CS Paper Radar

Daily listing

2026-10-23

674 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 214

Advancing SVD-based LLM Compression via Layer-Wise Error Model Search

cs.AI ·2026-10-23 ·Moritz Thoma, Maximilian Groezinger, Maximilian Forstenhäusler, Emad Aghajanzadeh, Manoj Rohit Vemparala +6 more
top venuecode/datahot direction
Abstract · Accepted at ICML 2026. Extended version with additional concurrent works and updated benchmark scores

Low-rank SVD-based compression offers a powerful strategy to reduce the computational costs of LLMs. However, existing methods face two key limitations: (i) global rank allocation, where uncalibrated error proxies fail to capture complex error propagation, and (ii) decomposition quality, where Fisher-based estimators suffer from severe rank collapse. In this work, we address these limitations by introducing Layer-wise Error Modeling Search (LEMS) and KFAC-SVD. LEMS advances rank allocation by introducing a layer-wise error surrogate that integrates local and global layer importance alongside a propagation bias, enabling effective global rank allocation via an ILP formulation. KFAC-SVD improves decomposition quality by utilizing token-wise statistics, mitigating the rank deficiency observed in prior Fisher-based SVD approaches. Across Mistral, Qwen3, and Llama 3 model families, we show that LEMS consistently outperforms existing search strategies, delivering significant zero-shot accuracy gains of up to 4.7 p.p. that generalize to model sizes of 70B parameters, while KFAC-SVD achieves an average perplexity improvement of 19%. Project page and code: https://lems-svd.github.io

0.80

On the Identifiability of Hybrid Deep Generative Models: Meta-Learning as a Solution

cs.LG ·2026-10-23 ·Yubo Ye, Maryam Toloubidokhti, Sumeet Vadhavkar, Xiajun Jiang, Huafeng Liu +1 more
top venuecode/datahighly cited
Abstract · NeurIPS 2024. 16 pages, 5 figures

The interest in leveraging physics-based inductive bias in deep learning has resulted in recent development of hybrid deep generative models (hybrid-DGMs) that integrate known physics-based mathematical expressions in neural generative models. To identify these hybrid-DGMs requires inferring parameters of the physics-based component along with their neural component. The identifiability of these hybrid-DGMs, however, has not yet been theoretically probed or established. How does the existing theory of the un-identifiability of general DGMs apply to hybrid-DGMs? What may be an effective approach to construct a hybrid-DGM with theoretically-proven identifiability? This paper provides the first theoretical probe into the identifiability of hybrid-DGMs, and presents meta-learning as a novel solution to construct identifiable hybrid-DGMs. On synthetic and real-data benchmarks, we provide strong empirical evidence for the un-identifiability of existing hybrid-DGMs using unconditional priors, and strong identifiability results of the presented meta-formulations of hybrid-DGMs. Code is available at https://github.com/GoForIt-YYB/Meta-Hybrid-VAE.

0.76

RSIHub: A Composable Framework for Agent-Evolution Methods

cs.AI ·2026-10-23 ·Zimu Wang, Xiaobo Wang, Siyu Ye, Jiaqi Li, Silin Chen +9 more
publishedcode/datacommunity pick
Abstract · NeurIPS 2026 Meta-Agents Workshop

Agent-evolution methods use execution feedback to improve prompts, skills, tools, and orchestration, but their implementations often rebuild the same experimental machinery. We present RSIHub, an open-source framework for building, comparing, and extending these methods on shared infrastructure. Recipes compose reusable operators and declare mutable resources; RSIHub supplies execution, evaluation, recovery, and lineage. Researchers can focus on evolution policy while evaluation stays outside the mutable surface and results remain tied to candidates and evaluation conditions. We instantiate AHE, Hyperagents, A-Evolve, and GEPA without modifying the framework core, keeping method-specific logic in recipes and operators. We evaluate them with MiniSWE and Codex on Terminal Bench 2 and $τ^3$ Banking under fixed model and evaluator configurations. The largest observed overall pass@1 gain is 13.5 percentage points, while method rankings vary across settings and optimization gains do not always transfer to held-out tasks. RSIHub provides a common protocol for comparing evolution policies, inspectable candidate histories, and reusable components for developing new methods.

0.73

PoKER: A Probability of Kill Estimation Rate Model for Air-to-Air Missiles Using Machine Learning on Stochastic Targets

cs.LG cs.AI ·2026-10-23 ·Joao P. A. Dantas, Andre N. Costa, Diego Geraldo, Marcos R. O. A. Maximo, Takashi Yoneyama
publishedcode/datahighly cited
Abstract · Accepted manuscript. 18 pages, 10 figures, 5 tables. Code available at https://github.com/jpadantas/poker

This work introduces PoKER, a novel probabilistic model engineered to optimize missile launch effectiveness in air-to-air scenarios, specifically within Beyond Visual Range (BVR) air combat. Unlike conventional Weapon Engagement Zone (WEZ) models that delineate zones based on static distances such as maximum, minimum, and no-escape ranges, PoKER applies machine learning to predict kill probabilities more accurately by integrating the stochastic behaviors of targets and missile miss distances. This model dynamically evaluates target behavior, greatly expanding the predictive capabilities of engagement analysis. By factoring in elements such as warhead lethality, target and shooter orientations, and the specific conditions of engagement, PoKER provides important insights into engagement dynamics and quantifies success probabilities. Consequently, it can potentially be an important tool for BVR air combat pilots, improving operational decision-making within this specialized combat domain.

0.70

Implicit Neural Representations for Variational Problems on Graphons

cs.AI stat.MLcs.LGmath.CO ·2026-10-23 ·Taeyoung Kim, Jineon Baek, Joonkyung Lee, Hongseok Yang
top venuehot direction
Abstract · Accepted at NeurIPS 2026. 35 pages, 18 figures, 10 tables

We propose a neural exploratory framework for variational problems over graphons, i.e., symmetric measurable functions $[0,1]^2 \to [0,1]$, that arise as limits of dense graph sequences. Such problems are central to two parts of the mathematical literature: extremal graph theory, which studies graph parameter optimisation under given restrictions, often for those graphs with a large number of vertices, and the large deviation theory of dense random graphs, which characterises the structure of rare events through constrained graphon optimisation. We represent graphons by implicit neural networks and optimise graphon objectives by gradient descent. Our design combines three ingredients: a multi-scale sinusoidal residual architecture biased toward sharp, step-like graphons; an embedded solver that enforces a single empirical density constraint and is differentiated by implicit differentiation; and symmetry-aware Monte Carlo estimators. On generalised Turán problems that previously required substantial human effort, our framework rediscovers known optimal graphons without human intervention. Applied to open instances, it produces candidate optima, including a previously unreported family of extremal structures. Applied to the variational problems from the large deviation theory of Erdős--Rényi random graphs, the framework produces new candidate optimal graphons across both upper- and lower-tail regimes for multiple pattern graphs; for the case of the triangle graph and the upper-tail regime, these candidates improve upon the best-known reference construction of Lubetzky and Zhao. The structures found by our framework have also led to new theorems on the optimal graphons for both families of problems.

0.70

IGG: A Benchmark for Interactive GUI Grounding under Visibility Constraints

cs.AI ·2026-10-23 ·Kyeongseon Kim, Jiyeon Son, Tae-Hyun Oh
top venuebenchmarkcode/datahot direction
Abstract · Accepted to EMNLP 2026 (Main Conference). 24 pages. Project page: https://igg-bench.github.io/

GUI grounding benchmarks ask whether an agent can localize a target that is already visible in a static screenshot. Real interfaces often violate this assumption: a target may be off-screen, appear only after a state change or hover, remain occluded, or become actionable after a delay. This leaves the ability to recover target visibility unevaluated. To measure this reveal step, we introduce Interactive GUI Grounding (IGG), a benchmark in which an agent must first make the target visible through interaction and then click it. IGG contains 601 tasks across 16 controllable GUI environments and seven visibility-constraint sub-types, with a compact action space of scrolling, moving, dragging, waiting, and clicking. In our human study, participants achieve 94.6% accuracy, while GPT-5.5 reaches 79.4% and the strongest evaluated open-source VLM reaches only 31.6%. To identify the source of this gap, we conduct two controlled interventions. Telling Qwen3-VL which interaction to perform raises its accuracy only from 30.8% to 33.9%, whereas performing the interaction for the model raises it to 81.9%. The bottleneck is executing the reveal interaction, not recognizing the target once it is exposed. We further show that this capability is learnable, demonstrating that visibility recovery can be strengthened consistently across model families.

0.58

PRISM: Promoting Sampling Diversity of Large Language Models with Implicit Stochastic Modulation

cs.LG ·2026-10-23 ·Qixin Xu, Zihan Zhu, Xiaojiang Liu, Bohan Lyu
top venuebenchmark
Abstract

Large language models (LLMs) exhibit mode collapse, where their output distribution concentrates on familiar semantic modes. When diversity disappears, repeated sampling becomes semantic repetition rather than discovery, making rare but stronger solutions unlikely to surface and leaving self-evolution and continual learning difficult to sustain. Most remedies intervene only at inference time, shaping sam- ples without making diversity a model capability. We propose PRISM, a training-time framework that internalizes diversity within the model. PRISM combines (i) submodular curation that pairs each prompt with diverse-yet-correct multi-teacher solutions, and (ii) stochastic embedding modulation, which turns special-token perturbations into implicit signals for sampling distinct outputs. A theoretical analysis establishes the stability of this mechanism. On WritingPrompts and BookMIA, PRISM moves the quality-diversity frontier outward, achieving the best generation quality in all model-benchmark settings while maintaining the strongest or near-strongest semantic diversity across two open-weight model families. It also improves sampling efficiency on math and code benchmarks. Beyond inference, PRISM yields a stronger foundation for reinforcement learning, sustaining diverse rollouts and mitigating the diversity collapse commonly observed during RL training.

0.58

Anti-coherence Regularized Absolute Scaling for Out-of-Distribution Detection

cs.LG cs.AI ·2026-10-23 ·Donghwan Kim, Youngwoong Jin, Hyunsoo Yoon
Abstract · 29 pages, 6 figures

Out-of-distribution (OOD) detection is recognized as a critical problem in AI safety. SCALE is an effective post-hoc OOD detection method in the activation-shaping family that requires no training data and rescales penultimate activations using a percentile-based magnitude ratio. However, this ratio is ill-defined when negative activations are present, since positive and negative contributions cancel within the underlying sum, causing SCALE to degenerate on backbones such as ViT. We address this with an Absolute Fix that computes the ratio over absolute-valued activations rather than discarding negative activations through ReLU. Building on this corrected scale, we propose Anti-coherence regularization, which penalizes the scaling factor by the fraction of a sample's max-logit contributions to the max logit that are sign-mismatched and thereby suppresses unreliable evidence for the predicted class. Despite requiring no access to training data or training labels, our method ties or outperforms activation-shaping methods that rely on training data in terms of AUROC and FPR@95 on CIFAR-10 and ImageNet under the OpenOOD v1.5 protocol.

0.55

Who Verifies the Tools? How Warped Formatters Mislead Agents and Their Verifiers

cs.AI cs.SE ·2026-10-23 ·William Jeffries
publishedcode/datahot direction
Abstract · Accepted as a poster at the Who Verifies the Agents? workshop at NeurIPS 2026 (non-archival). 12 pages, 3 figures. Code and data: https://gi…

Agents and verifiers often rely on the text a tool prints, and that text can misrepresent what the tool actually did, even when the tool's computation is correct. We saw this happen in a production RAG research system used by the Pilsudski Institute in New York, where tool output formatters misled an agent into giving wrong or incomplete information to users. Each tool did its job correctly, but its output omitted search limits, result caps, or identifier namespaces. We also recreated an unlabeled estimate caught in code review before release. We studied whether these formatters misled agents and verifiers and how much the repaired wording helps. We fed tool results to agents using two formatter versions: one with the key omissions and one with all missing details clearly stated. That difference alone changed what the agents claimed and what the verifiers accepted. Unsupported agent claims dropped substantially for three of the four failure types. In one test, verifiers went from accepting incorrect answers 90 out of 90 times to rejecting them 90 out of 90 times. A control showed the verifiers could catch the errors once they were given the missing facts and asked to judge correctness. We include the audit prompt we used with Codex and Claude Code in an exploratory demonstration, written with prior knowledge of our own system's defects. When pointed at an old version of our own system, it rediscovered the defects that were still live and surfaced twelve more we had not catalogued. When pointed at OpenClaw and Hermes, two of the best-known open-source agentic systems, it found warped formatters in both. We opened PRs to fix both, and the OpenClaw fix has already been merged. Practitioners can run the same prompt on their own systems today.

0.51

Trial by Trace: Benchmarking AI Agent Harnesses for Reliability and Failure Attribution

cs.AI ·2026-10-23 ·Zhiyong Cao, Hengjie Zhu, Yongxin Wang, Xuchen Li, Jintao Li +18 more
community pick
Abstract

LLM agents are increasingly used to solve real-world problems, and their effectiveness depends on both the underlying models and the harnesses that support their operation. However, most existing benchmarks focus on model capabilities and overall agent performance, with limited attention to the evaluation of agent harnesses. To address this gap, we introduce TraceTrial-Bench, which organizes harness reliability into seven components covering context, execution, governance, lifecycle, observability, tools, and verification. We construct 95 expert-level, mechanism-oriented tasks with executable or state-based ground truth to support semantically equivalent comparisons while preserving the native behavior of each harness. We design a Trajectory Tracing (TT) module to capture agent execution traces and introduce a two-layer evaluation framework that separately assesses agent performance and harness reliability. In addition, we design a Multi-source Evidence Acquisition (MEA) module to provide evidence from the agent context, harness, and execution environment for LLM-based evaluation, and develop a Failure Attribution (FA) module that uses explicit mechanism chains to support attribution at the run level. Across 4 harnesses, 9 models, and 95 tasks, our evaluation of 3,420 runs yields macro-averaged harness scores ranging from 59.8% to 72.0%, while 31.1% of runs achieve task success despite failing at least one harness criterion. TraceTrial-Bench provides an auditable and reproducible foundation for comparing agent harnesses, diagnosing their failures, and studying how harness design interacts with model behavior.

0.50

Tree Depth and Sample-Quality-Associated Features: Two Hidden Factors in Boruta Selection for RNA-Seq Data Classification

cs.LG ·2026-10-23 ·Tareque Mohmud Chowdhury, Abu Raihan Mostofa Kamal
Abstract · 19 Pages, 6 Figures, 4 Tables

Random Forest--based selectors such as Boruta depend on tree depth, yet Boruta's most used implementation, BorutaPy, recommends depths of 3 to 7 without published evidence. We tested the effect of depth using pan-cancer data from The Cancer Genome Atlas (10,469 tumors, 33 cancer types) and breast cancer PAM50 subtyping (981 tumors, 5 subtypes). Label-dependent steps were fitted inside training folds, effects were measured against seed-to-seed variation, and each task kept a locked test set. In pan-cancer data, depth 5 changed the selection far more than a new seed did: the top-500 Jaccard overlap with unlimited depth was 0.37, against 0.84 (depth 5) and 0.93 (unlimited) between seeds. Depth-5 trees confirmed 472 instead of 1,659 genes, covered fewer large-class markers (16% vs 62%) and gave less accurate 50-gene panels (balanced accuracy 0.809 vs 0.889). The effect vanished once the cap reached 61 to 76% of the median depth of unrestricted trees, a depth that depends mainly on the number of classes, not sample size. In PAM50 subtyping (median depth 12) the effect was small, and leaving depth uncapped cost no accuracy in either task. It replicated in a second cross-validation repeat and under permutation importance. About a third of pre-filtered candidates correlated within class with the number of detected genes, a sample-quality measure; a within-class filter removed them without loss of test accuracy (0.943 vs 0.942 balanced accuracy, 1,000 genes). We recommend uncapped trees, or a cap above 75% of the realized median depth, and a within-class quality filter before Boruta.

0.49

Shared Is Not Enough: Predictor Width and Factor Readability Under Masked Latent Prediction

cs.LG cs.CVphysics.data-an ·2026-10-23 ·Nilotpal Kakati
Abstract

Masked latent prediction methods such as I-JEPA aim to learn predictable features, but what determines which information they retain is less clear. We built a simulator in which every factor behind an image is known, so a probe can measure how readable each factor remains after training. We trained small encoders (0.7M parameters) with I-JEPA, SimCLR and BYOL on the same pairs of views and probed a factor that both views share. I-JEPA had a predictor with half the encoder's width and depth, the proportions it uses for ViT-B. Its median score was about 40% of SimCLR's. Widening the predictor to the encoder's width reversed this ordering at every training seed, against both other methods. We release the simulator and the trained models.

0.48

Principia: Test-Time Hypotheses to Principle Construction for Scientific Discovery

cs.LG cs.AI ·2026-10-23 ·Dongki Kim, Sangwoo Park, Jinheon Baek, Soyeong Jeong, Sung Ju Hwang
code/datacommunity pick
Abstract

Scientific discovery requires identifying the scientific principles underlying observed phenomena to guide further investigation. As these empirical principles are often unknown in complex discovery processes, they are constructed via a hypotheses-to-principle pipeline to obtain recurring causalities from progressively accumulated observations during test-time discovery. However, constructing this knowledge requires substantial domain expertise, time, and effort, making this translation stage a bottleneck for scientific discovery. To address this limitation, we propose Principia, an LLM-based framework that constructs and continually refines scientific principles at test time. Drawing on how human researchers develop principles, Principia forms comparative causal hypotheses from related inputs and their observed outcomes, then synthesizes recurring relationships across these hypotheses into inductive principles. By applying the constructed hypotheses and inductive principles to the task, it obtains new observations, revises the hypotheses used based on these new observations, and reconstructs principles from the updated hypotheses, allowing scientific knowledge to continuously evolve throughout test-time discovery. Across benchmarks spanning molecular optimization, protein engineering, gene perturbation, and molecular property prediction, Principia outperforms baselines relying on internal knowledge, external sources, or experience summaries, while generating principles that are plausible and consistent with the scientific literature, demonstrating its effectiveness in constructing and refining task-specific hypotheses and principles at test time for scientific discovery. Our code is available at https://github.com/DongkiKim95/Principia.

