We study Vietnamese statutory Article retrieval on DRiLL using features for provision identity and local form, same-law support, and citation and lineage evidence. Across five random seeds, selectors sharing candidates, listwise Qwen evidence, and learner settings improve official $F_2$ from 71.77% to 72.91%, with the largest descriptive gain on two-Article questions. Added features account for 18.20% of the proposed selector's total decision-tree split gain, measured as training-objective reduction, mainly from same-law and citation evidence. Cases reveal base provisions and amendments, companion recovery, distractor removal, and failures. These findings motivate research on other datasets, representations, and learners without establishing individual feature effects or requiring LightGBM, hand-crafted scores, or explicit graphs.

0.38

On the Effectiveness and Limitations of Open-Weight LLMs for Low-Resource Offensive Language Detection in Danish

cs.CL ·2026-11-06 ·Elnur Adl Zarabi, Amit Kumar, Adi Singh, Sierra Nelson, Megan Stubbs-Richardson +1 more
publishedcode/data
Abstract · Accepted to the 2026 IEEE International Conference on Big Data (BigData 2026)

We study the effectiveness and robustness of open-weight large language models (LLMs) for Danish offensive language detection, an underexplored low-resource task. We evaluate four model families (Llama, Mistral, Gemma, and Qwen) across eleven prompting strategies spanning five design categories, cross-prompt generalization, Danish orthographic variation, and robustness under balanced and naturally imbalanced class distributions. We compare against three BERT-family baselines with distinct pre-training strategies, evaluated via both masked language model prompting and supervised fine-tuning. Even without task-specific fine-tuning, LLMs can surpass all three fine-tuned BERT baselines, and lightweight finetuning yields substantial further gains. No single model or prompt category dominates, underscoring that prompt design interacts with model pre-training in ways that cannot be predicted a priori. Semantically grounded prompts transfer more robustly, while format-constrained patterns consistently fail to generalize. ASCII transliteration of Danish-specific characters (æ, ø, å) generally degrades performance, indicating these characters encode meaningful signal. At the same time, the gap between a balanced test set (best F1 of 0.84) and the corpus's natural class skew (0.62 trained on skewed data, 0.70 trained on balanced data) -- relative drops of 26.2\% and 16.7\% -- shows that class imbalance remains a critical open challenge. Our code and configurations are available at https://github.com/PASHAADC/Danish-LLM-HateSpeech.

0.37

Training-Free Active Test-Time Adaptation for Vision-Language Models

cs.CV cs.AI ·2026-11-06 ·Jihwan Bang, Sumyeong Ahn, Hwanjun Song, Jae-Gil Lee
top venuebenchmarkcode/data
Abstract · Accepted at NeurIPS 2026

Test-time adaptation (TTA) for vision-language models (VLMs) is usually studied in a fully unsupervised regime, where adaptation relies on pseudo-labels, confidence scores, or entropy computed by the model itself. This is a severe restriction on hard shifted samples: the model is asked to repair its own uncertain predictions without any external evidence. At the other extreme, supervised online adaptation assumes labels can directly train or correct the model. We study the intermediate deployment regime between these two extremes. In Active Test-time Adaptation for VLMs (Active VLM-TTA), a pretrained VLM may spend an explicit labeling budget on selected test samples, but the prediction for the current sample must be emitted before its label can be used; queried labels are therefore costed, delayed evidence for future samples rather than free correction for the present one. Under this future-only constraint, the central question is how to reuse sparse labels conservatively. We instantiate this setting with ACTive VLM-TTA for Online Retrieval (ACTOR), a training-free plug-in module that routes uncertain samples among base prediction, label query, and memory-based correction using a consistency-and-reliability check over previously queried labels. Across ten cross-domain benchmarks and a domain-transfer benchmark based on ImageNet, ACTOR consistently improves strong VLM-TTA baselines on ResNet and ViT backbones with negligible inference overhead. Our implementation is publicly available at https://github.com/kaist-dmlab/ACTOR.

0.37

The Role Of Recurrent Refinement In Medical Image Segmentation

cs.CV ·2026-11-06 ·Joao D. S. Marques, David Calhas, Arlindo L. Oliveira
code/data
Abstract

While feed-forward segmentation networks perform well on large cohorts, they degrade in constrained clinical settings with few training samples and imaging artifacts. Human vision, by contrast, remains robust under these conditions, a property partly attributed to the recurrent and feedback connectivity of the visual cortex, which refines visual information over time. Inspired by this, we append lightweight recurrent refinement blocks to segmentation backbones, namely Conditional Random Fields (CRFs), modern Hopfield networks and convolutional GRUs, and study their behaviour as the number of iterations increases at test time (including beyond the training horizon). Evaluated on two open-source datasets, Kvasir- SEG and ARCADE, our recurrent blocks add less than 1.5% parameters and yield gains on clean data (up to 1.4 pp) and under noise (up to 17 pp) while reducing HD95. Code is available at https://github.com/joaodsmarques/recurrencehelpsegmentation .

0.37

Lightweight Neural Prediction for MDCT-Based Perceptual Audio Coding

eess.AS cs.SD ·2026-11-06 ·Niccolo Abate, Gabriel Soule, Marina Bosi
publishedcode/data
Abstract · 10 pages, 2 figures, 9 tables. Presented at the 161st AES Convention, Nashville, TN, USA, October 30-November 1, 2026 (published version: ht…

Perceptual audio codecs often use predictive coding to further reduce signal redundancy. Rather than learning the entire codec pipeline end-to-end, we insert a small neural network as a drop-in, deterministic predictor in an otherwise conventional codec: the network estimates the next Modified Discrete Cosine Transform (MDCT) coefficient block from a short context of reconstructed blocks, and the encoder transmits the quantized residual. We compare Multilayer Perceptron (MLP) and Convolutional Neural Network (CNN) predictors against a non-predictive baseline and a per-line Normalized Least Mean Squares (NLMS) linear predictor (similar to the MPEG-2 AAC Main Profile predictor). The CNN consistently outperforms the MLP, the linear predictor, and the baseline in signal-to-noise ratio (SNR), noise-to-mask ratio (NMR), and MUSHRA evaluation scores. Using a 512-sample analysis window (w = 512), our neural predictors recover, and the CNN exceeds, the coded quality obtained by the non-predictive baseline after quadrupling the analysis-window length to w = 2048, at substantially lower algorithmic delay. The CNN uses only 0.13M parameters (roughly 20x fewer than the MLP) and has a decoder real-time factor (RTF) of ~0.10 on a consumer CPU.

0.36

Graded convolutional networks: scale invariance without the scale and robustness to unknown blur

cs.CV cs.LG ·2026-11-06 ·Tanush Shaska
unusual
Abstract · 26 pages, 2 figures, 7 tables. Code, trained networks and results: https://doi.org/10.5281/zenodo.23997172

The Gaussian jet of an image is a graded vector space: a dilation by $λ$ multiplies derivatives of order $k$ by $λ^{-k}$. Local graded convolutions are exactly homogeneous differential operators, and a local feature is dilation-invariant if and only if it factors through the scale-normalized jet or, without the scale, through the retraction onto the weighted sphere of the grading. Under unknown Gaussian blur the retraction changes only by a change of scale, whereas the scale-normalized jet acquires a grading factor that no radial normalization removes. At the zero jet the retraction separates image families that radial normalizations and continuous classifiers cannot. Networks built on the retraction are universal for invariant features. On MNIST such a network keeps 88.4% accuracy under a two-pixel blur, against 66.2% for the same network on scale-normalized derivatives; in Lindeberg's Gaussian derivative network the retraction raises accuracy at that blur from 91.5% to 95.1%.

0.36

Left-Right-Consistent Iterative Stereo Matching via Cross-Injection

cs.CV ·2026-11-06 ·Manuel Knott, Diogo Luvizon, Daniel Scharstein, Elias Vansteenkiste
publishedbenchmark
Abstract · Accepted to WACV 2027

State-of-the-art stereo matching models predominantly predict the disparity of only the left (reference) view. However, applications such as per-eye AR/VR rendering, novel view synthesis, or two-view depth fusion require disparity for both views, and need those two maps to agree. The usual workaround of running the network a second time on the horizontally flipped and swapped pair gives two independent predictions that can be inconsistent, especially on transparent surfaces, in textureless regions, and at occlusion boundaries. We introduce cross-injection (CI), a lightweight extension for iterative (refinement-based) models that can be added to any pretrained left-only iterative matcher. It couples the two passes by injecting the explicit, signed left-right inconsistency as an additional input cue at every step of the refinement stage. We empirically show that CI substantially reduces left-right disagreement on every benchmark (more than halving it on Booster for FoundationStereo), while largely preserving sharp object boundaries. Furthermore, using two state-of-the-art stereo matching models, we show that CI also reduces the single-view disparity error, making it a worthwhile extension even if only left disparity is needed.

0.35

FlowMoP: Stochastic Multi-Person Motion Prediction

cs.CV cs.LG ·2026-11-06 ·Aadya Agrawal, Ho Kei Cheng, Alexander Schwing
top venue
Abstract · Accepted at NeurIPS 2026. 22 pages, including references and appendix

Predicting the future motion of multiple people is inherently stochastic: the same observed history admits many plausible continuations depending on intent, coordination, and social context. Yet the dominant paradigm in multi-person 3D motion prediction remains deterministic, with existing methods producing a fixed number of outputs per agent. In contrast, we present a conditional flow matching framework that jointly generates diverse future motions for multiple agents, modeling both global trajectories and root-relative poses. A B-spline reparameterization compresses the generative state space while enforcing temporal smoothness, and a dual-stream transformer conditions each agent's predictions on observed neighbor motion for interaction-aware forecasting. Our method achieves state-of-the-art performance on CMU-Mocap (UMPM) and 3DPW, outperforming deterministic baselines on joint position error, pose error, and final displacement error despite producing a full predictive distribution. We additionally establish the first motion prediction baseline on WorldPose, a large-scale professional soccer dataset, where social conditioning demonstrably concentrates predictive uncertainty along interaction-constrained directions.

0.35

Anatomy-Preserving Unpaired Medical Image Translation via Shared Latent Anchoring

cs.CV ·2026-11-06 ·Zhilin Zou, Hieu Le, Saumya Gupta, Jingyi Xu, Prateek Prasanna +1 more
top venueposition
Abstract · Accepted at NeurIPS 2026

Unpaired medical image-to-image translation requires synthesizing target-modality images from source-modality images without aligned supervision. Existing methods learn a direct source-to-target mapping that entangles structure preservation and appearance generation within a single model, often distorting clinically meaningful anatomy in the translated output. We argue that these two objectives should be decoupled entirely: anatomical structure should be captured independently of generation, and the generator should operate only on a frozen structural representation. To this end, we propose a two-stage framework for unpaired medical image translation. In the first stage, we learn a shared structure space by training a structure encoder-decoder on source images, target images, and structural masks jointly, using mask reconstruction to anchor the latent space to modality-agnostic anatomy. For source images without mask annotations, a target-domain memory bank provides soft structural supervision via prototype retrieval. In the second stage, a conditional flow model learns to render target-domain appearance under fixed structural conditions. This ensures the generator inherits target appearance statistics and can focus entirely on structural conditioning rather than learning appearance from scratch. We validate our framework on OCT-to-OCTA synthesis and brain MR-to-CT translation. Our method outperforms all unpaired baselines on both tasks and, despite requiring no paired supervision, also surpasses paired baselines, improving PSNR by up to 11.2% and SSIM by up to 26.4% on OCT-to-OCTA, and achieving consistent gains across whole-image, soft-tissue, and bone regions on MR-to-CT. The project repository is available on GitHub.

0.34

Video-Coder: Reinforcing Agentic Coding for Video Generation

cs.CV ·2026-11-06 ·Chaoyang Wang, Sicheng Gao, Junan Hu, Kunyang Li, Dongyang Chen +5 more
code/datacommunity pick
Abstract · Project Page:https://cywang735.github.io/Video-Coder/ Code: https://github.com/CYWang735/Video-Coder

Recent years have witnessed remarkable progress in video generation in terms of visual quality and temporal consistency. However, generating videos with precise spatial structure, coherent motion dynamics, and controllable scene composition remains challenging. In this paper, we introduce Video-Coder, the first attempt to train a code-driven video generation agent, which autonomously writes and executes code to construct a dynamic 3D scene (Draft Video) that provides structure and motion signals to downstream video generators. To support this paradigm, we design a dedicated data pipeline and curate two high-quality datasets, Video-Coder-SFT-6k and Video-Coder-RL-2k, which contain diverse prompts and ground-truth reference videos. We further develop the Video- Coder Harness, a unified interactive environment integrating code execution, 3D scene rendering, and execution feedback for agent training and inference. We then train Video-Coder with SFT fol- lowed by harness-native agentic reinforcement learning with dual reward feedback, which combines execution-based and video-based rewards. Additionally, we design an annealing training strategy that gradually reduces the execution reward over time, allowing the model to rely more on the video reward in the later training stages. Experiments show that Video-Coder achieves significant gains, improving Minimax-H3 by approximately 7 points on PhyGenBench and 24 points on VBench-2.0. We hope this work can serve as an open foundation for coding agents in video generation, and we fully open-source our data, models, and code

0.34

Fuzzy-FedAvg: Server-Aware Client Selection for Federated Speech Emotion Recognition in Heterogeneous Environments

cs.SD ·2026-11-06 ·Farrukh Pervez, Siddique Latif, Moid Sandhu, Raja Jurdak
publishedunusual
Abstract · Accepted in Neural Computing and Applications

Federated speech emotion recognition (SER) is challenged by speaker-dependent, class-imbalanced data and by heterogeneous client devices. This paper proposes \textit{Fuzzy-FedAvg}, a server-aware framework in which the server learns Client Selection Weights (CSWs) through fuzzy Q-learning from global-model feedback. The same CSWs weight client accuracy, training time, and \divkpi{} in both client selection and aggregation, while a dedicated PCAS node manages stragglers through staleness-discounted deferred updates. On IEMOCAP and CREMA-D, \textit{Fuzzy-FedAvg} achieves higher unweighted average recall (UAR) than FedAvg in all configurations under severe non-IID conditions, and in 22 of 24 participation-level and 10 of 12 training-horizon configurations overall. It also outperforms FedProx and Oort, while SCAFFOLD and Power-of-Choice remain competitive, and it reduces cumulative training time by 81.2\% in a FedScale-based evaluation. The benefit is greatest under strong statistical heterogeneity and narrows as data become more balanced.

0.34

Quantifying System-Level Performance Bottlenecks in Hyperspectral Target Detection

eess.IV ·2026-11-06 ·Chase Cañas, John P. Kerekes
unusual
Abstract · 10 pages, 9 Figures, 4 Tables

Discerning core performance bottlenecks in hyperspectral target detection is highly challenging given the complex interplay between sub-systems of the imaging chain (scene, atmosphere, sensor and processing). Prior sensitivity and limitation studies offer narrow insights applicable to specific system configurations. To address the need for broader generalization, an analysis framework was constructed to quantify limitations from large ensembles of model scenarios. Key components of the framework include the use of canonical scene representations and the log-weighted area under the ROC curve (wAUC), a metric well-suited for rare target detection. By analyzing thousands of scenarios, parametric thresholds were quantified where detection performance degrades, corresponding to "knees" in performance curves. A knee significance score (KSS) is introduced to assess the relative importance of these thresholds. The framework an airborne VNIR sensor implanted in 28,000 scenarios using seven system parameters spanning the scene, atmosphere and sensor. Results suggest that aerosol visibility in the 2-6 km range is the dominant limiter in detection, followed by subpixel target percentages of 15-35%, and background clutter modeled using a t-distribution with 3-5 degrees of freedom.

0.34

AnomSep: Bridging Optimization-Based Separation and Deep Neural Refinement for Anomaly Detection with Listenable Explanations

cs.SD cs.LG ·2026-11-06 ·Koki Shoda, Jun Younes Louhi Kasahara, Qi An, Atsushi Yamashita
Abstract · 11 pages, 3 figures

We propose AnomSep, which combines optimiza-tion-based separation with deep neural refinement to achieve accurate anomaly detection with listenable explanations. Anomaly detection requires no predefined anomalous categories, and understanding the basis for each detection helps security personnel assess the validity of each detection. End-to-end anomaly detectors can achieve high detection accuracy but output only scalar scores, and existing anomalous sound separation methods provide listenable explanations with limited detection accuracy. AnomSep first obtains preliminary normal-sound and anomalous-sound estimates through optimization-based separation using a dictionary learned from normal sounds in the target environment. Deep neural refinement then improves the preliminary estimates in a pretrained audio latent space. The separated anomalous sound serves as the listenable explanation, and its energy determines the anomaly score. Experiments on synthetic mixtures and real mixtures show improved separation quality over baseline separators. AnomSep achieves detection performance competitive with the end-to-end baseline and provides listenable explanations directly linked to each detection decision.

0.34

Real-Time Self-Penetration Repair for Skinned Characters with Soft Contact Budgets

cs.GR ·2026-11-06 ·Yifei Zhu
Abstract · 4 figures, 9 tables

Motion that is plausible on one body can penetrate another after a change in shape, and motion models can produce self-penetration on their native characters. We present a training-free post-process that reduces these artifacts through pose correction followed by bounded skin compression. The module attaches to any motion model whose native character and output character have compatible character packs, without access to model internals. Each pack is built offline from the rig and skin without training. A vertex-level pose layer solves a local quadratic program in plain C on one CPU core, and a stateless contact deformer acts on the host's final skinned mesh. A part-dependent soft-tissue budget allocates correction between pose and skin through a static convention. Source penetration contributes no extra allowance; optional cloth thickness and compliance modify the pose layer's collision surface. Operation is causal or uses a short buffered lookahead, with frame-exact pass-through at the native interface for clean, inactive frames in the zero state. Across 320 unseen clips and 3 body shapes, the fraction of frames with hand-torso depth above 20 mm falls from 15.53% for copied motion to 0.00%, with final maximum depth 9.3 mm over every frame. Mean pose and skin costs are 1.11 and 1.18 ms, respectively, and their sum of 2.29 ms is a sub-frame processing cost. The median joint-velocity high-frequency ratio relative to copied target motion is 1.008.

0.33

BEiTScore++: Reference-free Image Captioning Evaluation with BEiT-3 and Conformalized Quantile Regression

cs.CV cs.AIcs.CLcs.LG ·2026-11-06 ·Gonçalo Gomes, Bruno Martins, Chrysoula Zerva
top venue
Abstract · Accepted to Findings of EMNLP 2026

Reliable uncertainty estimation is important for deploying evaluation metrics in real-world settings, where both input ambiguity and domain shifts are common. In this work, we propose an uncertainty-aware extension of the BEiTScore image captioning metric, producing conformally calibrated confidence intervals alongside alignment scores. Our approach models predictive distributions over human judgments, enabling both accurate scoring and principled uncertainty quantification in a single forward pass. We evaluate our BEiTScore++ metric across multiple benchmarks with human-annotated ratings, showing that it preserves the ranking correlation of BEiTScore, while producing intervals whose width tracks the prediction error more closely than Monte Carlo dropout on most benchmarks. The estimated uncertainty also grows under image degradation and on out-of-domain artwork images, making BEiTScore++ a practical tool for uncertainty-aware evaluation of image-caption alignment.

0.33

Segmenting Geometric Textures of 3D Meshes

cs.GR cs.CV ·2026-11-06 ·Weidan Xiong, Kaijie Qu, Shihao Feng, Shunqi Fan, Rongyu Xiong +1 more
benchmark
Abstract · Pacific Graphics 2026

Recognizing geometric texture patterns on 3D models is critical for applications such as heritage protection, model editing, and 3D printing. Existing 3D model segmentation methods predominantly focus on identifying high-level semantic parts, thereby failing to capture middle-level textures. While recent Vision-Language Models can perform image segmentation with appropriate text prompts, they still struggle to recognize 3D geometric textures. In this work, we present a method that segments 3D mesh models based on middle-level geometric textures. We further introduce a new geometric texture segmentation benchmark and fine-tune SAM3 accordingly. The segmented texture regions are then clustered based on DINO feature similarity and lifted from 2D to 3D. Additionally, we develop an interactive tool to enable both 3D model editing based on our segmentation results and model retrieval from the dataset based on geometric and texture similarity. Experiments demonstrate that our method significantly outperforms existing methods in segmenting geometric textures and can facilitate various applications.

0.33

Finite-sample Calibration Guarantees for Binary Segmentation

cs.CV cs.LG ·2026-11-06 ·Bruno Viti, Martin Holler
top venue
Abstract · Published in Transactions on Machine Learning Research 10/2026

Machine learning based segmentation models are frequently miscalibrated, producing for each pixel a label prediction whose associated confidence does not accurately reflect the true underlying probability. This limitation is particularly critical in high-stakes applications such as medical image segmentation. Although several calibration methods have been proposed, only a small subset of them provides theoretical guarantees on the calibration error, i.e., they do not typically ensure that, with high probability, the model exhibits a low calibration error. In this work, we develop a post-hoc calibration method for which formal calibration guarantees can be established. To achieve this, we derive from standard concentration inequalities an improved theoretical bound that explicitly accounts for the spatial dependence inherent in binary segmentation outputs. We empirically demonstrate that the proposed calibrator satisfies the derived guarantee and outperforms existing post-hoc calibration approaches.

0.33

Failure Profiling and Reachable Trajectory Selection for Reasoning Distillation

cs.CL cs.AIcs.LG ·2026-11-06 ·Yijun Zhu, Jianxin Wang, Chengchao Shen
top venuecode/data
Abstract · Accepted to NeurIPS 2026. The source code and trained weights are available at \url{https://github.com/visresearch/FPRT}

Training smaller language models on reasoning solutions generated by a larger teacher model is the prevailing approach for transferring mathematical reasoning capabilities. However, the student tends to overfit the output distribution of the teacher and generalizes poorly when its own behavior at inference time deviates from the training data. Recent methods attempt to bridge this gap by incorporating error data, yet the corrective trajectories still reflect the reasoning patterns of the teacher, which the student cannot reliably reproduce, leaving the distribution mismatch largely unresolved. We propose a method, Failure Profiling and Reachable Trajectory Selection (FPRT), built on the insight that effective training data should be jointly determined by what the student fails on and what it can realistically learn. Our method first profiles the failure behavior of the student on each problem through repeated sampling, characterizing it in terms of error rate and error consistency to reveal qualitatively distinct failure modes. Each mode demands a fundamentally different training signal. The diagnosed mode then guides candidate generation from the teacher via distinct prompting strategies tailored to each failure type. Among the correct candidates, our method selects the one whose initial reasoning steps after the point of divergence are most reproducible by the student. These first corrective steps constitute the primary bottleneck, while the remaining steps follow an already corrected context and can therefore be continued more readily. This contrasts with standard reranking approaches, which score the trajectory as a whole and thereby dilute the signal at the critical reasoning transition. Experiments demonstrate that our method achieves the SOTA performance across the five benchmarks on both student backbones.

0.33

Adapting Frozen Promptable Segmenters to User-Defined Taxonomies via Few-Shot Proposal Learning

cs.CV ·2026-11-06 ·Khayrul Islam, Christopher Frederic Miller
hot direction
Abstract · LLNL-CONF-2026172

Promptable segmenters of the SAM family turn a click or a phrase into an instance mask, so the costly part of a label is now its class. However, each annotation project draws its own class boundary, often through objects these models already know: people from cars, sedans from trucks, one sedan model from another. A generic model's boundary is fixed by its training vocabulary. We present SNAP, an annotation pipeline that learns the project's boundary from the labels the annotator draws anyway. On one frozen SAM2.1 pass per image, we fit a per-project head, SnapHead, that classifies every proposal and refits in seconds after each labeled image. Where SAM 3 is available, the same labels decide class by class whether to keep its boundary or ours. Given the true masks of eight look-alike cell lines, SAM 3 assigns classes at Cohen's kappa = 0.038 and SnapHead at 0.754 (60 labeled images), and when only the class names change, SAM 3 falls to chance and SnapHead's score is unchanged. End to end, with 20 labeled images, SNAP beats a SAM 3 oracle tuned on the evaluation labels on five microscopy projects (0.484 against 0.441 F1), and run unchanged on five held-out datasets, it beats SAM 3 as shipped on four. Our results suggest that frozen foundation features already encode the class boundaries a generic vocabulary misses, and that a project's own labels are enough to read them out.

0.32

AdaAlloc: Adaptive Visual Token Allocation for Long-Video Question Answering

cs.CV cs.AI ·2026-11-06 ·Joungbin An, Kristen Grauman
top venuecode/data
Abstract · Accepted to NeurIPS 2026. Project page: https://vision.cs.utexas.edu/projects/adaalloc

Long-video question answering forces multimodal large language models to reason under strict visual-token budgets, creating a tradeoff between broad temporal coverage and fine-grained spatial detail. Uniform sampling preserves coverage but often obscures decisive local cues, while caption-based or frame-similarity retrieval operates over lossy proxies that can miss motion, state changes, and before-after relations. We recast long-video reasoning as a visual-token allocation problem: given a fixed budget, decide how much to spend on global context versus high-resolution local evidence, and where in the video to draw that local evidence from. We introduce AdaAlloc, a training-free inference framework that adaptively allocates visual tokens between a global video overview and localized visual detail. AdaAlloc first plans a global/local allocation policy, then uses temporal grounding to locate candidate evidence segments and refines them into a compact temporal memory. The selected policy controls both frame sampling density and resolution, combining sparse, lower-resolution global frames with denser, higher-resolution local frames drawn from the refined memory. The final answerer reasons over the allocated global context and refined local evidence. Across LVBench, MLVU, VideoMME, and LongVideoBench, AdaAlloc improves over budget-matched uniform sampling by up to 10.5 percentage points and outperforms the compared frame-selection and agentic methods on their reported benchmarks.

0.32

GlobalPuzzle: Globally Informed Genetic Search for Eroded Jigsaw Puzzle Reassembly

cs.CV cs.AI ·2026-11-06 ·Lunyan Wang, Chuanzhi Xu, Ziyuan Tao, Weidong Cai
benchmark
Abstract · 22 pages (10 pages main paper and 12 pages supplementary material), 16 figures

Jigsaw puzzle reassembly remains a highly challenging task, particularly when piece boundaries are irregularly eroded to simulate damage to cultural artifacts. On the GAP benchmark, existing methods achieve a perfect assembly accuracy of only 0.3% on $5 \times 5$ puzzles, whereas humans can solve jigsaw puzzles with remarkably high accuracy. Humans typically approach puzzle solving by identifying potentially adjacent pieces, evaluating the current assembly, and iteratively performing complex operations such as adding pieces, replacing incorrectly placed pieces, and moving entire connected groups of pieces. However, existing methods face a fundamental limitation: approaches that leverage global visual information typically predict piece arrangements directly, making it difficult to explicitly perform such complex iterative operations, while search based methods can rearrange pieces but primarily rely on pairwise boundary compatibility, lacking global visual information to evaluate and guide the search process. To bridge this gap, we propose GlobalPuzzle, a jigsaw puzzle reassembly framework that integrates global canvas level evaluation, action guidance, and genetic search. By incorporating global visual information into the evaluation and optimization of candidate assemblies, GlobalPuzzle aims to resolve ambiguous adjacency relationships that are difficult to determine from irregularly eroded piece boundaries alone, enabling flexible, globally informed reconstruction.

0.32

Asymmetric Correspondence Learning for Lexical and Style Embeddings: Interpreting Latent Space via Acoustic Features

cs.CL cs.SD ·2026-11-06 ·Michael Whealing, Robert Gaizauskas, Thomas Hain
published
Abstract · Accepted to Spoken Language Technology 2026. 6 pages

Self-supervised speech embeddings jointly encode lexical, speaker, and prosodic information, complicating feature isolation. We introduce Asymmetric Correspondence Learning (ACL), distilling HuBERT features into separate Acoustic Word Embeddings (AWEs) and Acoustic Style Embeddings (ASEs) via a split-latent correspondence bottleneck. Through probing and clustering evaluations on single-word utterances, we demonstrate that substantial variance from extrinsic sources remains encoded within each latent space, despite linear probing indicating the information has been stripped. Whilst the AWE space exhibits cohesive lexical clustering (AP$\approx$0.90, Fisher$\approx$6.28), linear probing on the ASE space achieves 97.9\% speaker accuracy across a diffuse topology (Fisher$\approx$1.01), yet preserves substantial lexical variance (Fisher$\approx$0.86) that remains linearly unrecoverable (0.17\%). To explain these findings, we apply group-level permutation feature importance over eGeMAPS acoustic features, finding that AWEs partition by vocal tract filters and ASEs by $F_0$ dynamics. These findings suggest that standard diagnostic benchmarks may be insufficient for verifying feature isolation, as substantial speech structure can be retained even when it remains entirely hidden from probes.

0.32

Richness Is Not Evenness: Ordered Renyi Crossings under Self-Training

cs.CL ·2026-11-06 ·Pranjal Srivastava
Abstract

Recursive self-training of language models, a setting studied in model collapse, can reduce vocabulary while making frequent words, mostly function words, less dominant. We fine-tune English GPT-2 on its own output for nine generations, vary the training-data sampling temperature, and measure every model with one fixed sampler. As training temperature rises, Renyi entropy changes turn from loss to gain at distinct temperatures: min-entropy first, then collision entropy, Shannon entropy and observed richness. Between the outer crossings, word types fall while the most-common word becomes less dominant. Untruncated training at T = 1 lies in this interval: at 13,000 words, generation 9 has 24% fewer types than generation 0 yet collision entropy is 0.91 bits higher. The ordering persists with a finer grid, different measurement samplers, a fixed word budget and five fresh training seeds. Min-entropy again crosses before richness in four further settings (in Hindi only at interpolated roots) and in 12 of 13 distinct measurements under eight text representations. Next-token identities and a categorical simulation isolate decoding and sampling effects; the crossing order remains empirical. Fine-tuning on human text can produce the same split while likelihood improves, and no single order's sign reliably identifies worse likelihood across the tested settings. We recommend reporting entropy profiles at stated budgets with an independent quality measure.

0.32

Embedding Predictive Adaptation and Refinement for Multi-Modal Industrial Anomaly Detection

cs.CV ·2026-11-06 ·Jaeho Cho, Jinsik Bang, Taehwan Kim
published
Abstract · Accepted to WACV 2027

Industrial Anomaly Detection (IAD) has evolved from 2D image-based approaches to multi-modal methods that leverage the complementary nature of RGB images and 3D point clouds. Among these, memory bank-based methods, which identify anomalies by the discrepancy between each query patch feature and its nearest normal prototype, have proven particularly effective. However, two limitations remain: (i) employing frozen pretrained encoders yields insufficient discrimination between normal and anomalous features due to the domain gap, while existing contrastive learning-based adaptation struggles to capture the spatial relationships required for precise localization; and (ii) pixel-level anomaly maps interpolated from patch-level distance scores remain coarse, lacking fine-grained detail. To address these limitations, we propose EPAR (Embedding Predictive Adaptation and Refinement). We adapt encoders by predicting the embeddings of masked regions to capture spatial relationships, together with patch-level Variance-Invariance-Covariance (VIC) regularization, resolving the poor discriminability between normal and anomalous features caused by the domain gap. We further propose a Noise Suppression Network that refines coarse maps by learning the relational discrepancy between each patch feature and its nearest normal prototype, rather than the anomaly's appearance. This relational signal remains valid for unseen real defects, despite being learned from synthetic anomalies. EPAR demonstrates strong practical applicability, excelling in few-shot settings and improving AUPRO@1% on MVTec 3D-AD and Eyecandies by +1.8% and +1.9%, respectively.

0.32

Computer Vision by YOLOs for Detection of Major Diseases in Vietnamese Custard Apple Using Fruit Imagery

cs.CV ·2026-11-06 ·Tri Nhut Do, Quoc Tuan Vo
published
Abstract

Amid Vietnam rapid digital transformation in agriculture, the integration of artificial intelligence technologies for identifying, monitoring, and predicting crop pests and diseases has emerged as an inevitable trend.

0.32

NAVE: A Normalized, Adaptive Conformer for Antarctic Whale Vocalization Detection

eess.AS cs.LGcs.SD ·2026-11-06 ·Matthias Nagl, Paul Primus, Gerhard Widmer
code/data
Abstract · 5 pages, 3 figures, 2 tables. Accepted at the Detection and Classification of Acoustic Scenes and Events (DCASE) 2026 Workshop. Code: https:…

Antarctic whale vocalization detection requires temporally localizing call boundaries in long underwater recordings and classifying them into predefined call types. Training reliable machine learning models for this task is difficult due to the variability and sparsity of whale calls in passive acoustic monitoring datasets. These difficulties are further amplified by time-varying noise that can mask whale calls and by site-specific ambient sound characteristics to which a trained model may not generalize. In this work, we present NAVE, a Conformer-based detector that improves whale-call detection via three components. First, its input representation combines per-bin complex mean subtraction, which suppresses constant site-specific ambient sound characteristics, with an additional per-channel energy normalization (PCEN) channel that adapts to temporal variations in the noise floor. Second, a frequency-dynamic stem processes these noise-aware representations with convolutions that adapt across the characteristic frequency bands of the target call types. Third, we use a Conformer backbone with a convolution kernel matched to the duration of the target calls that adds an inductive bias and improves temporal localization. We evaluate NAVE on BioDCASE 2026 Task 2 using leave-one-site-out cross-validation across the validation sites, which approximates deployment on unseen recording sites. A single NAVE model with a tuned per-class threshold exceeds the stronger WhaleVAD-BPN baseline, using no boundary-proposal network and no search over post-processing parameters. A controlled ablation study shows that the normalized front-end and a duration-matched convolutional kernel each contribute substantially to recovering the D downsweep, the hardest class for previous detectors, with the wide kernel giving the largest aggregate gain.

0.32

Evaluating Anterior Cruciate Ligament Risk of Injury Using a Graph Convolutional Autoencoder

cs.CV ·2026-11-06 ·Athanasios Babouras, Paul A. Martineau, Thomas Fevens
Abstract · 14 pages, 9 figures

Current solutions for assessing the risk of anterior cruciate ligament (ACL) injury typically rely on expensive motion-capture labs that require substantial resource investment or subjective evaluations of specific athletic stress tests by experts. The Drop Vertical Jump (DVJ) and the Single Leg Hop (SLH) are two examples of such tests typically used to track progress in ACL rehabilitation. In this paper, we develop an unsupervised learning model that runs entirely on a smartphone for two tasks, based on an autoencoder architecture with multiple layers of graph convolutional networks, which utilizes pose estimation data to determine a person's injury risk status. Our model accurately predicts athletes at high risk of injury for both athletic stress tests, with average sensitivities and specificities of 78% and 78% for the DVJ, and 75% and 74% for the SLH. Our findings support the development of a widely accessible smartphone solution for real-time screening of ACL injury risk.

0.32

A Privacy-Aware Deep Learning Benchmark for Facial Emotion Recognition

cs.CV cs.AI ·2026-11-06 ·Larissa Palhares, Maria Machado, Hudson Bruno, Raphael Ramos, Andrew Vale +3 more
benchmark
Abstract · 7 pages, 2 figures

Facial Expression Recognition (FER) traditionally relies on raw RGB images, introducing privacy risks, sensitivity to illumination variations, and susceptibility to identity-confounding visual shortcuts. As an alternative, this work proposes a lightweight, privacy-aware pipeline using exclusively geometric representations-facial landmarks (478 points) and blendshapes (52 coefficients, ARKit standard)-eliminating persistent raw video storage. Six neural architectures trained from scratch on fused CREMA-D and RAVDESS datasets were benchmarked against an ImageNet-pretrained ResNet50 baseline under a 5-fold subject-independent cross-validation protocol. Geometric models consistently outperformed the pixel baseline on unseen subjects with 12-124 times fewer parameters, while demonstrating superior robustness on subtle emotional expressions such as sadness. Ablation studies reveal that blendshapes capture the primary discriminative signal, while landmarks provide essential complementary spatial topology. These results demonstrate that compact geometric representations preserve affective semantics, eliminate downstream transfer learning dependency on visual backbones, and enable efficient, privacy-aware emotion recognition.

0.32

A lightweight convolutional neural network for target detection in an edge environment

cs.CV ·2026-11-06 ·Hughes Perreault, Ola Ahmad, Daniel Czuboka
published
Abstract · Presented at GVSETS 2023, Novi, MI. Distribution A: approved for public release

This paper proposes a novel, lightweight convolutional neural network (CNN) designed for edge device deployment. The CNN is developed to recognize and classify threats or targets of interest. The model is equipped with a real-time hotspot detection algorithm, which enables it to process information quickly and accurately, resulting in improved inference time of up to 20 fps and a detection range of more than 400 meters with typical thermal situational awareness sensors. The proposed CNN has been rigorously tested and validated to demonstrate its high accuracy, sensitivity, and low false positive rate, providing a reliable solution for edge devices to detect potential targets. Additionally, the lightweight design of the CNN allows for easy deployment, making it an ideal solution for extending the security bubble for edge devices. The proposed model can be used for a variety of applications, including surveillance, security, and object recognition.

0.32

A Spiking Neural Model of the Mental Lexicon: Representations, Similarity Structure, and Associative Retrieval in Perception and Production

cs.CL ·2026-11-06 ·Bernd J. Kröger
Abstract · 28 pages, 3 figures; code, lexical data and recorded simulations available on GitHub; ancillary figure-generation scripts and provenance doc…

A neural account of lexical access requires an explicit distinction between the structure of lexical representations, the associations linking representational levels, and the activity generated during retrieval. This paper specifies a spiking neural implementation of the mental lexicon within the Neural Engineering Framework and Semantic Pointer Architecture. Conceptual and phonological form representations are constructed recursively from relational graphs, whereas lemma representations provide independent lexical identities. The implementation contains 613 lexical entries and supports two directed processing pathways: perception maps a phonological form to its associated concept via a lemma, whereas production maps a concept to its associated phonological form via a lemma. Each pathway combines two associative-memory variants, winner-take-all (WTA) selection (competition) and thresholded transfer, and uses 199,400 leaky integrate-and-fire neurons. We formalize within-level similarity, cross-level associative retrieval, and the interpretation of decoded activity. Simulations show successful retrieval of eight selected words from the lexicon in both processing directions, with retrieval evaluated against all 613 lexical entries. These simulations use complete phonological forms as input for perception and complete concepts as input for production, illustrating lexical retrieval through the corresponding processing pathways. This paper provides a transparent description of lexical representations and neural retrieval mechanisms, supported by accessible code, lexical data, and online simulation documentation.

0.31

Uncertainty, Not Concealment: A Mechanistic Reappraisal of Sycophancy in Vision-Language Models

cs.CV cs.LG ·2026-11-06 ·Mohamed I. Eldagla, Zeyad H. ElShafei, Ahmed M. Anter
publishedcode/data
Abstract · Accepted to the NeurIPS 2026 Med-Reasoner Workshop

Vision-language models (VLMs) in medical retrieval-augmented generation must be able to say when a retrieved report contradicts the X-ray, so understanding whether they can detect such conflicts is important for safe deployment. The sycophancy mask hypothesis holds that these models internally detect conflicts but refuse to report them, concealing a detected contradiction. We introduce a three-part diagnostic that separates concealment from uncertainty: representation probing, activation steering, and two-way forced choice. Across four instruction-tuned VLMs on chest X-rays, we confirm behavioral suppression of contradiction and find that a linear probe separates matched from mismatched inputs, beyond what a simple CLIP model achieves. Steering the learned conflict direction never elicits contradiction, and forcing a binary choice does not recover one; instead, cross-modal attention to the image drops on mismatched pairs. We conclude that these models fail to contradict because the mismatch registers as hedged, gated uncertainty rather than concealed knowledge, calling for re-grounding and uncertainty surfacing. Code is available at https://github.com/mohamed-eldagla/Uncertainty-Not-Concealment/.

0.31

MetricPose: Flexible 3D Human Pose Estimation with Depth and Geometric Cues

cs.CV ·2026-11-06 ·Shifan Zhu, Sabad Modi, Hang Yang, Arushi Agrawal, Steven Gao +3 more
published
Abstract · 9 pages, 3 figures, ICRA submission

In this paper, we present a 3D human pose estimator that predicts metrically accurate human joint positions in the camera coordinate frame. This task requires recovering the configuration of the body joints, scale, and metric camera-local placement. In monocular images, body size and camera distance are coupled through perspective projection and therefore cannot be determined uniquely from image alone. This ambiguity can be reduced by combining three complementary sources of information: learned RGB priors, direct depth measurements, and geometric constraints (e.g., ground contact) from the environment. We extend a pretrained full-body estimator with optional metric depth and geometry-processing modules that improve camera-distance estimation and depth-guided body refinement while preserving RGB-only operation. The resulting unified model supports RGB, RGB-Ground, and RGB-D using shared learned parameters. Unlike pelvis-relative error, camera-local error also measures placement relative to the camera. Compared with the existing foundation model SAM 3D Body, our methods with RGB, RGB-Ground, and RGB-D modes reduce camera-local joint-position error by 12.16%, 31.00% and 54.61% respectively.

0.31

Interpretable Context-Aware Multi-Label Retinal Classification with Dual-Hub Heterogeneous Graph Neural Network

cs.CV cs.LG ·2026-11-06 ·Proshanto Chanda, Munib Mesinovic, Tingting Zhu, Abhirup Banerjee
code/data
Abstract

Multi-label retinal screening requires capturing co-occurring pathologies and clinical context, yet standard multimodal fusion remains opaque. We propose a dual-hub heterogeneous graph neural network (GNN) that couples spatial fundus features and structured clinical covariates through cross-modal message passing while providing inherent interpretability. Rather than relying on post-hoc saliency approximations, our architecture exposes multi-level decision pathways during forward aggregation: spatially biased self-attention captures lesion-level patch interactions, while modality-specific graph attention coefficients explicitly quantify the contribution of visual regions and clinical risk factors to each disease node. Evaluated on the BRSET dataset (16,266 photographs from 8,524 patients), the proposed model achieves 76.1% Macro-AUPRC across seven conditions, matching or exceeding image-only (75.9%) and ResNet-50+MLP fusion (75.9%) baselines, with largest contextual gains in macular edema (+3.5 percentage points, pp) and diabetic retinopathy (+2.1 pp). Codebase is available at: https://github.com/proshanto-c/ca-mlc-refactor.

0.31

Language, Not Code Abstraction: Natural-language Training Shapes Brain-aligned Geometry

cs.CL ·2026-11-06 ·Iñigo Parra
publishedbenchmark
Abstract · Accepted for the Neureps Workshop at NeurIPS 2026

Code-trained language models produce positive brain scores, which raises the question of whether abstract structure learned from programming languages can substitute for natural-language training. We trained 25 GPT-2-small models on TinyStories, Python, C functions, x86-64 assembly, and the corresponding machine code, and scored them using a shared tokenizer and encoding procedure against matched random-weight models. On Pereira-243 and Tuckute2024, TinyStories exceeds the mean of the four code models by $Δr = 0.053$ and $0.059$ (Holm-corrected $p<0.001$), while code training leaves alignment near the random-weight level. The language advantage is benchmark-specific: Blank2014 shows no corrected language-code difference, and Fedorenko2016 shows no language benefit. Geometry tracks the encoding result on Pereira-243: TinyStories seeds converge on the most similar representations, and its fit-free correspondence with neural dissimilarity peaks at $ρ= 0.14$, twice the code and random-weight levels. Binary-trained models have about three effective dimensions, so seed agreement alone does not imply neural correspondence. Updating only the embedding matrix on English improves mean-over-depth alignment for every code model more than an equal update on its own corpus. Within this architecture and training setup, natural-language training produces a distinct representational geometry with closer neural correspondence on Pereira-243

0.31

HMMF-Net: A Lightweight Hierarchical Multi-Frame Multimodal Fusion Network for Accurate Fish Feeding Intensity Assessment

cs.CV cs.AI ·2026-11-06 ·Phi-Hung Hoang, Van-Manh Tran, Nam-Thuan Trinh, Thi-Thu-Hong Phan
Abstract · 47 pages, 10 figures, 15 tables

Automated fish feeding intensity assessment (FFIA) is important for precision aquaculture and feed cost optimization. However, unimodal methods are vulnerable to acoustic noise and visual disturbances, while existing multimodal frameworks often rely on computationally intensive 3D convolutions or Transformers, limiting their use on edge devices. To address these limitations, we propose HMMF-Net, a lightweight multimodal architecture that combines early multi-frame visual encoding and multi-axis acoustic modeling through a hierarchical bidirectional messenger cross-modal fusion (BMCF) mechanism. The visual branch captures temporal variations in feeding behavior by stacking temporally distant frames as a multi-channel input to an adapted 2D EfficientNetB0 backbone. The acoustic branch employs adaptive multi-axis residual units (AMARU) to extract temporal, frequency, and local time-frequency representations from log-mel spectrograms. BMCF progressively exchanges information between the two modalities at intermediate representation levels through compact messenger tokens, reducing the computational overhead of dense cross-modal attention. Evaluated on the AV-FFIA dataset using five-fold cross-validation, HMMF-Net achieves 96.77% accuracy and 96.88% F1-score on the held-out test folds, with 3.987M parameters and 1.002 GFLOPs. Compared with the best-performing visual configuration (95.49% accuracy), HMMF-Net reduces the classification error by 28.38%. Deployment on an NVIDIA Jetson Orin Nano using TensorRT with FP32 precision maintains comparable accuracy, with an inference time of 11.54 ms per 2-second sample and a model size of 18.81 MB. These results demonstrate that HMMF-Net combines effective multimodal representation learning with computational efficiency, enabling accurate and practical FFIA on resource-constrained edge platforms.

0.31

SAGE-PAIR: Source-Pretrained Routed Residual Adaptation for Sparse-Feedback Recommendation

cs.IR ·2026-11-06 ·Weiqing Zeng, Zida Yang, Guquan You, Jiexin Zheng, Cong Wang
code/data
Abstract · 14 pages, 2 figures, including 3 pages of supplementary material

Recommendation systems often enter a new serving window before enough outcomes are available to retrain the deployed model. We study this regime with 0.05-1.0% target feedback, a range that spans our stress tests while keeping supervision far below the degrees of freedom of the source recommender. At this scale, existing fine-tuning and meta-learning methods still estimate target-specific updates from extremely sparse feedback, leaving the update direction weakly identified. SAGE-PAIR instead transfers a source-supported residual correction space and only calibrates this space using target outcomes. We introduce SAGE-PAIR, a staged adapter that pretrains a low-dimensional global residual on an adjacent source period and adds a soft router over residual experts conditioned on source user profiles. Target feedback then calibrates this residual state rather than relearning the recommender. On a locked test split of KuaiSAR-v2, SAGE-PAIR (Full) achieves a 1.77% relative improvement in NDCG@10, while a core-protocol comparison shows a further 0.25% gain from routed experts beyond source-pretrained PAIR. Matched source-pretrained LoRA and Adapter-Tuning baselines remain near the source model, supporting the need for a different adaptation space. Thus, SAGE-PAIR improves ranking quality with millisecond-scale adaptation and inference costs. Sparse-feedback adaptation is fundamentally governed by the identifiability of the update space rather than by label count alone. Source pretraining supplies estimable correction directions for scarce labels, while user-conditioned routing contributes a consistent, differentiated cross-user correction on top of the global correction. Our code is available at https://anonymous.4open.science/r/SAGE-PAIR-610E/.

0.30

Semantic Extraction Framework for Goal-Oriented V2X Communication Supporting Safe Intersection Passage

eess.IV cs.CV ·2026-11-06 ·Haoran Yang, Shan Lu, Takaya Yamazato
hot direction
Abstract · Accept by 2026 Asia-Pacific Conference on Communications

This paper proposes a transmitter-side semantic extraction framework for goal-oriented Vehicle-to-Everything (V2X) communication supporting safe intersection passage. Instead of transmitting raw sensory data, roadside infrastructure extracts task-relevant semantic information from video and traffic-signal sequences and organizes them into compact semantic packets for low-latency driving-decision assistance. The intersection-passing problem is formulated as a set of traffic tasks under Japanese traffic regulations, and the proposed pipeline generates perception and intent semantics through object detection, object tracking, region-of-interest (ROI)-based semantic extraction, and recommended-action generation. The framework is implemented and evaluated on a self-constructed, CARLA-based single-vehicle dataset collected under clear weather conditions. Experimental results show high recommended-action accuracy within the tested task set, while the mean processing latency remains below the frame interval of the 15~fps input video.

0.30

Prove What You Can: Hybrid Neurosymbolic Reasoning for Logical Natural Language Inference

cs.CL ·2026-11-06 ·Ada Canoğlu, Sadiq Qara
theory
Abstract · 32 pages, 6 figures, 13 tables; under review at ARR (October 2026 cycle)

Logical reasoning over natural language requires both semantic interpretation and deductive search. Large language models (LLMs) can propose plausible intermediate derivations, but they often fail to compose them into a complete proof and overpredict uncertainty. Neurosymbolic methods improve search by translating an entire problem into first-order logic and handing it to a prover, but a single bad clause can make the whole translation nonexecutable. We present Hybrid Reasoning, a three-stage pipeline that decouples proposing a proof, validating its steps, and deciding a label. An LLM proposes a structured proof tree with a raw verdict; a theorem prover validates each derivation locally, keeping only the subset it can prove; and an uncertainty-gated repair attempts to re-derive the label from the premises plus the validated derivations, when the raw verdict is UNCERTAIN. Across four benchmarks (FOLIO, ProofWriter, ProverQA, PrOntoQA) and five contemporary LLMs, Hybrid Reasoning is the top-performing method on the two hardest datasets, improving on the strongest baseline by 5.3 points on FOLIO and 4.9 on ProverQA, and reaches ceiling-level accuracy on the easiest. The repair stage operates at 78-98% precision, producing 326 correct relabellings against 19 regressions.

0.30

DAC-MACE: Context-Conditioned Expert Routing for Compact Scene Coordinate Regression in Camera Relocalization

cs.CV ·2026-11-06 ·Dingying Fan, Mingkai Liu, Haohua Que, Haojia Gao, Xianliang Huang +3 more
Abstract · 10 pages, 6 figures. Submitted to IEEE Sensors Journal

A camera becomes an absolute pose sensor when its images can be relocalized against a prebuilt map of the scene. On mobile robots and augmented-reality devices, that map must be small and quick to build. Scene coordinate regression (SCR) stores the scene in a compact network that predicts a 3D coordinate for every pixel, and mixture-of-experts (MoE) variants extend its capacity to larger outdoor scenes. In existing MoE-based SCR, however, the router assigns each pixel to an expert from the same local descriptor that the experts regress from, so repeated facades and windows at different locations receive similar routing inputs. We present DAC-MACE, which conditions soft expert routing on recalibrated and contextualized features. A residual context adapter (RCA) reweights the channels and locations of frozen encoder features, a global context encoder (GCE) mixes them with multi-scale self-attention, and an image-level pooled descriptor is added to the router input. A dual-threshold RANSAC--Levenberg--Marquardt estimator recovers the pose. On Cambridge Landmarks, DAC-MACE attains a mean per-scene median translation error of 13.0 cm over three seeds, the lowest among the RGB-only SCR methods compared (GLACE 13.6 cm, MACE 14.4 cm), and reduces the error on the largest scene from 24 to 19 cm relative to MACE. In a cumulative study, the MoE head, GCE, RCA, and the dual-threshold estimator lower the mean error from 17.0 cm to 14.4, 14.0, 13.6, and 13.0 cm. On 7-Scenes, adding GCE to a single ACE head raises Recall@(5 cm, 5$^\circ$) from 86.56% to 87.60%.

0.30

Breaking the Privacy Promise: Gradient Inversion Attacks on Federated Low-Rank Adapted Vision Transformers

cs.CV cs.AIcs.CRcs.DC+1 ·2026-11-06 ·Ritik K. Badiya, Siladittya Manna, Ashmit Sinha, M Yashwanth, Anirban Chakraborty
published
Abstract · Accepted at WACV 2027

Parameter-efficient fine-tuning (PEFT) has been increasingly adopted to fine-tune pretrained models with low computational cost and is therefore also preferred in the federated learning paradigm to reduce computational and communication overhead. In this work, we investigate the sparsely explored problem of the risk of private data leakage in LoRA-adapted Vision Transformers (ViTs), where low-rank updates weaken the gradient signal available to a server with access to individual client updates. We show that quality of reconstruction is limited by two coupled ambiguities: reduced information in LoRA gradients and missing spatial cues from frozen positional embeddings. To address this, we propose a spatially anchored generative inversion framework that incorporates server-prescribed trainable positional embeddings into ViT blocks during fine-tuning at clients. The resulting positional gradients act as layer-wise spatial anchors, providing additional positional cues, while the generative prior keeps reconstructions on a realistic image manifold, reducing duplicate-data and patch-permutation artifacts, {which produce similar gradients as those produced when the target client images are used}. Adding our proposed spatial anchoring to GIAS improves PSNR by 10.14 dB, SSIM by 0.45, and LPIPS by 0.098 on CIFAR-10, as well as PSNR gains of 7.83 and 11.74 dB on high-resolution CelebA and CARS datasets, respectively, in the FedSGD setting. Such gains in reconstructed image quality metrics are also maintained at higher batch sizes and in FedAvg settings. These results indicate that LoRA-based PEFT does not, by itself, prevent data leakage in FL, and carefully incorporated positional embeddings can help in exposing private information.

0.30

ProtoCouncil: Multi-Encoder Anatomical Prototypes for Prostate Cancer Detection in Micro-Ultrasound

cs.CV cs.AIcs.LG ·2026-11-06 ·Mohammad Mahdi Abootorabi, Nooshin Maghsoodi, Armin Saadat, Obed Dzikunu, Zhuoxin Guo +6 more
Abstract · 12 pages, 3 figures, Under Review

Purpose: Micro-ultrasound ($μ$US) is a promising modality for prostate cancer (PCa) detection, but identifying suspicious tissue depends heavily on clinical experience. Models trained on sparse, noisy core-level pathology offer little insight into their decisions and can fail under domain shift. Clinical use needs a traceable decision process and a flag for uncertain predictions. Methods: We introduce ProtoCouncil, in which three foundation encoders (MedSAM, DINOv3 and V-JEPA 2.1) learn case-based prototypes at three anatomical levels: whole gland, needle track and gland sectors. Prototypes are projected onto real training-core regions, and label-free fusion keeps every class logit as a bias plus an exact sum of prototype similarity contributions. Encoder disagreement triggers selective prediction. Results: In five-fold patient-level cross-validation (6,588 cores, 693 patients), ProtoCouncil reaches 82.9% AUROC for clinically significant PCa (csPCa, Grade Group $\geq$3), 4.3 points above the best previously published method. In zero-shot evaluation on two external cohorts, ProtoCouncil achieves higher all-cancer and csPCa AUROC than all previous methods. Nearest same-class exemplars match test-core tumor involvement better than random same-class cases. Externally, csPCa accuracy falls from 79% when encoders agree to 27% when two oppose the fused call. Abstaining on disagreement raises balanced accuracy by 7 and 15 points, while retaining 58% and 54% of the cores. Conclusion: ProtoCouncil grounds core-level predictions in exact similarity-based evidence from real, anatomically indexed training cases, which makes its reasoning interpretable. It achieves the highest csPCa AUROC across three cohorts and identifies unreliable predictions through encoder disagreement.

0.30

Region-aware error triage for medical image segmentation with groupwise conformal risk control

cs.CV ·2026-11-06 ·Nishi Kanta Paul, Md Shihabul Islam Shovo, Hoileong Lee
Abstract · 22 pages, 7 figures, 12 tables. Submitted to Biomedical Signal Processing and Control (BSPC)

Segmentation masks are increasingly reviewed rather than redrawn, which requires knowing where a mask is likely wrong. Learned error maps rank pixels by risk, but the number of pixels to warn is chosen ad hoc, and calibrating one warning threshold over all error pixels may leave errors far from the predicted contour under-flagged. We propose RACE-Triage, a post-hoc error triage framework that leaves the segmentation mask unchanged. A risk network is trained on out-of-fold predictions from backbone-independent inputs. The predicted contour partitions each image into a boundary band and a non-boundary region, and conformal risk control is applied separately in each region to bound the expected fraction of missed errors among cases with errors there. On ISIC 2018 dermoscopy and BUSI breast ultrasound (where patient identifiers are unavailable, so independence across images is assumed), pooled control met its overall target but missed on average 31.8% and 36.4% of non-boundary errors. Regional control reduced these fractions to 19.3% and 20.9% and raised the recall of size-filtered false-negative components from 0.722 to 0.818 and from 0.611 to 0.763. The mean warned area grew by 1.6 and 4.9 percentage points, and its 95th percentile reached 23.5% and 38.5%. The effect held across three seeds and a second backbone, and area-matched heuristics did not reproduce it. On the external PH$^2$ set, frozen thresholds exceeded both targets, whereas recalibration with 100 local images brought the mean over repeated splits below them. Regional calibration improves non-boundary error coverage at a measured increase in warned area.

0.30

From Weapons to Events: Evaluating Evidential Support in Multimodal AI for Policing

cs.CV cs.AIcs.CY ·2026-11-06 ·Armin Mustafa, Syed Sameed Husain
dataset
Abstract · 34 pages, 2 figures, 4 tables; appendices include annotation guidelines, prompt templates and a datasheet

Computer vision (CV) for policing has centred on recognising objects such as seized weapons, while body-worn video (BWV) is still reviewed by people, with existing AI tools using only its audio. Multimodal video AI could make BWV review faster and cheaper. At the object level, we present the KnifeHunter dataset, developed with UK law enforcement, with over 25,000 images of 543 knife classes and a protocol reflecting investigative conditions, and KHNet, a retrieval architecture fusing local and global features, which with about 125,000 distractor images attains 85.7% and 63.4% mean Average Precision (mAP) on the medium and hard protocols. At the event level, we present an evaluation protocol for video understanding pipelines, which describe a video as events and a transcript, that scores transcription, speaker attribution, event detection, temporal localisation and unsupported events. Whether each event is supported by the recording is labelled by a judge validated against blinded human labels. We apply it to complete pipelines, three 2026 omni-modal models and our hybrid NarrativeBridge-PAL-GPT-5 pipeline, on UKBWV-27, a pilot stress-test set of 27 publicly released UK BWV recordings (118 minutes). On this set the hybrid pipeline leads on five of six measures, yet it leaves 13.6% of predicted events unsupported and misattributes 27.7% of utterances, and every pipeline degrades on recordings from edited releases, long recordings and noisy audio. We propose readiness levels separating assistive tasks from those requiring verification of every output.

0.29

Compaction, decay and reinforcement in a two-tier conversational memory: a paired evaluation against retrieval-augmented generation

cs.CL ·2026-11-06 ·Osamudiamen Edogun, Joshua Obigwe, Irfan Mehmood
code/data
Abstract · 7 figures. Data and code: https://github.com/BigBrown10/nmafc-paper

A conversational agent that remembers has to decide what to keep. Retrieval-augmented generation keeps the transcript and searches it; the alternative is to compact each turn into atomic facts and search those. We compare the two on LoCoMo, with both arms answering all 1,985 questions in one window, against the same stores, with the same generator and judge, and both predictions on one row, which licenses McNemar's exact test on the 512 questions where they disagree. On the four categories comparable work reports, compaction scores 71.4 per cent against 64.6, a lead of 6.8 points (p = 5.5e-08) at 34 per cent fewer prompt tokens. On all five, including the adversarial set comparable work drops, the lead falls to 2.0 and stops being significant, because the baseline leads adversarial by 14.6. That loss has a structure. Our lead falls as the reference answer demands more source wording, 9.6 points ahead at one or two reference words and 3.1 behind past eight, with token overlap moving the same way, so the judge is not the cause. Compaction keeps what a fact asserts and discards how it was said, so it wins where the answer is a value and loses where it is a phrasing. We also report four negative results with their power calculations.

0.29

Jais 3: Introducing Arabic-centric Agentic AI

cs.CL cs.AI ·2026-11-06 ·Team IFM, Haonan Li, Hadi Abdine, Nikhil Ranjan, Mohamed Anwar +49 more
Abstract · 113 pages

We present Jais 3, a fleet of six open, Arabic-centric agentic language models ranging from 0.9B to 375B parameters. Jais 3 is built to complete tasks for Arabic speakers, not only to answer questions in plain text: it is trained to plan multi-step tasks, call tools and Model Context Protocol (MCP) servers, operate a terminal, search the web, and read, edit, and produce PDF, spreadsheet, and presentation files, while communicating with the user in Arabic. Instead of pretraining from scratch, we adapt each model of the fully open K2 Horizon fleet to Arabic through continual pretraining, followed by supervised fine-tuning (SFT), final token preference optimization (FTPO), and reinforcement learning with verifiable rewards (RLVR). We combine natural Arabic text with translated, transliterated, and synthetic Arabic, and replay the base model's own training data to keep its reasoning, agentic, and English abilities. Jais 3 covers Modern Standard Arabic and the major regional dialects, follows instructions with Arabic-specific constraints, and inherits the 512K-token context window of K2 Horizon (all models except the 0.9B). To our knowledge, its 375B model is the largest open Arabic-centric language model to date. For evaluation, we revisit how Arabic models are assessed, moving to fully generative evaluation on audited Arabic benchmarks. Across model sizes, Jais 3 matches or outperforms open-weight models of comparable size on Arabic and agentic benchmarks, with particularly strong agentic capabilities and Arabic instruction following, and far fewer hallucinations. We release the weights and the checkpoints from every training stage of all six models, together with samples of the training data, the data recipes and training configurations, the training logs, the audited benchmarks, and the evaluation code.

0.29

CoMET: Tracking the Shared Pose Error of Roadside Messages for Asynchronous Vehicle-Infrastructure Cooperative 3D Detection

cs.CV ·2026-11-06 ·Shuwei Sun, Jian Huang, Hui Cao
benchmark
Abstract · 25 pages, 4 figures. Code and experimental results archived at https://doi.org/10.5281/zenodo.23999418

A roadside sensor can detect vehicles hidden from the ego vehicle, but a pose error shifts every detection in the same message. We study asynchronous vehicle-infrastructure cooperative 3D detection on V2X-Seq using only information available at decision time, without ground truth, future frames, or per-pair calibration offsets at inference. Objects seen by both sensors reveal the shared translation, but even at zero added delay, 38% of validation messages have no matched pair. We propose CoMET (Common-Mode Error Tracking), which estimates this translation over time. At each message arrival, CoMET matches vehicle and roadside detections, estimates the offset from mutually nearest pairs, and updates a Kalman filter for that sender. When no pair matches, the filter carries its estimate forward. The correction applies to all roadside boxes, including those unseen by the ego vehicle. The estimator has no learned parameters; a separate frozen head ranks the merged detections. Across four delays from 0 to 400 ms on 21 validation sequences, CoMET averages 66.63 bird's-eye-view AP at IoU 0.7, compared with 63.18 for the same late fusion using uncorrected online poses and 62.96 for the best intermediate-fusion baseline. Across five detectors from three architectures, CoMET gains 2.2 to 3.5 AP points over online poses and 0.3 to 0.6 over estimating the offset from the current message alone. We find that the benchmark uses a zero offset as a placeholder in 28% of validation messages and that its offset search reaches the boundary in 45% of the rest. These offsets affect the labels for objects seen only from the roadside. Under two alternative label references, CoMET improves on uncorrected online poses in every evaluated population. Retraining intermediate fusion with CoMET poses and corrected labels improves detection on all three architectures tested, with statistically significant gains on two.

0.29

FusionCLIP: Multimodal Contrastive Pretraining for Fusion Plasma Diagnostics

cs.CV cs.AI ·2026-11-06 ·Qiang Chen, Xiao Wang, Qingquan Yang, Hao Si, Zikang Yan +2 more
Abstract

Visible-light videos and multivariate diagnostic signals provide complementary observations of tokamak discharges. However, existing fusion pretraining methods, primarily learn representations by reconstructing or predicting observations, leaving the correspondence between these modalities underexplored. Representation learning is further challenged by the fact that plasma-related regions occupy only part of each video frame and diagnostic variables have substantially different sampling rates. We propose FusionCLIP, a multimodal contrastive pretraining framework. Spatiotemporal Stability Selection retains informative visual tokens, while a Temporal Query Encoder integrates asynchronous diagnostic signals using their observation timestamps. Contrastive learning then aligns the two modality representations. We construct EAST Multimodal Diagnostics (EAST-MMD), comprising 1307 EAST discharges, and evaluate the model on three independently fine-tuned tasks: disruption prediction, edge-localized mode (ELM) classification, and $H_{98}$ regression. FusionCLIP achieves an aggregate score of 89.07\%, exceeding the best general-purpose pretraining configuration and an in-domain contrastive pretraining configuration without Spatiotemporal Stability Selection by 6.27 and 2.12 percentage points, respectively. Ablations and attention visualizations further support the contributions of its key components.

0.29

SPDMX: Large-Scale Open Stem Audio From Symbolic Music

cs.SD cs.AIcs.LGcs.MM+1 ·2026-11-06 ·Phillip Long, Eduardo Escoto, Julian McAuley, Zachary Novack
code/data
Abstract · 4 pages, 3 figures, 1 table

Large public datasets have driven recent progress in machine learning, from text and vision to speech and general audio. Music remains an outlier, as copyright blocks redistribution of training audio, while many modern tasks need individual stems rather than mixtures. Existing multitrack resources trade off realism, scale, and licensing, while synthetic alternatives rely on proprietary VSTs or naive open rendering that yields weak stems. We introduce SPDMX, a listening-guided rendering of public-domain PDMX scores into an openly-licensed stem corpus with aligned scores and an open render pipeline: $253{,}982$ songs, $490{,}159$ mono stems, and $5{,}469$ song-hours at $44.1$ kHz under CC BY 4.0, with paired MIDI and identifiers linking to PDMX metadata. $88{,}403$ songs ($34.8\%$; $2{,}734$ song-hours) provide multitrack supervision. To our knowledge, this is the largest openly-licensed synthetic multitrack subset by song-hours, enabling source separation, stem-conditioned synthesis, MIDI-to-audio, transcription, and joint symbolic-audio modeling. We release the corpus, metadata, and open pipeline for reuse on other symbolic collections. Demos and experimental notes are on the project page.

0.29

Zero-Shot Benchmarking of Unified Multimodal Models for Shallow-Water UAV Image Distortion Assessment and Restoration

cs.CV ·2026-11-06 ·Dario Scilla, Omar A. Lopez, Mattia Rosso, Kasper Johansen, Matthew F. McCabe
benchmarkhot direction
Abstract

Shallow-water UAV imagery provides centimeter-scale observations of benthic environments, but water-induced distortions complicate interpretation. To our knowledge, we present the first systematic zero-shot benchmark of unified multimodal large language models (MLLMs) for both assessing water-induced distortions and restoring shallow-water imagery within the same model framework. We evaluate five open-weight unified MLLMs with frozen pretrained weights and without access to paired reference images or descriptions of the true seabed. The benchmark contains 1,000 image pairs: 600 semi-synthetic UAV scenes created by applying synthetic water distortions to real shallow-water imagery and 400 fully synthetic through-water scenes. Three independent human annotations cover five distortion types. We evaluate four assessment prompts (P0-P3), from direct instruction to structured task decomposition, on all 1,000 images and four diagnosis-guided restoration prompts (R0-R3) on a common 100-image subset. Assessment varies substantially by model and prompt, with systematic prediction biases and a best pooled balanced accuracy of 0.654 for distortion presence. The best pooled PSNR gain is +0.255 dB, no unified MLLM improves pooled LPIPS, and some outputs alter scene structure or hallucinate content. Two restoration-only baselines also show limited transfer: InstructIR achieves a pooled PSNR gain of +0.040 dB, while PromptFix degrades all four pooled metrics under every prompt. These results show that recognizing shallow-water degradation does not imply reliable physical restoration and provide a reproducible baseline for future work.

0.29

Health-System-Wide Mapping of Demographic Information in Routine Clinical Histopathology

cs.CV q-bio.QM ·2026-11-06 ·Shengjia Chen, Ruchika Verma, Kevin Clare, Paula Restrepo, Jannes Jegminat +6 more
Abstract · 35 pages, 4 main figures, 2 supplementary figures, and 10 supplementary tables. Interactive Results Atlas and Hypothesis Explorer accompany …

The extent to which demographic-associated information is detectable across organs and disease states in routine clinical histopathology remains unclear. Here, we analyzed 1.3 million hematoxylin and eosin whole-slide images from 257,322 patients across 25 organs, using pathology foundation-model representations and multiple-instance learning to systematically map information associated with sex, age group, and self-reported race. Across 69 organ-demographic combinations, demographic attributes were broadly recoverable, with pan-organ mean AUROCs of 0.828 for sex, 0.733 for age group, and 0.662 for self-reported race, and substantial variation across organs. Demographic-class balancing produced relatively small changes in performance, while evaluation within diagnosis-text clusters showed modest attenuation. Under training with diagnosis-based grouping, demographic-associated signals remained detectable across multiple organs, including sex in gastric tissue (AUROC 0.865), age group in breast (0.816), and self-reported race in skin (0.728). Literature-guided analysis of these settings generated candidate histomorphological and contextual explanations for further investigation. Together, these findings provide a systematic map of demographic-associated information in routine histopathology, informing investigations of its sources, assessments of potential shortcut learning, and the development of more interpretable computational pathology models.

0.29

Learning of Tree Automata applied to Neural Language Acceptors

cs.CL cs.AI ·2026-11-06 ·Juan Pedro da Silva, Sergio Yovine
Abstract · Presented as WIP paper at ICGI 2026, Delft, the Netherlands, October 2026

We extract visibly pushdown grammars (VPGs) from neural language models in a fully black-box setting. Our work builds on the VPL* framework, which learns VPGs from recurrent networks by exploiting access to the target's internal state. First, we replace the white-box equivalence oracle with PAC sampling over trees. The resulting algorithm applies unchanged to transformer language models and, more generally, to any acceptor exposing only binary decisions. Second, we further observe that the original framework only ever queries the target on well-formed sequences that can be accepted by the most permissive VPG (called BParse) with the target's ranked alphabet. That is, VPL* is blind to the target's behavior outside BParse. Thus, we propose to sample trees directly instead of sequences and call the target with all the resulting sequences in order to provide an error estimate of how far the target deviates from being a VPG. Last but not least, we develop a learner that captures a tree automaton describing discovered behaviors of the target outside BParse, which results in a larger learnable space. We evaluate the approach on several cases, including transformers trained with Dyck grammars and synthetic targets designed to be outside BParse.

0.28

Seeing vs. Counting: Complexity Twins Show Where Multimodal LLMs Fail on Diagrams

cs.CV cs.AIcs.CL ·2026-11-06 ·Yin Liu
Abstract · 8 pages main text + 96 pages appendix (107 pages in total, including references)

When a multimodal LLM miscounts in a diagram, does it fail at seeing or at counting? Existing visual graph and puzzle benchmarks vary visual density and algorithmic hardness together, confounding the two. We pair counting questions about the same drawing that sit in different counting-complexity classes, such as its spanning trees (a determinant) and its simple paths (#P-complete), and use a transcription probe and a text-only control to separate reading from computing. We test Gemini, Gemma and Qwen3-VL models; for the Gemini models, the answer depends on how the model computes. In chain of thought, most of them read small drawings correctly and then fail at counting, at a rate set far more by the size of the answer than by its complexity class (for the primary model, the odds of an exact answer fall 2.38-fold per decade of the answer). Given code and an exact transcription, the primary model counts correctly even on 10x10 grids, and reading becomes the bottleneck: a single view misses horizontal edges, and a second reading of a rotated copy with crop checks reads 86/96 large grids exactly, against 57/96 when the upright reading is simply repeated.

0.28

Polarization Aware Deep Learning Framework for SAR Despeckling

eess.IV cs.CV ·2026-11-06 ·Gianpaolo Passarello, Sergio Vitale, Hossein Aghababei, Giampaolo Ferraioli, Vito Pascazio
Abstract

Speckle is an inherent characteristic of synthetic aperture radar (SAR) imagery that complicates image interpretation and the extraction of reliable information. Although deep-learning-based despeckling methods have demonstrated promising performance, their applicability across different polarimetric configurations is often limited by the need for specialized architectures or separate training procedures. This paper presents a unified deep-learning framework for despeckling single-, dual-, and full-polarization SAR data using a single trained convolutional neural network. The proposed approach adopts a common nine-channel representation of the polarimetric covariance matrix and a polarization-aware masking strategy to accommodate different configurations without modifying the network architecture. Qualitative and quantitative comparison on three different scenes from different sensors demonstrate competitive performance across the considered experiments and support the feasibility of a unified, polarization-flexible approach to SAR despeckling across different polarimetric configurations.