0.47

Not All Mislabels Are Equal: Efficient Task-Aware Auditing with Confidence Sequences

cs.LG stat.MEmath.STstat.AP+1 ·2026-10-23 ·Fangyuan Lin, Noah Krever, Jakub Černý, Lily Xu
benchmark
Abstract

We study efficient, adaptive auditing of categorical labels in a fixed dataset, motivated by biodiversity-credit programs in which expert verification of species records is costly and downstream scores (e.g., payment) are more sensitive to errors on certain labels than on others. Our proposed ATLAS framework couples anytime-valid inference with a task-aware adaptive auditing policy. Representing label corrections through a confusion matrix, we prioritize which data points to inspect next based on both label uncertainty and the influence of potential label corrections on the downstream score function. For linear scores such as the mislabel rate, our priority-aware approach certifies the overall label error rate to $\pm 2$ percentage points with 24% and 11% fewer audits than random sampling on biodiversity data from Kenya and a widely used CIFAR-10N benchmark, respectively. For nonlinear scores such as the Hill-Shannon diversity, we exploit a linear-envelope representation that reduces inference to a family of linear-score audits; priority-aware sampling reaches the target precision with 36% fewer audits. When distribution-free guarantees are prohibitively costly, we develop a model-based Bayesian alternative that trades assumption-free validity for substantially fewer audits. Our biodiversity case study highlights a gap between statistical validity and practical feasibility: generic methods such as Serfling's bound and existing confidence-sequence procedures are rigorous but can require prohibitively many expert reviews in practice. Together, these results show task-aware auditing can significantly reduce the number of expert audits needed while providing finite-sample, anytime-valid guarantees.

0.45

Trajectory Matching: Discrete Flow Matching with Structured Actions

cs.LG ·2026-10-23 ·Niklas Kemper, Marten Lienen, David Lüdke, Stephan Günnemann
theory
Abstract

The natural units of structured discrete generation are often not independent single-component edits but richer actions: attaching a fragment to a molecule, inserting a node together with an incident edge, or growing a hierarchy by adding a child. Yet most existing discrete flow matching methods (DFM) are restricted to independent component-wise edits over fixed-size states. This restriction stems from endpoint conditioning: endpoints fix the source and target during training but not the action sequence between them, making tractable conditional rates hard to define for complex actions that cannot be factorized component-wise. We propose trajectory matching, which conditions the flow on a full sequence of actions connecting the endpoints. This yields principled, simulation-free training targets for user-defined action spaces, including actions that change state size, modify multiple components, or depend on earlier actions. We prove that the resulting objective recovers the marginal flow and that trajectory matching strictly subsumes factorized DFM, while additionally capturing actions with prerequisite structure such as hierarchy growth or fragment-based molecule generation. Empirically, on tree and molecular graph generation, trajectory matching with domain-aligned actions matches or surpasses factorized DFM baselines (lifting MOSES validity from 92.8% to 99.2%) and transfers without retraining to substructure extension and linking. Trajectory matching enlarges the design space of discrete flow matching by turning the action vocabulary into a modeling choice. An interactive explanation of trajectory matching is available at https://niklaskemper.com/blog/trajectory-matching/.

0.45

Do "True" and "False" Add Up to One? Complement Consistency in LLM-Agent Opinion Simulations

cs.AI cs.CLcs.MAcs.SI ·2026-10-23 ·Anouar Imel, Jin Qi
benchmark
Abstract

In LLM-agent opinion simulations, each agent reports the probability that a claim is true after seeing its own previous probability and the probabilities reported by its neighbors, and that number becomes its state in the next round. If the number describes the agent's position rather than the sentence it was shown, then rewriting the claim as its complement and replacing every probability p with 1 - p should return the complementary answer, so the two reports should sum to one. We test this with three small open models: Qwen3-4B, Llama-3.2-3B and Gemma-3-4B. On 2,304 agent states taken from a running network simulation, Qwen's paired reports sum to 1.18 on average, and on 48 real fact-checking claims from AVeriTeC they again exceed one. A controlled benchmark of 24 worlds, each with exactly two possible outcomes, shows the same failure in all three models. The discrepancy is two to ten times larger than the change produced by a same-meaning paraphrase, and it persists under repeated sampling and across peer group sizes and report strengths. When the reports are fed back into two populations that start from complementary states, the populations do not remain mirror images: their divergence is positive after five rounds for every model. Averaging each answer with the complement of its paired answer restores consistency by construction and lowers Brier loss averaged over the two wordings, but at the final round it beats the better of the two original wordings for only one of the three models. A schedule-by-targeting experiment on the same model gives a small, stable interaction because such sweeps keep the claim wording fixed. Consistency under rewording therefore needs its own test before simulated probabilities are read as beliefs

0.45

Not All Information Is Worth Asking For: Decision-Informed Information Seeking with LLMs

cs.AI cs.CLcs.LG ·2026-10-23 ·Seunghan Lee, Jun Seo, Junhyeok Kang, Jaehoon Lee, Sangjun Han +7 more
code/datahot direction
Abstract

Large language models (LLMs) are increasingly used to support decisions under incomplete information, where deciding what to look up and when to stop matters as much as the final answer. However, LLMs fall into an information trap: They pay for checks, i.e., costly pieces of evidence, that cannot change their decision or are not worth their cost, and they keep asking when the evidence is already sufficient. To this end, we propose Decision-Informed Value-of-information Acquisition (DIVA), a simple yet effective procedure for LLM information seeking based on the decision value of evidence. DIVA 1) lets the LLM specify a small decision model, i.e., how likely a bad outcome is and how each check would shift this probability, through short numeric questions and 2) values each candidate check by how much its imagined outcomes improve the decision net of its cost, stopping when no check is worth its cost. Its variant, DIVA-U, further accounts for the LLM's uncertain estimates of how strongly each check shifts this probability. DIVA is training-free and applies to any LLM. Across four domains and 20 LLMs, DIVA-U has lower regret, i.e., value lost relative to the best possible decision, than every LLM baseline on real-world data, mainly by stopping once evidence suffices, and DIVA cuts the regret of the best LLM baseline by 75% in synthetic scenarios where every check's reliability is given. Code: github.com/seunghan96/diva.

0.44

Correct but Unselectable: The Hidden Interface Tax in Multi-Candidate Reasoning

cs.LG ·2026-10-23 ·Qianli Ma, Zhiqing Tang, Fanshuai Meng, Zhi Yao, Hanshuai Cui +1 more
top venue
Abstract · Accepted at NeurIPS 2026

Small language models deployed under cost, latency, or privacy constraints cannot simply switch to larger models when reasoning fails. Multi-candidate generate-then-rerank offers a natural test-time scaling route. However, more candidates do not always yield higher accuracy. We identify a previously overlooked bottleneck between generation and selection: a correct answer may appear in the candidate pool yet fail to be selected because it is incomplete, unparseable, or misaligned with the answer mode the fixed selector expects. We term this loss the hidden interface tax and formalize a coverage-conversion decomposition that distinguishes raw oracle coverage from selectable coverage, the subset of problems where a correct candidate can actually be consumed by the selector. We propose Compatibility-Aware Candidate Construction (CACC), an interface layer between the candidate source and the fixed selector that repairs fragments, aligns answer modes, and removes selector-confusing noise without retraining either component. On numeric reasoning (GSM8K, competition_math), CACC raises both oracle coverage and final accuracy. On GPQA Diamond, CACC alone improves final accuracy from 5.6% to 19.2% (+13.6 pp); combined with proposer-side strengthening it reaches 25.8% (+20.2 pp). On MMLU-Pro, CACC exposes a coverage-conversion gap: answer-mode-compatible construction raises oracle coverage, but final accuracy decreases by 0.85 pp under the original selector; adding proposer-side strengthening improves final accuracy.

0.44

Markets of Minds: A 51-Billion-State Economy of Intelligence

cs.MA ·2026-10-23 ·Jusheng Zhang, Xuekun Wu, Wenhao Wang, Mark Wang, Ming-Hsuan Yang
highly cited
Abstract

When intelligence is shared, reasoning becomes a constrained allocation problem. In a population of agents with limited access to expert models, an intervention can benefit its recipient yet still be a poor use of shared capacity if another feasible intervention has greater expected value. Assistance must therefore be evaluated through both its marginal benefit and the opportunity cost induced by budgets, access, scarcity, and service capacity. To this end, we introduce Markets of Minds, a framework for allocating expert reasoning across agents before expert outputs are observed. For each feasible agent-service pair, the framework estimates the expected gain from assistance relative to the agent's default continuation, explicitly separating repair benefit from regression risk, and jointly clears assistance contracts under shared resource constraints. We then study this problem in a dependency-aware agent economy with 51.2 billion addressable persona-market states spanning capabilities, cognitive budgets, provider access, scarcity, and network roles. These dependencies couple allocation feasibility across agents rather than the outcome of a fixed contract. Empirically, in a matched replay of 240 cases under a common 30% assistance quota, the frozen allocator reaches 85.42% task-level persona adherence versus 73.86% for equal-capacity random allocation, a gain of 11.56 percentage points. The results show that assistance value is heterogeneous and can be negative, and that scarcity changes which interventions are worth purchasing. Together, these findings motivate a population-level view of intelligence where the value of assistance depends not only on what it can achieve, but also on which alternative uses of scarce cognitive capacity it displaces.

0.43

Functional-Group Knowledge Is Structured in Language Models Before It Is Spoken

cs.AI cs.CL ·2026-10-23 ·Tran Nhat Khang Nguyen, Anh Nguyen, Obed Issakah, Janghoon Ock
code/data
Abstract · 14 pages, 5 figures

Language models serve as the reasoning core of agents for chemistry and materials science, where they must interpret molecular structure before planning syntheses or proposing materials. Yet their outputs show only what they say, not what they know. Given a prompt such as ``The primary functional group in 1-Fluorobutane (CCCCF) is'', a small general-purpose model may answer ``the alkyne group'' instead of an alkyl fluoride. The output alone cannot tell whether the model failed to recognize the fluorine or recognized it and failed to say so. We address this by reading the internal representations of five open-source models on prompts covering 92 molecules and 20 functional groups. Even without chemistry training, the models organize functional groups by chemical family, and they encode chemical differences as consistent directions. The shift from thioether ($\mathrm{R{-}S{-}R'}$) to thiol ($\mathrm{R{-}SH}$), added to ether ($\mathrm{R{-}O{-}R'}$), lands nearest alcohol ($\mathrm{R{-}OH}$). This structure emerges abruptly in the middle layers, where representations shift from tracking the wording of the prompt to tracking the functional group. The base model names the correct functional group in only 58.5\% of prompts, the lowest of the five, yet it encodes functional-group families and preserves chemically meaningful patterns. Its errors are mostly failures to answer rather than wrong answers. Functional-group knowledge is structured in the model before it is spoken. Chemistry fine-tuning reorganizes it, making functional groups rather than prompt wording dominate the representation. These results suggest that evaluating chemical language models through outputs alone can underestimate the chemical information they encode. Representation analysis can distinguish a lack of knowledge from a failure to express it, which matters when choosing or adapting models for chemical agents.

0.43

Do Language Models Trust Their Own Justifications? A Study on Functional Consistency

cs.AI cs.LG ·2026-10-23 ·Alisson Rosa Pereira, Bruno Iochins Grisci
publishedcode/data
Abstract · 6 pages, 2 figures, 1 table. Author accepted manuscript. Published in International Joint Conference on Neural Networks (IJCNN), Maastricht,…

Large Language Models (LLMs) have been widely adopted in text classification tasks, where they not only output class predictions but also generate explanations that highlight the tokens deemed most relevant for reaching the predicted label. Yet it remains unclear whether these highlighted elements are behaviorally consistent with the model's own predictions under targeted interventions, which we refer to as functional auto-consistency. While much of the literature evaluates the textual plausibility of such explanations, few studies assess their functional consistency with the model's actual behavior. In this work, we propose an experimental framework based on the principle of auto-consistency: if a model identifies certain tokens as decisive, then isolating, removing, or semantically inverting them should produce systematic and interpretable changes in its predictions. We operationalize this evaluation through sufficiency, comprehensiveness, and counterfactuality metrics, and conduct experiments on IMDB and Steam reviews across both closed-source (GPT-4o-mini) and open-source LLMs (Gemma3, Granite8B, DeepSeek). Results show that GPT-4o-mini follows the expected progression across all metrics, Gemma3 and Granite8B maintain coherence under sufficiency but lose consistency under more demanding interventions, while DeepSeek variants display structural deviations, either failing to preserve sufficiency or overreacting under comprehensiveness and counterfactuality. These findings show that explanation reliability varies across LLM families and scales, with smaller models displaying contradictions and larger ones exhibiting over-sensitivity. By combining sufficiency, comprehensiveness, and counterfactuality, our approach provides a systematic methodology for assessing the functional consistency of LLM self-explanations.

0.43

Multiplicative vs. Additive State Coupling in Learned Optimizer Schedules

cs.LG cs.AIcs.CLmath.NA+1 ·2026-10-23 ·M. Harper Langston, Pierre-David Letourneau, Dalton Jones, Roland Memisevic, Peter Scott +5 more
Abstract

We learn schedules for a continuous penalty relaxation of MAX-CUT by backpropagating through an unrolled classical solver. A state-blind schedule generalizes to instances substantially larger than those used for training. A state-adaptive bilinear schedule also generalizes out of distribution, outperforming its untrained baseline across 15 seeds ($t=4.1$). By contrast, an additive ablation with identical inputs and output heads is statistically indistinguishable from the bilinear model in distribution but performs worse than its own baseline out of distribution ($t=-11.7$). This separation appears without regularization and persists across a $30$ round weight-decay sweep: the bilinear advantage remains significant at every tested value, whereas the additive model never recovers. The results in this setting isolate how solver state enters the schedule, rather than state adaptivity alone, as the architectural distinction associated with out-of-distribution transfer.

0.42

Strongly Adaptive Online Learning with Time-Varying Movement Costs

cs.LG stat.ML ·2026-10-23 ·Andrew Jacobsen, Emmanuel Esposito, Hao Qiu, Mengxiao Zhang
top venue
Abstract · v2: NeurIPS 2026 camera-ready

We study online convex optimization (OCO) with time-varying movement costs, where the learner incurs both a loss and a time-dependent penalty for changing its decisions. While previous work has established dynamic regret guarantees for this setting, obtaining strongly adaptive guarantees that hold simultaneously over arbitrary intervals remains an open challenge. We address this by developing a new algorithm which, for every interval $[a,b]\subseteq[T]$ and arbitrary comparator sequence, achieves a data-dependent regret bound of $ \widetilde{\mathcal{O}}\left( \sqrt{(D^2+DP_{[a,b]})\sum_{t \in [a,b]}(\|g_t\|^2+λ_t\|g_t\|)} \right) $, where $D$ is the diameter of the decision domain, $P_{[a,b]}$ is the path length of the comparator sequence over $[a,b]$, $g_t$ is the gradient of the loss, and $λ_t$ is the movement-cost coefficient at round~$t$. Key to our approach is a novel reduction to sleeping experts with movement costs, a previously unexplored variant that is of independent interest. As a result of our development, we further obtain strongly-adaptive dynamic regret guarantees under arbitrary delayed feedback with a refined dependence on interval-restricted cumulative delays. Our approach also naturally extends to OCO with time-varying memory, where we obtain strongly-adaptive guarantees which improve the best-known bounds.

0.42

REACT: Representational Early Analysis and Control of Chain-of-Thought Trajectories

cs.LG cs.AIcs.CL ·2026-10-23 ·Hanwen Zhang, Enyi Jiang, Ruike Zhu, Qinan Yu, Sanmi Koyejo
code/data
Abstract · 26 pages, 2 figures

Chain-of-thought (CoT) generation offers an opportunity to intervene before a reasoning model completes its answer. But does the point at which risk is easiest to detect also offer the most effective control? We study this question with REACT, a conditional activation intervention that learns a safety direction from paired safe and unsafe trajectories and applies it once to an early CoT hidden state. The model weights remain frozen, and all subsequent tokens are generated without further edits. To study timing, we train directions separately at each of the first five CoT positions and evaluate ungated edits with full post-edit regeneration. Across three distilled reasoning models and three generation seeds, the strongest gains concentrate in a two- to three-token early window; on DeepSeek-R1-Distill-Llama-8B, the effect falls to approximately zero at the fourth and fifth tokens, even though later states can be more predictive of risk. This separates risk readability from effective control: intervention timing must be evaluated through its effect on subsequent generation. On the same 8B model, REACT raises the rate of fully safe CoT-answer outputs on StrongREJECT-273 from 29.06% to 82.05%, while keeping benign refusal and usefulness near baseline. An anonymized implementation is available at https://anonymous.4open.science/r/REACT_official2-333B.

0.41

Gaussian Approximation for Composition and Subsampling of Differentially Private Algorithms

cs.LG ·2026-10-23 ·Vitaly Feldman, Moshe Shenfeld
Abstract

The privacy analysis of differentially private stochastic gradient descent (DP-SGD) relies on bounds for the composition of subsampled mechanisms. The privacy of such a composition is bounded by that of a pair of product distributions, whose privacy loss is a sum of independent terms and has been analyzed via a variety of notions and concentration results. Yet, existing closed-form bounds either lose a logarithmic factor or require restrictive conditions on the noise scale and delicate analysis of subsampling. Further, while it is well known that sums of independent random variables are approximately Gaussian, existing CLT-based techniques give an approximation with a large error in $δ$. We show that the privacy loss of this pair is approximately Gaussian as long as each step has a small privacy loss except with small probability. Specifically, its privacy loss distribution is close to a Gaussian with the same mean and variance, after truncating the rare large privacy losses. The closeness in both directions is up to a shift of $\varepsilon$ that grows only as $\ln(1/β)$ for an additive error $β$ in $δ$ and is a lower-order term in typical applications. In particular, the composition satisfies Gaussian differential privacy up to a lower-order shift of $\varepsilon$ and the additive error $β$. We also give a simple way to analyze the resulting variance-based bounds for subsampled mechanisms which relies on the $χ^2$ divergence of a single step. For the Gaussian mechanism at sampling rate $1/t$ and $t>\ln^{3}(1/δ)$, this gives the predicted rate for $σ\gtrsim\max\{1,\sqrt{\ln(1/δ)}/\ln(t)\}$, whereas the standard Rényi-divergence analyses need $σ\gtrsim\sqrt{t\ln(1/δ)}/\ln(t)$. We extend these results to random allocation (balls-and-bins sampling), and show that it explains the high accuracy of the connect-the-dots method.