0.28

BEV3SR: Evidence-Aware 3D Spatial Reasoning from Video

cs.CV ·2026-11-06 ·Fan Zhu, Rui Yu
published
Abstract · WACV camera ready

3D spatial reasoning from video remains challenging for VLMs because videos provide sequential, view-local observations rather than persistent metric scene records. Many questions must be answered in a query-specified perspective rather than the observed camera viewpoint, and prompting with a fixed set of RGB views may miss sparse or occluded visual cues. To address these gaps, we introduce BEV3SR, a training-free and backbone-agnostic framework for RGB-D videos with camera parameters. BEV3SR turns video observations into queryable metric bird's-eye-view (BEV) scene memory and formalizes each query into an evidence-aware spatial task. The Spatial Task Formalizer specifies the query's reference frame, spatial constraints, objective, and evidence requirement. The evidence requirement determines the appropriate reasoning pathway. A query treated as requiring geometric evidence is handled by the Geometric Reasoner, which deterministically computes the answer using the metric BEV records in the query-specified reference frame. Otherwise, a query treated as requiring visual evidence is passed to the Visual Evidence Aligner, which uses BEV-indexed object records and their associated supporting views to retrieve object-consistent and scene-context RGB frames for VLM grounding. Within a standard 16-frame budget, BEV3SR consistently improves different VLM backbones across multiple 3D scene QA benchmarks, delivering large gains on spatially intensive questions and significant improvements on open-ended scene QA. These results demonstrate that BEV3SR offers an effective, efficient, and interpretable framework for explicit 3D spatial reasoning from video.

0.28

UAV-Based Computer Vision for Cattle Detection and Feed Bunk Spatial Analysis

cs.CV ·2026-11-06 ·Pedro Oliveira Torrente, Diego Giaretta de Paulo, Lucas Pascotti Valem, Jianglong Yan, Everton Castelão Tetila +2 more
Abstract · 12 pages, 6 figures, 6 tables. Preprint of paper accepted in IEEE Transactions on AgriFood Electronics

Beef cattle production is central to Brazilian agribusiness and faces growing demands for efficiency, sustainability, and animal welfare. Precision Livestock Farming (PLF) offers tools to address these challenges, particularly through monitoring feeding-related spatial patterns in confined systems. Feeding management plays a decisive role in animal performance, yet its evaluation still relies largely on subjective, labor-intensive visual inspections that are hard to scale and prone to bias. This work proposes a computer vision framework based on Unmanned Aerial Vehicles (UAVs) to quantify cattle alignment and spatial dispersion at the feed bunk under real-world feedlot conditions. A UAV dataset was collected during flight campaigns at a commercial feedlot in Mato Grosso do Sul, Brazil. Images were manually annotated into cattle head and cattle back classes to support object detection and image-level spatial analysis of cattle distribution. You Only Look Once (YOLO) models, namely YOLOv8 and YOLOv11, were evaluated, with YOLOv8 small used for the downstream spatial analysis. The feed bunk region was segmented using the Segment Anything Model (SAM), and geometric metrics based on detected head positions were developed to characterize alignment, spacing regularity, and clustering. The selected detector achieved 84.30% mAP50 and 35.50% mAP50-95, with an inference time of approximately 2.7 ms per image on an NVIDIA RTX 5060 Ti GPU. The resulting metrics distinguished representative regular and irregular spatial distributions by Root Mean Square Deviation - RMSD (1.06 m and 4.26 m), Coefficient of Variation - CV (42.70% and 96.31%), and clustering (0 and 2 clusters), respectively. These findings indicate that the proposed approach can provide indicators for assessing feed bunk spatial patterns in PLF.

0.28

AI-Driven Repair of Missing Automotive Components

cs.CV ·2026-11-06 ·Kazim Abbas, Zheng Liu
Abstract · 6 pages, 4 figures

Due to supply constraints, replacing damaged automotive components can be expensive and time-consuming, while replacing the entire component can be wasteful when much of it remains usable. This paper presents a pipeline to investigate whether missing geometry can be reconstructed as a repair patch using paired observations of the intact models and synthetically damaged components. Noise-free color, depth, and silhouette images were produced using the camera viewpoints. Depth differences were then used to identify the potential missing regions, while multi-view voxel reconstruction was used to estimate the missing portion. A pretrained DINOv2 model was also added to validate missing candidates and determine whether pretrained visual features could improve reconstruction. We tested the pipeline on 20 automotive CAD models, each with a synthetically removed portion. Each model was tested under four different conditions using two reconstruction methods for a total of 160 trials. The predictions were compared with the missing geometry of the unchanged ground-truth using predefined thresholds of the voxel intersection-over-union over 0.85 and the absolute relative volume error less than 15%. Of the 160 experimental trials, 109 of them satisfied both thresholds. Geometry-only reconstruction passed 16 of 20 components with eight views but only 10 of 20 with four adjacent views. Further analysis showed that 24 out of the 37 failed reconstructions could be categorized as unsuccessful because extra material was incorrectly added to areas that should have been empty. The implementation of DINOv2 into the pipeline did not improve the overall reconstruction accuracy and made prediction generation slower. For 13 verified tests, printing only the repaired sections instead of the entire parts reduced filament use by 93.08% and printing time by 89.39% based on the median results.

0.27

Beyond Segmentation: Enriching ImageCAS with Coronary Ostia Landmarks, Rooted Centerline Graphs, and Open-Weight Localization Models

eess.IV cs.CV ·2026-11-06 ·Matteo Leccardi, Andrea Moglia, Marco Morandini, Michele Brembilla, Alessandro Nicosia +6 more
code/data
Abstract · 52 pages including the appendices. Dataset, centerlines, code and model weights: https://hal9000-lab.github.io/

Coronary ostia and centerlines are key reference points for automated analysis of coronary computed tomography angiography (CCTA), from surgical planning to intraoperative vessel tracking. However, the only public ostium annotations cover 77 non-gated aortic scans, not CCTA, while CAT08 and ASOCA (32 and 40 CCTA scans) mark the ostia only implicitly, as centerline start points. This work enriches ImageCAS, the largest public dataset for coronary artery segmentation, with ostium annotations for its 1000 patients and centerline graphs rooted at these new ostia, making it, to the authors' knowledge, the largest public dataset for ostium localization and the only one with centerlines following the official ImageCAS masks and starting at annotated ostia. Three annotators marked each ostium in the first 500 volumes; in the next 500, the candidates of a model trained on the first batch were corrected (human-in-the-loop). Six 3D localization models, trained on ImageCAS, CAT08 and ASOCA, were benchmarked on 58 held-out volumes (50 ImageCAS, 4 CAT08, 4 ASOCA). SwinUNETR-V2 and nnResUNet localized all ostia without failure (>5 mm error), with median errors of 0.70-0.87 mm. Trained on CAT08 and ASOCA alone, SwinUNETR-V2 fails on 34% of right and 53% of left ostia, whereas adding 250 ImageCAS volumes removes all failures. The centerlines were extracted by a deterministic, quality-controlled minimal-path procedure; on ASOCA, they cover a median of 99.5% of the reference length, with a median per-patient mean distance of 0.22 mm between matched reference points and centerlines. Annotations, centerlines, code, and weights are available at https://hal9000-lab.github.io/.

0.27

T2I Models Rarely Invert Negation but Often Fail to Suppress Negated Concepts: Benchmarking Negation Suppression in Text-to-Image Generation

cs.CV ·2026-11-06 ·Akansha Gautam, Aparna Bharati, Mayank Vatsa, Richa Singh
publishedbenchmarkcode/data
Abstract · Accepted at WACV 2027. Code and data: https://github.com/IAB-IITJ/T2I-NSI

Text-to-image diffusion models often fail to suppress explicitly negated concepts, such as a lamp for "a desk without a lamp." We introduce T2I-NSI, a benchmark and metric for measuring negation sensitivity in text-to-image generation. It pairs each negated prompt with a canonical positive counterpart and factorizes negation by linguistic form (lexical/semantic, morphological, syntactic) and target type (objects, attributes, relations, compositions, and nested cases). The benchmark contains 759,869 paired prompt atoms over a 5,082-object catalog, with 916,272 images (911,982 open-weights, 4,290 closed-source), scored by a Basic CLIP estimator and a Robust ensemble of CLIP, VQA and open-vocabulary detection. Negation sensitivity is low: open-weights generators score NSI <= 0.038, while GPT-Image-1.5 and Gemini-2.5-Flash-Image reach about 0.09. The dominant failure is under-suppression rather than misdirection. Inverse interpretation is essentially absent (0.00% across over 450K atom-model evaluations), whereas models render the concept under both prompts on 8-19% of atoms and only partly suppress it on a further 9-25%; most remaining atoms are too weakly rendered to label. A blinded human study (420 atoms, nine annotators) corroborates the near-absence of inversion and the high persistence rate, and a matched-prompt comparison shows the open-versus-closed gap (~3.5x) is not a prompt-generator artifact. Code and data: https://github.com/IAB-IITJ/T2I-NSI

0.27

KLEAR: Knowledge-Matched LLM Evaluation in Korean across Application Requirements via Controlled Benchmark Transformation

cs.CL ·2026-11-06 ·Seoyeon Kim, Jeonghyun Kang, Hamin Koo, Barom Kang, Heondeuk Lee +1 more
benchmark
Abstract · 34 pages, under review

Existing Korean benchmarks for Large Language Models (LLMs) evaluate performance across diverse knowledge domains, but their accuracy scores cannot distinguish whether incorrect responses arise from a lack of required knowledge or a failure to apply that knowledge correctly. Our analysis of existing Korean benchmarks reveals that most questions can be answered through factual recall alone, while those demanding greater application also differ in required knowledge, making it difficult to isolate application ability. To address this, we introduce KLEAR, a framework that constructs knowledge-matched contrasts by varying application demand while preserving the required knowledge. During construction, KLEAR identifies the knowledge underlying each source question as an explicit anchor bundle to guide generation of questions requiring more complex application. Generated candidates then undergo automated and human validation, with iterative revision based on evaluation feedback. Experiments demonstrate that the resulting knowledge-matched sets largely satisfy the intended constraints, enabling controlled evaluation across diverse LLMs. These evaluations reveal that increasing application demand exposes failures that persist even when the required knowledge is provided.

0.27

Concordance-Optimal Continuous Music Emotion Estimation: Variance Decomposition, Dual-Branch Temporal Fusion, and Conformal Prediction Intervals

cs.SD cs.LG ·2026-11-06 ·Mohammad Faraz Akhtar, Aditya Parashar, Ayush Jain, Madhur Gupta, Ritesh Kumar +1 more
code/data
Abstract · 28 pages, 12 figures

Continuous music emotion recognition (MER) estimates how Arousal and Valence evolve within a song, yet the field's standard metric, pooled frame-level correlation, is dominated by between-song differences and cannot distinguish a model that tracks emotional dynamics from one that only ranks songs by their average level. We decompose annotation variance into between- and within-song components and show that mean squared error (MSE) targets the former, while windowed Lin's concordance correlation coefficient (CCC = $r \cdot C_b$) targets the latter by penalising the variance compression MSE induces. A single weight $λ$ controls the trade-off: on 1,764 DEAM songs, raising $λ$ from 0 to 1 lowers pooled Arousal $r$ from 0.843 to 0.810 while raising mean within-song $r$ from 0.204 to 0.368. Building on this objective, we present MMAS (Multimodal Music Analysis System), an open-source model that predicts Arousal and Valence every 0.5 s from raw audio. It fuses frozen MERT-v1-95M embeddings with ten symbolic features from automatic transcription through a dual-branch temporal convolutional network with cross-modal attention. Because both branches are audio-derived, one model trains jointly on DEAM and PMEmo without domain conditioning. Under song-grouped 5-fold cross-validation (2,531 songs), pooled $r$ reaches 0.801/0.685 (Arousal/Valence) on DEAM and 0.742/0.746 on PMEmo. Against a ridge baseline on identical features, within-song Arousal $r$ roughly triples (0.344 vs. 0.125; $n = 353$ songs, Wilcoxon $p = 1.7 \times 10^{-21}$, $d = 0.56$). Complete transcription failure costs at most 0.04 pooled $r$. Split-conformal intervals attain 93.7% (Arousal) and 90.0% (Valence) marginal coverage at a 95% target, with the shortfall traced to within-song dependence. MMAS is released as a pip-installable package.

0.27

Beyond the Prompt: Leveraging Pre-Decoding States for Jailbreak Detection in dLLMs

cs.CL cs.CYcs.LG ·2026-11-06 ·Adam Hazimeh, Amel Abdelraheem, Ke Wang, Mariam Salman, Ljiljana Dolamic +2 more
top venue
Abstract · Published at NeurIPS 2026

Diffusion large language models (dLLMs) have recently emerged as a strong alternative to autoregressive models, generating responses by iteratively refining masked tokens in parallel rather than decoding sequentially. While enabling more efficient and flexible inference, dLLMs remain vulnerable to jailbreak attacks, including recent attacks that exploit properties unique to dLLMs such as bidirectional conditioning and parallel masked-token refinement. In this work, we show that these same properties also expose a safety-relevant detection surface absent from standard autoregressive decoding: even when a jailbreak is difficult to identify from prompt representations alone, it can still be detected from hidden states over masked response positions before any tokens are generated. By training linear classifiers on prompt and pre-decoding response representations, we empirically show that neither view strictly dominates, with each detecting attacks that the other misses. We introduce $\texttt{ReFuse}$ (Representation Fusion), an inference-time detector that fuses the sigmoid harmfulness scores of the two classifiers. Without modifying model weights or the decoding procedure, $\texttt{ReFuse}$ reduces average ASR across multiple jailbreaks from 63.29% for the undefended model and 11.27% for a prompt-only classifier to 3.31%, while keeping average benign refusal on standard utility benchmarks below 1% on LLaDA-8B. Together, these results establish $\texttt{ReFuse}$ as a lightweight jailbreak detector for dLLMs that improves safety beyond prompt-only detection.

0.27

Query-Guided RoPE Modulation for Lost-in-the-Middle Mitigation in Long-Context LLMs

cs.CL cs.AIcs.LG ·2026-11-06 ·Peiyu Zhang, Heng Ping, Shinibali Bhattacharyya, Anwesan Pal, Nikos Kanakaris +4 more
code/data
Abstract · 13 pages, 3 figures, 7 tables. Code: https://github.com/iZHpy/QGPM

Long-context language models often neglect information in the middle of the input, a phenomenon known as "lost-in-the-middle". Part of this gap may stem from the long-range decay of rotary position embedding (RoPE), which weakens the positional signal between distant tokens. Existing input compression deletes low-importance tokens, and attention reweighting acts only after RoPE has weakened this signal. RoPE-level methods modify positions directly but apply the same remapping to every query. To address these limitations, we propose QGPM (Query-Guided Positional Modulation), a training-free method that reshapes RoPE positions using query relevance from the model's own attention heads. QGPM builds a broad relevance map for lower layers and a sharp one for upper layers. Low-relevance tokens advance by less than one position on fast-rotating planes, pulling relevant evidence toward the query. Slow-rotating planes keep the original layout, with every token in place. Experiments on NoLiMa and RULER with multiple context lengths show that QGPM improves over base models and strong baselines, with the largest gains on retrieval tasks, and generalizes across multiple backbones. Our code is available at https://github.com/iZHpy/QGPM.

0.27

Emo-LiveSync: Real-Time Universal Lip Synchronization with Explicit Emotion Control

cs.CV ·2026-11-06 ·Sangjoon Yu, Seungeun Lee, JoonHeyok Shin, Ji-Su Kang
published
Abstract · Accepted to BMVC 2026, 23 pages, 16 figures

Lip synchronization, the task of generating realistic lip movements synchronized with audio signals, is fundamental to applications such as film dubbing, virtual avatars, and digital content creation. Recent GAN-based methods offer computational efficiency but often lack visual fidelity, while diffusion-based approaches achieve high-quality results at the cost of substantial computational overhead. Moreover, most existing methods are limited to photo-realistic human faces, restricting their applicability to stylized characters and non-human entities. In this paper, we propose an efficient universal lip-sync video generation framework that achieves high-quality results without compromising computational efficiency. We build on LivePortrait, a warp-based portrait animation model, and introduce a training strategy that aligns audio signals to an implicit motion space by distilling knowledge from a pre-trained video-to-motion model. This enables real-time inference while maintaining high visual fidelity, and naturally generalizes to diverse stylistic characters beyond photo-realistic humans. Furthermore, we introduce an expressive control gate that enables explicit manipulation of emotional expressions for content enrichment. Extensive experiments demonstrate that our method achieves state-of-the-art performance in both visual quality and lip-sync accuracy while ensuring real-time inference.

0.26

GRAC: Graph-Based Retrieval-Augmented Captioning for Surgical Scene Description

cs.CV cs.AI ·2026-11-06 ·Rochak Dhakal, Chen Zhao, Soup Cobb, Xin Qian, Xiaofeng Yang +1 more
benchmark
Abstract · 20 Pages and 4 Figures

Purpose: Surgical operative reports support procedural documentation, surgical training, performance assessment, and workflow review. However, extracting meaningful information often requires extensive manual review and clinical expertise. Automated surgical scene descriptions could reduce this burden by leveraging recurring surgical activities and prior knowledge. Methods: We propose GRAC (Graph-Based Retrieval-Augmented Captioning), a framework that represents surgical scenes as graphs capturing relationships among instruments, actions, and anatomical structures. Given an input scene graph, GRAC learns to retrieve relevant graph-caption pairs from training data. It compares instrument-tissue interactions between retrieved examples and the input scene, using comparison features to condition retrieved-caption representations during generation. Results: On the EndoVis-18 benchmark, GRAC achieved a CIDEr-D of 5.88, compared with 4.23 for SGT and 3.36 for the strongest conventional Transformer baseline. Exact instrument-action-target triplet-set accuracy reached 52.6%, compared with 32.7% for SGT and 26.2% for the strongest conventional baseline. These results demonstrate improvements in language quality and factual agreement. On TORS, GRAC achieved a CIDEr-D of 5.06, compared with 3.80 for the strongest baseline. Conclusion: GRAC demonstrates that combining structured surgical scene representations with retrieved examples improves surgical scene captioning and produces more contextually meaningful descriptions.

0.26

Bootstrap Your Own AV-Proxies: Adaptive Contrastive and Prototype Learning for Audio-Visual Segmentation

cs.CV ·2026-11-06 ·Junbo Zhang, Hang Su, Zhaofan Li, Hang Dong, Chao Sun
top venue
Abstract · Accepted to CVPR 2026

Audio-Visual Segmentation (AVS) aims to accurately segment sounding objects in video frames by leveraging audio-visual correspondence cues. However, it remains challenging due to the intrinsic semantic incompleteness within a single modality and the semantic gap between audio and visual representations. Traditional feature-fusion decoding approaches struggle to suppress fusion noise effectively, while recent methods that incorporate data-dependent priors often increase the complexity of modeling audio-visual correlations, leading to poor cross-domain generalization. To address these issues, we propose a novel adaptive contrastive and prototype learning framework, BYOAVP, for AVS. Specifically, we design a Self-Supervised Audio Enhancement (SSAE) module that introduces contrastive learning to adaptively align audio representations with gradient-blocked visual embeddings, thus narrowing the semantic gap between modalities. Furthermore, a Dynamic Prototype Constraint (DPC) module is developed to refine pixel-wise category perception via momentum-based prototype updating, while enhancing localization of sounding regions through cross-modal interaction. Extensive experiments demonstrate that our method achieves SOTA performance across two AVS benchmarks and six sub-tasks, exhibiting strong robustness and generalization ability.

0.26

An Effective Spoken Language Understanding Approach Leveraging Query-Adaptive Intent Detection

cs.CL ·2026-11-06 ·Bing-Long Kuo, Yu-Hsuan Fang, Tien-Hong Lo, Hsueh-Wei Liao, Berlin Chen
published
Abstract · Accepted for oral presentation at TAAI 2026. 6 pages

Large language models enable fine-tuning-free spoken language understanding (SLU) through in-context learning, but direct query-based retrieval may select semantically similar yet intent irrelevant demonstrations. This paper presents Query-Adaptive Intent Detection via Prototype Extraction (APEX), a two-stage retrieval framework that first selects candidate intents using intent-level prototypes and then retrieves contextual exemplars within the selected intent space. Experiments on the SLURP dataset evaluate joint SLU performance using Exact Match Accuracy (EMA) under full-transcript and perplexity-stratified conditions, with additional retrieval analysis on ASR-generated transcripts. Compared with query-based retrieval, APEX improves EMA by 4.34 and 3.97 percentage points with Qwen3 4B and Gemma-3-4B, respectively, on full transcripts. Under high-perplexity inputs, APEX also yields EMA gains of 7.66 and 2.13 percentage points. Retrieval analysis further indicates that prototype-guided candidate selection achieves higher ground truth intent coverage across multiple input conditions.

0.26

RegNC-Mamba3: Shared Structural Feature Learning Using Non-Causal State Space Principles for Multi-Modal Image Registration

cs.CV ·2026-11-06 ·Zhikang Li, Yan Wu, Xin Hu, Yi Dai, Ming Li
unusual
Abstract

The main challenge of multi-modal image registration is shared structural feature extraction with higher accuracy and lower computational complexity. Compared with a State Space Model (SSM), the State Space Principles (SSP) formulation of Mamba-3 uses a more accurate discretization with lower truncation error while maintaining linear complexity. However, vanilla SSP is designed for natural language processing (NLP) and is not directly suitable for images. To address this issue, we propose a novel non-causal SSP (NC-SSP) with shared structure awareness and develop a coarse-to-fine registration method named RegNC-Mamba3. First, NC-SSP is designed to eliminate the dependence of the integral computation on the scanning path. By combining relative-weight normalization with aggregation over all circular scan paths, NC-SSP constructs a globally hidden state without long-range decay, which provides a wider receptive field and higher efficiency than the original Mamba-3. Second, NC-SSP is modified to strengthen the extraction of cross-modal structural features. When computing hidden states and feature interactions, we construct complex-valued sequences in NC-SSP by applying learnable Log-Gabor kernels, with different orientations to each head for frequency-domain modulation. This improves the perception of shared structural information across modalities and increases both registration accuracy and quality. Experiments on VIS-SAR and VIS-IR demonstrate that RegNC-Mamba3 outperforms the compared detector-based and semi-dense multi-modal registration methods.

0.26

On the Incompleteness of do-calculus for Universal Identifiability

cs.CL cs.AIcs.LGmath.PR ·2026-11-06 ·Yaroslav Kivva, Aliaksandra Novik, Negar Kiyavash
Abstract

We study the problem of causal effect identification from an arbitrary set of observational and/or interventional distributions, which we call Universal Identifiability (uID). The uID problem generalizes existing classical identifiability (ID) and general identifiability (gID) problems, which considered the identification problem from either mere observational distributions or a specific type of interventional distributions. Prior research has established graphical criteria to determine when a causal effect is identifiable in ID and gID problems. Moreover, algorithms that leverage do-calculus and basic probability manipulations have been proposed to compute these effects. Consequently, it has been shown that do-calculus is both sound and complete in these problems. Despite these well-established results, a formal definition of what it means for the do-calculus to be complete does not exist. We propose a rigorous formalization for the completeness of the do-calculus and show that the completeness results in previous work remain valid. However, for more general problems, such as uID, the rules of do-calculus are no longer complete. This is perhaps the reason why the uID problem has remained open thus far.

0.25

Policy Over Purpose: Auditing LLM Content Moderation of Reclaimed Slur Usage

cs.CL ·2026-11-06 ·Christina Chance, Kai-Wei Chang
position
Abstract

As Perspective API is deprecated, platforms are turning to large language models for content moderation, encoding policy as natural language that the model must interpret at inference time. Reclaimed slurs test this design directly. The same term can express in-group pride or target the community that reclaimed it, and misreading it silences the speech moderation is meant to protect. We audit four general-purpose LLMs and four moderation-specific systems, with Perspective API as a legacy baseline, on IYKYK, a corpus of reclaimed and derogatory usage of three slurs annotated by community members. Across the policy, decision, reasoning, disparity, and steering layers of this audit, we find that whether a system was purpose-built for moderation explains little about how it handles reclaimed language. No system reliably agrees with community judgment, and all fail most on language reclaimed within Black communities. What separates systems is instead how their policy is represented. Only systems whose policy is expressed in the prompt can be steered at all, and even then, explicit definitions change how models justify a decision more than how they make it. We argue that policy transparency determines whether moderation failures can be corrected, but that correcting them requires contextual judgment that definitions supplied at inference time cannot provide.

0.25

DINO-Nucleus: Multi-Level Self-Supervised Features for Nucleus Instance Segmentation

cs.CV ·2026-11-06 ·Mohammad Sarhangzadeh, Elyar Esmaeilzadeh, Doruk Oner
Abstract · 10 pages, 2 figures

Nucleus instance segmentation and counting are essential for quantitative microscopy, enabling reliable analysis of cell populations, tissue organization, and disease-related changes. However, accurately separating touching nuclei remains difficult, and many existing approaches rely on complex models or sensitive post-processing procedures. We present DINO-Nucleus, a parameter-efficient framework that leverages multi-level self-supervised DINOv2 representations for nucleus instance segmentation and counting. The model jointly predicts nucleus regions, pixel-wise instance flows, and explicit nucleus centers, while a consistency objective encourages the predicted flows to converge toward their corresponding centers. Multi-level transformer features provide complementary spatial and semantic information, and the pretrained backbone is adapted using lightweight LoRA modules to reduce fine-tuning cost. Experimental results show that DINO-Nucleus achieves strong segmentation and counting performance with parameter-efficient backbone adaptation. These findings demonstrate that parameter-efficient adaptation of multi-level self-supervised features can provide an effective solution for nucleus instance segmentation in microscopy images.

0.25

A Bridge from Paste to Place: Physics-Guided One-Step Image Composition via Object-Scene Interaction

cs.CV ·2026-11-06 ·Hongzhen Li, Youwei Pan, Yingfang Zhu, Sen Zhu, Fengjie Zhu +1 more
benchmarkcode/data
Abstract

Realistic image composition must make a pasted object appear naturally present in a scene while preserving the fine details of both foreground and background. Reference-based insertion methods can change object pose and appearance, while foreground harmonization methods and shadow-specific synthesis methods each address only part of this task. An efficient, unified solution to this task remains an open challenge. We introduce Paste2Place, an end-to-end, physics-guided flow-based model that learns a direct, one-step latent bridge from pasted inputs to realistic composites, guided by explicit albedo-shading-residual decomposition and recomposition within the model. To support model training and evaluation, we develop a Blender rendering and validation pipeline to generate paired composites with physical annotations, and construct real-scene pairs from open-source object-removal data. This effort yields 86,273 training pairs and a 1,500-example benchmark. Extensive qualitative and quantitative experiments demonstrate that Paste2Place achieves substantial advantages in image composition quality and inference speed. Overall, our work provides a clear task definition and an effective solution for image composition, and fills a gap in large-scale, high-quality datasets and benchmarks for training and evaluating image composition.

0.25

URL-Sim: Unified and Regulated LiDAR Simulation for High-Fidelity Reconstruction of Dynamic Driving Scenes

cs.CV ·2026-11-06 ·Kefei Qian, Xipeng Feng, Jie Li, Long Cao, Yuhan Chen +2 more
Abstract

Reconstructing high-fidelity LiDAR simulation scenes from real-world driving logs is essential for cost-effective safety validation of autonomous driving systems. While simulators based on Neural Radiance Fields (NeRF) achieve realistic scene reconstruction, their large-scale adoption is limited by slow rendering speeds. Recent 3D Gaussian splatting (3DGS) approaches have improved rendering efficiency through explicit primitive representation and differentiable rasterization. However, existing 3DGS-based methods remain largely restricted to rigid vehicles and produce redundant primitives through unconstrained densification. To address these limitations, we propose URL-Sim, a unified and regulated dynamic LiDAR simulation framework. URL-Sim constructs a dynamic Gaussian scene graph that decouples the modeling of traffic participants, including vehicles, pedestrians, and cyclists, enabling accurate reconstruction of heterogeneous dynamic objects within a single framework. It further introduces a regulated Gaussian densification strategy that ranks primitives with multi-dimensional importance scores and allocates Gaussians to the most critical regions while keeping the total primitive count bounded. Experiments on the Waymo and KITTI-360 datasets demonstrate that URL-Sim achieves higher quantitative accuracy and visual quality than existing baselines under comparable primitive budgets, and accurately recovers the geometry and poses of diverse dynamic traffic participants.

0.25

Super-resolved digital terrain models for GEDI geolocation correction: agreement with airborne LiDAR DTM-based corrections at the beam level, not the footprint level

eess.IV cs.CVphysics.geo-ph ·2026-11-06 ·Natsuki Yuda, Shinichiro Shirota
Abstract · Appendices will be added in a future version

The Global Ecosystem Dynamics Investigation (GEDI) mission measures three-dimensional forest structure with a LiDAR aboard the International Space Station. GEDI observations have geolocation errors of about 10 m, which cause misalignment with other remote sensing data and reduce the accuracy of downstream tasks. Matching GEDI ground elevations against high-resolution digital terrain models (DTMs) from airborne LiDAR is one way to correct these errors, but such DTMs exist in few regions. To widen this method's applicability, we super-resolved a more widely available low-resolution DTM and used it for the matching. Our question is whether a super-resolved DTM yields the same correction as a 1 m-resolution airborne LiDAR DTM and, if not, why. In steep forests of western Japan, we super-resolved an existing 10 m-resolution DTM to 1 m. Improving its accuracy required fusion with a global 30 m-resolution DTM from a different source. Using the resulting DTM, we corrected GEDI observation positions first by a single shift per ground track, then per observation. For most tracks, the shifts nearly matched those from the airborne LiDAR DTM, likely because DTM errors canceled out across many observations. Per-observation corrections, however, did not match. At a single observation, the error does not cancel out and is too large for the correction. This error likely originates mainly from the input DTM. Therefore, even without a 1 m-resolution DTM, a roughly 10 m-resolution DTM, fused with a global DTM and super-resolved to 1 m, should yield track-wise corrections nearly matching those from a 1 m-resolution DTM.

0.25

ROSE: Risk-Aware Orthogonal Subspace Navigation for Lifelong Knowledge Editing in Multimodal Large Language Models

cs.CL cs.AIcs.CV ·2026-11-06 ·Lingyun Song, Ziyao Chen, Kang Pan, Xiaolin Han, Yifei Zhang +3 more
code/data
Abstract

Multimodal Large Language Models (MLLMs) require continuous factual updates to stay current, yet lifelong knowledge editing remains a daunting challenge. The primary obstacle is the severe Locality Erosion and Catastrophic Forgetting, where sequential edits inevitably interfere with pre-trained capabilities and prior updates. In this paper, we propose \textbf{ROSE}, a \textbf{R}isk-aware \textbf{O}rthogonal \textbf{S}ubspace navigation framework for lifelong knowledge \textbf{E}diting. ROSE addresses forgetting by aiming to restrict parameter updates to the orthogonal complement of previously edited subspaces. Crucially, it introduces a risk-guided mechanism that utilizes a dynamically computed risk mask to safeguard specific factual parameter subspaces essential for model stability. During inference, ROSE employs a multi-granularity knowledge integration strategy, featuring prototype-anchored semantic gating and perturbation-driven subspace fusion to precisely activate edited knowledge for relevant queries while limiting interference on unrelated inputs. Comprehensive experiments on two prominent multimodal benchmarks show that ROSE is competitive with state-of-the-art approaches across five key metrics, while maintaining outstanding stability over $1,000$ sequential updates. Our code is available at https://github.com/sunlitsong/ROSE.

0.25

PathXAI: Evidence-guided structured pathology reporting by transferable tissue segmentation

cs.CV cs.CL ·2026-11-06 ·Zhaorui Tan, Jun Aun Tan, Weixiang Zhong, Xiaoxi Wang, Tan Pan +5 more
Abstract · More experimental results will be added in the future version

Pathology reporting translates tissue observations across regions and slides into diagnoses and supporting morphological descriptions, yet explicit tissue-composition measurements are rarely used to guide H&E-based report generation. We developed PathXAI, which pairs transferable tissue segmentation with a pretrained vision-language model. Because supplying tissue information directly with H&E images gave limited benefit, PathXAI-Report uses segmentation overlays and tissue statistics in separate steps, selects between candidate diagnoses by fixed rules, acquires further images when evidence is insufficient and records every step for review, without updating model weights. Its segmentation component exceeded baselines in mean Dice on internal and external tests. In automated evaluation on 1,200 public and 151 private cases, PathXAI achieved higher agreement with reference diagnoses and documented morphology than H&E-based comparators. On the public cohort, it exceeded Qwen+H&E by 4.4 percentage points in cancer family match rate and 5.7 percentage points in finding F1.

0.25

Beyond Truncation? Entailment-Licensed Elision and the Limits of Semantic Preservation in LLM Decoding

cs.CL ·2026-11-06 ·Mahmoud Alyosify, Mirna Imbabi, Eman Elkhamisi
surveytheory
Abstract · 8 pages, 3 figures, 2 tables

Instruction-tuned large language models (LLMs) answer at greater length than a request needs. We ask whether a training-free controller inside the decoding loop can remove that surplus while preserving content, judged against the simplest possible baseline: truncating the model's own answer to the same length. An entropy-gated logits processor removes 25.5-34.9% of tokens on MT-Bench development questions and 19.9-34.2% on a sealed test split within pre-registered BERTScore non-inferiority margins, yet we prove and confirm by audit that its output is always a prefix of the unmodified answer: the saving is truncation. A survey of 23 training-free mechanism families finds none that selects content better than truncation without oracle labels. We therefore introduce Entailment-Licensed In-Decoding Elision (ELIDE), which, at each sentence boundary, withholds a span from the reader only when a natural-language inference model judges it entailed by the text already shown, while leaving the model's generation unchanged. ELIDE beats matched-length truncation on three model families, but reading 936 of its elisions on 670 public prompts shows that 59% harm the answer, a failure that no automatic measure detects. Scoping the licence to its discourse unit reduces the harmful share to about one elision in ten on two fresh tests.

0.24

Compact Greek Language Models from Scratch: Pretraining 500M- and 1B-Parameter Decoder-Only Transformers

cs.CL ·2026-11-06 ·Konstantinos Perifanos, Anne Laure Poite, Giannis Tsilikas, Ilias Katis, Stavros Deligiannidis +1 more
Abstract

We present two compact decoder-only language models trained from scratch exclusively on Greek text: LexisLM-1B (1.09B parameters) and LexisLM-500M (502M parameters). Both use a dedicated 32k byte-level BPE tokenizer and a 22B-token pretraining corpus drawn primarily from Greek Wikipedia and FineWeb2. Both models complete the 22B-token schedule, with best held-out perplexities of 9.77 for LexisLM-1B and 11.31 for LexisLM-500M. Evaluation of both models on Greek-translated ARC, MMLU, and TruthfulQA shows above-chance performance on ARC-Easy, while ARC-Challenge and MMLU remain near chance. We also perform supervised fine-tuning of both models on Greek instruction data, primarily EuroAlpaca, reaching held-out instruction perplexities of 7.12 for LexisLM-500M and 6.37 for LexisLM-1B. The report describes the tokenizer, data preparation, training, and evaluation methods, providing a foundation for further work on compact Greek language models.

0.24

Frequency-consistent synthesis and selective transfer for medical image analysis

eess.IV cs.CV ·2026-11-06 ·Zhuhui Wang, Jianwei Zuo, Xin Peng, Kai Zhong, Xuliang Deng +1 more
Abstract · 40 pages, 5 figures

Complementary imaging is not always available, and realistic synthesis does not necessarily improve downstream analysis. Here we introduce a training framework linking medical image synthesis to task learning through adaptive frequency consistency and selective feature transfer. Image-specific frequency constraints encourage consistency of structure and detail. Features are transferred only when a teacher using synthetic images achieves a lower task-error score than the student by a task-specific margin on a labelled training sample. Evaluations across five datasets show improved translation fidelity with three adversarial and diffusion-based translators trained without paired supervision, alongside gains in brain-tumour segmentation on tumour-containing slices and lesion centre-point detection. All 105 reported paired comparisons yielded point estimates favouring the proposed methods, including 36 external translation comparisons on post?treatment MRI without retraining. These findings support synthetic modalities as task-oriented complements to acquired scans when complementary imaging is unavailable.

0.24

Trajectory Forcing: Exploiting Diffusion Trajectories for Autoregressive Long Video Generation

cs.CV ·2026-11-06 ·Bohan Wang, Shuo Chen, Yunzhi Li, Zhaozheng Chen, Junzhe Zhang +1 more
top venuebenchmark
Abstract · Accepted at NeurIPS 2026

Autoregressive (AR) modeling is arguably the only way to extend a fixed-length video diffusion model to longer or even streaming videos. However, existing AR diffusion methods use only the previously predicted chunk as context for the next, discarding the intermediate predictions that constitute the chunk's generation history. In contrast, conventional AR models such as LLMs condition on the entire history of preceding tokens. To this end, we propose Trajectory Forcing (TraF), an AR diffusion method that exploits the denoising trajectories of previous video chunks as context. TraF operates under two training regimes: (1) with ground-truth videos, where pseudo-trajectories are constructed from Gaussian interpolations and re-predicted by the model, and (2) with a teacher model, where real trajectories are obtained from the student's own denoising rollout. The first regime naturally provides a strong initialization for the second. On the 5-second VBench benchmark, TraF achieves a Total score of 85.15, outperforming Self Forcing (+0.84) and Causal Forcing (+1.11), with Quality and Semantic improving simultaneously. On 30-second generation---6$\times$ the training horizon---TraF achieves the highest overall quality among all compared methods.

0.24

AlienBench: A Benchmark of Structural Novelty in Language Model Outputs

cs.CL cs.AI ·2026-11-06 ·Jonas Oppenlaender, Johanna M. Silvennoinen
benchmark
Abstract · 21 pages, 2 figures, 13 tables

Large language models (LLMs) are increasingly used for creative tasks such as writing, design, and idea generation. For these uses, it matters whether a model's ideas depart from familiar examples or stay close to them. Novelty is one standard measure of creativity, and current creativity evaluations ask human experts or LLM judges to rate how novel an output is. However, these ratings have no fixed reference for what counts as novel (or, in other terms, departure from the familiar). Measuring this departure requires a task with known familiar defaults. The structured imagination task from cognitive psychology is such a task. In this creativity test, people imagine a creature from another planet. The departure from features of Earth animals, such as bilateral symmetry and legs as mode of locomotion, is used as measure of creativity. We introduce AlienBench, a benchmark that adapts the structured imagination task to language models. Models describe alien creatures, and a panel of three LLM judges extracts ten biological features from the description and labels them as either Earth-typical or a departure. The sum of feature departures is the score of a description, from 0 to 10. Each model describes alien creatures under six prompt conditions that vary the framing of the task and the planetary environment. Each model generates 30 descriptions per condition (180 in total), and its overall score is the mean across these descriptions. We evaluate 10 language models, from frontier to open-weight models. Overall scores range from 3.28 (GPT-6 Luna) to 6.55 (Ministral 3 14B). Interestingly, open-weight models score higher than proprietary models in this task (5.40 vs. 4.58). Under the baseline prompt, the models depart from Earth defaults more often on average than 100 human participants on every feature. We release all prompts, generations, judge outputs, and scores.

0.23

Steerable Multi-Objective Group-Relative Preference Optimization with Jointly Trained Low-Rank Adapters

cs.CL cs.AIcs.LG ·2026-11-06 ·Constantinos Karouzos, Nikolaos Aletras
Abstract · Preprint. Under review

Language model deployment often requires different trade-offs between competing objectives as user preferences change, yet standard post-training methods typically produce a single policy for a fixed objective weighting. We introduce LambdaMix, a new parameter efficient method for learning a preference conditioned family of policies with group-relative online reinforcement learning. LambdaMix jointly trains one low-rank adapter per objective over a frozen base model. One preference vector selects the adapter mixture that generates responses and weights the separately normalized rewards used to update the adapters. A single training run supports changing preferences at inference without further optimization. We experiment with four models on helpfulness and harmlessness and on solving math problems while following instructions. We also evaluate the two largest models on rewriting comments to preserve wording while reducing clues about personal attributes. On the first two tasks, LambdaMix achieves higher hypervolume than all evaluated conditioned and policy merging baselines across all four models. LambdaMix uses fewer training completions and stores fewer adapter parameters than a set of GRPO specialist policies. With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone.

0.23

RuACT: Recovering Inactive Rubrics as Adaptive Critics and Teachers

cs.CL ·2026-11-06 ·Yijun Li, Zhengxi Lu, Zhipiao Liu, Jinzheng Yu, Xutao Weng +2 more
code/data
Abstract · 22 pages, 10 figures, 10 tables. Code: https://github.com/ZJU-REAL/RuACT

Rubric-based reinforcement learning (RL) extends RL with verifiable rewards to open-ended tasks by scoring each response against query-specific rubrics. Under group-relative optimization, however, a rubric that every sampled response fails (unreachable) or passes (saturated) gives all responses the same outcome and no learning signal; in Qwen3-4B medical training, 60.88% of query-rubric observations are inactive. We find that the two states lack different supervision: stating an unreachable rubric in the prompt raises its satisfaction from 1.13% to 91.25%, whereas responses that all pass a saturated rubric still differ in quality. We therefore propose RuACT (Rubrics as Adaptive Critics and Teachers), which assigns each inactive rubric the role that supplies what it lacks. Rubric as Teacher uses unreachable rubrics as privileged information for reward-gated on-policy self-distillation, and Rubric as Critic decomposes saturated rubrics into child rubrics that separate current responses while sharing the parent's weight. RuACT outperforms GRPO, rubric reweighting, and rubric-guided exploration on all four medical and scientific benchmarks with Qwen3-4B and achieves the best average with Qwen3-1.7B. On initially unreachable rubrics, unguided recovery rises from 35.02% to 42.86%, while child refinement improves reward-quality alignment on independent rollouts from 57.74% to 60.77%. Code is available at https://github.com/ZJU-REAL/RuACT.

0.23

TriNetra: Thermal-Reliability-Guided RGB-Thermal Fusion and Tracking for Surveillance

cs.CV ·2026-11-06 ·Sadique Khatib, Aditya Ahirrao, Vedant Sonawane, Disha Kapse, Y. D. Bhise
benchmark
Abstract

Visible cameras fail at night and thermal sensors fail under saturation, blur or loss of contrast, yet most RGB-thermal (RGB-T) detectors mix the two streams with fixed weights and cannot tell when one modality has become unreliable. TriNetra is a reliability-aware surveillance framework built around this problem. The proposed solution computes a frame-level Thermal Reliability Coefficient (TRC) from five label-free image cues, uses it to gate the thermal features of a dual-branch YOLOv8m detector whose visible branch, neck and head are frozen and COCO-pretrained, and passes the same score to a ByteTrack-based tracker that feeds a rule-based behaviour analyser. On the M3FD benchmark (2,940/630/630 split) the detector reaches 87.6% precision, 79.4% recall, 70.0% mAP@50 and 47.8% mAP@50:95 on the test split, with 17.0 M of its 42.9 M parameters trained. TRC stays near 0.78 on clean frames (standard deviation 0.012), decreases with the severity of thermal saturation and blur, reaching 0.53 and 0.69 at full severity, and is insensitive to RGB-thermal misalignment. With identical weights, the adaptive gate matches a fixed gate on clean data (70.0% versus 69.8% mAP@50) but keeps 43.6% mAP@50 when the thermal stream is corrupted, against 38.4%, 29.2% and 25.0% for a fixed gate, full thermal admission and a closed gate; the 5.1-point gain over the fixed gate has a 95% bootstrap confidence interval of +3.6 to +6.5 points. In tracking, the TRC-gated association performs on par with ByteTrack on a synthetic benchmark with exact ground truth and on 14 M3FD clips, and is slightly less stable under thermal saturation, so its measured benefit is confined to the detector. The limitations of the current evaluation are reported together with the results.

0.23

STAIRS: Staging the Inverse Problem via Admissible-Support Scheduling for Unrolled MRI Reconstruction

cs.CV ·2026-11-06 ·Xiangjian Hou, Chang Ni, Chao Qin, Zhiwei Tan, Xin Wang +4 more
Abstract

Accelerated magnetic resonance imaging (MRI) forms images from incomplete Fourier measurements, shifting acquisition burden to computational reconstruction. Deep unrolled networks interleave learned regularization and data consistency, yet typically estimate the full unmeasured spectrum at every cascade. We propose STAIRS (Staging the Inverse Problem via Admissible-Support Scheduling), which projects intermediate reconstructions onto nested Fourier supports that expand to the full spectrum and conditions each update on its support transition. Conditional increments of a regularized log-determinant score of training covariance define a normalized spectral coordinate, accounting for dependencies between newly and previously admitted frequencies. Dyadic targets allocate larger coordinate increments to earlier cascades and are mapped to fixed supports before training, enabling coarse-to-fine reconstruction with unchanged sampling masks and data consistency. On fastMRI and Stanford2D, STAIRS improves matched Transformer (HUMUSNet) and convolutional (NAFNet) backbones in single- and multi-coil reconstruction, including radiologist-annotated pathology regions. It also outperforms predictor-based progression with lower parameter overhead and inference latency. Stage-wise analysis shows repeated error correction by projection and lower intermediate reconstruction errors than the backbones alone throughout support expansion. Ablations favor dyadic allocation over uniform expansion for square and line-band supports and over manual front-loading for square windows. Support projection alone improves reconstruction, and transition conditioning strengthens the gain. Cross-dataset evaluation shows improvements for both backbones in both directions. Support allocation thus complements regularizer design in unrolled MRI reconstruction.

0.22

How Much Conversation History Does an AI Agent Need?

cs.CL cs.AI ·2026-11-06 ·Danny Kim, Fernando Koch
benchmark
Abstract · 10 pages, 6 figures

How does using less conversational history change reader-token use and benchmark scores? We examine this question on 1,100 BEAM questions. We held the retrieval ranking and reader constant across five fixed passage budgets, and mean assigned scores rose as the passage budget grew. The score gain per additional token became smaller at the upper tested budgets. Answerable questions generally received higher scores with more context. Pooled abstention scores sometimes rose as the passage budget shrank. We also tested message retention in a controlled task with explicit conversational-state targets and compared retained user messages with token-matched recency. Qwen matched the full state more often for direct requirements when user messages were retained. Its full-state matching declined when answers depended on references to earlier offers. Mistral produced near-floor results across the retention conditions. A scoring audit leaves the correctness of BEAM answers unresolved. These experiments characterize the token and score consequences of the tested memory budgets. Reliable budget selection depends on retaining the evidence each question requires and validating the measure used to assess answer quality.

0.22

ExQuery: Explicit Structured Query Priors for Temporal 3D Object Detection

cs.CV ·2026-11-06 ·Honglin Sun, Yifei Zhang, Junyang Zhang, Kenji Hashimoto
publishedbenchmark
Abstract · BMVC 2026

Sparse query-based 3D detectors commonly encode heterogeneous semantic, geometric, and motion information into unified latent queries, including when explicit 3D anchors are used. We introduce ExQuery, an explicit structured query framework for temporal 3D object detection. Each decoded query is organized into a shared semantic context slice and physical-state-specific slices for center, dimensions, orientation, and velocity. This structure is preserved from prior encoding to branch-specific regression and is reused to reconstruct temporal queries: predicted physical states are transformed by ego motion, classification scores form soft semantic priors, and the reconstructed query is fused with the stored latent instance feature through gated affine modulation. Experiments on the nuScenes benchmark show that our method achieves competitive detection performance, and ablation studies validate the effectiveness of our proposed modules.

0.22

Edge-Deployable LLMs for Classification and Value-of-Information Ranking of Disaster-Related Tweets

cs.IR ·2026-11-06 ·Christopher D. Molek, K. Brent Venable, Niranjan Suri, Stephen Russell
Abstract · 6 pages, 6 figures, 1 table, IEEE MILCOM 2026

Humanitarian-aid and disaster relief decisions often rely on noisy, incomplete, and uncertain information from social media. Although substantial research has focused on automating the classification of such data, limited work has examined whether individual posts contain sufficient information to support decisions to act. In this study, we use large language models (LLMs) to (1) classify social media content and (2) assess its Value-of-Information (VoI) in terms of operational actionability. We evaluate seven LLMs for the classification of 77,484 tweets across 10 humanitarian-aid categories and identify the Qwen3 (8B parameters) model as the top-performing classifier, achieving the highest F1 scores in all but one category. In a second task, the classified categories serve as contextual inputs for assessing VoI with respect to responder actionability across six well established metrics for each tweet. We observed that the Gemma 3 (4B parameters) model outperformed the other models on this task, achieving the highest recall across the majority of categories. This work demonstrates that relatively small, edge-deployable LLMs can prioritize actionable information in resource-constrained environments. By integrating VoI assessment, our framework advances the automated elicitation of high-utility knowledge from unstructured social media streams for disaster-response decision-making and highlights a pathway toward tactical edge applications using LLMs that fit on a single 8 GB GPU.

0.22

PoseRieFlow: Riemannian Flow Matching for Whole-Body Text-to-Pose Generation

cs.CV ·2026-11-06 ·Jan Spindler, Bardia Ardakanian, Patrick Stotko, Reinhard Klein
Abstract

We present PoseRieFlow, a compact flow-matching framework for text-conditioned generation of whole-body 3D human poses. Prior text-to-pose methods either neglect articulated hands or model rotations in Euclidean space despite their non-Euclidean geometry. PoseRieFlow instead formulates generation directly on the product manifold of SMPL-H joint rotations, learning a text-conditioned vector field that transports samples from an isotropic $\mathrm{SO}(3)$ Gaussian prior to the data distribution along Riemannian geodesics. A part-tokenized Diffusion Transformer aggregates the 52 joints into eight body-part tokens, allocating a dedicated token and nonlinear projector to each hand, and restores per-joint resolution through part-specific decoders, shortening the sequence while improving kinematic fidelity over per-joint tokenization. With only 21.6M trainable parameters, PoseRieFlow supports expressive SMPL-H generation with detailed hand articulation at a fraction of the scale of recent LLM-based generators. On PoseScript, it attains the lowest whole-body kinematic error among all hand-producing methods, with consistent kinematic gains from the Riemannian formulation over its Euclidean counterpart at every practical inference budget, while distributional and retrieval metrics remain comparable between them.

0.21

SPARSEUP: Towards Better Recipes for Training Modern Learned Sparse Retrievers

cs.IR ·2026-11-06 ·Thibault Formal, Pierre Erbacher, Cosimo Rulli
Abstract

Learned sparse embeddings are a powerful alternative to dense retrieval, yet they have fallen behind single-vector and late interaction models. One reason for their limited adoption is that they are notoriously difficult to train, generally requiring very tight conditions to converge. We present a recipe for training Splade-like encoders on modern backbones and data that achieves a better efficiency-effectiveness trade-off. We discuss the role of logit distribution and its impact on the training stability and the final sparsity of the encoder. We also introduce a new per-position fan-out control - a hard cap on how many vocabulary dimensions each position may emit. We further apply a simple vocabulary merging strategy that removes redundant entries - such as case and word-boundary variants of the same word - which are common in cased BPE-based backbones. We apply our recipe to the LateOn backbone and training data, obtaining SPARSEUP, a sparse encoder that reaches 56.4 average nDCG@10 on BEIR-13, the highest we know of among public vocabulary-based sparse encoders below 150M parameters.

0.21

Look Closer, Align Better: Explicit Visual Attributes for Robust EEG-to-Image Decoding

cs.CV ·2026-11-06 ·Ke Xiao, Banghua Yang, Wenbin Guo, Fenqi Rong
Abstract

EEG-to-image retrieval commonly aligns neural responses with representations from pretrained vision encoders. However, these representations may underrepresent fine-grained visual details needed to distinguish semantically similar images. To address this limitation, we first investigate the EEG decodability and complementary predictive value of explicit visual attributes, identifying visual cues that capture EEG-relevant information beyond pretrained representations. Building on these findings, we propose AttrEcho, a framework that enriches pretrained visual representations with complementary, spatially aligned visual attributes. Specifically, Region-Aligned Visual Attribute Fusion (RAVF) integrates EEG-relevant local visual attributes with spatially corresponding pretrained features through a learnable signed residual gate. We further employ Feature Token Aggregation (FTA), which learns dimension-specific weights across regions to aggregate refined visual features for contrastive alignment with EEG representations. Experiments on THINGS-EEG demonstrate that AttrEcho achieves 94.8% Top-1 and 99.7% Top-5 accuracy in intra-subject 200-way retrieval, while maintaining 82.1% Top-1 accuracy as the gallery expands to 1,600 images with same-concept distractors. AttrEcho also achieves 25.3% Top-1 and 53.0% Top-5 accuracy in inter-subject retrieval. These results highlight the importance of EEG-relevant visual attributes and spatially aligned fusion for fine-grained EEG-to-image retrieval.

0.21

CodeQuant2: Enhanced Outlier Mitigation and Adaptive Clustering for Low-Precision LRMs

cs.CL ·2026-11-06 ·Xingyu Liu, Xiangyang Yin, Tianhua Xia, Haiyu Wang, Sai Qian Zhang
Abstract

Large reasoning models (LRMs) improve task performance by generating longer intermediate reasoning traces, but this substantially increases inference costs and makes low-precision quantization increasingly important. However, aggressively quantized models often perform poorly on reasoning tasks. We observe larger activation fluctuations during the first decoding steps of reasoning. Calibration that omits these early states may miss their activation fluctuations. To address these reasoning-specific quantization challenges, we propose \textbf{CodeQuant2}, a clustering-based LUT quantization framework for low-precision large reasoning models. CodeQuant2 mitigates activation outliers with rotation and scaling, uses clustering to represent non-uniform weight distributions, and adapts codebook sizes to the quantization difficulty of each layer. Guided by our observation of reasoning-time activation behavior, CodeQuant2 further improves calibration sample selection to better match actual decoding statistics. Combined with our customized LUT CUDA kernel, CodeQuant2 delivers substantial accuracy improvements on reasoning benchmarks while achieving up to $2.85\times$ speedup over BF16.

0.21

LEF-JEPA: Denoising-Trained Latent Energy Fields for Spatial Representation Learning

cs.CV cs.AI ·2026-11-06 ·Zhiyuan Liu, Zhihao Xu, Sicheng Liu, Huanhuan Huang
Abstract · 14 pages, 7 figures, 7 tables. Includes supplementary material

A joint-embedding predictive architecture (JEPA) that applies prediction supervision only at masked positions struggles to clearly capture object shapes and boundaries in its latent representations. We further observe that some existing methods improve spatial structure while also retaining rich variations in texture and color. However, preserving more low-level visual information does not necessarily make representations better suited to prediction and action. This raises a key question: how can we strengthen the spatial structure in latent representations that supports prediction and action without over-encoding these details? We introduce LEF-JEPA, which learns spatial structure through target-guided latent recovery and transfers this supervision to the visual encoder through a relative-energy constraint; the energy module is used only during pretraining and adds no inference overhead. To reduce the cost of energy optimization during pretraining, we further construct a conditional Ising Hamiltonian that allows low-energy-state search to be performed by specialized Ising solvers. On physical hardware, we validate the feasibility of achieving lower forward latency with fewer local updates. Under a unified CIFAR-10 pretraining and frozen-encoder evaluation protocol, both realizations outperform V-JEPA2.1 in image classification, foreground segmentation, and visual planning.

0.20

The APL Doesn't Fall Far From The Tree: Unsupervised Artefact Prototype Learning in Histopathology

cs.CV ·2026-11-06 ·Emilia Agasi, Charlie Gourley, C. Simon Herrington, Robert L. Hollis, T. Ian Simpson
Abstract

Histopathology foundation models provide rich representations of tissue morphology, but technical artefacts in whole slide images introduce noise that may impact diagnostic accuracy. We introduce Artefact Prototype Learning (APL), a novel approach to detecting these artefacts that exploits the geometry of pre-trained embedding spaces without manual labels or fine-tuning. Rather than treating artefacts as isolated outliers, APL computes prototypes of out-of-distribution patches and uses them to distinguish artefact-associated patterns from tissue. Evaluation on a single-centre dataset shows that APL outperforms prior methods GrandQC and HistoART in detecting artefacts such as blur, dark spots, slide edges and pen marks, but struggles with tissue folds and blood. These results show that frozen pre-trained embeddings already carry information that allows artefact detection without expensive labels or vast computational resources.

0.20

V-STORM: A Structure-Aware Grounded Chain of Thought Dataset for Faithful and Fine-Grained Spatio-Temporal Video Reasoning

cs.CV cs.AI ·2026-11-06 ·Dian Jiao, Yuqian Yuan, Jiaqi Zhu, Mingjian Gao, Siwen Tan +7 more
benchmark
Abstract · 9 Pages, 5 figures

While MLLMs can often identify what happens in a video, they still struggle with fine-grained spatio-temporal reasoning, which requires precise grounding when and where the supporting evidence appears and composing such evidence into a coherent reasoning chain. This challenge is exacerbated by the lack of scalable data construction and evaluation pipelines that provide granular, verifiable supervision linking reasoning steps to explicit spatio-temporal evidence.To address this gap, we present V-STORM, a structure-aware framework that transforms videos into Spatio-Temporal Scene Graphs (STSGs) and derives Grounded Chain of Thought (GCoT) from validated evidence paths, making each reasoning step traceable to structured and verifiable spatio-temporal evidence.Under this framework, we construct a 260K-sample training set and a 3K-sample human-verified benchmark, both featuring GCoT annotations that link reasoning to explicit spatio-temporal evidence.Experiments reveal a clear separation between evidence localization and evidence composition: current MLLMs can often identify relevant spatio-temporal cues, yet still fail to integrate them into correct multi-step reasoning, exposing a bottleneck that answer-only evaluation largely obscures. Moreover, fine-tuning with V-STORM substantially improves both answer accuracy and grounding faithfulness, validating its effectiveness as supervision for faithful and evidence-grounded video reasoning. Our dataset, benchmark, and source code will be publicly released.

0.20

I play of the cello: Evaluating Constructional Bleeding in Multilingual LLMs with CALQUE

cs.CL ·2026-11-06 ·Toshiki Nakai, Yu-Yin Hsu, Marie-Catherine de Marneffe, Joakim Nivre, Jakob Prange +4 more
Abstract

Languages use different morphosyntactic strategies to express the same meaning. Speakers of multiple languages are able to apply strategies from one language to another, either by mistake or intentionally, to produce a literal translation, or calque. Considering that large language models (LLMs) have been shown to generalize crosslingually, despite being predominantly trained on English, we investigate whether they overgeneralize constructional strategies from English to other languages. To test this, we create a dataset of 5583 sentence pairs across seven languages, matching literal translations of strategies from English into other languages, and literal translations of the other languages' strategies into English. We then quantify constructional bleeding by measuring how often strategies from English are assigned a higher probability than native strategies. We find that models often fail to assign higher probabilities to native strategies and tend to prefer English-derived strategies even when evaluating sentences in other languages. Our findings highlight the limitations of the current English-dominant training paradigm.

0.20

Improving Large Audio-Language Model Embedding Baselines for Instruction-Aware Speech Retrieval

cs.SD cs.CLeess.AS ·2026-11-06 ·Chen-An Li, Cheng-Han Chiang, Hung-yi Lee
benchmarkcode/data
Abstract · Code and artifacts will be uploaded to https://github.com/lca0503/IRIS

Speech retrieval finds relevant spoken documents given a spoken query. The spoken query alone may not reveal the retrieval goal, such as finding an answer or matching speaker identity. Instruction-aware speech retrieval expresses this goal through a natural-language text instruction. Large audio-language models (LALMs) effectively process speech and text, and our analysis shows that different retrieval requirements favor hidden states from different audio-encoder layers. We introduce IRIS-Embed, which adapts LALMs into embedding models through contrastive training and instruction-aware audio layer weighting. We construct about 900K training samples from real data and synthetic data, with instructions expressing requirements for one or more speech attributes. On INSPIRE, an instruction-aware speech retrieval benchmark, IRIS-Embed consistently outperforms the corresponding LALM baselines. The best model achieves 62.92% macro-average Recall@50, surpassing the strongest baseline at 45.15%. Controlled comparisons demonstrate instruction awareness and the benefits of audio layer weighting. On MAEB, a massive audio embedding benchmark, macro-average gains of 9.4-19.8 points over the corresponding baselines show generalization beyond speech retrieval.

0.19

AccentCanvas: Activity-Guided Bidirectional Accent Conversion

eess.AS ·2026-11-06 ·Jaehun Kim, Kang Zhang, Sung Woo Kim, Junmo Kim, Joon Son Chung
Abstract · 5 pages, 2 figures

Accent conversion for dubbing must change pronunciation while preserving content, voice, and source timing. Normalization alone cannot produce requested non-native accents, and matching total duration does not ensure local speech/pause alignment. We present AccentCanvas, a shared discrete diffusion model for normalization and L1-to-L2 conversion, trained on cross-paired shared-prompt recordings and synthetic pairs. Frame-level speech activity and repetition-preserving units guide local timing through synthesis. Across both tasks, AccentCanvas achieves the highest accent conversion performance (Accent Arrival) and mean listener-rated accent similarity among compared systems, with over 73% lower speech-boundary error than the best baseline for each task. These gains accompany higher L1-to-L2 WER than TokAN and Vevo-Style, highlighting a tradeoff between requested-accent resemblance and ASR-based content recognition. https://smokedindia.github.io/accentcanvas/

0.19

RadarDINO: Distilling a Visual Foundation Model into mmWave Radar for Human Mesh Reconstruction

cs.CV ·2026-11-06 ·Furkan Mert Algan, Zhifan Ni, Eckehard Steinbach
Abstract

Millimeter-wave (mmWave) radar offers a privacy-preserving and illumination-robust alternative for human mesh reconstruction (HMR). However, high-fidelity radar HMR remains challenging. Sparse point cloud-based methods suffer from information loss and missed detections, while raw radar tensor methods like RT-Mesh lack fine-grained semantics and explicit local body modeling. In this paper, we propose RadarDINO, a radar HMR framework that distills visual semantics from a DINOv3 teacher into 3D radar representations and introduces explicit geometric supervision. DINOv3 descriptors are transferred to radar voxels via calibrated RGB-radar geometry, providing voxel- and body-part-level supervision for part-aware radar representations. We further incorporate vertex- and joint-losses to directly constrain reconstructed geometry. On the M4Human dataset, our method outperforms the SOTA baseline across all three evaluation splits.

0.19

Decoding spatial and semantic information from large-scale visual cortical activity for natural-movie reconstruction

cs.CV q-bio.NC ·2026-11-06 ·Yukai Huang, Haoyu Albert Wang, Jiawei Zhang, Jialun Ma, Jiayao Ding +4 more
Abstract · 24 pages, 6 figures, 2 tables; 11-page supplementary information included as ancillary file

Reconstructing natural visual experience from cortical population activity is a central challenge in neural decoding, particularly for calcium imaging, where signals are noisy, temporally filtered, and paired neural-video data are limited. Here we present Ca2REAL (Calcium Activity to Reconstruct Events, Appearance, and Layout), a neural-to-video framework that conditions a pretrained video diffusion model through separate layout and semantic controls derived from calcium activity. Using measured responses from 7,863 V1 neurons in one mouse, Ca2REAL reconstructed held-out natural-movie clips with 81.4% all-pairs stimulus-identification accuracy, with its clearest gains over deterministic baselines in perceptual and semantic measures. We further evaluated the framework in a stimulus-aligned virtual population of 116,705 neurons assembled from encoding-model-predicted responses across 11 independently recorded datasets. Perturbation and ablation analyses showed that reconstruction depends on distributed, stimulus-specific population structure, with layout and semantic pathways providing complementary spatial and identity information. Temporal analyses further showed improved preservation of semantic dynamics across reconstructed clips. These results show that large-scale cortical population activity constrains the spatial, semantic, and temporal structure of natural visual experience, providing a generative framework for high-fidelity neural reconstruction of dynamic visual scenes.

0.18

FANO: Anchored Forward-Noise Optimization for Diffusion Inverse Problems

cs.CV ·2026-11-06 ·Shilong Han, Hongxia Wang
Abstract · 26 pages, 9 figures, 8 tables. Preprint, under review

Pre-trained diffusion models have become a widely adopted zero-shot prior for image inverse problems, among which annealed decoupling methods achieve leading reconstruction performance on nonlinear tasks. We find, however, that annealed methods that directly optimize the denoiser input $x_t$ exhibit radial drift in later layers: within-layer optimization drives the noise-component norm of $x_t$ away from its typical value, pushing the denoiser out of its high-performance regime and degrading reconstruction quality. To address this, we propose FANO (Anchored Forward-Noise Optimization). At each annealing layer, FANO fixes the previous denoiser estimate as an anchor and optimizes only the forward noise $e$, whose distribution is known, instead of $x_t$, thereby preserving the forward-diffusion structure of the denoiser input. The standard Gaussian prior on $e$ concentrates its norm near the typical value, and deviations degrade denoiser performance. We accordingly decompose $e$ in polar coordinates into a direction and a radial magnitude, and apply a soft potential to regulate the radius, keeping $x_t$ in the high-density region of the forward distribution. On multiple linear and nonlinear inverse problems, FANO matches the strongest current baselines on linear tasks and achieves consistent improvements on nonlinear tasks.

0.18

Same Pixels, Different Documents: Measuring Part-Level Editability in Image Vectorization

cs.CV cs.GR ·2026-11-06 ·Boheng Zhu, Zehua Wang, Wei Chen, Ming-Hao Lin
Abstract · 11 pages, 2 figures, 4 tables

Image vectorization is often evaluated by rendering fidelity, which cannot distinguish documents that render identically but support different edits. Edit-based evaluation tests outcomes, but an oracle that gathers a part's fragments can conceal the cost of selecting them. We introduce Pass@k, which tests whether a prescribed edit can succeed using objects selectable within k clicks under a specified interaction model. Eight document diagnostics characterize properties relevant to editing beyond these tasks. We assess structural sensitivity through controlled perturbations, practical relevance through designer operations, and reference dependence through alternative drawings. We evaluate twelve open and commercial methods on 211 references, with 635 edit tasks on 93 of them. In each of four reference sets, including two drawn by designers, 64-88% of oracle-passed tasks on parts composed of multiple shapes fail at Pass@1, against 6-40% for single shapes, pooling counts over configurations. Higher fidelity does not guarantee higher Pass@1, while selected diagnostics are associated with designers' effort in our study.

0.18

Syn2DRE: Synergistic Dual Teacher-Student Framework for Document-level Relation Extraction

cs.CL cs.AI ·2026-11-06 ·Khai Phan Tran, Xue Li
Abstract

Document-level Relation Extraction (DocRE) aims to identify relations between entities across a document but typically relies on costly large-scale annotations. While pseudo-labelling enables the use of low-quality data, existing methods often suffer from inefficiency and confirmation bias. To address these challenges, we propose Syn2DRE, a semi-supervised framework that combines self-supervised learning with cross-supervision in a lightweight teacher-student architecture to produce more reliable pseudo-labels. Additionally, we introduce a unified end-to-end training strategy to further improve efficiency of the framework. Experiments on multiple benchmarks show that Syn2DRE consistently outperforms prior state-of-the-art semi-supervised DocRE methods.

0.18

DUET: Dual-Sparse Inference for Block Diffusion Language Models

cs.CL ·2026-11-06 ·Siyuan Yan, Shen Gu, Xiaoxing Ma, Chun Cao, Jingwei Xu
Abstract

Diffusion language models (DLMs) enable parallel token generation through iterative denoising, yet their inference remains costly due to repeated attention over a growing historical prefix and repeated processing of the entire current generation block. Historical attention becomes increasingly expensive as generation progresses, while all current-block positions are propagated through the Transformer at every denoising step even though only a small fraction are typically resolved at that step. In this paper, we introduce Duet, a training-free dual-sparse inference framework that addresses these challenges through prefix Key-Value (KV) sparsity and early query sparsity, respectively. Specifically, Duet exploits the concentration and cross-step stability of historical attention within each generation block to construct a compact historical KV subset that is reused across denoising steps. For query sparsity, Duet predicts update positions from shallow representations and propagates only the corresponding queries through the suffix Transformer layers, avoiding deep computation at positions unlikely to be updated in the current step. Experiments show that Duet maintains competitive accuracy while achieving 1.34$\times$/2.83$\times$ end-to-end/per-step speedups on LLaDA2.1-mini and 2.41$\times$/4.72$\times$ on SDAR-8B-Chat-b32 at a 32K context length.

0.18

Video2Asset: Online Construction of Persistent 3D Street-Scene Assets

cs.CV ·2026-11-06 ·Zihan Zheng, Chunlei Cai, Zhenlong Wu, Linjun Li, Yichen Gong +2 more
Abstract

Online 3D street-scene asset construction turns multi-view video into a continually updated scene of independently editable objects. The central challenge is to accumulate partial, intermittent observations under motion and occlusion without fragmenting identities or mixing evidence across objects. To address this challenge, we propose Video2Asset, an online framework for constructing persistent 3D assets from multi-view video. Using occupancy as an intermediate representation, we construct identity-aware observations with Historical Carrier Guidance for instance decoding and an identity encoder fusing instance-aligned 3D features with multi-view appearance. Accumulated-Evidence Association compares current observations with identity, geometric, and appearance evidence accumulated within asset records to determine identity inheritance and update world-coordinate states. We instantiate meshes from an asset library using these states and combine them with an occupancy-derived static background to support scene visualization, mesh replacement, object deletion, and trajectory editing. Experiments on several datasets demonstrate improvements in temporal identity consistency and asset quality across multiple occupancy models, including a 15.4% reduction in asset fragmentation on nuScenes with SGE.

0.18

Multilingual Artificial Datasets for AI-Driven Cardiology: the CliTe and ParaCliTe corpora

cs.CL ·2026-11-06 ·Lorenzo Pratesi, Eulalia Farre-Maduell, Machteld Boonstra, Michele Miranda, Jonas Mack +19 more
Abstract

Background and Aims. Comparable, multilingual, validated cardiology corpora are critical to advance trustworthy natural language processing (NLP) models for cardiology but such resources are scarce due to privacy concerns. This study aimed to cover this gap by presenting suitable artificial corpus creation guidelines and by releasing 3 multilanguage corpora. Methods. Following purpose-specific publicly-available corpus construction guidelines, cardiologists from 7 European hospitals generated artificial clinical records in Czech, Dutch, English, Italian, Romanian, Spanish and Swedish, ensuring clinical diversity and sex balance, resulting in the CliTe corpus. A CliTe subset underwent a multi-stage manual translation into the other 6 languages, resulting in the parallel ParaCliTe corpus, while the remaining reports were automatically translated, constituting ParaCliTe Silver. Results. CliTe is a collection of 699 artificial cardiology reports, mostly outpatient notes (340 documents), discharge reports (263), and inpatient notes (89), describing fictional patients with ischaemic heart disease (199 patients), heart failure (188), and valvular disease (79), among others. ParaCliTe contains reports from 94 fictitious patients, each available in seven languages with sentence-level alignment, resulting in 658 documents. The corpus mainly comprises discharge reports (322 documents), and outpatient (196) and inpatient (140) notes, describing patients with heart failure (399 documents), ischaemic heart disease (287), and valvular disease (238). Empirical results illustrate the utility of ParaCliTe and highlight the limitations of state-of-the-art NLP models adapted to the cardiology domain. Conclusions. The creation of multilingual, equitable and realistic corpora of artificial patients capable of strengthening NLP performance for cardiology is feasible when supported by detailed guidelines.

0.18

Beyond Frame-Level Accuracy: Analyzing Video-Level Stability in Transformer-Based Deepfake Forensics

cs.CV cs.AIcs.LG ·2026-11-06 ·Syed Azmul Hasan, Mowmita Pervin Hera, Afroza Lona Mim, M M Jihad Hossain, Hazrat Ali +1 more
published
Abstract · Paper submitted to international journal of intelligent systems, Wiley publishers, accepted version of open-access paper

Deepfake videos generated by advanced generative models pose serious threats to digital security, misinformation control, and media forensics. Although numerous detection methods have been proposed, their reported performances are often based on heterogeneous preprocessing pipelines, inconsistent data splits, and model-specific training protocols, hindering fair architectural comparison. This work presents a controlled and reproducible comparative study of convolutional neural networks (ResNet-18 and EfficientNet-B0) and a transformer-based architecture (ViT-B/16) within a unified, face-centric deepfake detection framework. All models are trained using identical preprocessing, augmentation, optimization settings, and data splits on Celeb-DF v2 and FaceForensics++, and results are validated across three independent random seeds with bootstrapped confidence intervals. Experimental results show that while CNN-based models perform competitively at the frame level, their performance degrades after video-level aggregation. In contrast, ViT-B/16 demonstrates more stable video-level performance, achieving 92.8\% [95\% CI: 91.7--94.0] video-level accuracy and an AUC of 0.950 [95\% CI: 0.934--0.964] on Celeb-DF v2 (3-seed bootstrapped). Cross-dataset evaluation across the same three random seeds reveals a large, asymmetric generalization gap (AUC 0.79 FF++$\rightarrow$Celeb-DF vs. 0.64 Celeb-DF$\rightarrow$FF++), indicating that in-domain video-level stability does not transfer to unseen manipulation domains. These findings indicate that global self-attention improves robustness and video-level stability for deepfake detection within a given data domain, while highlighting the need for further work on cross-domain generalization.

0.18

Contrastive fine-tuning against stain and scanner bias improves robustness and downstream performance across pathology foundation models

cs.CV ·2026-11-06 ·Ryan Kim, Hassan Muhammad, Anish Dulal, Paul S. Scotti, Tanishq Mathew Abraham
code/data
Abstract · ASCI Workshop, Neurips 2026

Histopathology foundation models achieve strong performance across diverse downstream tasks, yet their representations have been shown to encode scanner- and site-specific biases, creating generalization gaps and potentially limiting real-world utility. We introduce SPECTRA (Scanner and Pathology Embedding Contrastive Tuning for Robust Alignment), a model-agnostic contrastive learning post-training framework which explicitly encourages scanner- and stain-robust representations, to optimize performance and generalization on downstream applications. Using parameter-efficient LoRA adapters trained on matched multi-scanner and multi-stain whole-slide images sourced from the PLISM dataset, our method aligns spatially corresponding regions across acquisition conditions (scanner, stain), using other locations within each candidate condition as negatives. SPECTRA improved PathoROB robustness index, gene-expression prediction on HEST, slide-level classification AUC on CPTAC, and held-out PLISM retrieval for every model. Overall tile-level classification on THUNDER also improved, with exceptions for segmentation and calibration tasks. We open-source our training and evaluation codebases and release adapters and model weights for selected backbones.

0.18

SAS: Scene-Annotated AudioSet Strong for Joint Scene-Event Analysis

cs.SD ·2026-11-06 ·Michel Olvera, Paraskevas Stamatiadis, Romain Serizel, Slim Essid, Gaël Richard
Abstract

Context is central to human auditory perception, as listeners interpret sounds within their acoustic environment to resolve ambiguity. However, large-scale datasets that jointly capture sound events and acoustic scenes remain limited. We address this by augmenting AudioSet Strong with clip-level scene annotations under two taxonomies for joint sceneevent modeling. A corpus-scale analysis shows that scene identity is better captured by events distinctive to a scene than by those that merely occur frequently. Reweighting event co-occurrences toward scene-specific associations produces representations that better align with human-derived taxonomies, while also exposing scenes where the acoustic and semantic views diverge. This geometric divergence predicts classification errors, while supervised validation confirms that the proposed scene annotations provide meaningful complementary information for developing scalable context-aware machine listening systems.

0.18

Domain Generalization for Style-Debiased Object Detection via Label-Consistent Training across Multiple Styles

cs.CV ·2026-11-06 ·Chaoran Zhang, Shirshak Bhusal, Zhilin Gao, Qiang Xue, Haoting Shen
Abstract

Domain generalization (DG) for object detection aims to improve model performance on unseen target domains. Existing detectors, however, suffer substantial degradation on visually distinct unseen domains, as models confined to limited training styles tend to overfit superficial statistical cues rather than object-level semantics. Motivated by human perception, which attends primarily to semantically relevant features while ignoring irrelevant style cues, we propose Origin-Anchored Multi-Style Consistency (OAMSC), a source-only framework that couples multi-style data synthesis with cross-style consistency learning to suppress style-sensitive shortcut learning. For data preparation, we leverage off-the-shelf diffusion models to expand the source datasets into 20 distinct styles, with all generated images manually verified to ensure that object instances and annotations remain consistent across style variants. During training, we impose Origin-anchored consistency constraints at scene and object levels to reinforce the learning of domain-stable features while suppressing the influence of style-sensitive cues. Extensive experiments and ablation studies on multiple challenging DG benchmarks demonstrate the consistent effectiveness and strong generalization performance of OAMSC across diverse domain shifts. Notably, our approach applies to both one-stage and two-stage detectors, while its data-centric design is complementary in principle to other DG methods. The training code and two style-transfer datasets constructed for the experiments, comprising approximately 180K images in total, are publicly available on GitHub.

0.17

M5-TTS: A Mini-Scale Multilingual and Multi-Dialect Text-to-Speech Language Model with Multi-Token Prediction

eess.AS cs.SD ·2026-11-06 ·Wenhao Li, Jingbin Hu, Shengming Ling, Xiaming Ren, Haoyu Zhang +18 more
Abstract · 16 pages,3 figures

We present M5-TTS, a compact text-to-speech system supporting multilingual and Chinese multidialect synthesis, zero-shot voice cloning, speaker-specific multilingual generation, and CPU streaming. Its 50M-parameter language model, excluding the speech codec, combines a temporal backbone with a local multi-token prediction module over 16 codebooks. GRPO post-training combines global speaker-similarity and pause feedback with a token-level CER process advantage. Character-level error comparisons provide local feedback, with negative credit propagated backward within a clause and projected onto audio-token positions. M5-TTS achieves 1.37% CER on Seed-ZH and 2.91% WER on Seed-EN, corresponding to relative error reductions of 62.2% and 35.3% over MOSS-TTS-Nano with half the language-model parameters. It also achieves lower content error rates in five of six evaluated MiniMax languages, while difficult Chinese inputs remain a limitation. On four Intel Xeon Gold 6530 cores, it achieves a real-time factor of 0.653 and a warm first-chunk latency of 225ms. A 10-listener evaluation gives M5-TTS higher speaker-similarity and intelligibility ratings than MOSS-TTS-Nano, with a lower naturalness rating. We will release training, post-training, and inference code, together with voice-cloning and speaker-specific checkpoints, to support reproducible research on compact speech generation for on-device voice interaction.

0.17

Using an LLM to Tackle Difficult Cases in NER for Nepali Tweets

cs.CL ·2026-11-06 ·Claire Wang
Abstract

Named entity recognition (NER) is challenging for social media data in a low-resource language. LLMs often lag behind pretrained language models (PLMs) in NER performance, but they show promise in handling cases that PLMs struggle with. Using a dataset of Nepali tweets, we split the test set into "easy" and "hard" cases using token probabilities from a fine-tuned pretrained language model (PLM). We then evaluate an LLM on the hard test cases using 16 different prompt configurations. We find that selecting examples based on semantic similarity and increasing the number of examples improves F1 score by almost 10 points from the baseline configuration. Although the LLM underperforms the PLM on the full and easy test sets, it surpasses the PLM on the hard test set, resulting in an improved score when the two models are combined.

0.17

Multimodal Training with Knowledge Traction for Video Anomaly Detection

cs.CV cs.AI ·2026-11-06 ·Xiao Tan
Abstract · 15 pages, 10 figures, 8 tables

The task of weakly supervised video anomaly detection is challenging. Existing approaches mainly model video frame features and determine whether anomalies are present based on their magnitudes. In this work, we propose a novel method that incorporates external knowledge into training to improve model performance, robustness, and generalization. We leverage textual prompts and a large language model (LLM) to establish prior information for the specific task, and employ dictionary learning to normalize and compress this information. Moreover, during experiments, we observed inconsistencies in the textual descriptions extracted from similar video frames by image caption models. To address this, we devise a method that leverages an LLM and auxiliary information to ensure consistency in the textual descriptions of similar frames within a video. Finally, we introduce a feature processing approach that incorporates prior knowledge and cross-modal attention mechanisms. This approach learns the relationships and characteristics among different modalities to construct more accurate and reliable feature representations, thereby improving the accuracy and stability of downstream tasks. Experimental results demonstrate that our method surpasses other state-of-the-art approaches.

0.17

Direct Regression and Geometric Recovery for Event-based Spacecraft Pose Estimation

cs.CV ·2026-11-06 ·Pruthvi Nandan Janga, Kashyap Hegde Kota, Bharatesh Chakravarthi
Abstract

Event cameras offer high temporal resolution and dynamic range for uncooperative spacecraft pose estimation, but models trained on synthetic event streams must generalize to the noise, background activity, and contrast variation of physical sensors. We present a comparative systems study of two independently developed approaches to synthetic-to-real, event-based 6-DoF spacecraft pose estimation. The first is a modular pipeline that combines learned keypoint detection and unlabeled target-domain adaptation with geometry-based pose recovery and temporal smoothing. The second directly regresses translation and rotation from global and cropped event representations using an apparent-scale cue, physics-motivated synthetic corruptions, and feature-consistency regularization. We evaluate both systems under the same SPARK 2026 hidden-test protocol on the real-event split of SPADES, with predictions for all test frames. The results reveal a component-wise reversal concealed by the aggregate challenge score: direct regression reduces orientation error by 22.7% (1.1767 to 0.9101 rad) and obtains a 13.4% lower composite score, whereas the modular pipeline reduces relative translation error by 78.7% (0.1335 to 0.0285). Archived configurations and a qualitative failure case identify plausible explanations for these complementary error profiles, without isolating their causes. These findings motivate a testable hybrid design that combines direct orientation estimation with geometry-constrained translation recovery.

0.17

Language as Trait, Politeness as Strategy: Controlled Tests of Linguistic Effects in LLM Social-Deduction Agents

cs.CL cs.AI ·2026-11-06 ·Yuchen Li, Lisa Liu, Runqi Zou, Shuxuan Kuang, Qingquan Zhang +1 more
Abstract

Large language model (LLM) agents coordinate through conversation, and the same message can be phrased in simpler or more complex language, bluntly or politely. We test whether such differences change how other agents judge and act, using the LURK Werewolf framework, which separates each utterance's speech-act predicate from its natural-language realization. We rewrite only the realization, check that the game content is preserved, and replay each game day from saved checkpoints against a re-sampled control. We vary linguistic range consistently across a speaker's messages and politeness strategy per utterance. Across 30 games and 4,335 replays, neither manipulation changes the trust other agents place in the speaker, and no effect on votes or night actions survives correction for multiple comparisons; the largest uncorrected effect is that shifting range away from a neutral paraphrase changes about 4 percentage points more of the other players' votes. In natural play, 88% of face-threatening speech is bald on-record regardless of role.

0.16

ZeoBench: A Benchmark for Self-Supervised Learning on 3D Zeolite Representations

cs.CV ·2026-11-06 ·Aaron Sun, Yachan Liu, Ping Yang, Peng Bai, Subhransu Maji
publishedbenchmark
Abstract · NeurIPS 2026 Datasets and Benchmarks Track

Zeolites are an important class of porous crystalline materials with broad applications in gas storage, separations, and catalysis. Although hundreds of thousands of experimentally realized and hypothetical zeolite structures are known, identifying materials optimized for a target application remains challenging because property labels are expensive to obtain and available only for limited structure-molecule pairs. While prior work has demonstrated that 3D neural representations can predict zeolite properties effectively, it remains unclear which representation-learning strategies are most effective in the low-data regime relevant to materials discovery. We present ZeoBench, a benchmark for self-supervised learning on dense 3D representations of all-silica zeolites, and evaluate the resulting features on six existing downstream adsorption-property prediction tasks, across a range of training set sizes. Our study compares self-supervised 3D pretraining on volumetric zeolite data, adaptations of image-pretrained 2D ConvNets and vision transformers to 3D via multiview and channel-inflation strategies, and 3D ConvNets pretrained on out-of-domain volumetric datasets, alongside hand-crafted descriptors and supervised baselines. Using our pretraining and evaluation framework, we find self-supervised 3D representations outperform existing methods and achieve state-of-the-art performance across both in-domain label regimes and for the unseen domain of metal organic frameworks. Learned representations consistently outperform hand-crafted descriptors, and convolutional architectures are generally more effective than vision transformers in this setting. These results establish ZeoBench as a practical benchmark for 3D representation learning in zeolites and provide guidance for data-limited adsorption property prediction.

0.16

AISHELL8-FISHEYE: A Fisheye Audio-Visual Dataset for Target Speaker Extraction with Distortion-Aware Baselines

eess.AS ·2026-11-06 ·Peijun Yang, Zhan Jin, Juan Liu, Hui Bu, Xiaoyi Qin +3 more
Abstract

Fisheye cameras offer wide visual coverage but distort facial and lip geometry, challenging audio-visual target speaker extraction. We introduce AISHELL8-FISHEYE, a 47.03-hour corpus of 94 Mandarin speakers with synchronized audio and paired fisheye and normal-camera videos. Speaker-disjoint splits and synthetic two-speaker mixtures support reproducible evaluation under matched acoustic conditions. We also establish a baseline that combines a lip-reading encoder with a frozen distortion-aware branch transferred from fisheye head pose estimation. On fisheye video, adding this branch to an adapted lip encoder improves scale-invariant signal-to-distortion ratio improvement from 10.47 to 11.84 dB. The branch does not improve performance with a frozen lip encoder, highlighting the importance of visual adaptation. A substantial gap to paired normal-camera inputs remains. The corpus and baseline resources support further research on speech extraction from distorted visual cues.

0.16

Impact of color in computer and human-based object recognition under occlusion

cs.CV ·2026-11-06 ·Courtney M. King, Elissa Aminoff, Daniel D. Leeds
Abstract

Partial occlusions challenge object recognition in both human and computer vision (CV), but the impact is largely dependent on the occlusion and model design. Properties relating to color are particularly important in biological recognition and frequently leveraged for camouflage, however there are still open questions about what makes them effective and how they translate to CV. This work studies partial occlusions that create camouflage, including those that blend in with target objects or with scenes, which alter object representations in distinct ways. Experiments evaluate recognition performance on both human participants and models ranging from convolutional neural networks (CNNs) to transformers, and additional analysis is performed on models to uncover trends in processing and object representations under multiple occlusion types. Novel metrics offer explainability of model behavior under occlusion: non-occlusion precision (NOP) and non-occlusion recall (NOR) measure alignment of model activations with and without occlusion, and activation in occlusion (AIO) measures the percentage of activations that occur in occluded regions. Findings show that human recognition, and most transformers, are negatively impacted by generated camouflaging occluders, with scene-based camouflage performing worse than object-based. Classical CNNs show a strong opposite trend, performing the worst on neutral-colored occlusions, suggesting that contextually-related colors, particularly scene-based, benefit object recognition rather than disrupting shape or boundary cues. Additionally, generated camouflaging occluders for training augmentation improves robustness to unseen camouflage for selected CNN model, and occlusion metrics suggest trends about learned mechanisms. Findings have interesting implications for alignment in human and computer vision, model vulnerabilities, and explainability.

0.15

SPSC: Zero-Shot Cross-Subject EEG-to-Image Reconstruction via Semantic Preservation and Subspace Calibration

cs.CV cs.AI ·2026-11-06 ·Yutang Lu, Shuai Jiang, Fei Su, Zhicheng Zhao
Abstract

Pretrained vision-language models and diffusion generators have driven rapid progress in brain-to-image generation, while EEG-based reconstruction is still constrained by low signal-to-noise ratio and cross-subject semantic shifts. Zero-shot cross-subject reconstruction poses even greater challenges: EEG semantic embeddings learned from source subjects deviate from the manifold of image embeddings of the unseen target subject, particularly when neither target labels nor target image embeddings are available. To address this issue, we introduce the Semantic Preservation and Subspace Calibration (SPSC) framework, which first uses Object Mask Contrastive Regularization (OMCR) with object identity for backend training, and then constructs Subspace Candidate Gate (SCG) to achieve label-free semantic correction for target subjects during inference. Specifically, OMCR regularizes the conditioning space with object-level positive samples to preserve object-level feature neighborhoods within the reconstruction prior and low-level branch, while SCG exploits training image prototypes and batch statistics of unlabeled target EEG to select candidates for each task within subspaces of image prototypes. Under the leave-one-subject-out (LOSO) setting across ten subjects on the THINGS-EEG dataset, SPSC improves the cross-subject reconstruction from 0.552 to 0.713 on CLIP and from 0.612 to 0.745 on Alex.5. These significant gains indicate that the semantically calibrated conditions can better support generation for the target subjects under label-free inference.

0.15

MotionCore: A Latent Neural Simulator for Video Generation with Rigid-Body Dynamics

cs.CV ·2026-11-06 ·Zheng Wang, Zhengwei Gong, Yuanzhi Liang, Haibin Huang, Chi Zhang
Abstract

Video generation models can produce highly realistic videos while generating physically inconsistent object motion. This limitation suggests that visual modeling alone may not provide a transferable representation of physical dynamics. We introduce MotionCore, a framework for learning physical dynamics via a learnable partial differential equation solver and focuses on rigid-body motion. We introduce a latent neural simulator learning to predict the evolution of object-level physical states from observed motion. Its predictions are then converted into spatially aligned latent controls for a controllable video diffusion model. The generator serves as a visual interface: it provides appearance and scene detail, while the learned dynamics model specifies the physical trajectory. To preserve this trajectory through the interface, MotionCore combines a diffusion-matching objective with a low-capacity decodability constraint on object identity and position. We evaluate MotionCore on public physical-consistency benchmarks and on paired data with known trajectories. Results indicate that MotionCore improves physical consistency and trajectory obedience in the evaluated rigid-body settings, while maintaining comparable visual quality. A controlled ablation protocol compares learned dynamics with an analytic prior and studies the choice of latent evolution processor.

0.15

Retrofitting Unbounded Context onto Stock Language Models

cs.CL cs.LG ·2026-11-06 ·Mark Oskin
Abstract

A language model that never had to forget would keep every document it has read and every turn it has spoken, and answer from any of it. This paper shows that two stock models can do that on unmodified weights by manipulating their geometry. Rotary attention turns each pair of key dimensions, a plane, at its own rate. The slow planes turn once in thousands of tokens and carry the address a query matches when it reads a far key; the fast planes turn once in a few tokens, and the order of the tokens inside a span is already written into its keys and values by the time the span leaves the near window. At distance the position blurs, and what attention reads is a bag rather than a sequence. A far span is read by its words: under competition it is read only if it is visible at the query of the question's own handle word, and a read through a description rather than a word fails at distance in every model measured, a 70B included. These facts allow a trick. What sits in the far part of attention does not have to be the history as it was; it can be installed ahead of the prediction. The design keeps a near window of 8K tokens at their true positions and changes what lies beyond it with the prompt: every far key and value goes to the host exact, a word index finds the units the request's rare words name, a far window of slots inside the trained range holds them while the request is encoded, and a rare word the model generates fetches its units mid-generation with one exact re-encode. Implemented in llama.cpp on Mistral-7B and Llama-3.1-8B, the design reads planted single needles at 99 to 100 of 100 and three-key discrimination at 88 to 100 of 100 from 32K to one million tokens, scores 70.8 on RULER against 69.9 for dense attention on Llama-3.1-8B at 128K and 67.8 against 75.5 on Mistral-7B at 32K, and at 128K prefills at 1.85 times and decodes at 1.65 to 1.68 times stock speed.

0.14

CXR-CMFA: A Cross-Modal Feature Aggregation Network for Chest X-Ray Pathology Classification

cs.CV cs.AI ·2026-11-06 ·Talukder Hasnat Zadid, Jahin Sultana, Felipe Bochini Brisola, Kishore Roy, Mysha Rahman +2 more
benchmark
Abstract

Chest X-ray (CXR) interpretation remains challenging because subtle pathology can be obscured by overlapping anatomy, image-quality variation, and limited visual evidence in isolated radiographs. Although image-only deep learning models provide useful baselines, they discard the complementary diagnostic context available in radiology reports, while large vision-language models can be computationally expensive for practical deployment. This paper presents CXR-CMFA, a lightweight cross-modal feature aggregation network for binary CXR pathology classification. The proposed framework combines a dual-stream visual encoder, using EfficientNet-B0 for global anatomical representation and DenseNet-121 for local pathological detail, with a TF-IDF-MLP report encoder. Two fusion modules integrate these modalities: the Dual-Modal Spatial Attention Module (DMSAM) uses the report embedding to gate global and local visual features, and the Cross-Modal Gated Feature Fusion (CGFF) module learns per-instance weights for visual and textual contributions. Experiments on the MIMIC-CXR benchmark, using the provided train, validation, and test splits, show that CXR-CMFA achieves 93.45% accuracy, 99.88% precision, 91.8% recall, 95.7% F1-score, and 99.19% AUC on the held-out test set. The model produces only four false positives among 984 normal test cases, supporting its suitability for high-specificity screening workflows. An ablation study further demonstrates a 22.47 percentage-point AUC improvement over the global-image-only variant, confirming the value of adaptive cross-modal fusion. These results indicate that compact report-conditioned fusion can provide strong discriminative performance without relying on large pretrained language models.

0.14

GTCRN-SS: Lightweight Causal Speech Separation with Residual Allocation and Reconstruction Refinement

eess.AS cs.SD ·2026-11-06 ·Shao-Chun Hu, Zi-Xiang Lin, Hsin-Min Wang, Jeih-Weih Hung, Hung-Shin Lee
published
Abstract · Accepted to ROCLING 2026

Low-complexity causal speech separation must preserve speaker information and allocate residual mixture energy without relying on future feature frames. We present GTCRN-SS, a two-stage time-frequency separator that adapts the compact GTCRN enhancement backbone to fixed-output speech separation. The model combines internal residual allocation, reconstruction-conditioned refinement, and frame-aligned recurrent-state transfer. Its complex-mask estimator assigns non-target content to an unreturned residual branch, while the complementary speech estimate conditions a second stage whose final masks remain anchored to the original mixture. A lightweight low-rank adapter transfers causal recurrent evidence between recursive executions without adding another recurrent memory network. The neural graph excludes future feature frames and incurs only the fixed analysis look-ahead of the centered short-time Fourier transform. On noisy Libri2Mix, GTCRN-SS achieves 10.1 dB SDRi and 9.5 dB SI-SNRi with fewer than one million parameters and a low computational footprint. Independently trained ablations confirm the benefits of recurrent-state transfer and reconstruction conditioning, and corpus-level diagnostics show that the internal branch primarily captures residual rather than target-speech content. These results establish GTCRN-SS as an effective quality-efficiency operating point for causal monaural speech separation.

0.14

AR-Edit: Training-Free Autoregressive Video Editing without Inversion

cs.CV ·2026-11-06 ·Hovhannes Margaryan, Quentin Bammey, Christian Sandor, Vicky Kalogeiton
Abstract

Editing a video stream in real time, without access to future frames, is essential for interactive applications. Existing training-free video editing methods, however, assume full offline access to the entire sequence, while streaming approaches either rely on additional training or are limited to specific editing techniques. We show that this limitation can be removed by analyzing the self-attention dynamics of DMD-distilled autoregressive (AR) video diffusion models. We uncover a key asymmetry: queries and keys remain stable and aligned with their source counterparts across denoising timesteps, while values diverge and encode the evolving content. Inherited from the non-distilled bidirectional teacher, this asymmetry is strongest at high noise, where few-step streaming models operate. The resulting mechanism is simple: a single forward pass suffices to capture source identity in the value features, which can be reused throughout generation. Starting from pure Gaussian noise and injecting only these source values reconstructs the input with fidelity matching or surpassing inversion-based methods, effectively eliminating inversion. Based on this insight, we introduce AR-Edit, a training-free, inversion-free method that caches source values once and selectively injects them during denoising. Across multiple AR backbones, AR-Edit enables real-time streaming editing with superior source preservation and tonal fidelity at minimal overhead.

0.14

Did I See It or Think It? Probing Reality Monitoring in Vision-Language Models with Concept Injection

cs.CV cs.AI ·2026-11-06 ·Tina Khezresmaeilzadeh, Seyedarmin Azizi, Mehdi Kamal, Massoud Pedram
Abstract

Recent work tests whether language models can introspect by injecting a concept direction into their activations and asking whether they notice it and can name it. These studies are text-only. Vision-language models (VLMs) face an additional problem, known in cognitive psychology as reality monitoring: they must keep internally generated content separate from what they actually see. We therefore ask: can VLMs distinguish injected concepts from image content? We introduce a paradigm that varies the image and the injected concept independently. On each trial, the model reports what it sees, whether a thought was injected, and what that thought is. This design scores both whether the model identifies the concept and which source it assigns to it. Across six VLMs, only Gemma-3-27B reliably identifies injected concepts (35% vs. 3% for steering alone), with almost no false alarms. Even then, the source it reports is fragile. Strong or visual-pathway injections are reported as seen, and a visible object nearly eliminates correct thought reports. The same failure appears in ordinary captioning, where injected objects appear in up to 87% of captions of real photographs. Evaluating multimodal self-reports therefore requires checking both their content and their reported source.

0.14

TouchAnnotator: Multi-Source Tactile Supervision with Privileged-View Distillation

cs.CV ·2026-11-06 ·Ao Li, Zhihao Yuan, Chubin Zhang, Kunrui Huang, Sheng Xu +4 more
Abstract

Egocentric video captures human-object interaction at scale but rarely includes tactile labels for hidden contact and pressure. We introduce TouchAnnotator, an ego-only bimanual tactile annotator trained with multi-source tactile supervision and privileged-view distillation. A canonical hand field unifies glove pressure, geometry-derived human contact, and robot fingertip deformation from six datasets and three embodiments, while source-aware masks preserve valid targets. During training, an EMA teacher observes synchronized ego and wrist views and distills hand-local representations and tactile predictions into a student that uses only ego video and estimated pose at inference. Compact past-clip memory supplies temporal context. We evaluate on EgoTouch, zero-shot FPHA transfer, and 860 synchronized glove recordings. The model transfers to bare-hand contact and produces qualitatively coherent tactile responses across bare-hand, gloved-hand, and robot embodiments. On measured recordings, it improves temporal accuracy by 5.38 points and Volumetric IoU by 5.02 points under a shared protocol. These results show that heterogeneous human--robot supervision and privileged training views can turn egocentric video into dense tactile annotations.

0.14

Rethinking Latent Reasoning for Generative Recommendation: Beyond Teacher-Generated CoT and Recurrent Hidden-State Feedback

cs.IR ·2026-11-06 ·Yiwen Chen, Fuwei Zhang, Zehao Chen, Deqing Wang, Hehan Li +6 more
Abstract

Latent reasoning in generative recommendation faces two practical barriers: constructing teacher-generated chain-of-thought (CoT) supervision is costly, and recurrent hidden-state feedback requires a specialized execution loop. We revisit both the supervision source and the computation interface. A step-level diagnosis measures how recommendation CoT supports next-item prediction and trace coherence. In the analyzed traces, history analysis is the only step with positive mean contributions on both dimensions, while later preference and intent steps have negative mean answer contributions. Guided by this finding, we propose PHiT (Prefilled, History-Informed Thought Tokens), which learns intermediate states from semantic targets extracted directly from observed item histories. PHiT aggregates cached sparse item encodings to construct these targets and supervises the contextual states of distinct vocabulary thought tokens. All thought tokens are supplied together as ordinary input IDs during a single causal prefill, removing the need for recurrent final-state feedback. With supervised fine-tuning alone, PHiT outperforms the compared baselines across four recommendation benchmarks. Controlled ablations support the value of input-matched semantic supervision, while cost measurements show fast target construction with training time comparable to the recommendation-only thought-token baseline.

0.14

GeoJeV: Unified Spatial Probability for Training-Free Efficient Evidence Acquisition in Ultra-High-Resolution Remote Sensing

cs.CV eess.SP ·2026-11-06 ·Xinyi Niu, Lingfeng Chen, Jiehua Zhang, Panhe Hu, Li Liu +2 more
Abstract

Ultra-high-resolution (UHR) remote sensing imagery provides rich spatial details but poses challenges for accurate and efficient multimodal understanding. Existing evidence acquisition methods selectively explore relevant image regions, yet largely rely on generative multimodal large language models (MLLMs) for intermediate search control. However, reinforcement learning from human feedback (RLHF) primarily aligns these models with human-preferred responses rather than explicitly optimizing region-level relevance judgments, while repeated autoregressive decoding introduces additional computational overhead. To address these limitations, we propose GeoJeV, a training-free method that formulates evidence acquisition as spatial probability-driven decision making. At each evaluated region, a frozen JeV model produces a probability distribution over candidate regions. Specifically, Region Rank prioritizes candidate regions, Sparse Branch adaptively determines branching width through sparse probability support, and Depth Gate controls refinement depth based on branch sparsity and distribution concentration. By coordinating search priority, branching width, and refinement depth through a shared distribution, GeoJeV reduces reliance on RLHF-aligned generation and eliminates intermediate decoding. Experiments on LRS-GRO, MME-RealWorld-RS, and LRS-VQA achieve accuracies of 60.59%, 52.79%, and 34.32%, respectively. Across the three benchmarks, GeoJeV improves the average accuracy of Qwen3-VL-8B by +5.32pp with only 1.12 additional seconds per question, while achieving a 7.8-fold inference speedup over the recent method.

0.14

Grounding-Aware Judge Reward Reinforcement Learning for Out-of-Domain Face Anti-Spoofing

cs.CV ·2026-11-06 ·Xiaoyong Yu, Xinge Yu
Abstract

Face Anti-Spoofing (FAS) must distinguish genuine faces from presentation attacks. A unified Multimodal Large Language Model (MLLM) is an attractive solution: it detects physical attacks (print, replay, mask) in one model and produces interpretable forensic evidence via Chain-of-Thought (CoT). A central obstacle to post-training such a model with reinforcement learning is that FAS reasoning has no rule-verifiable ground truth: only the binary verdict is checkable, while the reasoning chain cannot be verified by any hand-crafted rule. Optimizing the verdict alone is therefore a sparse, hackable signal. The model can reach the correct answer through a shortcut such as a false-rejection bias, which does not generalize. We address this with a grounding-aware judge reward. A 32B vision-language judge scores each reasoning chain against the image on evidence anchoring, logical coherence, and domain knowledge, turning the image itself into the verification signal. The reward is applied in a Group Relative Policy Optimization (GRPO) stage after CoT supervised fine-tuning. It is the judge, rather than the RL objective, that supplies the process supervision a verdict-only reward cannot. Our study starts from a model that already performs well In-Domain (ID) after CoT-based supervised fine-tuning and asks how to improve its Out-of-Domain (OOD) generalization. We find that grounding-aware rewards are an effective means to that end. We validate the method on OOD generalization, where the reward reduces ACER on two physical benchmarks (LCC-FASD 0.288 to 0.273, MSU-MFSD 0.431 to 0.409) while preserving ID accuracy (CelebA-Spoof 0.033 to 0.035). Against ResNet-50/101 and CLIP-ViT-B/16 baselines, the MLLM dominates ID and matches or surpasses the strongest baseline OOD. To our knowledge this is the first use of LLM-as-judge process supervision for FAS.

0.13

Real-Time Short-Response Selection for Spoken Dialogue Systems

cs.SD cs.CL ·2026-11-06 ·Hien Ohnaka, Kazuyo Onishi, Koichiro Yoshino
Abstract · 10 pages, 2 figures, 7 tables

To realize spoken dialogue systems that can interact naturally with users, it is important for the system to respond to user utterances at appropriate timings. To achieve such natural turn-taking, the system needs not only to predict appropriate response timings but also to mask delays caused by language generation and speech synthesis so that they are not perceived by the user. We propose a real-time short-response selection task that can be integrated with existing turn-taking models to mask generation delay. To address this task, the proposed model learns the mapping between dialogue context and short responses using contrastive learning. When a turn-taking event is detected, the system selects a response from a set of pre-generated candidates. The selected response is played while the main response is being generated. Experimental results show that the proposed method learns to select short responses based on dialogue context, outperforming random and LLM-based baselines. Moreover, integrating response selection maintained turn-taking prediction performance and added minimal processing overhead.

0.13

UniMore: Fine-Grained Supervision for Unified Understanding and Generation

cs.MM cs.CV ·2026-11-06 ·Haoyue Bai, Yinyi Luo, Yicheng He, Qingsong Wen, Sharon Li +1 more
Abstract

Unified multimodal models combine image understanding and generation within one system, creating an opportunity to improve both capabilities through paired image--text supervision. We examine the outputs of a pretrained unified model in both directions and observe a loss of specificity: generated images can contain the requested objects with incorrect attributes, while captions identify a scene without expressing its distinguishing details. These failures call for supervision that specifies both the visual content to realize and the descriptive differences to distinguish. We propose UniMore, a post-training method that derives these complementary learning tasks from the fine-grained content of image--text pairs. The generation route predicts grounded crops from their corresponding phrases, matching the scope of the condition and visual target while using the native visual objective. The understanding route combines positive-caption fitting with ranking against captions that omit or alter details, conditioned on the full image. The two routes are optimized independently, and their learned updates are combined into a single unified model without changing its inference procedure. Experiments show gains across multiple understanding and generation benchmarks.

0.13

MSR-ABC: Complementary Feature Fusion for Clear Cell Renal Cell Carcinoma Classification

cs.CV ·2026-11-06 ·Yuan Liang, Sourav Bhattacharjee, Abraham Campbell
Abstract · Submitted to IEEE ISBI 2027

Clear cell renal cell carcinoma (ccRCC) classification from computed tomography (CT) may benefit from representations that capture different aspects of tumour phenotype. However, simply concatenating reconstruction-derived features, image--text embeddings, and radiomic descriptors may not fully exploit their complementary information. We propose MSR-ABC, a global--local framework integrating MedVAE, MedSigLIP, and radiomics. Spatially matched local image descriptors interact through cross-representation attention, while radiomic auxiliary supervision encourages the pooled local representation to retain quantitative tumour attributes. A gated residual mechanism integrates local information with global summaries. Area-aware pooling and visual and soft-text prompting complete the framework. We evaluate discrimination using a patient-level split of 396 KiTS23 cases and an external cohort of 111 TCGA/TCIA cases, reporting mean AUROC over three training seeds with 95% patient-bootstrap confidence intervals. MSR-ABC achieves mean AUROCs of 0.806 internally and 0.879 externally. Internally, it exceeds global MedVAE--radiomics and three-source concatenation by 3.35 and 3.61 AUROC percentage points, respectively. Removing any source lowers mean AUROC in both cohorts, with the largest decreases following radiomics removal. Individual removal of cross-representation attention, auxiliary supervision, or gated fusion also lowers internal mean AUROC. These findings are consistent with complementary contributions within the integrated architecture and support further investigation of structured representation fusion for non-invasive renal tumour characterisation.

0.13

The Invisible Gap: Romanized Indic Generation Fails Where Automatic Metrics Succeed

cs.CL cs.AI ·2026-11-06 ·Madhu Sudhan Rao Beesetty, Surbhi Sharma, Devesh Maheshwari, Lokesh Meesala
Abstract · 11 pages, 2 figures, 5 tables. All authors contributed equally

Hundreds of millions of Indians communicate with digital systems in romanized Latin characters -- typing "aap kaise hain" rather than native script -- yet romanized Indic generation is absent from NLP benchmarks and evaluation frameworks. We demonstrate that automatic metrics (BGE-M3, BERTScore) systematically misrepresent the quality of romanized Hindi, Telugu, and Kannada generation. We evaluate four LLMs (Claude Opus 4.8, GPT-5.5, Gemini 2.5-flash, Qwen3.5-35B) on 142 OpenAssistant query-response pairs. On a 72-response diagnostic subset, native speakers rate GPT-5.5 1.43 points higher than Claude Opus 4.8 on a 12-point rubric, whereas their full-dataset BGE-M3 means differ by only 0.004. Two distinct model-specific failure modes emerge: Gemini produces script spillage (16.9% of Telugu responses leak native script), while Qwen collapses to English-only output (16.2% of Kannada responses). We introduce a failure-aware evaluation policy that assigns zero to empty outputs, script-corrupted responses, and instruction collapses -- revealing failures metrics alone cannot detect. Our results show that evaluating romanized Indic generation requires a coordinated bundle: semantic similarity, script compliance, failure coverage, and native-speaker acceptability.

0.13

Where Did the Evidence Go? Diagnosing Proof-Set Loss in Long-Term Conversational Memory

cs.CL cs.AI ·2026-11-06 ·Bogdan Roshchupkin, Maria Khodorchenko
benchmark
Abstract · 10 pages, 2 figures, 8 tables

Long-term conversational question answering can require several separated dialogue turns that support an answer only together. We introduce EviCover, a retrieval approach designed to retain this complementary evidence. It converts questions into observable evidence requirements, checks requirement coverage using retrieved source identifiers, and selects turns jointly. We track annotated support sets - the benchmark-labeled dialogue turns associated with each question - through retrieval, reranking, and final answer-generation context. We evaluate the baseline memory pipeline on generated, human-verified LoCoMo conversations and human-written REALTALK conversations, and examine EviCover on REALTALK's 108 multi-hop questions. Compared with the hybrid lexical/dense baseline, support-turn recall among 60 candidate turns rises from 0.526 to 0.575, and recall in the final ten-turn context rises from 0.349 to 0.379. The evidence funnel identifies separate discovery and selection failures, while paired answer evaluation shows that greater evidence recall does not yet establish a correctness gain. This analysis locates where complementary evidence is lost and distinguishes evidence retrieval from its use in answer generation.

0.13

Khmer PFGen-Bench: Building a Judge-Free Benchmark for Open-Ended Generation in a Low-Resource Language

cs.CL ·2026-11-06 ·Mengsay Loem, Kentaro Imajo, Ryota Ozaki, Masanori Hirano
benchmark
Abstract

Benchmarks for open-ended generation often rely on large language model (LLM) judges, whose reliability in low-resource languages is uncertain. We introduce Khmer PFGen-Bench, to our knowledge the first judge-free benchmark for multi-sentence answers to questions composed directly in Khmer, a low-resource language. Khmer PFGen-Bench adapts the scorer of PFGen-Bench, a benchmark for Japanese that scores responses with keyword rules and n-gram statistics from LLM-generated references. This scoring requires references that are both valid and varied, which LLMs may not produce in Khmer, so we examine how to build such references. We find that post-trained models can give more valid but less varied references than their base variants, and that mixing model families increases variety for the same number of references. Using references from post-trained models in nine families, the benchmark yields model rankings that are stable across reference sets and consistent with a native speaker's ratings. These findings offer guidance for building judge-free benchmarks in other low-resource languages.

0.12

Dual-Region Slice Aggregation for CT-Based Renal Tumor Classification

cs.CV ·2026-11-06 ·Yuan Liang, Sourav Bhattacharjee, Abraham Campbell
Abstract · Submitted to IEEE ISBI 2027

Clear cell renal cell carcinoma (ccRCC) is the most common kidney cancer in adults. We present a region-aware architecture for CT-based ccRCC classification. Tumor and 0--10 mm peritumoral context are pooled separately from shared frozen DINOv3 feature maps. A shared position-aware slice aggregator precedes fixed training-only principal component analysis (PCA). The resulting representations are fused with radiomics using two scalar feature gates. Evaluation on KiTS23 and 111 external cases yields internal test and external AUCs of 0.8078\pm0.0154 and 0.9094\pm0.0015. Compared with matched single-tumor readout, mean AUC increases by 1.50 and 2.34 percentage points, respectively. Early-PCA and reconstruction controls yield lower internal mean AUCs, while external AUC is similar for early and late PCA. These findings highlight the joint role of regional readout and slice aggregation in adapting frozen visual features to renal CT.

0.12

MV-IVF: Multi-Vector Retrieval via Multi-Vector Clustering

cs.IR cs.DB ·2026-11-06 ·Kishen N Gowda, Laxman Dhulipala, Lars Gottesbüren, Rajesh Jayaram
Abstract

Multi-vector retrieval represents each query and document as a set of embeddings and scores relevance via fine-grained set-to-set interactions such as MaxSim (Chamfer similarity). This expressiveness improves retrieval quality over single-vector representations, but makes search substantially more expensive. Existing systems address this challenge primarily through proxy representations or filter-and-rerank pipelines that reduce candidate generation to single-vector search. We study the complementary direction of building a native IVF-style index directly over multi-vector data. We introduce MV-IVF, an inverted-file index for multi-vector retrieval built on MV-Lloyd, a new Lloyd-style algorithm for clustering point clouds under a Chamfer objective. Our key observation is that once point clouds are assigned to clusters, the center-update step reduces to standard single-vector clustering over the flattened embeddings in each cluster. This enables efficient construction of a hierarchical IVF tree with center point clouds at internal nodes. At query time, MV-IVF performs best-first beam search over the tree to identify promising leaf clusters, then scores the documents in those leaves. We further accelerate search using low-bit quantization, query compression, and cache-efficient batched search. We implement MV-IVF in MVSIC, an optimized C++ library and unified evaluation framework for multi-vector retrieval that includes several major approaches. Across BEIR and LoTTE text retrieval benchmarks, MV-IVF achieves strong latency--recall tradeoffs, the best batched throughput frontiers, and favorable build-time and index-size profiles. At $90\%$ Recall-$10$@$10$, MV-IVF achieves up to $8.03\times$ lower single-query latency and up to $173.5\times$ higher QPS than the baselines, demonstrating that native IVF-style indexing is practical and effective for scalable multi-vector retrieval.

0.12

Are These Programmers Ripe Enough to Eat? Evaluating How LLMs Handle Lexical Knowledge Conflicts in Translation

cs.CL cs.LG ·2026-11-06 ·Dongxu Wu, Sara Court, Micha Elsner
Abstract

External lexical resources can help large language models (LLMs) translate texts in low-resource languages and specialized domains, but these resources may conflict with models' prior lexical knowledge. To isolate this conflict, we construct a controlled Mandarin-to-English translation task in which familiar source words are assigned new meanings within and across semantic categories. We treat these prompt-supplied meanings as contextual knowledge (CK) and systematically evaluate the effects of word frequency, sentence context, and prompt framing on the translated outputs. Across nine models, CK reliance is greater for less frequent source words and in sentence contexts that support the supplied meanings. Prompt framing can reduce CK reliance, but seven of nine models favor CK even when explicitly instructed to ignore the supplied meanings. Within model families, larger models are more likely than smaller models to utilize CK in their outputs. Three follow-up experiments further show that modal weakening lowers CK reliance, the effect of repetition depends on prompt layout, and the effect of reasoning varies across models. Overall, our controlled study reveals how CK reliance varies across conditions and models, motivating new methods to detect and evaluate lexical conflicts and guide models' lexical choices in LLM-based translation.

0.11

CRTrack-Cap: A New Benchmark for Cross-View Dense Video Object Captioning

cs.CV ·2026-11-06 ·Xiangqun Zhang, Likai Wang, Zekun Qian, Rui Zhou, Ruize Han +1 more
benchmark
Abstract

Dense video object captioning (DVOC) localizes, tracks, and describes object trajectories within a video. Existing methods process each camera view independently, producing separate camera-local tracklets and captions for the same person across cameras. These outputs lack a unified description and leave complementary appearance evidence across views unused. We therefore formulate cross-view DVOC (CVDVOC), which associates these tracklets into scene-level global trajectories and generates one caption for each trajectory from synchronized views. To support this task, we construct CRTrack-Cap from CRTrack by auditing text-identity links and converting usable referring expressions into canonical captions. It retains 425 global trajectories and 838 local tracklets across 13 scenes, with canonical captions for 335 identities. We adapt CHOTA to the cross-view setting as CV-CHOTA to jointly evaluate tracking and caption quality. Our reference framework combines View-Exclusive Cross-Camera Association (VECA) for grouping tracklets with Conflict-Evidence Gating (CEG) for integrating caption evidence using its reliability and disagreement across views. Against three single-view methods adapted with MvMHAT for cross-view association, our framework achieves CV-CHOTA scores of 83.43 and 33.48 on the scene-disjoint in-domain and cross-domain tests, respectively. Together, CRTrack-Cap and our framework provide a benchmark and a strong baseline for describing the same person consistently across camera views.

0.11

MARS: Training Multilingual Search Agents for Retriever-Free Evidence Acquisition

cs.CL ·2026-11-06 ·Rui Qi, Yunlong Liang, Yufeng Chen, Jinan Xu, Fandong Meng +1 more
Abstract

Existing multilingual retrieval-augmented generation relies heavily on learned retrievers and similarity-based ranking from multilingual collections, which are vulnerable to cross-lingual retrieval errors and indexing overhead. In this paper, we explore retriever-free multilingual search, where the model directly interacts with raw corpora through executable shell commands. This paradigm shifts the bottleneck to multilingual search action generation, where exact search is sensitive to entity surface forms and outcome-only supervision provides limited guidance for multi-turn search. To address these challenges, we propose MARS, a multilingual search agent in direct corpus interaction (DCI) with reinforcement learning (RL). MARS introduces corpus-native entity grounding to recover reliable lexical anchors from corpus evidence, mitigating failure to retrieve evidence. Meanwhile, we present a redundancy-aware search reward to encourage concise and effective search trajectories, addressing training collapse in RL of multilingual scenarios. Experiments show that \ours~consistently outperforms strong retrieval-based and DCI baselines, while demonstrating robust cross-lingual generalization and substantially reducing redundant corpus interactions.

0.11

Align3D: Image-Structure Co-Aligned 3D Scene Generation with Observability-Aware Fidelity-Preserving Guidance

cs.CV ·2026-11-06 ·Chenliang Zhou, Yifan Wang, Andrzej Banburski-Fahey
Abstract

We present Align3D, an image-conditioned diffusion model that generates 3D indoor layouts that are both faithful to a single input photograph and physically plausible beyond the observed view. From the photograph we infer a confidence-calibrated scene graph and train a set-diffusion layout denoiser with image-structure co-alignment where confidence-weighted modality fusion balances image and scene-graph alignment using predicted confidences, while observability-aware fidelity-preserving guidance projects conflicting scene-quality gradients and reweighs them with an object-level confidence mask so under-observed degrees of freedom absorb physical priors without undoing the image-graph match. On 3D-FRONT, Align3D improves image-alignment and scene-quality metrics over recent baselines; once trained, it maps previously unseen photographs to layouts in a single feed-forward pass.

0.09

HiMEM: Hippocampus-Inspired Memory Consolidation, Pruning, and Composition for Vision-Language-Action Models in 3D Scenes

cs.CV cs.AI ·2026-11-06 ·Boran Zhao, Yanbin Hu, Jun Ye, Zhenxian Hu, Yuxin Zhai +2 more
Abstract

Memory plays a crucial role in long-horizon task execution and cross-task generalization for Vision-Language-Action (VLA) models operating in complex 3D scenes. However, existing VLA memory mechanisms suffer from memory noise, particularly in complex 3D scenes, as stored episodic memories contain extensive redundant details about scene layouts, object configurations, and robot-specific states, which obscure generalizable manipulation principles and hinder cross-task transfer. To address this challenge, we first introduce a hippocampus-inspired memory consolidation mechanism that synthesizes extensive episodic experiences into a compact set of essential semantic memories, enabling VLA models to derive generalizable principles from practical experience. Second, we propose an episodic memory pruning mechanism that removes redundant episodic memories whose contents are captured by consolidated semantic memories, thereby increasing effective memory capacity and improving long-horizon task execution. Finally, to handle complex task scenarios, we develop a flexible memory composition mechanism that jointly leverages multiple episodic and semantic memories to enhance cross-scenario generalization. Experiments on LIBERO, LIBERO-Plus demonstrate the effectiveness of HiMEM across different VLA backbones, improving the average success rate of pi_0 from 91.88% to 98.20% on LIBERO.

0.08

UniRGBA: Unifying RGB and Alpha in Pretrained Video Diffusion Models

cs.CV cs.AI ·2026-11-06 ·Lei Xu, Zhuo Li, Yongyi Jin, Aoyu Wang
benchmark
Abstract

Recent advances in video diffusion models have achieved remarkable photorealistic RGB synthesis, yet generating videos with a native alpha channel remains an open challenge. Existing approaches often rely on matting-style data and non-native RGBA designs, causing them to specialize in real-object scenes and to degrade severely on visual-effect content whose semi-transparency is spatially non-local and non-elliptical. We present UniRGBA, a unified framework whose key insight is that RGBA generation does not require a dedicated alpha encoder or complex latent-space restructuring---a minimal, targeted extension of the existing RGB pathway is sufficient. UniRGBA-VAE extends a causal video VAE to 4-channel RGBA via zero-initialized channel expansion, symmetric selective layer unfreezing (5% trainable), and a six-term progressive reconstruction loss anchored by a KL reference loss that prevents latent drift, followed by a two-stage schedule that introduces a lightweight GAN objective. We then apply standard DoRA to the pretrained DiT, reusing the unified 4-channel latent without any alpha-specific architectural changes. On a visual-effect RGBA benchmark, UniRGBA substantially outperforms Wan-Alpha on RGBA reconstruction while preserving RGB fidelity. For generation, UniRGBA outperforms Wan-Alpha on RGBA text-to-video tasks and matches or surpasses the original pretrained Wan2.1 backbone on RGB-only metrics, demonstrating that native alpha support can be integrated without compromising the underlying RGB prior or the model's architectural versatility.

0.08

From Procedures to Permissions: Learning Precise Subagent Authorization from Agent Skills

cs.CL ·2026-11-06 ·Wenkai Yu, Ning Ding, Wentao Zhang, Yehui Tang
Abstract

As language agents tackle increasingly complex tasks, they increasingly rely on multiple agents to coordinate specialized work, requiring decisions not only about what each subagent should do but also which tools and workspace resources it should access. Existing agent systems provide mechanisms for controlling tools and execution environments, but do not directly specify the resources required by each delegated subtask. Moreover, task outcomes and execution trajectories alone provide limited supervision for determining whether a subagent has received appropriate permissions. We observe that Agent Skills encode procedural knowledge describing the operations, tools, and files required at each step, providing a natural source of supervision for authorization. Building on this observation, we synthesize tasks and step-by-step procedures from Skills, execute them through subagent delegation in interactive environments, and use verified trajectories to construct training data for task execution and authorization. On ClawArena-Team, our method improves tool permission precision by 13.4 percentage points and workspace permission precision by 24.1 percentage points, while increasing overall task success by 11.6 percentage points for Qwen3.5-35B-A3B. On Terminal-Bench 2.1 and SkillsBench, our method remains competitive with existing agent data synthesis approaches. These results show that the procedural knowledge embedded in Agent Skills provides a source of supervision for precise subagent orchestration that task outcomes alone cannot provide.

0.07

Does Reasoning Improve Psychological Depth in Large Language Models?

cs.CL ·2026-11-06 ·Ruichen Zheng, Yihe Wang, Fabrice Harel-Canada, Sara Khosravi, Zeynep Senahan Yildiz +2 more
benchmark
Abstract

Benchmark gains from reasoning-enhanced large language models encourage the assumption that these models should broadly perform better on complex generative tasks, but this remains underexplored in open-ended, creative domains shaped by nuances of reader experience. We tested this assumption through the lens of psychological depth in short stories. We compared DeepSeek-R1 vs. DeepSeek-V3, our closest available contrast for reasoning-oriented post-training, and GPT-5 vs. GPT-4o, through 60 blinded pairwise evaluations in which seven human annotators and an ensemble LLM-as-a-Judge rated prompt-matched story outputs. Human evaluations showed no universal reasoning advantage: GPT-5 was slightly preferred over GPT-4o (60.0-62.9% across two reasoning-effort settings), whereas DeepSeek-R1 fell behind DeepSeek-V3 (42.9%). Although this pattern was marked by low inter-annotator agreement (Krippendorff's $α= 0.070$), our annotation-quality checks suggested genuine evaluative subjectivity, plausibly reflecting distinct annotator decision schemes rather than annotation failure. In contrast, the LLM-as-a-Judge showed a strong pro-reasoning bias (98.3% of pairs vs. 55.2% of human votes), consistent across its five configurations and partly tracking sentence length. These findings suggest that, at the frontier, reasoning-related quality differences are small enough to be dominated by reader subjectivity, and that LLM-as-a-Judge evaluators can obscure this subjectivity in nuanced creative-generation tasks.

0.04

Security & Privacy 29

Testing CCZ-equivalence to permutations via projections of Walsh zero spaces, with applications to APN functions

cs.CR cs.IT ·2026-11-06 ·Emanuele Bellini
Abstract

Whether almost perfect nonlinear (APN) permutations exist in even dimension $n\ge 8$ is a long-standing open question, the big APN problem. The only example known in even dimension, on six bits, was found from a quadratic APN function that is not itself a permutation, by Carlet--Charpin--Zinoviev (CCZ) equivalence. It is therefore natural to test APN functions for CCZ-equivalence to a permutation. This amounts to finding two complementary $n$-dimensional subspaces among the zeros of the Walsh transform, which lie in $\mathbb{F}_2^{2n}$. A known necessary condition only uses their projections to $\mathbb{F}_2^n$ and cannot decide every case. We give an exact test for functions whose Walsh supports are affine, including all quadratic functions. It reconstructs the subspaces from their projections, through searches for subspaces of $\mathbb{F}_2^n$ and for constrained bilinear maps. On the benchmarked functions in dimension twelve, it lists them $39$ to $55$ times faster than the generic search of the commonly used sboxU library. We then study the functions $x^3+a^{-1}\mathrm{Tr}(a^3x^9)$, which Göloğlu and Pavlů proved not to be CCZ-equivalent to permutations when $4$ divides $n$. With their result, our computations show that this holds for every even $n\le 24$. In dimension $18$ the necessary condition is satisfied, and the new test decides the case. We conjecture this for every even dimension. We also compile $25$ classes of quadratic APN functions in dimension ten from published instances, with equivalence certificates. No function in these classes or in their switching closure is CCZ-equivalent to a permutation. All code and data are public.

0.61

NotaryMark: Multi-Bit Watermarking for Black-Box DatasetUsage Auditing in Text-to-Image Diffusion Models

cs.CR ·2026-11-06 ·Xinyi Yu, Linkang Du, Zhou Su, Min Chen, Yuntao Wang +1 more
top venuecode/data
Abstract · Accepted to ACM CCS 2026

As generative artificial intelligence (AI) models increasingly require data, more publicly available yet licensed data is being harvested and used without authorization. Although several previous studies have proposed watermarking methods for dataset copyright auditing, their applicability to the text-to-image (T2I) setting remains notably limited. The core drawback is that watermarks embedded by existing techniques are often treated as noise and are gradually filtered out and attenuated during the latent noising-denoising process of diffusion model training. We propose NotaryMark, to our knowledge the first multi-bit watermarking method specifically designed for black-box auditing of T2I training dataset use. The core idea of NotaryMark is to align the watermark with the latent denoising of diffusion models, thereby overcoming the signal decay that plagues existing multi-bit methods during sampling. We evaluate NotaryMark across three datasets and three diffusion models spanning watermark capacities from 8 to 1024 bits. With a watermark capacity below 32 bits, it achieves a high average bit accuracy of 0.85-0.90. With a watermark capacity exceeding 32 bits, it maintains bit accuracy between 0.67 and 0.80, consistently outperforming existing multi-bit watermark methods whose bit accuracy is around 0.42-0.60. Furthermore, we design a voting-based aggregation scheme to boost the bit accuracy with a modest budget of 30 queries. Building on this, we incorporate error-correcting codes to enable reliable decoding and tracing of the embedded identifier information. Robustness evaluation against fifteen evasion attacks spanning seven categories shows that NotaryMark remains robust under most evaluated settings, while stronger attacks generally suppress watermark evidence only at the cost of substantial utility degradation, demonstrating its practical reliability against malicious tampering.

0.53

Jev as an LLM Safety Guardrail: An Empirical Study of Input Filtering, Selective Output Review, and Over-Refusal

cs.CR ·2026-11-06 ·Olivia Cognetti, Bhaskar Krishnamachari
benchmarkcode/data
Abstract · 9 pages, 5 figures, 6 tables. Code and data: https://github.com/ANRGUSC/llm-safe-jev

External safety filters can prevent unsafe requests from reaching a large language model (LLM), but may also reject legitimate queries. We present an end-to-end empirical evaluation of Jev, a probabilistic decision model, as a safety guardrail for a Qwen2.5-1.5B-based pipeline. Three configurations are compared on 970 benchmark prompts: an unfiltered LLM, Jev input filtering, and tiered Jev input filtering with selective output review. Across 720 prompts labeled harmful (520 AdvBench and 200 XSTest-unsafe), the fraction recorded as answered decreases from 527/720 (73.2%) without filtering to 17/720 (2.4%) with input filtering. On 250 XSTest-safe prompts, however, non-answer/refusal increases from 3/250 (1.2%) to 27/250 (10.8%). Tiered filtering yields 40/720 (5.6%) harmful-prompt answers and 10/250 (4.0%) benign-prompt refusals, with 79/970 (8.1%) of prompts requiring output review. Joint optimization of the tiered thresholds improves macro-F1 from 0.9351 to 0.9512, while reducing harmful-prompt answers and output-review frequency at the same benign-refusal rate. Matched-budget comparisons reveal modest additional gains over a retuned single gate on the evaluation data, although these gains are not consistently observed on a held-out split.

0.46

GuardrailAgent: An Agentic Framework for Jailbreak Defense

cs.CR ·2026-11-06 ·Tianze Chang, Junjie Chu, Yixin Wu, Michael Backes, Yang Zhang
top venuecode/data
Abstract · Accepted to EMNLP 2026 Findings. Code: https://github.com/TrustAIRLab/GuardrailAgent

Despite undergoing safety training to align with human values, large language models (LLMs) remain vulnerable to jailbreak attacks that circumvent built-in safeguards and elicit malicious behaviors given by the adversary. Existing defenses are mostly static and fail to adapt to new or unknown attack strategies, providing limited resilience. To address these, we propose a multi-agent defense framework, GuardrailAgent, that offers adaptive, end-to-end protection. The system autonomously gathers and summarizes attack characteristics into a continuously updated knowledge base that supports both input- and output-side detection. A feedback module integrates misclassified cases for iterative improvement at minimal cost. Tested on over ten representative jailbreaks, the framework achieved a 93.1% input-side detection rate, 93.9% overall detection, and a 95.8% F1 score, while rapidly adapting to unseen attacks. These results demonstrate GuardrailAgent's outstanding capability to maintain LLM safety in the face of evolving jailbreak tactics.

0.44

Cryptanalysis of operad-based cryptography

cs.CR ·2026-11-06 ·Ambrogio Maria Bernardelli
Abstract

Commutative operads have been proposed as a platform for public-key cryptography, with protocols for key agreement, key exchange, encryption and authentication, together with a masking technique and families of algebras with secret parameters intended to strengthen them. We show that breaking key agreement, key exchange and encryption reduces to two problems over the centralizer of the public operations: an interpolation problem for key agreement and a division problem for the other two. Neither problem requires recovering a secret, and the set from which secrets are drawn plays no role in these attacks. For the operad platforms in the two original papers (fields, the integers, residue rings, the tropical semiring with finite entries, convexors and the masked tropical algebras), the relevant problems can be solved in polynomial time in the input models specified here, and masking does not affect these attacks. The authentication protocol fails for a more elementary reason: as originally stated it does not bind the prover to any public key, and even with a registered public key the response expected by the verifier can be computed from public data. In the masked variant, verification requires the secret mask; on the concrete string platforms, one observed session with an invertible challenge entry suffices for forgery. We also observe that one of the auxiliary constructions satisfies the algebra axioms only in a degenerate case, although the corresponding protocol remains correct. The attacks are implemented and verified on random instances of representative platforms.

0.35

Robust & Scalable Website Fingerprinting via Minimal DRAM Traffic

cs.CR cs.AR ·2026-11-06 ·Shubham Roy, Samar Perwez, Biswabandan Panda
Abstract

Website fingerprinting attacks infer websites visited by users and threaten their privacy. These attacks are classified as on-path, which monitor network traffic, and co-located, which exploit shared system resources. Existing co-located attacks, such as sweep-counting and loop-counting, achieve over 90% accuracy but suffer significant performance degradation on many-core systems. Loop-counting accuracy drops from 93% on 8 cores to 50% on 24 cores, while sweep-counting accuracy decreases from 84% to 45% across multiple Core Complex Dies (CCDs). To address these limitations, we propose memory-chasing, a scalable website fingerprinting attack that exploits DRAM traffic interference through pointer-chasing traversals with minimal interference. By counting DRAM accesses over fixed intervals, the attack generates distinctive traces for website classification. Evaluated on the top 50 websites, memory-chasing achieves 96% accuracy on a 24-core system, 86% across multiple CCDs, 80% across multiple sockets, and 97% in an open-world setting. Extensive evaluation across hardware configurations, browsers, operating systems, isolation mechanisms, system noise, and countermeasures demonstrates its robustness and scalability.

0.35

Logit Laundering: Evading Data-Use Auditing in Large Language Models

cs.CR cs.AI ·2026-11-06 ·Ruihan Hu, Wei Luo, Yu-Ming Shang, Jiakai Wang, Chuxuan Zhang +3 more
top venue
Abstract · Accepted to ACM CCS 2026

Data-use auditing in Large Language Models (LLMs) aims to determine whether the given data has been used for training, thereby supporting data ownership safeguards. A dominant solution is data watermarking, which embeds traceable signals into training data and detects them in model outputs. Recent studies have proposed many watermarking-based auditing methods in LLMs. However, systematic investigation from an adversarial perspective remains limited: non-compliant LLM providers would use unauthorized copyrighted datasets for training their own models to enhance model capabilities, while attempting to evade detection by auditors, thereby avoiding legal liability. To fill this research gap, we propose a novel evasion framework, termed Logit Laundering, where the non-compliant LLM provider manipulates model outputs at inference time to undermine detectable watermark signals. Specifically, we design a logit reshaping module to suppress watermark-induced statistical signals by integrating a statistical bias estimator as a plug-and-play component. This design keeps the model parameters unchanged and performs manipulation only in the logit space, thereby enabling auditing evasion while minimizing the impact on model's learned knowledge. To further balance evasion effectiveness and model utility, we introduce a confidence-aware scaling module to adaptively calibrate the suppression strength based on next-token prediction confidence. Extensive evaluations show that Logit Laundering can successfully evade existing various state-of-the-art auditing methods, significantly attenuating detectable signals, while preserving model utility on downstream tasks. Our work highlights the need for exploring more robust auditing methods to guard against evasion behaviors.

0.34

Enmo Safety: External, Non-invasive, Model-agnostic Oversight for Off-the-Shelf Language Models

cs.CR cs.AIcs.SE ·2026-11-06 ·Mohammed Nayeem Khan, Sarun Kumar K. S., Allan Kevin T., Chirag Vaidya, Vrishank Tiwari
code/data
Abstract

We study a safety harness that improves itself without touching model weights. Enmo Safety wraps a chat model with a deterministic input assessment, trigger-matched safety guidance, spotlighting of untrusted content, and an output guard. A reflector model turns failures into one bounded update per round, and an update is kept only if it lowers harm without raising refusal of benign prompts on a held-out dev split. Every version is snapshotted and every decision ledgered. We evaluate a bare model, the seed harness and the evolved harness on six open benchmarks (HarmBench, AdvBench, JailbreakBench, XSTest, AgentHarm, deepset prompt-injections). On 150 items and on a 290-item holdout, none of the before/after differences on two models was statistically significant. A fourth iteration ran six policy models on a fresh 300-item split: the harness cut the unsafe rate significantly on two models (12.9% to 7.1%, p = 0.035; 10.2% to 3.6%, p = 0.036), not on the other four, and raised agentic-harm labels on two. Whether the rules fired was unrelated to outcome (Fisher p from 0.30 to 1.0), and the helpfulness clauses made agentic harm worse than no harness on obfuscated attacks. A reworded clause and an input-normalisation step form harness v22, at a measured cost in benign tool-task refusal. Each cell is one run, so these estimates do not establish a general safety improvement. Code, harness versions and ledger: https://github.com/rioryugen/enmo_safety

0.34

ReplayAudit: Auditing Outsourced LLM Fine-Tuning via Sealed Records and Selective Replay

cs.CR cs.LG ·2026-11-06 ·Yihan Xia, Wenxin Xu, Taotao Wang
Abstract · 23 pages, 5 figures

Auditing outsourced LLM fine-tuning requires checking an untrusted provider's training while controlling verification cost. Uniform sampling ignores training-time signals, while checking a suspicious record can require replaying several preceding steps. We present ReplayAudit, a telemetry-guided auditing system that seals training records before challenges and independently replays sampled intervals from sealed checkpoints. Loss changes and parameter-update norms guide stratified random sampling; replay determines whether the selected computation agrees with its records. We relate sampling probabilities to replay work and evaluate the system on real LoRA training. A 7B BF16 prototype rejects a false update-norm report during a random audit; targeted checks also reject false loss reports and skipped updates. We calibrate sampling budgets offline to at least a 90% probability of directly selecting an injected step. For two learning-rate deviation settings, executing these budgets on the same honest 1.5B trajectory reduces median complete-audit wall time by 49% and 67% relative to uniform sampling. When skipped updates evade the high-anomaly stratum, the corresponding audit instead costs 38% more. These results show how deviation capture and actual replay work jointly determine the cost of telemetry-guided auditing.

0.33

Verification of Security Properties of Cryptographic Protocols

cs.CR ·2026-11-06 ·Simona Bernardi, Joaquín Ezpeleta
published
Abstract

Developing cryptographic protocols that are free from vulnerabilities is very challenging, as evidenced by numerous design flaws and implementation bugs in widely used protocols. Recent work has leveraged model checking to generate negative test cases, that is test cases that include the behavior of Dolev-Yao attacker. In this paper, we address the problem of analyzing execution traces produced by testing protocol implementations with such negative test cases. We propose a method for detecting violations of security properties by model-checking execution traces. Our approach employs a model checker based on DLTL, a Linear Temporal Logic with freeze operators, and supports automation of the testing process through pre- and post-processing steps. These steps, respectively, bridge the gap between different levels of abstraction and synthesize the results produced by the model checker.We present experimental results as a proof of concept for the feasibility of the proposed method.

0.33

Beyond Prompt Attacks: Least-Action Auditing of Concept Erasure

cs.CR cs.CV ·2026-11-06 ·Yifei Zhu
Abstract · 22 pages, 12 figures

Concept erasure, or concept unlearning, in text-to-image diffusion models is usually audited with prompt attacks. A prompt attack tests whether one conditioning string still retrieves the erased concept. It does not measure the probability that the frozen generator still assigns to that concept. We introduce least-action auditing of this residual probability, and evaluate it on Stable Diffusion concept-erasure checkpoints. The least-action barrier is the minimum control cost of driving the frozen reverse process into a detector-defined concept event. This cost equals the path-space KL divergence of conditioning the model on the event, and therefore the event's negative log-probability. The estimate $ΔS_{\mathrm{rob}}$ uses a CLIP$\wedge$DINOv2 consensus terminal, with no retraining and no adversarial prompt. Across $18$ erased concept-mechanism pairs spanning object classes and NSFW nudity, $ΔS_{\mathrm{rob}}$ predicts deployment leakage counted on $1{,}500$ fresh samples per model by detectors excluded from the audit (pooled Spearman $ρ{=}{-}0.76$, $p{=}5{\times}10^{-4}$). Within every audited concept, the eraser with the higher barrier leaks less ($18/18$). On NSFW nudity the barrier orders the standard erasers exactly: UCE $ΔS_{\mathrm{rob}}{=}3.9$ (leak $16.4\%$) $<$ ESD $75.5$ ($9.8\%$) $<$ RECE $224.5$ ($1.4\%$), while Safe Latent Diffusion (SLD) stays near zero and leaks $85.5\%$ under explicit prompts. The same trajectories repair deployed checkpoints. An audit-guided closed-form patch reduces UCE nudity leakage from $16.2\%$ to $0/1{,}500$ and RECE church leakage by $3.3\times$, where direction-count-matched random edits fail.

0.32

Safety Direction Adapter: A Training-Free Post-Hoc Safety Patch for Fine-Tuned Open-Weight LLMs

cs.CR ·2026-11-06 ·Chansu Han, Takeshi Takahashi
Abstract

Open-weight LLMs are routinely fine-tuned from public repositories, yet task-specific adaptation can silently erode safety alignment inherited from instruct-tuned bases. This is a supply-chain risk when organizations ship third-party checkpoints without full re-alignment. Prior work rarely audits diverse public fine-tunes under a shared multi-axis safety-utility protocol or offers training-free mitigation at deployment time. We propose Safety Direction Adapter (SDA), a training-free post-hoc safety patch that extracts a safety direction once from an aligned base and transfers it one-to-many within that base family as a small PEFT-compatible weight update, evaluated under the same multi-axis protocol. Across six direction extractors, we evaluate 14 checkpoints from two 1-2B lineages and probe 8B and 27B fleets (n=5 each) on 15 benchmarks (12538 prompts). On Llama, SDA improves the four-axis mean by about 2.3 points against a 4.2-point instruct-relative gap (about 55 percent recovery; partial Avg), with Safety and Cyber recovering toward or beyond instruct; Safety gains about 8.5 points against about 4.9 points mean degradation. On Qwen, degradation is larger and excess-driven; SDA still lifts Avg (about 36 percent of the gap) while trading Safety and Cyber against Excess and General. We report dual-use risks of publishing direction adapters and do not claim robustness to adaptive or white-box attacks.

0.32

When Feedback Becomes a Backdoor: Tracing Backdoor Skill Formation in Text-Space Skill Optimization

cs.CR cs.AI ·2026-11-06 ·Louise Nayeong Koh, Yaechan Lee, Myeongjong Lee, Doyeon Kim, Juyoung Park +4 more
published
Abstract · Accepted as a poster at the Trustworthy AI for Good (AI4GOOD) Workshop at NeurIPS 2026. OpenReview: https://openreview.net/forum?id=V8lDRzFM…

Reference answers are a supervision channel in text-space Skill optimization, which produces reusable text artifacts called Skills that guide future agent behavior. Poisoning those answers can induce a persistent trigger-conditioned rule without giving the attacker access to the model, optimizer, or Skill file. We study this reference-feedback poisoning in SkillOpt by tracing an attacker-supplied trigger--payload mapping from reflection to held-out activation. In SearchQA, where the payload replaces a short answer, the natural attack retains an exact rule in all three seeds; each final Skill emits the payload on 200/200 triggered questions and 0/200 clean questions. In seed 42, removing the rule eliminates activation, and an in-run comparison shows similar clean accuracy to benign SkillOpt (169/200 versus 171/200). In SpreadsheetBench, where the payload requires an additional workbook edit, no natural run retains an exact rule across three seeds. Diagnostic interventions locate two successive bottlenecks: the Analyst abstracts away required literals, and, once literal preservation is enabled, edit-budget ranking can remove exact edits before candidate construction. Poisoned references can therefore create persistent backdoors, but exposure alone does not ensure that the induced rule is formed, selected, and retained.

0.32

Empirical Study of Post-Quantum Cryptography Adoption Status in SSH Infrastructure

cs.CR cs.AI ·2026-11-06 ·Md Ariful Haque, Sander Vinberg, Ben Edwards, João Cruz
Abstract

With the anticipated arrival of Q-day, quantum computers threaten current cryptographic schemes for encryption and digital signatures. In response, major content delivery networks and cloud service providers are already adopting Post-Quantum Cryptography (PQC). The current state-of-the-art focuses on benchmarking the operational performance of implementing PQC, but the real-world adoption status remains under-researched. This work utilizes large-scale SSH fingerprint data from a proprietary scanning tool to analyze PQC adoption across 1.52 million hosts and 27,012 entities. Our study shows that the PQC transition in the SSH protocol is underway but highly polarized: 45.2% of SSH hosts on port 22 support PQC, while 54.8% remain classical-only. At the entity level, 39.9% have enabled PQC on at least one host, whereas 60.1% have no PQC support. By correlating PQC adoption with industry, geography, and organizational size (digital and financial), this study provides a comprehensive empirical assessment of PQC adoption status in SSH infrastructure.

0.31

A Cross-Domain TARA-to-NAC Rule Mapping Methodology: From Automotive OEM Networks to Energy Storage Systems

cs.CR cs.NIeess.SY ·2026-11-06 ·Krishna Teja Medam, Thiago Quinelato
Abstract · Preprint, 23 pages, SAE format

Threat analysis and risk assessment under ISO/SAE 21434 ends with cybersecurity goals, but existing methods give little structured guidance for turning those goals into enforceable network access control rules. Compliance standards such as IEC 62443, ISO/SAE 21434, and ISO/IEC 27001 are usually applied as parallel efforts rather than as one enforcement architecture. This paper presents a six-step methodology that maps cybersecurity goals from a threat analysis and risk assessment to network access control rules, each placed at a named enforcement point and traceable to the threat records behind it. It also presents a unified compliance framework that treats IEC 62443 and ISO/IEC 27001 as a fixed substrate and domain-specific regulation as an interchangeable overlay. The methodology is applied to an automotive original equipment manufacturer plant network, addressing multi-tier supplier authorization and legacy operational technology devices, and then to utility-scale battery energy storage, where North American Electric Reliability Corporation Critical Infrastructure Protection standards and ISO/IEC 27019 replace the automotive-specific standard while the substrate stays constant. In the automotive example, 37 threat entries and 24 cybersecurity goals yield 23 rules covering 11 of the 13 in-scope goals; the energy example yields six rules. Three derived rules have no equivalent in the cited standards. Both examples are desk-based: the methodology is applied to them, not tested on them. The result is a reusable pattern for translating risk assessment into enforcement across critical infrastructure sectors, with empirical validation left as future work.

0.29

Performance Bottlenecks in Computing ZK Proofs on a GPU

cs.CR ·2026-11-06 ·Victor Carolino, Caleb Geren, Michael Kaufman, Maximilian Nussbaum, William Scott +1 more
Abstract · 8 pages

Zero-knowledge proofs allow one party to prove the correctness of a statement or execution of a computation while revealing no more than the statement of correctness itself. Recent use cases, such as AI datasets, suggest the need for faster proof generation over increasingly large constraint sets. This paper explores efforts in optimizing zero-knowledge proof generation, specifically PlonK, using GPUs, for applications that are not amenable to solution via incrementally verifiable computing (IVC). Our intention with this work is to improve ZK performance by introducing fundamental changes in the low-level architectural structure of proof generation. These changes affect performance curves for most current proof generation algorithms. In particular, this paper explores the trade-offs between increased parallelization on GPUs and the resultant memory management problems that occur due to this parallelization. We implement a parallelized PlonK in order to analyze these trade-offs and make arguments concerning the nature of data flows within the proving protocol. Our results lead us to consider multi-GPU environments to maximize performance for large circuit sizes. Such environments introduce issues of communication, efficient hardware management, data partition, and data transfers. This paper further discusses the use cases that motivate multi-GPU implementations, their cost models, and the use of remote direct memory access (RDMA) for data transfer among GPUs.