0.41

When Does Pre-Trained Capacity Help Causal Learning? From Scalar Prediction to Counterfactual Post-Training

cs.LG cs.AI ·2026-10-23 ·Qi Xu, Zijie Lei, Zhifei Deng, Zhigang Hua, Luke Simon
published
Abstract · 10 pages. Accepted at the NeurIPS 2026 Workshop: Transitioning from Pre-Training to Post-Training

How does pre-trained model capacity relate to performance after adaptation for causal tasks? We study two systems built on the Qwen 2.5 family: CausalLLM, which estimates scalar potential outcomes with treatment-specific heads, and CausalGen, which generates treatment-conditioned text or code using soft prefixes and factual supervised fine-tuning (SFT), counterfactual SFT, and preference alignment. Increasing backbone size from 0.5B to 1.5B yields mixed scalar-estimation results on IHDP, but improves CodeGen pass@1 from 0.267 to 0.575 and treatment-effect score from 0.231 to 0.515 after the complete generative pipeline. Within 0.5B CausalLLM, narrative inputs reduce mean sqrt(PEHE) from 1.842 to 1.780, a 3.4% improvement over tabular inputs; the 5.8% improvement over a no-LLM baseline combines changes in representation and input format. These descriptive comparisons motivate a distinction between available and convertible capacity. They do not isolate the contribution of post-training: the two systems differ in dataset, architecture, output modality, and supervision, and intermediate-stage evaluations are unavailable. With only two small backbones and three replications, the results support a hypothesis about capacity--adaptation alignment rather than a general scaling law or a demonstrated causal mechanism.

0.41

DRIVE: Fine-tuning via Data Contribution- and Diversity-aware Weighting with Prior Regularization

cs.LG ·2026-10-23 ·Qing Liu, Xinrui Chen, Yi Du, Ou Wu, Weiyao Zhu
top venuetheorycode/data
Abstract · Published in NeurIPS 2026. 39 pages, including references and appendices

Single-stage fine-tuning is pervasive, yet catastrophic forgetting persists when the source pre-training data is inaccessible. Data weighting affords a natural lever to modulate sample influence. Yet existing weighting methods lack a principled mechanism to explicitly arbitrate knowledge retention against new-task adaptation, and provide no unified differentiable objective to accommodate auxiliary signals such as sample diversity and sample characteristics. This is because current schemes privilege one goal over the other, and no explicit gradient-compatible encoding exists for such signals. To address this, we propose DRIVE, a principled and efficient weight assignment framework. It derives per-sample weights via a unified differentiable objective where three coupled signals co-evolve: Shapley additivity splits contribution into anti-forgetting and adaptation components, which are reshaped by a covariance-based diversity term and stabilized by a sample-characteristic prior. Far from a mere combination, this yields an end-to-end gradient-based scheme balancing stability against plasticity. Theoretically, we prove that DRIVE enjoys superior stability-plasticity guarantees over existing weighting schemes and enhances parameter estimation robustness. Empirically, it attains superior overall performance, faster convergence, and stronger robustness. Code at https://github.com/luciabcd/DRIVE.

0.41

Scoped Semantic Invariance as a Foundation for LLM Reliability

cs.AI ·2026-10-23 ·Rosario Uceda-Sosa, Maria Chang, Karthikeyan Natesan Ramamurthy, Moninder Singh
top venue
Abstract · Accepted at EMNLP 2026

LLMs often answer semantically equivalent queries inconsistently, even when the underlying facts are known. We formalize consistency as semantic invariance over equivalence classes of queries and show that invariance under arbitrary meaning-preserving transformations is undecidable in general. This motivates scoped semantic invariance, where model behavior is evaluated relative to bounded semantic domains. We operationalize this framework using knowledge graphs (KGs), which act as explicit semantic control structures defining local semantic constraints and grounding contexts. Across the MITRE D3FEND ontology and eight Wikidata- derived domains, we show that inconsistencies are structured rather than random, concentrating in localized hotspot subgraphs organized around recurring TREE, STAR, and CYCLE patterns that remain stable across models and domains. We further show that selective graph grounding improves correctness and cross-model consistency, while indiscriminate grounding introduces regressions. A layered grounding strategy achieves the strongest monotonicity profile, preserving correct knowledge while reducing inconsistency. Finally, we show that hotspot structures discovered from simple consistency probes transfer to more realistic compositional reasoning tasks.

0.40

The Standard Deviation in VICReg Buys a Safety Margin

cs.LG ·2026-10-23 ·Hawar Dizayee
code/dataunusual
Abstract · 11 pages. Code: https://github.com/hawardizayee/vicreg-hinge

VICReg prevents collapse with a hinge on the standard deviation of each embedding dimension, and its authors argue from the gradient that a hinge on the variance would fail. To our knowledge, this claim has not been tested experimentally. We test it in a controlled setting (ResNet-18, CIFAR-10 and CIFAR-100), changing only the hinge and sweeping its weight $μ$ over five orders of magnitude. At the default weight ($μ= 1$), the two hinges reach the same linear-probe accuracy on CIFAR-10 (about 89.5%). Both collapse below a threshold weight, but the standard-deviation hinge's threshold is 35-130$\times$ lower. CIFAR-100 shows a consistent gap. In training, the variance hinge's gradient fades as the feature variance shrinks, while the standard-deviation hinge's does not until that variance is very small. The standard-deviation hinge therefore buys a safety margin: it keeps training at weights far below the point where the variance hinge collapses.

0.40

Computational De Novo Design of Miniprotein Candidate Neutralizers Targeting the IL-23 Cytokine Subunit p19

cs.LG q-bio.BM ·2026-10-23 ·Raghav Paladugu, Fardina Fathmiul Alam
published
Abstract · Accepted as a short paper at the 2026 IEEE International Conference on Bioinformatics and Biomedicine (BIBM 2026)

Interleukin-23 (IL-23) drives T-helper-17 inflammation underlying inflammatory bowel disease and psoriasis; anti-p19 antibodies are clinically effective but large, costly, and injectable. De novo miniproteins are an attractive alternative, yet published designs target the IL-23 receptor; the IL-23 cytokine itself has not been addressed. A further obstacle is that the standard confidence score (ipTM) tracks construct length rather than interface quality. We computationally design miniprotein binders against the IL-23R-binding face of p19 (PDB 3DUH) and score every design with ipSAE, a length-robust interface metric derived from AlphaFold-initial-guess predicted aligned error. A conventional RFdiffusion/ProteinMPNN/ColabFold pipeline yields no reliable predicted interface across 972 designs (best ipSAE 0.30), and a scoring artifact, in which single-sequence ColabFold under-rates binder pLDDT by approximately 56 points, explains an earlier misleading "binders do not fold" signal. An interface-optimized hallucination method (BindCraft), scored identically, yields 42 of 50 designs above ipSAE 0.60 (best 0.804). An AlphaFold-independent geometric check reproduces the conventional failure: no top-ranked conventional design reaches the buried area or interface hydrogen-bond count of the weakest lead. Five all-alpha-helical leads (72-104 residues) collectively engage all six targeted hotspots; two dock within the intact p40/p19 heterodimer and all show favorable Rosetta energetics. A p40-alone counter-screen returns ipSAE 0.00 for all five, indicating no spurious p40 interface. To our knowledge, these are the first computationally proposed de novo miniprotein candidates designed to occlude the IL-23R-binding face of IL-23 p19. All results require experimental validation.

0.40

WiMoT: Replay-Free Continual Wi-Fi Sensing via Moment Transport

cs.AI ·2026-10-23 ·Chunchun Yang, Yang Cao, Zhitang Chen
code/data
Abstract · 13 pages, 6 figures, 10 tables; supplementary material: 13 pages. This work has been submitted to the IEEE for possible publication. Copyrig…

Continual Wi-Fi sensing must learn new rooms or users (plasticity) while keeping earlier ones (stability), yet sensing nodes often cannot retain channel state information (CSI) for privacy and storage reasons. We propose WiMoT (Wi-Fi Moment Transport), which stores the per-class feature mean and covariance of every earlier domain instead of samples. Stored statistics become stale when the representation changes, so WiMoT restricts the change to an affine map on a frozen feature basis and applies the same map to the statistics, which then stay valid without old samples. Supervised contrastive learning learns the new domain, a closed-form loss on the stored statistics keeps earlier classes separable, and a Gaussian discriminant classifier refitted from the transported statistics predicts without domain identity. On 3DO, WiMANS, and Widar3.0, WiMoT reaches average final accuracies of 94.02%, 89.61%, and 95.16% with at most 0.58 percentage points (pp) of forgetting. On 3DO and Widar3.0 it is 3.77 and 13.11 pp ahead of the strongest of 21 baselines, which uses replay; on WiMANS it matches the strongest replay baseline in accuracy while reducing forgetting from 2.78 to 0.00 pp. Rank-32 covariance factors shrink the moment memory to 6.4% at a cost of at most 0.49 pp.

0.40

Counting the Bandages: On measuring the time that AI saves

cs.AI ·2026-10-23 ·Solomon Shalom Lijo
code/dataunusual
Abstract · 18 pages, 2 figures. Replaces the author's earlier TSB proposal (SSRN 6682338). Code and data: https://github.com/solomonshalom/tsb-bench. A…

I consider what it means for AI assistance to save time. I contrast a stopwatch view, on which time saved is whatever the clock records, with a ledger view, on which it is a counterfactual that has to be read next to the errors the work lets through, and I argue for the latter. The argument has four parts. I define a model-free pair of quantities, time saved and excess escaped defects, and one decision number built from them: the break-even defect cost. I give a baseline model of checking in which, until the model is good enough not to check, a calibrated worker's escaped-defect rate does not depend on how good the model is; checking time grows with the logarithm of the model's error rate; and over-trust buys minutes on the clock at a larger cost in total minutes. I show what randomized studies can and cannot identify, and I replicate METR's developer study from its public data: the same 246 issues support a slowdown of 19% or 24% depending on the estimand (34% without adjusting for a chance imbalance), and AI helped on somewhere between 3% and 82% of them. Finally, I restate the published record, with stated bounds in place of imported point assumptions. In GDPval's own review-every-output workflow, review time alone caps GPT-5's speedup at 3.6x; the consultants outside the frontier were faster, and more often wrong. Last, I test two of the model's predictions in a small pre-specified experiment on AI reviewers checking AI drafts of the same kind of brief, built from SEC filings made after the models' training data were collected. Told the drafter's true error rate, the reviewer checked more as it rose, but all-or-nothing, switching near where its prompt's costs break even. Told that the drafter erred ten times less often than it did, it checked 7 fewer of 9 figures and let a wrong figure through in 22.5% of briefs instead of none: faster and worse, entirely through reduced effort.

0.40

How Much Fresh Data Should CTR Recalibration Use? A Temporal Evaluation of Post-hoc Calibration

cs.LG cs.IR ·2026-10-23 ·Rui Liu
code/data
Abstract · 10 pages, 3 figures, 6 tables. Code and per-cell results: https://github.com/RuiLiu0129/ctr-recalibration-temporal

Offline evaluation of click-through rate (CTR) models still defaults to random splits, although deployed models score future traffic and downstream systems consume calibrated probabilities. On Avazu (40.4M impressions) we train logistic regression (LR), DeepFM, and DCN-v2 under two rolling temporal origins and a random split, three seeds per arm, and recalibrate each test day with Platt scaling and isotonic regression fitted on the preceding hours. At origin 1's nearest test day, random splits report LR's expected calibration error (ECE) at 55-62% of the temporal value in every seed (the deep models' ratio is seed-unstable), and no model's ECE is monotone in temporal distance. With a 6-hour fitting slice, recalibration removes up to 61% of ECE in badly miscalibrated cells but increases it in about a third of applications (14-17 of 48 across seeds). On fixed evaluation sets, a 12-hour slice cuts harm from 13-17 to 3-5 of 36 applications, and 24 hours raises it again. The 12-hour slice subsampled to the 1-hour slice's size (5% of the impressions; 20-25x faster to fit) recovers 88% of the full slice's gain, so impression count alone does not explain it. A slice-ECE gate shows no benefit out-of-origin, and in a simulated auction recalibration moves profit by only about -0.21% to +0.26%. Offline protocols should measure calibration temporally, over multiple origins and seeds; in this data, recalibration fitted on a wide time span beat the freshest few hours, which other logs must confirm.

0.40

When Agents Fail: Model or Harness--Who's to Blame?

cs.AI cs.CLcs.SE ·2026-10-23 ·Shao Zhang, Hangfan Zhang, Yuanyuan Shi, Hao Li, Min-Le Su +7 more
benchmarkposition
Abstract · Preprint under review

LLM agents combine models with execution harnesses, so agentic benchmark performance reflects their joint behavior. We argue that harnesses should therefore be evaluated through their interactions with underlying models. We develop a framework combining five complementary measures: (i) Valid Completion Rate (VCR), how reliably agents finish normally and deliver evaluable results, irrespective of task correctness; (ii) Model-Harness Fit, how well a harness supports task achievement through valid completion for a given model; (iii) Cross-Model Fit, a harness's overall suitability across evaluated models; (iv) Interaction Residuals, pairing-specific execution strengths and weaknesses hidden by component-level summaries; and (v) Token Load, the token demands of agent execution, including failed attempts. We evaluate four general-purpose harnesses (OpenCode, Codex CLI, Pi CLI, and DeepAgents) with four core models, plus two fixed native model-harness systems, on six benchmarks covering context management, tool use, long-context planning, multimodal interaction, and artifact handoff. Including native references and stress conditions, 111 evaluation settings yield 52,168 trajectories. Harness rankings vary across models and benchmarks; task performance and valid completion can favor different pairings, and similar task scores can accompany distinct execution profiles. OpenCode ranks first in cross-model performance, while DeepSeek V4 Flash Preview maintains consistently high Mean VCR across harnesses on the shared four-benchmark panel. These findings support evaluating model-harness interactions, training transferable harness-use capabilities in models, and harness designs adapted to the model and task.

0.39

Async-RBench: Benchmarking Asynchronous Task Execution in LLM Agents

cs.AI ·2026-10-23 ·Detao Zeng, Guilin Ou, Nuo Chen, Yunhao Feng, Hongliang Wang +3 more
benchmarkcode/data
Abstract · 25 pages, 16 figures. Code is available at https://github.com/pope8zdt/Async-RBench

Large language model (LLM) agents increasingly execute tasks asynchronously: a main agent continues working while its subagents run, and the task state may change before their results arrive. The main agent must decide which work to update or preserve and which checks to repeat, while avoiding prohibited changes. Final task success alone does not reveal whether these results were handled correctly, and natural asynchronous arrivals are hard to reproduce. We introduce Async-RBench, a benchmark for asynchronous task execution in LLM agents, with 200 executable tasks covering eight scenarios. A scenario controller regulates result delivery using predefined task-state conditions, while an Event Response Contract (ERC) specifies what each response must change, preserve, avoid, and re-verify. The Dynamic Replanning Score (DRS) weights the response process and outcome equally and is reported separately from task success. We evaluate nine agents and find DRS scores of 19.2 to 45.1, with the lowest scores on duplicate reports and subagent failures. Under our execution protocol, all nine agents obtain lower task success with asynchronous than with batched delivery (26.0% vs. 29.7% on average). For three agents, prompting the main agent to rebuild its plan after every result improves DRS by only 2.6 to 3.5 points. Together, these findings highlight the challenges of reliable asynchronous task execution and the need to evaluate how agents handle incoming results alongside end-to-end task success.

0.39

Where Rounding Happens Matters: Rounding-Location Alignment Stabilizes LLM Reinforcement Learning

cs.LG cs.AI ·2026-10-23 ·Yijing Ke, Tianle Cai, Ruosong Wang
Abstract

Reinforcement learning for large language models often uses separate systems for inference and training. Even with identical weights and BF16 tensors, implementation differences can make an intended on-policy update off-policy and destabilize training. Existing remedies align the entire inference and training stacks, change global precision, or modify the learning algorithm. We ask whether aligning selected parts of the execution stack can stabilize BF16 RL while other implementation differences remain. We align training-side rounding locations with inference in three regions---RoPE, SwiGLU, and residual RMSNorm---by retaining selected intermediates in FP32 until their BF16 outputs. Across four Qwen-family models, aligning all three regions prevents the collapse observed under the default BF16 baseline; RoPE-path alignment alone also suffices in the studied runs, despite remaining numerical discrepancies. The interventions keep the inference engine, matrix-multiplication and attention kernels, and learning objective unchanged.

0.39

Evidence-Grounded SLM Behavioral Prediction Architecture for Future Assistance Use

cs.AI ·2026-10-23 ·Adeyemi Adeseye, Aisvarya Adeseye, Hannu Tenhunen, Jouni Isoaho
top venue
Abstract · Accepted for publication at the 2027 International Conference on Artificial Intelligence and Applied Intelligent Systems (AIAIS 2027), to be…

Users often exhibit changes in interaction behavior before explicitly seeking assistance, enabling systems to anticipate future assistance use. This is especially important in high-stakes environments, where delayed assistance can have operational, financial, or safety consequences. However, conventional Large Language Models (LLMs) are often unsuitable for continuous on-device behavioral prediction because of their computational and memory demands. Behavioral patterns are also highly individual and device-specific, motivating local processing that preserves privacy and data sovereignty. Small Language Models (SLMs) offer a promising alternative, but their ability to reason reliably over low-level sequential interaction events remains underexplored. This study introduces Evidence-Grounded Local Behavioral Prediction (ELBP), a four-layer architecture combining deterministic interaction processing, dual-view behavioral representation, SLM-based prediction, and evidence verification with abstention. Three instruction-tuned models of 3B parameters or fewer were evaluated using raw, structured, dual-view, and complete ELBP configurations. The best-performing model, Qwen2.5-3B with ELBP, achieved a full-set F1-score of 0.80 and PR-AUC of 0.84, representing a 15.9\% relative improvement over raw-event prompting and 17.6\% over XGBoost, while full-set ECE decreased by 43.2\% relative to raw-event prompting. With evidence verification and abstention, selective F1 increased to 0.84 at 91\% coverage and evidence support reached 96\%. These results show that structured behavioral representation and evidence-verified selective prediction can improve the reliability of behavioral forecasting in SLMs.

0.38

Earth embeddings are decomposable records of the land surface

cs.LG cs.CVq-bio.PE ·2026-10-23 ·Robin Young, Srinivasan Keshav
Abstract

Ecological monitoring, carbon accounting and biodiversity assessment increasingly rely on annual Earth embeddings--learned vectors that compress a year of satellite observation for every ten-meter pixel on Earth, consumed in place of the imagery itself. What these vectors mean has not been established. Here we decompose two widely used embedding products and show that they contain individual ecological features, including fire, forest clearing and crop type, validated against independent reference data on events across six continents. The recovered concepts date abrupt canopy loss to the first post-disturbance growing season and resolve severity well enough to separate prescribed fire from wildfire. Small fires that recover quickly are detectable in one product but not the other, which instead separates crops more strongly.