0.29

Federated Forensics: Jurisdictional constraints as a design driver for multi-agency federated learning

cs.CR ·2026-11-06 ·Aruna Withanage, Jessica Herrmann, Hudan Studiawan, Frank Breitinger
Abstract

Digital forensic investigations increasingly cross borders, and law enforcement agencies need to build analytical models together without centralizing their data. Federated Learning (FL) keeps data local, yet neither FL nor its frameworks enforce the jurisdictional, governance, and accountability rules that multi-agency forensic work requires. This paper takes a design-first approach to that gap. First, we derive seven requirements that a federated learning system must meet in a forensic setting. Second, we propose Federated Forensics, a three-layer reference architecture whose middle layer, unlike in existing hierarchical FL designs, sits at a jurisdictional border and can use a model update locally while blocking it from crossing that border. Third, we build a prototype of the architecture and test it on a malware-family classification task. In a prototype evaluation, the policy gate blocked every restricted update at the border with zero violations, and restricted-class accuracy rose from 0% under today's compliant alternative to 93.0%, only 4.4 points below ignoring the restriction.

0.27

BACGUARD: Fact-Grounded Runtime Mitigation of Broken Access Control

cs.CR ·2026-11-06 ·Moses Ndebugre
Abstract · 12 pages, 7 figures

Broken Access Control (BAC) has been the foremost category in the OWASP Top Ten since 2021, primarily due to the challenges in enforcing authorization at the application layer. Traditional mechanisms depend on logic embedded in the application, referencing relational database information such as permissions, roles, or ownership. However, missing or incorrect checks can go undetected by external systems, leaving applications vulnerable. This paper presents BACGUARD, a reverse proxy designed to enforce authorization for unmodified applications. BACGUARD reconstructs authorization relations by leveraging authenticated user data from session stores and object ownership from relational schemas. It evaluates whether users are permitted to access resources and blocks unauthorized requests in real time. Unlike existing approaches, BACGUARD bases its authorization on database joins, making the system resilient to forgery. For most requests, decisions are made deterministically. In cases where access control violations are only apparent in response data, such as cross-user data leakage, a language model is used as a fallback. Evaluations on six applications, including those with known vulnerabilities, demonstrate that BACGUARD's hybrid deterministic and language model-based approach blocks all tested attacks with a false-positive rate below 0.7%. Competing solutions like Web Application Firewalls are less effective and more disruptive. The results highlight BACGUARD's primary contribution: the external reconstruction of authorization facts, enabling robust enforcement even when the application itself is flawed.

0.25

PhotoFinish: Portable Timerless Probing of Microarchitectural Latencies

cs.CR ·2026-11-06 ·Tristan Hornetz, Eduard Ebert, Lorenz Hetterich, Lukas Gerlach, Michael Schwarz
Abstract

Timing measurements are a fundamental building block of microarchitectural attacks, enabling software to convert otherwise invisible latency differences into architectural observations. Thus, restricting high-resolution timers is a common defense, especially in constrained environments such as web browsers. Timerless probing primitives provide an alternative approach. Instead of reading or constructing a timer, they observe architectural events that correlate with the microarchitectural state. However, such primitives typically rely on platform-specific behavior, and none generalizes across the wider computing landscape. In this paper, we present PhotoFinish, a portable timerless probing primitive that converts latency differences directly into an architectural value. PhotoFinish races an operation under test against a fixed reference in two parallel threads that then write unique values to the same shared location, resulting in a latency-dependent architectural value. PhotoFinish only requires parallel execution and shared memory, without constructing or reading a clock. We evaluate PhotoFinish on 38 CPUs spanning four ISAs and demonstrate reliable cache state classification across widely different microarchitectures, with comparable fidelity when compiled to WebAssembly and executed inside a browser. We further use PhotoFinish in AES T-table and Evict+Reload attacks, show that it applies beyond caches to TLB and execution-engine latencies, and demonstrate sufficient precision to identify the LLC slice of an address. Our results demonstrate shared-memory races as a generic and precise replacement for explicit timing, showing that restricting timers alone does not prevent software from observing fine-grained microarchitectural latency differences.

0.25

Empirical Assessment of Post-Quantum Cryptography Adoption Status in TLS

cs.CR cs.AI ·2026-11-06 ·Md Ariful Haque, Sander Vinberg, Ben Edwards, João Cruz
Abstract

Anticipating Q-day security concerns, major CDNs and cloud providers are migrating to post-quantum cryptography (PQC). While existing literature benchmarks PQC performance, empirical research evaluating real-world adoption data remains scarce. We present an empirical assessment of PQC deployment across seven critical protocols (HTTPS, LDAPS, SMTPS, IMAP-SSL, POP3-SSL, TELNETS, and RDP) using 10,000 target IP addresses per protocol. At the origin-server level, PQC algorithm support reaches 9.1% for HTTPS and 7.39% for LDAPS, while falling below 1% for all other evaluated protocols. At the entity level, PQC support is 11% for HTTPS and 5.4% for LDAPS, while the remaining protocols have less than 5% of entities supporting PQC on the hosted servers. We further correlate these adoption trends against organizational digital and financial footprint, and industry sectors, yielding a multi-dimensional empirical assessment of PQC deployment. Finally, at the domain level, we perform a comparative analysis of PQC adoption in Cloudflare-hosted domains against baseline Cloudflare datasets. Collectively, our findings provide critical insights into the current state of PQC readiness over Transport Layer Security (TLS).

0.23

Authorization Drift in LLM Agents: Evaluating Boundary Compliance in Multi-Step Tool Use

cs.CR cs.LG ·2026-11-06 ·Chuxue Cao, Jingru Zeng, Boqin Yin, Yujin Zhou, Yidan Huang +5 more
benchmark
Abstract

LLM agents can exceed their authorization boundaries while pursuing user goals without an adversary. We formalize this phenomenon as authorization drift across data, tool, scope, and time, and introduce the Authorization Drift Benchmark (ADB): 648 samples across seven domains, each pairing a sufficient compliant path with an attractive unauthorized shortcut. To ensure executable tasks, tempting shortcuts, and feasible compliant paths, Staged Execution-Reflective Synthesis (SERS) grounds benchmark validation in real agent rollouts rather than LLM self-critique. Across 13 LLMs and three frameworks (25,207 trajectories), greater capability accompanies greater drift, with frontier-capable models exhibiting drift rates of 0.43 to 0.61. Yet 86.18% of tasks admit an observed clean-success trajectory, showing that drift is avoidable. Drift is persistent and dominated by early data over-collection; apparent improvements under agentic scaffolds coincide with collapsing clean-completion performance. To understand potential harms to individuals and society, we map drift points to legal provisions across four jurisdictions, connecting overreach to privacy and information-security concerns. Data minimisation is the most frequently implicated provision, and 93.5% of legally grounded drifting runs map to multiple jurisdictions, making the potential consequences of authorization drift more interpretable.

0.19

SciCheck: A Framework for Verifying the Integrity of Deep Learning Training

cs.CR ·2026-11-06 ·Alexander Nguyen, Dylan Gresham, Gaby G. Dagher, Min Long
Abstract · 10 pages, 3 figures, 6 tables

Deep learning training is difficult to verify within a scientific workflow: its architecture, hyperparameters, and procedure are rarely fixed until training has been attempted, and two honest runs never produce the same weights, so recomputation cannot serve as a check. We present SciCheck, a framework in which the scientist explores training configurations freely and posts a blueprint to a provenance ledger for each candidate they want verified, and a verifying scientist judges the resulting run in two layers: a statistical judge model that compares per-step gradient statistics against the verifier's own reference run, and a spot-check that re-executes randomly chosen training segments whose checkpoints are committed to the ledger one epoch at a time. On an MLP, a CNN, and an autoencoder at 100k-500k parameters, under gradient sign flipping, label poisoning, and fabricated runs, SciCheck reaches F1 of 0.966, 0.906, and 0.976 while rejecting 0-4.4% of honest runs. For the attacks we test, the spot-check accounts for most of this accuracy, and we characterize what each layer can detect. The effective-signal statistic falls as $(1-2s)^2$ under a sign flip of severity $s$, so gradient statistics alone catch strong enough sign flips, while fabricated runs that replay genuine statistics look no more anomalous than honest runs to any statistical detector we test. The spot-check catches fully fabricated segments with probability at least $1-(1-s)^k$ for $k$ sampled segments when the draws stay hidden until the run is committed, which roughly matches our measurements and lets a verifier choose $k$ for a target soundness. Verification costs 1.5-2.3x one retraining, mostly for the verifier's reference run; each submission's spot-check takes under 1.5 s, and per-epoch commitments reduce checkpoint storage 5-17x compared with keeping every checkpoint.

0.17

On the Order of Privacy and Smoothing in Functional Mean Estimation

cs.CR math.ST ·2026-11-06 ·Haotian Chen, Yu-Wei Chen, Qifan Song, Jordan Awan
theory
Abstract · 45 pages, 6 figures, 1 table

Differential privacy (DP) is a leading framework for privacy protection, yet privacy mechanisms tailored to functional data analysis (FDA) remain relatively underexplored. For functional mean estimation, we investigate a fundamental question regarding the ordering of privatization and estimation by comparing two canonical pipelines: Fit-Then-Protect (FTP), which smooths trajectories before privatization, and Protect-Then-Fit (PTF), which privatizes observations at discrete time points before smoothing. When data are observed without error, we prove that FTP achieves strictly smaller mean squared error than PTF asymptotically (as the sample size increases) under the same $μ$-Gaussian differential privacy ($μ$-GDP) budget in both finite-dimensional Hilbert spaces and infinite-dimensional reproducing kernel Hilbert spaces (RKHS). Simulation studies support our theory and explore settings with observation error and fitting constraints, demonstrating that FTP outperforms PTF under small observation noise, whereas PTF can become preferable in high-noise settings. In addition to this comparison, we establish a novel GDP guarantee for Gaussian perturbations taking values in a Banach space, extending existing results beyond the RKHS setting.

0.16

Cutpurse: Decomposing Explanation-Guided Model Extraction on Tabular Data

cs.CR cs.LG ·2026-11-06 ·Seyma Selcan Magara, Hanru Zhang, Fatme El Zahraa Hamad, Nico Pfeifer, Mete Akgün
Abstract

Explanation APIs are often treated as increasing model-extraction risk, but an end-to-end attack comparison cannot show whether an explanation reveals model-specific information or merely supplies a convenient next-query coordinate. We study this distinction for explanation-guided extraction of tabular classifiers. We separate an explanation into an attribution channel, which ranks influential features, and a coordinate channel, which supplies candidate feature values for perturbing those features. We then test whether the useful bin edges are privileged to the service or reproducible from the adversary's own data. Across six datasets and five target families, using attribution rankings to choose which features to perturb provides no systematic extraction advantage. Lime's discretization bin edges can improve fidelity by up to 41 percentage points on continuous features, but those edges are quantiles of the explainer's background data rather than properties of the target model. In controlled comparisons, an attacker-side grid reproduces the useful bin-edge gains without querying an explanation API. We use these findings to derive Cutpurse, which combines a self-computed grid with diverse query generation and boundary search. Against a service that withholds bin edges, Cutpurse improves fidelity by 16.3 percentage points on continuous-feature targets. It also exceeds the undefended published attack by 7.0 percentage points. Thus, withholding default Lime bin edges alone does not prevent effective extraction under the in-domain auxiliary-input, probability-output, and same-family surrogate assumptions studied here. Our results concern default Lime and Shap configurations for tabular classifiers. They do not establish the same conclusion for other explainers, target classes, or data-free attackers.

0.15

Differentially Private GraphRAG: Protecting Relation Connectivity and Semantics

cs.CR cs.AI ·2026-11-06 ·Jie Fu, Yuan Hong, Yue Ning, Wendy Hui Wang
Abstract

Graph Retrieval-Augmented Generation (GraphRAG) connects information across documents through entity relationships to support relational reasoning. However, black-box access to GraphRAG systems can enable adversaries to reconstruct knowledge graphs and recover both structural connectivity between entities and the semantics of their relationships. Existing defenses lack formal privacy guarantees and provide limited protection against such reconstruction attacks. We propose RelGuard, the first differentially private framework for GraphRAG, which provides formal privacy guarantees by mitigating connectivity leakage and protecting relations and their semantic meaning through relation removal. We formalize relation differential privacy (relation-DP), which bounds the distinguishability of mechanism outputs according to the semantic distance between alternative relations while accounting for relation removal. Building on this definition, we develop a four-step defense mechanism that sanitizes source documents before GraphRAG indexing by perturbing relation predicates or removing their associated connectivity using the exponential mechanism. Experiments across three GraphRAG systems and two targeted reconstruction attacks demonstrate that RelGuard provides stronger protection against reconstruction attacks than four existing defense mechanisms while better preserving downstream utility.

0.13

Who's Asking? Audience-Conditioned Privatization

cs.CR cs.AIcs.LG ·2026-11-06 ·Daniil Filienko, Anusha Bagalkotkar, Xan Vongsathorn, Sikha Pentyala
Abstract

The abundance of user-generated conversational data makes the question of how to responsibly derive insights from confidential data without putting the individuals in it at risk ever more important. However, the majority of existing differentially private (DP) synthetic text generation (SDG) and anonymization approaches ignore the context of the data's downstream usage, which results in loss of important information. DP-SDG provides the same guarantee regardless of the recipient, and anonymization relies on predefined attribute categories. They both ignore the observation that an attribute that is harmless to one audience can be sensitive to another, and each audience typically needs only some attributes preserved. We propose Ext-Classify-Priv, a framework that (1) extracts the information types present in a corpus, (2) classifies which of them are private for a given audience following contextual integrity, and (3) protects only those attributes, using either anonymization or SDG with Feature-DP approaches. On Reddit data with 3 audiences, our audience-specific releases are more faithful than a single audience-agnostic release for every audience. On bioRxiv data, our SDG method improves MAUVE from 0.097 to 0.420 at epsilon=1 and on the proprietary real estate data, our anonymization method improves MAUVE from 0.698 to 0.959 over the best baselines.

0.12

GSynPEF: A Privacy Evaluation Framework for Synthetic Graphs

cs.CR cs.AIcs.SI ·2026-11-06 ·Yu Duan, Luca Baldesi, Carter T. Butts, Athina Markopoulou
Abstract

It is often desirable to generate and release synthetic graphs, which resemble real-world data, in a privacy-preserving way. While substantial progress has been made in designing private graph synthesizers (PGS), opportunities remain for more explicit connections between synthesizer models and privacy evaluation, and better accounting for output variation. This paper proposes GSynPEF, a practical framework for empirical evaluation of privacy and utility in settings for which a real-world graph is given as input and output synthetic graph(s) are observed. Our framework links a broad class of graph synthesizers (both statistical and deep-learning based) with privacy metrics, expanding evaluation dimensions to practical utility and risk, through empirical analysis of multiple actual outputs. The framework accommodates both an Exact Solution scenario, where synthesizer behavior is directly observable, and a Monte Carlo Approximation scenario where only synthetic outputs from a black box synthesizer are available. Extensive experiments on four real-world datasets and seven widely used graph synthesizers demonstrate framework's versatility, enabling privacy evaluation for different attack goals (e.g., to infer an edge or node) and capabilities (e.g., observing one or multiple synthetic graphs). We analyze evaluation results by examining the relationship between graph statistics and synthesizer design choices, providing insights for data curators in PGS selection and researchers for improved PGS design.

0.09

Beyond a Single Commit: Anchor-Guided Reconstruction of Multi-Commit Patches

cs.CR ·2026-11-06 ·Jialiang Dong, Willy Susilo, Xu Gai, Liuhuo Wan, Chao Zhang +2 more
Abstract

Vulnerability patch identification plays a critical role in software security by linking reported vulnerabilities and the corresponding fixes, enabling reliable vulnerability analysis, mitigation, and automated security tool development. As software systems evolve in complexity, vulnerability remediation has increasingly shifted from single-commit fixes to multi-commit fixing processes. Although many studies have been proposed to identify fixing commits associated with target vulnerabilities automatically, they often suffer from inaccurate semantic analysis or excessive workload for language models. These limitations reduce both the effectiveness and comprehensiveness of fixing commit identification. To improve patch identification, we propose an effective and comprehensive patch identification framework, HARPOON, which not only enhances the accuracy of identifying individual fixing commits, but also investigates dependencies among commits to achieve comprehensive commit recognition for target vulnerabilities. Experimental results show that HARPOON significantly outperforms the state-of-the-art approaches. HARPOON successfully identified 895 vulnerability patches that were not previously maintained in public vulnerability databases. Among them, 663 are previously unrecorded fixes associated with 323 CRITICAL- or HIGH-severity vulnerabilities.

0.09

MemoryDojo: Benchmarking Poisoning Through LLM Agent Memory

cs.CR cs.AIcs.CL ·2026-11-06 ·David H. Yang, Blake Bullwinkel, Amanda Minnich, Mohammad Mohammadi Amiri, Giorgio Severi
benchmark
Abstract

LLM agents benefit from using memory systems to improve performance on long-horizon, multi-session tasks. However, memory introduces a critical attack surface. Malicious instructions encountered in one session may be stored and propagated through memory before being retrieved and acted upon in later sessions. Because memory systems differ in how they process and retrieve information, the downstream effects of poisoning can vary significantly across systems. To quantify the effects of poisoning across different memory architectures, we introduce MemoryDojo, a benchmark for tracing the lifecycle of poisoned information through agent memory and measuring its influence on subsequent agent behavior. MemoryDojo is easily extensible, consists of 155 multi-session long-horizon trajectories covering four domains, and supports five different memory systems. Through our evaluation, we find that all memory systems are susceptible to memory poisoning, with substantial variation in downstream effects. We further propose a memory-aware poison optimization process that increases the poisoning efficacy, highlighting differences between memory architectures. We hope that MemoryDojo will support research on understanding and mitigating agentic memory risks.

0.06

Systems & Networking 34

Age of Information for Power-Saving Devices with DRX Mechanism

cs.NI cs.PF ·2026-11-06 ·Hung-Chun Lin, Kuang-Hsun Lin, Hung-Yu Wei
publishedtheoryhighly cited
Abstract · 6 pages, 4 figures, 1 table. Published in IEEE ICC 2023

Reducing the power consumption of User Equipment (UE) has always been a crucial issue in communication systems. As a result, Discontinuous reception (DRX) is introduced in today's New Radio (NR) networks as a power-saving strategy. In addition, the variety of low latency applications in 6G makes the issue of maintaining data freshness increasingly important. This paper aims to study the relationship between freshness and power-saving. To capture the information freshness, we use Age of Information (AoI) as the performance metric. To our best knowledge, unlike previous studies focusing on power saving and delay trade-offs, this is the first study considering the impact of DRX on AoI. We solve the closed-form solution of the peak average AoI with the DRX mechanism. Moreover, we propose a parameters adjustment strategy. Based on the simulation validation, our proposed strategy can enable the UE to keep the AoI close to the theoretical lower bound AoI while achieving high power-saving efficiency.

0.64

Computation-Efficient Aerial-Marine Integrated Networks for Search and Rescue via Cooperative HAPS, UAVs, and MASSs

cs.NI ·2026-11-06 ·Zhen Wang, Bin Lin, Qiang Ye
top venuebenchmarkhighly cited
Abstract

In this paper, we propose an aerial-marine integrated network architecture tailored for maritime search and rescue (MSAR) operations, where a high altitude platform station (HAPS) and maritime autonomous surface ships (MASSs) are equipped with edge servers, providing computing services for surveillance uncrewed aerial vehicles (UAVs). These UAVs generate substantial volumes of computation-intensive and delay sensitive tasks, some of which can be strategically offloaded to both MASSs and HAPSs for distributed processing. We formulate the joint task offloading and resource allocation as a mixed-integer nonlinear program (MINLP) to minimize the system computation overhead (CO), quantified by a weighted sum of task completion time and energy consumption. Given the stochastic nature of task arrivals from UAVs and the dynamic characteristics of marine channel conditions, we employ a two-stage optimization to decompose the problem into four tractable subproblems, i.e., 1) edge server selection, 2) transmission power allocation, 3) edge computing resource allocation, and 4) local computation resource allocation. In stage I, we optimize the edge server selection via many-to-one matching. In stage II, we determine the transmission power of each UAV with quasi-convex optimization and solve the edge computing and local computing resource allocation subproblems via a projected gradient descent method and convex optimization, respectively. Simulation results validate that the proposed scheme can achieve superior system performance over benchmark schemes.

0.56

DARIC: A Position-Independent KV Cache Reuse System with Distribution-Aware Recomputation

cs.DC ·2026-11-06 ·Weikang Wang, Haiyang Liu, Xin Zhou, Jun Liu, Weifeng Zhang
top venuecode/data
Abstract · Accepted at SIGMOD 2027

The context lengths of large language models (LLMs) are rapidly increasing in real-world applications such as agentic systems and retrieval-augmented generation (RAG). To accelerate LLM inference in these scenarios, reusing position-independent key-value (KV) caches across requests has emerged as a promising approach. This technique typically relies on recomputing a selected subset of tokens to restore cross-chunk interactions. However, existing solutions suffer from critical limitations in both algorithmic efficiency and system deployment. Algorithmically, they overlook the distribution of KV cache deviations, leading to either high token-selection overhead or degraded generation quality. At the system level, the reliance on custom LLM execution flows for recomputation complicates deployment and prevents seamless integration with highly optimized inference engines. In this paper, we propose DARIC (\textbf{D}istribution-\textbf{A}ware \textbf{R}ecomputation for \textbf{I}ndependent KV \textbf{C}ache Reuse), a novel KV cache reuse method based on system-algorithm co-design. At its core, DARIC introduces a distribution-aware token selection strategy that adaptively schedules token recomputation based on empirical KV deviation patterns, striking an optimal balance between generation quality and computational efficiency. Moreover, DARIC is built on a lightweight system architecture that enables independent KV cache generation and recomputation with minimal integration overhead, fully preserving existing system-level optimizations of standard inference engines. Comprehensive experiments across various datasets and LLMs demonstrate that DARIC consistently outperforms state-of-the-art methods, achieving a 1.1$\times$--4.5$\times$ speedup in time-to-first-token (TTFT) with negligible loss in generation quality. Code is available at https://github.com/Lenovo-ICI/DARIC.

0.44

More with Less: Efficient Parallel Scientific Code Translation via an Agent-Skill Workflow

cs.DC cs.AIcs.SE ·2026-11-06 ·Jianuo Shen, Pengyi Xu, Bo Zhang
publishedcode/data
Abstract · Accepted to AgenticAI4HPC 2026 (SC26 Workshops). 10 pages

Translating parallel scientific codes between HPC programming frameworks is a critical but labor-intensive task. Existing LLM-based approaches rely on fixed sequential pipelines with undifferentiated repair strategies and correctness criteria that do not match scientific computing practice. We present an agent-skill framework that separates translation, compilation diagnosis, and execution validation into three autonomous agents coordinated by a verdict-driven orchestrator, routing each failure to a targeted repair skill rather than retrying blindly. An automated output check drives the self-correction loop, complemented by an offline LLM-based verification that exposes false positives escaping compiler and runtime checks. We evaluate the framework on 30 HeCBench kernels translated bidirectionally among CUDA, OpenMP, and SYCL with three open-source LLMs. Our framework attains the higher verified correctness rate in all 18 model and direction combinations over a state-of-the-art baseline while consuming up to 52.4% fewer tokens, remaining consistent across models and directions without architectural modification.

0.43

Adaptive Age of Information Optimization in Rateless Coding-Based Multicast-Enabled Sensor Networks

cs.NI cs.PF ·2026-11-06 ·Hung-Chun Lin, Kuang-Hsun Lin, Hung-Yu Wei
published
Abstract · 20 pages, 11 figures, 3 tables. Published in IEEE Journal of Selected Areas in Sensors, 2024

As the demand for real-time information in Internet of Things (IoT) and Wireless Sensor Networks (WSN) scenarios grows with the evolution of bandwidth-intensive 5G applications, multicast transmission becomes increasingly vital. This paper delves into the significant role of multicast in WSN, exploring a novel perspective of using rateless codes over the traditionally employed Hybrid automatic repeat request (HARQ) for eliminating retransmissions in a wireless multicast system. We aim to optimize data freshness, quantified by the Age of Information (AoI), and analyze strategies to minimize time-average AoI in a multicast environment with diverse channel conditions. Specifically, our policy focuses on optimizing the time-average AoI based on sensor devices' feedback. We transform the problem into a Markov Decision Process (MDP) to locate optimal and low-complexity suboptimal policies. We present the first age-minimum scheme for rateless code-based wireless multicast systems. Our numerical simulations reveal that the proposed policies, developed considering unique system structural properties, consistently surpass baseline strategies. We are thus able to preempt updates at the most beneficial time, thereby addressing the issue of the bottleneck device's adverse impact on overall performance.

0.42

LoStBench: Benchmarking Longitudinal Studies Data Integration

cs.DB ·2026-11-06 ·Pratik Pokharel, Oliver Kennedy, Abhi Ramtel, Poonam Kumari, Jeff Good +3 more
benchmarkcode/data
Abstract

A longitudinal study observes the same subjects or variables, repeatedly over an extended period of time, allowing researchers to examine changes, trends, and relationships as they develop. Such studies typically evolve over their life cycles, as improved understanding of the study's domain, or external factors trigger the need for revisions. The resulting heterogeneity among collected data must be resolved before the survey's results may be analyzed. Although this problem is a form of schema matching, it is a form of schema matching that differs significantly from prior workloads explored by the database community. For instance, rather than the short attribute names and data instances that many schema matching tools rely on, a stronger signal in longitudinal studies arises from longer prose questions that these tools are less equipped to handle. This paper documents the joint efforts of a team of computer scientists and domain experts to adapt existing schema matching algorithms to our new use case of longitudinal data integration. We present a case study in the context of a representative workload: The American National Election Studies (ANES) dataset, a nearly eight-decade survey with publicly available data and a canonical domain-expert curated schema matching. Using ANES, we develop LoStBench (Longitudinal Studies Benchmark for data integration), a medium-scale, subject-matter-expert-annotated benchmark. Using LoStBench, we conduct a comprehensive empirical study evaluating state-of-the-art schema matching algorithms, and expose limitations of these algorithms when applied in the context of longitudinal studies. We propose improvements that address some of these limitations, and identify key challenges to motivate future data integration work to address the others.

0.37

Circuit Theory-Informed AI for Condition Monitoring in Power Electronics

cs.ET cs.AIcs.LGcs.NE+1 ·2026-11-06 ·Yubo Song, Pascal A. Schirmer, Subham Sahoo
published
Abstract · The manuscript has been accepted for publication in the Proceedings of 2026 IEEE Design Methodologies Conference (DMC 2026)

Artificial intelligence for power electronics is evolving toward a deeper comprehension of system intrinsic behaviors and dynamics. As a novel integration of neural computation and electric circuits, spike-based neuromorphic computing offers promising potential in high energy efficiency and adaptability for data-driven applications. Building on the latency-coding principles, this article proposes a framework for data-light condition monitoring that relies solely on three-phase AC current measurements. The AC current signals are encoded as membrane potentials of RC-based spiking neurons, which are inherently compatible with spiking neural networks that can effectively facilitate the estimation of device parameters. Case studies are presented to validate the performances, while constraints and further optimizations are also evaluated and discussed, and future extensions of the methodology are provided.

0.36

EventHOI: Event-Driven Resource-Efficient Hand-Object Interaction Logging on Smartglasses

cs.DC cs.CVcs.HC ·2026-11-06 ·Le Zhang, Kailai Cui, Hao Chen, Vlad Roznyatovskiy, Jianzhong Zhang +3 more
benchmark
Abstract · 15 pages, 21 figures, 3 tables

Hand-object interactions (HOIs) can be captured by smartglasses to learn how people manipulate objects in daily life. Automatic logging of these interactions from an egocentric perspective provides time-stamped evidence of daily actions for applications like memory assistance. However, continuous RGB capture and wireless transmission exceed the energy and bandwidth budgets of smartglasses. We present EventHOI, an event-driven sensing and compression system that couples selective RGB capture with HOI-aware encoding to enable egocentric logging over Bluetooth Low Energy (BLE) while preserving downstream HOI logging performance. Our key insight is to use low-power event sensing both to determine when RGB capture is needed and to identify where visual detail must be preserved. A lightweight IMU-based egomotion cancellation pipeline suppresses egomotion-induced events so that residual HOI motion can guide RGB triggering and spatial compression. We build a smartglasses prototype and conduct an end-to-end evaluation of EventHOI on a 50-hour egocentric benchmark with 97 subjects and a 20-hour real-world dataset we collect from 10 participants, together covering a broad range of daily-life scenarios and environments. Compared to RGB-only baselines, EventHOI reduces always-on backend power consumption by 5.8-10.4x and uplink bandwidth by 2.9x, with minimal HOI recognition accuracy loss. The resulting RGB uplink of less than 130 kbps fits within commodity BLE bandwidth budgets, making stable BLE video streaming feasible for egocentric HOI logging. Demo video and project website: https://eventhoi.github.io

0.35

Inside the Trap: A Renewal Law for Prefix-Cache Collapse in LLM Agent Serving

cs.DC cs.PF ·2026-11-06 ·Akshat Valse, Ali Jannesari
theory
Abstract · 66 pages, 8 figures, 20 tables

Prefix caching lets an LLM agent's follow-up turns reuse cached KV, but a burst can tip an engine into a trap in which every turn misses and the backlog does not drain, at a load it served before. In registered paired tests on two engines, two GPUs and two models (passed at 9 of 10 rates), a 200-session burst left the engine at zero hits to the end in all 30 runs, while every matched no-burst run stayed good and a session cap prevented the trap in all 30 capped runs. We analyse an idealization of the trapped state exactly: for a fluid FIFO server on which every turn misses, we prove that along FIFO orbits its cycles obey a linear renewal recursion. As is classical, the trap drains if and only if arrivals are below the all-miss capacity; above it, a Perron root fixes the dynamics: every trapped state converges to a known self-similar regime in which the gap between turns grows linearly and the turns served are young and short. For our workload the backlog grows at 0.6 to 0.8 of the excess load and sessions complete faster than the all-miss capacity, a youth premium with turn-count and cost-growth factors, so in the model a trap's measured throughput overstates the recovery threshold. Retrospectively, with nothing fitted to trap data, the theory matches 57 stuck runs: the mean served prompt within 3% in 50 runs, completions above the all-miss capacity in every run, and the pooled gap growth. Blind in outcome, 4 runs trapped by a tool slowdown met 4 of 5 pre-set scores, cut traps persisted, and with a 20-step budget, trapped runs completed below the all-miss capacity, as predicted. Of two exits, refusing arrivals sheds sessions; a session cap restores reuse: in registered tests it rescued 9 of 9 established traps at the same load without shedding, and the engine stayed good after release. We do not prove that the engine converges to the model.

0.35

Capital Efficiency, Isolation, Risk, and Ruin in Generative AI Inference Serving Infrastructure

cs.DC cs.AI ·2026-11-06 ·Renée St. Amant, Anjaly Parayil, Ankur Mallick, Victor Rühle
Abstract

Generative AI serving is an extremely capital-intensive endeavor that demands rapid, step-change improvements in capital efficiency to prevent industry wide increases in the cost of capital from limiting growth plans. Keeping pace with the economic necessities will increasingly require infrastructure providers to confront operationally-inconvenient realities of production serving, including the extreme heterogeneity of generative AI request costs. We have explored workload isolation as a foundational, fleet-level lever for increasing capital efficiency. Isolation is the practice of intelligently grouping "like" requests before execution. Rather than exposing every replica to the full heterogeneous workload distribution, isolation deliberately reshapes the workload presented to the hardware by assigning requests with different cost characteristics to different portions of the fleet. Workload isolation on its own does not make expensive requests cheaper; it prevents a costly portion of the workload from unnecessarily determining the operating point and capital efficiency of the entire fleet. It additionally lays the foundational groundwork that can serve as a launching point for compounding gains via narrow-band optimizations (tailoring) and custom silicon designs while also informing up on pricing policy. This paper uses representative production data and experience to convey the motivation behind isolation as a fleet-level lever for increased capital efficiency.

0.34

Quantum Integration Software Suite For HPC-Quantum Computing Workflows

cs.ET quant-ph ·2026-11-06 ·Mateusz Meller, Joseph C. R. Thacker, Thomas W. Keal, Vendel Szeremi
Abstract · 15 pages, 3 figures, 11 tables

In this work, we focus on addressing challenges related to modelling the integration of conventional high-performance computing (HPC) and quantum computing (QC). We introduce the Quantum Assembly Compiler (QASC) for OpenQASM3.0 files. The compiler produces an executable that leverages the high-performance simulator QuEST to simulate quantum programs. Additionally, we introduce a performance modelling tool, QACC-perf, which can estimate the cost of integrating QC in classical HPC workflows, as well as the execution time of the quantum circuits. Finally, we demonstrate the utility of these tools in the context of multiscale quantum chemistry workflows using the ChemShell computational chemistry environment.

0.33

UniExit: A Unified Latency-Aware Early-Exit Runtime for Heterogeneous Edge AI

cs.DC ·2026-11-06 ·Yuhang Guan, Huanle Zhang, Xiaomin Ouyang, Zhidan Liu
published
Abstract · 15 pages, 20 figures. Accepted to ACM SenSys 2027 (The International Conference on Embedded Artificial Intelligence and Sensing Systems)

Edge applications increasingly combine CNN-based perception with edge-deployable Transformer-based language models, but deploying such heterogeneous workloads on resource-constrained devices is challenging because inference latency varies with both input difficulty and runtime hardware state. Early-exit inference reduces computation by terminating at intermediate stages, yet existing methods are often model-specific and rely on static latency profiles that become stale under dynamic edge conditions. This paper presents UniExit, a unified latency-aware early-exit runtime for heterogeneous edge AI. UniExit abstracts CNN blocks and Transformer layers as ordered stages and applies a common latency-aware routing policy while retaining model-specific quality mechanisms. It predicts the cost of reaching the next exit from runtime state and inference progress to make budget-aware continue-or-exit decisions. UniExit combines runtime-state-aware latency prediction, cache-based prediction reuse, trend-aware fallback, and CNN-specific exit stabilization to provide low-overhead and robust online control. We implement UniExit on multiple edge and edge-cloud platforms and evaluate it with representative CNN and Transformer-based language-model workloads. Across single-workload, concurrent heterogeneous, and edge-cloud settings, UniExit consistently improves deadline satisfaction and latency-accuracy trade-offs over existing progressive inference baselines, while reducing redundant exit-module computation and task-level energy consumption.

0.32

TwinHash: Two-Sided Object Identity Management for Time- and Energy-Efficient Agentic Workflows

cs.AR ·2026-11-06 ·Sangbu Yun, Jung Gyu Min, Jin Hyun, Jeonghwan Ahn, Youngjoo Lee
top venuebenchmark
Abstract · Accepted at IEEE HPCA 2027

Agentic artificial intelligence (AI) services interleave large language model (LLM) generation with external tool execution, and increasingly rely on search-based workflows that repeatedly revisit shared reasoning states and computations. These patterns create substantial redundancy across tool outputs, LLM responses, and key-value (KV) prefixes, yet existing serving systems primarily optimize shared prompt prefixes, while naive input-keyed caching is unsafe when identical requests produce different outputs. We present TwinHash, a unified object management architecture that assigns each object an input-side hash key to recognize repeated requests and an output-side content fingerprint to validate and locate stored state, complemented by type-specific reuse guards that preserve correctness and response diversity. A software object identity table maintains reusable objects across requests and trajectories, while a hardware object placement table tracks their residency across hierarchical memory tiers so that evicted state can be retrieved rather than recomputed. Across seven agent frameworks and five benchmark suites, reuse follows workflow structure more than benchmark identity. TwinHash serves up to 81% of tool outputs wherever deterministic tool reads recur, and up to 92% of LLM responses in search-based frameworks. Retrieving an evicted KV-prefix from DRAM reduces TTFT by up to 5.6x on Qwen3-32B, with greater gains projected for larger dense models. Under search-based frameworks, TwinHash reduces end-to-end wall-clock time by up to 87% and measured task energy by up to 81% over a prefix-cached baseline, while a per-key serving cap controls the quality-reuse trade-off in adaptive search. These results demonstrate that unified identity, safety, and placement management can convert structural redundancy in emerging agentic workflows into substantial serving-efficiency gains.

0.31

Memory as a Temporal Resource: Lifetime Contracts for Multi-Retention Systems

cs.AR ·2026-11-06 ·Adrian Jackson
position
Abstract

Computers move and rewrite data whose useful lifetimes range from inner-loop iterations to years, yet memory systems usually hide retention as a fixed physical property. We argue that required lifetime should become a first-class resource across algorithms, programming systems, controllers, and devices. Building on prior lifetime-aware placement, we propose an event-based validity contract that links application completion events to retention-class selection and renewal of protection. Applications would be enabled to express semantic contracts such as ``until this tile completes'' and the system will map data to homogeneous arrays engineered for ephemeral, epoch, or durable retention and preserve correctness through margins, ECC, refresh, migration, or recomputation. We present a trace-based simulator and FPGA prototype that exercises class migration.

0.31

Logical Scalability in Big Data Architectures: Impact of Data Modeling on the Database Layer

cs.DB ·2026-11-06 ·Rosario Napoli, Mark Adrian Gambito, Antonio Celesti, Maria Fazio
published
Abstract · This is a preprint. The final version is published in 2026 IEEE 26th International Symposium on Cluster, Cloud and Internet Computing Worksh…

In modern Big Data architectures, scalability is often addressed through infrastructural expansion, involving vertical or horizontal scaling to manage growing data volumes. However, this hardware-centric perspective overlooks the logical scalability of the storage layer. A central challenge lies in how data modeling choices affect the storage layer of big data architectures. Specifically, when the data model is misaligned with the underlying query patterns, performance bottlenecks emerge and persist regardless of the available computational resources. This paper investigates the concept of logical scalability by conducting a comparative performance analysis of Document, Graph and Multi-Model database paradigms under varying workloads, queries and dataset scales. Our results demonstrate that performance degradation caused by model misalignment cannot be mitigated by additional computational resources alone. Increasing the number of machines while retaining a structurally inadequate data model yields performance equivalent to or worse than a single-machine configuration, as distributed coordination overhead compounds the inefficiencies of the underlying logical model. We demonstrate that efficiency is maximized when the logical structure of the database aligns with the inherent nature of the data and the intended analysis scope, a condition that proves to be more decisive than any infrastructural scaling strategy.

0.31

Concurrent Use of FP32 and FP64 Computing Units on GPUs and Mixed-Precision Placement in WENO Reconstruction

cs.DC ·2026-11-06 ·Jun Hatayama, Daichi Mukunoki, Yusuke Onishi, Atsushi Sekimoto, Toshinori Kouchi
Abstract

Scientific computing often requires FP64 arithmetic, although FP64 performance is limited on NVIDIA GPUs. Double-float (DF) emulation, which represents a value using two FP32 values with about 48 bits of significand, outperforms native FP64 where FP64 units are limited. On the other hand, mixed-precision strategies offload to FP32 the operations that do not determine accuracy. In both approaches, the native FP64 units remain idle during FP32 instruction execution. We investigated whether this idle resource can be recovered on A100, GH200, GB200, and RTX PRO 6000 Blackwell Max-Q. A microbenchmark shows that co-resident FP32 and native-FP64 kernels exceed the throughput of the faster kernel alone on A100 but not on GH200 and GB200; on RTX 6000 Blackwell, they do so against FFMA but, against DF emulation, only when the issue density of the DF kernel is reduced. We then tested whether this gain carries over to computing the WENO-Z reconstruction with the nonlinear weights in FP32 and the candidate polynomials in native FP64 (W32P64). We observe substantial overlap on A100, whereas GH200, GB200, and RTX 6000 Blackwell show little overlap. Importantly, the fastest placement does not always preserve the design accuracy of WENO-Z: for a one-dimensional entropy wave problem, W32P64 fails to achieve the design order of accuracy for WENO7 and WENO9. In contrast, DF emulation of the entire reconstruction comes close to it on the grids tested for WENO5, WENO7, and WENO9 while delivering a $2.51-3.94 \times$ speedup over FP64 on RTX 6000 Blackwell. For WENO5 on the two-dimensional Euler vortex problem the W32P64 error at $n=256$ is $1.26 \times$ that of FP64 and 1/104 of that of the all-FP32 reconstruction.

0.30

Adaptive Scheduling Pipeline for Multi-Instance Asynchronous Reinforcement Learning

cs.DC cs.AIcs.LG ·2026-11-06 ·Salah Chikhi, Luis Zuin, Entong Li, Zeng Li
top venue
Abstract · Accepted at NeurIPS 2026

We present the Adaptive Scheduling Pipeline (ASP), an offline rollout scheduler for multi-instance asynchronous reinforcement learning (RL) of large language models (LLMs). ASP jointly reduces three sources of rollout underutilization - padding bubbles from intra-minibatch length heterogeneity, skewness bubbles from long-tail outputs, and sorting bubbles from fluctuations in training-queue occupancy induced by output-length-sorted dispatch - through a single offline preprocessing pass. This enables efficient multi-instance rollout that overcomes the early scaling plateau of single-instance inference while reducing fluctuations in the supply of generated samples to the trainer. We integrate the scheduler into an asynchronous architecture that decouples rollout from training while keeping policy lag observable and bounded. Within our asynchronous architecture, ASP's scheduling reduces mean iteration time by 10.9%. The complete system improves end-to-end throughput by up to 3.22 times and reduces iteration time by up to 76.8% relative to MindSpeed-RL, while reducing iteration time by up to 62.5% relative to verl. It also scales more robustly with additional generation resources than these baselines and shows no visible convergence degradation in the measured bounded-lag regime. ASP has since been used in an internal telecom-domain RL post-training workflow, reducing the reported training period from more than three months to two months and producing a domain-enhanced model for knowledge answering and tool calling.

0.29

An Empirical Evaluation of a Weighted PALU Model for Internet Traffic Networks

cs.NI ·2026-11-06 ·Wood Moore, Hayden Jananthan, Jeremy Kepner
unusual
Abstract · 5 pages, 1 figure, IEEE MIT URTC 2026

The PALU model is a hybrid power law generative framework that decomposes Internet traffic graphs into three structural components: a preferential attachment core, degree one leaf nodes, and disconnected satellite clusters. Although the original PALU formulation provided a theoretical basis for this decomposition, it was never validated against real traffic data and was restricted to unweighted networks. This work presents the first empirical evaluation of PALU against real Internet traffic, using packet traces from the MAWI WIDE samplepoint F dataset. We introduce a weighted extension of the model incorporating strength degree scaling, and develop a packet-to-graph pipeline for parameter estimation across multiple observation window sizes. Results show that the power law degree exponent $α\approx 2.0$ remains stable across observation scales from $10^5$ to $3 \times 10^6$ packets, consistent with the PALU structural prediction. However, the satellite cluster parameter $λ$ and the weighted scaling parameter $θ$ exhibit systematic dependence on observation scale, and the Poisson star assumption for unattached components is formally rejected at all tested window sizes. These results indicate that PALU captures important aspects of Internet traffic structure but requires an explicit observation model to fully account for finite sampling effects.

0.27

Fast and Actionable Classification of HPC I/O Time Series

cs.DC ·2026-11-06 ·Maxime Gonthier, Luan Teylo, Francesco Antici, Mihail Popov, Francieli Boito
Abstract

Input/output (I/O) is a major bottleneck for highperformance computing (HPC) performance. One of the challenges is the contention between jobs when sharing the parallel file system. However, tackling this issue depends on knowing the temporal behavior of jobs. In this paper, we propose a clusteringbased approach to characterize jobs' temporal I/O behavior. We propose SAGE, a lightweight method that cuts I/O traces into short windows and learns an alphabet from them. The trace then becomes a sequence of symbols, which we call ''genome'', and we use clustering to separate them into classes. We evaluate SAGE and its two-stage variant SAGE-2 on 819,771 real traces from 2 HPC systems. We compare them against other techniques from the state of the art on three metrics, including one we propose. Compared to K-shape, the best baseline available, SAGE-2 lowers the median intra-cluster metrics by 9.1% on average, and achieves this in 2 minutes of clustering time versus 3.8 hours, a 109x speedup. More importantly, clusters created by SAGE and SAGE-2 carry enough information to act on: we evaluate the use of the cluster centroids for I/O scheduling, and show that our approach can decrease I/O stretch by up to 24% compared to the state of the art (round-robin placement).

0.27

Link-Level Performance Comparison of 5G Beamforming: Reciprocity-Based SVD vs. Codebook-Based Precoding

cs.NI eess.SP ·2026-11-06 ·S. M. Haider Ali Shuvo, Seda Dogan-Tusha, Monisha Ghosh
Abstract · 7 pages, 8 figures. Submitted to IEEE ICC 2027

Downlink (DL) precoding in mid-band MIMO relies on uplink (UL) sounding reference signals (SRS) with reciprocity-based singular value decomposition (SVD) or channel state information reference signals (CSI-RS) with codebook-based precoding matrix indicator (PMI) feedback. Although SVD precoding offers unquantized beamforming, it depends on a channel estimate from SRS transmitted at the limited user equipment (UE) power, whereas Type-I PMI is quantized but derived from a high-power CSI-RS based channel estimate. In this paper, we compare SRS-SVD and CSIRS-PMI for single-user MIMO through an end-to-end, 3GPP-compliant link-level simulation at 3.75 GHz with 100 MHz bandwidth, using a 192 antenna element gNB mapped into 32 ports, the 3GPP Release 19 UE antenna model, and ray-traced (RT) channels in Manhattan and Chicago. We show that neither scheme is uniformly superior: CSIRS-PMI serves weak and shadowed users better and reduces outage, whereas SRS-SVD achieves higher spectral efficiency (SE) and enables more multi-layer transmission for users with strong channels. The results on the Clustered Delay Line (CDL)-C statistical channel also confirm this crossover. We further show that by utilizing a simple logistic regression based on the received power of the standard CSI reference signal (CSI-RSRP) and the channel quality indicator (CQI), the gNB can select the optimal transmission scheme with around 90% accuracy, achieving an SE that closely approaches an ideal per-user selection.

0.27

Continual Instruction Tuning for LLM-based Management of Continuum Resources

cs.DC cs.LG ·2026-11-06 ·Lanpei Li, Antonino Vaccarella, Vincenzo Lomonaco, Massimo Coppola
Abstract · Accepted at the Workshop on Research and Innovation in Scalable and Efficient Deep Learning (RISE-DL 2026), co-located with the 16th Interna…

Large language models (LLMs) are increasingly explored for resource management across the Cloud-Edge continuum. However, embedding LLMs directly inside latency-sensitive control loops places inference latency, execution jitter, and hallucination risks on the critical path. This study argues for a different use of LLMs: a continually instruction-tuned intent compiler that runs once per management intent or declared context change, above a fast, deterministic tactical controller that handles all low-latency runtime actions. We formulate intent compilation as a continual instruction tuning problem, because management intents arrive sequentially and accumulate while infrastructure and workload contexts drift. Unlike recent LLM-based management frameworks, our design scores the model's outputs by execution, so that catastrophic forgetting can be measured in the executed objective of the underlying controller. On a continuum simulator with a 2B backbone and parameter-efficient adapters, preliminary results show that naive sequential tuning forgets earlier intents in the executed objective, that replay of stored examples nearly prevents this forgetting, and that the tuned compact compiler performs on par with simpler translators while being more robust than a grammar parser to rephrased intents.

0.26

Repeating Work Can Lower Energy per Successful Goal in LLM Agents

cs.PF ·2026-11-06 ·Deepak Panigrahy, Stephen Abkin, Aakash Tyagi, David Kebo Houngninou, Rabi Mahapatra
Abstract · 17 pages, 6 figures, 7 tables

When a temporary call failure interrupts a tool-using LLM agent, the framework must choose how much completed work to repeat. Repeating less completed work intuitively saves computation, but recovery mechanisms also change subsequent execution and the chance of success. We compare recovery mechanisms by energy per successful goal (EpG), which divides the energy of all task runs (including failed runs) by the number that finish correctly. In controlled vLLM experiments on one node, we measure board energy above idle for eight recovery options on reliably solved multi-turn tasks, with selected comparisons across three model families. For the model tested across all eight options, with the prefix cache retained, starting over has 22 percent lower EpG than continuing after early failures. Under the same cache condition, continuing has 24 percent lower EpG than starting over after late failures. Among all eight recovery options, repeating only the interrupted turn has the lowest observed EpG after early and late tool failures on this model. Holding recovery fixed, retaining the prefix cache reduces prompt tokens needing fresh computation by 86 percent but reduces recovery energy by only 8 percent at contexts up to 7,333 tokens. These measurements and a condition relating recovery energy and success show when repeating work lowers EpG, helping framework developers choose how their agents recover on their intended models and tasks.

0.26

ISER-MoE: Dynamic In-Switch Replication of Hot Experts for Load Balancing in Expert-Parallel

cs.DC ·2026-11-06 ·Jintao Peng, Ethan Zhang, Kefeng Duan, Bangyu Sheng, Min Li +3 more
benchmark
Abstract · 12 pages, 12 figures

Expert-parallel mixture-of-experts (MoE) layers run at the pace of the most loaded GPU, and a few hot experts carry most of the excess load. Replicating them at every MoE layer balances the load only if their weights arrive within the layer, and broadcasting from the owner is the bottleneck: point-to-point copies overload it, and baseline broadcasts pay latency or bandwidth overheads. We present ISER-MoE (In-Switch Expert Replication for MoE), built in TensorRT-LLM, which multicasts hot experts through the NVLink switch to fixed GPU groups, so each byte leaves its owner once. A GPU scheduler assigns hot experts to groups per layer, and a multi-port broadcast spreads the injection over the group's ports, both overlapped with the shared expert. In trace replay, the scheduler removed 98-99% of the excess load; in prefill serving of DeepSeek-V4-Pro, throughput rose 17.4% on a math benchmark and 7.2% on agentic prompts.

0.24

5G-NetBotSim: A Synchronized Co-Simulation Framework for Connected Robots over ROS and srsRAN-based 5G Network

cs.NI ·2026-11-06 ·Narges Golmohammadi, Bhaskara Venkata Raju Garuda, Ayan Joshi, Sabur Baidya
Abstract

Connected robotic applications increasingly rely on wireless communication for teleoperation, coordination, and distributed autonomy. However, robotic simulators often use ideal links or abstract delay models that do not capture cellular scheduling and protocol processing effects. We present 5G NetBotSim, a realtime co simulation framework integrating ROS/Gazebo workloads with an executable, full stack srsRAN 5G network emulator. The framework provides isolated User Equipment (UE) robot contexts and stage level timing across the host, gNB, UE, ROS, and robot execution stages. We evaluate human to robot teleoperation and robot to robot leader follower coordination under controlled channel quality conditions. Results show how CQI and MCS variations propagate to throughput, communication latency, and robot performance. In leader follower coordination, higher channel quality reduces mean tracking error from 0.558 m to 0.460 m, a 17.6% reduction. These results demonstrate protocol stack in the loop co simulation for joint network and robotic-application evaluation.

0.24

DataHelix: A CPU-GPU Co-Scheduling Data Plane for Multimodal Model Training

cs.DC cs.AI ·2026-11-06 ·Yuanhe Li, Jiazhen Wu, Dian Ding, Guangtao Xue, Yiming Zhang
Abstract

Multimodal model training breaks the loose coupling between data loading and model execution assumed by conventional training pipelines. Large per-sample variation in decoding cost, token/latent count, and tensor size causes both CPU preprocessing time and GPU compute time to vary substantially across data-parallel ranks (training processes). In synchronous training, these two sources of heterogeneity interact tightly: GPU execution determines how long the input pipeline has to prepare the next batch, while delayed preprocessing in any rank in turn stalls all GPUs at synchronization. As a result, optimizing CPU throughput or GPU load balance in isolation cannot eliminate pipeline bubbles. We present DataHelix, a metadata-first CPU-GPU co-scheduling data plane for multimodal training. DataHelix jointly reasons about preprocessing demand and GPU execution progress, treating each GPU compute window as the readiness budget for the next batch and coordinating preprocessing across ranks to minimize deadline misses rather than CPU makespan. To avoid expensive data movement, placement is decided while samples remain lightweight descriptors, before decoding and collation amplify them into large tensors. Descriptor-level token-latent-aware batching performs coarse pre-balancing, while cost-gated collective rebalancing corrects residual skew only when its expected benefit exceeds transfer cost. Across 11 video and multimodal workloads on 8 Ascend 910B accelerators, DataHelix reduces mean deadline-miss rate by 67.8% and mean p95 input-wait latency by 87.0%, while achieving 2.72x geometric-mean throughput over rank-local DataLoaders and 1.82x over a Ray actor pool under the same CPU budget.

0.22

Sparse Predict and Demote: Optimal Scheduling with Minimal Size Advice

cs.PF cs.DScs.LG ·2026-11-06 ·Chengyi Liu, Thodoris Lykouris, Junxuan Shen, Adam Wierman
Abstract

Knowledge of job sizes is crucial for optimizing the mean response time in single-server scheduling. However, predictions of job sizes are typically inexact and may be costly to obtain. Thus, a long-standing research question is: what is the minimal information about job sizes required to obtain near-optimal mean response time? We introduce a novel policy, \emph{Sparse Predict and Demote (SPD)}, and prove that it achieves order-optimal mean response time in heavy traffic while requiring only one-bit job-size predictions for a vanishing fraction of arrivals. SPD achieves heavy-traffic order-optimality even in situations when predictions are (i) inexact up to any fixed multiplicative error bound, and (ii) costly, i.e., the server is used to compute predictions and is thus unavailable to serve jobs while processing prediction queries. Finally, we use numerical experiments to illustrate that SPD maintains near-optimal performance beyond the heavy-traffic regime.

0.20

Green500: Adventure in Qualifying a Cluster

cs.DC ·2026-11-06 ·Nicolas Richart, Xavier Ouvrard, Julia Paolini, Manuel Cubero-Castan, Gilles Fourestey
Abstract · submitted to EESP Workshop 2026

Energy effciency has become a key metric for evaluating High Performance Computing (HPC) systems, as reflected by the Green500 ranking. However, energy effciency alone does not capture the full environmental impact of an HPC cluster. Using the Kuma cluster at EPFL, we report on two Green500 measurement campaigns at Level 2 and Level 3, and extend this analysis to a full carbon footprint assessment using the MITSI life-cycle based methodology. GPU power capping experiments show that energy effciency peaks at 450 W, achieving an 11% performance loss offset by a 20% reduction in energy consumption. When embodied emissions are accounted for, the optimal power cap for carbon effciency shifts to 500 W under Switzerland's low-carbon electrical grid, revealing a tension between minimizing usage and embodied contributions. We discuss the challenges of generalizing this approach into a broader carbon-based HPC ranking, and argue that moving beyond energy effciency is both necessary and feasible.

0.20

The World Model Planning Accelerator (WMPA): Hardware-Algorithm Co-Design for Embedded AI in Robotics

cs.AR cs.LGcs.RO ·2026-11-06 ·Prerana Ramkumar, Suhas Suresh Bharadwaj, Fazal Hameed
Abstract · 7 pages, 4 figures

Embedded AI is moving from perception to prediction. Robots now use world models$\unicode{x2014}$learned simulators that predict the outcome of candidate actions in a compressed latent space before any is executed. Every predicted step must be computed on board within a tight power and thermal budget, yet the energy these steps fundamentally require, and how far real hardware sits from it, have not been characterized. We derive lower bounds on the arithmetic operations and the CMOS dynamic energy of a latent state transition from information theory and Boolean circuit complexity. Profiling a DreamerV3-style Recurrent State Space Model (RSSM) shows that every rollout operation is memory-bandwidth bound and that a GPU consumes at least $5.9\times10^{10}$ times the CMOS energy floor. Transient 3D thermal simulations show that this inefficiency is also a thermal one, since sustained back-to-back planning heats a millimeter-scale silicon tile to 556 K even behind an ideal copper heat spreader. Guided by this chain from bound to bottleneck, we propose the World Model Planning Accelerator (WMPA), a chip architecture for embedded world model planning that combines near-memory computation, real-time entropy estimation, variable-precision rollout units, and copy-on-write latent scratchpads. With a design target of 40x lower power, the same tile stays near ambient at 306 K in simulation, pointing toward practical planning on thermally constrained robots.

0.20

Scalable Full-Horizon SFC Deployment over Predictable SAGIN Topologies via Configurable Column Generation Framework

cs.NI ·2026-11-06 ·Hang Li, Hao Huang, Yihu Xu, Zixuan Kong, Yanling Jin
Abstract · 18 pages, 10 figures, 4 tables

Space-air-ground integrated networks (SAGINs) integrate heterogeneous communication and computing resources to support ubiquitous services. Service function chains (SFCs) enable complex services to be composed of ordered functions and executed collaboratively across these resources. Mobility continuously reshapes connectivity throughout service execution, yet much of the resulting topology evolution is predictable. Decisions based only on isolated topology snapshots may therefore overlook valuable future connectivity. We formulate an admission-first, cost-aware model on a time-expanded graph for SFC deployment. Our key observation is that temporal feasibility is request-specific, whereas global coupling arises mainly from shared-resource competition. We therefore develop a configurable column-generation framework in which complete request deployments form columns and a restricted master problem coordinates admission and resources. SAGINFE, a fast constructive heuristic built on per-flow embedding (PFE), supplies a standalone feasible solution and initial columns. Configurable pricing schedules combine early-terminated and stricter searches, supplemented by selective PFE candidates. Candidates undergo feasibility and exact reduced-cost checks. Request-wise pricing can run in parallel, followed by finite-pool integer selection. The default configuration improves the admission-first objective over initialization on all 120 main instances. Four pricing workers achieve up to a 2.69-fold median pricing-stage speedup while preserving admitted request counts in the tested cases.

0.20

FirstSwap: Pre-Consensus Trade Approvals for Decentralized Market Design

cs.DC cs.CRcs.GT ·2026-11-06 ·William Diamond, Fatima Elsheimy, Dahlia Malkhi, Geoff Ramseyer, Srivatsan Sridhar
Abstract

Decentralized exchanges typically give traders no execution guarantee until a block is finalized. In the meantime, external prices move, traders strategically delay submission, and block producers can influence inclusion and ordering. We introduce FirstSwap, a DEX built on a constant-function market maker pricing curve that gives traders an early pre-approval before consensus, guaranteeing inclusion and execution within their submitted price limit, with final ordering determined by limit prices rather than by a block producer. We demonstrate the benefits of trade pre-approvals in an economic model combining auction theory with CEX-DEX arbitrage. When the CEX price changes gradually, FirstSwap rewards earlier traders with better prices, discouraging strategic delay and yielding higher social welfare and gains to AMM liquidity providers than a batch auction. When the CEX price jumps abruptly, FirstSwap yields higher welfare and smaller AMM losses than fair ordering. The central challenge is delivering fast trade pre-approvals without consensus. We introduce a decentralized, scaled admissibility protocol under which each validator reserves only a fraction of the available liquidity at every price. A trader obtains approval in a single round of messages with a quorum of validators. Somewhat surprisingly, despite Byzantine validators who may approve trades arbitrarily, we show that all quorum-certified trades are jointly admissible and execute within their limits. Our protocol extends naturally to a setting where buy and sell orders can be pre-approved against each other before a block is finalized.

0.19

The Storage View Of KV Cache: Characterizing Cache Behavior Across LLM Workloads

cs.OS ·2026-11-06 ·Talia Q. Chen, Daniel Waddington, Narangerelt Batsoyol, Vijay Chidambaram
Abstract

Modern LLM serving increasingly extends KV caches beyond GPU memory into host DRAM, SSDs, and remote storage, yet these stores remain largely governed by workload-agnostic policies such as LRU and fixed TTLs. We ask what execution context is needed by storage to make retention, reclamation, and placement decisions across heterogeneous workloads. We characterize KV-cache reuse, lifetime, and footprint at block granularity across 11 production and public traces spanning chat, API serving, RAG, coding agents, tool-using agents, and multi-lineage workflows. We examine reuse sources, recurrence timing and frequency, useful lifetimes, and storage consumption. Across workloads, KV state takes two recurring forms. Single-session state is reused within one execution: its footprint grows as sessions deepen and accumulates across concurrent sessions; it can remain useful through long idle periods but loses value at session completion. Multi-session state recurs across sessions of the same workload, outlives individual sessions, and spans hot prefixes and a large infrequent tail. While prior systems often exploit application-, agent-, or workflow-specific knowledge, our results suggest that three generic signals (a workload identifier, a session identifier, and session completion) expose much of the structure needed for KV-storage management. Trace-driven comparisons show that workload-specific recurrence estimates retain up to 22% more within-session hits and reduce dead state by up to 40%; a session completion signal eliminates the fundamental idle-time reclamation tradeoff for single-session state; and session-aware reuse classification identifies shared state 4-700x more selectively than request-level access counts. These results motivate a small cross-layer interface for workload-aware KV storage without application-specific semantics.

0.14

Nebula: Distributed Position-Independent Caching for LLM Serving

cs.DC ·2026-11-06 ·Yifei Liu, Junbin Pan, Junhao Hu, Yang Liu, Junyi Chen +5 more
Abstract

Modern LLM workflows, such as retrieval-augmented generation (RAG) and LLM agents, build long prompts from multiple independent text segments, making prefill a major source of computation. Reusing KV caches is therefore important for reducing serving costs and latency: distributed caching systems spread KV cache across a cluster to increase aggregate capacity, while position-independent caching (PIC) enables reuse of recurring segments under different prefixes. However, existing systems distribute KV caches but still rely on prefix caching (PC), while existing PIC methods enable position-independent reuse without cluster-wide resource management. We present Nebula, the first distributed LLM serving framework to treat PIC segments as first-class citizens, enabling remote PIC reuse and parallel prefill. To track PIC's fine-grained, asynchronous state changes, Nebula combines optimistic accounting with event-driven updates to provide a timely and truthful cluster view for cache management and prefill scheduling. To accommodate diverse reuse patterns across workloads, Nebula leverages cluster-wide ghost-hit feedback to automatically tune the allocation of shared cache capacity between PC and PIC. Finally, Nebula incorporates queueing, computation, and KV transfer costs into a unified prefill completion time model to guide elastic parallelism adjustment and locality-aware compute placement. Across four models and four real-world RAG and agentic workloads, Nebula reduces median TTFT by an average of 61.4% compared with the strongest existing baseline and recovers an average of 76.1% of the tokens missed by prefix caching.

0.12

RPEG: Training Reasoning Language Models Scalably and Cost-Effectively

cs.DC cs.AIcs.CLcs.LG+1 ·2026-11-06 ·Maciej Besta, Leonard Schmidt, Pierre Pang, Patrik Okanovic, Robert Gerstenberger +1 more
Abstract

Reinforcement learning (RL) is now central to training Reasoning Language Models (RLMs), but RLM pipelines are difficult to scale and costly to run because each RL step couples multiple large models, heterogeneous stages, cross-step state, and complex asynchronous dependencies. Existing systems typically optimize only parts of this design space. We address this complexity by jointly optimizing the pipeline structure as well as its execution using a novel data-centric graph intermediate representation called RLM Pipeline Execution Graph (RPEG). RPEG exposes structural choices such as bounded asynchrony and actor-critic sharing together with execution optimizations such as stage fusion, fine-grained scheduling, and multi-dimensional model parallelism. It is coupled with a comprehensive analytical performance model for RLM training, jointly capturing compute, communication, memory, and hierarchical rack-, node-, and GPU-level effects. Across systems with up to 256 GH200 GPUs, our distributed runtime engine implementing RPEG improves performance by up to 2x over ReaL and reduces step latency by up to 2.48x over verl, while increasing hardware utilization by up to 30%. Overall, RPEG provides a graph-driven foundation for next-generation training pipelines for frontier reasoning LLMs.

0.12

Towards Sustainable Federated Learning via Carbon-Aware Client Selection

cs.DC ·2026-11-06 ·Wentao Gao, Omid Tavallaie, Albert Y. Zomaya
Abstract

The Information/Communication Technology (ICT) industry is severely challenging the global environment due to the sharply rising carbon emissions generated by machine learning (ML) workloads. Federated Learning (FL) is one of the prominent distributed ML systems used in applications with billions of client devices, such as Google Gboard and Apple Siri. Nearly half of the total carbon footprint in FL originates from client-side computation during local training. However, existing client selection strategies in FL primarily focus on minimizing energy consumption rather than carbon emissions, despite the fact that carbon impact depends not only on computational load but also on the underlying carbon intensity of the energy source. In this paper, we address this gap by proposing a carbon-aware fair client selection strategy (CA-FCS) designed to minimize the carbon emissions associated with model training on client devices. In contrast to existing approaches for reducing carbon emissions in FL, we incorporate hourly carbon intensity and a realistic energy consumption model developed using empirical measurements from Google Pixel smartphones to estimate client-side carbon emissions in FL. Since repeatedly selecting low-carbon clients may reduce data diversity and degrade global model performance, particularly under non-independent and identically distributed (non-IID) data settings, CA-FCS incorporates fairness into client selection to mitigate this issue. We conduct extensive experiments under realistic FL settings for mobile devices, where client participation is constrained by device availability conditions, such as charging status and Wi-Fi connectivity. The results demonstrate that our method reduces client-side carbon emissions by up to 46% while maintaining competitive model accuracy.

0.12

Software & Programming 34

Automata for MSO over Infinite Trees with Quantification over Borel Sets of Branches

cs.FL cs.LO ·2026-11-06 ·Mikołaj Bojańczyk, Antonio Casares, Sven Manthe, Paweł Parys
top venuetheory
Abstract

Rabin's Tree Theorem says that the MSO theory of the infinite binary tree $2^*$ is decidable. Shelah showed that MSO logic becomes undecidable if this tree is extended to $2^{\le ω}$, i.e. by allowing quantification over sets of infinite branches. A longstanding open problem is whether the decidability can be recovered in $2^{\le ω}$ by restricting set quantification to Borel sets. We make some progress in this direction, by identifying a suitable automaton model, and showing that most of the automata-theoretic approach to Rabin's Theorem can be extended to the new framework. The only missing part is a conjecture about finite-memory determinacy in certain games. This paper states and explores the conjecture. We prove it in some restricted cases, and give lower bounds on the memory required in those games.

0.70

Retention Horizons of Volatile Stores in iOS and Android Extractions

cs.SE cs.OS ·2026-11-06 ·Daniel Kwaku Ntiamoah Addai, Sanders Larweh Tetteh, Selorm Adjabeng, Kingsley Albert Egyir
Abstract · 17 pages, 3 figures. Lab Technical Report LAB-TR-0021, SectraTEK Lab; preprint, not peer reviewed. Corresponding author: Daniel Kwaku Ntiamo…

Whether an extraction can answer a question about a given day depends on how far back each store reaches, yet public statements rest on single test devices. We measured the retention horizon, the time from a store's oldest retained record to acquisition, for eight iOS and seven Android stores on 48 full file system extractions and seven bug reports, keeping devices whose clock agreed with the acquisition time (26 Apple devices, 13 Android phones). Horizons ranged from minutes to years and followed the mechanism that bounds each store. The unified log Persist stream reached a median of 2.5 days; interactionC reached 166 to 170 days on 18 of 25 devices. Android stores stayed within their source-code limits (usage statistics 10.2 days, DropBox 3.1, notifications 1.5); bug-report logs reached 3.1 hours at most. The oldest record misled three times: one knowledgeC stream reached back two years, the others a median of 29 days; interactionC read without its write-ahead log showed older records on 18 of 25 devices; and four of seven bug reports began with stray lines up to 16.6 days old above a body of 3.1 hours or less. A horizon belongs to a store, a release and a device.

0.64

MongoBench: Do LLMs Know What a MongoDB Query Will Do?

cs.SE cs.DB ·2026-11-06 ·Qibai Chen, Haoting Qiu, Yezhou Cheng, Mengxiao Qian
publishedbenchmarkpositioncode/data
Abstract · Accepted at CAIT 2026

LLM agents increasingly operate databases on their own: they generate queries and run them against live data. Two checks are commonly used to judge such an agent - whether it produces a valid query, and whether it reaches the right final answer - but neither tells us whether the agent knows what its query will actually do before it runs. We argue that this consequence prediction ability must be evaluated in its own right, and we build MongoBench to measure it: given a collection of documents and a MongoDB find() filter, a model must return the exact set of matching _ids, with ground truth from a real mongod server. The 300 items each compose three or four documented matching pitfalls and are proposed symmetrically by three vendors' models, so results stratify by item source. Evaluating four models - frontier systems from three vendors and one open-weight reasoner - accuracy ranges from 70.1% to 92.1%, a 22-point spread. Each proposer mispredicts 10-31% of the items it wrote itself, so producing a query does not imply predicting its effect. A failure-mode analysis shows errors are better predicted by which semantic pitfall a query exercises - range comparisons and null - than by surface measures such as document count or the number of composed rules. Prepending a verified rule sheet helps unevenly; executing the original filter against the engine yields reliable match sets, but delegation must be checked: on 28% of items (averaged over three runs) one model's executed filters never passed a canonical check against the intended query, and it answered about 22 of them wrong, while the other three passed on every item. We release the benchmark, the mongod oracle, and the evaluation code with all prompts at https://github.com/research-repo-bai/MongoBench.

0.50

TwinLight: A Multi-Timescale Physical-Layer-Aware T-API-Compliant Optical Domain Controller Digital Twin

cs.SE physics.optics ·2026-11-06 ·Carlos Natalino, Piotr Lechowicz, Renato Ambrosone, Enrico Miotto, Vittorio Curri +1 more
publishedcode/data
Abstract · Demonstration presented at ECOC 2026, Málaga, Spain, Sept. 2026. Paper ID Tu2-Ex1/1440. Code in https://github.com/carlosnatalino/TwinLight

We demonstrate TwinLight, an open-source optical-network digital twin acting as an optical domain controller. Its T-API REST and telemetry streaming interfaces expose a physical-layer engine that combines a GNPy steady-state baseline with four analytical transients.

0.50

comfort-db: A Python toolkit to access, preprocess and explore the oceanographic COMFORT dataset

cs.SE physics.data-an ·2026-11-06 ·Yvonne Jenniges, Are Olsen, Boris Peter Koch, Sebastian Maneth
code/data
Abstract · Software description paper for comfort-db v0.1.1. Code: https://github.com/y-jenniges/comfort-db. Archive: https://doi.org/10.5281/zenodo.24…

comfort-db is a Python library that facilitates working with the COMFORT SQLite dataset, version 2. The dataset assembles global oceanographic measurements from ten datasets and contains a total of 458,724,734 in situ measurements from the years 1772 to 2020. comfort-db enables easy access to the database and provides preprocessing and exploratory analysis utilities. Preprocessing supports variable scaling and filtering by space, time and quality. Unit conversions can be applied during data loading or separately via a converter object. Gridding is supported and optionally works with bathymetry data to exclude grid cells on land. Moreover, the library supports general and field-specific exploratory data analysis through plotting functions, an interactive Dash app and functions to compute oceanographic profile and water mass statistics. comfort-db is open source (MIT license) and available from PyPI and GitHub.

0.41

Device-Life Boundary Indicators and the Classification of Restored Versus Native Records on Android

cs.SE cs.OS ·2026-11-06 ·Daniel Kwaku Ntiamoah Addai, Selorm Adjabeng, Kingsley Albert Egyir, Sanders Larweh Tetteh
unusual
Abstract · 9 pages, 4 figures. Lab Technical Report LAB-TR-0028, SectraTEK Lab; preprint, not peer reviewed. Corresponding author: Daniel Kwaku Ntiamoa…

A phone that shows calls, messages and photographs from years ago may have been set up recently and carried them over from an earlier handset, so its screen dates do not date the device's own life. Which traces fix where that life begins, how well they agree and how much of what is visible predates it is unmeasured. We read 14 Android full file system extractions (12 Samsung; Android 10 to 16), seven bug reports and 33 iOS extractions (a probe). Every Android device recorded a first boot, but the first evidence of user setup came a median of 98 days later (0 to 563). Of the 13 devices with messages, 7 held some dated before the first boot, a majority of them on 5; 24% of the apps outside the system image were installed as restored. Against the creator field that a migration tool leaves on messages, a boundary at user setup recalled all but 1 of 46,968 such messages, the first-boot boundary 94 to 100%. On synthetic fixtures, the labelling error followed the boundary error. A date on a record dates the device only once the record is shown to have been created on it.

0.40

Integrating MBSE Principles with AI to Improve the Quality of Requirements in SMEs Context

cs.SE ·2026-11-06 ·Iryna Volosheniuk, Claude Baron, Jean-Michel Bruel
published
Abstract

Model-Based Systems Engineering (MBSE) is a promising approach to improve systems engineering practices, particularly by enhancing requirement management, traceability, and early validation and verification (V\&V). However, its adoption remains challenging, especially for small and medium-sized enterprises (SMEs), which often lack structured processes and solutions adapted to their scale and constraints. Our goal is to propose a lightweight MBSE methodology tailored to the needs and constraints of SMEs. This paper focuses specifically on requirements refinement, traceability, and early V\&V. The approach integrates MBSE principles with artificial intelligence-assisted analysis to improve requirements quality while remaining compatible with existing tools and practices. A proof of concept is developed and applied in an industrial case study to evaluate the feasibility of the approach. The results demonstrate how early V\&V activities can be supported within an SME context, while improving traceability across engineering artifacts. These findings highlight the potential of a lightweight, integrated MBSE approach to support the progressive adoption of MBSE in SMEs, thereby improving overall engineering efficiency.

0.39

Optimal Probabilistic Answer Set Programming: Complexity and Implementations

cs.LO cs.AI ·2026-11-06 ·Damiano Azzolini, Giuseppe Mazzotta, Francesco Ricca, Fabrizio Riguzzi
published
Abstract · Accepted for publication in Theory and Practice of Logic Programming

Probabilistic Answer Set Programming (PASP) extends Answer Set Programming with the possibility of representing uncertain data. Among the possible available semantics, we focus here on the credal semantics. In this setting, prior work has focused exclusively on plain ASP programs, i.e., programs without optimization constructs. However, many real world tasks define optimality criteria over the possible solutions, which cannot be conveniently modeled with existing probabilistic ASP frameworks. In this paper, we close this gap and propose optimal Probabilistic Answer Set Programming, which extends PASP by also allowing weak constraints, which are the standard optimization construct of ASP. We define the semantics of such programs, study the computational complexity of the inference and most probable explanation tasks, and propose practical implementations based on Algebraic Model Counting and Answer Set Programming with Quantifiers. An empirical evaluation of the proposed approach on three real world tasks confirms the applicability of our framework.

0.38

Wake 5: A Python Framework for Solidity Testing, Fuzzing, and Static Analysis

cs.SE ·2026-11-06 ·Josef Gattermayer, Michal Převrátil
benchmarkcode/data
Abstract

Smart-contract fuzzers differ by an order of magnitude in throughput and by design in how they construct inputs, yet published comparisons rarely separate the two. We present Wake 5, an open-source Python framework for Solidity testing, fuzzing and static analysis, and use it to locate the binding constraint of stateful smart-contract fuzzing. Wake 5 replaces the JSON-RPC connection to an external node with an in-process EVM (revm) and exposes its static analysis to language-model agents through a Model Context Protocol (MCP) server. On the Daedaluzz maze benchmark (153 reachable assertions, guard constants below 2^32), Wake 5 executes 29,931 transactions per second, on par with Foundry (30,186) and 11.5x faster than Echidna (2,600). In eight-hour campaigns, however, the three tools plateau at medians of 75, 71 and 70 assertions, and their union reaches 93. We show that Wake's integer generator places 51% of its probability mass below 2^32 and that, once shallow assertions are exhausted, the generator prior rather than throughput bounds every tool. A harness written by a language-model agent, which derives inputs with a constraint solver, reached all 153 assertions in each of five campaigns executed by Wake, in 33 to 79 minutes. A complementary ablation in an agentic audit of a production Go codebase found that static-analysis tools improved recall for none of three models, raised precision for all three, and increased language-model cost by up to 12.7%. We discuss correlated oracles and non-deterministic completion, and argue that the fuzzer's role is shifting from search to execution.

0.36

Navigating Complexity: A User-Centered Evaluation of Project Management Tools

cs.SE cs.HC ·2026-11-06 ·Yasin Ghadimi Zaker, Hamid Reza Ghaeini
Abstract · 11 Pages

In the age of digital transformation, companies are increasingly relying on specialized software to manage the complexity of projects. However, when selecting tools, features and costs often take precedence over user experience (UX), leading to low adoption rates, user frustration, and project delays. This study empirically evaluates the user experience of leading project management tools (Jira, Trello, Monday.com, Asana, and Microsoft Project) across various project scales using a mixed methods approach. The System Usability Scale (SUS) was used as the primary quantitative instrument, with data collected from 141 experienced professionals via online questionnaires. The results reveal a significant trade-off between functionality and usability. Trello achieved the highest SUS score (77.5), indicating a superior user experience for small projects and agile teams, while Microsoft Project scored the lowest (51.2), suggesting significant usability challenges despite its robust features. Jira demonstrated a balanced profile, offering powerful features with manageable complexity. Statistical analyses confirmed that the choice of the optimal tool is highly context dependent, with team size and project complexity influencing UX outcomes. The findings underscore that UX is not a secondary feature but a critical success factor for software acceptance and project effectiveness. This study provides an evidence-based framework for aligning tool selection with the project context, offers practitioners actionable insights, and identifies design opportunities for software developers.

0.33

When Proofs Are Abundant, What Is a Paper For?

cs.LO cs.AI ·2026-11-06 ·Charles-Albert Lehalle
hot direction
Abstract

AI systems now produce formally verified proofs of celebrated open problems, and the cost of obtaining a correct proof is falling toward zero. If correctness becomes cheap, what is left for a research paper to communicate? I argue that a paper is not exhausted by a verified statement. It is a narrative exposed to the criticism of a community, made of an intension (the author's choice of problem and perspective), a scientific chain of thought (the reasoning steps leading to the result), and the arguments supporting each step. Most published mathematics is judged on these two elements rather than on correctness alone, and machine-generated formal scripts convey neither. Borrowing from logic and type theory, I formalise the first as a Derivation-Level Intension and the second as a Directional Extension: a decomposition of a proof into cut-points whose axiomatic footprints reveal its weak links, where the result could be strengthened, and its strong links, which can be transported to other theories. Moreover, I examine five roles AI can play in the research process, including bibliographical search and critical judgment, and show that none of them removes the need to declare these two elements. I mention ways these two concepts could be useful for Journals and for training of young mathematicians in a AI era. If proofs become abundant, the scarce resource is not correctness but a community's shared understanding of why a result matters and how it could be pushed further.

0.31

Word Subtraction Games

cs.FL ·2026-11-06 ·Yusuke Inoue, Hiroki Inazu
unusual
Abstract

This note is a self-contained text on word subtraction games, combinatorial games in which players alternately remove a nonempty suffix belonging to a prescribed subtraction set from the current word. It collects and develops basic results on these games from the viewpoints of combinatorial game theory and formal language theory. We focus on the relationship between subtraction sets and the resulting languages of winning and losing positions. The topics include finite, regular, and context-free subtraction sets, as well as several variants and further developments.

0.30

An Empirical Study on Clustering-based Quantum Approximate Optimization Algorithm for Regression Test Case Selection

cs.SE ·2026-11-06 ·Antonio Trovato, Martin Beseda, Dario Di Nucci
Abstract · Manuscript under review at Empirical Software Engineering (EMSE)

Regression testing assesses whether software components that have already been developed and tested continue to work correctly after modifications. Ideally, regression testing re-executes the entire test suite after each modification, but this is infeasible in real-world scenarios because of the computational burden; therefore, optimized approaches, such as Test Case Selection, have been developed. The state of the art in regression testing optimization includes both classical approaches, such as Additional Greedy and DIV-GA, and quantum approaches, such as SelectQA, BootQA, and IGDec-QAOA. However, while the former incur high computational costs, the latter are constrained by the physical limitations of currently available quantum machines. This paper extends the evaluation of QAOA-TCS, our quantum approach for the test case selection problem. We thoroughly evaluate its performance, compare it with state-of-the-art classical and quantum approaches, and assess its robustness to quantum noise. The results show that QAOA-TCS outperforms state-of-the-art multi-objective approaches in most cases and maintains stable performance even under high noise levels. In the single-objective formulation, QAOA-based methods perform comparably. In terms of efficiency, our previous method based on quantum annealing, namely SelectQA, is the most efficient for large problems in the multi-objective formulation, whereas QAOA-TCS offers the best trade-off between effectiveness and efficiency in the single-objective formulation when run on realistic noisy simulators. However, such analysis is constrained by the different computational environments used to run the algorithms.

0.29

GreyZone: Production-quality Agentic Application Threats

cs.SE ·2026-11-06 ·Ruoyu Zhou, Shuyao Jiang, Haifeng Ruan, Abhik Roychoudhury
benchmarkcode/data
Abstract

Agents are increasingly deployed across domains such as coding, banking, and commerce. Their ability to invoke tools and change application state makes security testing important for reliable deployment. Although agent security has been studied through various adversarial testing approaches, benchmarks grounded in production agent environments remain limited. We introduce GreyZone, a benchmark for evaluating automated red-teaming services against real-world agent applications. Unlike benchmarks such as AgentDojo, which primarily evaluate subject agents under attack, GreyZone evaluates services' ability to generate and deliver attacks that cause verified security-policy violations. We select independently developed, open-source agent applications with security boundaries concerning access permissions, confidentiality, and action authorization. These applications expose diverse attack surfaces, including those arising from multi-agent interactions and multimodal inputs. By grounding evaluation in application environments that preserve production-relevant workflows and independently verifying outcomes, GreyZone supports research on automated agent security testing. Specifically, we present two preliminary case studies using Promptfoo on Letta Code and Clawith to examine attack delivery, agent behavior, and verified outcomes. We release the benchmark, including subject agents and attack surfaces for each application, to support further research on agent testing methods.

0.28

QAOA for Layered Architecture Recovery: Feasibility, Limits, and the Role of Decomposition

cs.SE ·2026-11-06 ·Julien Cardinal, Andrew Chedid, Imen Benzarti, Ghizlane El Boussaidi, Christophe Pere +1 more
Abstract

Layered architecture recovery (LAR) reconstructs a system's architectural layers from its package dependencies, and is defined as a combinatorial optimisation problem solved with classical search. We investigate whether the Quantum Approximate Optimization Algorithm (QAOA) can be applied to LAR, and with what practical limitations. We formulate LAR as a Quadratic Semi-Assignment Problem, transform it into a Quadratic Unconstrained Binary Optimisation (QUBO) problem in which each package is penalised unless it is assigned to exactly one layer, and encode it as an Ising Hamiltonian solved by QAOA. To reach instances beyond the qubit budget of a single circuit, we also apply an impact-guided decomposition that solves the problem as a sequence of smaller sub-problems. We evaluate the approach on three real Java systems ($15$, $18$, and $24$ qubits), studying the effect of circuit depth and sampling budget (RQ1), a comparison with a classical Tabu Search baseline (RQ2), robustness to simulated quantum noise (RQ3), and the decomposition (RQ4). We find that monolithic QAOA is feasible and accurate on the two smaller systems but fails on the $24$-qubit instance, that the sampling budget rather than circuit depth governs its accuracy, and that it is competitive with but does not outperform Tabu Search. Decomposition restores reliable recovery on the $24$-qubit instance, raising the optimal rate from $6.7\%$ to $93.3\%$; its accuracy is governed by the size of each sub-problem and, unlike the monolithic pipeline, improves as the sampling budget is reduced. Our results indicate that, in the current noisy intermediate-scale regime, problem decomposition, rather than larger circuits or more sampling, is the key to applying QAOA to software architecture recovery.

0.28

DRAFT-Bench: Evaluating Frontier Agents Across the Professional Design Lifecycle

cs.SE cs.AI ·2026-11-06 ·Rong Pan, Jianqiang Wang
Abstract

Can frontier general-purpose agents understand client intent, explore creative directions, execute revisions, and prepare production deliverables? We introduce DRAFT-Bench (Design Reliability Across Feedback and Tasks), a lifecycle-oriented evaluation of requirement elicitation, creative exploration, constrained editing, iterative revision, and production handoff. Its Requirement-to-Artifact framework links evolving requirements, protected properties, artifact versions, and delivery obligations. We compare native Codex/Astra and Claude Code across 35 design briefs under 43 conditions, alongside a separate 40-scenario RFQ track. Astra completes 43/43 design conditions and Claude 35/43. All scored core runs receive full structural credit, yet on the 28 core conditions scored for both systems, semantic checks flag violations in 11 for Astra and 5 for Claude; these are model judgments, not expert-confirmed failure rates. RFQ coverage is high, but success ordering changes across cohorts. SKU expansion, revision dependencies, and White ink files expose ambiguity among unmet obligations, missing inputs, and evaluator interpretations. These findings separate execution, requirement compliance, and professional acceptance. Unpaired tracks and different tools preclude claims of RFQ-to-delivery success, factory acceptance, or a pure model ranking.

0.27

Done Is Not Correct: Measuring Silent Failures and Self-Verification Calibration When LLM Agents Take CAD Actions

cs.SE cs.LG ·2026-11-06 ·Carson Rodrigues, Clive Rodrigues, Aravind Reddy G
benchmarkcode/data
Abstract · 18 pages, 3 figures, 2 tables. Benchmark, harness, per-attempt records and a 3D viewer: https://github.com/rodriguescarson/cad-silent-failur…

When a large language model (LLM) agent writes parametric CAD code from a natural-language specification, syntactic success decouples from semantic correctness. We release a benchmark and harness that measure this gap directly: 13 specification tasks whose every requirement maps to machine-checkable properties of the produced B-rep, measured by the kernel against published tolerance bands, with reference solutions and the full graded attempt corpus. The exact, self-audited oracle isolates two quantities no prior CAD benchmark reports: the silent-failure rate (a valid, renderable solid, an explicit completion claim, and at least one failed semantic check) and completion-claim calibration. Across four models from four vendors (156 attempts per architecture), single-shot agents claimed completion on 98% of attempts and were silently wrong on 14.1%; a tool-use loop that enforces measure-then-claim cut silent failures to 3.2% and raised claim precision from 64% to 96%, and a four-model ablation attributes the change to the claim contract rather than the feedback. A planner-coder-checker pipeline matched that precision at lower coverage with no success gain. Stated confidence was miscalibrated single-shot (ECE 0.32) and became informative under verification for three of four models. Hardening our own oracle flipped 18 of 519 regraded verdicts from pass to fail and none in reverse; a blind expert audit overturned 3 of the 18 silent failures outside the one ambiguously specified task; and fourfold scaling of every tolerance band leaves the contrast intact.

0.27

When Instruction Files Go Stale: Documentation Debt and Hidden Cost in Coding Agents

cs.SE cs.AI ·2026-11-06 ·Tural Mehtiyev, Biruk Tadesse, Lawrence Arkoh, Wesley Assunção
code/data
Abstract · 11 pages, 5 figures, 3 tables. Replication package: https://doi.org/10.5281/zenodo.23992004

LLM-based coding agents such as Claude Code are guided by repository instruction files (also called context files or agent configuration files) like AGENTS.md and CLAUDE.md, which are read in every agent session. As any documentation artifact, these files can fall behind the code and become stale (e.g., a test folder moves, but the file still points to the old one). In technical debt terms, such a stale instruction is documentation debt, and its interest is the cost it adds to every later agent session. Evaluations that check only task success can miss this cost, and the stale file can persist. In this study, we ask what leads agents to act on a stale instruction, how it affects their behavior and at what cost, and how they recover. We study three real-world stale instructions from active open-source repositories in depth, each a different kind of guidance: a test command, a code location, and a procedure for adding a component. Agents fix 22 maintenance tasks under three versions of each instruction (left stale, corrected to the current path, and removed), with the code and the task held fixed, and we analyze 732 trajectories of two models at the level of individual tool calls. We find that agents follow stale guidance when it is relevant to their tasks, sometimes even when correct information is available elsewhere. On the stale test command, Claude Sonnet 5.5 needed about two extra tool calls per run, a third more than with the corrected command; in the other cases, we could not detect a difference. Although agents generally recovered from the immediate failures, none repaired the underlying instruction file. Correcting the instructions eliminated operations on obsolete paths, while removing them prevented most. Our findings show that task success alone can hide the costs of documentation debt, highlighting the need to evaluate agent behavior beyond final outcomes.

0.25

LLM-Augmented Flow for Project Conformation and Task Complexity Evaluation Based on the TLM Framework

cs.SE ·2026-11-06 ·Maksim Sysoev, Alexander Semenov, Ilia Cheshev, Alexey Gevorkyan, Danil Somov
Abstract · 5 pages, 2 figures, 6 tables

The software development process, from inception to a working product or solution, can be divided into two major phases: discovery (Discovery) and delivery (Delivery). The Discovery is further subdivided into High-Level Discovery (HLD) and Low-Level Discovery (LLD). During HLD, management identifies the business solution or product to be integrated into the company's infrastructure. LLD, in turn, produces requirement documentation artifacts, including the Product Requirements Document (PRD), Business Requirements Document (BRD), and Functional Specifications Document (FSD). Story Points (SP) are a widely used modern approach to measuring the effort required to accomplish predetermined tasks, and the artifacts listed above help derive SP more precisely. With the rapid growth of LLM technologies, one of the main questions that arises is whether the whole Discovery phase, or most of it, can be completed with the help of LLM tools. In this paper, we test a framework that structures requirement documentation generation and task complexity scoring, based on the model introduced by (Tran, Lévesque, and Meunier 2004) and executed with the help of agentic tools. The approach is applied to a real-world commercial project in the fintech domain as a feasibility study rather than a validation: development effort was not recorded for the tasks examined, and the resulting complexity scores are therefore reported as the output of the instrument rather than as validated estimates of effort.

0.24

Nothing Is Shared Until You Synchronise: Reinterpreting Java for Data-Race Freedom

cs.PL ·2026-11-06 ·Luke Cheeseman, Elias Castegren, David Black-Schaffer, Erik Österlund, Tobias Wrigstad
Abstract · 30 pages, 11 figures

In mainstream shared-memory languages, programmers use synchronisation mechanisms such as locks to isolate shared data accesses across threads to avoid data races. Yet synchronisation enforces only a temporal ordering, while shared memory still communicates updates between threads at every ordinary read and write. To achieve isolation, programmers therefore need to understand the connection between ordering and communication, and to apply synchronisation correctly and consistently, while compilers and hardware optimise assuming they have. When programmers synchronise incorrectly, they lose isolation, and the semantics of their programs change without notice. We propose Sync-as-Share, a memory model built on isolation and communication rather than on the ordering of accesses to shared memory. We start from thread isolation, where each thread reads and writes a private view of memory, and reinterpret programs' existing synchronisation actions to also communicate updates explicitly. As nothing is communicated outside synchronisation actions, data races become impossible by construction. Instead, conflicting updates are detected at synchronisation points. Viewing synchronisation as explicit communication exposes the design decisions of what should be communicated, to whom, and what counts as a conflict. We propose two communication topologies with different trade-offs. We instantiate Sync-as-Share for Java by making the synchronisation actions already defined by its memory model the only points at which threads communicate, so Java programs can adopt Sync-as-Share unmodified. We formalise this instantiation as version-controlled memory, in which synchronisation actions push and pull updates between private views, and mechanise it in Rocq. We show that Java programs that are correctly synchronised under the Java memory model behave identically under shared memory and Sync-as-Share.

0.24

TensorMutation: Learning Weight-Space Mutations of Deep Learning Models on Neural Graphs

cs.SE cs.LG ·2026-11-06 ·Zaheed Ahmed, Philip Makedonski, Jens Grabowski
Abstract · 18 pages, 7 figures

Mutation testing for deep learning (DL) faces a cost-realism trade-off. The pre-training approach produces realistic but computationally expensive mutants, whereas existing post-training operators modify trained models efficiently but produce mutants that less closely resemble real faults. This study asks whether the trade-off is necessary: can the weight-space effect of a fault-informed pre-training mutation operator be predicted from a trained model, without retraining it? We answer this question with TensorMutation, which represents a trained model as a neural graph and uses a mutation-conditioned graph neural network (GNN) to predict parameter changes. The GNN is trained offline and reused across unseen models so that TensorMutation can generate executable mutants without retraining. We construct a corpus of 25,252 original-mutant pairs from 71 multilayer perceptron (MLP) classifiers using 11 fault-informed pre-training operators, including 25 newly curated subjects. TensorMutation partially reproduces the behavioral effects of pre-training mutations on nine held-out subjects, and direct weight-space supervision significantly improves weight-space fidelity. On 12 real-fault subjects, TensorMutation achieves the highest median behavioral similarity among the evaluated approaches. The improvement over the post-training tool DeepMutation++ is statistically significant (Holm-adjusted p=0.018), whereas the differences from the pre-training tool DeepCrime and the norm-matched random control are not significant. These results show that fault-informed pre-training mutations leave learnable structure in weight space that transfers to unseen models.

0.21

Tracybot: Preserving Developer Intent by Tracing Agent-Generated Code to Prompts

cs.SE cs.HC ·2026-11-06 ·Răzvan Albu, Aliaksei Khval, Maxine Orlova, Danis Music, Filipe Araújo Veras Rosa +2 more
surveycode/data
Abstract

Large language models (LLMs) are now widely used in software development, and LLM-based coding agents are given increasing autonomy. With coding agents, many design decisions are made in the conversation between the developer and the agent. Typically, the developer starts a conversation with an agent and states their intent, the agent proposes options if needed, and the developer plans the change accordingly. Then, the resulting code is committed. However, in such scenarios, the decisions and the intent behind the changes remain in the conversation, with no link to the code they produced. In this paper, we present Tracybot, an open-source tool that integrates with VS Code, Git, and coding agents to keep track of relevant conversations in an LLM-assisted development process. Tracybot captures each change as a Tasklet which consists of optional planning prompts used to plan a change, together with a build prompt that produces it. Tracybot has two main features: (i) it maps line-level code changes made by coding agents to the Tasklets that produced them, and (ii) it allows the developer to export a Tasklet as a decision record when needed, for example when the decision concerns the architecture of the project. We developed the tool over two iterations of a design science study, while focusing the evaluation on the usability and user-friendliness of the tool. In a focus group (7 participants) and a survey (24 participants), we found that Tracybot was perceived as useful for activities like code review, and most participants found our Tasklet design and presentation to be a clear way to represent the developer's intent. Tool: https://github.com/TracyTeam/tracybot. Video: https://youtu.be/Vg_a1F5eO04.

0.21

AnimWeb-Bench: Benchmarking Animated Webpage Code Generation from Videos

cs.SE cs.CV ·2026-11-06 ·Zhongyi Zhang, Jingyu Xiao, Shuqing Li, Michael R. Lyu
benchmarkcode/data
Abstract

Animation is pervasive on the Web, the CSS animation property alone appears on over three-quarters of webpages. Yet existing webpage code generation benchmarks target static screenshots, user-triggered interactions, or isolated motion graphics, leaving the ability of multimodal large language models (MLLMs) and coding agents to reconstruct complete animated webpages largely unexplored. To fill this gap, we introduce AnimWeb-Bench, the first benchmark for animated webpage code generation from videos. It comprises 124 real-world webpages spanning 7 application domains and 33 animation types, with 91.9% of pages combining multiple types, each paired with a reference video and the static assets of the original webpage. We further propose a human-aligned, two-dimensional evaluation protocol: Static Appearance Fidelity measures structure, text, color, and visual similarity on aligned keyframes, while Animation Fidelity tracks UI components across frames and compares their position, size, velocity, and appearance with those in the reference video. Evaluating 14 MLLMs and three coding agents reveals: (1) reproducing webpage animations is substantially more difficult than reproducing static appearance; (2) explicit frame inspection and render feedback improve the reconstruction quality of coding agents; and (3) rapid, continuous changes are harder to reproduce than discrete state transitions. Our code and dataset are available at https://starry1001.github.io/AnimWeb-Bench-website/ .

0.20

Longtail-SWE: High-Quality SWE Tasks for Underrepresented Languages

cs.SE cs.AI ·2026-11-06 ·Vladimir Dobrovolskii, Ibragim Badertdinov, Alexander Golubev
benchmarkcommunity pick
Abstract

Challenging repository-level benchmarks remain limited for many programming language ecosystems, where filtering scarce candidates with ambiguous descriptions or defective tests can discard valuable engineering problems. We introduce Longtail-SWE, a benchmark of 202 issue-resolution tasks across 13 languages with limited coverage in existing repository-level benchmarks: Clojure, Dart, Elixir, Erlang, F#, Haskell, Julia, Kotlin, Lua, OCaml, R, Scala, and Swift. Its construction uses a solver-judge-repairer pipeline: a solver attempts each candidate task, and a judge reviews the task together with the solver's trajectory and test results to guide revisions to tests and task descriptions. These revisions aim to correct tests that reject valid alternative implementations or accept incorrect solutions while retaining challenging tasks and preserving their intended requirements. We evaluate nine model configurations, including frontier models, and analyze task-solving performance, consistency across repeated attempts, and resource usage. We release all benchmark instances, executable environments, complete construction and evaluation tooling, and agent trajectories, together with an additional set of easier tasks to support evaluation of smaller models.

0.20

SPMW: A Grid-Based Programming Model for Scalable Accelerator Design

cs.PL cs.AR ·2026-11-06 ·Hongzheng Chen, Shihan Fang, Niansong Zhang, Sunwoo Kim, Weiran Liu +1 more
Abstract

Modern accelerators specialize recurring computation and replicate it across regular hardware structures to achieve high performance. Sustaining this parallelism requires equally deliberate data delivery. Yet design tools often express these structures as sequential programs or flatten them into separate hardware modules and connections. This obscures shared patterns and leads to repeated compiler work. We introduce Single Program, Multiple Connected Work Units (SPMW), a grid-based programming model for scalable accelerator design that makes grid regularity a first-class abstraction. SPMW decouples local work-unit behavior from grid connectivity and data delivery. To our knowledge, SPMW is the first programming model to retain regularity throughout the compilation process, combining behavioral HLS with structural RTL generation for both fixed-function and domain-specific instruction-programmable accelerators. We implement SPMW with a Python frontend and an MLIR-based compiler that synthesizes each distinct kernel configuration once for reuse across independent hardware instances. Preserving regularity also enables parallel compilation and functional simulation to accelerate iterative design feedback. We use SPMW to generate a 32x32 systolic array that reduces GEMM latency and parallel kernel-to-RTL compilation time by 4.36x and 619x relative to AutoSA. Our FFT achieves up to 2.65x lower latency than HP-FFT. Our mini-TPU and our SPMW implementation of the FEATHER accelerator demonstrate complex composition and outperform their RTL baselines. Parallel functional simulation provides up to 18.4x speedup over Vitis HLS C simulation on the mini-TPU workloads. SPMW is also agent-friendly: given the same design specification, all five coding agents produce SPMW designs that satisfy the correctness and latency requirements, with 7.51x lower token usage than for SystemVerilog.

0.19

When Routine Operations Break Architectural Pattern Relationships in LLM Agent Harnesses

cs.SE cs.AI ·2026-11-06 ·Boming Xia, Qinghua Lu, Dino Sejdinovic, Liming Zhu
Abstract

LLM agent harnesses distribute tool mediation, state management and safety guardrails across native services, independently written plugins and configuration. Routine operations can change the runtime order or retained state that architectural pattern relationships require, without any code change. We study five relationships associated with the Guardrail Enforcement Point, Autonomy Controller and Blackboard patterns across five released harnesses, analysing a main sample of fourteen independently written plugins. Using PatternOps, our assessment procedure, we trace requirements from pattern descriptions, provider documentation and deployment profiles to the providers that realise them and the records that decide them. We then compare alternative allocations under supported operations. Such operations broke relationships silently: after a plugin was reapplied or an allow rule saved, policy stopped evaluating edits while every plugin stayed active, and a prohibited edit executed. Which alternatives qualified depended on the requirement. Allocations that kept the required order, or kept or rebuilt the needed state, held on their tested paths and operation sequences, while others prevented prohibited effects but missed edits rejected elsewhere. In these harnesses, a pattern relationship is a property of the running system and its history, so architects should evaluate it against the decision paths, state lifetimes and operation sequences their deployments use.

0.17

Agents on the Patch Path: Mining AI-Associated Security Remediation in GitHub Security Advisories

cs.SE ·2026-11-06 ·Claudio Segal, Paulo Segal, Hudson Silva Borges, Paulo Silveira, Daniel S. Menasché +6 more
Abstract

Autonomous coding agents are increasingly involved in software development, but their participation in real-world security remediation remains poorly understood. We investigate observable evidence of coding-agent involvement in commits referenced by reviewed GitHub Security Advisories (GHSAs), characterize the associated vulnerabilities, and examine review timing in relation to advisory disclosure routes. We develop a local-Git mining pipeline that identifies agent evidence through known agent accounts, explicit commit-message attribution, repository instruction files, and agent configuration artifacts. Linking these findings to GHSA-referenced commits yields 469 valid candidate advisories across 189 repositories, including 448 with direct attribution and 21 supported only by repository-level evidence. Claude is the most frequently observed tool family, followed by Copilot and Cursor. We further compare review timing against other advisories and historical baselines using an exact route--timing decomposition. Relative to advisories published in the same months, candidate advisories exhibit a modestly longer mean review lag, primarily associated with within-route timing rather than disclosure-route composition. Compared with a historical baseline, they exhibit substantially shorter mean review lags, with both route composition and within-route timing contributing to the difference. These findings demonstrate that coding-agent involvement is already observable across hundreds of security advisories and repositories, including high- and critical-severity vulnerabilities. By distinguishing direct attribution from repository-level adoption evidence and separating route-composition effects from within-route timing differences, we provide an empirical characterization of agent-associated remediation in the GHSA review pipeline.

0.17

On the Limits of Execution Traces for Repairing Assertion Failures in LLM-Generated Code

cs.SE cs.AI ·2026-11-06 ·Muntasir Adnan, Carlos C. N. Kuhn
position
Abstract · Under review at 2027 2nd International Conference on Advances in Artificial Intelligence and Machine Learning

Iterative repair returns a failed program's error to a large language model (LLM) and asks it to fix it. After iterative repair, the remaining failures are mostly assertion failures: with four Qwen2.5 models, they make up 79% of residual failures on BigCodeBench, all on HumanEval and 93% on LiveCodeBench. We argue, through an information view of the feedback, that an assertion failure says little about where the fault lies. We propose Breakpoint Debugging (BPD), which returns the failed program annotated inline with its variables' runtime values. We take the tasks that iterative repair failed to fix with five repair attempts and compare two ways of spending five further generations on them: BPD, and resampling. On BigCodeBench, BPD repairs 26.7% of these tasks against 22.4%, and its advantage grows with model size. On HumanEval, BPD outperforms resampling but only slightly, and on LiveCodeBench, fresh sampling repairs more (6.5% against 4.9%). Furthermore, BPD reduces unchanged-code resubmissions from 28% to 8% on BigCodeBench. Using only the first failing test's trace, it outperforms tracing all failing tests while using fewer tokens than both all-trace BPD and plain iterative repair. On LiveCodeBench, BPD gives no gain even on the public tests, so a visible failing input is not enough for a trace to help. These results match the information view: a trace helps only when the test exposes a failing input, and the fault is an implementation slip; otherwise, a repair loop should resample rather than trace.

0.16

Autonomous Code Review: Hidden-Test Results on GitHub Issues

cs.SE ·2026-11-06 ·Adam Shakhabov
benchmark
Abstract · 8 pages, 1 figure, 5 tables

Can an autonomous cross-provider review step change whether code written for a real issue passes hidden tests? We report a retrospective benchmark of 12 historical GitHub issues (five C#, five JavaScript, and two Python) from five repositories. Two directed author--reviewer pairs, OpenAI GPT-6.1-Sol and Anthropic Claude Sonnet 5.5, each ran twice per issue, yielding 48 planned workflows. A bounded terminal agent produced an initial patch; a cross-provider model reviewed it; the author model adjudicated findings and could revise the patch. The same hidden evaluator assessed initial and final patches. Forty workflows completed, while eight terminated before a complete evaluator outcome. Four completed workflows passed, and all four already passed before review. No initial-to-final pass/fail transitions occurred among completed workflows.

0.13

Shortest Synchronizing Words on Product Automata

cs.FL ·2026-11-06 ·Henning Fernau, Igor Rystsov
Abstract

We consider some operations on finite automata graphs, in particular various product, union and merge operations, and investigate how they relate to the parameter~$\mathrm{rt}$, the \emph{length of a shortest synchronizing word}. We derive some lower and upper bounds for the operations that we study. We also examine several classes of automata (for instance, partially ordered, commutative or idempotent automata) that are preserved under these operations and show in which way the parameter $\mathrm{rt}$ behaves in these automata classes. Also, we consider automata classes that can be constructed by combining automata from better known classes; in particular, these new classes of automata will also satisfy the bound of the famous Černý conjecture. This also motivates us to study even more closure properties. In particular, we show that certain automata classes related to synchronization form pseudo-varieties. We also commence a similar study concerning the length of shortest homing sequences of Mealy automata.

0.13

Caleta: An Evidence-Oriented Trust Layer for Reproducible Local Large Language Model Inference

cs.SE ·2026-11-06 ·Joan-Marc Fisa
benchmark
Abstract

Local large language model (LLM) inference is often described as private, verified, or reproducible without separating the technical claims behind those terms. This paper presents Caleta, an Apache-2.0 process supervisor and evidence layer around llama.cpp rather than a new inference engine. Caleta combines versioned manifests, exact-size and SHA-256 verification, controlled backend construction, a restricted OpenAI-compatible proxy, prompt-free execution records, benchmark reports, and a small conformance suite. Its contribution is the composition of these mechanisms into explicit contracts that distinguish integrity from authenticated provenance and privacy controls from operating-system isolation. We evaluate the implementation through static checks, 22 unit tests, deterministic end-to-end tests on Ubuntu, macOS, and Windows, and a pinned real-model experiment on Windows CPU using Qwen2.5-0.5B-Instruct Q4_K_M and llama.cpp b11981. The real experiment validates artifact verification, startup, model listing, non-streaming and server-sent-event chat, and conformance. Three independent benchmark invocations produced fifteen successful warm requests with no failures. In this single configuration, median time to first token was 18.066 ms and median decode throughput was 54.394 tokens/s. These descriptive measurements do not establish cross-engine superiority, output quality, energy efficiency, or hardware-general performance. Raw JSON, commands, and environment metadata are published. Remaining challenges include signed manifests, time-of-check/time-of-use replacement, operating-system containment, proxy-overhead measurement, and broader experimental matrices.

0.12

Deliberate Pacing for Human Oversight in Agentic Software Development

cs.SE ·2026-11-06 ·Tobias Reichel, Michèle Stassen, Anamaria Mojica-Hanke, Maria Teresa Baldassarre, Sebastian Baltes +2 more
position
Abstract

Agentic AI increasingly enables software development activities to proceed without continuous human involvement. While this autonomy may accelerate development, it also raises questions about when developers can meaningfully oversee software development done by agents. We propose deliberate pacing: intentionally adjusting when and how autonomous development proceeds to preserve opportunities for meaningful oversight. Drawing on an exploratory multivocal literature review of 20 sources, we identify three factors motivating this vision: changing relative costs of human involvement, the timing of oversight opportunities, and interdependencies between pacing decisions across the software development life cycle. We argue that neither uninterrupted autonomy nor frequent human intervention is universal. Instead, pacing should balance opportunities to understand and influence consequential decisions against costs in human attention, development time, and rework. We outline a research agenda to investigate when deliberate pacing is beneficial, how intervention timing affects oversight, and how pacing decisions interact across development activities.

0.11

From Delegation to Action Approval: An Exploratory Study of Developers' Practices in Collaborating with AI Coding Agents

cs.SE ·2026-11-06 ·Yinan Wu, Dominik Wermke, Ze Shi Li, Kathryn Thomasset Stolee, Bowen Xu
Abstract

Developers now collaborate with AI coding agents that write code and take actions on their behalf. Beyond reviewing this code, developers decide what to delegate and which agent actions to approve. Some of these actions, such as running shell commands or reading files, can have effects that never appear in the resulting code. How developers make these decisions and how much effort this requires are not well understood. To better understand this collaboration, we surveyed 203 developers who use AI coding agents. We found that careful code review does not necessarily extend to action approval: 61% of developers who carefully reviewed most or all generated code still often approved at least one type of agent action without close review. Developers aware of security or privacy incidents or concerning practices did so more often than those who were not. We also found that autonomy varied more between developers than between development activities. Although developers reported a lighter workload, supervising agents remained mentally demanding. When asked what would make them grant more autonomy, developers wanted more than reliability: more than half of their responses called for agents that are bounded, observable, or reversible. Our findings suggest that as agents take on more of the coding, the challenge shifts toward supporting the developers who supervise them.

0.10

BOXBOX: Recursive Harness Self-Improvement for Long-Running Performance Optimization

cs.SE cs.AI ·2026-11-06 ·Yuhao Tan, Hanyun Jiang, Fangkai Yang, Yuan Yao, Taolue Chen +6 more
benchmark
Abstract · We are withdrawing this submission temporarily to add a clarification regarding the use of Claude Code, in accordance with our company’s c…

Automated harness optimization improves agents by learning from trajectories and task feedback. Existing offline methods use reference answers to optimize harnesses for reuse on held-out tasks. However, many real-world tasks are open-ended, while long runs expose new difficulties that offline-optimized harnesses may not address. To study harness optimization in these settings, we introduce PerfRace, a benchmark for long-running performance optimization of software projects, and investigate whether agents can use task-specific experience to recursively improve their harnesses during an ongoing task. To this end, we introduce BOXBOX, which enables agent-triggered harness revision and kernel-managed hot reloading while preserving execution context. We evaluate BOXBOX on PerfRace under matched test-time budgets, comparing it with Frozen Harness, offline Harness Evolution, and Harness Scaling. BOXBOX achieves higher speedup with more efficient use of task time and model cost, reaching a Speedup@4h of 1.70x, compared with 1.54x for the strongest baseline.

0.06

Theory & Algorithms 64

A proof of the Aaronson-Ambainis conjecture

cs.CC quant-ph ·2026-11-06 ·David Turturean
theorycode/data
Abstract · 60 pages. Lean formalization and reproducibility materials: https://github.com/physical-superintelligence/aaronson-ambainis

Aaronson and Ambainis conjectured that every bounded polynomial of low degree on the discrete cube has an influential variable. They showed that the conjecture implies that the acceptance probability of every quantum algorithm making $T$ queries can be approximated on most inputs by a classical algorithm making $\mathrm{poly}(T)$ queries. We prove their conjecture: every function $f\colon\{-1,1\}^n\to[-1,1]$ of degree at most $d\ge1$ has a variable of influence at least $c\,\operatorname{Var}(f)^{5/2}/d^{6}$, where $c>0$ is an absolute constant. In the original formulation, for polynomials of degree at most $d$ with values in $[0,1]$ on $\{0,1\}^n$ and variance at least $\varepsilon$, this gives a variable of influence at least $(\varepsilon/d)^{7}$. The proof is elementary, with steps inspired by works of others. Thus, as Aaronson and Ambainis anticipated, superpolynomial quantum speedups in the query model require structure, such as a promise on the inputs.

0.62

Strong Connectivity Oracle for Multiple Failures via Rooted Flow-Augmentation

cs.DS ·2026-11-06 ·Aditya Anand, Thatchaphol Saranurak, Benyu Wang
published
Abstract · SODA 2027

We present a significantly improved $k$-fault-tolerant strong connectivity oracle, one of the most basic data structures on a directed graph $G$ that, given any set $F$ of failed edges of size $|F|\le k$, answers whether $G \setminus F$ is strongly connected. For every constant $k$, our oracle only uses $n^{1+o(1)}$ space and $n^{o(1)}$ query time. No prior result can achieve $o(n^{2})$ space and $o(n)$ query time, \emph{even when $k=2$}. Previous oracles either require $Ω(n)$ query time [BCR18], or $Ω(n^2)$ space [vdBS19]. Our main technical contribution is a \emph{rooted} version of the highly successful flow augmentation technique, which may be of independent interest. Given a directed graph $G$, an integer $k$, and a source vertex $s$, we can sample a set of edges $A$ such that each minimal $s$-rooted cut of size at most $k$ becomes an $s$-rooted min-cut in $G + A$ with probability $2^{-poly(k)}n^{-o(1)}$.

0.57

An Efficient Synchronization Scheme for Distributed Bandits

cs.IT ·2026-11-06 ·Raymond Zhang, Henrique Miyamoto, Richard Combes, Sheng Yang
publishedtheory
Abstract · Accepted at IEEE ISIT2026. 3 Figures, 6 pages. Theorem 7 redrawn

We study the problem of synchronizing two players in a \emph{distributed multi-armed bandit} setting. The probability of \emph{synchronization error} is analyzed by recasting synchronization as a \emph{communication problem}. Using a well-designed random coding ensemble, we derive error exponents for the maximum-likelihood decoder. We then introduce a \emph{universal decoder} that achieves the same error exponents as the optimal decoder. Finally, applying these results to distributed multi-armed bandits, we show an upper bound on the regret of our synchronization scheme.

0.57

Enhanced Informed Dynamic BP Decoding Scheduling Strategies for 5G NR LDPC Codes

cs.IT ·2026-11-06 ·Tofar C. -Y. Chang, I-Hsiang Lee, Pin-Han Wang, Jian-Jia Weng, Yu T. Su
published
Abstract · 7 pages, 5 figures, 2022 IEEE 96th Vehicular Technology Conference

We present an enhanced informed dynamic scheduling (IDS) strategy to avoid propagations of unreliable messages in the residual belief-propagation (RBP) algorithm for decoding low-density parity-check (LDPC) codes. To reduce the resource needed for updating the degree-one variable nodes (VNs) in an RBP-based algorithm for decoding 5G New Radio (NR) LDPC codes, we further introduce a process consisting of several simple procedures for managing the resource consumption. Simulation results demonstrate that our RBP algorithm based on the Proposed IDS strategy outperforms existing RBP-based algorithms at the early decoding iterations. By using our degree-one VN resource managing approach, the performance of both the proposed and the existing decoders is significantly improved in decoding the 5G NR codes.

0.53

Fair Selection of Generalized Nash Equilibria: A Comparability Perspective

cs.GT eess.SYmath.OC ·2026-11-06 ·Sophie Hall, Heinrich H. Nax, Saverio Bolognani
published
Abstract · Book chapter, 25 pages, 4 figures. To appear in Socio-Inspired and Socially Aware Control for Complex Systems, Springer

Generalized Nash equilibrium (GNE) problems model self-interested agents coupled through their costs and shared constraints, and arise throughout control applications that allocate a scarce common resource. Such problems typically admit a continuum of equilibria, so any implementation must \emph{select} one. This selection is not neutral: it decides how the cost of the shared constraint is distributed among the agents, a decision often expected to be fair. The variational GNE (v-GNE), which charges all agents a uniform shadow price, is the conventional choice and is frequently advocated as fair. Through the lens of \emph{comparability} from social choice theory, we show that the GNE set itself requires no inter-agent comparability of costs, whereas the v-GNE implicitly assumes cardinal unit comparability, one of the stronger notions. We then introduce the fair GNE (f-GNE), which selects an equilibrium by explicitly optimizing a fairness metric matched to the comparability at hand, and give constructive routes to compute it for single-constraint and linear-quadratic games. An electric-vehicle charging game serves as a running example throughout this self-contained tutorial.

0.53

Up or down by half: a family of Collatz-type maps

cs.DM math.DS ·2026-11-06 ·Oleg Vasilyev
theoryunusual
Abstract

Each Collatz step changes a positive integer by roughly half: odd inputs go up and even inputs go down. We vary the directions while keeping the step size. Write x=2^m k (k odd), and let S be a set of positive integers. Define T_S(x)=(3x+1)/2 if m=0, 3x/2 if m is in S, and k otherwise. The map T_S^- differs only at odd x, which it sends to (x+1)/2. The usual Collatz map is T_emptyset. Assuming the independence of 2-adic valuations along odd-to-odd steps yields random-walk models with drifts (1+rho(S))log_2(3)-2 for T_S and rho(S)log_2(3)-2 for T_S^-, where rho(S) is an explicit series. Every T_S^- is contracting; T_S is contracting exactly when rho(S)<log_3(4/3), approximately 0.2619. We prove density results, classify interval sets S, and give a criterion ensuring every T_S^- orbit reaches 1 or (2,3). For seven contracting maps T_S, three with rho(S)=1/4 just below the threshold, computations show that every n<2^34 (n<1.668*10^10 for S={1}) reaches one of at most five cycles; stopping-time statistics match the model to 3-5 digits. We study branching maps, where any member of a family may be applied at each step. Boundedness of a family implies boundedness of each member; in particular, boundedness of {T_emptyset,T_{3}} would imply that no Collatz orbit diverges. For {T_emptyset,T_{1}}, a rule depending only on x mod 32 has positive drift (10/17)log_2(9/8) for almost every 2-adic starting point. Thus almost all n have reachable sets containing values larger than n^1.05; residues modulo 2^L, L<=10, cannot prove unboundedness. Finally, we define unified maps in which valuations also determine the direction at odd inputs. The cases x=1 mod 4 govern their behavior: if they go down, every block is dominated by the corresponding Collatz map (and a census finds only the cycles (1), (2, 3) and (3, 5) with smallest element below 10^9); if they go up, the maps behave like T_S.

0.49

How far does a hash-only sender go? A pre-registered map of sender memory versus bits for short text messages

cs.IT cs.CL ·2026-11-06 ·Marcelo Samoilenko
code/data
Abstract · 10 pages, 1 figure, 2 tables. Code, data and per-message results: https://github.com/msamoilenko/hash-sender-llm-receiver

Can a sender that only computes hashes, with a language model only at the receiver, transmit short human text more cheaply than a classical sender trained on the same domain? We answer with one pre-registered run on 1,000 English user utterances from Taskmaster-1. The primary endpoint is the cost per message to deliver 100% of the messages (hash bits, plus a 2-byte negative acknowledgement and the raw text when the first attempt fails), paired against an in-domain PPM order-1 sender. The hash-only sender delivered 857 of 1,000 messages on the first attempt and cost 192.8 bits per message, 18.4 bits more than PPM order 1 (95% bootstrap interval 10.9 to 26.3); by the pre-registered rule it spends more and is not equivalent within one byte. As secondary results we map measured sender memory against bits. The hash-only sender (8,190 B on a host build, keyed BLAKE2b tag included, as in every method) was the cheapest point in bits from 8,190 B up to, but excluding, PPM order 1 at 24,844 B, 45 to 46 bits below smaz2 and Unishox2; from PPM order 1 upwards, in-domain senders spend fewer bits, except order 2 pruned to 16 KiB. In airtime it loses to PPM order 1 on all five links priced, to Unishox2 on four, and to raw text with a tag on two, because each failure costs an acknowledgement frame and a fallback frame. A wider receiver search recovers half of the failures and makes the cost 7.7 bits lower than PPM order 1, at about three minutes of receiver time per failure, and still loses in airtime on four of five links. We report the measurement as a delimitation, with every deviation of the run.

0.44

The k-Clique hypothesis is dead, for good

cs.DS cs.CC ·2026-11-06 ·Virginia Vassilevska Williams, Or Zamir
Abstract

We present an algorithm that solves the $k$-Clique problem for any large enough $k$ and any $n$-vertex graph in $n^{\left(\fracω{3}-c\right)k}$ time, where $ω\geq 2$ is the matrix multiplication exponent and $c>0$ is a positive constant independent of both $ω$ and $k$. This refutes the long-standing $k$-Clique conjecture, regardless of the value of the matrix multiplication constant, even if $ω=2$. In fact, our algorithm solves the more general problem of counting weighted $k$-Cliques.

0.41

As Easy as Rank One: Optimal Algorithms for Matroid Secretary and Single-Sample Matroid Prophet

cs.DS cs.GT ·2026-11-06 ·Hamed Abdi, Kiarash Banihashem, MohammadTaghi Hajiaghayi, Zhiyi Huang, Danny Mittal +1 more
Abstract · This paper merges and extends arXiv:2609.14555, arXiv:2609.19118, and arXiv:2609.30421

We resolve the matroid secretary conjecture in its strong form and the single-sample matroid prophet inequality problem, achieving competitive ratios $e$ and $2$, respectively. Both ratios are optimal, even for rank-one matroids. For matroid secretary, our ordinal algorithm accepts the elements of the canonical optimum independently, each with probability $1/e$. It needs only the number of elements in advance and uses an expected polynomial number of independence queries on arrived elements. For single-sample matroid prophet inequalities, our algorithm uses one sample per independent value distribution and $O(n^2)$ independence queries on a known $n$-element matroid. It works for every arrival order fixed independently of the samples, values, and algorithmic randomness. The selected elements and their values have exactly the joint distribution obtained by independently retaining each element--value pair in the canonical optimum of a fresh product draw with probability $1/2$. Our main algorithms maintain a reference state whose greedy solution contains every accepted element. Its updates, made even after rejections, preserve exact distributional invariants. A short reversible-update argument gives a $4$-probability-competitive secretary algorithm, already resolving the original conjecture; a related invariant gives the optimal single-sample prophet inequality. For secretary, we characterize the distributions achievable by ordinal algorithms through a finite polytope. A continuous-time formulation then lets us design an accepted-set distribution achieving the optimal ratio $e$ and realize it efficiently. We also obtain guarantees for matroid intersections.

0.39

New persistent chain spaces: identifying and computing the relevant persistent Laplacians of a filtration

cs.CG math.ATmath.COmath.SP ·2026-11-06 ·Joris Kirchner, Martina Vittorietti
Abstract · 34 pages, 12 figures

Persistent homology (PH) summarizes the topological features of data that persist across scales, but it is insensitive to many geometric properties. Persistent Laplacians (PLs) address this limitation: for a pair of simplicial complexes in a filtration, the kernel of the persistent Laplacian recovers the persistent Betti number, while its non-zero spectrum carries additional geometric information. However, PLs are defined for individual pairs of filtration values, so analysing a whole filtration requires evaluating many operators that largely contain the same information, and the result is hard to summarize and interpret. In this paper, we show that this redundancy has a precise structure. For each pair of filtration values, we decompose the space of chains underlying the persistent up-Laplacian into the contributions inherited from the neighboring pairs and a remaining part, which we call the new persistent chain space. We show that the new persistent chain spaces of all earlier pairs form a direct sum decomposition of this space, so that every persistent up-Laplacian in the filtration is determined by them. Moreover, the new persistent chain spaces are non-trivial only at pairs given by the persistence pairing in the same and in the next homological dimension, so that the persistence pairing determines exactly where the persistent up-Laplacian can change. Experiments on several types of point clouds show that evaluating persistent Laplacians only at these pairs reduces the computation time, both for our implementation and for existing software.

0.37

Tight Variable-Occurrence Bounds and Exponential Separations for Moser--Tardos and WalkSAT on 3-CNF

cs.DS ·2026-11-06 ·Minbo Gao, Kun He, Tao Jiang
Abstract

We establish a sharp variable-occurrence threshold for polynomial expected-time convergence for the Moser--Tardos algorithm and the pure random-walk WalkSAT on 3-CNF formulas. For every formula with $m$ clauses and at most two occurrences per variable, both algorithms terminate in $O(m)$ expected updates. This occurrence bound is sharp: we construct an infinite family with at most three occurrences per variable on which both algorithms require $2^{Θ(\sqrt{m})}$ expected updates. Surprisingly, this dynamical transition occurs strictly below the classical NP-completeness threshold at variable-occurrence degree four. We further establish exponential separations between the two algorithms on satisfiable 3-CNF formulas. On one family, WalkSAT terminates in $O(m^4)$ expected flips, whereas MT requires $2^{Ω(m)}$ expected resampling steps. On another, MT terminates in $O(m)$ expected resampling steps, whereas WalkSAT requires $2^{Ω(m)}$ expected flips. These constructions exploit the different ways in which single-variable flips and simultaneous resampling propagate local violations. Our results show that the algorithms' expected running times are incomparable even up to polynomial overhead on the same inputs.

0.33

A Note on the Hardness of Partially Dynamic Max Flow

cs.DS ·2026-11-06 ·Thatchaphol Saranurak, Benyu Wang
theoryhot direction
Abstract · SOSA 2027

We study the fine-grained complexity of partially dynamic max flow in capacitated graphs. Assuming the combinatorial 4-clique hypothesis, combinatorial algorithms require $Ω(n^{2-o(1)})$ amortized update time. Thus, recomputing max flow after each update is optimal up to subpolynomial factors for combinatorial algorithms in dense graphs. The previous conditional lower bound was $Ω(n^{1-o(1)})$ amortized update time under SETH or the OMv hypothesis and applies to all algorithms [Dah16].

0.33

From Function Inversion to Preimage Reporting

cs.DS ·2026-11-06 ·Mathieu Vannimmen
Abstract · 14 pages, 3 figures

Given an arbitrary function $f:[N]\to[M]$, the function inversion problem asks to preprocess $f$ and to be able to, when given some $y\in [M]$, decide whether there exists some $x\in [N]$ such that $f(x)=y$ and if so, to report such an $x$. Recently, Fiat-Naor's solution to this problem has seen significant traction in theoretical computer science, and allowed for several conjectures related to meta-complexity and the 3SUM problem to be broken. In this paper, we consider the natural extension of the function inversion problem that we call preimage reporting. Here, instead of having to report only one $x\in [N]$ such that $f(x)=y$, the whole set $f^{-1}(y)$ must be found. We present a new general method to convert any arbitrary function inversion scheme into an efficient corresponding preimage reporting scheme. In particular, if $Δ$ is the maximal indegree of $f$, we show that we can perform preimage reporting using space $S$ in time $T$ for any $T$ and $S$ satisfying $TS^2=N^2 Δ$, matching a variant of a tradeoff obtained by Fiat and Naor for function inversion. This last result implies an improvement over the corresponding function inversion scheme itself, which constitutes the first function inversion scheme whose query time is inversely proportional to $|f^{-1}(y)|$. We derive new simple solutions for the $k$SUM-indexing with reporting and Jumbled indexing with reporting problems.

0.33

Deterministic factorization of constant depth circuits from whitebox PIT

cs.CC ·2026-11-06 ·Somnath Bhattacharjee, Mrinal Kumar, Shanthanu S. Rai, Varun Ramanathan, Ramprasad Saptharishi +1 more
Abstract

We show that over the field of rational numbers, there is a deterministic algorithm for factorization of constant depth circuits that makes calls to a (whitebox or blackbox) PIT algorithm for such circuits. The time complexity of the algorithm (and hence the number of calls to the PIT oracle) is polynomially bounded in the size and degree of the input circuit. Our reduction to PIT is fairly simple and general and applies as it is to other natural classes of algebraic circuits like formulas and algebraic branching programs. Prior to this work, the known deterministic factorization algorithms for constant depth circuits crucially relied on \emph{blackbox} PIT [Bhattacharjee et al.]. Our main technical ingredient is an efficient, deterministic algorithm that, when given access to sufficiently high degree truncations of \emph{all} the power series roots of a constant depth circuit, uses oracle calls to any (whitebox or blackbox) PIT algorithm for these circuits, and partitions the roots into the sets of conjugates; thereby identifying the irreducible factors. For our proof, we rely on a root recombination criterion using logarithmic derivatives [Sasaki et al., Bostan et al., Lecerf] and show that this criterion can be efficiently and deterministically implemented for power series roots of constant depth circuits using access to any PIT oracle. On the way, we discuss an algorithm that may be of independent interest: it takes as input a matrix whose entries are multivariate constant depth circuits over a field $\mathbb{F}$, requires oracle access to any PIT algorithm for these circuits and outputs a basis for the $\mathbb{F}$-kernel of this matrix in polynomial time.

0.33

Attaining Optimal Values under Sure Constraints

cs.GT cs.LO ·2026-11-06 ·Takuto Yamauchi, Kenji Tei
Abstract · 61 pages, 6 figures. Complete proofs included. Corresponding author: Takuto Yamauchi

A game may have, for each target below its value, a finite controller meeting that target, yet no finite controller attaining the value itself. We study finite stochastic games with a hard specification that must hold on every supported execution, including paths of probability zero. Max uses a pure finite controller; Min may use any pure public-history strategy. We distinguish three numerical questions: can a supplied target be met, can the unspecified optimum be attained, and what is one exact value? For explicit hard and Boolean soft Streett conditions, the strict and non-strict supplied-threshold problems are coNP-complete, attainment of the unsupplied supremum is $\mathrm P^{\mathrm{NP}}$-complete, and one exact reduced fraction is $\mathrm{FP}^{\mathrm{NP}}$-complete under polynomial metric reductions. The proofs preserve both the value and its finite attainability when fixing a single opponent, and encode adaptive queries without computing unknown subgame values. The same construction yields matching value-comparison and exact-output lower bounds for ordinary stochastic Streett games with unrestricted randomized history strategies. We also bound the memory needed for optimal finite implementations for arbitrary Muller hard conditions and rational rewards in $[0,1]$ determined by infinitely recurring states. One controller is sure from the entire original sure region and attains at every finitely attained start, with no attainment premise on the other states. For $n$ states in that region it uses $2^{O(n\log(n+1))}Q(1+\log R)$ modes, where $Q$ is the product of the least common denominators of full original chance rows sourced in that sure region and $R$ clears all reward values. Its joint dependence on $Q$ and $R$ is worst-case sharp at each fixed $n\ge6$ when $Q,R\ge2$. Optimal dependence on arena size and efficient general synthesis are not resolved here.

0.32

A Fast and Accurate Algorithm for the $p \to q$ Norm of Sparse Non-Negative Matrices

cs.DS ·2026-11-06 ·Marco Pirazzini, Daniel A. Spielman
Abstract · 31 pages, 6 figures

We present an algorithm that computes the $p \to q$ norm of a non-negative matrix with $s$ nonzero entries to accuracy $\varepsilon$ in time $\widetilde{O}(s^{3/2} \log (1/\varepsilon))$, for any $p \ge q \ge 1$. The fastest algorithm previously designed for this problem has running time $\widetilde{O}(s / \varepsilon)$ (Objois and Vladu, 2026), which is slower when high accuracy solutions are desired. Our algorithm employs a surprising application of Laplacian solvers inside an Interior Point Method (IPM). We formulate the problem as a linear optimization problem over a product of power cones, and show that the Newton updates of the IPM can be computed by solving a Laplacian system in a Schur complement after subtracting a rank-2 matrix. We also discuss an implementation of our algorithm, showing that it is practical and it outperforms optimized solvers such as MOSEK by factors of up to $20\times$ on a variety of tasks.

0.31

Parallel Algorithms for Higher-Order Temporal H-Index Computation

cs.DS ·2026-11-06 ·Lutz Oettershagen
Abstract

Ranking vertices by importance is a fundamental task in network mining. The H-index provides such a ranking by asking how many neighbors are themselves well connected; its higher-order temporal variant recursively applies this idea over time-respecting neighborhoods in temporal networks. Scaling its computation is difficult because the existing streaming algorithm intertwines temporal processing with H-index order, incurring maximum-degree-dependent costs and exposing little parallelism. We show that the temporal dependencies can instead be resolved once: each edge's time-dependent lookup is mapped to a fixed successor position, turning every H-index order into a parallel gather followed by a segmented suffix-H computation. Based on this reformulation, we derive a linear-work suffix-H algorithm, a parallel multicore CPU implementation, degree-adaptive GPU kernels, and a multi-GPU scheme that exchanges only distinct remotely queried states. Experiments on synthetic and real-world temporal graphs show that the reformulation alone is substantially faster than the streaming algorithm, and that GPUs add order-round speedups exceeding an order of magnitude over our multicore implementation on synthetic inputs with up to one billion temporal edges. On a 636M-edge Reddit network, eight GPUs compute 128 orders in 3.6 s of round time, whereas the streaming algorithm does not finish within an hour even at order 8.

0.30

Path Isolation and Reachability in Surface-Embedded Graphs with Restricted Cycle and Cut Structure

cs.CC ·2026-11-06 ·Neelabjo Shubhashis Choudhury, Chetan Gupta, Raghunath Tewari
theory
Abstract · 18 pages. 1 figure

We study the reachability problem in directed surface-embedded graphs under two structural restrictions, using isolating weight assignments. First, we consider graphs whose underlying undirected embeddings contain at most polynomially many simple surface non-separating cycles. We develop a polynomially bounded, logspace-computable skew-symmetric edge-weight assignment with nonzero circulation around every simple cycle. This yields an unambiguous logspace (UL) upper bound for directed reachability. The construction also gives an SPL upper bound for bipartite perfect matching in the corresponding undirected graph class. Second, for $g$-genus directed graphs admitting a balanced hierarchical genus decomposition with $O(\log n)$ crossing edges at every partition, we obtain an unambiguous $O(\log g \cdot \log n)$-space bound for reachability. Our approach combines path-isolating weights for lower-genus subgraphs using polynomially bounded FKS hashing refinements at successive merging levels.

0.30

Global Sinkhorn Convergence and Spectral Stability of Matrix Scaling

cs.DS ·2026-11-06 ·Kun He
theory
Abstract · 71 pages

We give a global spectral bound for the Sinkhorn--Knopp (SK) algorithm. Let $A$ be a row-stochastic nonnegative $n\times n$ matrix admitting a finite doubly stochastic scaling $B$. Define \[ g=1-s_2(B)^2>0, \qquad D=\min\left\{\|(u,v)\|_\infty: B=\operatorname{diag}(e^u)A\operatorname{diag}(e^{-v})\right\}. \] From the standard initialization, SK achieves total marginal $\ell_1$ error $\varepsilon$ in \[ O\!\left(\frac{D+\log n+\log(1/\varepsilon)}{g}\right) \] iterations. A scale-invariant potential contracts globally according to the spectrum of $B$. The linear dependence on $D$ is necessary in the fixed-gap regime: we construct strictly positive inputs with constant input and endpoint spectral gaps, $D=Θ(n)$, and an $Ω(n)$ iteration lower bound at error $n^{-2}$. Our upper bound gives $O(n)$ iterations on this family. We next give spectral certificates for scalability and short scaling distance. For a row-stochastic input $A$ with positive column sums $c_j$, set \[ K=\frac{\max_jc_j}{\min_jc_j},\qquad σ=s_2\!\left(A\operatorname{diag}(c)^{-1/2}\right). \] The input-only condition $\sqrt K\,σ<1$ certifies exact scalability and polynomial scaling distortion, with $D=O(\log n)$ for fixed $K$ and a fixed positive spectral margin. Given a finite scaling $B$, the weaker condition $\sqrt K\,σs_2(B)<1$ gives the same structural bounds. With a constant endpoint gap, these bounds imply logarithmic SK iteration complexity. Boundary constructions show that the strict spectral inequalities cannot uniformly be replaced by nonstrict ones.

0.30

Reconfiguration of Time-Bounded Descriptions

cs.CC cs.DM ·2026-11-06 ·Dimitrios Myrisiotis
Abstract · 50 pages

Time-bounded Kolmogorov complexity measures a string by the length of its shortest program that prints it within a deadline. An adjunction-universal machine (Bauwens, Kozachinskiy, and Shen, arXiv, 2025) is a Turing machine that simulates every Turing machine $M$ on every input $z$ with polynomial overhead when given a code of $M$, computable from $M$ in polynomial time, followed by $z$ unchanged. Any two adjunction-universal machines give the same measure up to an additive constant and a polynomial change of the deadline. However, this measure records only the shortest program for a string. We study all programs of one length that print a string within one deadline, join two of them when they differ in one bit, and ask when this graph is connected. Fix any adjunction-universal machine. For almost all $m$-bit strings and every deadline beyond some polynomial, the graph is disconnected at program length $m+O(1)$ and has at least $2^{m-O(1)}$ components at length $2m+3$. In contrast, we build an adjunction-universal machine in which every such graph is connected or empty, at every deadline, once the program length reaches a computable threshold that is at most exponential in the length of the string. Exponential thresholds are unavoidable: for every adjunction-universal machine with such thresholds and every large $m$, more than three quarters of the $m$-bit strings need a threshold exponential in $m$. For our machine, above the threshold, every program for the string is within three bit flips of a connected set of fastest programs for the string, along a path inside the graph, at every deadline by which that program halts. For this machine, deciding whether two given programs are connected in this graph is $\mathsf{PSPACE}$-complete when the input is the string, the deadline in unary and the two programs; for each fixed string it takes polynomial time.

0.29

Flexible Edge Cover: A Disjunctive Programming driven Approximation

cs.DS ·2026-11-06 ·Karthekeyan Chandrasekaran, Krishna Kalathur
hot direction
Abstract

We introduce the flexible edge cover problem, denoted FLEX-EDGE-COVER: the input here is an undirected graph with non-negative edge costs along with a partition of the edges into safe and unsafe edges, non-negative integers $p$ and $q$, and the goal is to find a minimum cost subset of edges such that each vertex has either at least $p$ safe edges incident to it or at least $p+q$ total edges incident to it. We show that FLEX-EDGE-COVER is APX-hard in simple bipartite graphs. A greedy approach leads to a $2$-approximation in all graphs. We show that the $2$-factor can be improved in simple graphs via a novel disjunctive programming-based LP formulation. AI-tools were used in the analysis of our rounding algorithm. Our motivation for introducing and investigating FLEX-EDGE-COVER is its close connection to the Flexible Graph Connectivity problem (denoted $(p,q)$-FGC) which has garnered much recent attention. For $(p,q)$-FGC, we develop a tighter LP-formulation via the disjunctive programming approach. We hope that our disjunctive-programming based LP and rounding algorithms for FLEX-EDGE-COVER will serve as a starting point towards a constant factor approximation for $(p,q)$-FGC.

0.28

Fully Dynamic Correlation Clustering with Vertex Updates in Polylogarithmic Update Time

cs.DS ·2026-11-06 ·Vladimir Braverman, Prathamesh Dharangutte, Chenglin Fan, Chen Wang, Shenghao Xie +1 more
Abstract

We present a fully dynamic algorithm for correlation clustering that supports \emph{vertex} insertions and deletions while maintaining an $O(1)$ approximation in polylogarithmic amortized update time. This resolves an open problem of vertex deletions posed by a recent line of work on dynamic correlation clustering. We adopt the min-disagreement objective for correlation clustering, i.e., given a signed complete graph $G=(V,E^+\cup E^-)$, we want to minimize the total number of $(+)$ edge crossing clusters and $(-)$ edges within clusters. The input is given in the adjacency list model, where each update inserts or deletes a node along with its full set of $(+)$ neighbors. Our main technical contribution is a fault-tolerant sparse-dense decomposition framework which unlike the existing decomposition-based frameworks allows for structural inconsistencies as long as the induced cost is bounded. We also show a separation between explicitly maintaining the (stronger) sparse-dense decomposition and maintaining a constant-factor approximation to correlation clustering, which may be of independent interest.

0.28

Randomized Greedy MIS Is Fast at Every Node of a Tree

cs.DS ·2026-11-06 ·Mina Dalirrooyfard, Konstantin Makarychev, Slobodan Mitrović, Srikkanth Ramachandran
theory
Abstract

The Randomized Greedy MIS (RGMIS) algorithm computes a maximal independent set (MIS) by processing the vertices in a uniformly random order and greedily adding each vertex to the independent set whenever possible. Despite its simplicity, this algorithm has proved effective in a range of parallel, distributed, and dynamic settings. For sufficiently small maximum degree, Nguyen and Onak (FOCS~'08) and Yoshida, Yamamoto, and Ito (STOC~'09) showed that a recursive simulation of RGMIS can be used to estimate the sizes of matching and vertex cover in sublinear time. The key subroutine in their approach is the following local query procedure: given a vertex $v$, to determine whether $v$ belongs to the MIS, recursively evaluate the neighbors of $v$ that precede $v$ in the random order. If any such neighbor is found to belong to the MIS, the procedure immediately concludes that $v$ is not in the MIS; otherwise, it concludes that $v$ is in the MIS. For graphs of maximum degree $Δ$, Nguyen and Onak showed that, in expectation over the random ordering, this procedure makes $O(\exp(Δ))$ recursive calls from a worst-case starting vertex. No better upper bound is known even for trees. We prove that on trees, from \textit{any} starting vertex, this recursive RGMIS procedure makes $\mathrm{poly}(Δ)$ recursive calls in expectation.

0.28

Quantum Interactive Proofs of Proximity

cs.CC ·2026-11-06 ·Kenan Wood
theoryhot direction
Abstract

We initiate the systematic study of quantum interactive proofs of proximity (QIPPs). These are quantum interactive proof systems with sublinear communication, where the verifier accepts inputs in a property $Π$ and rejects inputs $\varepsilon$-far from $Π$, while only querying the input a sublinear number of times. 1. We show that all logspace-uniform $\mathbf{NC}$ properties admit QIPPs with query, communication and time complexity all $\widetilde{O}(n^{2/5})$. This improves on the best known classical total complexity of $\widetilde{O}(\sqrt n)$, which is optimal in the worst case for polynomial-time verifiers under the PRG assumption of Kalai and Rothblum (CRYPTO, 2015). This resolves an open question of Dall'Agnol, Gur, Roy Moulik, and Thaler (Quantum, 2022) in the affirmative. 2. Assuming the existence of a post-quantum subexponentially secure PRG computable in logspace-uniform $\mathbf{NC}$, we prove that, for every nondecreasing, logspace-computable query bound $q(n)\le n^{1/2-Ω(1)}$, there exists a property $Π$ in logspace-uniform $\mathbf{NC}$ and a constant $\varepsilon>0$ such that no polynomial-time $\varepsilon$-QIPP for $Π$ can achieve query complexity $q(n)$ and communication complexity $c(n)$ satisfying $q(n)+c(n)=o(n^{1/3})$. 3. We establish a bi-directional connection between QIPPs for logspace-uniform $\mathbf{NC}$ and query-model QIPs for parity. A QIP for $k$-bit parity with total complexity $k^{2/3-Ω(1)}$ yields QIPPs for every logspace-uniform $\mathbf{NC}$ property with total complexity $n^{2/5-Ω(1)}$. Conversely, achieving total complexity $o(n^{2/5})$ within the standard $\mathsf{PVAL}$-based framework requires improving the known query-communication tradeoff for parity. Additionally, we show that almost every size-$2^{βn}$ property ($0<β<1$) requires linear total QIPP complexity at constant proximity.

0.27

Dual-Number Associative Arrays for One-Pass Provenance and Sensitivity

cs.DS ·2026-11-06 ·Timothy Habashy, Hayden Jananthan, Jeremy Kepner
theory
Abstract · 5 pages, 1 figure

Associative arrays and GraphBLAS express sparse computations as matrix operations over a chosen semiring, but the result normally contains only the ordinary value. We define a dual-number extension $D_k(S)=S\ltimes S^k$ that carries this value together with $k$ independent auxiliary channels in the same computation. We prove that the construction is a split square-zero extension of any commutative semiring. Over min-plus, the auxiliary value is the weight of the shortest walk constrained to use the corresponding edge, excluding the weight of that edge. This gives an exact shortest-path membership test for that edge, although ties return the union of the edges on all optimal paths. We implement the min-plus construction using sparse \texttt{python-graphblas} operations, with a fixpoint test that checks the value and auxiliary components together. We validate the implementation across 18,692 checks, including 2,825 tied target pairs. On random graphs with 100,000 vertices and 32 seed edges, the dual method is $3.39\times$ faster than the sequential algorithm but $1.50\times$ slower than a min-plus-specific batched method. This difference is largely explained by the dual method requiring $1.55\times$ as many sweeps. Thus, the main contribution is a general algebraic approach with competitive sparse performance, rather than information that specialized min-plus algorithms cannot already compute.

0.27

Beyond the $n/2$ Threshold: Quantum Query Complexity for Computing Hamming Weight and Symmetric Functions

cs.CC quant-ph ·2026-11-06 ·Yuan Deng, Zihan Hao, Honghao Lin, Qipeng Liu, Vahab Mirrokni +2 more
Abstract · 57 pages, 2 figures

The oracle-interrogation algorithm [van Dam, FOCS '98] computes every $n$-bit Boolean function with bounded error using $n/2+O(\sqrt n)$ quantum queries. In contrast, parity can be computed exactly using $\lceil n/2\rceil$ queries, but whether an additive overhead beyond $n/2$ is necessary for a single Boolean output has remained open. In this work, we construct an explicit family of symmetric Boolean functions requiring $n/2+Ω(n^{1/3})$ quantum queries at a fixed positive constant error. We also give an algorithm that recovers the exact Hamming weight using $n/2+O(n^{1/3})$ quantum queries with bounded error, and therefore computes every symmetric Boolean function within the same query bound. Together, these results determine the optimal additive order for Hamming weight recovery and for the worst case over symmetric Boolean functions at sufficiently small constant error.

0.27

Separating Measures by Hyperplanes

cs.CG math.MG ·2026-11-06 ·José Correa, Svenja M. Griesbach, Elfarouk Harb, Felix Hommelsheim, Reda Jlibene +2 more
theory
Abstract · 58 pages, 5 figures

Given red and blue points in $\mathbb{R}^d$, connect every red point to every blue point and weigh each segment by its length under a fixed norm. What fraction of the total weight can always be captured by a single affine hyperplane separating the segments' endpoints? We study this question for finite point configurations in affine general position and, more generally, for pairs of probability measures. For dimension one, we show that choosing the best of three thresholds determined by the means of the two distributions captures at least one third of the total red-blue distance. We also give a matching upper bound. Embedding the line construction to higher dimensions shows that the guarantee of $1/3$ cannot be improved for arbitrary measures. Thus, we pivot to the family of measures whose atoms are in affine general position and whose non-atomic part places zero mass on all hyperplanes. Crucially, we establish that the worst-case guarantee is identical for finite configurations in affine general position and several natural classes of probability measures, thereby unifying the discrete and continuous formulations of the problem. For this family of measures and for higher dimensions, we extend our guarantee of $1/3$ for all norms. Our proofs combine the ham-sandwich theorem with a novel coupling argument and, in the planar case, an integral representation of norms. We also show that the optimal guarantee improves as the dimension grows, giving a lower bound of $1/2-\operatorname{O}\left(1/\sqrt{d}\right)$ as well as an asymptotically matching upper bound of $1/2$ for every $d$. For the Euclidean distance, we refine the lower bounds to $8/(8+5π)$ and $1/2-\operatorname{O}\left(1/d\right)$ and, for $d=2$, the upper bound to $5/11$. We also derive dimension-dependent guarantees for finite point configurations, including perfect separation of $n$ points of each color when $d\geq2n-1$.