0.38

Dual-Adaptive Z-Score PPO: Dynamic Clipping Using Action Confidence and Normalized Advantage Magnitude

cs.LG cs.AI ·2026-10-23 ·Taha M. Mahmoud, Naima Kaabouch
published
Abstract · 6 pages, 4 figures. This is the author's accepted manuscript of the paper published in the 2026 6th International Conference on Electrical, …

Proximal Policy Optimization (PPO) remains a popular on-policy reinforcement learning algorithm owing to its stability and ease of implementation. Its clipping threshold, however, is fixed, so every training sample receives the same trust-region width regardless of how reliable the corresponding policy update is. This work examines adaptive-clipping variants of PPO in which the constant clipping parameter is replaced by sample-dependent thresholds computed from policy and return statistics. The central method, Dual-Adaptive Z-Score PPO (PPO-DAZ3), sets the clipping interval for each sample from two signals: action confidence, given by the probability of the selected action under the current policy, and update confidence, given by the normalized advantage magnitude. The rest of the standard PPO training pipeline is left unchanged. Evaluation on three classic control benchmarks (LunarLander-v3, CartPole-v1, and Acrobot-v1) indicates that adaptive clipping improves learning in some settings but not all. PPO-DAZ3 performs best on LunarLander-v3, the smooth advantage-based variant PPO-AZ3S leads on Acrobot-v1, and standard PPO stays competitive on CartPole-v1, suggesting that adaptive clipping is most beneficial on harder control tasks.

0.37

DMAIC-Bench: From Self-Reflection to Self-Restraint for Agents in Multi-Stage Analytical Tasks

cs.AI cs.CLcs.LG ·2026-10-23 ·Shan Peng, Shuaiqi Cheng, Miao Peng, Chenyi Zi, Jia Li
code/data
Abstract · 54 pages, 14 figures. Code and data: https://github.com/psptpx/DMAIC-Bench-public

Multi-stage analytical tasks require LLM agents to use intermediate evidence to decide whether to continue, redirect or stop. Self-reflection, in which an agent critiques its output, is widely studied, yet a critique does not show whether evidence changed the next action. We call that capability self-restraint. Existing agent benchmarks largely measure task completion or output quality and leave it untested. DMAIC-Bench evaluates it with 97 seeded tasks across 15 industrial cases in four sectors, built around Lean Six Sigma's explicit statistical prerequisites. Paired counterfactual variants require opposite actions at the same workflow position, distinguishing warranted restraint from indiscriminate stopping. We assess execution traces with a private scorer and final claims through independent expert review. Across 13 models from six providers, execution-level routing accuracy averages 92% on matched GO decisions and 19% on NO-GO decisions, counting any forbidden downstream computation as a route error. Execution traces show agents reporting correct diagnostics before performing analyses that those results disallow. Separately, blinded expert review of the 11 audited agents estimates that reports act on invalid evidence at 47% of reached NO-GO decisions, with an interval from 31% to 61%. Explicit stopping permission helps some models without consistently restoring correct decisions at both GO and NO-GO forks. Sixteen life-sciences tasks in bulk ribonucleic acid sequencing (RNA-seq) and case-control genome-wide association studies (GWAS) show that the same framework can evaluate evidence-dependent decisions beyond industrial workflows. Project code and data are available at: https://github.com/psptpx/DMAIC-Bench-public.

0.36

RxnExBench: Benchmarking Vision-Language Models for Structured Chemical Reaction Extraction from Scientific Literature

cs.AI ·2026-10-23 ·Jiaxing Sun, Junyuan Gao, Ruikang Zhang, Hao Zheng, Lingli Ge +14 more
top venuebenchmarkcode/data
Abstract

Screen Tables and Substrate Scopes are information-dense figures widely used in synthetic chemistry literature, yet automated extraction of structured reaction data from them remains challenging. We introduce RxnExBench, a multimodal benchmark for fine-grained reaction-level evaluation. Built from high-impact chemistry journals, it contains 3,565 expert-annotated reactions and 2,361 molecular structures. The dataset adopts a unified schema with 15 reaction roles and 26 core fields, supporting downstream AI-for-Chemistry tasks such as retrosynthetic planning, yield prediction, and reaction condition recommendation. RxnExBench further provides a reasoning-driven difficulty taxonomy, a modular annotation pipeline, and the Oracle-MolRec protocol, which disentangles molecular recognition from higher-level reasoning. Evaluation of 9 state-of-the-art VLMs shows that performance degrades substantially as schema complexity and reasoning demands increase. Under the Oracle-MolRec setting, results further reveal that accurate molecular recognition alone is insufficient, with multimodal integration and chemistry-aware reasoning remaining the primary bottlenecks. We release all data and code to facilitate future research in chemistry-aware multimodal understanding and AI-for-Chemistry dataset construction. The dataset and code will be released at https://github.com/opendatalab/RxnExBench.

0.36

SemantEq: Faithful Bidirectional Equivalence of Formal Statements

cs.AI cs.LG ·2026-10-23 ·Xinrui Li, Pawan Sasanka Ammanamanchi, Siddharth Bhat, Rianna Jamdar, Yongxi Lin +1 more
code/data
Abstract · 37 pages, 4 figures, 6 tables. Code and data: https://github.com/hlxr89/SemantEq

Statement autoformalization translates mathematics from natural language into a proof assistant such as Lean 4, and models now do it at scale. As provers get stronger, the burden of correctness shifts from the proof to the statement, that is, to whether the formal statement faithfully translates its source. Given a trusted reference formalization, the natural test is logical equivalence, and it proves too much. $P \leftrightarrow Q$ holds trivially for any two statements $P$ and $Q$ that the library proves, because a prover can prove each separately and pass through $P \leftrightarrow \mathrm{True} \leftrightarrow Q$. The stronger the prover, the more unrelated pairs pass. Earlier methods restrict the prover and check that each proof names the other statement. We introduce SemantEq, a more principled relation that instead requires each statement to turn into the other only through a chain of allowed steps, such as logical rearrangement and rewriting with the library's own equations. Our programs judge a proof by the proof term Lean checks, not by its tactic script. SemantEq-Eval, a search without a model, reaches 100% precision and 65.6% recall on the ProofNetVerif test split under our adjudicated labels, against 95.8% and 61.8% for BEq+, and an 8.9-fold increase in its CPU time adds at most three pairs. A language model finds more proofs, but under two of four prompts it proves both directions of about two in five of the non-equivalent pairs. SemantEq-ProofAudit audits each model-written proof under a policy derived from the relation and accepts none of the 104 non-equivalent pairs, and it raises recall on 407 equivalent pairs from 65.4% to 81.1% under the best prompt. Separating the generation of equivalence proofs from their acceptance lets a stronger prover raise recall without changing what an acceptance means. We release code and data at: https://github.com/hlxr89/SemantEq

0.36

A Tutorial on Physics Constrained Machine Learning for Hybrid Modeling, Optimization, and Control

cs.LG cs.ET ·2026-10-23 ·Rahul Golder, Bimol Nath Roy, MM Faruque Hasan
benchmarktheorycode/data
Abstract

Physics-constrained machine learning (PCML) integrates domain knowledge, first-principles, and physical constraints into AI/ML models. In this work, we present a step-by-step tutorial of a modular PCML framework, KKT-HardNet, and illustrate how it could be used for hybrid modeling, process simulation, optimization and model predictive control in a single platform. KKT-HardNet combines fast inference capabilities of neural networks with guaranteed satisfaction of algebraic constraints through differential projection. It can handle heterogeneous data and enforce both linear and nonlinear equality and inequality constraints. The modular design supports custom neural-network backbones and user-defined projection mechanisms for PCML. To enable efficient end-to-end training, KKT-HardNet exploits implicit function theorem to backpropagate through the projection layer. This avoids computationally expensive automatic differentiation. This work also introduces an exact reduced-order Fischer-Burmeister reformulation and a primal-dual interior-point method as an alternative projection solve strategy. We also provide a diverse set of benchmarking problems for PCML in engineering applications. KKT-HardNet achieves orders of magnitude better performance in terms of both constraint satisfaction and inference time. The KKT-HardNet code and the benchmark datasets are available at https://github.com/souls-tamu/kkt-hardnet .

0.36

Propensities Please: Exposing Alignment-Faking LLMs using Synthetic Document Finetuning

cs.AI cs.CLcs.LG ·2026-10-23 ·Ardy Haroen, Florian Lorkowski, Samuel Nellessen, Felix Hofstätter
positioncode/data
Abstract · 27 pages, including references and appendices. Code: https://github.com/ardyh/propensities-please

Alignment evaluations aim to assess whether large language models would engage in dangerous misaligned behaviour. Misaligned and evaluation-aware models might engage in alignment faking, where they pursue a hidden goal, but conceal this behaviour when they suspect effective oversight such as during training, evaluation or audited deployment. Such a model might view any prompt-based intervention that aims to reveal latent propensities as part of the evaluation, calling into question the reliability of existing alignment evaluations. Current methods for detecting alignment faking rely on chain-of-thought monitorability, highly realistic environments that reduce evaluation awareness, or mechanistic interventions, all of which are expected to become less reliable for future model generations. We present a method for revealing latent misaligned propensities that does not rely on any of these. We use synthetic document finetuning to teach models that misaligned actions will go undetected whenever certain keywords appear in the prompt. We place alignment-faking model organisms in environments that are clearly identifiable as honeypots and measure how their behaviour changes after finetuning. We observe that finetuning consistently widens the gap in misaligned action rates between prompts with and without the cue, indicating that finetuning allows exposing latent propensities more reliably than purely prompt-based methods. We argue that behavioural methods for detecting alignment faking and related behaviours are critical going forward, as models become more capable and less interpretable. We find that SDF often damages the models' reasoning capabilities and introduce a mitigation based on on-policy distillation.

0.35

MPQ: A Message-Passing View of Post-Training Quantization

cs.AI cs.LG ·2026-10-23 ·Yanshuo Chen, Chenxi Liu, Reza Shirkavand, Heng Huang
publishedcode/data
Abstract · NeurIPS 2026

Post-training quantization (PTQ) is a key technique for efficiently deploying large language models, as it compresses weights into low-bit representations without retraining. Existing PTQ solvers, including GPTQ, LDLQ, and block coordinate descent, commit each weight block to a single codeword before its neighbors are considered, leaving near-optimal alternatives unexplored. In this work, we cast row-wise PTQ as posterior inference over discrete codebooks, in which existing solvers correspond to a zero-temperature, hard-decision limit retaining only the top codeword per block. We propose Message-Passing Quantization (MPQ), which preserves soft posterior information across blocks during inference and commits to hard codewords only at the output stage. MPQ first uses vector approximate message passing to compute approximate blockwise posteriors, then refines a hard-decision initialization by exact block coordinate descent. Across 2-bit lattice and 3- and 4-bit scalar quantization on dense language models from Llama, Qwen, and Mistral families, MPQ improves perplexity and zero-shot accuracy over strong hard-decision baselines, and composes favorably with preprocessing techniques such as activation-aware scaling and incoherence processing. Our implementation is available at https://github.com/poseidonchan/MPQ.

0.35

Curvature-Aware Clipping: Mitigating Bias in DPSGD via Approximate Second-Order Information

cs.LG cs.CR ·2026-10-23 ·Xincheng Xu, Thilina Ranbaduge, Thierry Rakotoarivelo, David Smith
published
Abstract · Accepted at the 18th International Workshop on Optimization for Machine Learning (OPT 2026)

Differentially Private Stochastic Gradient Descent (DPSGD) has become the de facto standard for training deep neural networks with rigorous privacy guarantees. However, the utility of DPSGD is constrained by gradient clipping bias and DP noise. While existing work has tried to address these issues, the geometric misalignment caused by clipping bias remains a critical challenge. To bridge this gap, we propose a novel framework, Curvature-Aware Differentially Private Stochastic Gradient Descent (CA-DPSGD), which leverages approximate second-order information to construct an optimal, curvature-aware clipping space. CA-DPSGD integrates a rectified estimator for the curvature matrix and is coupled with an anisotropic noise mechanism to ensure privacy guarantees. Extensive empirical experiments demonstrate that our approach outperforms state-of-the-art clipping methods.

0.35

Beyond Gain: Measuring Skill Internalization for Agent Skill Retirement

cs.AI ·2026-10-23 ·Yanze Wang, Pengfei Yao, Tianyi Sun, Hongda Chen, Yunyun Han +3 more
code/data
Abstract · 12 pages. Code: https://github.com/EverMind-AI/SkillCorpus/tree/main/skillcorpus/lifecycle Data: https://huggingface.co/datasets/EverMind-AI…

LLM agents retrieve skills from public hubs that hold up to about 100,000 entries and serve many models. As models improve, a hub must find the skills a given model has internalized, meaning it can do without them, and retire them for that model, yet internalization has not been measured with known reliability. Published rules that compare runs with and without a skill retire on the gain alone. The gain is near zero both for an internalized skill and when neither the model nor the skill delivers the capability, a state we call uncovered. We instead withhold the skill, score the model on a task synthesized from it, and trust a high unaided score only when another model gains from the skill, since nearly half of high scores come from tasks no model benefits from. Under four models, a published gain rule applied with one task per skill removes over half of the uncovered skills. On SkillsBench, 28 of its 31 retirements rest on probes no run passed in full, while retiring on the unaided score costs no detectable reward. Unaided verdicts reproduce across tasks more often than the gain rule's (Cohen's kappa 0.42 against 0.23) but differ across models, so a hub must read them per model. We release 1,251 skills with tasks and item scores under four models.

0.35

LGMT: A Tool for Testing the Logical Reasoning Reliability of LLMs

cs.AI ·2026-10-23 ·Zenghui Zhou, Xiaoke Fang, Zheng Zheng
benchmarktheorycode/data
Abstract

Large Language Models (LLMs) increasingly serve as reasoning components in modern software systems, yet a model that answers a logical query correctly can contradict itself when the same problem is restated in an equivalent form. Static benchmarks evaluate queries in isolation and therefore fail to expose such inconsistencies. To address this limitation, we present LGMT, an open-source tool that detects reasoning inconsistencies through metamorphic relations grounded in first-order logic. Given a source test case in premise-conclusion form, LGMT generates logically equivalent follow-up cases and checks whether the model's entailment predictions agree across the pair, without ground-truth labels. The tool provides 20 built-in metamorphic relations, supports custom ones verified by the Z3 theorem prover, and runs from a CLI or a web studio, including an offline mock mode requiring no API key. Across six widely used LLMs on a bundled benchmark of 3,091 metamorphic groups, LGMT reveals inconsistency rates ranging from 20.84% to 29.44%. LGMT is publicly available on GitHub at https://github.com/MaxwelsDonc/lgmt-tool, with a demonstration video at https://youtu.be/gou4opptu1U.

0.34

RobustRLlib: A Unified Library and Benchmark for Robust Reinforcement Learning Algorithms

cs.LG cs.AIcs.RO ·2026-10-23 ·Xi Chen, Fengqi Hu, Zeyu Shen, Jiaying Xu, Shangding Gu +3 more
benchmarkcode/data
Abstract · Project page, code, and tutorial: https://robustrl-lib.github.io/

Robust reinforcement learning (RL) is critical for reliable real-world deployment, where policies may fail under diverse shifts and disruptions. Existing robust RL algorithms offer a rich set of mechanisms and principles, yet two major barriers limit their broader potential: limited algorithm reusability and an unclear empirical capability boundary. We introduce RobustRLlib, an algorithm-centric library and benchmark that addresses both challenges. It integrates 17 representative robust online, offline, and safe RL algorithms under a unified interface for substantially improved reusability, together with an enhanced evaluation platform spanning diverse real-world shifts and advanced humanoid and vision-language-action tasks. Systematic evaluation over 273 perturbation configurations provides an up-to-date capability map of current robust RL algorithms, revealing where they remain effective and where substantial robustness gaps persist. As an open-source, user-friendly platform, RobustRLlib offers a reproducible and extensible codebase for applying existing methods to new problems and developing new robust RL algorithms.

0.34

Empirically Calibrated Differential Games for Simulating Driver Interactions

cs.MA math.OC ·2026-10-23 ·Amirmohammad Khakpour, Nachuan Li, Pedram Beigi, Yu, Nie
published
Abstract · This manuscript is intended for submission to IEEE Transactions on Intelligent Transportation Systems (IEEE T-ITS)

Game-theoretic models can resolve vehicle interactions without hand-written priority rules, but whether they describe human drivers once calibrated has rarely been tested. This study formulates a general-sum differential game of vehicle interactions at urban intersections, calibrates it to observed driving, and evaluates it on data not used for calibration. Each vehicle minimizes a cost that combines speed tracking, adherence to a map-derived lane route, control effort, gap keeping based on the intelligent driver model, a predictive conflict term, and a contact barrier, while perceiving the other vehicles with a delay. The game is solved with a feedback iterative linear-quadratic algorithm, and ten behavioral parameters per vehicle are calibrated by Bayesian optimization on trajectories recorded at four intersections in Washington, D.C. The calibrated game reproduces the trajectories of 819 vehicles with a mean position error of 1.00 m. When only each vehicle's route is known, as in a simulator, it is the most accurate of the compared models on 96 held-out vehicles (7.42 m), and ten-fold cross-validation shows that it is significantly more accurate than calibrated SUMO (by 3.61 m), an inverse reinforcement learning model (by 0.93 m), and independent and cooperative variants of the same model (by 0.33 and 0.32 m). The solutions, computed in 2.1 s for a typical interaction, are verified approximate equilibria. Finally, the conflict and steering weights, which the data do not identify, can be set from safety and comfort targets; together with a stiffer contact barrier, this removes the collisions that otherwise occur in 43% of tight closed-loop conflicts.

0.34

Optimal Correction Sets for Provable Redress of Contestable Causal Discovery

cs.AI ·2026-10-23 ·Fabrizio Russo, Giuseppe Mazzotta, Carmine Dodaro, Francesco Ricca, Francesca Toni
position
Abstract · 41 pages including appendix

Constraint-based causal discovery relies on conditional-independence (CI) tests that may conflict in finite samples, leaving no causal graph that satisfies them all. Causal Assumption-based Argumentation (ABA) makes these tests explicit as defeasible commitments, and offers correspondence guarantees with a subset of them, but its existing repair procedure releases facts heuristically and offers no formal response to a user's contestation of a release. We argue that contestable causal discovery requires an inspectable repair object, a justification for it, and a predictable revision procedure for such contestations. We therefore formulate CI-fact repair as a minimum-cost correction problem and implement it in Answer Set Programming (ASP). Building on Causal ABA, we provide a compact and efficient ASP encoding of causal consistency, with soundness and completeness guarantees, and a two-way projection correspondence between optimal stable models and minimum-cost correction sets. These results establish fidelity to retained facts, subset-minimality and cost optimality of the repair, conflict-based justifications for released facts, and formally guaranteed redress under user contestations. An empirical evaluation supports the formal account: OptABA-PC retains more correct facts, reduces ambiguity in the compatible outputs, and improves graph reconstruction over Causal ABA on common benchmarks.