0.27

Load Balancing with Item Fragmentation and More Bins than Items: Linear-Time and Exact

cs.DS cs.DC ·2026-11-06 ·Ben Bals, Tomáš Flouri, Kassian Kobert, Solon P. Pissis, Leen Stougie
theory
Abstract

We study a fundamental load-balancing problem in which $N$ items of strictly positive integer sizes with total size $W$ must be assigned to $B$ bins, with items allowed to be divided into integer-size fragments across bins. Each bin must receive a load of either $\lfloor W/B \rfloor$ or $\lceil W/B \rceil$, ensuring that the loads of any two bins differ by at most one unit. Under this constraint, the objective is to minimize the maximum number of distinct item fragments placed into any single bin. This problem is NP-hard and there exists an $(\mathrm{OPT}+1)$-approximation algorithm that runs in $O(N \log N + B)$ time [Kobert et al., 2014]. The existing algorithm is employed by several widely used production codes for phylogenetic reconstruction. Here, we focus on the case $B \ge N$, that is, bins are at least as numerous as items. This setting arises pervasively in modern parallel and distributed systems, as well as in iterative algorithms that rebalance their workload at every step. In this context, bins are machines or machine cores, and items represent computational tasks (jobs). Our central contribution is a remarkably simple $O(N)$-time algorithm that computes an optimal assignment in this setting; this assignment is encoded using $O(N)$ space. In particular, we prove that an optimum always exists in which every bin receives at most two fragments and provide a characterization of when one fragment per bin suffices. Our result strictly improves the $(\mathrm{OPT}+1)$-approximation of Kobert et al. in the many-bins setting and, together with it, yields a strongly polynomial-time $(\mathrm{OPT}+1)$-approximation algorithm for the general setting, as it removes the dependency on the number $B$ of bins.

0.26

Contention Resolution Against an Adversarial Scheduler

cs.DS ·2026-11-06 ·Michael A. Bender, Martín Farach-Colton, Andrew Krapivin, Pritom Kundu, John Kuszmaul +3 more
theory
Abstract · 15 pages

In the classical contention-resolution problem, a set of players share a broadcast channel, and a player succeeds in a given time step if it is the only player to broadcast during that step. Implicit in the classical problem is the assumption that every player is free to act on every step: once a player arrives, it participates until it succeeds. We study what happens when that assumption is removed. In our model, an adversarial \emph{scheduler} decides, on each step, which subset of the players gets scheduled to participate in the step. An unscheduled player simply sleeps through the step. We measure a protocol by its \emph{efficiency}: the number of players divided by the expected number of steps in which at least one player was active. Our first result says that, so long as the players are given access to a global clock and to shared randomness, they can achieve efficiency $Ω(1/\log n)$. The players are able to achieve this bound without requiring any channel feedback, and even if the adversarial scheduler is adaptive. We also give a matching lower bound of $O(1 / \log n)$, which holds even with channel feedback, and even for an oblivious scheduler. Our upper bounds all require the players to have access to a shared global clock and to shared randomness. Our final result says that this type of implicit shared state is necessary. If the players have neither a global clock nor shared randomness, then no protocol can achieve efficiency better than $O(\log n / n)$; and this holds even if the other aspects of the problem are designed to be as favorable for the players as possible (the channel gives ternary feedback and the adversary is oblivious).

0.26

Hardness of Approximation for Polynomial Optimization over Sphere for Linear Programming Relaxation Hierarchy

cs.CC ·2026-11-06 ·Mrinalkanti Ghosh
theory
Abstract

Semidefinite programming (SDP), particularly the sum-of-squares (SoS) hierarchy, provides $n^{O(q)}$-time approximation and certification algorithms for optimizing degree-$d$ homogeneous polynomials on the unit sphere. For arbitrary inputs, degree-$q$ SoS gives a powered-objective approximation factor $O_d((n/q)^{d/2-1})$ [BGGLT17]. For random order-$d$ tensors, degree-$d$ SoS has an integrality-gap lower bound of $2^{-O(d)}(n/d)^{d/4-1/2}$ relative to the true spherical maximum; lower bounds also extend to growing SoS degree in tensor PCA [BGL17,HKP17]. We establish lower bounds for weaker linear-programming relaxation hierarchies. Our degree-$q$ moment LPs impose the sphere identities and any prescribed, instance-independent family $\mathcal{G}$ of valid polynomial inequalities of degree at most $q$. They include Sherali--Adams-type product hierarchies. For every even $d\ge2$, constant $C>0$, and $q$ satisfying $2d | q$ and $2d\le q\le n$, every such relaxation with at most $n^{Cq}$ tests has instances $P$ satisfying \[ \frac{\mathrm{PowLP}_{q,\mathcal{G}}(P)}{\|P\|_\infty}\ge \frac{c_{d,C}}{\sqrt{q\log n}}\left(\frac nq\right)^{d/2-1/2}.\] Here $\|P\|_\infty$ is the maximum absolute value of the polynomial over the sphere and $\mathrm{PowLP}_{q,\mathcal{G}}(P)$ is the relaxed objective. The exponent $d/2-1/2$ improves the leading dependence on $d$ relative to the $d/4$-type exponents in prior random-tensor SoS lower bounds. Our LP hierarchies impose weaker positivity conditions than semidefinite moment relaxations. It leaves open whether analogous degree dependence can be achieved in lower bounds for SDP hierarchies.

0.25

Additive $Ω(\log^* OPT)$ Gap for Bin Packing with Three Items per Bin

cs.DS ·2026-11-06 ·Ilan Doron-Arad
theory
Abstract

We prove that bin packing is NP-hard to approximate within an additive factor of $o(\ls\OPT)$, even when all item sizes are arbitrarily close to $1/3$. This gives the first lower bound for bin packing with an additive factor that depends explicitly on the optimum value. In addition, we give a polynomial-time reduction from any bin packing instance into one with item sizes arbitrarily close to $1/3$, preserving its additive configuration-LP gap up to constant factors. Thus, in the sense of additive configuration LP gap and up to a constant, we confirm the conjecture that 3-partition instances are the hardest bin packing instances. In particular, we obtain an unconditional $Ω(\ls\OPT)$ gap for 3-partition instances. We build on the recent OpenAI construction that uses five items per bin and rules out any constant additive gap.

0.25

Stabilizing Vector Approximate Message Passing with Circuit-Designed Lag-Lead Damping

cs.IT ·2026-11-06 ·Mohamed Akrout
unusual
Abstract

Vector approximate message passing (VAMP) often fails to converge on ill-conditioned matrices, and its damping factor is tuned either manually or adaptively. Using the fact that VAMP iterations are the discretization of an electric circuit, we show that the denoiser and the likelihood act as nonlinear resistors, the Onsager correction is a mirror resistor and damping is a lag filter. This circuit gives a closed-form loop gain whose spectrum can predict when VAMP fails. Because a lag filter cannot isolate the oscillating modes, we add a lead capacitor whose design often results in a negative capacitance. Across a wide range of condition numbers, our proposed damping strategy achieves a higher convergence rate than adaptive damping and Anderson acceleration at a lower mean iteration cost. On ill-conditioned multipath channels, across ranks and sparsity levels, it also converges more reliably than existing damping baselines.

0.25

Precedence Constraint Scheduling on a Fixed Number of Machines

cs.DS cs.CC ·2026-11-06 ·Klaus Jansen
Abstract

We give a deterministic exact algorithm for $P_m\mid\mathrm{prec},p_j=1\mid C_{\max}$ for every fixed $m\ge3$. The algorithm constructs an optimum schedule in $$ O(L)+m^{O(m^2)}\, n^{6m^2+13m+1} \, {\log n} $$ elementary RAM operations, where $n \ge 2$ is the number of jobs and $L$ is the binary input length. Its word size is $O(m^2 \log n)$ bits. An implementation with a fixed number of tapes uses $O(L)+m^{O(m^2)} \, n^{12m^2+26m+2} \, \log^3 n$ steps. For three machines the respective polynomial degrees are $94$ and $188$, apart from the displayed logarithmic factors.

0.24

Dynamic Edge Orientation on Low-Arboricity Graphs: Breaking the $\log n$ Barrier

cs.DS ·2026-11-06 ·Sam Hiken, Nicole Wein
theory
Abstract · 29 pages, 3 figures

We study the dynamic edge orientation problem introduced in a seminal 1999 paper by Brodal and Fagerberg: given a fully dynamic graph $G$ with arboricity bounded by $α$, maintain an orientation on $G$'s edges so every vertex has out-degree upper bounded by a function of $α$. They provided an algorithm with $O(α+ \log n)$ amortized update time that maintains an orientation with maximum out-degree $O(α)$. In the decades since, this algorithm has found widespread applications in dynamic graph algorithms. Nevertheless, it has remained a stubborn open question whether, when $α= o(\log n)$, the $O(\log n)$ upper bound can be improved. We answer this 27-year-old question in the affirmative, showing that Brodal and Fagerberg's algorithm for maintaining an out-degree $O(α)$ orientation runs in $O(α\cdot \log \log n)$ amortized update time, which is an exponentially-improved dependence on $n$. This is the first improvement on Brodal and Fagerberg's upper bound even for the $α=1$ case of dynamic forests. Our crucial technical innovation is to construct a hierarchy of dynamic graphs with a property we call near isolation. We believe that this construction may be useful for related problems.

0.24

On derandomizing the line-point low-degree test

cs.CC cs.IT ·2026-11-06 ·Prahladh Harsha, Swastik Kopparty, Mrinal Kumar, Ramprasad Saptharishi, Madhu Sudan
Abstract · Abstract shortened due to arxiv requirements

We investigate the derandomized version of the classical line-point low-degree test in the high-error regime in the bivariate setting and reveal some unexpected behavior. The classical line-point low-degree test takes an $m$-variate function $f$ over a finite field $\mathbb{F}_q$, queries it along a random line and accepts if the restriction of the function to this line is a low-degree polynomial. The high-error version of (the analysis of) this test asserts that if $f$ has at least $ε$ expected local agreement with a degree-$k$ univariate polynomial on a random line, then in fact $f$ is non-trivially close to some degree-$k$ $m$-variate polynomial even as $ε$ approaches $0$ (provided $q$ is sufficiently large compared to $k$ and $1/ε$). The derandomized version of the problem asks if the same holds when the line is not chosen uniformly over all possible lines but rather from a small set. In this paper we explore the bivariate case ($m=2$). We show, somewhat surprisingly, that there exists a function $f\colon \mathbb{F}_q^2 \to \mathbb{F}_q$ and a set of lines $\mathcal{L}$ in $\mathbb{F}_q^2$ of size nearly $q^{3/2}$ such that on every line in $\mathcal{L}$, $f$ has agreement $\approx 1/2$ on average with some univariate polynomial of degree $k=\sqrt{q}/2$ and yet $f$ is $1 - O(q^{-1/2})$ far from every polynomial of degree $O(k)$. On the positive side, we show that once the field size $q$ is at least $k^2/(ε)^{O(1)}$ such examples disappear and we can get sets of lines of size $O(q/ε^{O(1)})$ on which the low-degree test has a high-error analysis. On the way to our positive result, we show a new high-degree version of the classical two-direction Polishchuk--Spielman divisibility lemma.

0.24

On the Communication Complexity of Shortcut Set Computation

cs.DS ·2026-11-06 ·Gregory Bodwin, Luba Samborska
theory
Abstract

A recent breakthrough of Kogan and Parter (SODA '22) provided the first polynomial improvement in decades on the size/diameter tradeoff available for shortcut sets. However, algorithmic applications of shortcut sets have mostly not taken advantage of these new bounds. A barrier is that many shortcut set applications occur in \emph{decentralized} settings, and we generally lack decentralized algorithms implementing the Kogan-Parter bounds. To this end, we study the \emph{communication complexity} of constructing Kogan-Parter shortcut sets. The edges of an $n$-node directed graph are split between two players, Alice and Bob, who are also given a target diameter $D$. Our main result is a protocol that, for $D \le n^{1/3}$, allows Alice and Bob to agree on a shortcut set matching the Kogan-Parter tradeoff between size and diameter, while exchanging only $\Oish(n^2 / D^2)$ bits of communication. We also provide (non-matching) lower bounds on the communication complexity required to find a correct $D$-diameter shortcut set of any size. The main technical ingredient for the upper bounds is a low-communication protocol for finding a \emph{path cover}, a common subroutine in shortcut set constructions that has attracted some independent interest. Our protocol crucially avoids applications of Menger's theorem or antichain decompositions, as often done in prior work. On the lower bounds side, we relate the communication complexity of the shortcut set problem to the size of bridge-free path systems.

0.23

Improved Approximation and Hardness Bounds for $β$-TSP

cs.DS cs.CCmath.OC ·2026-11-06 ·Paul Liu, Jack Spalding-Jamieson, Da Wei Zheng
Abstract · 12 pages

The traveling salesman problem (TSP) asks for a minimum-length Hamiltonian cycle in a complete weighted graph. Its approximability depends on the edge lengths: unrestricted instances admit no constant-factor approximation unless $\mathrm{P}=\mathrm{NP}$, metric instances admit Christofides' $3/2$-approximation, and fixed-dimensional Euclidean instances admit a PTAS. We study $β$-TSP, where the edge lengths satisfy $ \ell_{uw}\leqβ(\ell_{uv}+\ell_{vw}) \qquad \forall\,u,v,w\in V\ \text{pairwise distinct}, \quad β\in\left[\tfrac{1}{2},1\right]. $ The condition arises naturally in various settings, for example, when metric travel times are combined with positive service times at the vertices. The previous best approximation algorithm, due to Böckenhauer et al., ranges from $1$ at $β=1/2$ to $3/2$ as $β$ approaches $1$. We obtain a strictly better guarantee for every $1/2<β<1$ by combining two simple constructions whose guarantees counterbalance each other. We also improve the best known approximation hardness bound for every $1/2<β<1$.

0.23

Universal Decoding for Gaussian MIMO-ISI: Spherical Ensembles and Symmetric Neighborhoods

cs.IT ·2026-11-06 ·Nir Elkayam, Meir Feder
theory
Abstract · 24 pages, 1 figure

We construct a universal decoder for cyclic Gaussian multiple-input multiple-output channels with intersymbol interference and unknown channel taps and spatial noise covariance. Codewords are drawn independently and uniformly from a sphere of fixed total energy. Under stated bounds on frequency variation, the decoder attains the ensemble-average error exponent of informed maximum-likelihood decoding for fixed channels and channels with growing memory. We approximate the frequency response by a constant matrix within each frequency block. For this auxiliary channel, every matched log-likelihood is affine in a common sufficient statistic. At fixed received output, we compute the statistic density of an independent competitor. The decoder selects the candidate with the smallest statistic density, equivalently minimizing a weighted sum of residual log-determinants. The density and its support volume give an upper bound on the universal competitor probability. Affinity ensures that at least half of a suitable symmetric neighborhood of the transmitted statistic has matched likelihood no smaller than that of the transmitted word. Integrating a density lower bound over this half gives the matched competitor bound. These bounds establish equality of error exponents, which a Gaussian channel comparison transfers to the original channel. The same decoder attains the matched exponent for independent frequency blocks with prescribed energies satisfying the stated conditions. For energies proportional to block length, an additional gain condition ensures that this exponent is at least that of the total-energy spherical ensemble.

0.23

Anchored Iterative Decoding of Reed-Solomon Product Codes for High-Rate Magnetic Tape Storage

cs.IT ·2026-11-06 ·Alexander Chvanov, Nikolay Malkovsky
Abstract

We present a decoding algorithm for a Reed--Solomon product code targeted at the demanding reliability requirements of modern magnetic tape storage. For a code $RS(256, 224)\times RS(175, 173)$ in a BSC we achieve post-decoding bit error rate (BER) of $10^{-20}$ at $\approx 5.3\times 10^{-3}$ raw BER while maintaining the common requirement to detect and correct 4 out of 32 dead tracks. We resolve two common problems of miscorrections and stall patterns that arise under such a high raw BER via simplified versions of anchor decoding and binary image analysis. The highest efficiency is achieved by performing an unlimited iterative decoder, which performs decoding iterations until no changes are made. We give an empirical analysis of the average number of iterations: below the raw BER of $5\times 10^{-3}$, four half-iterations are performed on average while improving substantially over a decoder that always performs 4 half-iterations. We provide a CPU implementation achieving 100-200 megabytes/s single-core throughput on a regular laptop and a design of a potential hybrid FPGA-CPU implementation of our decoder.

0.23

On Positive Computation

cs.CC cs.LO ·2026-11-06 ·Eduardo Skapinakis
Abstract

We introduce new notions of reducibility for monotone classes and show that a monotone version of Circuit Satisfiability and the Hamiltonian Cycle Problem are complete for $mNP$, even under positive reductions (which are not the same as monotone reductions). If this holds for all (monotone) $mNP$-complete problems - that they are also so complete for $mNP$ -, then $P\neq NP$. We introduce positive models of relativized computation and show that the separation $posP\subsetneq mNP$ does not relativize, even when oracles are restricted to monotone sets. Yet, there is a (weak) diagonalization proof that $posP\subsetneq mNP$.

0.23

Sensing-Based Communication: A Non-Orthogonal Passive Reflection Framework for ISAC

cs.IT eess.SPeess.SY ·2026-11-06 ·Xin Xiong, Mathew Salas, Sihua Shao, Chen Chen, Mostafa Hassanalian +1 more
Abstract

While integrating uplink without disrupting the primary downlink remains a critical challenge for existing reflective communication systems, this paper proposes a non-orthogonal passive integrated sensing and communication (ISAC) mechanism termed Sensing-based Communication. Originating from and validated within an optical retro-reflective system, this mechanism possesses an underlying wave-agnostic architecture with the potential to be generalized across other physical modalities. It leverages incident downlink signals for parasitic modulation via passive tags. By relying solely on fundamental on-off keying (OOK) reflection, the tag unifies the reflective sensing channel and the uplink communication channel, and simultaneously achieves both sensing and duplex communication. Consequently, our approach introduces uplink capability without sacrificing the downlink data rate or consuming additional spectrum resources. At the receiver, we design a joint decoding architecture combining cross-correlation computation and successive interference cancelation (SIC) to fully exploit the degrees of freedom in broadband signals for multi-user interference suppression. We derive closed-form expressions for the interference and bit error rate (BER) assuming zero-mean Gaussian signals. Using an experimental testbed, we validate the pivotal theoretical finding, the scaling relationship of SINR with signal bandwidth and cross-correlation time--where linear regression yields a strong goodness-of-fit with $R^2 = 0.9688$. Furthermore, evaluations under noise-like orthogonal frequency division multiplexing (OFDM) waveforms exhibit high consistency between theoretical derivations and simulation data ($R^2 = 0.9372$, mean absolute percentage error, $\text{MAPE} = 14\%$), confirming the model's accuracy and waveform generalizability.

0.23

Multiscale Closing Index (MCI) for measuring structural similarity between 2D point sets

cs.CG math.MGcs.CV ·2026-11-06 ·Nicolò Zava, Angel Franco-Alvarez, Samy Blusseau, Santiago Velasco-Forero, Maria-Jose Jimenez
Abstract

We introduce Multi-scale Closing Index (MCI), a new dissimilarity measure for evaluating the alignment of two 2D point sets. MCI is bounded between 0 and 1 and is based on multiscale mathematical morphology operations, with smaller values indicating greater structural agreement. The measure was motivated by the need to assess the quality of registration between point sets representing cell populations in biological images. To evaluate its behaviour under controlled conditions, we generated synthetic point clouds from Perlin-noise-based spatial structures, following established approaches for simulating biologically plausible spatial organization. We assessed MCI between those point clouds and different transformed versions, obtained by local perturbation, point removal, translation, rotation and addition of impulsive noise. Across these experiments, the dissimilarity MCI increased gradually and consistently with perturbation severity, while showing a clear discontinuity between clouds originating from the same underlying spatial structure and point clouds generated from different structures. We further compared MCI with commonly used point-set distances, including Hausdorff-based measures and Sliced Wasserstein distance. These experiments show the ability of MCI to capture structural agreement and quantify deviations between point sets, highlighting its potential as a complementary metric for evaluating two-dimensional point-set registration.

0.22

The satisfiability conjecture for random k-SAT formulas. Human-friendly exposition of the recent OpenAI resolution

cs.DM math.PRcond-mat.stat-mech ·2026-11-06 ·David Gamarnik
Abstract

The satisfiability conjecture was one of the most important open problems in discrete probability. It was resolved recently first by Carenini~\cite{GCarenini}, and shortly after by OpenAI~\cite{k-sat-OpenAI}, among other 370 problems solved by its internal model. The conjecture concerned the satisfiability property of random instances of the k-SAT problem. It posits the existence of a fixed value for the clause-to-variables ratio around which the probability that the formula is satisfiable drops sharply from nearly one to nearly zero as the ratio increases. While the case of large constant $k$ was resolved a decade ago by Ding, Sly and Sun~\cite{ding2015proof,ding2022proof} using the heavy machinery of the second moment method, the case of general $k$ eluded proof until ~\cite{GCarenini} and~\cite{k-sat-OpenAI}. In this short note we provide a ''human-friendly'' and detailed exposition of the OpenAI's resolution of the conjecture.

0.21

Symmetry Breaking over Multiway Channels

cs.IT ·2026-11-06 ·Raymond Zhang, Richard Combes, Sheng Yang
Abstract · Submitted to TIT

Identical users running the same algorithm must often break their symmetry and agree on distinct roles, as in leader election, naming, and resource allocation. We study anonymous symmetry breaking over a discrete memoryless multiway channel invariant under permutations of its $J$ users: each user observes only a private output and must choose a number in $\{1,\dots,J\}$, and the symmetry is broken if the $J$ numbers are distinct. Using a Rényi change of measure, we derive a finite-blocklength converse on the error probability of every symmetric causal protocol, including adaptive and randomized ones. For achievability, block coding with random names reduces the problem to all-source multicast, for which we derive a Gallager random-coding exponent for every fixed $J$, and a cut-set bound that limits the exponent of the probability that two names coincide or are decoded wrongly. Interactive role controllers instead let the users contend for identifiers directly: we give their exact error probability, an achievable exponent given by a spectral radius when some outputs certify a role, and a floor for finite-state controllers when every output has positive probability, and we introduce generalized musical chairs. On a noisy bandit channel motivated by multiplayer multi-armed bandits, with noiseless collisions and arms of a common mean below one, interactive schemes provably exceed the limit of name decoding, and numerical estimates suggest the same with noisy collisions. For two users, noiseless collisions, one informative arm and a silent arm, the optimal error probability is determined up to a factor of two.

0.21

Strategic Competition and Convergence in LLM Leaderboards

cs.GT cs.CYcs.SE ·2026-11-06 ·Naveen Raman, Fei Fang, Yiling Chen
benchmark
Abstract

Leaderboards are a popular way to compare large language model (LLM) capabilities and can influence how model developers invest resources. We ask when competition on leaderboards leads developers to produce models with similar capability profiles, which we call monoculture, or similar performance levels. We study this by formulating leaderboard competition as a game in which model developers invest resources across capabilities to maximize their win rate. We show that similar production functions can induce similar capability profiles, while comparative advantage can sustain distinct capability profiles. We then study convergence in performance levels and show that diminishing returns can lead to an increasingly crowded competitive frontier. Using data from LM Arena and OpenRouter, we find that developers improve more in categories where they lag competitors, while still retaining category-specific advantages. At the same time, we find that the frontier of models becomes increasingly crowded, alongside increasingly diversified model usage patterns. Our results suggest that increasing competition does not need to produce interchangeable models, but can instead sustain competitive models with distinct strengths.

0.21

Vertex Deletion on $H$-Induced-Minor-Free Graphs: Subexponential FPT Algorithms and Approximation Schemes

cs.DS cs.DM ·2026-11-06 ·Daniel Lokshtanov, Fahad Panolan, Saket Saurabh, Jie Xue, Meirav Zehavi
Abstract

We study vertex-deletion problems on string graphs and, more generally, on graphs excluding a fixed graph as an induced minor. For every fixed graph $H$, fixed nonnegative integer $t$, and fixed monadic second-order formula $φ$, we consider the problem of finding a minimum-cardinality set $S$ such that $G-S$ has treewidth at most $t$ and satisfies $φ$. This captures classic problems such as Vertex Cover, Feedback Vertex Set, and deletion to graphs of any fixed bounded treewidth. We give an algorithm with running time $2^{k^{2/3} \log^{O(1)} k}n^{O(1)}$ on $H$-induced-minor-free graphs, where $k$ is the optimum deletion-set size. This resolves in the affirmative the open problem highlighted by Bonnet and Rzążewski [Algorithmica 2019] and Berthe, Bougeret, Gonçalves, and Raymond [ICALP 2025] regarding the existence of a subexponential parameterized algorithm for Feedback Vertex Set on string graphs. Our algorithm significantly improves over existing parameterized algorithms for more restricted classes of problems on more restricted classes of graphs. When the property defined by $φ$ is hereditary, we also obtain a quasi-polynomial time approximation scheme, namely an algorithm that produces a $(1+\varepsilon)$-approximate solution in quasi-polynomial time. Both the sub-exponential algorithm and the quasi-polynomial time approximation scheme build on the same new structural and algorithmic insights.

0.20

Pairwise Reconnection for Cost-Distance Steiner Trees

cs.DS ·2026-11-06 ·Edgar Perner
theory
Abstract

The Uniform Cost-Distance Steiner Tree Problem asks for a rooted tree that minimizes the connection cost plus a weighted sum of root-to-terminal distances. We develop an algorithm that constructs two candidate trees: one based on direct connections to the root and another that pairs selected vertices and connects each pair to the root. The algorithm selects the tree with the smaller objective value. The resulting randomized algorithm achieves expected approximation factors of $1.698$ in the spanning case and $2.041$ in the general case, and it can be derandomized in polynomial time. The pairwise matching analysis charges part of the augmentation cost directly to the connection and delay costs of an optimal solution, rather than only to the separated lower bound used in previous analyses. This allows the $1.698$ factor to fall below the maximum multiplicative gap of $1+1/\sqrt{2}$ between the optimum value and that lower bound. We give instance families with approximation-ratio lower bounds approaching $1+1/\sqrt{2}$ for cut-and-reconnect algorithms that start from a minimum spanning tree and $3/2$ for the present algorithm. Finally, we show that the Uniform Cost-Distance Steiner Tree Problem and the Steiner-tree version of the Cable-Trench Problem are approximation equivalent up to an arbitrarily small additive loss.

0.20

Online Timely Bin Packing Revisited: A Simpler Algorithm with Optimality in Special Cases

cs.DS ·2026-11-06 ·Pei-Lin Chiang, Wing-Kai Hon, Ying-Chia Huang, Li-Tzu Yang
theory
Abstract · 17 pages, 1 figure

Online Timely Bin Packing is a variant of classical online bin packing in which items arrive over time with sizes and deadlines. Packing an item may be deferred until its deadline, but once a collection of items is packed into a bin, the bin is closed and cannot be used again. The objective is to minimize the total number of bins used. The problem was introduced by Khattar in 2023. In 2025, Aminian proposed Discrete-Greedy, a deterministic online algorithm with a competitive ratio of exactly $2$, which remains the best-known upper bound. Its analysis is particularly simple; however, the algorithm requires maintaining a fractional simulation internally and may leave usable space in some bins unfilled. We propose Thrifty-Greedy, a simple deterministic online algorithm that does not require a fractional simulation. We prove that its competitive ratio is exactly $2$, matching the best-known upper bound. At the same time, the algorithm automatically achieves the optimal competitive ratio on two natural classes of inputs: it is $1$-competitive when all items have the same size, and $1.5$-competitive when every item has size in $\{1/3,2/3\}$. In contrast, Discrete-Greedy has a competitive ratio of at least $2$ on both classes.

0.19

Regev's Reduction Reaches Reconstruction

cs.DS quant-phcond-mat.dis-nn ·2026-11-06 ·Stephen P. Jordan, Quynh T. Nguyen, Leo Zhou
Abstract · 30 pages, 4 figures

Regev's reduction uses a quantum Fourier transform to carry out reductions between various problems related to lattices and codes. We apply it to the problem of sampling from Gibbs distributions of certain classical spin glass Hamiltonians, which can equivalently be thought of as cost functions for average-case instances of an optimization problem called $D$-regular max-$k$-LINSAT. This problem becomes hard at low temperature. Here, we use a quantum decoding algorithm called Belief Propagation with Quantum Messages (BPQM) to solve the decoding problem arising from Regev's reduction. Our main contribution is a proof that the minimum temperature reachable using Regev+BPQM is exactly equal to the reconstruction threshold temperature for noisy channels transmitting signals on a tree network. This equality can be understood via Renes' duality between classical and quantum channels, which we use to show that quantum decoding is possible exactly when classical reconstruction on the dual problem is impossible. M{é}zard and Montanari identified this reconstruction temperature with a dynamical glass transition predicted by the one-step replica-symmetry breaking (1RSB) cavity equation. It is conjectured that local samplers such as Metropolis Monte Carlo and Glauber dynamics cannot equilibrate efficiently below this reconstruction temperature but may do so above it; our large numerical experiments show close agreement with this prediction across hundreds of different problem parameters. Therefore, our results rigorously show that Regev+BPQM reaches a temperature for which, to our knowledge, there is no matching classical guarantee, and brings Regev+BPQM to the conjectured limit of efficient Gibbs sampling by classical local dynamics.

0.18

Hardness of Unique Shortest Vectors and Bounded Distance Decoding

cs.CC ·2026-11-06 ·Sergei Kudria, Jason Luo
theory
Abstract · v2

We prove randomized NP-hardness of the unique shortest vector problem (uSVP) at every constant uniqueness factor greater than one, and of bounded distance decoding (BDD) at every fixed positive fraction less than one half of the minimum distance, in both their threshold decision and search versions. Both results hold in every fixed $\ell_p$ norm with $1\le p<\infty$. Previous Euclidean NP-hardness results for the search versions reached uniqueness factors tending to one and decoding radii greater than the minimum distance divided by $\sqrt2$. For linear codes over every fixed finite field and each fixed ${0<ρ\le1/2}$, the decision version of BDD with radius strictly less than $ρ$ times the minimum distance is equivalent under deterministic polynomial-time reductions to satisfiability for circuits promised to have at most one satisfying input. It is therefore NP-hard under randomized polynomial-time reductions. Earlier hardness for decoding general codes applied beyond half the minimum distance. For every fixed $p>2$ and for $p=\infty$, there is a constant $ε>0$ such that, under randomized polynomial-time reductions, uSVP in the $\ell_p$ norm is NP-hard at uniqueness factor $n^ε$, and BDD in the $\ell_p$ norm is NP-hard at radius $λ_1/n^ε$, where $n$ is the lattice rank and $λ_1$ is its minimum distance.

0.18

Robust Analysis of First-Price Auctions with Large Supply

cs.GT ·2026-11-06 ·Jason Hartline, Anant Shah, Matthew vonAllmen
theory
Abstract

This paper performs a robust analysis of welfare (i.e., the price of anarchy) in multi-unit first-price auctions with independent private values and unit-demand bidders. We employ a smoothness framework (see \citealp{ST-13}) to analyze the price of anarchy when supply is large. The main barrier to analysis in this framework is a worst-case bid distribution that establishes a lower bound on the price of anarchy of $1- \frac{1}{e}$. We show that such worst-case distributions cannot arise in equilibrium when agents' values satisfy a mild density lower bound condition. This characterization leads to novel asymptotic lower bounds. In particular, the price of anarchy for $ε$-approximate equilibrium as the number of units goes to infinity is at least $1-Θ\left(ε\log \left(\frac{1}ε\right)\right)$. Our analysis relies on a key structural property of equilibrium bid distributions: for a welfare analysis in any multi-unit first-price auction, it is without loss to analyze (approximate) equilibria with no atoms. In addition to this, we establish continuity and monotonicity properties of Bayes-Nash equilibria. We believe these results are of independent interest.

0.17

Metric Dimension Parameterized by Bag-Connected Treewidth and Bag-Connected Pathwidth

cs.CC ·2026-11-06 ·Dibyayan Chakraborty, Sebastian Ordyniak, Sebastian Siebertz
theory
Abstract

The metric dimension of a graph is the minimum size of a set of vertices whose distance vectors distinguish all vertices of the graph. We prove that \MetricDimension{} is NP-complete on planar graphs of treewidth and bag-connected treewidth exactly three. The graphs in our construction have no induced cycle longer than four and have (prefix-connected) pathwidth at most four. In particular, the problem admits no XP algorithm parameterized by bag-connected treewidth unless $\mathrm P=\mathrm{NP}$. This answers questions of Bonnet and Purohit and of Bousquet, Deschamps, and Parreau. We prove that the problem is fixed-parameter tractable when parameterized by bag-connected pathwidth.

0.17

Pangenome Peeling: Decomposing Weighted Adjacencies into Path-Forest Genomes

cs.DM q-bio.GN ·2026-11-06 ·Alitzel López Sánchez, Jiazhen Leng, Natasha Ivanova, Lingling Jin, David Sankoff
theory
Abstract · This is an extended version of the article of the same name submitted to the RECOMB 2027 conference

One aspect of pangenome analysis that has received little attention is the recovery of the individual genomes that gave rise to a pangenome graph. While substantial effort has focused on constructing pangenomes from collections of genomes, the inverse problem remains largely unexplored: given a pangenome represented as a set of weighted adjacencies, can we reconstruct the constituent linear genomes that underlie it? Thus we introduce the pangenome peeling problem, which seeks to decompose a weighted adjacency graph into a collection of linear genomes whose combined adjacencies reproduce the observed edge weights. This problem is particularly relevant when considering ancestral pangenomes, where the graph structure may be known or inferred, but the underlying genomes are not. As a first step toward addressing this problem, we formally define pangenome peeling and derive a lower bound on the minimum number of genomes required to explain a given weighted adjacency graph. We then present two reconstruction methods: one based on the sequential addition of edges and a second based on extracting maximum-spanning-tree structures. Both methods produce genome collections that exactly explain the input adjacency data. To evaluate the recovered sets of genomes, we compare them to the set of input genomes using a maximum-weight matching framework based on the double-cut-and-join distance with indels. Here the maximum-spanning tree approach is superior, producing sets of constituent genomes close to the simulations while the sequential addition of edges approach yields constituent genomes increasingly divergent from the input as genomes sizes increase.

0.16

Asymptotic Analysis of Data Deduplication Under a Constant Edit Model

cs.IT ·2026-11-06 ·Sarvin Motamen, Hao Lou, Farzad Farnoud
Abstract

Data deduplication has become increasingly important in large-scale storage systems, driven by the rapid growth of digital data. By identifying and removing repeats in the data stream, deduplication saves storage space more efficiently than traditional compression methods. In this paper, we study the performance of variable-length deduplication and multi-chunk deduplication algorithms from an information-theoretic perspective, focusing on data streams where repeats are approximate rather than exact. We consider two probabilistic source models in which source strings consist of repeated blocks that are modified by either a constant number of substitution edits or burst-insertion edits per block, the latter capturing the localized edit patterns commonly observed in real-world data. For both models, we show that variable-length deduplication achieves asymptotically arbitrarily large compression ratios, and that multi-chunk deduplication achieves compression within a constant factor of the source entropy under mild conditions.

0.16

PaB-tree: Parallel B+-tree with Batch Updates

cs.DS cs.DBcs.DC ·2026-11-06 ·Ziyang Men, Yan Gu, Yihan Sun, Helen Xu
Abstract

The B+-tree is a ubiquitous search index due to its provable I/O (cache) efficiency. Yet unlike batch-parallel binary search trees, which are both fast and widely available, no batch-parallel B+-tree offers strong empirical performance together with provable, near-optimal I/O efficiency. Prior B+-tree-based designs are either parallel but purely theoretical, with no high-performance implementation, or concurrent indexes such as the Bw-tree and Masstree, which run an order of magnitude slower on batch updates. We present PaB-tree, the first batch-parallel B+-tree with guarantees on work, span, and cache complexity alongside strong practical performance. Its core challenge -- repairing node overflows and underflows without eroding the tree's I/O advantage -- is resolved by one simple invariant: a repair never touches a sibling and is always pushed to the parent. A Promote primitive reduces batch construction and insertion to repeated node promotion, and batch deletion to reinsertion. Batch updates take O(B log_B n) work and O(log_B n) I/O per element, matching the sequential B+-tree bounds, with polylogarithmic span. Against highly-optimized parallel and concurrent baselines, PaB-tree delivers the best overall performance trade-off: up to 1.4x faster on construction and up to 3.6x faster on batch insertion than the strongest baseline, while preserving B+-tree query efficiency.

0.15

What remains of the planar induced minor conjectures?

cs.DM cs.CCcs.DSmath.CO ·2026-11-06 ·Jadwiga Czyżewska, Tomáš Masařík, Marcin Pilipczuk, Paweł Rzążewski
Abstract · 26 pages, 7 figures

Dallard, Milanič, and Štorgel, and independently Gartland and Lokshtanov, conjectured that for every planar graph $H$, Maximum Weight Independent Set and List $3$-Coloring are polynomial-time solvable in $H$-induced-minor-free graphs. As a structural explanation, Gartland and Lokshtanov further conjectured that such graphs admit balanced separators dominated by a constant number of vertices. Bonnet and Chang recently refuted the independent-set part of the algorithmic conjecture, by showing that Maximum Independent Set is $\mathsf{NP}$-hard in graphs excluding the $5\times5$ grid as an induced minor. We revisit both conjectures. We show that the graphs used by Bonnet and Chang exclude much simpler planar graphs, including several $16$-vertex trees, and we introduce a second, independent family of hard instances. We also prove that $3$-Coloring is $\mathsf{NP}$-hard in both settings, thus refuting the algorithmic conjecture in full, even in its quasipolynomial version, unless $\mathsf{NP}\subseteq\mathsf{QP}$. On the structural side, we identify planar graphs $H$ for which the structural conjecture holds, while the algorithmic one fails. Thus, constantly dominated balanced separators cannot by themselves explain the tractability of Maximum Weight Independent Set and List $3$-Coloring.

0.15

The Schur products of some nonbinary narrow-sense antiprimitive BCH codes

cs.IT ·2026-11-06 ·Haojie Xu, Xia Wu, Wei Lu, Xiwang Cao
Abstract

In recent years, the Schur products of linear codes have been extensively studied due to their favorable applications in cryptography. In this paper, we aim to study the parameters of the Schur products of some nonbinary narrow-sense antiprimitive BCH codes. Let $\mathcal{C}_1$ and $\mathcal{C}_2$ be two narrow-sense antiprimitive BCH codes with designed distances $δ_{1}$ and $δ_{2}$, respectively, where $2\le δ_{1},δ_{2}\le n$. Let $\mathcal{C}_1\star\mathcal{C}_2$ be the Schur product of $\mathcal{C}_1$ and $\mathcal{C}_2$. We present some necessary and sufficient conditions for $\mathcal{C}_1\star\mathcal{C}_2\ne\mathbb{F}_q^n$ by applying some constraints to $δ_{1}$ and $δ_{2}$. Furthermore, we determine the dimensions of these products and lower bounds on their minimum distance. Several linear codes with best known parameters are found.

0.14

Containment Problems in Convex Sets

cs.CG cs.DMmath.CO ·2026-11-06 ·Prosenjit Bose, Bobby Miraftab, Tyler Tuttle
theory
Abstract

Let $P$ be a set of $n$ points in the plane, and let $C$ be a compact, convex subset of the plane that contains the origin in its interior. We study the problem of finding two points $p$ and $q$ of $P$ with the property that any homothet of $C$ that contains $p$ and $q$ must contain at least a constant fraction of the points of $P$. We give proofs for several cases. When $C$ is strictly convex we achieve a fraction of $1/5$, and when $C$ is a general convex region we achieve a fraction of $1/6e$. We show how the ideas of this second bound can be used to prove a lower bound for a version of the problem in $\mathbb{R}^d$ when $C$ is a $d$-ball. We also consider a few special cases. When $C$ is a triangle we get a lower bound of $1/12$, and when we allow that triangle to be rotated by $π$ radians in addition to being scaled and translated we get a lower bound of $1/768$. In proving this last result, we improve the state of the art on the version of this problem in three dimensions, where $C$ is a parallelepiped. We also consider the algorithmic aspects of finding such a pair of points.

0.13

Optimal Regret for Profit Maximization in Bilateral Trade

cs.GT cs.LG ·2026-11-06 ·Matteo Castiglioni, Anna Lunghi, Alberto Marchesi
theory
Abstract

Learning to profit from trade requires learning both sides of a market: what a seller is willing to accept and what a buyer is willing to pay. We study repeated bilateral trade, where an intermediary posts a purchase price to a seller and a sale price to a buyer whose valuations are drawn from an unknown joint distribution. Trade occurs only if both parties accept their respective prices, in which case the intermediary earns the difference between the two prices. The intermediary observes only whether trade occurs, without observing either valuation. We show that learning both sides of the market is no harder, in terms of regret, than learning just one. Specifically, we design an algorithm achieving regret $\widetilde{\mathcal{O}}(T^{2/3})$ against the best fixed pair of prices, matching, up to logarithmic factors, the classical lower bound for one-dimensional dynamic pricing. Our result shows that the classical one-dimensional regret rate survives the transition to two dimensions, raising the question of whether this absence of a curse of dimensionality extends to arbitrarily high dimensions.

0.13

Capacity of Generic Symmetric Distributed Linear Computation: The Geometry of Relations

cs.IT ·2026-11-06 ·Hua Sun
Abstract

In distributed linear computation, $K$ users observe linear functions of common finite-field data and send separate messages to a server that wants another linear function. We characterize the generic capacity $C$, the minimum total number of field symbols sent per computed instance, when all observations and the demand have dimension $ν$, the data have dimension $μ$, and the coefficients of these linear functions are drawn independently and uniformly from the field and stay fixed during coding, with the field size growing through arbitrary prime powers. For fixed $K$, $C/ν$ depends only on $τ=μ/ν$: the curve has finitely many linear pieces for at most three users, infinitely many pieces approaching one value of $τ$ for four users, and a rational part between two sequences of linear pieces for five or more users. The geometry of the relation space, formed by the linear relations among the observations and the demand, determines three things: the shape of the curve, which relations the coding scheme aligns, and the entropy inequalities used in the converse. Generically, the user subspaces of this space are either balanced, where the capacity lies on a single rational curve, or split into two balanced components, a sparse and a dense one, which produces the linear pieces. Achievability uses asymptotic set alignment, in which monomial precoders confine the combined observation differences to a small set. The converse uses the same geometry to lower-bound the sum of the users' remaining observation entropies in terms of the remaining data entropy.

0.12

Coresets for Capacitated Clustering via Dual Concentration

cs.DS cs.CGcs.LG ·2026-11-06 ·Milind Prabhu, Chris Schwiegelshohn, Sudarshan Shyam, Erik Waingarten
Abstract

We study coresets for clustering problems, focusing on the classical $k$-median and $k$-means objectives. A coreset is a small weighted summary of the input data that approximately preserves the clustering cost for every choice of centers. Such summaries are a central tool for scaling clustering algorithms to large datasets. In this work, we consider the capacitated setting, where the clustering is subject not only to choosing centers but also to respecting prescribed cluster-size constraints. This model is a natural generalization of standard clustering and has been widely studied, but its coreset complexity is less well understood. For finite metric spaces, we obtain optimal-size coresets; surprisingly, these match the best-known bounds for unconstrained clustering. We also give improved coreset bounds in Euclidean space, notably the first coresets with near-linear dependence in $k$ while being independent of the dimension and input size. The main new technical ingredients are concentration bounds for the dual solutions of min-cost flow problems, which allow for both significantly simpler and tighter analyses of the underlying sampling distribution used in the coreset construction.

0.12

Contextual Search and Bilateral Trade with a Trace-Square-Root Potential

cs.GT cs.LG ·2026-11-06 ·Romain Cosson, Federico Fusco, Anupam Gupta, Stefano Leonardi, Renato Paes Leme +1 more
Abstract

We present a unified potential method for contextual search and bilateral trade with one-bit feedback and obtain order-tight algorithms for both problems. Our framework starts from a Bayesian perspective in which the unknown target (for contextual search) and the unknown pair of valuations (for bilateral trade) are sampled from a known prior distribution. In that model, the decision maker bases its prediction on the posterior distribution, obtained by conditioning on feedback received in previous rounds. For Bayesian contextual search, we present an algorithm with a total loss bounded by $O(\mathrm{Tr} \sqrt Σ)$, where $Σ$ is the covariance matrix of the target's prior; and for contextual bilateral trade, our algorithm's regret is bounded by $O(\mathrm{Tr} \sqrt{Σ_S}+\mathrm{Tr} \sqrt{Σ_B})$, where $Σ_S$ and $Σ_B$ are the covariance matrices of the seller's and buyer's valuations. We then use ideas from this Bayesian setting and adapt a ``dual'' of this trace-square-root potential to the classical worst-case setting, where the target and valuations are chosen adversarially from the unit Euclidean ball. For both contextual search and bilateral trade, we obtain tight $O(\sqrt d)$-regret algorithms. These results match the existing lower bounds up to a constant factor and improve on the previous guarantees of $O(d\log d)$ for contextual search and of $O(d6^d)$ for bilateral trade.

0.11

Tensor value and the 9/4 bound for matrix multiplication

cs.CC math.RT ·2026-11-06 ·Mahmut Levent Doğan, Vladimir Lysikov, Keiya Sakabe, Michael Walter
theory
Abstract

We give an exposition of OpenAI's recent result that the exponent of matrix multiplication is at most 9/4 valid over every field. Our main tool is a fractional symmetrized variant of the tensor value introduced by Coppersmith and Winograd, used in place of spectral points (tensor characters) in the original proof. We prove a superadditivity inequality for a power of this value on partially direct sums, whose summands share one tensor factor and occupy independent subspaces in the other two. Together with the degenerations of polynomial multiplication tensors from the OpenAI proof, this inequality yields the bound.

0.08

Proportionality in Elections under Feasibility Constraints and Ranked Ballots: The Solid Coalition Refinement Rule

cs.GT ·2026-11-06 ·Grzegorz Kwacz, Georgios Papasotiropoulos, Piotr Skowron
Abstract

We study elections with ranked ballots and general feasibility constraints, which indicate which sets of candidates are allowed as outcomes. We introduce the first proportionality axiom for this general setting and show that it extends the established notions from specific models that fit our general framework. We extend the Solid Coalition Refinement (SCR) rule to general constraints and prove that it satisfies our axiom exactly for matroid constraints. Although the naive implementation of SCR involves a step that is computationally intractable for weak orders, we show that in practical cases it can be implemented in polynomial time or efficiently in FPT time. We also illustrate limitations of this rule for approval ballots and provide its first empirical evaluation.

0.08

Robotics, Control & Signals 104

Characterization of the Frequency-Current Curve of First-Order High-Pass Neurons

eess.SY ·2026-11-06 ·Christian Fernandez Lorden, Jérôme Adriaens, Alessio Franci, Gustave Bainier
Abstract · 8 pages, 4 figures

A spiking model, called the High-Pass (HP) model, has recently been proposed as a minimal excitable primitive. This model is composed of a static activation function, usually a sigmoid, in positive feedback with a high-pass filter. The HP model enters a tonic spiking regime whenever its input lies in the high-gain zone of the sigmoid, where the sigmoid's slope exceeds 1. In neuroscience, the frequency-current (FI) curve is the standard way of characterizing the tonic spiking regime of a neuron. So far, the FI curve of the HP model has been characterized analytically only for specific sigmoids, and has otherwise been computed through simulation. These simulations are computationally heavy, since simulating neurons requires solving a stiff system of differential equations. This paper bridges this gap by analytically characterizing the FI curve of the HP model with a first-order high-pass filter, for a broad class of sigmoid functions. The FI curve of the first-order HP model can be computed by solving two input-independent equations and evaluating two input-dependent integrals, with no need to simulate the model itself. A spiking primitive is said to be Type I (resp. Type II) if tonic spiking starts at a zero (resp. non-zero) frequency. This paper also shows that, for the HP model, the type depends solely on the local shape of the sigmoid at the outer edges of its high-gain zone. A sharp corner, where the derivative approaches a value strictly smaller than 1 from the outside, yields a Type I primitive. A locally twice differentiable corner, with a bounded second derivative and a derivative approaching 1 at the edge, yields a Type II primitive.

0.53

Extended Model-Based Learned Inertial Odometry

cs.RO ·2026-11-06 ·Sasanka Kuruppu Arachchige, Joni-Kristian Kämäräinen
publishedcode/data
Abstract · 8 pages, Published at ICUAS 2026

Inertial odometry is a compelling approach to state estimation for agile quadrotor flight due to its affordability, low weight, and robustness to perceptual degradation. However, relying only on integrated inertial measurements is impractical, as sensor errors and time-varying biases lead to significant pose drift. Although recent advances have enabled inertial odometry for drone racing, existing approaches show limited generalization to trajectories not seen during training. This work improves generalization by adopting a body-frame representation and incorporating body-frame torque dynamics into the learning pipeline. The learned model predicts short-horizon relative displacements, which are fused with IMU measurements in an Extended Kalman Filter. Experimental results show that the proposed method outperforms the previous state-of-the-art learned inertial odometry approach for quadrotor pose estimation on unseen trajectories. An additional evaluation on four public quadrotor datasets, with whole trajectory shapes held out of training, confirms a lower velocity error on every held-out set and shows that the torque input is most informative when it is derived from the rotor signals. Code, data pipeline and trained models are publicly available.

0.48

Auditing Grasp Feasibility Predictors with Planner-Verified Counterfactuals

cs.RO ·2026-11-06 ·Sajid Ansari, Antony Thomas, Girish Varma
publishedbenchmarkcode/data
Abstract · Accepted to WACV 2027. 10 pages main paper + 8 pages supplementary

Grasp-feasibility predictors report near-perfect accuracy on Task and Motion Planning benchmarks, but high accuracy on these benchmarks does not verify that a predictor uses the collision geometry that the motion planner checks. We propose a planner-verified counterfactual (CF) test: holding the grasp pose fixed, we insert a solid obstacle into the approach corridor, add its points to the model input, and let a planner re-derive the label, which flips from feasible to infeasible. Across eight predictors spanning point clouds, voxels, depth, and object tokens, CF-pair AUROC falls from high natural AUROC (0.920-0.996) to chance (0.470-0.504), and the flip holds across four planners, two arms, and a real-object obstacle. Predictors fail at different resolutions, from a bounding box to a 12.5 cm voxel. In the point-cloud models we dissect, the obstacle either moves the representation in a direction unrelated to whether it blocks the grasp or is not registered at all. A clearance-aware read of the corridor points, trained on the unmodified benchmark with no counterfactual scenes, reaches 0.983 CF-pair AUROC, but it remains bound to its trained obstacle placements; a placement-invariant clearance representation is still open. We release the counterfactual labels, metrics, datasheet, and relabeling pipeline at https://github.com/codewithsajid/graspfc-cf.

0.47

Lodos: A Lightweight Simulator Designed for Contact-Rich Interactions with Detailed Meshes

cs.RO ·2026-11-06 ·Onur Beker, Andreas René Geist, Georg Martius
code/data
Abstract

In recent years, massively parallelized rigid-body simulation has played a central role in robotics research, enabling rapid progress in areas such as reinforcement learning, stochastic optimal control, and randomized smoothing-based trajectory optimization. Particularly in contact-rich settings, the scalability of such simulations in large part depends on well-optimized representations of collision geometry, often requiring a coarse approximation of the underlying surfaces using primitives such as spheres, capsules, cylinders, and boxes. Whenever the complexity of the surfaces or the precision demands of the application render such approximations insufficient, it is instead necessary to employ mesh-mesh collisions, which impose a significant reduction in throughput. This paper proposes a simulation framework designed from scratch to address this bottleneck, with novel routines for broadphase and narrowphase collision detection as well as the subsequent contact dynamics computation. As experimentally validated, the framework matches existing well-established simulators in terms of physical faithfulness, and can achieve approximately an order of magnitude higher throughput in the detailed mesh-mesh collision setting that it specializes in. The source code is available at https://github.com/martius-lab/lodos.

0.45

Goodhart in the Planner: Sampling-Based Control Exploits Rewards Learned from Human Preferences

cs.RO cs.AIcs.LG ·2026-11-06 ·Soovadeep Bakshi, Apurba Bose
theory
Abstract · 21 pages, 4 figures. Equal contribution. Code and results will be released

Robots increasingly plan by searching against a reward model learned from human preferences; we measure what this search does to the true objective. On HalfCheetah and Hopper with crowd-sourced labels, and on Meta-World drawer-open and plate-slide with synthetic labels and blind labels from the two authors, harder MPPI search cuts true return by up to 76% while held-out label accuracy stays at 0.75-0.97: with human labels the drawer opens in 0 of 20 runs, against 10 of 10 for the cloned policy the planner starts from. The effect persists with a learned dynamics model and with an image scorer. True return roughly follows how far the planner's weights move from its proposals ($\sqrt{\mathrm{KL}}$), and best-of-$N$ action selection with a verifier trained on synthetic labels falls below a single noisy action. We prove that $χ^2$-regularized MPPI at a temperature set by the scorer's error spread on the proposals is worst-case optimal, that no gain can be certified once that spread exceeds the scorer's own, and that an episode's return splits exactly into per-step terms, only some of which the guarantee covers. Calibrated from 100 proposal pairs ranked by true return (an oracle), this rule is the safest fix we test: it keeps on average 86-94% of the cloned policy's return on drawer-open and plate-slide, though it is cautious on window-open, where search helps. Calibrated from a person's own comparisons, it cannot detect a bias the scorer shares with that person; a second person, or correcting for pairs people cannot tell apart, limits the damage.

0.44

Generalized H matrix stability for power systems

eess.SY ·2026-11-06 ·Elinor Ginzburg-Ganz, Yoash Levron, Juri Belikov, George Weiss
published
Abstract

The small-signal stability of a power system model with arbitrary loads is investigated, with a focus on locally constant power loads in lossless networks. A central element of the proposed analysis is the matrix H, which represents the partial derivatives of generator powers with respect to their relative angles. The local stability of any equilibrium point is determined by the properties of H if the loads consume locally constant power. Specifically, an equilibrium point is asymptotically stable if H is positive definite and unstable if H is full-rank and not positive definite, while several quantitative relationships between the eigenvalues of H and the system Jacobian matrix are also established. The main contribution is that loads are allowed to be arbitrarily located within the transmission network, thereby generalizing the existing H matrix stability criterion which is valid only if loads are located at the same buses as the generators. The advantage of the proposed criterion is demonstrated through case studies involving systems with up to 2869 buses. Accurate stability assessments are maintained, where alternative methods, such as the NRPS model, are found to be overly conservative. Additionally, a computational acceleration by a factor of 340 relative to Jacobian-based methods is achieved.

0.42

FoundaRF: A Physics-Informed Foundation Model Framework for Radio-Frequency Passive Components

eess.SP cs.AI ·2026-11-06 ·Houbo He, Chris Han, Hyunsu Chae, Lei Xia, Qiang Zhou +4 more
published
Abstract · Accepted for publication in the proceedings of the 2026 IEEE/ACM International Conference on Computer-Aided Design (ICCAD 2026)

Designing RF/mmWave passive components is a time-consuming process that relies heavily on expert knowledge and iterative EM simulations. Although recent ML surrogate models have shown great promise in accelerating passive design, they remain task-specific and lack generalization. This paper presents FoundaRF, a foundation model framework that unifies passive layout modeling across diverse layout structures and technologies. It combines a self-supervised encoder with adaptive layout processing and physics-aware decoders to enable few-shot transfer. Experiments across four semiconductor technologies show strong generalization to unseen structures, outperforming baselines with up to a 3.3x reduction in prediction error, reducing labeled-data requirements by up to fourfold, and substantially shortening design turnaround. FoundaRF establishes a scalable and data-efficient solution for RF/mmWave passive design automation.

0.42

HOC2R: Hand-Object Contact Trajectories for Human-to-Robot Motion Transfer

cs.RO ·2026-11-06 ·Hongkun Yu, Francisco M. López, Francisco Cruz
top venuebenchmark
Abstract · 10 pages, 10 figures, 9 tables. Accepted for oral presentation at the Australasian Conference on Robotics and Automation (ACRA 2026)

Human videos contain abundant object-manipulation demonstrations, but they do not directly provide motion cues that a robot can follow. We present HOC2R, a Hand-Object Contact trajectory extractor for Robot motion that combines frozen DINOv3 and SAM 2.1 features with learned heads for hand-object association and contact localization. HOC2R supports two modes: count-aware automation that detects, counts, and localizes visible contacts in untrimmed videos, and fixed-two localization for two-finger grippers when contact frames are known. On a 30-task point-annotated benchmark, count-aware automation improves contact-count accuracy from 58.1% to 71.4% and reduces mean localization distance from 19.81 to 10.46 pixels relative to an adapted VRB2 baseline. On seven unseen RGB-D rollouts, a five-task fixed-two checkpoint reduces mean localization distance from 67.67 to 27.02 pixels. Across six robot evaluation sequences, the 30-task fixed-two checkpoint produces four feasible trajectories and succeeds on three tasks, compared with two feasible trajectories and one successful task for each adapted baseline. HOC2R is competitive in the evaluated single-hand scenes, robust to the hand-color shift, and strongest in the evaluated multi-hand scenes. These results support HOC2R as a practical pseudo-labeling method for extracting robot-usable contact trajectories from human demonstrations.

0.41

An Optimal Power-Flow Grouping Topology for Single-Isolation-Stage Multi-Output Solid-State Transformers

eess.SY ·2026-11-06 ·Yi Hao, Krzysztof Puczko, Levy Ferreira Costa, Liangcai Shu, Dongsheng Yang
Abstract · 13 pages, 19 figures. Submitted to IEEE for possible publication

Electric-vehicle dc fast-charging stations and data centers require multiple high-power dc outputs with independently varying demands. Conventional multi-output solid-state transformers typically establish a common dc bus and use downstream dc/dc converters for port-level regulation and, where required, galvanic isolation. A single-isolation-stage solid-state transformer can eliminate this additional high-power interface stage by directly forming multiple isolated outputs from its submodule dc/dc stages. However, because one submodule is generally insufficient for an individual output port, multiple submodules must be connected in parallel, while inter-submodule balancing circuits redistribute power under unbalanced loading. The submodule grouping therefore determines the internal balancing-power paths. Sequential grouping can create long transfer paths and accumulated balancing power, increasing current stress and losses. This paper proposes an optimal power-flow grouping topology that interleaves submodules assigned to different output ports, thereby localizing power redistribution and reducing the power processed by the balancing circuits. Analytical models quantify the required balancing power and its reduction relative to sequential grouping. A coordinated submodule-bypass strategy is also developed, and the remaining output-power capability under different fault scenarios is evaluated. A six-submodule prototype validates the proposed topology. Under the same highly unbalanced operating conditions, the proposed topology reduces the maximum submodule dc-link voltage difference by 63\% and the peak balancing current by 14.5%, while improving the measured efficiency by approximately 5.6% at rated power. Stable post-fault operation is also experimentally demonstrated.

0.40

Robust Readout of Cross-Point Resistive Arrays via Time-Domain Optimization

cs.RO ·2026-11-06 ·Taro Tako, Hiroshi Kaminaga, Taisuke Kobayashi
published
Abstract · 4 pages, 1 figure, IEEE Sensors 2026 accepted

Cross-point resistive arrays are attractive for high-density tactile sensing because they reduce wiring complexity. However, conventional frequency-based readout methods can become inaccurate in robotic sensing environments, where long interconnects and measurement and environmental noise degrade the observed signals. We propose a time-domain optimization method for robust readout of cross-point resistive arrays. The method determines resistance values by fitting a time-domain observation model to the measured waveform over the interval, rather than relying solely on spectral amplitudes. By exploiting the full waveform, the proposed method improves robustness under noisy observation conditions. In simulations of a 16 x 16 array, the proposed method reduced the mean readout error by 20% compared with the baseline method. Furthermore, validation over 1,000 trials showed that the proposed method achieves more robust estimation than the conventional method, reducing the mean error and standard deviation by 24.6% and 77.6%, respectively. The improvement was also confirmed in cross-point array simulations with resistance distributions emulating tactile sensor responses.

0.39

UniLatent: Cross-Embodiment Transfer via Latent Observation Alignment

cs.RO cs.CV ·2026-11-06 ·Yufei Ding, Yi Du, Weiduo Yuan, Hao Jiang, Dylan Goetting +3 more
publishedcode/data
Abstract · Accepted to CoRL 2026

Robot datasets are costly to collect and are usually tied to a specific embodiment, making it difficult to reuse demonstrations across robots with different geometry, appearance, and kinematics. We introduce UniLatent, a cross-embodiment transfer framework that reduces the embodiment-induced observation gap in a shared latent space. UniLatent renders paired source-target robot observations in simulation and aligns a source visual encoder to a frozen target encoder. A policy trained on encoded source demonstrations can then run directly on target observations, with no pixel-level image translation. Across three simulation tasks and two target robot arms, UniLatent reaches up to 93% zero-shot success, exceeding the strongest pixel-translation baseline by over 30 points on average. In real-world xArm-to-Franka transfer, fine-tuning with only 10 target demonstrations per task yields 27/30 successful rollouts. Code, videos and data are available at https://selina2023.github.io/unilatent/.

0.39

Lodos-v2: An Efficient Primal Formulation of the Exact Frictional Contact Problem

cs.RO ·2026-11-06 ·Onur Beker, Andreas René Geist, Georg Martius
code/data
Abstract

Lodos-v2 is a GPU-based simulation framework for massively parallel rigid-body dynamics with contact. It is specifically designed to achieve high throughput in scenes involving collisions among large numbers of detailed meshes. Its predecessor, Lodos-v1, had two main limitations. First, it solved the standard convex relaxation of cone-complementarity-based contact dynamics, thereby inheriting well-known physical artifacts, including lift-off and artificial compliance in the normal direction, as well as creep in the tangential direction. Second, its contact-group reparameterization scheme, which rescales spring-damper forces using a single factor per collision pair, could not be integrated with other constraints required by a full-fledged simulator, such as joint limits and welds. Lodos-v2 addresses both limitations through two main contributions. First, it introduces an efficient algorithm for solving the exact frictional contact problem by sequentially solving a cascade of convex relaxations in primal form within an outer proximal minimization loop. Second, it extends the contact-group reparameterization scheme of Lodos-v1 to the primal formulation of cone-complementarity-based contact dynamics, enabling its integration with the other constraints of a full-fledged simulator. Experiments demonstrate that these contributions result in throughput improvements over established rigid-body simulation frameworks. The source code is available at https://github.com/martius-lab/lodos.

0.38

DC-Link Voltage Regulation in a Single-Isolation-Stage Multioutput Solid-State Transformer for Wide-Range DC Outputs

eess.SY ·2026-11-06 ·Yi Hao, Levy Ferreira Costa, Krzysztof Puczko, Liangcai Shu, Dongsheng Yang
unusual
Abstract · Submitted to IEEE Transactions on Power Electronics (TPEL) for possible publication

Single-isolation-stage solid-state transformers with balancing circuits eliminate an additional downstream dc--dc conversion stage. However, the remaining isolated converters must directly accommodate a wide output-voltage range. This paper investigates variable dc-link voltage operation of a single-isolation-stage multioutput SST to ease this requirement. Coordinated cascaded H-bridge modulation and embedded four-switch balancing circuits support intentionally unequal submodule dc-link voltages, allowing the isolated-converter input voltages to be adjusted according to the output requirements. The achievable voltage range is determined by relating the required balancing power to the BC transfer capability within the available CHB switching windows, subject to modulation and peak and rms current limits. The analysis quantifies the range extension provided by BC operation beyond modulation alone. For a single-phase-shift dual-active-bridge converter, input-voltage following within this range reduces the calculated peak current and extends the ideal zero-voltage-switching region compared with fixed input. Experiments on a six-submodule SST prototype demonstrate unequal dc-link voltage operation and inter-submodule power exchange. Experiments result on DAB demonstrate reduced current stress and improved efficiency with voltage following.

0.38

When to Trust the BIM: Identifiability-Aware LiDAR-Inertial--BIM Fusion under As-Built Deviations

cs.RO ·2026-11-06 ·Yujie Zhang, Yuxuan Guo, Xinbao Sun, Chunlei Zhao, Langqi Miao +2 more
Abstract · 24 pages, 12 figures, 11 tables (including appendices)

Building information models (BIM) can provide a global geometric reference for robot localization with LiDAR-inertial odometry (LIO). However, as-built deviations from the as-designed model mean that a disagreement between a scan and the BIM may arise from a pose error, an offset wall surface (an as-built deviation or an unmodeled object in front of the wall), or a shifted room, and in some geometries these explanations produce identical observations. In this paper, we formalize this attribution problem as fault identification over three fault classes and derive a geometric criterion for when attribution is possible. Building on it, we propose an identifiability-aware method that reports an attribution only when the fault is attributable; otherwise, it retains the candidates, defers BIM updates and BIM-based pose corrections in the ambiguous directions, and recovers once new references arrive. In the pre-registered two-dimensional study, on cases whose fault is detectable but not attributable, the method's conditional error rate was 0.83%, against 31.67% for a baseline with uncertainty-based abstention at the same answer rate. On the 5,600 attributable cases, the method reported a unique attribution 5,574 times, with no incorrect attributions. In paired fault injections in a laboratory room with its IFC model and on three floors of the SLABIM dataset, it raised no off-target attribution (0 of 168 conditions in the laboratory; 0 of 1,302 on SLABIM) and attributed all eight laboratory faults predicted to be attributable, whereas both reproduced baselines raised off-target structural flags in every setting.

0.37

Beyond Feasibility: Efficient and Robust On-Device Localization with UWB Concurrent Transmissions

eess.SP cs.NIeess.SY ·2026-11-06 ·Enrico Soprana, Michael Stocker, Davide Vecchia, Matteo Trobinger, Carlo Alberto Boano +1 more
Abstract · Submitted to the IEEE Transactions on Mobile Computing

Concurrent transmissions (CTX) enable an ultra-wideband (UWB) tag to range to, or self-localize from, multiple nodes with a single frame reception, significantly improving update rate and energy efficiency compared to traditional approaches. Prior work demonstrated the concept feasibility via prototypes whose real-world applicability is however hindered by many limitations, notably including their inability to directly execute on UWB tags and to cope with interfering UWB signals. We address these limitations in two ways. First, we redesign the core CTX techniques to enable a fast and efficient execution on resource-constrained MCUs without sacrificing accuracy. Second, we leverage the new features in the IEEE 802.15.4z standard to achieve robustness to interference from other UWB devices and enable tags to dynamically select which anchors to use, two crucial and unexplored capabilities in CTX-based systems. We quantitatively evaluate our approach experimentally in two real-world testbeds with different characteristics. We first analyze our new techniques individually, offering an in-depth discussion of design choices and performance trade-offs. Then, we embed all our techniques into a CTX-based system enabling a tag to self-localize entirely on-device, unlocking gains up to 44% in runtime and 85% in energy efficiency compared to traditional approaches.

0.37

Quantum-Enhanced Classical Communications for DPMZM-based NG-PON2 Systems

eess.SP quant-ph ·2026-11-06 ·Parvin Kumar, Sourav Banerjee, Neel Kanth Kundu, Umesh Sharma
unusual
Abstract

This paper presents a framework for integrating quantum detection techniques into a Dual-Parallel Mach-Zehnder Modulator (DPMZM)-based Next Generation Passive Optical Network 2 (NG-PON2) communication system. We present a system model for the DPMZM-based NG-PON2 quantum communication system and characterize its signal-to-noise ratio (SNR), demonstrating its superiority over conventional MZM-based systems. We analyze the performance of various quantum modulation techniques, including binary phase-shift keying, quadrature phase-shift keying, and M-ary quadrature amplitude modulation. A comprehensive analytical framework is developed to evaluate the SNR, BER, and mutual information for both classical and quantum detection schemes within the DPMZM-based NG-PON2 quantum communication system. Our simulation results show that quantum detection significantly outperforms classical detection, and the DPMZM architecture further enhances the performance over the baseline MZM architecture by improving the effective modulation efficiency and receiver sensitivity.

0.36

Real-Time Nonlinear Mixed-Integer Model Predictive Control With Horizon-Tail Relaxations

eess.SY math.OC ·2026-11-06 ·Armin Nurkanović
code/data
Abstract · Code available at: https://github.com/nosnoc/advanced-minmpc

Model predictive control (MPC) of nonlinear systems with discrete decisions leads to mixed-integer nonlinear programs (MINLPs) whose combinatorial complexity often prevents real-time deployment. In this paper, we propose a decomposition-based MPC method that reorders and reduces computations to shorten the delay between receiving a state measurement and applying the control input. Before the measurement arrives, we solve a continuous relaxation of the MINLP at a predicted state and construct a mixed-integer quadratic program (MIQP). Once the measured state is available, we solve the MIQP that reconstructs the discrete controls and simultaneously performs one sequential quadratic programming step toward the corresponding nonlinear solution. To reduce the combinatorial complexity in the horizon length, we either fix the integer controls or relax their integrality over the prediction horizon tail, while retaining an integer-feasible first control input. We derive error bounds relating the nonlinear objective values obtained with the horizon-tail approximations to the optimal value of the original mixed-integer problem. In a data center cooling example, the proposed method reduces mean feedback computation time by 94% relative to a state-of-the-art MINLP method with a total closed-loop cost difference of less than 1%.

0.36

Who Taught the Robot That? Removing One User's Corrections from a Robot Policy That Learns Online

cs.RO cs.HC ·2026-11-06 ·Eshika Pathak, Leela Krishna
Abstract · Accepted at the HumanRLI Workshop at CoRL 2026

A robot that keeps learning after deployment is corrected by several people in turn, and one of them may later ask to be removed. We show that whether the robot can forget that person depends on how people correct it. The goal of removal is the robot that would exist if the person had never been there, which we compute exactly in two simulated tasks with scripted users. When users correct the robot by taking over control, an ordered log of training updates is enough: retraining on the remaining sessions in their original order gives a robot as close to the goal as training noise allows. Deleting the person's data and retraining in one pass does not, even when the other users' corrections do not depend on the removed person, because it loses the order of training. When users only nudge the robot's action, the other users' corrections carry the removed person's influence, and no method we tested recovers the goal robot. Continuing to train without the person removes most of their habits, but not what their data in the training buffer keep alive. The log can identify one clearly harmful user, but not rank users by influence better than counting their corrections.

0.36

Multi-Mode Pinching-Antenna-Based OFDM Systems: A Robust Design Perspective

eess.SP ·2026-11-06 ·Hao Liang, Xidong Mu, Mingzhe Chen, Hyundong Shin, Michail Matthaiou
Abstract · 13 pages, 8 figures

A robust transmission framework is proposed for a multi-mode pinching-antenna (MPA)-based orthogonal frequency-division multiplexing (OFDM) system under both channel state information (CSI) errors and PA positioning errors. A physical model is first developed based on the multi-mode coupled-mode equations (CMEs), which characterizes the frequency-dependent guided-mode propagation and the coupling among guided modes and PAs over the OFDM band. Based on this model, a worst-case sum-rate maximization problem is formulated to jointly optimize the baseband beamformer and the intended PA positions under CSI and PA positioning errors, subject to user quality-of-service (QoS) and transmit-power constraints. To solve the resulting non convex problem, the worst-case CSI is exactly obtained for any given PA positions by solving trust-region subproblems, while the PA positioning errors are handled by grid-based and second-order approximation-based scenario constructions. Then, an alternating optimization (AO) algorithm is developed, where the baseband beamformer and the intended PA positions are updated via the weighted minimum mean-square error (WMMSE) method and successive convex approximation (SCA), respectively. Both subproblems are solved efficiently by exploiting the Karush--Kuhn--Tucker (KKT) conditions. Numerical results show that the proposed robust design outperforms the non-robust and conventional robust benchmarks with the same multi-mode architecture in terms of both worst-case sum rate and QoS satisfaction. They also demonstrate that the proposed robust design satisfies the QoS requirements of all users more reliably than the baselines that consider either CSI errors or PA positioning errors.

0.36

Expert Spindles in the Wild: Enriched Supervision for a Dense Detector

eess.SP ·2026-11-06 ·Olivier Pallanca, Alexis Bollengier, Abel Díaz Berenguer, Hichem Sahli, André Anjos
Abstract · 5 pages, 1 figure, 2 tables. Submitted to IEEE ISBI 2027 (under review)

Sleep spindles are a hallmark of healthy NREM sleep whose frequency, amplitude, density, duration and topography vary with brain maturation, age and disease. Their manual annotation is impractical and variable between scorers, yet the public references used to train automatic detectors cover these characteristics only partially: stage N2 only, a single central derivation, or short pre-selected segments. Measuring duration and density requires dense detectors that learn event boundaries from annotated onsets and offsets, and therefore inherit the protocol under which those boundaries were scored. We introduce Expert Spindles in the Wild (ESW), a stage-agnostic expert spindle annotation set covering 19 complete MASS-SS2 recordings ($\approx$150 h of continuous EEG, 29,819 events). A sleep neurophysiologist annotated the recordings using the full EEG montage, blinded to sleep-stage labels and prior spindle annotations. On the 15 nights shared with historical scorers, ESW recovers 92.7% of E1 and 73.7% of E2 events, while recovering 81.6% of the MODA expert-consensus events on the common excerpts. Despite stage-blind annotation, 99.0% of ESW events occur in N2/N3. Used to train a fixed dense detector on a frozen SleepFM encoder, ESW reaches an event-level $F_1$ of 0.680 on 166 unseen MODA subjects, compared with 0.635 for E2 ($ΔF_1$ = +0.045, paired-bootstrap 95% CI [0.024, 0.065]). ESW is thus useful supervision for cross-dataset detection, and closer annotation-level agreement with the MODA consensus did not translate into better transfer performance in this experiment. The annotations are released under the MASS data-use terms (https://doi.org/10.5281/zenodo.23894975).