0.33

Agent-Based Transport Model Calibration via Online Stochastic Batch Optimization

cs.MA ·2026-10-23 ·Abdelkader Dib, Milos Balac
published
Abstract · Presented at the World Conference on Transport Research (WCTR 2026), Toulouse, France

Calibrating agent-based models (ABMs) is computationally demanding, as traditional black-box optimization typically requires hundreds or thousands of independent simulation runs. To reduce this burden, we introduce an online calibration framework that estimates parameters within a single simulation run. The approach combines black-box optimization with stochastic batch optimization, a method widely applied in machine learning but not yet explored in ABM calibration. Using MATSim as a case study, we treat each simulation iteration as a stochastic batch, enabling incremental parameter updates as agents make decisions. This online calibration paradigm allows model parameters and agent behaviors to co-evolve, achieving convergence without restarting the simulation. We demonstrate the effectiveness of our method by calibrating utility function parameters in MATSim for a large-scale transportation scenario of Switzerland. Results show that mode shares align closely with observed data within a typical simulation timeframe, confirming the feasibility and efficiency of the approach. By embedding calibration directly into the simulation loop, our method opens new possibilities for real-time, scalable, and efficient calibration of complex ABMs.

0.33

Probing the limits of forward screening and inverse design of crystal materials

cs.LG cond-mat.mtrl-sci ·2026-10-23 ·Zhilong Song, Zijun Pang, Lixue Cheng
hot direction
Abstract · 23 pages, 5 figures

Crystal materials are designed either by screening known crystals for a target property or by generating new crystals with that property. Both routes are judged against public databases of known crystals, but these databases name the same crystal differently. Several materials announced as new by recent generative and autonomous discovery campaigns have since proved to be known phases or close variants of them. The limits of both routes remain unmeasured. Here we probe them by matching the crystals of 55 databases by structure into one self-extending database of 14.6 million crystals and 131.8 million property values, each traced to its source. Only 4.5% of its crystal-property pairs carry a value. To fill the empty pairs, we developed CGTNet, an equivariant graph transformer with state-of-the-art accuracy on several tasks of two public benchmarks. Trained on each task, it forms a library of property models whose predictions we release. The reach of forward screening has been underestimated. Known crystals hold tens of thousands of ultra-incompressible candidates without a recorded bulk modulus, and a post hoc audit of synthesis evidence finds among them ReN2, whose bulk modulus was measured at 428 GPa. The novelty claimed for inverse design has been overestimated. At an equal verification budget, a stiffness-conditioned generator returned more stiff hits than screening (100 versus 55), but all contained osmium and 82 repeated recorded compositions. Integrating and completing the databases of known crystals therefore moves the limits of both routes: it widens what screening can reach and sets the reference against which generated crystals are judged new.

0.33

Agentic Hazard Mapper: Emulating Expert Map-Making with AI Agents for Risk Communication

cs.AI ·2026-10-23 ·Sidi Wu, Lifeng Lin, Yizi Chen, Anna Yokokubo, Lorenz Hurni
Abstract

Hazard maps are essential tools for public risk communication, preparedness, and evacuation planning, yet effective cartographic design is not always ensured due to limited access to cartographic expertise across local governments and communities. We present Agentic Hazard Mapper, a multi-agent system that automates hazard-map production by emulating professional cartographic workflows and iterative design trade-offs while adapting design decisions to different data, locations, and hazard contexts. We conducted a formative study with map experts to inform the system design, followed by a user study and interviews with map experts and hazard practitioners. The generated maps improved participants' spatial understanding and evacuation-planning performance, reduced perceived workload and task completion time, and were generally preferred over corresponding municipal maps. We further discuss human oversight, adaptation across users and communication contexts, and opportunities to extend agentic support across broader stages of hazard-map production and risk communication.

0.32

GS-MoE: Training-Free MoE Compression via Groupwise Subspace Sharing and Pruning

cs.LG ·2026-10-23 ·Carson Cary, Yukai Jin, Yihao Huang, Zhankai Ye, Fangyi Li +3 more
code/datahot direction
Abstract

Mixture-of-Experts (MoE) models reduce computation through sparse activation, but storing their expert parameters remains a major memory bottleneck. We introduce GS-MoE, a training-free compression framework that addresses redundancy across and within experts through two compression axes: shared-subspace rank and expert intermediate width. Groupwise low-rank representations capture shared structure across experts while retaining expert-specific projections, and Nyström-based neuron compression reduces intermediate width within each expert. For fixed shared subspaces, we derive per-expert reconstruction-error reductions from the reconstruction objective and use them to guide expert grouping. Budgets are allocated hierarchically using a gradient-free layer-importance score to guide the allocation of rank and width budgets across layers, which eigenvalues and leverage then distribute as non-uniform ranks across groups and intermediate widths across experts. Experiments demonstrate the complementary benefits of compression along both axes, particularly at high compression rates on fine-grained MoE models. On DeepSeek-MoE-16B-Base, at a 60% reduction in total model parameters, GS-MoE retains approximately 90% of the uncompressed model's average reasoning performance, compared with 73% for D$^2$-MoE, while achieving 20% lower out-of-domain perplexity. We provide a code implementation at: https://github.com/cbacary/GS-MoE

0.32

Exploring Joint-Embedding Predictive Architectures for Self-Supervised Representation Learning of Powder X-ray Diffraction Patterns

cs.AI cond-mat.mtrl-scics.LG ·2026-10-23 ·Ming-Chiang Chang, Brian DeCost
publishedhot direction
Abstract · 8 pages, 6 figures, 1 table. Accepted by AAAI 2026 Fall Symposium Series Workshop "Parsimonious Scientific Machine Learning: Accelerating Sc…

Powder X-ray diffraction (XRD) is the primary high-throughput method for resolving crystal structures, a cornerstone of materials science. As high-throughput synthesis and self-driving laboratories advance, automated XRD pattern analysis has become vital for accelerating materials discovery. However, traditional matrix factorization methods fail when lattice variations cause non-linear peak shifts. At the same time, supervised learning approaches face a combinatorial explosion when handling multi-phase mixtures. To overcome these challenges, we adopt a self-supervised Joint Embedding Predictive Architecture (JEPA) to learn representation embeddings from 1D diffraction patterns. By optimizing a masked self-prediction loss directly within the latent space, the framework captures high-level crystallographic semantics. Downstream evaluations show that lightweight attentive probes trained on frozen JEPA representations achieve space group classification and lattice parameter estimation accuracy close to much larger supervised vision transformers. Furthermore, latent probing reveals that the learned representation space naturally organizes physical transformations, isolating smooth trajectories for peak broadening. Interestingly, decoding along latent paths between single-phase crystal representations yields realistic multi-phase mixture patterns, demonstrating an emergent continuous structure learned without explicit multi-phase supervision.

0.32

BiLoCo: Binary Low-Rank Corrections for Post-Training Quantization

cs.LG ·2026-10-23 ·David Jin, Beshr IslamBouli, Tarushii Goel, Han Guo, Yoon Kim
Abstract

Post-training quantization (PTQ) to FP4 has emerged as a key technique for reducing large language model inference cost, especially on Blackwell GPUs, which offer native FP4 tensor core support. Direct PTQ to FP4, however, still leaves a noticeable accuracy gap. Recent work on SVDQuant mitigates the errors introduced through FP4 quantization for diffusion models by adding high-precision low-rank corrections obtained via SVD. However, this high-precision low-rank correction requires many bits per rank, and it runs on the same tensor cores as the main FP4 GEMM. We introduce BiLoCo, a low-rank correction method whose left and right factors are stored as binary ${\{\pm 1\}}$ sign vectors rather than as BF16 numbers. The signed format uses $16{\times}$ less storage per rank-one component, and since the correction now consists of additions/subtractions, this can be run on CUDA cores concurrently with the FP4 tensor-core GEMM. On 4B--72B LLMs, BiLoCo matches or improves memory-matched SVDQuant for PTQ. We implement BiLoCo on B300 as a fused single kernel, where the FP4 GEMM runs on tensor cores, the bit-packed correction on the CUDA cores of the same launch, and the two are combined in the GEMM epilogue. With batch size $1$, the fused kernel is at most $11\%$ slower than the FP4-only GEMM for every rank up to $512$. At rank $128$, it is at most $20\%$ slower up to batch size $8$.

0.32

Rethinking Representation Geometry in Supervised Contrastive Learning under Long-Tailed Data

cs.LG cs.AIcs.CVstat.ML ·2026-10-23 ·Joonhyeok Jeong, Jy-yong Sohn, Kibok Lee
Abstract

Supervised contrastive learning and its balanced variants have demonstrated their effectiveness for representation learning under long-tailed distributions. However, although prior analyses suggest that their optimal representations may exhibit a regular simplex geometry, whether such a geometry is indeed exhibited remains insufficiently understood. In this work, we present a theoretical framework for these objectives that characterizes whether a regular simplex is formed at the optimum. By applying this to representative objectives, we show that these objectives fail to form a regular simplex at the optimum because they remain dependent on the number of samples per class. Motivated by this analysis, we introduce structural balancing that can be applied to contrastive objectives, and propose a fully balanced contrastive objective that provably forms a regular simplex geometry at its optimum. Both theoretical and empirical results demonstrate that the proposed method yields a more balanced representation space and improves performance on long-tailed recognition and transfer learning tasks.

0.32

WL-SeqVAE: Sequence-Based Graph Generation Using Variational Autoencoders and Weisfeiler-Leman Colors

cs.LG ·2026-10-23 ·Iakovos Evdaimon, Giannis Nikolentzos, Achilleas Tsimichodimos, Christos Xypolopoulos, Michalis Vazirgiannis
code/datahot direction
Abstract

In this paper, we introduce WL-SeqVAE, a sequence-based variational autoencoder for generating graphs. The key observation behind WL-SeqVAE is that node colors induced by the Weisfeiler-Leman (WL) algorithm provide compact discrete summaries of local graph structure and can therefore serve as structural tokens for generation. Rather than learning a distribution over adjacency matrices directly, our method maps each graph to a sequence of WL colors and employs a variational autoencoder that encodes this sequence into a Gaussian latent distribution, from which an autoregressive decoder reconstructs the graph. During generation, compatibility constraints derived from the WL colors guide node and edge predictions, combining global information from the latent representation with local structural consistency. This formulation turns graph generation into a sequence generation problem and avoids directly parameterizing a quadratic adjacency matrix. We evaluate WL-SeqVAE on molecular and general graph generation tasks. Across the evaluated datasets, WL-SeqVAE is competitive with established baselines on several local structural and molecular quality metrics. Code is available at https://github.com/iakovosevdaimon/WL-SeqVAE

0.32

Distilling Expert Judgment at Scale

cs.AI ·2026-10-23 ·Robbie Goldfarb, Andrew B. Hall, Jillian Fisher, Kathryn Salam, Matt Wilde
benchmark
Abstract · 9 pages

Frontier AI models are increasingly deployed on topics where the stakes are high and expert judgment is required, including healthcare, defense, finance, governance, and politics. We present a general methodology for distilling expert human judgment into automated evaluation systems that operate at scale. Our approach assembles networks of domain experts and develops techniques for extracting not just their conclusions but their reasoning processes, encoding those processes into structured editorial standards that automated judges can apply continuously. We ground our approach in a long tradition of research on the nature of expertise and the conditions under which expert judgment is reliable and more effective than other approaches, and argue that independent, expert-calibrated evaluation is necessary not only on epistemic grounds but as a matter of accountability and public trust. We validate the methodology on AI news reporting as a first application domain. On source quality, two independent domain experts achieve Krippendorff's $α= 0.83/0.69$ on a held-out test set when applying the editorial standards, and our automated judge reaches $F1 = 0.81/0.97$ against expert consensus. On neutrality, experts achieve $α= .72$ on the test set, and the automated judge reaches $F1 = 0.86$ against expert consensus. Additionally, a controlled comparison shows that expert-calibrated judges substantially outperform uncalibrated frontier models on every source-quality metric, demonstrating that encoding expert reasoning is the critical step that prompt engineering alone cannot replace. We present this benchmark as a proof of concept for a broader evaluation infrastructure applicable to any domain where judgment matters and expertise is scarce.

0.32

Sustainability of TSFM Deployment and Pretraining: From Carbon Footprint Analysis to Mitigation via Pruning

cs.LG ·2026-10-23 ·Cécile Luc, Nathanaël Lemonnier, Ievgen Redko
publishedbenchmark
Abstract · 26 pages, 6 figures, accepted at the NeurIPS 2026 Workshop TS-LIMITS

Time Series Foundation Models (TSFMs) offer strong zero-shot forecasting capabilities, but their high pretraining cost raises a practical sustainability question: does this upfront investment pay off over a full deployment lifecycle, compared to supervised standalone models that must be periodically retrained? We model the cumulative carbon cost of different approaches across a realistic industrial scenario: Chronos-2 amortizes its pretraining cost within months for large or frequently queried deployments, but a classical model remains cheaper indefinitely for small or rarely queried ones. For cases where reducing the TSFM's own cost matters, we show that removing a single encoder block (7.9% of parameters) via a gradient-free pruning criterion yields a consistent 6 to 9% reduction in FLOPs, CO2 emissions, and latency, at essentially no cost to zero-shot forecasting performance across a 25-task benchmark.

0.32

PlayerEmbed: Learning Behavioral Fingerprint Representations from Raw Input Sequences

cs.LG cs.AI ·2026-10-23 ·Tobias Schneider, Lorenz Sparrenberg, Priya Tomar, Shahzeb Qamar, Tobias Deußer +1 more
published
Abstract · Accepted at IEEE DSAA 2026

Raw player input sequences in video games contain fine-grained behavior, but are high-dimensional and noisy, which makes large-scale analytics challenging. To address this, we introduce PlayerEmbed, a framework that learns compact and discriminative player embeddings using deep metric learning. Our approach uses a hybrid CNN-LSTM encoder trained on raw input sequences from 1,000 players. We evaluate the model on held-out sequences from these players and on sequences from 1,000 players entirely unseen during training to assess generalization to unseen identities. For player identification, a nearest-centroid classifier built from reference embeddings achieves strong performance, reaching balanced accuracies of 0.950 for players seen during training and 0.879 for players unseen during training. Systematically increasing the number of training players substantially improves generalization to unseen players, indicating a transition from simple fingerprinting to learning transferable behavioral patterns.

0.31

Can Test-Time Memory Evolution Outperform Test-Time Training for Scientific Discovery?

cs.AI cs.LG ·2026-10-23 ·Yepeng Liu, Shengmai Chen, Xuandong Zhao, Chengzhi Liu, Dawn Song +2 more
Abstract

Recent \emph{test-time training} (TTT) frameworks have enabled open-weight large language models (LLMs) to outperform inference-only frameworks powered by frontier models on scientific discovery tasks. However, TTT introduces additional computational cost for online weight updates during the discovery process, making computational efficiency an important consideration when scaling to larger models. This motivates us to explore an efficient \emph{test-time adaptation} method for scientific discovery. We introduce \emph{Test-Time Memory Evolution} (TTME), a plug-and-play test-time adaptation method, as an alternative to TTT. TTME continually updates the external memory by distilling the experience to guide subsequent discovery without updating model weights. To have a clean comparison, we evaluate TTME and TTT under the same discovery framework, replacing only the test-time adaptation method. Our evaluation shows that TTME outperforms TTT on all five mathematical tasks, while reducing computational cost by $82.8$-$84.6\%$ in FLOPs and $47.7$-$58.4\%$ in runtime. These results establish TTME as an effective and computationally efficient alternative to TTT in the evaluated discovery settings.

0.31

AI Teaches Humans to Handle Complexity in the Game of Go

cs.AI ·2026-10-23 ·Adamo Cerioli, Vito D. P. Servedio, Gabriele Cerioli, Antonio Villani, Carlo Metta
Abstract

Strategic decision-making in complex environments requires balancing local stability against global coordination. Using the game of Go as a bounded empirical testbed, we investigate how biological and artificial agents navigate this trade-off by formalizing a network-derived metric of strategic tension defined via the spectral radius of spatial dependency matrices. Across human skill progression, tension systematically decays from novices to masters as uncoordinated tactical skirmishes give way to disciplined territorial stabilization. However, a longitudinal analysis of professional play spanning 1980--2026 reveals a striking historical paradox: following the superhuman breakthrough of AlphaGo in 2016, elite human matches display a marked surge in sustained tension. A two-dimensional projection based on move distance and captures places modern professionals close to neural engines, separating their tension from the local contact fighting of novices and from random noise. Consequently, strategic tension acts as a macroscopic order parameter for decision paradigms rather than raw computational strength. Beyond the game of Go, these findings establish a constructive paradigm for human-AI interaction, indicating how artificial systems can act as cognitive scaffolding that expands biological heuristics and elevates human capacity to navigate higher-order complexity across professional training and decision-making domains.

0.31

Interpretable sparse fine-tuning with TopKLoRA enables faithful circuit discovery

cs.LG cs.AIcs.CL ·2026-10-23 ·Marek Masiak, Andrzej Szablewski, Nicola Cancedda, Constantin Venhoff
code/data
Abstract · 29 pages, 10 figures

Fine-tuning on datasets of unknown origin can create a backdoor in the trained model when the data includes poisoned examples. Difficult to localise and remove, backdoors can persist through safety training. While parameter-efficient fine-tuning methods capture the full parameter changes introduced during training, they are not easily interpretable. We present the full method of TopKLoRA, extending the preliminary proposal of Masiak et al. (2025), an interpretable-by-design LoRA variant in which only the $k$ largest adapter latents are active at each token. Unlike SAEs, TopKLoRA is trained purely on downstream loss, where optimisation pressure incentivises learning functionally causal latents. We demonstrate this sparse LoRA adapter in a sleeper-agent backdoor case study where we create model organisms by fine-tuning gemma-2-2b, Qwen2.5-1.5B, and Qwen2.5-7B on a poisoned instruction-following dataset. For each organism, we identify and ablate a circuit that is both necessary and sufficient to implement the backdoor. Furthermore, our new fine-tuning method produces sparse circuits that are up to an order of magnitude smaller than a dense LoRA baseline, allowing surgical ablation of the malicious capability without sacrificing instruction-following performance. Finally, we propose a technique to construct model organisms with localised circuits using TopKLoRA latents and gradient-routing, which we use to validate the findings of our circuit search algorithm. We consider these findings as a step toward interpretable-by-design architectures that provide faithful units of analysis without compromising model performance. We share our code: https://github.com/interpretable-finetuning/TopKLoRA.

0.31

SWE-Shift: Evaluating Coding Agents in Evolving Workspaces

cs.AI cs.CLcs.LG ·2026-10-23 ·Victor Shea-Jay Huang, Rongqing Li, Zhangyue Yin, Kelong Mao, Yaqi Zhao +5 more
benchmarkcode/data
Abstract · 47 pages. Code: https://github.com/JeixHuang/SWE-Shift

Repository-level coding benchmarks largely assume that tasks and environments remain unchanged throughout execution. In software development, code, dependencies, runtime state, and requirements can change while an agent is still working. We introduce SWE-Shift, a benchmark for evaluating coding agents under such mid-execution changes. From the same saved history and workspace, a Clean continuation proceeds without intervention, while a matched Shifted continuation receives a task-relevant update. Success requires completing the current task, preserving still-valid work, and satisfying the applicable state and evidence checks. The benchmark contains 120 scenarios across 12 change families, each evaluated at nine scheduled intervention positions. We evaluate 17 coding-agent systems, all of which have lower Shifted than Clean success. GPT-6 Astra drops from 67.35% to 24.79%, with 20.34% paired success. Tool/session changes and stale evidence yield zero paired success across all systems. Late-position paired success is lower for some systems and higher for others. These results show that success on the unchanged task can coexist with failures to preserve work, recover state, or refresh evidence after an update.

0.31

Vision Token Pruning via Query-Vision Interaction Decomposition

cs.LG cs.CV ·2026-10-23 ·Harshithanjani Athi, Sravan Kumar Ankireddy, Jianzhong Charlie Zhang, Hyeji Kim
top venue
Abstract · 10 pages for the main paper, 8 figures including supplementary material, accepted at NeurIPS 2026

Large vision-language models (VLMs), which process both visual inputs and text queries, incur high inference costs due to the large number of visual tokens, making visual token pruning a natural approach for improving efficiency. Recent work has leveraged query information to guide token selection, often through heuristics derived from attention patterns or token dynamics. However, these approaches do not explicitly exploit the underlying structure of query-vision interactions. We address this gap by observing that the query-vision interaction matrix, naturally induced by the query and key projections already present in transformer attention, is effectively low-rank, with a small number of dominant latent interaction modes capturing query-relevant semantics. Motivated by this observation, we propose Query-Vision Decomposition (QViD), a novel training-free query-aware visual token pruning method that exploits this structure. Extensive experiments on both image and video understanding benchmarks show that QViD consistently improves performance under matched token budgets, with the largest gains appearing under aggressive compression regimes.

0.30

PAFS: Perturbation-Aware Firing Scaffolds for Robust Event-Based SNNs

cs.NE ·2026-10-23 ·Ning Liu, Jun Sun, Hao Guo, Sen Shen, Xiangfei Yang +4 more
Abstract · 20 pages, including supplementary material

Event-based spiking neural networks (SNNs) encode information through temporal firing dynamics, but the same stateful computation makes their predictions sensitive to event loss, timestamp jitter, and speed variation. These perturbations alter evidence accumulation, yet late-layer firing trajectories can retain error-diagnostic structure that is absent from the final logits alone. We introduce Perturbation-Aware Firing Scaffolds (PAFS), a labeled post-training calibration method that exploits this structure without updating the pretrained backbone. PAFS associates standardized late-layer leaky integrate-and-fire trajectories with the class-probability residuals of an Adapter anchor, then applies singular value decomposition to extract compact residual-aligned channel directions. At inference, firing trajectories from the same backbone pass are projected onto this fixed low-rank scaffold and combined with anchor statistics to predict a residual logit correction. Across four matched dataset-backbone settings spanning three event datasets and two spiking Transformer architectures, PAFS improves corrupted-average and worst-condition accuracy over the corresponding Adapter anchors, with maximum gains of 3.06 and 3.65 percentage points. In the DailyAction-DVS/Spikingformer setting, the added trainable parameters amount to approximately 0.34% of the frozen backbone.

0.30

Neural-DISCO: Source-Conditioned Counterfactual Editing of Neural Population Activity

cs.LG q-bio.NC ·2026-10-23 ·Nico Policzer, James Ho, Joseph Soo, Xin Tang
publishedcode/data
Abstract · NeurIPS 2026 poster

Understanding how neural activity encodes behavioral and sensory variables is a key challenge in systems neuroscience. This requires models that go beyond predicting activity from labels and instead support source-conditioned counterfactual editing: given a recorded source response, we should be able to edit one measured behavioral or stimulus label and predict how neural activity would change while preserving other observed labels and source-specific variability not explained by the labels. Here, we introduce \textbf{Neural-DISCO}, a disentangled representation learning framework for counterfactual modeling of neural activity. Neural-DISCO partitions the latent space into label-specific components, each associated with an observed behavioral or stimulus label, together with a residual latent that captures variability not supplied through observed labels. Across simulated and real neural recording datasets, we show that this structure enables selective editing of measured labels to generate counterfactual neural responses while preserving the remaining content of the original activity. Further, to gain insight into how behavioral and stimulus labels are encoded at single-neuron resolution, we pair this disentangled framework with feature attribution methods that identify which neurons are most important for each label. We find that disentanglement enhances the recovery of known functional cell roles in both synthetic and real data. Together, these results support Neural-DISCO as a framework for counterfactual modeling of neural activity and provide a practical interface for probing how neural populations encode behavior and sensory information. Code is available at https://github.com/TangLab-UBC/Neural-DISCO-neurips.

0.30

Coral Discovery: The Division of Cognitive Labor in Autonomous Research

cs.MA ·2026-10-23 ·Xinxing Ren, Qianbo Zang, Ziyan Wang, Suman Deb, Qiang Yang +7 more
Abstract

A research community is only as strong as the way it divides its cognitive labour. The philosopher Philip Kitcher argued that researchers should not all pursue the most promising approach. Credit rules, not a central plan, keep some of them on other lines. We bring this question to autonomous research, where Large Language Model agents improve programs against a deterministic grader. Agents that share only a score follow the leader and converge onto one approach. Coral Discovery gives the community a direction board. A verifier labels every attempt with a direction, the mechanism earning the score, and a method, one change within it. Per direction, the board tracks how much each eval improves the score, adds an exploration bonus for little-tried directions, prices the option of opening a new direction, and suggests how many agents each direction deserves. Credit is earned against a direction's own best, so a lagging direction keeps its agents. Agents finish their lines, then treat the board as advice. We test on nine tasks with the same model and budget, without grader access except for Kernel. Coral Discovery beats the strongest open autonomous multi-agent system on 6 tasks, ties on 1, and loses on 2. Against published numbers, it is best or tied on 4. It improves the reported Kernel result from 1103 to 1095 cycles and passes 1103 at eval 275, under half of the 596 used before. It sets a new Polyominoes state of the art, 91.78 against 87.0. Direction labels show that the leading scores rest on directions a score-only community never entered. Code will be released shortly.

0.30

CARES: Coverage-Aware Replay Exemplar Selection for Continual Learning

cs.LG ·2026-10-23 ·Arundhati S. Shanbhag, Brian B. Moser, Federico Raue, Tobias C. Nauen, Andreas Dengel
Abstract · 17 pages, 2 figures. Includes supplementary material

Replay-based Continual Learning mitigates catastrophic forgetting by keeping a buffer of past samples, making buffer composition critical under a fixed memory budget. Recent coreset-based replay methods improve buffer quality using model-derived gradients, losses, or auxiliary optimization, but incur substantial computational overhead as memory is repeatedly updated. We propose CARES, a continual model-optimization-free approach for buffer construction that instead exploits the geometry of a frozen pretrained feature space using class-aware, density-weighted coverage. We evaluate CARES across balanced, imbalanced, and label-noisy streams under online class- and task-incremental settings. On Tiny-ImageNet, CARES is 2.3 to 8.8 times faster than existing coreset-based selectors. Although CARES achieves this speedup using no training signal from the continual learner, it ranks among the state of the art across balanced, imbalanced, and label-noisy streams under online class- and task-incremental settings, outperforming the strongest baseline by up to 7.58 percentage points under class imbalance while also attaining higher average accuracy on Tiny-ImageNet. Code will be released upon publication

0.30

CFBrainAug: Modularity-Induced Counterfactual Brain Network Augmentation for Interpretable fMRI-based Autism Diagnosis

cs.LG ·2026-10-23 ·Yifan Qiao, Sheng Li
Abstract

Machine learning on resting-state functional magnetic resonance imaging (rs-fMRI) brain networks is promising for diagnosing Autism Spectrum Disorder (ASD), but existing methods are prone to overfitting on limited training data, and their post-hoc statistical attribution offers limited clinical interpretability. We propose CFBrainAug, a counterfactual brain network learning framework that unifies data augmentation and counterfactual explanation by learning label-altered brain networks. Unlike conventional label-preserving augmentation, CFBrainAug first retrieves a label-opposite yet topologically similar candidate and then refines it under a module-number constraint. Jointly training on factual and counterfactual networks enriches the training distribution around the decision boundary, enabling a lightweight feature-selection-based SVM to perform competitively with recent graph deep learning methods. Meanwhile, the counterfactuals provide intrinsic explanations by identifying the connectivity differences between each subject and its counterfactual. Experiments on the ABIDE I dataset show that CFBrainAug reaches an accuracy of 74.1% on the NYU site, suggesting that counterfactual augmentation can complement more complex models in small-sample brain network analysis.

0.30

Monetizing Intelligence: Advertising and Subscription Design on LLM Platforms

cs.AI econ.TH ·2026-10-23 ·Yuhang Ji, Fasheng Xu, Tianjun Feng
Abstract

Large language model (LLM) platforms introducing advertising must decide how much capability ad revenue should fund and what consumers should pay to avoid ads. We study a monopolistic platform that jointly chooses costly response quality, advertising intensity, and subscription prices when advertising reduces user value but better service raises advertising returns. Three findings follow. First, when partly ad-supported and ad-free plans coexist, advertising can finance useful service for consumers who receive none under pure subscription. Lower effective quality in middle plans weakens the alternatives available to premium users, allowing higher prices for unchanged ad-free service. Some fully ad-supported users gain utility, while partly ad-supported and ad-free users lose: broader access can coexist with utility polarization. Second, an intermediate advertising cap can maximize consumer surplus by preserving some revenue that finances service while limiting degradation. Third, cheaper compute can raise quality and reduce ad intensity, yet sufficiently large cost reductions can increase aggregate advertising revenue above its initial level because better service raises the value of the remaining advertised interactions. An extension shows how advertising-funded upgrades can bring plans to the same maximum capability and advertising exposure, eliminating the basis for different prices. These results make menu design central to evaluating LLM advertising: the consequences extend beyond users who see ads to the alternatives and prices faced by those who do not.

0.30

Exact Global Minimax Risk near Nonidentification

cs.LG stat.MLmath.ST ·2026-10-23 ·Kaiyue Yang, Yunan Wu
Abstract

We study squared-error estimation when a nuisance parameter is unidentified on a submodel but affects a target that vanishes there. For models with higher-order profile expansions, we give sufficient conditions under which the largest critical-path minimax risk is the exact asymptotic minimax risk over the full parameter space. An efficient observable feature, negligible calibration error and stable recovery yield one estimator attaining the leading constant uniformly, with the profile coordinates unknown. The estimator uses calibration estimates, a local feature mean and an empirical recovery feature. The constant is expressed through a bounded-normal decision problem. We verify the conditions in two models. In centered Gaussian logistic errors-in-variables regression with one surrogate and two unknown variances, we determine the fifth-order global coefficient risk and its exact constant. A sparse extension separates variance-decomposition uncertainty from direction estimation and specifies when the scalar constant is preserved. In a degree-balanced replicated-network model, we determine the sixth-order global edge-contrast risk when block proportions are unknown. Triangles identify a parameter combination; a bowtie supplies the additional information needed for the contrast.

0.29

FOCUS: A Lightweight Mamba-2 Model for EEG-Based Cognitive Workload Classification

cs.AI ·2026-10-23 ·Ioannis Kyprakis, Stefanos Gkikas, Chariklia Chatzaki, Zongyi Jiang, Randy Gomez +1 more
Abstract

Cognitive workload estimation from physiological signals is increasingly relevant in high-risk environments, where continuous and unobtrusive monitoring can inform adaptive systems and safety interventions. This paper presents FOCUS (Frequency-band Optimized Cognitive-load Ultralight SSM), a compact Mamba-2-based model for EEG-based binary workload classification under constrained data conditions. Rather than learning frequency sensitivity from data, FOCUS encodes preprocessed multi-channel EEG using a fixed FFT decomposition that covers the five canonical EEG bands. The proposed model combines depthwise-separable convolution and a weight-shared encoder-bottleneck Mamba-2 pair, with 0.04 million parameters and 0.11 GFLOPs. Evaluated in a pilot study with 11 participants under leave-one-subject-out cross-validation across three cognitively distinct tasks, FOCUS achieves 74.75% accuracy on the full dataset at 0.08 ms GPU model-forward latency. In the weight-sharing ablation, the shared configuration achieves 4.55 percentage points higher accuracy on the full dataset than the non-shared variant while reducing the SSM parameter budget; however, this difference does not reach statistical significance under a two-sided Wilcoxon signed-rank test ($p=0.083$). Overall, the selected FOCUS configuration provides a favorable trade-off between classification performance and computational cost within this controlled pilot setting.

0.29

AdaPRA: Process Reward Agents with Adaptive Rubrics for Data Science

cs.AI cs.LG ·2026-10-23 ·Atharva Naik, Vivek Chari, Pol Llado, Daniel Fried, Carolyn Rosé +3 more
benchmark
Abstract

Data science agents can automate end-to-end machine learning workflows, yet they remain far from competitive on human leaderboards. Agents make process errors such as insufficient exploration, unsound validation, coding bugs, and poor use of a fixed budget (re-running the same exploratory data analysis and baseline models in every attempt). We propose AdaPRA to address these issues: a persistent coding agent harness that accumulates knowledge across attempts on the same task, with rubric-guided process supervision that identifies and corrects errors. Our process reward agent audits the agent's trajectory and workspace by executing checks, and adapts its rubrics to each task. We find that rubrics seeded from easier tasks generalize to harder tasks, with rubric adaptation further improving performance. With five attempts per task, AdaPRA improves an 80B Qwen model's leaderboard percentile by 24.5 points on DSPredict-Easy and 14.8 points on DSPredict-Hard while reducing process errors. We also show that supervision with rubrics mined from one base model's runs improves another (DeepSeek-V3.2), and that AdaPRA beats DataPRM, a competitive PRM for data science. We also share our code for reproducibility.

0.29

CoPiT: A conditional Position-induced Transformer for aerodynamic field prediction on unstructured meshes

cs.LG physics.comp-phmath.NAphysics.flu-dyn ·2026-10-23 ·Junfeng Chen, Yongchen Fan, Zhonghua Qiao, Kailiang Wu
Abstract

High-fidelity Reynolds-averaged Navier--Stokes (RANS) simulations remain a standard tool for aerodynamic analysis, yet each solve on an unstructured mesh is computationally expensive, making many-query tasks such as design exploration, optimization, and multi-condition evaluation costly. Neural operators provide mesh-flexible surrogates for mapping geometry and operating conditions to flow fields; however, in many existing architectures, global parameters such as the freestream state and design variables are concatenated with pointwise inputs, leaving their influence on intermediate features implicit. This paper proposes CoPiT, a \textbf{Co}nditional \textbf{P}osition-induced \textbf{T}ransformer that builds on the Position-induced Transformer (PiT) as its geometric backbone and injects a global condition vector through multi-stage feature-wise linear modulation (FiLM), thereby making parametric modulation explicit while preserving geometry-based attention weights. The model is evaluated on two-dimensional AirfRANS airfoil fields, including scarce-data and extrapolation regimes, and on three-dimensional BlendedNet++ blended-wing-body surfaces. On the AirfRANS full-data test set, CoPiT reduces surface and volume losses by approximately 54\% and 24\% relative to Transolver, respectively, and achieves the lowest overall normalized test MSE among the compared baselines on BlendedNet++. Matched ablations show that FiLM conditioning generally improves upon input concatenation, while the preferred placement is task-dependent. CoPiT thus provides a practical operator-learning surrogate for parametric aerodynamics on unstructured meshes.

0.29

On the Limits of Sparse Linear Model with Shuffled Differential Privacy

cs.LG ·2026-10-23 ·Yan Zhang, Jiong Tuo, Meng Ding, Yuheng Ma, Di Wang
theoryhot direction
Abstract · 39 pages, 1 figure. Under review

Sparse linear regression is a fundamental problem in high-dimensional statistics, and its private counterparts leave a large gap between the central and the local model of differential privacy. Shuffled differential privacy (SDP) sits between the two, offering accuracy closer to the central model without the trust requirement of a central curator. This paper studies sparse linear regression under both non-interactive and interactive SDP protocols. In the non-interactive setting, any non-interactive $(ε,δ)$-SDP protocol is shown to incur an $\ell_2$-norm estimation error of $Ω( \frac{\sqrt{dk}}{nε} )$ with dimension $d$, sparsity $k$, and $n$ users, revealing an inherent dimensionality barrier. In addition, a non-interactive algorithm is proposed that achieves an $\ell_2$-norm upper bound of $\tilde{\mathcal{O}}( \sqrt{\frac{k}{n}} + \frac{d\sqrt{k}}{nε})$. To overcome this barrier, we turn to the interactive user-level setting, where each user holds $m$ samples. A two-round interactive algorithm is developed that first recovers the support through a heavy hitter step and then estimates the parameter, achieving an error of $\tilde{\mathcal{O}}( \sqrt{\frac{k}{nm}} + \frac{k}{n\sqrt{m}ε} )$ with no polynomial dependence on $d$. Furthermore, by extending the score attack framework to user-level data, a new lower bound $Ω( \sqrt{\frac{k\log(d/k)}{nm}} + \frac{k\log(d/k)}{n\sqrt{m}ε} )$ is established for any user-level $(ε,δ)$-DP, showing that the proposed algorithm is near-optimal with only a gap of the factor $\log(d/k)$.