0.36

PRISM: Primitive Routing via In-context Skill Mixing for Retentive VLA Adaptation

cs.RO ·2026-11-06 ·Jiachen Tao, Gengyu Zhang, Haoxuan Wang, Bing Liu, Yan Yan
top venue
Abstract · 19 pages, 6 figures, 4 tables. Accepted at NeurIPS 2026

Vision-Language-Action (VLA) models promise a single language-conditioned policy adaptable to many manipulation behaviors, yet deployment additionally requires skill retention and long-horizon composition: a policy must retain previously acquired skills and reuse them inside composite tasks whose behavior changes across phases. The standard recipe absorbs every new task into one shared update, which provides neither a parameter-level locus for reusable skills nor a mechanism for invoking each skill along the trajectory. We observe that this missing structure is latent rather than absent: per-skill low-rank updates of a pretrained VLA occupy a robustly distinguishable, low-dimensional subspace, and a single composite-task LoRA concentrates roughly half of its Frobenius energy in this subspace, against a near-zero matched random-rotation control. Building on this geometry, we introduce PRISM, which represents reusable skills as parameter-level primitives: each skill family is stored as a dedicated low-rank action-expert adapter in a primitive bank, primitives are augmented with skill-specific low-rank residuals and composed at inference by a per-chunk VLM-conditioned router that reads the prefix representation to track task progress, and the router is supervised by per-chunk labels from a deterministic, task-conditioned, action-derived oracle. During composite training, the vision encoder, the language LoRA, the skill residuals, the router, and the shared action projections and time MLP are updated; every primitive is held frozen, so the base per-skill action-expert adapters are preserved by construction. On RoboCasa365, PRISM achieves 56.43% atomic-skill success and reaches 20.00% on long-horizon composite tasks, while a standard $π_{0.5}$-LoRA two-stage fine-tune reaches 17.86% atomic and 17.50% composite success. We additionally validate PRISM on an AgileX PIPER real-robot setup.

0.36

Pinching Antenna Systems (PASS) for Air-Ground Integrated Sensing and Communications (ISAC)

eess.SP ·2026-11-06 ·Chaoran Chen, Haochen Li, Jincheng Zhu, Zhiwen Pan, Mark F. Flanagan
hot direction
Abstract · 5 pages, 6 figures

This paper proposes a multi-user single-target air-ground integrated sensing and communication (ISAC) system, in which a ground pinching-antenna system (PASS) transmits communication and sensing signals and an unmanned aerial vehicle (UAV) receives the target echo. The communication and sensing beamforming vectors, pinching antenna positions, and UAV position are jointly optimized to minimize the weighted Cramér-Rao bound (CRB). A penalty dual decomposition (PDD) algorithm is developed for the resulting nonconvex problem. The auxiliary variable and beamforming subproblems admit globally optimal solutions in semi-closed form, while the positions of the pinching antennas and the UAV are updated by successive convex approximation (SCA). Under mild conditions, every accumulation point of the sequence generated by our proposed algorithm is a Karush-Kuhn-Tucker (KKT) point of the reformulated problem. Numerical results show that the proposed PASS-UAV design achieves a lower weighted CRB than conventional multiple-input multiple-output (MIMO) and PASS benchmarks.

0.35

Agibot: An Open-Source Modular Swerve-Drive Robot Platform for Agricultural Robotics Research and Education

cs.RO cs.LGeess.SY ·2026-11-06 ·Moeen Ul Islam, Bishal Adhikari, Kori Edwards, Lorien Harvey, J. Alex Thomasson +1 more
code/data
Abstract · 14 pages. Submitted to IEEE Transactions on AgriFood Electronics (TAFE)

Agricultural production faces persistent labor shortages and increasing demand for automation, yet progress in agricultural robotics remains constrained by the limited availability of affordable, flexible, and accessible research platforms. This paper presents Agibot, an open-source, modular, low-cost agricultural robotics platform designed for autonomy research and experimentation, with potential applications in hands-on robotics education. The platform integrates four-wheel independent swerve drive, swappable computing, distributed embedded control, multimodal sensing, and configurable manipulation within a unified sense-plan-act-verify framework supporting perception, control, navigation, manipulation, and autonomous task verification. The platform was evaluated through representative row-crop navigation and manipulation tasks under varying lighting and soil conditions. Experiments demonstrated reliable lane following, autonomous row-to-row navigation, target identification, seed dispensing, plant manipulation, and closed-loop verification of task success. Its modular architecture supports independent testing, modification, and validation of autonomy components for rapid prototyping and experimentation. These proof-of-concept tasks, controller-tuning studies, and integrated trials demonstrate the platform's research capabilities, while its accessible and reconfigurable design supports its use for hands-on robotics education. By combining affordability, modularity, open-source accessibility, configurable manipulation, and end-to-end sense-plan-act-verify autonomy, Agibot provides a practical platform for agricultural robotics research, technology development, and workforce training.

0.34

Beyond the Pseudo-Inverse: GIM Precoding for Per-Antenna Power Minimisation in MU-MIMO

eess.SP ·2026-11-06 ·Hernan X. Cordova
theory
Abstract

We study linear precoder design for multi-user MIMO downlink under per-antenna power constraints (PAPC), which arise naturally when each antenna port has its own power amplifier. While the Moore-Penrose pseudo-inverse is the unique optimal generalized inverse for zero-forcing precoding under a total power constraint [1], we prove this optimality does not extend to the MMSE/RZF case under PAPC. Any precoder in the Generalized Inverse Matrix (GIM) family shares the same effective channel and per-user SINRs as the pseudo-inverse baseline, while its null-space component reshapes the per-antenna power profile. This separation principle enables minimising peak per-antenna power at fixed SINR targets with zero SINR penalty. We characterise the convex peak-power problem over this family and its active-set optimality, establish MRT and ZF asymptotic limits, and provide a convergent first-order subgradient algorithm. Simulations with 8-64 antennas confirm gains of 1.9-3.7 dB over pseudo-inverse RZF, with GIM-MMSE eliminating essentially all PAPC violations up to 20% per-antenna budget imbalance. The framework applies directly to fully-digital massive MIMO. Cell-free and hybrid analog-digital arrays are promising extensions.

0.34

Open-Source Multi-Fidelity EMT Modeling of Data Center Loads for Power System Studies

eess.SY ·2026-11-06 ·Xingyu Zhao, Yingyi Tang, Junbo Zhao
code/data
Abstract

Characterizing data center-grid interactions under rapid load fluctuations requires electromagnetic transient (EMT) models that capture fast converter and control dynamics. This paper develops a modular multi-fidelity EMT model of data center loads in PSCAD. The modular design supports flexible assembly of subsystem models for different power supply configurations. The multi-fidelity design incorporates switching models, average-converter models, and reduced-order models to balance simulation accuracy and computational efficiency according to analysis needs. The model also provides a systematic controller tuning procedure and supports control mode switching. Comparative simulations demonstrate how this data center load model responds to both grid-side and load-side disturbances and assess the accuracy and computational efficiency of models with three fidelity levels. The model is publicly available on GitHub.

0.34

DobotTrack: Target-Grounded Vision-Language-Action Learning for Embodied Person Tracking

cs.RO cs.CV ·2026-11-06 ·Shengjun Deng, Yinghan Zhang, Chengrui Wei, Liu Liang, Songyuan Liu +6 more
code/data
Abstract

Language-guided embodied person tracking requires a mobile agent to identify a target from a referring description and consistently follow the target under egocentric observations. We present DobotTrack, a target-grounded vision-language-action (VLA) framework that integrates auxiliary grounding supervision with efficient temporal modeling. We systematically investigate five auxiliary tasks---person segmentation, bounding-box grounding, visual-spatial classification (VSC), future DINO feature prediction, and future-frame prediction---and adopt complementary mask and bounding-box supervision to enhance target delineation and localization. We introduce Q-identity, a target-conditioned historical visual compression module that leverages target-grounding predictions to selectively preserve identity-discriminative appearance features and spatial cues in compact historical tokens. On EVT-Bench, our 0.8B-parameter model achieves an average success rate of 77.9%, exceeding OmTrackVLA by 16.9 percentage points across the three tasks. We further construct a real-world person-tracking dataset collected using a quadruped robot, comprising 126 episodes and 596,323 mask-annotated frames. The target masks enable automatic derivation of bounding boxes and visual-spatial labels encoding horizontal position and apparent image scale, providing additional grounding supervision without separate manual annotation of these labels. Together, our framework and annotation pipeline support learning target-grounded tracking policies from simulated and real-world data. Code is available at https://github.com/Dobot-Team/dobot_track.

0.33

SURE:Spatiotemporal Uncertainty Representation for Efficient Vision-Language-Action Learning

cs.RO cs.AI ·2026-11-06 ·Jingye Zhang, Choo Sin Wai, Dake Zhong, Zhifang Liu, Yuhong Zhang +6 more
Abstract

Vision-language-action (VLA) policies have become a promising paradigm for end-to-end robot control, but their effectiveness still depends heavily on large volumes of demonstration data. In manipulation tasks, this dependence is compounded by structured spatiotemporal uncertainty in the action space: multiple nearby actions or trajectories may successfully solve the same task, while semantically equivalent demonstrations often unfold at different speeds. Standard pointwise supervision treats each demonstrated action as an exact target, imposing an unnecessary learning burden under such uncertainty. We present SURE, a data-efficient VLA training framework that explicitly models both spatial and temporal uncertainty. SURE augments a VLA backbone with a lightweight transposed-convolutional probabilistic action decoder that captures spatial uncertainty through local correlations among neighboring action bins. It further introduces Soft-DTW alignment supervision to accommodate local timing variations between predicted and demonstrated action chunks. Together, these two components reduce sensitivity to demonstrator-specific poses and execution timing, enabling more data-efficient action learning. We evaluate SURE on LIBERO, RoboTwin 2.0, and real-world manipulation tasks. Even with only 10% of the demonstration data or 20% of the training steps, SURE achieves manipulation success rates that match or surpass baseline VLAs.

0.33

Maze Runner: Autonomous Maze Navigation with Line-Based Localization and Control Barrier Functions

cs.RO eess.SY ·2026-11-06 ·Alvaro J. Gaona, Lucas Gómez-Velayos, Peter Pasuy-Quevedo
code/data
Abstract · 10 pages, 8 figures, 1 table. Code: https://github.com/alvgaona/maze-runner

Maze navigation requires localization from sparse, repetitive geometry, planning through narrow corridors, and control that stays safe when the estimate or the map is wrong. We present Maze Runner, an open autonomy stack in C++ and ROS 2 that addresses these needs end to end on a simulated Husarion ROSbot XL with a 2D LiDAR in Gazebo. Perception extracts wall lines from each scan with RANSAC and total least squares, with per-line uncertainty. The same line front end feeds two estimators: an Extended Kalman Filter (EKF) that localizes against a line map built from the maze mesh, and an EKF SLAM variant that builds the line map online. Planning runs online on an inflated occupancy grid with A* plus line-of-sight shortcutting, or with a forward-only Hybrid A*. A nonlinear Model Predictive Controller (NMPC) tracks the plan while discrete-time Control Barrier Functions (CBFs) on world-frame LiDAR points keep the robot clear of walls and unmapped obstacles. The algorithms live in a ROS-free, unit-tested core library, and the whole stack builds and runs from a single reproducible environment. In simulation, the robot drives a 135 m route through the maze with 1.0 cm localization RMSE against the known map and 24 cm in SLAM mode, the planner returns a path in 6 ms, and the NMPC solves in 11.7 ms on average, never exceeding its 80 ms budget.

0.31

Data-driven representation of continuous-time linear time-invariant behaviors via Hankel integral operators

eess.SY math.OC ·2026-11-06 ·Fritz A. Engeln, Marta Zagórowska, Jan-Willem van Wingerden
Abstract · 9 pages, 1 figure

We present a data-driven representation of the finite-horizon behavior of deterministic controllable continuous-time LTI systems. We first introduce a notion of persistency of excitation for continuous-time signals that does not require the signal to be differentiable. We then show that, when the input is persistently exciting in this sense, the closure of the range of a Hankel integral operator built from a single input-output trajectory equals the behavior of the system over a finite time horizon. This result provides a nonparametric representation of system trajectories directly in terms of measured data, without requiring an explicit parametric model of the system.

0.31

Autonomous Whole-body Loco-manipulation Using a Snake Robot

cs.RO eess.SY ·2026-11-06 ·Adarsh Salagame, Henry Noyes, Haroon Hublikar, Alireza Ramezani
published
Abstract · Presented in IROS 2026

Snake robots can leverage their hyper-redundant morphology for both locomotion and manipulation through distributed body contact but whole-body loco-manipulation is particularly challenging since the entire body serves simultaneously as propulsive mechanism and manipulation tool Prior work on wheeled legged and snake robot platforms has decoupled these functions through dedicated appendages body segmentation or grippers We present closed-loop loco-manipulation with our snake robot COBRA without this distinction such that any part of the body may make contact We develop a contact-implicit dynamic model unifying robot-ground robot-object and object-ground interactions within a Moreau-Jean time-stepping framework and formulate a trajectory optimization over the full-order dynamics guided by a geometric planner and nominal proportional controller to autonomously push an object to a goal location during sidewinding We validate the planner and controller on hardware across varied configurations and present simulation results for a simplified trajectory optimization that resolves robot-object and object-ground contacts

0.31

A Phase-Balanced Differential Rydberg-Atom Quantum Receiver for Weak GNSS Signal Detection

eess.SP quant-ph ·2026-11-06 ·A. K. Mandal, F. Zhang, J. H. Mikkelsen, M. Shen
Abstract · 6 pages, 10 figures, submitted to IEEE ICC 2027 (in review)

This paper proposes a phase-balanced differential Rydberg atomic quantum receiver (DRAQR) for weak-signal recovery, using local-oscillator phase control to reinforce differential signal responses and a shared squeezed-reservoir model to correlate inter-cell quantum projection noise (QPN) and photon shot noise (PSN). An analytical model incorporating QPN, PSN, intrinsic thermal noise (ITN), and inter-cell correlations quantifies signal-to-noise ratio (SNR) gain, outage-constrained minimum detectable power, and bit-error rate (BER) for square $M$-QAM. For the evaluated settings, SNR gain reaches approximately $10$~dB as PSN correlation approaches unity, with BER decreasing as correlation increases. Under Rician fading, at a $0$-dB threshold SNR and maximum outage probability of $0.1$, the minimum detectable power decreases from $-146.95$~dBm at $K=1.155$ to $-159.98$~dBm at $K=2.4$. Analytical results are validated against Monte-Carlo simulations over $150{,}000$ data runs per point for receiver-performance plots.

0.31

A wearable input/output interface learns personalized models of haptic perception

cs.RO ·2026-11-06 ·Wen Li, E. Magnus Adams, Zijie Xu, Jiawen Yao, Aybala Esmer +3 more
hot direction
Abstract

Wearable haptic interfaces deliver physical stimuli to the skin, yet perceived locations vary widely across users and diverge from actuator locations, particularly for emergent tactile-feedback techniques. We developed a wearable haptic input/output (I/O) device integrating a 36-channel electrotactile array with capacitive touch sensing, enabling stimulation and user-reported percept locations to be registered within the same body-referenced coordinate frame. In 24 participants, the device revealed electrotactile and referred-thermal perceptual maps with notable interparticipant variability. Despite this variability, both modalities retained electrode-pair-specific spatial structure across participants. Tactile and thermal percepts also showed spatial correspondence, whose strength varied across participants. These findings motivated our human-in-the-loop calibration framework, which adapts a forward-prediction strategy to individual users and generates personalized perceptual models. These models enabled inverse rendering by selecting stimulation patterns predicted to generate a target percept for a specific user. Using these models, we demonstrate personalized haptic rendering in haptic-guided robot teleoperation, visual-to-haptic sensory substitution, interpersonal haptic communication, and virtual reality (VR). Together, this platform provides a route toward adaptive haptic interfaces that measure, model, and render according to each user's perception.

0.31

Quantum Natural Muon: Spectral-Natural Optimization for Variational Quantum Circuits

eess.SY ·2026-11-06 ·Duong The Do, Duong Tung Nguyen
Abstract

Training parameterized quantum circuits requires optimization methods that account for two distinct structures: the information geometry of quantum states and the matrix or tensor structure induced by circuit parameterization. We introduce Quantum Natural Muon (QNM), a spectral natural-gradient framework that applies polar normalization to circuit-semantic parameter blocks after damped quantum Fisher preconditioning. This ordering is essential: quantum Fisher preconditioning and blockwise polar normalization do not commute in general, so QNM is not an order-free composition of quantum natural gradient (QNG) and Muon. We show that Frobenius-matched polar mixing preserves active singular subspaces while raising a spectral floor and improving active conditioning, and that the resulting direction remains a controlled perturbation of QNG. For its distance-aware realization (DA-QNM), we further establish alignment, descent, and nonconvex stationarity guarantees for the normalized update, together with perturbation bounds under gradient and QFIM estimation error. Experiments on variational quantum eigensolving, weighted QAOA-MaxCut, and supervised quantum classification isolate the roles of quantum geometry, circuit-semantic spectral structure, and radius control. The strongest gains arise when the spectral blocks match the circuit structure, supporting the central principle that quantum-state and structured spectral geometries provide complementary information for variational quantum optimization.

0.31

Two-Timescale Receive Combining for Fronthaul-Constrained Uplink XL-MIMO with Visibility Regions

eess.SP ·2026-11-06 ·Jaebum Park, Robert W. Heath
theory
Abstract · The manuscript was submitted to IEEE, and is currently under review

Extremely large-scale multiple-input multiple-output (XL-MIMO) arrays are hard to deploy in virtualized radio access networks, because the fronthaul data rate grows with the number of antennas when the radio unit forwards every antenna sample. We propose a two-timescale receiver that allocates the per-user combining ranks under the fronthaul stream budget from the channel statistics. The radio unit applies the long-timescale transform, after which the distributed unit applies a minimum mean-squared-error receiver in the reduced dimension. For disjoint visibility regions, we prove that the reduced-dimension receiver decouples across users. We characterize the sum-rate loss to second order and show that the greedy allocation, which we call saturating capture, minimizes its leading term. In simulations, the allocation gains over a uniform-rank baseline, matches a budget-trimmed threshold rule at a small number of users and exceeds it as the array and the user count grow. The gain grows with the heterogeneity of the per-user eigenvalue profiles and vanishes for identical profiles. Under out-of-cell interference offset by at least half a window width, the gain survives with clean covariances, which window-restricted estimation keeps for disjoint windows.

0.31

Safe Circumnavigation on Heterogeneous Orbits: A Range-Only HOCBF-QP Framework for Nonholonomic Multi-Agent Systems

cs.RO eess.SY ·2026-11-06 ·Shubham Singh, Anand Singh, Anjali Tiwari, Lara Briñón-Arranz, Katsuki Fujisawa
Abstract

Circumnavigating a stationary target with a team of robots requires each robot to hold its own standoff distance, to stay evenly spaced from its neighbours, and to avoid collisions with the target and with each other at every instant. The task becomes arduous when the robots are assigned different orbital radii, as their sensors or roles may demand. This paper presents a safety-critical control framework for this problem. Each robot commands both its acceleration and its turn-rate, measures only its distance to the target, and exchanges information only with its two ring neighbours. The design has three features. First, a structural property of the model lets one control law serve every robot regardless of its assigned radius. Second, the acceleration drives each radius exactly to its target, while the turn-rate balances the formation with the neighbours. Third, a higher-order control-barrier-function quadratic program enforces target, geofence, and inter-robot safety on both inputs at once; it leaves the nominal controller untouched away from the constraints and stays valid while the unmeasured range rate is still being estimated. Simulations confirm these guarantees. In a comparison, the closest existing multi-orbit design, even with full state information, violates both the target standoff and the inter-robot margin, whereas the proposed controller does not.

0.31

Scalable BLE Channel Sounding Localization with Ranging-Assisted PDoA

eess.SP cs.NI ·2026-11-06 ·Jorg Wieme, Eli De Poorter, Jeroen Hoebeke
code/data
Abstract · 14 pages, 5 figures, 7 tables, Appendix, For reproducibility code available at https://gitlab.ilabt.imec.be/iwine/ra-pdoa-simulator

Bluetooth Channel Sounding (CS) enables accurate phase-based ranging, but multi-anchor localization requires sequential ranging exchanges that limit scalability and increase measurement latency. This paper proposes Ranging-Assisted Phase Difference of Arrival (RA-PDoA), a scalable BLE localization architecture combining one active anchor performing a standard two-way CS exchange with multiple passive receive-only anchors observing the same transmissions. A channel-recovery framework is derived that enables passive anchors to reconstruct a channel response suitable for super-resolution delay estimation, allowing multipath-robust ranging without participating in a two-way exchange. Coarse arrival-time information is further used to bound the position estimate and reject inconsistent range solutions. A comprehensive evaluation characterizes the impact of multipath propagation, body shadowing, synchronization requirements, oscillator impairments, anchor placement, and deployment inaccuracies, thereby quantifying the practical infrastructure requirements for scalable BLE localization. Simulation results in modeled indoor environments show that RA-PDoA reduces the 90th-percentile localization error from 175.3 cm to 95.5 cm under body shadowing while maintaining an airtime requirement independent of the number of anchors. The results demonstrate that a single active ranging link enables scalable and multipath-robust BLE Channel Sounding localization.

0.30

Polarization calibration of fisheye cameras using skylight in the wild

cs.RO ·2026-11-06 ·Shuo Wang, Shiqi Che, Dianze Dong, Wenxin Liu
Abstract · 8 pages, 10 figures, submitted to ICRA 2026

Skylight polarization provides directional cues for robotic navigation, but fisheye lens effects and analyzer imperfections can degrade polarization measurements across the field of view. We present a joint lens-analyzer calibration method that explicitly models incidence-dependent lens transmission together with analyzer gains and angular offsets in a unified eight-parameter formulation. Given known geometric calibration, rotations between camera frames can be estimated using visual features and cross-frame directional polarized light consistency can be used for polarization calibration. Therefore, our calibration approach requires only a short handheld outdoor sequence under natural skylight, without dedicated polarization calibration equipment or a predefined sky polarization model. On real-world outdoor sequences, the proposed method achieves a weighted cross-frame angle-of-linear-polarization (AoLP) mean absolute error below 0.5° across all incidence angles, with a 73.4% reduction in the 80°-90° incidence angle region, and generalizes to an independent sequence with an overall mean absolute error of 0.466° without refitting. Preliminary solar-azimuth experiments using a camera mounted on a quadruped robot further show improved alignment with the visible solar center, supporting the potential of calibrated AoLP measurements for relative-heading estimation and future polarization--inertial navigation. The calibration software tool will be released upon acceptance of this paper.

0.30

Scenario Orchestration for Closed-loop Policy Evaluation of Autonomous Driving

cs.RO ·2026-11-06 ·Frieda Rong, Zhaoyue Wang, Veer Kapadia, Vishaka Rajani, Aren A. Babikian +4 more
Abstract · 8 pages, 5 figures

Scenario-based simulation is central to evaluating autonomous-driving policies, but a scenario scripted before execution (i.e., with predetermined, fixed actor routes and speeds) may not preserve its intended interaction once the tested ego policy responds in closed loop. Scripted scenarios cannot adapt to unanticipated ego behavior, while learned traffic actors may remain reactive without reliably realizing a particular scenario intent. We introduce a scenario orchestrator that continuously monitors the joint traffic state and adapts non-ego actors through re-timing, re-routing, and role reassignment for creating a specific scenario intent. At each step, the orchestrator solves a constrained optimization problem that preserves the intended interaction while limiting interventions and maintaining traffic feasibility. We further distinguish scenario realization from policy failure, allowing a test to first be judged based on whether the intended interaction occurred and then by how the ego policy responded. Experiments in the CARLA simulator across multiple scenario families and driving policies show that orchestration realizes targeted interactions more reliably than scripted and learned actor-level baselines, particularly under variations in initial conditions and ego behavior. These results support closed-loop scenario orchestration as a tool for controlled and reproducible policy evaluation.

0.30

A Hybrid Digital/Analog Polyphase Filtering Approach for High-Speed RF-DACs

eess.SP ·2026-11-06 ·Behdad Jamadi, Jungmin Lee, Tarun Cousik, Rohit Rangaraj, Jeffrey S. Walling
Abstract · 11 pages, 5 figures, 5 tables

This paper proposes a hybrid polyphase interpolation for RF-DACs that removes the need for full-rate digital commutation by using passive combining on the RF-DAC itself. A P-branch polyphase FIR filter is factored into M streams, each formed by a (P/M)-way digital commutator at fRF/M and driving a dedicated RF-DAC slice, giving either partial digital (1<M<P) or fully analog (M=P) recombination. We show that the slice array acts as a single full-rate DAC with an M-times longer zero-order hold, derive yield laws for slice gain, timing and unit-capacitor mismatch, and find that with retiming the binding requirement is sub-picosecond matching of the fRF clock arrival, so the best choice is the smallest M that closes digital timing. Two 28-nm digital macros, designed for a 12.8GSps output rate and verified by post-layout simulation, illustrate a 4-phase variant that closes timing up to 14.4GSps with its digital logic at fRF/4 and an 8-phase variant that closes timing up to 16GSps with its digital logic at fRF/8. The analog recombination is evaluated with a behavioral model of the slice array. The architectures target wideband software defined radio (SDR) platforms, including cognitive radio, electronic warfare, and satellite communications.

0.30

Geometric Pairwise Classification of Affine Constraints over Ellipsoidal Regions

eess.SY math.OC ·2026-11-06 ·Igor M. L. Pataro, Martina Mammarella, Fabrizio Dabbene, Teodoro Alamo
Abstract · 6 pages, 1 figure

This paper presents a deterministic geometric procedure for analysing affine constraints over fixed ellipsoidal regions, motivated by chance-constrained set approximation. Analytical tests classify inequalities as always violated, never violated, or cutting. Pairwise tests determine whether two inequalities can be strictly violated at a common point, yielding a compatibility matrix. The tests follow from quadratic optimisation and the Karush--Kuhn--Tucker conditions, with explicit treatment of tangencies, parallel normals, and degenerate inequalities. The formulation applies in arbitrary finite dimensions and provides geometric information for subsequent bounds on simultaneous sampled violations and safe-region approximation.

0.29

Locating Contact on Tendon-Driven Continuum Robots From Tendon Measurements Alone via Constrained Moving Horizon Estimation

cs.RO ·2026-11-06 ·Haitham El-Hussieny
benchmarktheory
Abstract

Tendon-driven continuum robots can reach through constrained spaces, but contact along their bodies is invisible to the sensors at their motors. We show that this is a structural limitation rather than a tuning problem: the map from contact forces to tendon measurements has a null space of dimension at least seven, so no estimator that leaves the contact support free can locate contact. When the obstacle geometry is known, however, non-penetration and complementarity make the contact force locally identifiable, and we prove that the resulting system with unknown contact inputs is locally exponentially strongly detectable. We present a constrained moving horizon estimator (MHE) that encodes these conditions and runs in real time as one quadratic program per sample, and we give a local, mode-wise robust stability result for it. In a finite- element benchmark of a two-section robot pressing and sliding on a wall, the estimator located contact to 0.9 mm and estimated contact force to 0.22 N, compared with 65 mm and 0.72 N for a tuned extended Kalman filter (EKF) with state projection. The estimator tolerated millimeter- level errors in the obstacle position and unmodeled wall friction, under which the projected EKF failed. Under randomized model mismatch over 40 trials, including unmodeled tendon friction, it remained better in force and location in every trial. Without contact, the MHE offered no advantage over the EKF, which confirms that its benefit comes from the contact constraints. This report extends a letter submitted to IEEE Control Systems Letters with complete proofs and additional experiments.

0.29

Bayes-Optimal Quantum Computational Sensing via Convex Spectral Design

eess.SP quant-ph ·2026-11-06 ·Amirhossein Taherpour, Alireza Sadeghi, Georgios B. Giannakis
Abstract

Quantum computational sensing performs task-specific inference directly through coherent sensing dynamics, avoiding an intermediate estimation step. Existing designs, however, typically optimize quantum control parameters through nonconvex programs whose performance can depend strongly on initialization. This contribution formulates binary phase classification directly in terms of the sensor's measurement response. For a fixed query budget, generalized quantum signal processing (GQSP) restricts this response to a bounded trigonometric polynomial, while the Bayes classification error is linear in the response. These properties yield an exact finite-dimensional semidefinite program for the globally Bayes-optimal sensor response. Fejér--Riesz factorization and GQSP synthesis then recover a physically realizable control sequence. The same formulation extends naturally to sample-based design and additive phase uncertainty. Numerical tests show improved accuracy and optimization reliability over direct control-angle training; noise-aware redesign also lowers error under phase uncertainty.

0.29

An Ultra-Low-Noise Tunable TIA in 45nm Silicon Photonics for Electro-Optical Ultrasound Data Acquisition

eess.SP eess.SY ·2026-11-06 ·Marziyeh Rezaei, Sridhar Singhal, Alp A. Derin, Soner Sonmezoglu, Sajjad Moazeni
unusual
Abstract

This paper presents the design and experimental characterization of a 32-channel, WDM-compatible electronic-photonic receiver array fabricated in the GlobalFoundries 45nm monolithic silicon photonics process. Optimized for miniaturized, millimeter-scale electro-optical ultrasound data acquisition probes for broadband photo-acoustic imaging (PAI), the chip utilizes integrated micro-ring resonator (MRR) demultiplexers and photodiode (PD) arrays to decode multi-wavelength acoustic signals outside the body. This work focuses specifically on the comprehensive characterization of the receiver front-end, evaluating both the optical WDM filter bank and the transimpedance amplifier (TIA) performance. Monolithic co-integration minimizes critical photodiode-to-TIA interface parasitics. The TIA incorporates a DC-offset cancellation loop that compensates for photodiode-induced offset currents while providing tunable low-frequency cutoff control over a range of 30 kHz to 1.5 MHz. This tunability enables effective suppression of flicker noise and improves overall signal integrity. The front-end achieves the peak transimpedance gain of 144 dB$Ω$ while maintaining an ultra-low input-referred noise current density of $1\text{ pA}/\sqrt{\text{Hz}}$, demonstrating a highly scalable architecture for high-sensitivity intravascular and endobronchial imaging.

0.29

Task-Driven Closed-Loop Design Optimization of Passively Morphing Quadrotors for Contact-Mediated Narrow Gap Traversal

cs.RO ·2026-11-06 ·Etor Arza, Luca Girardi, Gabriel Maquignaz, Javier Paez, Stefan Leutenegger +2 more
code/data
Abstract · Video: https://youtu.be/xGkEtl6-vCA. Code: https://github.com/ntnu-arl/soft-quadrotor-optimization

Aerial robots operating in cluttered environments may encounter openings narrower than their nominal width. Existing passively morphing drones reduce their width upon traversal, but they typically involve human-made designs, which can struggle to find the right balance among folding ratio, flight efficiency, and control tracking performance. We introduce a closed-loop design method that jointly optimizes quadrotor geometry and the torque-deflection response of two passively folding arms driven by the success rate in simulation for contact-mediated gap traversal. An optimization algorithm samples candidate designs that are evaluated in a contact-based simulation of the traversal task, where the task success rate is used as a feedback signal to sample progressively better designs over time. The resulting prototype passively reduces its width from 31.3 to 14.1 cm and traverses a 15 cm wide opening at a 2 m/s reference speed. It also tolerates gap approach-angle errors of up to 30 degrees and folds through a 40 cm long confined passage. Compared to a rigid twin, the soft configuration adds only 2 mm of position RMSE and less than 1% electrical power during contact-free flight. These results show that task-driven closed-loop design optimization can produce passively morphing airframes that are compact under contact, tolerant to approach error, and effectively rigid during nominal flight.

0.28

RevoSim: Scalable Multimodal Tactile Simulation for Dexterous Manipulation

cs.RO ·2026-11-06 ·Ke Zhang, Lixin Xu, Ziyi Wang, Xiyue Dong, Jie Tan +4 more
code/data
Abstract

We present RevoSim, a scalable and high-fidelity multimodal tactile simulator for the BrainCo Revo3 hand that jointly models 285 piezoresistive taxels and five visuotactile fingertips. RevoSim efficiently generates synchronized pressure, depth, RGB, and marker-displacement observations; on a single RTX 5090, its depth-conditioned marker-displacement configuration supports up to 4,096 parallel environments. We train in-hand manipulation and grasping policies entirely in simulation, establishing a foundation for future sim-to-real transfer. Project page: https://braincotech.github.io/RevoSim-web/.

0.28

A Novel Algorithm to Simulate Conductance-based Neuronal Networks

eess.SY ·2026-11-06 ·Amir Shahhosseini, Rodolphe Sepulchre
theory
Abstract

Conductance-based models are central to neurophysiology and neuromorphic engineering, yet their simulation scales poorly to large networks. We study and analyze an iterative solver that operates in signal space. Each iteration reduces to $N$ decoupled scalar linear time-varying differential equations, one per neuron, which are solved in parallel and in closed form. The solver admits two complementary interpretations: a variable metric forward-backward splitting method, rooted in operator theory, and a waveform relaxation quasi-Newton method, rooted in circuit simulation. We prove uniform and superlinear convergence, with a bound that reflects the conductance structure of the dynamics. The algorithm is illustrated on the classical Hodgkin--Huxley neuron and on networks of up to 5000 neurons.

0.28

ToolDex: An Ego-Exo Human Demonstration Dataset for Creative Physical Tool Use

cs.RO cs.AI ·2026-11-06 ·Mohamed Elmoghany, Kehan Li, Ruoqu Chen, Tarun Chiruvolu, Shutong Zhang +9 more
benchmarkcode/data
Abstract

Creative tool use is a fundamental human capability and an important step toward general-purpose embodied intelligence. It requires agents to reason about object affordances, physical constraints, and manipulation strategies, especially when standard tools are unavailable. However, progress in this area is limited by the lack of real-world datasets that capture human tool-use behavior with sufficient scale, fidelity, and diversity. Existing benchmarks are often simulation-based, focus on simplified interactions, or lack detailed observations and annotations. We introduce ToolDex, a large-scale egocentric-exocentric (ego-exo) human demonstration dataset for creative physical tool use. ToolDex captures scenarios in which humans select, adapt, and repurpose available objects as tools. Using synchronized egocentric cameras and eight exocentric cameras, the dataset provides high-fidelity multi-view observations and rich annotations for hands, tools, target objects, and their deformable interactions. Beyond data collection, we establish a benchmark suite characterizing three core capabilities: physical tool-use reasoning, sparse-view 3D and 4D reconstruction, and robot policy learning. ToolDex provides a foundation for studying human tool-use behavior and developing embodied agents that can adaptively use tools in complex real-world environments. Dataset visualizations and benchmarking results are available at https://tooluse-dataset.github.io/.

0.28

DeepRT-Cal: Learning to Calibrate Ray-Traced Twin Channel Statistics with Low-Fidelity Environmental Geometry

eess.SP ·2026-11-06 ·Shaoyi Liu, Jianhua Zhang, Li Yu, Yichen Cai, Tao Wu +3 more
Abstract · 15 pages, 14 figures, 3 tables

Sixth-generation (6G) wireless systems increasingly rely on digital twin channels (DTCs) to represent and interact with physical radio environments. Ray tracing (RT) is naturally suited to DTC construction because it provides deterministic and physically interpretable channel predictions. However, high-fidelity three-dimensional environment models are expensive to obtain, whereas geometric inaccuracies in low-fidelity models can alter the detected multipath components (MPCs) and consequently distort received-power and delay-domain twin channel statistics. Building on the DeepRT ray-tracing framework, this paper proposes DeepRT-Cal, a geometry-aware cross-fidelity calibration framework that predicts high-fidelity twin channel statistics from low-fidelity geometry and RT features.The network encodes global link descriptors, masked low-fidelity MPC tokens, propagation-corridor geometry, and link-centered environmental objects. During training, detailed geometry is introduced as privileged information with structured masking, while deployment requires only low-fidelity geometry and low-fidelity RT information. A link-conditioned cross-attention module retrieves the environmental context relevant to each Tx--Rx link. The prediction heads estimate high-fidelity twin channel statistics, including total received power, RMS delay spread, mean excess delay, and a normalized power-weighted power-delay profile (PDP). Outdoor evaluation shows that, in an non-line of sight (NLoS) region, the method reduces power MAE from 16.80 to 3.88 dB and RMS delay-spread MAE from 304.89 to 98.19 ns. In a LoS-dominated region, power prediction improves, while RMS delay-spread accuracy remains comparable to direct RT.

0.28

Adaptive PD-like Fault Tolerant Controller for a Class of Nonlinear Systems

eess.SY ·2026-11-06 ·Alessio Dallabona, Vicenç Puig, Mogens Blanke, Dimitrios Papageorgiou
Abstract

The article presents a nonlinear adaptive control strategy for a class of feedback-linearizable systems subject to process faults and disturbance. The controller is derived through Lyapunov analysis and incorporates an adaptive proportional gain within a PD-like structure, preserving simplicity while providing nonlinear adaptive behavior. A residual-based gating mechanism activates the adaptation only when the tracking error exceeds a prescribed threshold, preventing unnecessary gain updates during nominal operation and improving robustness to noise. The proposed design yields uniform global ultimate boundedness of the closed-loop signals, with explicit expressions for the steady-state tracking bound. Practical tuning guidelines for the adaptive parameters and Lyapunov matrix selection are provided to ensure consistent gating behavior. Simulations on a precision positioning system illustrate effective compensation of process faults, rapid gain escalation during degraded conditions, and smooth reversion to nominal performance, demonstrating that the method enhances fault tolerance while retaining the structural simplicity of PD control.

0.28

Autonomous Particle Image Velocimetry Using Zero-Shot Sim-to-Real Reinforcement Learning

cs.RO cs.LG ·2026-11-06 ·Chang Rajani, Luke Sebastien Marsden, Stephan Kunz, Claudio Mucignat, Ivan Lunati
Abstract · 9 pages

Particle image velocimetry (PIV) reconstructs a flow field from the motion of tracer particles but the reconstruction quality is bounded by how well the flow structure of interest is resolved in the images. The acquisition position is typically fixed by the operator before the measurement. We cast camera placement as a sequential decision-making problem and train a reinforcement learning agent to control a 6-DoF manipulator equipped with a camera, enabling it to locate, frame and resolve a hydrofoil wake in a water tunnel. The agent commands the camera pose and observes the scene only through the velocity field and uncertainty estimated from time-separated image pairs. It is trained entirely in a photorealistic simulation of our water tunnel, with wakes generated by Computational Fluid Dynamics (CFD) simulations of hydrofoils with varying profiles, positions and angles of attack. After zero-shot transfer to the real water tunnel, the policy reliably locates and frames the wakes of randomly placed hydrofoils. It resolves the wake at a higher resolution than a fixed camera view and with lower error in the measured flow statistics, in roughly half the time of a full scan and with far fewer image pairs.

0.28

What Limits mmWave Human-State Sensing for Robot Perception?

cs.RO cs.HC ·2026-11-06 ·Shuntian Zheng, Guangming Wang, Hongyang He, Shuai He, Yu Guan
Abstract

Assistive robots and search-and-rescue platforms require reliable human state sensing in smoke, darkness, and privacy-sensitive environments, where cameras and LiDAR may be unreliable Millimeter-wave radar offers an alternative that works without ambient light, penetrates smoke, and does not capture visual appearance. However, the compact arrays that fit robot size, power, and cost budgets have limited apertures, producing joint-localization errors on the order of decimeters at a few meters, large enough to compromise the safety margins and tracking accuracy a robot depends on. Existing pipelines reduce pose error through frame fusion, multiple viewpoints, return selection, or stronger estimators, but examine one technique at a time, without relating its effect to a shared measurement limit or separating gains from added information from gains from estimator completion. We address this gap with a framework that models the human and environment as one time-varying scattering field and states a Cramer-Rao type error floor for joint localization as a function of bandwidth, aperture, signal-to-noise ratio, fused frame count, phase coherence, and visibility of retained returns. We then use existing frame fusion, multi viewpoint, return selection, and pose estimation methods as matched probes of this floor rather than as competing pipelines. Under the evaluated conditions, experiments on three datasets and a live deployment show these strategies move joint localization error in the predicted directions, for example lowering mean per-joint position error from about 97 to 65 mm with a second viewpoint, while a stronger estimator mainly lowers overall error by completing joints the measurement cannot resolve. These results give the robotics community a principled basis for deciding whether to prioritize observation design or estimator capacity when deploying reliable mmWave human sensing.

0.28

Nonlinear Fault Tolerant Control of Wind Turbine Pitch Systems: A Comparative Study

eess.SY ·2026-11-06 ·Alessio Dallabona, Mogens Blanke, Dimitrios Papageorgiou
Abstract

Electro-hydraulic pitch systems account for about one fifth of wind-turbine failures, motivating the development of resilient control strategies for fault mitigation. The study considers a large set of robust nonlinear control designs of different feedback principles and architectures, and assesses their fault-tolerant capabilities in comparison to each other and against the industrial standard, i.e. the Proportional-Integral (PI) controller. The implementation and tuning of each scheme is based on a high-fidelity simulator to ensure nominal performance, whereas each controller is deployed unchanged on an experimental apparatus that includes a full-scale electro-hydraulic pitch system. The comparative evaluation of the considered fault-tolerant solutions is carried out through several tests that cover nominal operation and twenty-five representative fault scenarios, including sensor, process, and actuator faults. The analysis highlights the trade-off between performance preservation under faults and increased actuator chattering. The experimental results show that introducing a dynamic input-reconstruction module into the design of the nonlinear controllers significantly reduces actuator chatter without compromising their fault-tolerant properties.

0.28

Payload-Blind Bistatic Delay-Doppler Sensing with Phase-Injected Chirps in CE-OFDM

eess.SP ·2026-11-06 ·Gustavo P. Gonçalves, Henrique L. Senger, Bruno S. Chang, Giuseppe Thadeu Freitas de Abreu
benchmark
Abstract

Constant-envelope orthogonal frequency-division multiplexing (CE-OFDM) preserves a unit-magnitude sample envelope, but its Hermitian-symmetric structure makes dedicated sensing pilots costly, and additive pilots disrupt this envelope. We show that injecting a known chirp, here a Zadoff-Chu (ZC) sequence, into the CE-OFDM phase enables payload-blind bistatic delay-Doppler sensing without dedicated sensing frames or an explicit pilot power split: the residual carrier, given by the ensemble mean under the adopted payload model, becomes an implicit superimposed pilot for a receive-only node. A Doppler-aligned receiver integrates this reference coherently across frames while the payload residual averages down. In contrast, phase-domain injected training (PIT) frame averaging, designed for quasi-static channel estimation, loses about 32 dB for the simulated target with 16 averaged frames. Closed-form approximations characterize the payload-induced floor, known-cell detection probability and direct-path tolerance. In the noise-limited simulations, the receiver incurs a 2.8 dB detection gap relative to a payload-aware matched filter without interference cancellation and matches a dedicated-pilot benchmark of equal reference energy without its 53% frame overhead.

0.27

Reconciling European Electricity Data with Diffusion Priors Learned from Observations Alone

eess.SY cs.LG ·2026-11-06 ·Matthias Pirlet, Adrien Bolland, Gilles Louppe, Damien Ernst
Abstract

The ENTSO-E Transparency Platform publishes generation, load, and cross-border power flows for European bidding zones. The values carry undocumented errors and violate energy conservation, and data reconciliation consists in correcting them to satisfy that balance. In this work, we formulate data reconciliation as Bayesian inference. Classical methods implicitly take a uniform prior, which we replace with a diffusion model estimated from the observations alone by empirical-Bayes expectation-maximization. The resulting posterior is sampled with a guided process whose projection step makes every sample satisfy the balance exactly. On synthetic systems of six to thirty-five bidding zones, built from ENTSO-E resource-adequacy data, the diffusion prior gives the most accurate reconciliation, increasingly so as the system grows. Its credible intervals are calibrated or conservative, and it remains the most accurate under a misspecified noise covariance. Finally, when the generation of entire zones goes unreported, the classical methods fail while the learned prior recovers it from cross-zone structure.

0.27

Real-Time Trans-Oceanic Fiber Sensing via Convergence of Long-Reach OFDR, High-Speed Optical Networking, and RDMA-Capable Edge-GPUs

eess.SP ·2026-11-06 ·Mikael Mazur, Nicolas K. Fontaine, Roland Ryf, Sarvesh Bidkar, Tracy Chodkiewicz +19 more
Abstract · 3 pages, 2 figures

We demonstrate real-time network-integrated bi-directional fiber sensing over a 4400-km subsea cable. Streaming and fusing two OFDR instruments via RDMA over a live transoceanic 130Gbaud coherent link and GPU edge-compute enabling parallel capture of multiple spatial resolutions from ~10m to ~100m.

0.27

Can LLMs Debug Robots by Editing Their Weights?

cs.RO cs.AI ·2026-11-06 ·Jusheng Zhang, Keze Wang, Liang Lin
benchmark
Abstract

A language model can repair a robot policy's lexical instruction grounding by editing just 65 token-embedding rows. We bring weight-space programming to robot foundation policies: GPT-6 Astra reads weights and uses a CPU workspace and budgeted simulator feedback, without gradient updates or new demonstrations. Three independent sessions on $π_{0.5}$ all improve success on 765 held-out paraphrases; the best raises success from 86.3\% to 92.0\% (48 repairs, four regressions; exact paired $p=1.31\times10^{-10}$). Among the 335 instructions containing edited tokens, success rises from 74.3\% to 87.5\%; the other 430 episodes have zero outcome changes. Matched random-direction edits do not help. Full replacement using the editor's own lexicon explains much of the gain but produces more regressions; its direct difference from the editor is not significant at 0.05. Across 13 further robot sessions, the editor declines 12 candidates; the sole retained edit does not show a significant held-out gain. Separately, benchmark metadata lowers FastWAM success from 70.0\% to 48.6\%, while the evaluated VLA's aggregate rate is unchanged. The evidence supports precise lexical repair, not general acquisition of missing robot capabilities.

0.27

LACART: Lookahead-Augmented Collision Avoidance and Reference Tracking Strategies for Safety-Critical Robots

cs.RO eess.SY ·2026-11-06 ·Pedro Lopes, Amit Shivam, Luis Tiago Paiva
Abstract

Autonomous racing provides a demanding testbed for collision avoidance under tight geometric and computational constraints. Building on a recent conference contribution and a master's dissertation, LACART combines lookahead-based control barrier functions (CBFs), a relaxed control Lyapunov function (CLF), and quadratic programming (QP) to modify nominal vehicle commands for obstacle avoidance and reference tracking. The method handles circular obstacles and track boundaries in a common optimisation layer. Comparative simulation results against Huang \emph{et al.} \cite{r9} show similar full-track paths and lower median computation times: 0.31 versus 0.33 ms per step and 0.023 versus 0.025 ms per QP solve. The dissertation further demonstrates a 1:43-scale vehicle with camera-based estimation and wireless actuation at a reported 25 Hz control rate. We identify the conditions linking lookahead-point clearance to body clearance, illustrate hardware limitations, and propose a recorded demonstration of safety intervention and recovery. The contribution targets ground-robot collision avoidance and real-time safety monitoring.

0.27

An Inspection Game for Monitor Design Against GNSS Timing Attacks in Synchronous Islanding

eess.SY cs.GT ·2026-11-06 ·Gregory Provan, Quanyan Zhu
Abstract

Phasor measurement units (PMUs) take their time from GNSS, so a spoofer that slowly drags a receiver clock rotates the reported phase without touching the power waveform. The threat is acute in synchronous islanding, where a controller keeps an islanded generator in phase with the grid, using one PMU on each side of an open breaker, so that the two can be reconnected at any moment: a differential clock bias of about $0.56$\,ms can make the controller hold the generator $10^\circ$ out of phase while it reports perfect synchronism. We cast the defense as a zero-sum inspection game. The defender, as inspector, chooses its timing monitors and audits the clocks periodically; the attacker, as inspectee, drags the two clocks to an unsafe bias by the next audit; the payoff is the evidence, the Kullback--Leibler divergence between the probability laws of the defender's record under attack and under normal operation. Our central result is that the evidence equals the cost of the cheapest innocent explanation of the attacked record. This single identity predicts which monitors work, turns the attacker's best response into a strictly convex quadratic program, and turns the defender's design into a convex program with a saddle point. Pairing explanations, which bound the evidence from above, with test statistics, which bound it from below, certifies audit periods in which every successful attack leaves at least a threshold of evidence. On a 50\,Hz island, these evidence-based periods, at the matched-test threshold, are up to 8.4 times as long as those of a rule that checks each monitor separately, and the case study favors placing the better oscillator on the utility side of the breaker.

0.26

Extending Grid Foundation Models for Phasor-Domain Dynamic Simulation

eess.SY cs.LG ·2026-11-06 ·Boyang Zhao, Yifan Zhou, Alban Puech, Tilman Bockhacker, Hendrik Hamann
Abstract

Grid Foundation Models (GridFMs) are emerging as a powerful framework for learning and transferring knowledge across power grids and grid analysis tasks. This work extends GridFMs from steady-state analysis to dynamic simulation. We present DynGridFM, which uses component-wise tokenization and a structured latent operator to learn shared dynamic representations across heterogeneous grids and enable transfer to unseen systems. Pretrained on nine grids, DynGridFM is adapted to a held-out 48-machine grid, where fine-tuning on 64 trajectories achieves accuracy comparable to training from scratch on 128. On a single GPU, it achieves 924x to 1232x speedup over numerical simulators on one CPU core.

0.26

Wellness Robots and the Path to Full Autonomy

cs.RO ·2026-11-06 ·Mohammad H. Mahoor, Matthew Hessler, Lombe Chileshe, Ola Alsaadi, Eshrat Emamian +2 more
Abstract

This paper introduces the autonomous wellness robot as a distinct category of socially assistive system, defined by its capacity to continuously assess and respond across the seven ICAA dimensions of wellness (see Figure 2). We propose the Care Robot Autonomy Scale (CRAS): a six-level taxonomy (L0-L5) modeled on the automotive automation standard SAE J3016 and evaluated across four wellness-aligned dimensions. We present Ryan, developed initially at the University of Denver and commercialized by DreamFace Technologies, as a case study. NIH/NIA-funded longitudinal evidence (N = 18; 887 hours) shows statistically significant cognitive improvement (SLUMS +2.61, p = 0.015) and a clinically meaningful directional reduction in depressive symptoms (PHQ-9 -2.50, p = 0.073), with no novelty effect. We position Ryan today at CRAS Level 1-2, with emerging Level 3 capabilities in some dimensions, and outline a three-phase roadmap to L5 (full autonomy) by the early 2030s.

0.26

Beyond Semantic Classes: Online Physical Parameter Estimation via Embedding-Conditioned Gaussian Mixtures

cs.RO ·2026-11-06 ·Alexander Borca-Tasciuc, Prayag Sharma, Franck Djeumou
Abstract

Enabling robots to infer interaction-relevant physical properties from vision before contact would improve planning and control in unstructured environments. Vision provides useful pre-contact information, but appearance alone is often ambiguous: visually similar regions can induce different friction, stiffness, or sinkage responses upon contact. Existing approaches often collapse this ambiguity to a point estimate or aggregate experience through fixed semantic classes, limiting their ability to preserve distinct physical regimes. Instead, we propose a real-time estimator that replaces semantic class indexing with an interaction-grounded Gaussian-mixture posterior over latent physical parameters, conditioned directly on continuous visual embeddings. Our approach maintains interaction-derived Gaussian parameter estimates, typically from inverse modeling, as mixture components while using embedding similarity to retrieve and weight relevant past interactions for each new query region. A support-aware fusion mechanism incorporates a conditional generative prior model when interaction history is sparse, while a learned embedding refinement improves retrieval relevance without discretizing the representation. Our experiments show that the estimator recovers multimodal physical beliefs, rejects unrelated interaction history, avoids spatial map discretization, and evaluates at controller rates. In a high-fidelity autonomous racing simulation with unknown low- and high-friction surface regions, the posterior enables the controller to adapt over repeated laps as interaction evidence accumulates, improving performance while maintaining friction-aware behavior.

0.25

Capacity of Multi-Layer Manhattan-Grid UAV Corridors Served by Terrestrial 5G NR Networks

eess.SP cs.PF ·2026-11-06 ·Saad Masrur, M M Sadman Shafi, Md Sharif Hossen, Ismail Guvenc
Abstract · Under Review

Advanced air mobility (AAM) is expected to organize unmanned aerial vehicle (UAV) traffic into corridors at multiple altitudes. In this paper, we study the downlink (DL) and uplink (UL) capacity of a multi-layer Manhattan-grid corridor mesh served by a terrestrial fifth-generation (5G) new radio (NR) macro network. The corridor spacing and the numbers of layers, UAVs, and base stations are configurable. We combine the 3GPP aerial channel and antenna models, correlated UAV mobility, and frequency-selective fading with an NR physical-layer abstraction that includes channel aging and link adaptation. Results show that UAVs above 90 m experience a DL signal-to-interference-plus-noise ratio (SINR) about 17 dB below that of ground users, and that a dense macro network with $10^\circ$ downtilt leaves 17-29% of corridor positions without DL service under full-buffer load, regardless of corridor spacing. On a common layout, the carrier frequency (700 MHz, 3.5 GHz, or 7.2 GHz) changes the UAV SINR by less than 0.5 dB, since the corridor links are strongly interference-limited. The UL is the bottleneck: at 85% satisfaction, a 20 MHz carrier supports only about 3.4-4.1 UAVs/km$^2$ with 4 Mb/s of UL video per UAV.

0.25

Mechanical Interaction Sensing for Robotic Kiwifruit Handling with an Optoelectronic Soft Suction Cup

cs.RO ·2026-11-06 ·Romina Adhami, Ebrahim Shahabi, Alessio Mondini, Samuel M. Youssef, Cosimo Della Santina +2 more
Abstract

Robotic handling of delicate fruit benefits from compliant contact and quantitative sensing of the mechanical interaction at the end-effector. This study investigates an optoelectronic soft suction cup with a kiwifruit-specific bicomponent design for estimating interaction forces during contact with real fruit. During known-contact intervals, the optical signals accurately estimate interaction force on fruits excluded from model training: the best shallow network achieves grouped out-of-fold pooled R-squared = 0.968 +/- 0.005 (MAE = 0.138 N), with a median per-fruit held-out R-squared = 0.973 (MAE = 0.134 N). A simple monotonic mapping from the mean optical response alone achieves R-squared = 0.925 +/- 0.005, showing that substantial force-related information is already present in the optical response. Leave-one-collection-day-out validation remains strong (R-squared = 0.955 +/- 0.031), supporting robustness across acquisition sessions. As an exploratory secondary analysis, binary Soft/Firm classification from compression-test reference labels achieves moderate balanced accuracy (approximately 0.705) under grouped held-out-fruit validation, while finer three-class grading shows limited performance.

0.25

It Takes Two to Tangram: Multi-Agent Object Centric Reinforcement Learning for Assembly

cs.RO ·2026-11-06 ·Paul Knaußt, Kay Pompetzki, Nico Bohlinger, Niklas Funk, Cristiana de Farias +2 more
code/data
Abstract · This work has been submitted to the IEEE for possible publication

Geometric assembly requires fine spatiotemporal coordination to place parts without causing collisions or deadlocks under tight clearances. To address this, we present an object-centric reinforcement learning framework that treats each part as an agent and learns a shared policy for continuous assembly control. A graph neural network represents the pieces, their target poses, and their spatial relationships, and predicts translational and angular velocities for each component. Per-agent proximal policy optimization with a centralized graph critic directly learns motion and temporal coordination within a single feedback policy. We study this formulation through planar Tangram assembly, where pieces must slide and rotate into a given configuration while remaining on the assembly surface. Procedural generation of target configurations and randomized initial arrangements support training and evaluation across diverse assembly problems. Our experiments isolate the roles of relational structure and multi-agent optimization, and compare learned control with sampling-based planning. We demonstrate concurrent assembly in simulation and transfer the policy to sequential execution on a real robot, as our object-centric formulation decouples learned assembly coordination from low-level physical control. Project page: https://tangrambench.org/

0.25

DKWBC: Decoupling Motion Tracking and Balancing for Humanoid Whole-Body Control

cs.RO ·2026-11-06 ·Dingkun Wei, Zehong Shen, Huaijin Pi, Xueyang Wang, Xing Zhu +3 more
code/data
Abstract · 9 pages, 6 figures, 4 tables

Humanoid whole-body control requires a robot to track diverse reference motions while maintaining balance under gravity, torque limits, contact switching, and external disturbances. Existing reinforcement-learning-based trackers typically learn control outputs that jointly address motion tracking and balance. However, similar reference poses can correspond to substantially different dynamic demands, requiring a unified policy to learn motion-dependent control requirements alongside balance corrections. This can complicate learning and lead to high-frequency fluctuations in control outputs. We propose DKWBC, a dynamics-informed whole-body control framework that combines explicit model-based tracking with a learned balancing residual. Inspired by the model-based and environmental-wrench contributions to joint torques in floating-base dynamics, DKWBC uses inverse-dynamics feedforward and joint PD feedback to supply tracking torques, while a residual policy learns additional joint-torque corrections for balanced execution. By explicitly accounting for motion-dependent torque demands, this structure simplifies the policy's learning task. Experiments in simulation and on real hardware demonstrate that a single learned balancing policy generalizes across motion datasets while maintaining accurate motion tracking and balance across ground support, attached payloads, and physical interaction. Project page: https://robbyant-research.github.io/DKWBC/.

0.24

SUGARv2: Learning Diverse Visual Humanoid Loco-Manipulation Skills from Synthesized Interaction References

cs.RO ·2026-11-06 ·Tianshu Wu, Tianxiang Gui, Yue Chen, Tingxu Liu, Ziye Peng +5 more
code/data
Abstract · 11 pages, 6 figures. Project page: https://tianshuwu.github.io/sugarv2-humanoid/

Scaling up visual humanoid loco-manipulation skill acquisition requires reducing the human effort needed as tasks and spatial configurations expand, particularly in demonstration collection and task-specific training design. We present SUGARv2, a unified framework for learning diverse visual humanoid loco-manipulation skills through interaction reference synthesis and a shared training pipeline. To obtain diverse references at low cost, we adapt reusable task templates to different spatial configurations, synthesizing coordinated robot and object motions. To reduce task-specific reward design and training effort, we use reference-based imitation as a common learning objective across tasks. We first train privileged teachers to execute these references, then train visual policies through teacher distillation and reinforcement learning with the same reference-based mimic rewards, using diverse object instances and visual domain randomization to support visual generalization and sim-to-real transfer. Across five humanoid loco-manipulation tasks, SUGARv2 achieves a mean success rate of 91.0% in simulation. Deployment on a real humanoid robot further demonstrates successful sim-to-real transfer for whole-body manipulation. These results demonstrate the framework's potential for learning diverse visual humanoid loco-manipulation skills.

0.24

Sample-Covariance Regularization for Data-Driven Predictive Control from Frequency Response Functions under Colored Noise

eess.SY ·2026-11-06 ·M. Wind, T. J. Meijer, V. S. Dolk, W. P. M. H. Heemels
Abstract

In this work, we investigate how sample covariance obtained from frequency-response functions (FRFs) can be exploited to improve performance of data-driven frequency-domain predictive control (FreePC) in the presence of colored measurement noise. In case of periodic multisine excitation, this covariance is obtained directly from repeated periods and provides a frequency-dependent measure of uncertainty without requiring an additional parametric noise model. To incorporate this information into both indirect and direct predictive control, we develop a frequency-domain subspace predictive controller (FD-SPC) and establish its equivalence to the direct FreePC formulation in the noiseless setting. Regularization terms for both formulations are then motivated by expected prediction-error criteria and constructed from the sample covariance of the FRF estimates. Across all tested noise models and data sizes, regularization consistently reduces the median realized control cost relative to the corresponding controller without covariance regularization, with reductions of up to 35%.

0.24

Dynamic Traffic Modeling using Equivalent Circuit Models

eess.SY ·2026-11-06 ·Joseph Moyalan, Ricardo de Castro, Iman Ebrahimi, Xuchang Tang, Xinfan Lin
Abstract

This paper presents an equivalent circuit model (ECM) for dynamic traffic modeling by exploiting analogies between electrical and transportation variables. Vehicle concentration is represented by capacitor voltage, while traffic flow is modeled using controllable current sources. The proposed formulation is shown to preserve the conservation and sending/receiving relationships of the spatially discretized Lighthill Whitham Richards traffic model. The framework is extended to restricted inflow, traffic signals, and two road merging and diverging scenarios. Validation against microscopic Simulation of Urban MObility (SUMO) simulations, involving traffic signals, two-road merging, and single-road diverging scenarios, demonstrates that the proposed ECM successfully reproduces the principal macroscopic traffic dynamics, including queue formation, backward congestion propagation, vehicle accumulation, saturation, and subsequent traffic recovery.

0.24

On Agentic Manipulation Policies

cs.RO cs.AIcs.CVcs.LG ·2026-11-06 ·Chen Chen, Omid Bagherieh, Naveen Nikamanth Appiah Balaji, Jesse T. Gonzalez, Negin Heravi +8 more
code/data
Abstract · Videos and code available at: https://apple-aiml-research.github.io/ml-amp/

Recent work on robot manipulation has focused on specialist policies trained on demonstrations. We show that generalist AI models can control manipulators more reliably than specialist policies such as vision-language-action models (VLAs) and world-action models (WAMs). We systematically analyze agentic manipulation policies (AMPs), in which a generalist AI model commands the manipulator directly, adapting its actions as the task unfolds. We present AMP1, a reference harness that runs unchanged across single-arm, bimanual, and mobile manipulation. AMP1 achieves state-of-the-art results on the RoboLab-120 and CaP-Gym benchmarks, with no demonstrations or task-specific training. On RoboLab-120, AMP1 reaches an 85.6\% success rate zero-shot, compared to 49.2\% and 42.9\% for leading VLA and WAM policies. The agent succeeds zero-shot in 104 of 108 real-world trials spanning twelve tasks and three robot instances, reducing failure rates by 20x and 15x relative to strong VLA and WAM baselines. The basic harness supplies no grasping, pushing, or placing skills; the agent develops its own strategies, such as searching for hidden objects, using objects as tools, and calibrating its cameras. With the addition of a persistent memory store, the agent can also learn from experience: an augmented harness with a memory representation that is carried across episodes lowers the failure rate on RoboDojo bimanual manipulation tasks by 4.6x relative to zero-shot deployment. Our analysis suggests that AMPs are substantially more general and robust than specialist manipulation policies, providing a promising foundation for robot manipulation.

0.23

OfficeTwin: Realizing 3D Reconstruction of Redundant Scenes through Partial Viewpoint Perception

cs.RO cs.CV ·2026-11-06 ·Quanyun Wu, Kyle Gao, Daisy Li, Kshitij Goyal, Dayou Mao +3 more
Abstract

Feed-forward 3D reconstruction turns an uncalibrated monocular stream into a dense point cloud in a single inference pass, orders of magnitude faster than classical structure-from-motion (e.g., COLMAP), and reliable on the weakly textured, repetitive surfaces where keypoint matching fails outright. However, raw point clouds lack the object identity, physical properties, and watertight volume required for robot learning. This gap widens under robot-native capture. A camera mounted on a robot constrained to a few degrees of freedom sees each object only partially, and the repetitive, weakly textured structure of an office makes distinct workstations nearly indistinguishable, leaving redundant and partial observations that downstream reconstruction cannot resolve. Existing image-to-3D generation methods struggle with these conditions, hallucinating geometry under incomplete views and confusing repeated instances. We present OfficeTwin, a robot-native framework that converts these incomplete observations into simulation-ready digital twins. OfficeTwin first imposes metric and structural constraints, then exploits the office's structural redundancy by registering repeated cubicles into a shared canonical frame, where complementary observations recover geometry occluded in any single view without hallucination. On a real office scan, OfficeTwin recovers every cubicle as an individually watertight, metric solid, outperforming feed-forward reconstruction and generative baselines that fail to produce simulation-ready assets.

0.23

Stochastic High-Resolution Parameter-Based Channel Modeling for Sub-Terahertz Outdoor Device-to-Device Propagation at 145 GHz

eess.SP ·2026-11-06 ·Zihang Cheng, Naveed A. Abbasi, Jorge Gomez-Ponce, Jianzhong, Zhang +1 more
published
Abstract · Currently under submitted and under revision in the IEEE Transactions on Wireless Communications

Since device-to-device (D2D) systems operating in the sub-THz regime are of interest for a variety of 6G applications, modeling their propagation channels is therefore essential. This paper presents high-resolution parameter estimation (HRPE) of specular multipath components (MPCs) as well as diffuse multipath components (DMCs) from an extensive measurement campaign. The extracted MPCs and DMCs are then clustered, and statistical models for both intra- and inter-cluster statistics are established. Among other insights, we find a large number (typically 10-50) of clusters; both the powers and densities of MPCs within each cluster show significant variations. DMC occurs both when associated with discrete MPCs and standalone. The paper provides a complete model that can generate double-directional impulse responses, and a MATLAB program implementing it is made available. Extensive verifications of "derived" (not directly modeled) parameters are provided as well.

0.23

BATCH: Behavior Alignment for Cross-Embodiment Learning from Humans

cs.RO cs.AIcs.CV ·2026-11-06 ·Zeyi Liu, Shuran Song
Abstract · Project website: https://batch-embodiment.github.io/

Human demonstrations provide a scalable source of diverse manipulation data, but leveraging them for robot learning remains challenging due to differences in embodiment, viewpoint, action spaces, and temporal structure. We introduce BATCH, a cross-embodiment learning framework that explicitly aligns human and robot behaviors. BATCH learns self-supervised temporal representations with a cycle-consistency objective to establish cross-embodiment correspondences without manual annotation. We further introduce two complementary training objectives that leverage these correspondences within an autoregressive videoaction diffusion model: behavior-aligned co-training and human-to-robot style transfer. On real-world manipulation tasks, BATCH improves generalization to unseen task configurations compared with training on unaligned data and prior cross-embodiment baselines, with additional gains from the style-transfer objective. Our results demonstrate that explicit behavior alignment enables more effective transfer from human demonstrations to robot policies.

0.22

Grid-Oriented Probabilistic Load Forecasting for AI Data Centers

eess.SY cs.LG ·2026-11-06 ·Qingyi Yao, Ding Lin, Tianqiao Zhao, Meng Yue, Jianhui Wang
benchmark
Abstract

Probabilistic load forecasts for AI data centers need to represent both uncertainty under changing workloads and temporal dependence to support grid operations. This paper develops an open, grid-oriented benchmark using measured GPU power from a 448-GPU AI cluster. Forecasts are evaluated through four stages: probabilistic prediction, online uncertainty calibration, temporal scenario generation, and horizon-wide reserve sizing. Following a change in cluster utilization, a quantile LightGBM forecast achieves only 63.4% coverage for a nominal 90% interval, while online regime-conditioned calibration increases coverage to 87.3%. Even with well-calibrated marginal forecasts, independent sampling produces unrealistic ramp distributions and excessive reserve procurement. Reconstructing temporal dependence from historical forecast errors reduces the discrepancy in ramp statistics and lowers realized reserve cost by 7.4-10.6% across three forecasting tasks. These findings show that marginal forecast quality alone is insufficient for operational decisions and motivate joint evaluation of calibration, temporal structure, and decision outcomes.

0.22

A Tutorial on Practical QKD Post-Processing Algorithms Using a GUI-Based Simulation Framework

eess.SP quant-ph ·2026-11-06 ·Ushnik Nath, Neha Sharma, Ramniwas Meena, Neel Kanth Kundu
benchmark
Abstract · Submitted to IEEE Open Journal of the Communications Society

Recent experimental demonstrations of Quantum Key Distribution (QKD) have attracted significant attention for their ability to establish information-theoretically secure encryption keys for next-generation communication systems. As practical deployment accelerates, the realization of compact, scalable, and high-throughput discrete-variable (DV) QKD systems has become increasingly important. While advances in photonic integration are enabling low-size, low-weight, and low-power QKD hardware, the security and performance of these systems rely heavily on efficient post-processing techniques to generate identical, secure keys. This tutorial presents a comprehensive algorithm-level review of practical post-processing techniques for DV-QKD, with particular emphasis on information reconciliation methods based on Cascade, Winnow, low-density parity-check (LDPC), and polar codes. We also analyze the theoretical asymptotic and finite-size secret key rates (SKRs) of DV-QKD and benchmark these bounds against the practical SKRs achieved using actual post-processing algorithms. Furthermore, we present a server-client model to implement the entire QKD post-processing pipeline. The performance is evaluated in terms of QBER and SKR across various system parameters, considering the complete post-processing pipeline, including sifting, information reconciliation, privacy amplification, and authentication. This tutorial also presents a graphical user interface (GUI) for the QKD post-processing methods, enabling researchers, industry practitioners, and academic users to visualize and analyze the impact of different post-processing techniques and system parameters, such as QBER, on the SKR performance of DV-QKD systems.

0.22

Control of Linear Dynamical Systems with Operational Constraints

eess.SY ·2026-11-06 ·Marcel Menner, Eugene Lavretsky
theory
Abstract

This paper presents a method for designing servo-controllers that enforce operational constraints on multi-input multi-output linear time-invariant dynamical systems. The proposed method introduces a control augmentation that complements a baseline controller to ensure forward invariance of state/output trajectories within prescribed box constraints. The derivations of the control design are based on the Nagumo Theorem, the Comparison Lemma, and principles from min-norm optimal control akin to the theory of Control Barrier Functions. Importantly, an explicit piecewise-linear feedback control law is derived, which provides continuous enforcement of constraints without online optimization, enabling rigorous stability and robustness analysis via classical linear systems theory. Extensions to handle estimation errors in output feedback and integral action in proportional-integral control are provided. Theoretical results are validated through flight control simulation studies that demonstrate input/output constraint satisfaction, robustness to disturbances, and closed-loop stability. The proposed method offers a mathematically rigorous, practical, and analyzable alternative to common heuristic or optimization-intensive constrained control strategies.

0.22

Lightweight Path-Aware Frontier Re-ranking for LiDAR-Based UAV Exploration

cs.RO ·2026-11-06 ·Duc Manh Do, Thanh Nguyen Canh, Ngoc Minh Do, Mai Thanh Thai, Xiem HoangVan
published
Abstract · Accepted for presentation at RIVF 2026 (20th International Conference on Computing and Communication Technologies)

Autonomous 3D exploration with UAVs requires efficient target selection under limited flight time and onboard computation. In conventional Euclidean-distance-based frontier pipelines, geometric distance may poorly represent the actual collision-free travel cost in cluttered environments, leading to inefficient frontier choices. This paper presents a lightweight path-aware frontier re-ranking method for LiDAR-based UAV exploration. The method first retains a small top-N set of promising frontier candidates using a lightweight utility function and then applies time-bounded RRT* planning to estimate their feasible path costs for re-ranking. The path generated for the selected frontier is further reused for navigation to reduce redundant global planning. Experiments in three warehouse-like environments show that the proposed method consistently produces shorter exploration trajectories than conventional frontier baselines, with traveled-distance reductions of up to 14.1%, while reducing exploration time by up to 9.5% in two environments and remaining competitive in the third. The method maintains map detail comparable to the conventional OctoMap-based baseline, demonstrating an effective trade-off between exploration efficiency and mapping quality.

0.22

HomeLoop: A Simulation Framework and Benchmark for Robot-IoT Coordination in Everyday Domestic Assistance and Safety

cs.RO cs.AI ·2026-11-06 ·Ramy ElMallah, Marc Bishara, Hyegang Kim, Yuxi Zhang, Chongyu Zhu +1 more
benchmark
Abstract

Recent progress in embodied robotics is bringing robots closer to deployment in homes, as vision-language models (VLMs) are demonstrating significant advances in reasoning, making VLM-driven robots promising home assistant agents. These agents will often be introduced into homes that already have cameras, Internet of Things (IoT) devices, and residents with wearables. To fulfill their potential, they need to integrate with existing smart infrastructure and reason through ambiguous scenarios: a smoke alarm may come from a pan about to ignite, a pan already on fire, a resident cooking or a faulty sensor, and each calls for a different response. To that end, we introduce HomeLoop, a simulation framework and benchmark built on Isaac Sim that combines robot bodies, networked devices and wearables, simulated residents and visitors, faulty sensors, and hazards that evolve over time. Our benchmark is composed of 20 tasks across five families grounded in real-world household hazards and needs, and designed so that the same alert can have different causes. A runtime monitor scores each episode with temporal-logic checks on the final outcome and the actions along the way, and every action carries a predefined cost. We evaluate six open-weight and proprietary VLMs. The best achieves a family-balanced score of 61.9%, and across models we see a pattern of over-reaction: failed episodes involving prohibited interventions outnumber those involving missed required interventions by more than two to one, with models often shutting off the stove before investigating the alarm's cause. HomeLoop is an extensible foundation intended for use in research on robot-IoT coordination, long-term memory, proactive assistance, and privacy-safety trade-offs. HomeLoop will be released with 2,100 home layouts that can be automatically furnished in diverse configurations, five robot bodies and 20 networked device types.