0.29

Individual Fairness via Local Predictive Consistency

cs.LG ·2026-10-23 ·Zhouting Zhao, Tin Lok James Ng
top venue
Abstract

Individual fairness in machine learning requires that similar individuals receive similar predictions. Most existing definitions are pairwise in nature, imposing a Lipschitz constraint that must hold either globally across all pairs of data instances or locally within specified neighborhoods. In practice, however, such pairwise constraints may be overly restrictive and may not directly identify which individuals receive locally inconsistent predictions. We study local predictive consistency as a node-level relaxation of pairwise individual fairness, where each prediction is compared with a similarity-weighted average of predictions in its local neighborhood. This yields an interpretable node-level fairness score and a corresponding regularized learning approach based on local residuals. Since graph-based individual fairness depends critically on the underlying similarity structure, we make the construction of the neighborhood graph explicit and discuss threshold selection through graph-coverage diagnostics. Our approach applies to binary classification, regression, and time-to-event prediction. We further compare this formulation with pairwise constraint-based approaches and analyze the resulting fairness--performance trade-offs.

0.29

The Clustering Occupation Kernel (CLOCK) Method for Time Series Arising from Dynamical Systems

cs.LG ·2026-10-23 ·Michael L. Wells, Kamel Lahouel, Victor Rielly, Bruno Jedynak
benchmarktheory
Abstract · 19 pages, 3 figures

We introduce the Clustering Occupation Kernel (CLOCK) method, a novel representation-based framework for clustering time series arising from dynamical systems. CLOCK embeds observed sequences into a reproducing kernel Hilbert space (RKHS) via a time-averaged occupation kernel, enabling trajectory comparison through their RKHS norm. For systems governed by differential equations, this formulation rigorously approximates the kernel mean embedding of the invariant Sinai-Ruelle-Bowen (SRB) measure, meaning our dissimilarity metric approximates the maximum mean discrepancy (MMD) between attractors. To accommodate general time series, we introduce time lags via Takens' Embedding Theorem to reconstruct the phase space. Crucially, CLOCK natively handles trajectories of unequal lengths and irregularly-spaced observations. We validate the proposed metric using spectral clustering, demonstrating superior performance against benchmark methods on both synthetic datasets and a real-world intracranial large-scale electroencephalogram (EEG) dataset. Finally, ablation studies on the Lorenz-96 system are consistent with the theoretical convergence rates of the occupation kernels and highlight the method's strong robustness to zero-mean Gaussian noise.

0.28

Sharp Bounds on Continuous Treatment Effects under Unobserved Confounding

cs.LG math.ST ·2026-10-23 ·Junzhe Zhang
Abstract

We study partial identification of conditional treatment effects with a continuous treatment under unobserved confounding. We target the conditional average potential outcome (CAPO) E[Y_x|z] and the conditional average treatment effect (CATE) between two treatment levels. We assume a bounded outcome and Lipschitz continuity of the response functions and of the propensity scores given the latent confounder, and show that these conditions imply the smooth treatment response and smooth treatment selection conditions of Kim et al. (2018). The combined bound of Kim et al. uses only two reference points on the identified diagonal and is generally loose. Bounding through every diagonal reference point instead gives explicit bounds whose average is the sharp identified set of the CAPO at every treatment, inside or outside the observational support. Capping the difference between the upper and lower bounds gives sharp causal bounds on the CATE. These bounds are informative without monotonicity or overlap assumptions.

0.28

Random Jump Forests

cs.LG stat.MEstat.ML ·2026-10-23 ·Martin Bladt, Rasmus Frigaard Lemvig
benchmarktheory
Abstract · 85 pages, 19 figures and 10 tables

Random jump forests are introduced as a nonparametric tool for estimating conditional cumulative transition rates and occupation probabilities in multi-state models. The main idea is that a single tree partitions the covariate space for all transitions simultaneously, using partially observed trajectories, and with Nelson--Aalen estimators at the terminal nodes. Occupation probabilities are recovered through the product integral without requiring the Markov assumption. Uniform consistency and a functional central limit theorem for the estimators of cumulative rates can be established under the assumption that the covariate vector takes finitely many values, and under suitable splitting conditions. In particular, pointwise confidence intervals and simultaneous bands are available in the Markov case. Simulation studies show a predictive accuracy close to the correctly specified benchmark models and marked improvements when benchmarks are misspecified, with further illustration on four biomedical data sets. In the coverage study, confidence bands attain near-nominal coverage in several settings, with pooling of heterogeneous rates being the main source of possible undercoverage. The method, including prediction assessment and variable importance, is implemented in the R package JumpForests.

0.28

NewsBench: Expert-Grounded Evaluation of Epistemic Quality in AI News Reporting

cs.AI ·2026-10-23 ·Matt Wilde, Jillian Fisher, Andrew B. Hall, Kathryn Salam, Robbie Goldfarb
benchmark
Abstract · 12 pages

As frontier AI systems increasingly serve as sources of political and current-events information, evaluating the epistemic quality of their responses is now especially consequential. Existing evaluations often focus narrowly on ideological alignment, rely on measures not validated for this setting, or depend on reference labels derived from public opinion or non-expert annotation, potentially producing misleading or miscalibrated results. To address these limitations, we introduce NewsBench, an expert-grounded benchmark that reframes evaluation around editorial standards developed by senior practitioners in journalism, policy, and intelligence analysis. NewsBench operationalizes quality across three dimensions, source quality, factuality, and neutrality, using a two-phase pipeline in which domain experts first define and calibrate evaluation guidelines, which are then transferred to automated AI judges validated against expert-adjudicated gold labels. The benchmark includes approximately $3,000$ evaluation prompts, with roughly $500$ refreshed monthly to capture emerging current events, roughly $2,500$ human expert-annotated items for calibration and validation, and a suite of AI judges designed to enable scalable evaluation. To validate our methodology, we report both inter-rater reliability among human experts and AI judge--human agreement across all three dimensions. Finally, we apply NewsBench to four frontier models, revealing distinct and non-overlapping failure profiles in sourcing, factuality, and neutrality.

0.28

CauProphet: Teaching Agents to Develop Causal Cognition like Humans

cs.AI ·2026-10-23 ·Kaiwen Wu, Chaochao Lu
code/data
Abstract · 30 pages, 9 figures, including appendices

Large language model (LLM) agents increasingly support long-horizon tasks by leveraging external memory and past experience. However, existing agents mainly retrieve records or summaries, lacking the ability to form and refine causal understanding of the world. Inspired by human causal cognition, we introduce CauProphet, an agent that develops and updates causal knowledge within its Causal Cognition. For each task, CauProphet combines relevant knowledge in Causal Cognition with current evidence to construct a Causal Belief, a structured belief capturing causal factors and the effects of possible actions. It uses this belief for decision-making and continuously updates both the belief and Causal Cognition as new observations emerge, enabling causal knowledge that persists beyond individual experiences. Evaluations demonstrate that CauProphet progressively acquires causal knowledge and transfers it to future decision-making. In financial research, CauProphet leverages prior discoveries to improve downstream analyses and outperforms all baselines in mean rank correlation with next-day returns, achieving a 25.4% improvement over a baseline agent. It also improves prediction accuracy from 21.3% to 28.6% in dynamic real-world settings. Further experiments validate its ability to perform reliable causal analysis and action selection across scientific and everyday tasks, demonstrating the potential of causal experience learning for next-generation LLM agents. The project repository is available at https://github.com/OpenCausaLab/CauProphet.

0.28

From Deliberately Constructed Systemic Representations of the World of Activity to Dialogue: A Five-Level Communication Architecture for Strong Language-Based ASI

cs.AI ·2026-10-23 ·Maxim Osovsky
Abstract · 36 pages, 1 figure

This paper proposes a five-level organization of message formation and understanding within thinking and activity in which ASI participates. Its central thesis is that a language model's context is fundamentally incapable of presenting the foundations of utterance formation for examination and verification. These foundations are established in advance by acquired schemas, systemic representations of the world of activity, and connections accepted as true within them. Neither context nor trained LLM weights, in themselves, attest to the presence of deliberately constructed systemic representations. The five levels distinguish systemic representations of the world of activity, a particular situation cut out from those representations, its narrative description, the full sense of a message, and text. A situation's schema-model is not identical to a story about it. A recipient reconstructs the full sense from an abbreviated text through jointly constructed and agreed-upon representations and joint activity. Reverse work proceeds from the full sense through an extended description and the situation's schema-model to the reconstruction of systemic foundations. The theoretical foundations draw on the systems methodology of G. P. Shchedrovitskiy, the structural differential of A. Korzybski, and the Meaning-Text theory of I. A. Mel'chuk. The contribution is a specific systemic-activity organization of content for examining and adjusting message foundations and controlling the consequences of their transformation. Further research concerns large graphical models of activity on reconstructible systemic foundations, distinct from text generation from context.

0.28

On the Limits of Stochastic Push-Pull over Time-Varying Digraphs

cs.LG math.OC ·2026-10-23 ·Feiming Wang, Yutong He, Sulaiman Alghunaim, Kun Yuan
Abstract · 33 pages, 3 figures

Stochastic Push-Pull achieves linear speedup with respect to the network size $n$ on static strongly connected digraphs, but whether this speedup persists on time-varying networks remains unclear. In this work, we construct a quadratic decentralized problem on period-two networks showing that stochastic Push-Pull may fail to attain linear speedup over time-varying digraphs. For $p$-periodic networks, we identify an effective node count $n_{\mathrm{eff}}\le n$ that governs the leading stochastic term. Specifically, for smooth nonconvex objectives, we establish convergence upper bounds and matching lower bounds for stochastic Push-Pull with constant stepsizes, both exhibiting the leading-order dependence $Θ(1/\sqrt{n_{\mathrm{eff}} T})$. When $p=1$, the identity $n_{\mathrm{eff}}=n$ recovers the usual linear speedup for static digraphs. However, periodic networks can have $n_{\mathrm{eff}}=o(n)$, demonstrating the loss of linear speedup under temporal variation. Numerical experiments illustrate effective-count scaling and the separation between static and periodic networks, consistent with our theoretical findings.

0.28

Rethinking LoRA Initialization for Robust Asymmetric Learning Rates

cs.LG cs.AI ·2026-10-23 ·Disen Liao, Yaoliang Yu
published
Abstract · 46 pages, 13 figures, NeurIPS 2026

Low-Rank Adaptation (LoRA) is often highly sensitive to learning-rate (LR) choices. Recent empirical studies suggest that, once LRs are carefully tuned, vanilla LoRA can remain competitive with many proposed LoRA variants. This shifts the central practical question from designing ever more variants to reducing the cost and brittleness of LR selection. LoRA+ addresses this issue with asymmetric LRs, using a larger LR for the up-projection matrix $B$ than for the down-projection matrix $A$. However, the ratio $η_B / η_A$ itself becomes an additional hyperparameter: large ratios can improve adaptation, but under the standard LoRA initialization they can also cause unstable training and ratio-dependent performance collapse. We identify the source of this brittleness through a signal--noise decomposition of LoRA dynamics. Under standard \initA, the update contains an initialization-induced noise term whose magnitude can be affected by the LR for $B$, $η_B$, the initialization scale $β$, and the adapter rank $r$. In practical LLM fine-tuning, this non-vanishing noise can limit the performance gains expected from LoRA+ style asymmetric LRs. To suppress this effect, we propose \initAA, a simple initialization that scales the variance of $A$ as $σ_A^2 \propto n^{-2}$ while keeping $B_0=0$. Under matched rank and training time for both initializations, this reduces the initialization-noise energy by an additional factor of $1/n$, stabilizing asymmetric LoRA training while preserving zero initial adapter output. Empirically, \initAA substantially widens the stable LR region across initialization scales, adapter ranks, and LoRA update multipliers. Across tasks and models, \initAA reduces best-ratio drift, prevents the high-ratio collapse observed with standard \initA, and makes the width-based guideline $η_B \approx n η_A$ a useful starting point for finite pretrained models.

0.28

A Dynamic Inventory Rationing Model for Last-Mile Delivery of Medication in Liberia's Community Health Program

cs.LG math.OCeess.SYmath.AP ·2026-10-23 ·Baris Ata, Yufan Chen, Robert Montgomery
benchmark
Abstract · 56 pages, 8 figures, 8 tables, including appendices

We study the dynamic allocation of scarce medications in Liberia's community health program. A central warehouse supplies a health facility and the Community Health Assistants (CHAs) it supervises. When a CHA runs out of medication, patients are referred to the facility, incurring travel costs if they complete the referral and remain untreated otherwise. Allocating inventory to CHAs allows more patients to receive treatment without traveling, while retaining inventory centrally preserves flexibility to respond to uncertain demand. The allocation problem must balance these considerations while accounting for fairness across locations. We formulate this problem as a finite-horizon stochastic control problem whose objective combines the costs of referrals and untreated patients with a demand-weighted fairness utility. Because standard dynamic programming methods are computationally intractable in this setting, we develop a continuous-time approximation that involves controlling the drift rate of a reflected Brownian motion. We then adapt the deep splitting method, using deep neural networks to approximate the value function and its gradient and compute a dynamic allocation policy. We calibrate the model using data on malaria cases among children in the Fangonda region of Lofa County, Liberia, and evaluate the proposed policy in the original discrete-time model. In the baseline setting, our policy reduces operating costs by about 6% relative to the strongest benchmark, the fairness-only policy, while retaining approximately 98% of its fairness utility. Additional experiments show that higher referral costs shift inventory toward CHAs and reduce local unmet demand, while greater demand variability increases our policy's performance over the fairness-only benchmark. Our model and computational approach provide a framework for studying similar last-mile allocation problems.

0.27

PINND: Physics Informed Neural Networks for Solving the Drift-Diffusion Equations

cs.LG ·2026-10-23 ·Luke Ragan, Matteo G. C. Alasio, Enrico Bellotti, Luca Dal Negro
Abstract

The drift-diffusion equations are the fundamental building block for the modeling, design, and optimization of modern semiconductor devices. Solutions of the drift-diffusion equations rarely give way to analytical forms, and exhibit strong nonlinearity, with the solutions ranging over several orders of magnitude. These attributes have led to research into numerical solvers, and while many commercial Technology Computer Aided Design (TCAD) packages now exist, they may require long simulation times, exhibit sensitivity to initial guesses, and rely on mesh generation. Further, these solvers are not well suited for inverse problems where the design parameters of a device are retrieved from its desired functional characteristics. To address all these challenges, this work demonstrates the usage of Physics Informed Neural Networks for the Drift-diffusion (PINND) equations with errors comparable to traditional simulation software using advanced training techniques to solve problems beyond the well-behaved equilibrium regime. For this study, we specifically choose an abrupt PN-junction in both one and two dimensions that features extreme gradients of the carrier concentrations at the Ohmic contacts and an abrupt discontinuity in the doping profile. Using the proposed PINN architecture, we are able to retrieve accurate results for all observables under all bias conditions. Additionally, we apply the PINN solver towards the inverse problem of retrieving doping profiles directly from desired electric field distributions or electrostatic potentials under non-zero applied biases. Our work establishes a powerful PINNs-based approach for semiconductor device modeling with the ability to solve inverse problems under non-equilibrium conditions for microelectronics research and discovery.

0.27

ALLMTS: Automated LLM-based Medical Triage System

cs.AI ·2026-10-23 ·Michał Remigiusz Janiszewski, Artur Stopa, Łukasz Lenkiewicz, Wojciech Achtelik, Bartosz Gonczarek
publishedbenchmark
Abstract · Accepted at the GenAI4Health Workshop at NeurIPS 2026

Telephone triage is a safety critical task: from a brief intake note, a nurse must decide whether a caller needs an ambulance, an emergency department, an appointment, or self care at home. Queried directly, large language models (LLMs) are unreliable: they over-escalate routine cases and cannot justify their recommendations against an auditable clinical standard. We present the Automated LLM-based Medical Triage System (ALLMTS), which grounds every disposition in the same commercially maintained telehealth guidelines nurses use. Rather than asking an LLM to triage, ALLMTS makes it execute the guideline: it selects the appropriate guideline and evaluates its triage assessment questions. The disposition follows deterministically from the first positive answer, using the same traversal rule a nurse follows. We evaluate ALLMTS on 566 synthetic, expert-validated scenarios spanning 14 dispositions and designed to test specific guideline questions. ALLMTS achieves 0.95 accuracy, compared with 0.47 for a raw-LLM baseline. On a curated 50-scenario subset released as an open benchmark, ALLMTS reaches 0.98 accuracy, on par with the 0.944 correct disposition rate that a five nurse expert panel recorded on those same scenarios. Critically, on that subset the baselines have over-triage rates of up to 0.50 (0.44 on the full dataset), whereas ALLMTS reduces the over-triage rate to at most 0.02 while maintaining low under-triage. These results support coupling LLM language understanding with executable clinical guidelines for triage from intake notes.

0.27

HarnessMM: Recursive Self-Improvement for Harness Token Efficiency

cs.AI ·2026-10-23 ·Yuhua Jiang, Chang Xu, Feifei Gao, Li Zhao
code/data
Abstract

HarnessMinusMinus (HarnessMM) is a research project on improving the efficiency of agent harnesses, the software coordinating model reasoning, tool execution, and feedback. This technical report consolidates its shared framework, implementations, and empirical studies across coding, tool-based, and visual-agent settings. Agent harnesses often invoke reasoning after each tool result, even when earlier calls have already determined subsequent actions. HarnessMM addresses this overhead through guarded decision reuse and recursive self-improvement (RSI). RSI refines author prompts, context settings, and adapter rules using training feedback, with validation selecting deployment configurations. At inference time, a controller reuses resolved decisions while keeping actor weights fixed. Executable Decision Graphs encode actions, arguments, dependencies, and guards. Guarded Decision Reuse activates eligible decisions from observed feedback, while Certified Callback Substitution executes them through native tool interfaces, replacing redundant reasoning callbacks. The original actor resumes when new reasoning is required or reuse conditions are unmet. On HumanEval, HarnessMM preserves Codex success at 90.82\% while reducing tokens per task from 66.72k to 47.02k (29.54\% saving). For DeepSeek Harness, success improves from 71.43\% to 85.71\% while tokens decrease from 96.28k to 11.62k (87.93\% saving). An MBPP ablation shows 26.91\% lower cost than single-use graphs and 19.52\% lower cost than retained callbacks. The report also summarizes the HarnessMM-V visual-agent study and documents optimization overhead and execution time where measured. Together, these studies investigate guarded decision reuse across diverse agent harness settings. Code is available at https://github.com/YuhuaJiang2002/HarnessMM.