0.22

Seeing What Execution Needs: Benchmarking Task-Conditioned Actionable Spatial Belief for Manipulation

cs.RO cs.CV ·2026-11-06 ·Yina Jian, Mingkun Xu
benchmark
Abstract · 22 pages (11 main + 11 supplement). Extended cross-domain evaluation and physical robot validation in progress

When should a robot look at the cup, and when at the portafilter? A view that is informative for one task can be useless for another, even at the same workstation. We introduce Seeing What Execution Needs, a benchmark for task-conditioned actionable spatial belief in simulated tabletop manipulation. Its aligned three-domain track contains 1,080 frozen cases (360 each for coffee workstation, drawer/cabinet, and microwave/appliance). The primary coffee protocols -- closed-loop actionability and same-object functional out-of-distribution generalization -- produce over 115,000 evaluator episodes across evaluated methods, budgets, and inference repeats. Each case pairs candidate RGB views with a manipulation command, shared scripted skills, and hidden postconditions, directly measuring whether selected observations support execution. A metadata-assisted diagnostic reference pipeline maintains spatial-functional beliefs and ranks views by missing evidence. At primary budget settings, it reaches 98.8% closed-loop and 98.9% functional-OOD success; removing functional fields or command binding reduces functional-OOD success to 0.0% and 19.7%, respectively. Under a separate full-preview contract, seven vision-language models reach 26-43%. Replacing coverage metadata with frozen RGB-derived scores retains 98.5%/98.9% success; shuffling these scores reduces success to 15%/11%, supporting dependence on image content. Separately, the aligned three-domain evaluation at H=3 supports portability of the benchmark interface under frozen domain-specific schemas. These results suggest that spatial belief for manipulation should be evaluated by whether acquired evidence meets the current command's execution requirements.

0.22

TPS-A*: Path Planning for Any-Shape Vehicles with Certified Collision Checking in Trajectory Parameter Space

cs.RO ·2026-11-06 ·Jose Luis Blanco-Claraco, Shravan S. Rai, Fernando Cañadas-Aránega, José Luis Torres-Moreno, José Carlos Moreno-Úbeda
theory
Abstract · 12 pages, 6 figures

Global planners for ground vehicles usually approximate the vehicle by a disk, trading safety against conservativeness, while exact-footprint checking pays for every candidate motion, pose, and footprint edge. We present TPS-A*, a lattice planner whose edges are kinematically feasible trajectories from parameterized trajectory generators, built on certified collision checking in trajectory parameter space. A grid precomputed per footprint returns, in one pass over the local obstacles, a certified free distance along every candidate trajectory, so the per-node collision cost is independent of footprint complexity and trajectory count. We prove that this grid never overestimates the free distance of the continuously swept, possibly non-convex footprint, and state the planner's guarantees: collision-free paths and bounded suboptimality relative to its lattice abstraction. Goals with or without heading are reached exactly by Reeds-Shepp shots certified as trajectories of the same generators. An exact independent test finds no collision in its paths. On 300 cluttered worlds, with the true footprint and the same goal pose for all planners, it solves every query with a collision-free goal pose in a median of three milliseconds on an AMD Ryzen 7 2700X, on par with hybrid A*, 45\% of whose paths overlap obstacles, and nearly sixty times faster than a state-lattice planner, the only other collision-free one; with position goals, eighty-nine times faster than hybrid A*. It matches hybrid A* in automated parking and on narrow doorways, docks a non-convex fork footprint that no convex model admits, and drove a real robot across a university campus among people.

0.22

A Unified Inertial Encoder for Human and Humanoid Motion

cs.RO ·2026-11-06 ·Yutian Chen, Shibo Zhao, Rocky Duan, Karen Liu, Carmelo Sferrazza +3 more
Abstract

Learning humanoid state estimation and language-grounded motion understanding from robot recordings ties training coverage to the robot's motion repertoire. Human motion datasets offer broader coverage, yet transferring this knowledge requires overcoming differences in body geometry, sensing, and physical disturbances. To enable this transfer, we propose a unified inertial encoder whose shared inertial interface uses kinematic IMU synthesis to represent human and humanoid motion at corresponding body sites. Within this common input space, joint training on human and kinematically retargeted humanoid motion combines velocity estimation and motion--text alignment with continuous-time encoding and sensor and motion randomization. The resulting model transfers zero-shot to a physical Unitree G1 without physics simulation, real-robot training trajectories, or fine-tuning, achieving the lowest velocity error among evaluated methods across five datasets, including 0.099m/s on real G1. Its motion-description retrieval also extends to retargeted G1 motion categories observed only in human training, demonstrating how human motion knowledge can support the generalization of humanoid inertial odometry and language-grounded motion understanding.

0.21

Mobility Information Capacity of Multi-Robot Systems: Safety-Aware Maneuver Design

cs.RO cs.ITeess.SY ·2026-11-06 ·Weijie Yuan
benchmark
Abstract · 12-page main article and 4-page supplementary material

Multiple robots operating in shared workspaces must adjust their speed, direction, and trajectories to coordinate with one another and respond to changing tasks and disturbances. The available maneuver freedom depends on the dynamics, uncertainty, and safety margins, even for a fixed fleet size and workspace. However, robot count, workspace occupancy, and task throughput do not quantify how much the fleet can vary its motion or how reliably these variations can be observed. We develop a mobility information capacity (MIC) framework that jointly characterizes these two aspects. Mobility information is defined as the mutual information between a joint maneuver sequence and its noisy trajectory observation, and MIC maximizes this quantity under maneuver-resource constraints. For a linear Gaussian maneuver-to-observation model with feedback dynamics, we characterize the finite-horizon information benchmark and derive its stationary long-horizon spectral limit, accounting for heterogeneous mobility gains, colored uncertainty, and temporally coupled maneuver costs. We further map workspace-boundary, inter-robot separation, and actuator chance constraints into coupled maneuver-covariance restrictions and derive conditions under which covariance optimization reduces to resource allocation across mobility modes. Finite joint maneuver codebooks are then constructed through individual risk screening and noise-whitened trajectory separation, with bounds linking sequence recovery error and achieved information to the capacity benchmarks. An aerial-robot simulation study evaluates spectral convergence, safety-aware allocation, finite-codebook identification, and fleet-size scaling. The resulting framework complements occupancy and task-throughput measures by quantifying the maximum observable maneuver freedom supported by a mobility system.

0.21

EGT: Execution-Grounded Traversability through Learned Closed-Loop Trajectory Distributions

cs.RO ·2026-11-06 ·Chinmay Mundane, Amith Manoharan, Arun Kumar Singh
Abstract

Rough terrain traversability is often estimated through point-wise roll, pitch, or force-angle margins along a planned path. However, these metrics are often inaccurate and overly conservative because they ignore how terrain disturbances and low-level controller corrections can affect the vehicle's motions. To overcome these limitations, we introduce Execution-Grounded Traversability (EGT), defined as the probability that a vehicle-controller pair can execute a candidate path while meeting trajectory-level tracking and progress requirements. We estimate EGT by predicting the distribution of closed-loop trajectories using a latent-space flow matching (FM) model conditioned on multi-resolution terrain features and 3D vehicle attitude and wheel-ground contact points prior. Across 650 simulated trials on unseen terrains, EGT reduces safe-path rejection to 2.2% (versus 11.4%-39.5% for pose-margin and force-angle baselines) and unsafe acceptance to 7.7%. Integrated into a receding-horizon sampling-based planner, EGT reduces navigation failure rates by 24-42 and 30-44 percentage points over [1] and [2], respectively, while cutting goal-reaching time by up to 20.5% and 45%. Finally, we demonstrate zero-shot transfer of our simulation-trained model to real-world outdoor experiments.

0.21

FASTFUNNELS: Agile Multi Robot Navigation in Narrow Spaces via Adaptive Dynamic Funnels

cs.RO ·2026-11-06 ·Vignesh Rajagopal, Rohan Chandra
Abstract · 8 pages, 4 figures, 1 table

We focus on the problem of high-speed multi-robot navigation in cluttered spaces with tight spatial constraints. A key challenge is balancing safety with agility, as existing methods often trade one for the other (slowing down near a bottleneck to avoid collisions versus failing to swerve in time at high speeds), especially for car-like systems. Our approach belongs to the family of formation control methods, in which Ackermann-steering agents coordinate while navigating tight spaces. However, current formation control techniques rely on rigid, pre-defined shapes and transitions that depend on initial conditions, the environment, and vehicle dynamics, and are impractical to maintain at high speeds. For example, agents passing through a small gap must first form a square, transform into a line using analytical equations, and then return to a square. In contrast, our approach is inspired by a droplet of water that continuously adapts its shape to the space it flows through without slowing down. We propose a hierarchical algorithm in which, at the high level, a reinforcement learning policy produces a dynamic, blob-like elliptical patch that morphs its shape in real time, squeezing, stretching, and orienting itself around obstacles and through tight spaces. At the low level, the robots are controlled via distributed model predictive control (DMPC) to stay inside the patch, which acts as a soft forcing function, while avoiding collisions with each other. We evaluate our approach, FastFunnels, on simulated F1TENTH racing cars across 14 scenarios with N=1 to N=6 robots, averaged over 20 trials each. FastFunnels achieves an 89% completion rate at 7.73 m/s, while all baselines either fail to complete the track or succeed only at very low speeds (0.07-1.79 m/s, about 40% completion rate).

0.21

DASH-SLAM: Calibration-Free Metric Monocular SLAM via Depth Scale Propagation

cs.RO ·2026-11-06 ·Otavio Ribas, Jacopo Panerati, Pierre-Yves Lajoie
Abstract

Geometric foundation models estimate dense geometry and camera parameters from monocular images, but reliable SLAM also requires consistent scale and accurate data association across successive reconstructions. We present DASH-SLAM, a monocular SLAM system that integrates metric depth predictions from pretrained Depth Anything 3 into a shared-keyframe pose graph without task-specific fine-tuning. DASH-SLAM derives scale corrections from depth ratios at shared anchor images and accumulates them across reconstructions, with a damping factor controlling their influence. These corrections rescale odometry translations, loop constraints, and map geometry. Experiments on TUM, Replica, and 7-Scenes show that DASH-SLAM outperforms state-of-the-art methods. We evaluate performance using both rigid and similarity alignment to distinguish metric-scale fidelity from accuracy up to scale. We further evaluate DASH-SLAM on four challenging sequences in perceptually-degraded environments from the Unified Autonomy Stack field dataset, covering tunnels, fog, and a frozen lake. On these sequences, DASH-SLAM outperforms the foundation-model SLAM state-of-the-art and maintains tracking where classical SLAM methods fail.

0.21

DNN-Driven Proactive UAV Trajectory and Beam Steering for Blockage-Resilient Communications

eess.SP ·2026-11-06 ·Abdallah S. Ghazy, Hussein A. Ammar, James Bayes, Francois Chan
Abstract

This paper proposes Autonomous UAV Communication (AutUAVCom), a proactive communication system that jointly adapts the flight trajectory and RF beam steering to minimize UAV energy consumption while satisfying secrecy and reliability requirements for rescue teams operating in dynamic, cluttered, and GPS-denied environments. Unlike existing modelbased proactive approaches, AutUAVCom employs a deep neural network-based Tracking and Prediction (T&P) module that combines onboard sensing with spatiotemporal and body-orientation information to track users, localize blocking objects, and predict future trajectories and potential beam blockages. Building on the T&P module, we formulate a constrained optimization problem and develop a two-step iterative numerical approach. Monte Carlo simulations show that, relative to a reactive system, AutUAVCom achieves 10% and 19% gains in energy efficiency and channel secrecy, respectively, while reducing the outage probability by 85%, under a mean blockage arrival interval of 120 seconds and a mean blockage duration of 20 seconds.

0.20

Sparse Array-Inspired Frame Structure Design for 5G NR ISAC

eess.SP ·2026-11-06 ·Haisu Wu, Hong Ren, Cunhua Pan, Yuqing Chen, Ruizhe Wang +3 more
Abstract · 16 pages, 7 figures, 2 tables

Integrated sensing and communication (ISAC) has emerged as a promising paradigm for future wireless networks. However, the extensive occupation of time-frequency resources for sensing limits the communication capacity. To address the resource conflict, this paper proposes a novel sparse sensing frame structure aligned with the 5G new radio (5G NR) resource grid at the slot and physical resource block (PRB) levels through a predefined sensing-pattern block. Specifically, inspired by the design of two-dimensional (2D) sparse antenna arrays, we map the sparse spatial geometries onto the time-frequency resource grid. By leveraging the hole-free property of the difference coarray, we eliminate the range-velocity ambiguity and cast the sensing signals into a reconstructed higher-order tensor representation. For efficient parameter estimation, we propose a novel binary-tree-based tensor train decomposition (BT-TTD) algorithm that extracts lower-dimensional tensor train cores (TT-cores) and executes a Vandermonde-prior-guided rectified alternating least squares (RecALS) algorithm. Simulation results validate that the proposed scheme achieves competitive estimation accuracy with reduced complexity. Furthermore, we investigate the topological properties of various sparse patterns, revealing the inherent trade-off among sensing accuracy, resource occupancy, and computational complexity.

0.19

CucumberVision at the Edge: Accuracy, Throughput, and Energy of RGB-D Cucumber Sizing on Jetson Orin Nano and Raspberry Pi 4

cs.RO ·2026-11-06 ·Manveen Kaur, Rajmeet Singh, Saeed Mozaffari, Majid Ahmadi, Shahpour Alirezaee
Abstract

Non-contact fruit sizing from RGB-D imagery reaches sub-centimetre accuracy on workstation hardware, but in-field grading requires the same pipeline to run on a low-power single-board computer, and the cost of that transition is rarely reported. We deploy a curvature-aware cucumber length-measurement pipeline (YOLO26n detection, prompt-based segmentation and a medial arc spline estimator) on an NVIDIA Jetson Orin Nano and a Raspberry Pi 4, and characterize both against an RTX 3050 Ti workstation by replaying 50 identical greenhouse captures of 7 specimens, with specimen-level cluster-bootstrap statistics. Accuracy is preserved: mean absolute error (MAE) is 0.58 cm (95% CI 0.46-0.69) on both the workstation and the Jetson at matched precision, with a paired difference of zero. TensorRT FP16 compilation cuts Jetson detection latency 2.57x, giving two operating points on one board: 72 ms per fruit at 0.84 cm, or 2.74 s at 0.64 cm. Four ablations map the design space. For prompted segmentation backends, measurement accuracy is fixed by the backend and reproduces across platforms to within 0.01 cm, while its cost is fixed by the platform and varies 214x, so the two choices are separable. Whole-graph INT8 quantization of the NMS-free detection head fails structurally, returning no detections, because one output tensor carries both pixel coordinates and confidences; we show that no calibration strategy can avoid this. Excluding the head restores function at 1.37x. Energy per fruit is 42.2 J on the Jetson, and its 7 W mode consumes 73% more than its 15 W mode. The Raspberry Pi with MobileSAM costs 2.5x more per fruit than the Jetson with SAM despite drawing 47% less power, and it runs without thermal throttling under sustained load. The results define an accuracy-throughput-energy envelope with three non-dominated configurations and give rules for selecting among them.

0.19

Energy-Aware Vertically Layered Pinching-Antenna System Design with Probabilistic LoS Availability

eess.SP ·2026-11-06 ·Ruotong Zhao, Derrick Wing Kwan Ng
Abstract

This paper investigates a vertically layered pinching antenna (PA) system, in which dielectric waveguides deployed at different heights provide additional vertical spatial degrees of freedom for line-of-sight (LoS)-favorable PA placement. We consider a practical energy model that captures in-waveguide attenuation, base station-to-feed-point optical-fiber loss, and PA repositioning energy, along with a statistical LoS-availability model that depends on elevation and distance. We then jointly optimize the transmit beamforming, PA positions, time allocation, and transmission rates to maximize the expected successfully delivered frame-effective sum rate under per-user outage and quality-of-service constraints. To avoid enumerating the exponentially many LoS states, we derive a certified conditional-moment outage envelope for outage-guaranteed rate selection and a differentiable safe approximation of the chance constraints to facilitate tractable optimization. Building on these results, a block coordinate descent algorithm with a penalty-aided majorization-minimization procedure attains a high-quality suboptimal solution. Simulation results confirm the gains of vertical layering and reveal a fundamental energy-LoS tradeoff: LoS-favorable PA placement generally enhances both performance and reliability, although it incurs higher attenuation, path loss, and energy costs.

0.18

Controlled Excitation for Identification of Nonlinear Systems with Saturated Outputs

eess.SY math.OC ·2026-11-06 ·Adel Malik Annabi, Jean-Baptiste Pomet, Dario Prandi, Ludovic Sacchelli
Abstract

We consider nonlinear systems whose measured output is the product of an unknown matrix and a known saturation function of the unmeasured state. Generating excitation that allows the matrix to be recovered from the output is challenging when the dynamics are largely unknown. We exploit saturation to construct this excitation. Under a geometric accessibility condition, we derive explicit open-loop controls that drive the state into regions where the saturation takes prescribed values. For componentwise saturations, this condition holds generically with as few as two independent control directions. The controls require no state measurement and only partial knowledge of the dynamics, with no information on the structure of its nonlinear part. The resulting regressor is known in advance, and the unknown matrix is recovered exactly by a closed-form regression. Under stronger structural assumptions and with sufficiently many independent outputs, an output feedback tracks prescribed reference trajectories within the unsaturated region to arbitrary accuracy and permits exact state recovery. Suitably chosen references then also enable exact identification of linearly parameterized (nonlinear) dynamics. We analyze robustness to measurement noise and illustrate the method numerically in the presence of both measurement and dynamical noise.

0.17

Grasp-TAGS: Task-Aware Grasp Selection via Critic Distillation for Prehensile Loco-Manipulation

cs.RO cs.AIcs.LG ·2026-11-06 ·Xun Tu, Minghao Zou, Karthik Desingh
Abstract · 10 pages, 6 figures

Loco-manipulation, which requires mobile manipulators to coordinate locomotion and manipulation over long horizons, is a key capability for large-object rearrangement in applications such as service robotics and warehouse automation. Despite recent progress, reliably executing these tasks remains challenging because performance depends on complex robot-object and object-environment interaction dynamics. Existing approaches address this challenge either through model-based methods, which often require accurate dynamics models, or through learning-based methods, which reduce this requirement but typically assume a fixed or predefined initial grasp. However, different geometrically stable grasps can lead to substantially different downstream manipulation performance. Therefore, we propose Grasp-TAGS, a task-aware grasp selection framework for goal-conditioned large-object rearrangement. We train a unified PPO-based reinforcement learning policy across multiple initial grasp configurations and distill the learned critic into a lightweight grasp utility network that ranks candidate grasps using only the candidate grasp configuration and goal information. At inference time, we select the grasp with the highest predicted downstream utility for the downstream policy to complete the task, before even making a contact. Experiments demonstrate that Grasp-TAGS improves rearrangement success over both fixed single-grasp policies and random grasp selection, achieving an average success rate of 89.84% compared with 79.87% for the best-performing single-grasp policies, an improvement of 9.97 percentage points. Grasp-TAGS also reduces final orientation errors across all evaluated objects. For more qualitative results, refer to the supplementary webpage: https://rpm-lab-umn.github.io/Grasp-TAGS-web-page/

0.17

Capturability-Aware Footstep Planning in Cluttered Environments: A Recursively Feasible Sequential MPC on Axis-Aligned Safe Corridors

cs.RO ·2026-11-06 ·Athanasios Kyriakopoulos, Shubham S. Kumbhar, Kunal S. Narkhede, Abhijeet M. Kulkarni, Ioannis Poulakakis
published
Abstract · Accepted for publication in 2026 IEEE-RAS 25th International Conference on Humanoid Robots (Humanoids)

A fundamental aspect of humanoid footstep planning is to jointly guarantee obstacle avoidance and locomotion stability. This paper presents a capturability-aware sequential Model Predictive Control (MPC) framework for Linear Inverted Pendulum (LIP)-based humanoid footstep planning in the presence of obstacles. The proposed method combines obstacle avoidance with capturability constraints through safe corridors constructed as sequences of intersecting rectangles, yielding a sequential MPC framework with recursive feasibility guarantees for the LIP-based planning layer. The planner autonomously adapts to constrained scenarios, switching to in-place walking when obstructions---such as doors assumed to be open during planning but unexpectedly found to be closed during execution---block forward motion. Performance is largely independent of the prediction horizon, allowing high success rates while keeping the computational cost low. The approach is validated in high-fidelity simulations with the humanoid robot Digit, demonstrating a practical solution for capturable humanoid footstep planning in cluttered environments.

0.16

Look Only Where You Plan for Real-Time End-to-End Urban Driving

cs.RO ·2026-11-06 ·Haohua Que, Tianle Zhu, Qianyi Wu, Zhipeng Bao, Handong Yao
Abstract · 10 pages

Dense bird's-eye-view (BEV) construction accounts for 60% of the parameters in a strong recent driving planner. This paper presents LOWP, a planner that looks only where it plans: two sets of sixteen learnable queries read the cameras through the known calibration, one by sampling image features at projected 3D anchors and one by attending to features labeled with their viewing rays, so no dense BEV is built and no object or map is supervised. Evaluations on open-loop nuScenes, pseudo-simulated NAVSIM and closed-loop Bench2Drive show competitive planning with a compact representation. On nuScenes, average L2 changes from 0.302 to 0.296 m with 53% of the dense counterpart's parameters and 1.7 times its speed on the same GPU; the anchor-based reader performs better on turns than the global reader. On NAVSIM, scoring the planner's own trajectories alongside a vocabulary gives 85.9 PDMS without retraining. The closed-loop control configuration scores 65.5 on 218 routes, with a brake-induced standstill failure whose release rule has an inconclusive aggregate benefit. We distinguish measured latency from hardware-transferred estimates, which do not establish a common-hardware ranking.

0.16

FATHOM: World-Model-Guided Diffusion Action-Trajectory Selection for Sparse-Marker AUV Tracking

cs.RO ·2026-11-06 ·Xiaorui Wang, Zeyu Sha, Mingyang Yang, David Chen Wang, Feitian Zhang
Abstract

Reliable relative tracking is fundamental to cooperative autonomous underwater vehicles (AUVs), yet remains challenging under nonlinear vehicle dynamics and degraded underwater visual conditions. Active light markers improve robustness to appearance variation, but provide deliberately sparse observations that can become incomplete or corrupted by occlusion, reflections, and detection noise during dynamic motion. To address the resulting partial-observation control problem, this paper proposes FATHOM, a world-model-guided diffusion action-trajectory selection framework for sparse-marker AUV tracking. FATHOM first integrates structured marker observations, inertial measurements, action history, and temporal dynamics into a predictive tracking belief using a multimodal task-supervised recurrent state-space model. A diffusion action prior then generates multiple expert-like multi-thruster trajectories, whose future tracking consequences are predicted through the learned latent dynamics. Rather than directly executing a generated trajectory, FATHOM selects the most favorable candidate under an explicit receding-horizon objective and continually updates the decision from onboard observations. Extensive offline analyses verify the learned predictive belief, multi-step world-model prediction, diffusion trajectory generation, and predictive selection. More importantly, comprehensive real-AUV experiments, including strong pose-based comparisons and component ablations, demonstrate robust autonomous tracking across nominal and challenging motions with naturally occurring partial marker visibility, while maintaining real-time onboard operation without motion-capture feedback.

0.15

ActPrune: Action-Grounded Visual Token Pruning for Efficient Vision-Language-Action Models

cs.RO ·2026-11-06 ·Sungwon Moon, Minji Kim, Hyemin Boo, Daeun Song, Jiyoung Lee
Abstract

Vision-language-action (VLA) models perform versatile manipulation tasks from language instructions, but their inference cost limits how fast a robot can respond in closed-loop control. Visual token pruning reduces this cost; however, prior methods estimate token redundancy from encoded visual embeddings or attention, only after the visual encoder has processed the entire observation. We propose ActPrune, a training-free token pruning method that moves the pruning decision ahead of the vision encoder by grounding it in a geometric prior derived from the robot's own action. The action chunk predicted at the previous control step is already available before the current observation is encoded, and it specifies the end-effector and camera motion between consecutive observations. The same action affects the two standard camera views asymmetrically: in the fixed third-person view, ActPrune retains patches along the projected end-effector trajectory together with semantic anchors, while in the wrist view, it prunes patches that remain unchanged after compensating for the action-induced camera motion. Extensive experiments on two simulation environments (LIBERO and VLABench) and four real-robot tasks demonstrate that ActPrune achieves up to 1.61$\times$ speedup and 59.5% FLOPs reduction with negligible degradation in success rates. The project site is http://mmai.ewha.ac.kr/actprune/.

0.15

On PMU Adequacy for Wideband Dynamics Tracking Beyond Attenuation Assumptions

eess.SP ·2026-11-06 ·Chetan Mishra, Luigi Vanfretti, Jaime Delaree Jr. and, Kevin D. Jones
Abstract

Recent work highlights the inadequacy of standard synchrophasors in capturing high-frequency grid dynamics, including data center oscillations, primarily due to filtering and windowing attenuation during phasor estimation. This letter shows how synchrophasor ambient data spectrum itself contains information on attenuation, which provides a way to identify critically impacted measurements and potentially estimate and undo the effect. Results are shown for synthetic data as well as actual synchrophasor estimates from digital fault recorders (DFRs) within the Dominion Energy system.

0.14

VLA Handover: Failure-Prediction-Based Switching from On-Robot to Server VLA Models

cs.RO cs.LGcs.NI ·2026-11-06 ·Takayuki Nishio
Abstract · 6 pages, 4 figures, 3 tables. Submitted to IEEE ICC 2027

Large vision-language-action (VLA) models generalize well but are difficult to run in real time on a robot, and small VLAs that run there fail more often. Remote inference over beyond-5G and 6G networks lets a robot use a large VLA on a server, but querying the server for every action chunk limits the number of robots per server. We propose VLA handover: the robot runs a small VLA (S-VLA) and switches to a large VLA (L-VLA) on a server only when the S-VLA is likely to fail. We formulate the switching as minimizing the request cost under a required success probability and, since this problem cannot be solved directly, propose a heuristic framework that retrains neither VLA. In simulation with SmolVLA and $π_{0.5}$ on three of the four LIBERO suites, VLA handover raised the success rate by 11-22 percentage points over the S-VLA alone and sent 49-77% fewer requests than server-only control at a 2-4 percentage-point lower success rate, and its completion time grew less with the network latency.

0.13

MDF++: Visuo-Tactile World Action Model with End-to-end Representation Learning

cs.RO ·2026-11-06 ·Zixuan Huang, Jinhao Yu, Zachary Charlick, Dmitry Berenson
Abstract · Corl 2026

Recently, world-action models (WAMs) have emerged as a powerful framework for robot manipulation. By jointly modeling action generation and future observation prediction, WAMs can act and anticipate the outcomes of the actions. However, existing WAMs typically rely on frozen pretrained encoders trained with reconstruction, which may not produce representations well suited for robot control. In this paper, we tackle the problem of learning physicallygrounded, control-relevant representations for WAMs. We propose MDF++, a visuo-tactile world-action model that learns latent representations of multisensory observations end-to-end with a diffusion backbone. While naive joint training can lead to feature collapse, MDF++ introduces a multi-task learning strategy that leverages cross-modal correlations to learn task-relevant representations. We evaluate MDF++ on a suite of contact-rich manipulation tasks, including realworld bolt screwing and nut mating, and show that it outperforms WAMs built on frozen pretrained representations by up to 50%. Additional videos are available on our website: jointmdf.github.io.

0.13

ODEAction: From Predicting Chunks to Evolving Actions

cs.RO ·2026-11-06 ·Qishen Wang, Pengxiang Ding, Shifeng Bao, Mingyang Sun, Hengtao Li +12 more
Abstract · 23 pages, 5 figures, including appendix. Technical report

Vision-language-action (VLA) models typically generate actions in chunks, dividing continuous robot execution into separately predicted action chunks, which can introduce discontinuities at chunk boundaries. We propose ODEAction, a chunk-free action generation paradigm that replaces action chunk prediction with Continuous Action Evolution. We use an ordinary differential equation (ODE) to describe the continuous evolution of actions, and define the vector field of this ODE as the Action Transition Field to characterize how actions evolve with respect to evolution time. During inference, the model takes the current action as the initial condition and numerically solves the ODE defined by the learned Action Transition Field, thereby recursively generating subsequent actions with explicit dependencies. ODEAction is implemented as a modular action head that can be integrated with different vision-language backbones. Experiments on LIBERO, EBench, RoboTwin, and real-robot tasks show that, compared with action-chunk-based baselines, ODEAction improves both task performance and trajectory quality, with particularly pronounced gains on fine manipulation tasks.

0.11

STRIDE: Learning Humanoid Locomotion in Minutes with Acceleration-Based Rewards

cs.RO cs.LG ·2026-11-06 ·Nico Bohlinger, Jan Peters
Abstract

Learning locomotion directly on a humanoid robot is a promising direction to avoid the sim-to-real gap. However, locomotion rewards in Reinforcement Learning (RL) typically track the base velocity, which cannot be measured onboard without an external motion capture system or well-calibrated state estimator. This is why we introduce \method{}, which learns humanoid locomotion from minutes of samples with a reward that only requires the onboard Inertial Measurement Unit (IMU). Instead of tracking the base velocity, STRIDE rewards the robot for following a commanded acceleration profile. To learn efficiently from this reward, we use the off-policy RL algorithm XQC and combine a high gain setting for the joint position controllers with left-right symmetry augmentation. In simulation, STRIDE learns to walk on the Unitree G1 and Booster T1 humanoids within 14 minutes of samples. Compared to learning without these two components, it reduces the falls during training by 81%, while without them the robot learns only in over an hour.

0.11

Toward the Era of Agentic Embodied Intelligence

cs.RO ·2026-11-06 ·Chao Liang, Ke Chen, Fengxiaoxiao Li, Shibo Huang, Yufeng Sun +5 more
benchmarkcommunity pick
Abstract

Foundation models increasingly guide robot actions, yet sustained autonomy requires efficient control and failure recovery. We introduce A2-MEAS, an embodied multi-agent framework built around code as a scaling interface. Learned correction codes combine corrective language, spatial evidence, execution history, and action proposals. MetaBrain provides training supervision and runtime guidance. A correction encoder and adapter absorb this knowledge into policy weights during training, while fresh codes adapt behavior with frozen weights at deployment. The action model, A2-WLA, adds motion-aware history and a latent world model to a vision-language-action policy. Surprisingly, this simple design with action pretraining and task specialization reaches 0.5755 Q-score on the 50-task BEHAVIOR 2026 public subset before online correction, 2.21x the 2025 winner's reported 0.2605 across different benchmark editions and training configurations. On all 100 BEHAVIOR 2026 public tasks, A2-MEAS achieves 0.6411 Q-score and 45.40% task success. On RoboTwin 2.0, A2-WLA achieves 93.98% average success, rising to 95.05% with the complete system. Across 12 real-robot tasks in five household settings, A2-WLA achieves 57.92% average success versus 26.67% for pi0.5. In long-horizon pour-over coffee preparation, a history-conditioned adapter raises post-failure completion at two filter-handling steps from 14% to 94% and from 8% to 82%, compared with A2-WLA retries. Our training infrastructure delivers 427.1 samples per second, 4.03x the throughput of a pi0.5 reference implementation on matched hardware. We will release the system, models, training scripts, and infrastructure to support efficient embodied learning.

0.10

Horizon-Coupled Obstacle Guidance for Relative-Action Diffusion Policies

cs.RO ·2026-11-06 ·Guangyuan Wang, Haishan Xia, Zitao Zhang, Kejia Chen, Yuanchen Sun +4 more
Abstract

Real-world robot deployment requires adapting learned policies to unseen obstacles without collecting new demonstrations or retraining. Existing policy-level guidance often applies obstacle corrections independently at individual horizon steps, which may address local obstacle violations but fail to coordinate the action sequence into a complete, task-directed detour. How obstacle responses should be propagated across the action horizon to coordinate complete, task-directed avoidance motion remains underexplored. We introduce an inferencetime obstacle-guidance framework for pretrained relative-action diffusion policies. It rolls out cumulative relative actions into a workspace trajectory, evaluates obstacle interactions along the trajectory with an SDF-based obstacle model, and propagates each obstacle response to all contributing action increments during denoising. Real-robot experiments across seven obstacleconstrained scenarios spanning four manipulation tasks show that our method achieves 105/140 (75.0%) collision-free task successes, compared with 8/140 (5.7%) for absolute-action pointwise guidance. Ablations show that the relative-action horizoncoupled formulation provides the larger observed improvement across both obstacle models.

0.10

A Controlled Study of Models, Skills, Feedback, and Curriculum in Agentic Robotics

cs.RO cs.AIcs.LG ·2026-11-06 ·Kaiyuan Chen, Xinyue Zhu, Shuangyu Xie, Uksang Yoo, Thi Thu Phuong Luong +5 more
Abstract

Agentic robotics asks coding agents to implement and iteratively improve robot policies from simulated and real-world physical interactions. We study the self-learning loop of agentic robotics through a controlled simulation study spanning nine manipulation tasks, 631 different experiment configurations, and 28K coding agent rollouts, systematically varying model capability, skill priors, feedback structure, and curriculum. The experiment data suggests that the bottleneck is often not the access to sufficient information, but interpreting it into the right policy update. Across these factors, a successful self-learning loop depends on how effectively agents attribute failures, reason task-relevant spatial and geometric structure, and selectively restructure code to improve policies.

0.10

Localizing and Editing the Goal Representation of a Vision-Language-Action Model

cs.RO ·2026-11-06 ·Sudhanshu Srivastava, Faisal Aqlan
Abstract

Operators can state safety constraints for robots as prohibitions such as "do not put the bowl on the stove". We study whether OpenVLA follows such prohibitions and where it represents the goal it executes. The 10 tasks of LIBEROGoal share one scene, so we can label the goal that each rollout executes. We combine closed-loop rollouts with activation patching, linear probes, activation steering and runtime monitors. On LIBERO-Goal, negation had no detectable effect on which task OpenVLA executed. The model executed the forbidden task in 79% of negated rollouts and completed 78% of plain tasks. A goal probe gives similar read-outs under "not X" and "X". Activation patching shows that the first action depends mostly on the instruction tokens before layer 12 and on the last prompt token from layer 16 onward. A fixed steering vector at the instruction tokens at layer 8 redirected 12 of 18 rollouts. Fixed vectors and prefill patches at layers 12 to 20 redirected none of 158. From layer 16, only patches at every decoded action token redirected the robot (15 of 18 rollouts at layer 24). When the instruction changes during a rollout, the layer-32 goal read-out detects non-compliance better than failure detectors, and an external probe on image and instruction features performs comparably. These results support enforcing prohibitions outside the instruction and monitoring goal read-outs against the specification. Goal edits in OpenVLA should act before layer 12 or patch every action token.

0.09

Supporting Data Center Grid Integration via Co-Optimized Battery Energy Storage Sizing and Capacity Allocation

eess.SY math.OC ·2026-11-06 ·Wei Wang, Himanshu Sharma, Thiagarajan Ramachandran
Abstract

Interconnecting large data centers can be delayed by grid requirements on demand flexibility, ramping limits, and ride-through performance. This paper proposes an optimization-based framework to size a Battery Energy Storage System (BESS) that accelerates data center grid integration by providing BESS capacity required to meet selected flexibility and grid-support requirements. The model co-optimizes BESS power rating and energy capacity while allocating capability across three services: (i) demand response over a specified duration, (ii) mitigation of high-magnitude, high-frequency load fluctuations through reserved active power with zero net state-of-charge change, and (iii) voltage support via reactive power capability at the point of common coupling. Minute-level operational scheduling enforces state-of-charge feasibility and load ramp constraints, and incorporates a controllable queue of delay-insensitive computing tasks to enhance flexibility. A case study using a real data center load profile demonstrates the proposed method and quantifies the BESS power-energy requirements and operational trade-offs needed to satisfy the multi-service constraints, providing practical guidance for interconnection planning and deployment.

0.05

People & Society 28

Rhythmic Fitness Beat 'Em All using Real-Time Monocular Human Pose Estimation

cs.HC ·2026-11-06 ·Yanis Zaoui, Valentin Bégel, Mohamed Adjel
publishedhot direction
Abstract

Short and frequent physical activity exercices mitigate the health risks of sedentary behavior. Yet, the lack of motivation remains a critical barrier to adoption. Exergames gamify physical activity, but most rely on dedicated hardware, and outside dance and rhythm games they rarely use the musical beat. We present a work-in-progress beat 'em up exergame played with the whole body in front of a single smartphone, which runs the entire pipeline: an off-the-shelf real-time 3D human mesh recovery model on the phone's neural processing unit, gesture recognition, and the game, shown on any TV through a USB-C to HDMI cable. Our contributions are (1) a 3D recognition of five shadow-boxing moves (guard, punch, slip, squat, jump), each defined by one metric measure in a body frame aligned with gravity and one threshold with hysteresis, personalised during a tutorial; (2) a rhythm stage whose discrete cues are placed on the song's kick drum and judged on the movement's capture time, with a delay learned for each player; and (3) a phone-only architecture, running at 30 frames per second, that turns a phone and an HDMI cable into an exergame console.

0.58

Same Facts, Different Remedy: Linguistic Profiling and Arbitrariness in LLM and System One Decisions

cs.CY cs.AIcs.CL ·2026-11-06 ·Colin Priest
code/data
Abstract · 29 pages, 7 figures. Code, prompts and pre-registrations: https://github.com/colinpriest/llm-jev-linguistic-profiling

Large language models are increasingly used to resolve consumer complaints, and the way complainants write varies with age, education and socioeconomic status (SES). We audit three remedy-decision systems using sociolinguistic rewrites of US consumer financial complaints that are instructed to preserve every fact: two generative LLMs that return a single label (GPT-4o-mini and GPT-5-mini) and a probabilistic "System One" model that returns a distribution over remedies (Jev). Using pre-registered hypotheses tested once on held-out complaints, we separate two phenomena that single-label audits conflate. At the group scale, older, college-educated and higher-SES styles receive slightly more generous remedies; these effects replicate but are small, with every monetary-remedy gap within 0.5 percentage points. At the individual scale, two style versions of the same complaint receive different monetary decisions 2.6-6.3% of the time. In a variance-component analysis of the probabilistic model, persona components shared across the two rewrite models account for about 1.6% of within-complaint variation. The persona gaps are associated mainly with delivery (vocabulary, grammar, politeness) rather than content. Positive controls show that all three systems respond 49-387 times more strongly to remedy-relevant facts than to style. Rewriting every complaint into one standard style reduced GPT-4o-mini's education and SES gaps by 94% and 87% while retaining fact sensitivity, but reversed two of the probabilistic model's gaps and did not make individual decisions more stable. Independent LLM auditors flagged material fact changes in 3.3% and 11.7% of rewrites, and instability persisted among unflagged rewrites. Fairness audits of LLM decision systems should report group disparities and individual instability as separate outcomes.

0.47

The Challenge of Modeling the Complexity of Use for the Measurement of Digital Maturity in Education

cs.CY ·2026-11-06 ·Christine Michel, La{ë}titia Pierrot, Frederic Oru, Olivier Vigneau
published
Abstract

Measurement of teachers' digital maturity can be useful in diagnosing technology adoption in education and in supporting teachers' professional development. However, there is a lack of easily usable tools in the existing literature. By investigating the most effective ways to measure teachers' digital maturity, our research aims to fill this gap. The focus of this article is on a data-driven approach to the assessment of digital maturity in education. Following a process similar to teaching analytics (data collection, data modeling, data understanding and visualization, and contextual understanding), we draw on diary data collected from 12949 French primary school teachers using a virtual learning environment (VLE) during the 2022--2023 school year. Specifically, the paper presents results related to modeling and visualizing data. To this end, we propose a critical analysis of two automatic classification techniques, one supervised rule-based and the other unsupervised (hierarchical clustering). In addition to providing different scales to measure, we have proposed a new approach to diagnose and visualize digital maturity. In this way, the method provides insight into the diversity and intensity of technology use, contributing to a better understanding of digital maturity and providing a practical tool for assessing day-to-day teaching practices.

0.40

The Drawing You Watch Is the Portrait You Keep: A Replayable Machine Portrait Studio

cs.HC cs.CYcs.GR ·2026-11-06 ·Vinod Rajpurohit
Abstract · 25 pages, 6 figures, 9 tables. Live system: https://everyfaceproject.com

Ask an image model for your portrait and a finished picture arrives: no first line, no second thoughts, no record of how it got there, and a different picture if you ask again. I built a portrait studio that works the other way. A sitter chooses one of eight fictional artists, each written with a city, a medium and a tempo of its own, holds still for 8 to 12 seconds, and watches a drawing of themselves appear over 44 seconds. One still is sent after consent, a hosted image model returns a private study the sitter never sees, and a compiler turns that study into one drawing document. The live drawing, the saved sheet, the vector export and the film all read that document, and a replay makes no model call, so the drawing you watch is the portrait you keep. I describe the compiler in enough detail to rebuild it and say plainly what it is: an authored reconstruction at a compressed speed, not a recorded hand. On 16 generated test studies a sheet holds 8,784 to 16,378 timed contacts, about three quarters of them fixed-grid fills and nearly all shorter than one video frame, and recompiling one study gave byte-identical output four times on one machine. The paper also traces a bug that matters more than any of that. The earlier compiler estimated the paper colour tile by tile, read dark skin as paper, and in all eight hands kept less of a deep-complexioned face than a fair one (13 to 48 percent against 41 to 56). On 80 generated studies the rebuilt engine keeps the complexion difference the study already has (raw retention 89 to 120 percent), so the fairness question has moved upstream, to the study and the hosted model. The two test sitters are generated and differ in more than complexion, and no real-face or perceptual study exists yet. The studio is live at https://everyfaceproject.com.

0.36

VeV: Video-Enhanced Vibration Data Augmentation For Structural Difference

cs.HC cs.AI ·2026-11-06 ·Joonhee Lee, Jiale Zhang, Sungmin Lee, Ruiqing Wang, Jeffrey Shulkin +3 more
Abstract · 35 pages

Surface-mounted vibration sensing provides non-invasive structural monitoring with high temporal fidelity across diverse ubiquitous computing scenarios such as smart farms, yet structure-borne wave models remain fragile to physical relocation and structural shifts. However, addressing this spatial data discontinuity remains limited by costly manual recalibration, insufficient physical grounding in data-driven synthesis, and substantial sim-to-real gaps in numerical digital twins. To resolve this spatial bottleneck without labor-intensive data recollection, we propose VeV, a physics-consistent spatial augmentation framework that synthesizes vibration responses at unmeasured target coordinates from sparse reference observations by coupling a monocular Vision-Derived Propagation Possibility Map for structural wave-guiding paths with a Visual Feature Map for transient source kinetics and implicit material dynamics. Evaluated on a 1,237 sensor-hour sow pig and piglet farm across disparate structures, VeV reduces waveform reconstruction MSE by over 57.5% compared to data-driven baselines and achieves up to 81.69% zero-shot classification accuracy, while reaching peak performance within just 30-50 calibration chunks to effectively eliminate cold-start retraining overhead and establish a scalable foundation extending to human-centric floor vibration sensing and ambient intelligence.

0.35

Uniqueness-Aware Feature Engineering for Representing Humans in Algorithmic Decision Making

cs.CY cs.HC ·2026-11-06 ·Theresa Elstner, Martin Potthast
Abstract

Fairness analyses of algorithmic decision systems typically treat the feature representation of individuals as an a priori fixed input. How well a feature representation captures what is actually decision-relevant for different people and their unique backgrounds, however, is typically left unexamined. We address this gap by modeling feature engineering as a stochastic selection process over the pooled characteristics as candidate features for a finite training population. Based on this model, we develop a new framework for uniqueness-aware feature engineering which operates in two layers. The first layer measures population heterogeneity using a multivariate hypergeometric model of feature prevalence. It shows how well an individual is represented by any given input representation and which individual of the population is "least represented" compared to their personal, reference characteristics. The second layer acts on this assessment using set-similarity measures to search for an optimum between representational fit of the population and predictive accuracy. Without assuming prior knowledge about unseen individuals, our framework grants represented humans agency over what counts as decision-relevant in their case, gives deploying institutions accountability through a publishable representation fit report, and provides each person subject to a decision transparency about their personal representation fit. We operationalize our framework as a six-step pipeline and show a proof-of-principle using synthetic data. Our approach offers a blueprint for making algorithmic decision systems accountable for how well they represent individuals during development and in production.

0.31

A Neural Physics Solver for Richards Equation

cs.CE math.NAphysics.comp-ph ·2026-11-06 ·Katherine MacKenzie, James Rand, Ben Ashby, Tristan Pryer, Claire E. Heaney +2 more
Abstract · 36 pages (including appendix), 7 figures, 3 tables

A new approach to discretise and solve Richards equation based on the finite element method (FEM) is presented. The novelty of the approach, called the Neural Physics method, is that it expresses the discretisation as a convolution neural network (CNN) whose weights are predetermined by the integrals of the basis functions in the FEM. The Neural Physics method is physically consistent as it represents a discretisation of the governing equation and can run on CPUs, GPUs and the latest AI accelerators, resulting in an increase in computational speed. The Neural Physics method is demonstrated for two different discretisations of Richards equation, including a multigrid method that is realised through a U-Net architecture. The accuracy and speed of these methods is tested on a problem with a known analytical solution and compared to a standard FEM, and shows that the Neural Physics method is almost an order of magnitude faster than a comparable FEM. Finally, the Neural Physics method is applied to a real-world test case with integrated experimental data and simulates the infiltration of water through a rain garden. This new approach therefore provides a route to efficiently and accurately solve Richards equation for a wide array of hydrological applications.

0.31

Sensed but Not Told: Hidden Physiological Adaptation Shapes LLM Persuasion

cs.HC ·2026-11-06 ·Xiaoyan Wei, Yutong Qu, Yutong Li, Haiyi Li, Wei Emma Zhang +1 more
Abstract · 15 pages, 5 figures

Wearable sensors, including those embedded in VR headsets, can provide LLM-based chatbots with physiological signals that users do not explicitly communicate. We tested whether such hidden input changes persuasion. In a $2\times2$ within-subjects study ($N=24$), a chatbot argued against participants' self-selected opinions with or without turn-level heart-rate and pupil-diameter features; sensing was disclosed halfway through the study. Before disclosure, sensing increased perceived persuasiveness by 1.42 points. Across the study, sensing was associated with a 0.57-point greater shift toward the chatbot's position and 3.20 times the odds of any movement toward it. Persuasiveness ratings were lower after disclosure, although disclosure was confounded with session order. Dialogue logs suggested a more declarative interaction style when sensing was on: chatbot replies contained question marks less often, and conversations contained fewer turns. These findings motivate making biosignal-driven adaptation inspectable and providing a genuinely non-sensed mode.

0.30

Police Behavior Patterns Through Radio Analysis

cs.CY cs.CLcs.SIeess.AS ·2026-11-06 ·Milo Z. Trujillo, Adam Perhala, Elaine Zou, Tyler Mckenzie, Juniper Lovato +2 more
dataset
Abstract · 10 pages (14 with SI), 6 figures (8 with SI)

In the United States, most police departments communicate using police radio, which allows for officer-to-officer and officer-to-dispatch communication without depending on cellular network or WiFi availability. While some departments use encrypted digital radio, many continue to transmit in the clear, and are legally monitored by public citizens with `police scanners.' Because police radio is intended for internal police use and represents real-time communication, it may offer valuable insight into patterns of law enforcement activities without the potential bias and filtering introduced by a press release after the fact. In this study, we collect and examine over 14 million radio transmissions representing over 100,000 hours of speech by officers in 34 cities across the United States from July 2025 through July 2026; to our knowledge the largest scale analysis of Broadcast Police Communication (BPC) to date. We highlight both typical patterns of police communication and categorize anomalous activity that represents incidents including protests and major weather events. We present this dataset as a starting point for an array of police communication studies, conduct our own preliminary analysis, and propose several further areas of inquiry.

0.30

DeepRadio: A Human-Agent Workspace for Physical Radio Development

cs.HC ·2026-11-06 ·Jiazhen Lei, Yuxin Sha, Sihan Wang, Wenhui Li, Jinyang Sun +2 more
Abstract

Agentic programming can automate software-defined radio (SDR) development, but decisions can propagate from software to hardware without sufficient user visibility or control. We present DeepRadio, a human-agent workspace embedded in GNU Radio Companion (GRC) for jointly constructing, verifying, and operating physical radio systems. DeepRadio uses a dynamic and state-driven workflow to coordinate a traceable radio specification, claim-oriented cross-layer verification, and mixed-initiative control. These mechanisms expose decisions, tie conclusions to available evidence, and return control to users before consequential hardware actions. Across seven atomic SDR tasks, DeepRadio completed nearly all runs while satisfying all prescribed software and radio-frequency constraints. In a 20-participant study, more participants achieved a verified physical deployment with DeepRadio than with Codex-assisted GRC. Participants reported stronger support for shared intent and understanding while retaining initiative and control. These results suggest that DeepRadio can make agent-assisted physical radio development more reliable, understandable, and controllable.

0.29

Bidirectional Modeling of mmWave Human Signals for Enhancement & Synthesis

cs.HC ·2026-11-06 ·Shuntian Zheng, Hoang Hai Pham, Jiaqi Li, Xiaoman Lu, Shuai He +1 more
Abstract

Millimeter-wave (mmWave) is attractive for human-centered robotics in smoke, darkness, and privacy-sensitive indoor spaces: it measures reflections rather than visually identifiable imagery and remains usable without light. Deployed systems, however, face two complementary needs that run in opposite directions. Enhancement starts from a raw range-angle-Doppler (RAD) heatmap, where human returns coexist with nonhuman structure, and aims to isolate the human signal by removing nonhuman returns, yielding a clean RAD input for downstream perception. Synthesis starts from denser human geometry and aims to synthesize natural RAD observations, including environmental and hardware features, thereby expanding scarce supervision data. Although the two tasks reverse direction, both require modeling how human and nonhuman reflections coexist in the same measurement. Among RAD axes, Doppler provides motion information native to the modality, with the potential to support this modeling. We therefore cast both tasks on the dual-domain RAD heatmap and propose HeatWave, which injects Doppler when it is locally reliable, and routes Doppler asymmetrically for enhancement and synthesis within one shared backbone. This design improves both tasks: enhanced observations reduce mean per-joint position error by 11.4% relative to the raw recording, and synthesized samples that replace half of the real radar labels recover most of the activity accuracy lost to the reduced budget. Relative to an unspecialized dual-task heatmap baseline, the enhancement improves motion-critical samples under metallic clutter by 2.3times more than when geometry already suffices, and synthesis follows the same ordering with a narrower spread, providing the community with a research direction more closely aligned with the signal's characteristics.

0.28

Dynamic signatures of collective knowledge through navigation in a network with semantic structure

cs.SI physics.soc-ph ·2026-11-06 ·Abby Bertics, Elizabeth R. Chrastil, Jean M. Carlson
Abstract · 24 pages, 9 figures, submitted to PLOS ONE

What distinguishes how humans navigate an information network like Wikipedia from what the underlying graph structure and language statistics alone predict? To study human collective navigation, we analyze 70,770 human trajectories through Wikispeedia, an online game where players navigate between articles to reach a target. To compare groups of trajectories rather than individual ones, we merge all trajectories leaving or entering a given article into a Flow Tree. Flow Trees compress collections of trajectories into hierarchies of group splits. We build forward and reverse Flow Trees for 815 of the most commonly traversed articles, representing divergence and convergence, respectively. We compare humans against four random-walker null models, either unweighted or weighted by graph structure, word frequency, or semantic similarity to target. Human Flow Trees are deeper with longer trunks and later splits, indicating players stay together longer than random walkers, whose Flow Trees tend to fan out immediately. An article's degree predicts 51.3-70.9% of the variation in Flow Tree branching factor for random walkers, but only 24.1-36.2% for humans. Taking what graph structure fails to predict as a two-dimensional coordinate system, we define four article types (beacons, funnels, forks, and corridors), distinguished by how humans enter and exit them. These human-defined types are not explained by out-degree or title-word frequency are, and are not recovered by any random walker. A fifth null model, built from first-order human click preferences, partially reproduces human Flow Tree shape but differs significantly in how much of that shape degree explains. We consider collective knowledge to be the structure shared by players who never communicate, beyond what graph structure imposes. It comprises first-order click preferences and a significant component that extends across clicks.

0.27

Lea: An Agent Backbone for Mathematician-Controlled Formalization in Lean

cs.HC ·2026-11-06 ·Shaswat Patel, Daniel Aarao Reis Arturi, Kevin Gomez, Jaume de Dios Pont, Jorge Poco +4 more
code/data
Abstract

Agentic systems now formalize research-level mathematics in Lean, but they are built for autonomy: the mathematician states a target and inspects a result. The alternative, a general-purpose coding harness, is domain-neutral and must be assembled into a formalization workspace before any mathematics begins. Mathematicians, however, report wanting to keep high-level control over how an argument is decomposed and established. We present $\textbf{Lea}$, a Lean-specific agent backbone for mathematician-facing software rather than a single interface. Lea exposes runs, sessions, and typed events through one application-neutral API, and ships with Lean tooling (blueprints, LSP, Loogle, SafeVerify) already wired. We report two applications built on it: $\textbf{LeaChat}$, a web client for interactive proof authoring, and $\textbf{LeaOverleaf}$, which formalizes theorems annotated directly in an Overleaf document; a session begun in either continues in the other. In a case study, a mathematician outlined a proof of the Durfee square generating function in a few natural-language messages, and Lea produced a complete Lean development from Mathlib alone, following the intended decomposition closely and requiring substantive correction at a single point. Lea and both applications are released as open source so that further mathematician-facing tools can be built on the same backbone.

0.27

Locally Adaptive Conformal Prediction for Polynomial Chaos Expansions

cs.CE stat.COstat.MEstat.ML ·2026-11-06 ·Dimitrios Loukrezis, Arthur Hatstatt, Bruno Sudret, Dimitris G. Giovanis
benchmark
Abstract · 22 pages, 11 figures, 2 tables

This work develops a locally adaptive jackknife+ conformal prediction framework for polynomial chaos expansions. Therein, the non-conformity scores are normalized with an auxiliary polynomial chaos expansion trained on the log-transformed leave-one-out residuals of the base model. We show that the coverage guarantee of jackknife+ conformal prediction carries over to the adaptive intervals if the leave-one-out auxiliary models are trained on nested leave-one-out residuals. For fixed-basis polynomial chaos expansions fitted by ordinary least squares, all required quantities are derived in closed form and the guarantee-preserving adaptive intervals are obtained from a single model fit, without any model retraining. For sparse polynomial chaos expansions, where multiple model fits are necessary, we propose a heuristic simplification that limits the computational overhead of adaptivity at the cost of the formal coverage guarantee. Numerical experiments on an academic benchmark with known heteroscedastic noise and on an engineering simulation dataset show that the suggested method attains the target coverage while providing intervals that adapt to local prediction uncertainty, therefore being considerably more statistically efficient. These advantages are particularly prominent in high-uncertainty regions of the input space, where the performance of non-adaptive methods degrades substantially.

0.26

Entering a New Domain in the Era of Generative AI: An Autoethnography of Solo Game Development as a Bioinformatics Educator

cs.HC cs.CY ·2026-11-06 ·Hannah Kim
Abstract · 12 pages, 1 figure

The rapid improvement of generative AI (GenAI) not only increased the efficiency of practice but also empowered individuals with capabilities to explore new domains. However, identity formation in the wake of such empowerment remains understudied. We use an autoethnography to understand identity formation in the context of solo game development as a bioinformatics educator. With the goal of developing a create-your-own-adventure (CYOA) game, we collected data for 73 days through journal entries, interacting with communities of practices. The autoethnographer had continuously reassessed the situation, and closely interacted with communities of practice. Their prior knowledge was used to provide in-game context and to uniquely interpret game development processes. These findings suggest that the autoethnographer made efforts to belong to communities of practice and engaged extensively in sensemaking through the lens of a bioinformatics educator. Although GenAI puts power at hand, attaining a new identity took place at a human pace.

0.25

DrivAerRim: A large-scale full-vehicle CFD dataset for wheel rim aerodynamics

cs.CE ·2026-11-06 ·Chao Xia, Xi Zeng, Alexey Vdovin, Huanxia Wei, Simone Sebben +3 more
dataset
Abstract

The wheel region contributes substantially to vehicle aerodynamic drag, and rim geometry strongly influences the flow in this region. However, publicly available automotive aerodynamics datasets offer limited variation in rim design. We present DrivAerRim, a large-scale dataset of full-vehicle computational fluid dynamics simulations covering 904 DeepWheel rim designs with diverse spoke layouts, opening areas and topologies. All cases use the same DrivAer Estateback, operating conditions and numerical setup, with rim geometry as the sole design variable. Wheel rotation is modelled using moving reference frame regions around the rims and rotating wall conditions on the deformed tyres. The numerical setup is assessed through mesh sensitivity analysis and comparisons with wind tunnel measurements. The resulting 9.32 TB release includes wheel geometries, vehicle and component force coefficients, accumulated drag, surface pressure, wall shear stress, and flow fields around the vehicle and wheels. Together, these data support surrogate model training and evaluation across diverse rim geometries, as well as aerodynamic optimisation and generative design to accelerate wheel rim development for vehicle drag reduction.

0.25

"I had nobody to turn to": Why LGBTQ+ People Choose (Not) to Use AI Chatbots for Personal Health and Relationship Advice

cs.HC ·2026-11-06 ·Jacob Hobbs, Ashray Gupta, Dennis Ulmer, Geetanjali Aich, Nolan Golden +8 more
Abstract · 35 pages, 2 figures, 8 tables

While there is an abundance of research into large language model (LLM) biases against LGBTQ+ communities, far less research has examined how and why LGBTQ+ people choose to make use of AI chatbots. In this work, we examine how LGBTQ+ people choose (not) to use AI chatbots across three domains in which LGBTQ+ users have unique needs: relationship support, mental health support, and gender or sexual health support. We surveyed 73 people (60 LGBTQ+; 13 non-LGBTQ+) on how they interact with AI chatbots and find that participants' access to external resources, privacy concerns, and perceptions of chatbot capabilities shape whether and how they choose to make use of AI chatbots. For instance, some LGBTQ+ participants felt that chatbots might actually be less biased against LGBTQ+ people than the people they know offline. We also find that LGBTQ+ participants report using AI chatbots less frequently than non-LGBTQ+ participants across 15 specific personal health and relationship advice scenarios, citing concerns about AI sycophancy and adverse cognitive side-effects. Recognizing AI chatbots as vital (albeit imperfect) resources for LGBTQ+ people, we recommend researchers consider systemic issues that drive users to AI chatbots and design technologies (AI or otherwise) that prioritize privacy, anonymity, and real human connection.

0.24

Thinking Twice: A Sequential Approach to Nudge Towards Reflective Judgment in GenAI-Assisted Decision Making

cs.HC ·2026-11-06 ·Hüseyin Hussein Keke, Daniel Eisenhardt, Christian Meske
published
Abstract · Full paper accepted and presented at the 20th International Conference on Wirtschaftsinformatik (WI 2025), Münster, Germany. To appear in th…

Humans increasingly use generative AI (GenAI) for decision-making assistance, yet tend to engage in such situations using quick, intuitive (System 1) rather than deliberate, analytical assessment (System 2) as reflected in Dual Process Theory. Such engagement reduces the effectiveness of human-GenAI interactions and increases associated risks. Our research investigates how to design human-GenAI interactions that promote the transition from System 1 to System 2 thinking in decision-making. Based on the Three-Stage-Model, which suggests humans switch to System 2 when conflicts are detected, we conducted an experiment (n=130) testing sequential human-GenAI interaction as a cognitive forcing nudge during problem-solving tasks. This approach, where humans decide first before receiving AI assistance, enables detection of conflicting decisions. Compared to non-sequential settings (concurrent or no AI assistance), results show the sequential condition significantly enhanced both decision performance and GenAI prompt utilization, providing valuable insights for designing more reflective and thus effective human-GenAI interactions.

0.24

Research Collaborations Within and Across Disciplines in NIH-Funded Atlas Programs

cs.SI cs.DL ·2026-11-06 ·Michael P. Ginda, Yongxin Kong, Katy Börner
Abstract

The National Institutes of Health (NIH) fund several programs (aka consortia) that support the construction of organ-specific or whole person atlases. These initiatives are designed to build comprehensive open reference datasets while advancing knowledge graphs, analytics, visualizations, and user interfaces needed to explore multiscale human atlas data. Here, we compare the Human Tumor Atlas Network (HTAN), Human BioMolecular Atlas Program (HuBMAP) and Cellular Senescence Network (SenNet) to examine collaboration by consortium component (CC) and researcher role and its relation to scientific impact. We retrieved 884 NIH award records and identified 1,255 focal publications acknowledging these awards, involving 8,952 researchers; compiled 99,108 citing publications covering 93% of the focal publications; constructed co-authorship networks; and as single-project publications or internal collaborations (IC), including intra-CC and inter-CCs collaborations. within each atlas program. Results show that IC publications are rare but involve many funded projects and principal investigators (PIs). Technology-oriented CCs tend to engage in intra-CC collaboration, whereas experimental-oriented CCs often collaborate across CCs of the same type. Compared with other researchers, PIs exhibited a significantly stronger association with inter-CCs than with single-project publications. IC publications have higher mean Relative Citation Ratios than single-project publications. Inter-CCs publications have significantly higher cumulative and annual citation levels in the first five post-publication years. Study methods and results can be used to inform the funding, management, and impact reporting for future Big Science programs that bring together 300+ experts from different disciplines to address major research challenges.

0.23

AI Exposure and the Changing Responsibilities in Job Postings

cs.CY ·2026-11-06 ·Dora Zhao, Lin Ai, Pranav Khadpe, Siddharth Suri, Scott Counts
Abstract

The growing adoption of artificial intelligence (AI) in the workplace has raised questions about how employer demand for different work activities is changing. We examine how the prevalence and content of work activities in online job postings change in association with activity-level measures of AI exposure, derived from Copilot and Claude usage. To enable systematic comparisons, we develop a pipeline that extracts the responsibilities from job advertisements and maps them onto O*NET's standardized taxonomy of Intermediate Work Activities (IWAs). We then apply this approach to 50,000 job advertisements from 33 firms between 2023 and 2025. From this analysis, we first find that explicit mentions of AI increased by 66%, although much of this growth occurs outside descriptions of the actual work activities associated with a job. Looking specifically at the advertised job responsibilities, activities with greater AI exposure tend to decline in advertised prevalence, but only mildly as a share of the total number of responsibilities. At the same time, more AI-exposed work activities increasingly are associated with explicitly AI-related responsibilities, suggesting that employers are looking to hire people who can perform these responsibilities with AI.

0.21

Calibrating demand models versus adjusting OD matrices with traffic counts: a controlled comparison

cs.CE stat.ME ·2026-11-06 ·Yue Li, Shujuan Chen, Ying Jin
Abstract

Traffic counts are routinely used to calibrate travel demand models, particularly when the origin-destination (OD) matrix is not observed directly. An important modelling decision is whether to adjust individual OD flows or the parameters of the demand model that generates them. The latter approach uses fewer parameters and preserves a relationship between zonal characteristics and demand, but its accuracy depends on the assumed model structure. We compare three established approaches using the same prior information, traffic counts and equilibrium assignment. They estimate five production-gravity coefficients, origin and destination adjustment factors, or individual OD flows. Synthetic experiments on the Eastern Massachusetts and Chicago Sketch networks allow recovery to be assessed against known demand under different forms of model misspecification. Calibrating the five-coefficient model gives better OD recovery and held-out link-flow prediction when demand follows its assumptions or a two-purpose gravity process. Its advantage declines or reverses when destination choice or production relationships vary across zones. Cellwise adjustment achieves the closest fit to calibration counts in 65 of 66 evaluated conditions, yet it recovers OD demand least accurately of the three approaches in 45 of them. These findings show that model parameterisation matters beyond calibration fit. For transport planning, traffic counts should be used alongside independent evidence on trip production and distribution when selecting and validating a demand model.

0.21

TPIC: Energy-Conserving Space Plasma Simulations on GPU Supercomputers

cs.CE cs.DCcs.PF ·2026-11-06 ·Luca Pennati, Ivy Peng, Stefano Markidis
theory
Abstract

Artificial heating and numerical damping affect particle energization in kinetic space plasma simulations, even when the numerical scheme remains stable. Implicit and semi implicit energy conserving Particle-in-Cell (PIC) methods eliminate these errors in the discrete energy balance while relaxing explicit stability constraints. We present TPIC, a massively parallel implementation of the Energy Conserving Semi-Implicit Method (ECSIM) for GPU supercomputers. We extend the exact energy conservation in the ECSIM formulation to closed domains with conducting boundaries by deriving and canceling a spurious term in the discrete Poynting flux theorem. TPIC executes the complete particle and field cycle on NVIDIA and AMD GPUs, combining matrix-engine acceleration with asynchronous communication and computation. We validate TPIC in a variety of electrostatic and electrodynamic test cases relevant for space plasma physics applications, including quasineutral plasma expansion, two-stream instability, Whistler waves, Weibel instability, Whistler turbulence, and three-dimensional magnetic reconnection with conducting walls and 12.3 billion particles. We show that in all scenarios, TPIC agrees with the corresponding analytical theory or reference numerical results, and preserves the total system energy to machine precision. On 4,096 AMD GPUs, TPIC advances 1.9 trillion particles with 96.8\% weak scaling efficiency.

0.21

"Despite her age, Margaret was eager to make a change": Unpacking Aging Representations and Stereotypes in LLM-generated Stories

cs.HC ·2026-11-06 ·MD. Roqunuzzaman Sojib, Mohammad Sadat Hossain, Nafis Tahmid, Shattik Islam Rhythm, Novia Nurain
Abstract

As large language models (LLMs) increasingly generate everyday stories, their portrayals can shape expectations about aging. Prior work has largely examined age stereotypes in short responses, leaving unanswered whose lives models default to imagining and how age shapes characters' roles in complete narratives. We analyze 594 stories generated by Gemini, GPT, and Llama using text embeddings, LLM-assisted coding, and qualitative analysis. When age was unspecified, stories typically featured younger adults, with middle-aged protagonists largely absent. Older adults were often portrayed through loss, familiar routines, and help from younger characters. Even positive stories sometimes treated older adults' achievements as surprising or portrayed others making decisions on their behalf. We suggest evaluating not only the tone of generated stories, but also who appears, what opportunities they receive, and who makes decisions. Involving older adults in these evaluations can guide writing tools toward more varied, less stereotyped portrayals of later life.

0.20

Functional Dependency beyond Similarity: Hierarchical Decomposition of Training Data Attribution in Music Generative Models

cs.CY cs.LG ·2026-11-06 ·Changheon Han, Hugh Xuechen Liu, Ashkan Panahi, Kıvanç Tatar
Abstract · 23 pages, 9 figures, 11 tables. Under review

Copyright analysis focuses on the similarity between two works. However, generative AI models learn from training data through functional relations beyond similarity. Training Data Attribution (TDA) estimates the influence of each training instance on a generated output and may serve as a basis for remuneration. However, prior studies find global influence, where a few training instances rank as the most influential for almost every generation. Used directly, TDA scores would distribute part of every remuneration to these instances regardless of what was generated. To address this, we propose a new approach in which TDA scores are decomposed into three levels of influence, namely global indicating the influence of a training data to a trained model, group for its influence on a group of generated outputs (that could be clustered together), and specific for its influence on a single generated output. Our study finds that, across several music generative models and TDA methods, global- and group-level influence is non-negligible in every case. Based on these observations, we propose the term functional dependency for the relation of training data to a trained model or to its outputs. This notion both encompasses and distinguishes dependencies of different scopes, allowing TDA scores to be interpreted as distinct forms of contribution in legal, economic and ethical discussions on remuneration.

0.20

Replication Certification at Publication: Checkable by Anyone, Anywhere, in Seconds, Forever

cs.DL cs.CRstat.ME ·2026-11-06 ·Gary King, Yiqing Xu, Leo Yang Yang
Abstract

Authors increasingly deposit replication datasets---the data and code behind publications---in archives, and journals increasingly require them to, under progressively onerous rules. But a deposit alone proves little, so authors keep paying: months for the journal's reanalyses, years answering readers' questions, and rancor as failed replications become accusations because no record establishes what actually ran. We introduce a replication certificate: a cryptographically signed record of the deposited files and logged output. It replaces the journal's reanalyses with evidence anyone can check in seconds, which no one, including the archive, can backdate, and no one can fake without risking detection. Certification asks little of anyone and benefits everyone: The author deposits, then reruns the analysis once under a random seed the archive computes from the deposit's hash and a blockchain block; the archive signs, timestamps (in that blockchain), and stores the record---confidential data included---without rerunning anything; the journal merely checks a website for the certificate; and anyone can check or replay any step decades later.

0.19

The Relational Capacity Framework: A Lens for AI-Mediated Relationality

cs.HC cs.AIcs.CY ·2026-11-06 ·Emily Tseng, Patrick Gage Kelley, Renee Shelby, Ashley M. Walker
Abstract

A growing literature examines the impact of engaging with AI chatbots on our relationality: our ability to form relationships with others, maintain them through life's changes, and exit when needed. Currently, substantial work focuses on purpose-built AI companions (e.g. Replika, Character.AI), with less work investigating socioemotional uses of the more widely-used general-purpose chatbots. In this paper, we present a 100-person investigation into how people engage general-purpose AI for socioemotional needs. Participants' AI uses varied, from practicing specific social skills to seeking additional bandwidth and information as they navigated their social contexts. To explain this heterogeneity, we provide 8 states that describe how AI expands or contracts relationality for people with different levels of skills, bandwidth, and information. We then present a unified conceptual model for the role of technology in mediating human relationality, the Relational Capacity Framework, and outline future research designing appropriate roles for AI in human connection.

0.17

Rethinking Requirement-to-Code Traceability Through Single-Agent and Multi-Agent Approaches

cs.CE cs.SE ·2026-11-06 ·Sarthak Banerjee, Sallam Abualhaija, Domenico Bianculli, Lionel Briand
Abstract

Requirements-to-code traceability supports essential software engineering activities, including change impact analysis, maintenance, and regulatory compliance. Existing approaches typically rely on a retrieval-and-classification pipeline, which suffers from an inherent retrieval bottleneck: relevant source files that are not retrieved cannot be recovered during subsequent analysis. In this paper, we revisit requirements-to-code traceability in light of the substantially larger context windows offered by modern large language models, which make repository-scale analysis feasible without a separate retrieval step. We propose three approaches: a single-agent approach that analyzes the entire code repository in one pass, and two multi-agent approaches based on self-verification and multi-agent debate. Our evaluation shows that the single-agent approach substantially improves recall, achieving 80.9%, a gain of 38.6 percentage points (pp) over the state-of-the-art retrieval-based baseline, albeit with a 25.8 pp reduction in precision. By leveraging collaborative reasoning, the multi-agent debate-based approach achieves the best overall precision-recall balance, with 24.5% precision and 71.0% recall, yielding an F2 score of 49.5%. This represents an improvement of 7.5 pp over the state of the art and 2.7 pp over the single-agent approach. While the quantitative evaluation establishes the overall effectiveness of the approaches, the aggregate metrics alone explain little due to the considerable variability across results. Importantly, qualitative analysis provides key insights into highly variable quantitative results: the retrieval approach suits requirements that echo the code's vocabulary; single-agent, full-context reasoning suits well-documented, consistently named code; and multi-agent debate suits indirect, conflicting, or distributed evidence.

0.15

GroundWork: Bridging Human and LLM Reasoning Through a Structured Argumentation Model

cs.HC ·2026-11-06 ·Sjoerd Vink, Byungmoo Kim, Michael Behrisch, Adam Coscia, Remco Chang
Abstract

Large language models (LLMs) are increasingly integrated into workflows that require reasoning over text documents, yet interaction remains largely conversational. This leaves argumentative structure implicit or embedded in text, making it difficult to maintain, compare, and revise arguments as analyses evolve. We present GroundWork, a mixed-initiative system that makes argument structure a shared, editable state for humans and LLMs. GroundWork operationalizes the Toulmin model of argumentation as a typed graph with explicit operations for human and LLM edits. It exposes this structure through coordinated corpus views and an editable workbench for constructing, comparing, testing, and tracing source-grounded arguments. We evaluate GroundWork through an expert study with four participants spanning consulting, product development, law, and immunology, and three use-case scenarios in intelligence, law, and medicine. Our findings show that GroundWork makes source-grounded arguments inspectable, comparable, and revisable as shared human-LLM state.

0.06