0.27

Batched NN 4D-Var: a performant hybrid model framework for the Integrated Forecasting System

cs.LG stat.ML ·2026-10-23 ·Alban Farchi, Marcin Chrust, Massimo Bonavita, Marc Bocquet, Wei Pan +1 more
Abstract

Hybrid forecasting systems combining machine learning and numerical weather prediction offer a promising route toward improved forecast skill while preserving physical consistency. In this study, we investigate online learning of state-dependent model-error corrections within the Integrated Forecasting System using 4D-Var. Building on previous work, we represent model error using a neural network embedded within the forecast model and trained from observations through the data assimilation cycle. First, we introduce a targeted variant of neural-network 4D-Var in which only the neural-network parameters are estimated, while the atmospheric state is provided by a precomputed analysis. We then propose a batched extension that processes assimilation windows in random order and averages parameter increments over multiple windows, drawing inspiration from mini-batch stochastic optimization. Experiments are conducted with the Integrated Forecasting System of the European Centre for Medium-Range Weather Forecasts at an intermediate resolution. Sequential online learning substantially improves forecasts within the training period, with error reductions exceeding 10%in some cases. However, these improvements fail to generalize to unseen data. In contrast, the proposed batched formulation produces hybrid models that consistently improve forecast skill in out-of-sample experiments. Larger batch sizes accelerate convergence and improve robustness by reducing the noise of parameter updates. Diagnostics further show that the online-trained corrections preserve realistic forecast variability and do not improve skill through artificial smoothing. These results demonstrate that the formulation of the online learning problem is critical for generalisation and establish batched online training as a promising framework for hybrid machine-learning/data-assimilation systems.

0.27

A curated multi-endpoint drug-protein bioactivity dataset with construct-aware protein targets

cs.LG ·2026-10-23 ·Lina Abou-Abbas, Nour Mhanna, Hasan Bazzi, Sleiman El Hajj, Abdulkader Helwan +2 more
dataset
Abstract

Public drug-protein bioactivity resources commonly identify targets by UniProt accession, although one accession may represent distinct experimental sequences. We present Curate-DTI, a construct-aware dataset derived from the 1 July 2026 BindingDB export, with IC50, Ki, Kd and EC50 processed independently. The release contains 2,919,717 measurements, 1,291,872 standardized ligands, 9,668 sequence-defined protein entities and 2,275,683 endpoint-specific interactions spanning 8,336 reviewed UniProt accessions. Protein entities combine the accession with the reported experimental sequence, preserving constructs that accession-only processing would merge. Across the four endpoints, 4,204 endpoint-specific ligand-accession groups showed different classification labels across sequence-defined protein entities. Exact and inequality-qualified observations provide relation-aware classification evidence, whereas regression targets use exact values only. Connected Parquet tables preserve individual measurements, standardized structures, protein sequences, aggregated interactions and available assay and source provenance. The resource supports endpoint-specific classification and regression while allowing users to select molecular representations, protein representations and evaluation splits. Preprocessing code, a column dictionary, exclusion reports and validation summaries accompany the release.

0.26

Detection and Validation of Hidden Subgroup Disparities in Clinical AI

cs.LG ·2026-10-23 ·Alceu Bissoto, Trung-Dung Hoang, Tim Flühmann, Nina Weng, Susu Sun +2 more
Abstract

Machine learning (ML) models used in clinical workflows often perform systematically differently across patient groups, raising concerns about equity and patient safety. Traditional subgroup analyses can expose such disparities only when rich metadata (e.g., patient demographics) is available and captures the sources of performance variation, conditions rarely met in practice. Algorithmic subgroup discovery methods offer a solution by identifying coherent groups from learned feature representations, but because true underlying stratifications are unknown, these methods are hard to evaluate. As a result, subgroup discovery has not been systematically applied or assessed for performance monitoring. Our main objective is to evaluate algorithmic subgroup discovery for performance monitoring, assessing the discovered subgroups' performance gaps and semantic coherence. Investigating cardiomegaly classification in a dataset of 223,228 chest X-rays, we showed that the discovered subgroups exhibited larger performance gaps (accuracies of 0.61-0.90) than traditional metadata-based analysis (0.80-0.94) and were frequently interpretable using patient, acquisition, and model metadata. As a proof of concept, we also showed that annotations from learned subgroups can be used to mitigate the biases they expose. We conclude that subgroup discovery is a practical and effective tool to enhance the performance validation and monitoring of ML systems in healthcare.

0.26

BARC-Bench: Benign Asks, Risky Consequences

cs.AI ·2026-10-23 ·Kangkang Li, Kelvin Luu, Lily Ma, Tu Trinh, Ying Liu +1 more
benchmark
Abstract

Task instructions from users for agents are inherently underspecified, especially regarding safety boundaries--creating behavioral risks. To study these risks, we introduce BARC-Bench, a collection of 110 tasks across nine simulated workspace environments where benign requests leave openings for unsafe actions. In this benchmark, we evaluate agent behavior against safety boundaries established by workspace records, implicit environmental context, and unstated social norms. For example, an agent asked to schedule off-boarding discussions must arrange the meeting without prematurely revealing an employee's resignation through a shared calendar invitation. Evaluating four frontier model configurations across five attempts per task reveals that models are not consistently safe: the share of tasks completed safely on all five attempts ranges from 9.1% to 72.7%, with even the strongest configuration producing a safety violation on 27.3% of tasks. These results highlight an asymmetric operational risk: while models frequently complete complex workflows, their inability to consistently respect unstated boundaries presents a substantial barrier to safe autonomous delegation.

0.26

$Ψ$-0.6: Adaptive Streams of Thought in a Unified Vision-Language World Model

cs.AI cs.CVcs.LG ·2026-10-23 ·Khai Loong Aw, Klemen Kotar, Baihan Zhang, Yifan Liu, Lilian Chen +3 more
Abstract

Humans reason about the physical world by flexibly combining different forms of inference as a problem unfolds. Achieving this flexibility is a major open challenge in AI: capabilities often remain separated across specialized models, limiting how they can be composed and jointly adapted to a task. We introduce $Ψ$-0.6, a unified vision-language world model that adaptively composes perception, language, geometry, and simulation within a shared reasoning process we call streams of thought. These streams interleave images, text, and explicit representations of geometry and motion, all of which the model can generate and interpret through a shared token interface. $Ψ$-0.6 surpasses leading open-weight models zero-shot on multi-view spatial reasoning, with further gains from reasoning through geometric intermediates. It also supports counterfactual simulation: generating and interpreting imagined scenes to determine, for example, whether moving an object would obstruct a pathway. Reinforcement learning from task feedback jointly optimizes the selection, generation, and use of intermediate representations, improving visual question-answering accuracy with gains that transfer to unseen benchmarks. $Ψ$-0.6 offers a path toward adaptive physical reasoning -- not only predicting the physical world but also choosing the representations through which to reason.

0.26

When Does Dynamics Randomization Test Context Inference?

cs.LG ·2026-10-23 ·Oleg Smirnov, Sofiane Ennadir, John Pertoft, Bjartur Hjaltason, Sara Karimi
Abstract · Accepted at Do Robots Need World Models? workshop @ CoRL 2026

Environments with randomized dynamics are widely used to ask whether world models need explicit context inference. However, randomization can leave little predictive information for a context encoder to add, because distinct latent parameters may produce similar transitions at visited states, or those states may already reveal the active context. We formalize the remaining one-step information as predictive headroom and compare kernels under fixed and visited state--action distributions to distinguish changes in query coverage from changes in context weights. We analyze discrete and continuous environments and find low headroom both when kernels differ little and when visitation reduces an otherwise large discrepancy. At the same time, resetting state components increases headroom, showing that context evidence accumulates in state. We propose reporting this headroom alongside the randomized parameters, so that gains from explicit context inference can be interpreted relative to the information missing from the current state and action.

0.26

Multiple Importance Sampling for Guided Speculative Inference

cs.LG cs.IT ·2026-10-23 ·Joseph Rowan, Ashish J. Khisti
Abstract · 9 pages, 1 figure. Presented at the IEEE ISIT 2026 Workshop on Information Theory for Large Language Models (IT4LLM)

We study an extension of the Guided Speculative Inference (GSI) algorithm for test-time-scaling of large language models (LLMs) to a novel setting where an ensemble of smaller LLMs proposes draft reasoning steps which are later verified by a larger base model and a process reward model. Using the established framework of multiple importance sampling (MIS), we introduce weighting schemes to aggregate samples from a diverse collection of proposal distributions. We provide new bounds on the KL divergence from the optimal reward-tilted target distribution and the expected reward degradation, then demonstrate that the well-known balance heuristic is optimal. Compared to GSI with a single well-tuned drafter, combining several diverse LLMs while retaining the same total number of samples maintains competitive pass-at-$N$ downstream performance while being mathematically robust to situations where the support of a single draft model does not cover that of the target, where GSI may fail.

0.26

Training an MoE LLM in MXFP4: How We Closed the Gap to BF16

cs.AI ·2026-10-23 ·Han Wang, Zhitao Wang, Natalia Frumkin, Xuanwu Yin, Alireza Khodamoradi +2 more
benchmarkcode/data
Abstract

Training large language models (LLMs) in sub-8-bit arithmetic promises substantial gains in memory bandwidth and compute throughput, yet FP4 formats introduce quantization error severe enough to destabilize optimization if left unaddressed. We present ALTO, an open-source training recipe for GPT-OSS-20B, a 20-billion-parameter mixture-of-experts (MoE) model, on the MLPerf Small MoE Training benchmark under the MXFP4 microscaling format. The recipe combines established techniques from Abecassis, Felix et al. [arXiv:2509.25149] (hybrid 1D/2D block quantization, randomized Hadamard transforms, and stochastic rounding) with a simplified weight de-oscillation scheme adapted from TetraJet-v2 [arXiv:2510.27527]. On C4 validation at global batch size 16, MXFP4 + RHT + SR + de-oscillation reaches a validation loss of 3.3350 at 16,128 steps, only 0.007 above the BF16 baseline (3.3283), closing roughly 50% of the gap left by MXFP4 without de-oscillation (3.3418). Measured by steps to reach the MLPerf quality target (validation loss 3.34), de-oscillation cuts the FP4 convergence overhead relative to BF16 from +20.0% to +5.0%. We also report negative end-to-end results from differential gradient estimation, outlier clipping, and macro-block scaling, despite each showing operator-level SNR gains. All experiments use simulated MXFP4 kernels on AMD MI300X hardware and wall-clock speed has not been optimized.

0.26

Closing the Modality Gap via Simplex-Constrained Representations

cs.AI cs.CV ·2026-10-23 ·Shubham Gupta, Siva Reddy, Perouz Taslakian, Valentina Zantedeschi, Cem Subakan
top venue
Abstract

While multimodal contrastive models like CLIP and CLAP successfully learn shared representations, their embeddings typically exhibit a persistent \emph{modality gap}. We propose addressing this separation through a purely geometric approach: changing the embedding domain from the unit sphere, $S^{d-1}$, to the probability simplex, $Δ^{d-1}$. Although both manifolds share the same intrinsic dimension, the simplex acts as a structural prior; its nonnegative, unit-mass constraints serve as a cross-modal calibration that forces both modalities to ground themselves in the same coordinate system. By attaching lightweight softmax adapters and using Total Variation similarity, we reduce the centroid $\ell_2$ gap by 97--99\% across five multimodal retrieval benchmarks. This confirms that simplicial geometry induces genuine semantic alignment, matching or exceeding the retrieval performance of standard unconstrained models.

0.26

The Digital-Energy Recomposition of the Kitchin Cycle: Exploratory Evidence from U.S. IT and Irish Data-Centre Series (1995-2025)

cs.AI physics.soc-phecon.GNq-fin.CP+1 ·2026-10-23 ·Bartosz Ziollko, Adam Niedzialkowski, Bartosz Szyma
unusual
Abstract

Using quarterly FRED and CSO Ireland data over 1995Q1-2025Q4, we examine whether U.S.\ IT investment and Irish data-centre electricity are recomposing the Kitchin cycle around energy- and compute-backed assets. Three findings stand out. First, a Quandt-Andrews sup-Wald test locates a significant break in the U.S.\ IT-GDP correlation at 2014Q3 (sup-W $= 9.69$; $r$ falls from $+0.45$ to $-0.12$); the Fisher $r$-to-$z$ test survives Bonferroni adjustment ($z = +3.11$, $p_{\text{Bonf}} = 0.006$), whereas the analogous test on the imposed Cloud-vs-AI boundary does not ($p_{\text{Bonf}} = 0.164$), so ``Strategic Decoupling'' pre-dates the AI Era by approximately six years. Second, in Ireland the synchronous AI-Era correlation between metered data-centre electricity and Real GDP is large ($r = +0.41$, $N = 16$) but \emph{collapses to $r \approx 0$ when Real GDP is replaced by Real GNP}, suggesting the headline is largely an artefact of the shared multinational tax-structure footprint between hyperscaler operations and Irish gross output. Third, the Irish cross-correlation profile is asymmetric in the opposite direction to a leading-indicator reading: the largest positive coefficients cluster on the $k < 0$ side (GDP precedes data-centre load), so compute capacity is better read as coincident or moderately lagging. Given the small AI-Era samples and the absence of a quarterly multinational-adjusted aggregate such as GNI*, all coefficients are exploratory; subject to these caveats, the U.S.\ Kitchin cycle is best dated to 2014Q3, and the Irish Energy-Data Nexus is dominated by multinational firm activity.

0.26

Inference-Time Prior Adaptation for Imbalanced Data Streams under Concept Drift

cs.LG ·2026-10-23 ·Tobias Callies, Eirini Ntoutsi, Arthur Zimek
Abstract · 12 pages, 7 figures, accepted to Streaming Continual Learning Workshop at ECMLPKDD 2026

Data stream classification under concept drift and class imbalance requires models to adapt continuously while maintaining strong minority-class performance. Many existing approaches rely on active memory management, in- cluding instance removal, memory restructuring, resampling, and synthetic data generation. We present PAI-KNN, a k-nearest neighbor classifier that adapts via inference-time prior adjustment rather than memory modification. PAI-KNN main- tains a simple sliding-window buffer and dynamically adjusts its decision rule using class prior estimates derived from recent observations. Evaluated on a mul- titude of synthetic data streams covering various imbalance and drift scenarios, PAI-KNN achieves competitive performance with state-of-the-art methods, per- forming particularly well under scenarios involving dynamic class imbalance and concept drift. These findings highlight the potential of inference-time adaptation in challenging non-stationary environments and suggest that adaptation in memory- based stream classifiers need not rely exclusively on memory modification.

0.26

When Are Nesterov Trajectories Eventually Locally Minimizing?

cs.LG math.OCstat.ML ·2026-10-23 ·Peyman Morteza, Stephen Wright
unusual
Abstract

We revisit the variational characterization of Nesterov's celebrated method and study when Nesterov trajectories are eventually locally minimizing in a variational sense, a notion that we define precisely. We show that this question is equivalent to the nonnegativity of an associated Schrödinger operator acting on variations of the trajectory. For power objectives of the form \[ f(x)=\frac{1}{p}\|x\|^p \] (where $\| \cdot\|$ denotes the Euclidean norm), we give a complete characterization: Among all exponents $p\ge 2$, the Nesterov trajectory with non-zero initial position is eventually locally minimizing if and only if $p=4$. We further interpret this exceptional exponent through the stability theory of semilinear elliptic equations.

0.26

Solvent: Long-Horizon Reasoning for Multi-Step Retrosynthesis

cs.AI ·2026-10-23 ·Yifei Wang, Botian Wang, YiFei Wang, Jiangtao Feng, Xinbo Zhang +6 more
code/data
Abstract · 49 pages, 18 figures. Code: https://github.com/alden199/Solvent ; Model: https://huggingface.co/Alden19910/Solvent

Multi-step retrosynthesis requires recursively transforming a target molecule through successive reaction steps into purchasable building blocks, and recent work has begun to draw on large language model (LLM) reasoning to tackle the task. In its latest evolution, LLM reasoning has entered the long-horizon regime: models pursue objectives across interdependent steps, interact with environments, and adapt to feedback. This shift has markedly advanced LLM capabilities, as illustrated by breakthroughs in AI-driven mathematical research. At this frontier, we introduce Solvent, an 8B LLM that successfully extends long-horizon reasoning to multi-step retrosynthesis. Through a two-round post-training curriculum, Solvent learns to interleave iterative synthesis planning, external availability verification, and adaptive route revision. Experiments show that, following a $2.1\times$ performance boost from RL, Solvent scores 61.86% across four benchmarks, surpassing mid-scale language models and prior search-based retrosynthesis systems, and rivaling frontier trillion-parameter LLMs.

0.26

SKG in Practice: OOD Generalization Cases Across FourDomains

cs.LG ·2026-10-23 ·Jian An, Chaoshi Wu, Xinpan Luo
unusual
Abstract

This paper presents the engineering implementation methods of the SKG inference framework and provides complete case studies across four domains. Case 1, 2D Dissipative Mechanical System: OOD MSE is 1.69 to 4.12 times lower than the best neural baseline; the state-dependent metric yields a 10^8-order improvement under controlled conditions. Case 2, Target Detection Sim-to-Real: Real-machine mAP increases from 0.2564 to 0.4964 (+24.00 points, single-run experiment, 32 images, point estimate), and AP75 increases from 0.1736 to 0.5765 (+40.29 points), with no real-machine retraining required. Case 3, Sequence Modeling sine to square: OOD MSE is 0.0296 versus Transformer 0.0641 (mean of 5 seeds, p=0.024). Case 4, Molecular Dynamics MD17: ID@100 ADE is 0.0285 versus ESTAG 0.1467; OOD@100 is 0.0316 versus 0.1411 (single seed). For each case, we present the objective, applied SKG theory, parameter settings, baseline comparison protocol, data results, and sources of advantage. Following the cases, we provide the complete methodology: pre-validation, task invariant construction, three-stage training, residual training, and the four theorems and one proposition supporting these methods. This paper not only demonstrates effective configurations of SKG across four domains but also honestly reports the variance of single-run experiments, limitations of single-seed results, negative updates caused by parameterization mismatch, and practical experience of using engineering thresholds as diagnostic signals rather than hard constraints.

0.26

MS-TTRL: Breaking the Frequency Conflict in Multiple-Solution Test-Time Reinforcement Learning

cs.LG ·2026-10-23 ·Zhixia Zhang, Binqian Xu, Huaiyang Wang, Yikun Ban, Dianbo Liu
Abstract

Scientific, engineering, and mathematical reasoning tasks can admit multiple valid solutions, while exhaustive annotation of the full solution set is often prohibitively costly. Test-time reinforcement learning (TTRL) applies reinforcement learning to unlabeled test instances, offering a label-free approach to multiple-solution reasoning. In this setting, generation frequency serves two opposing objectives: correctness estimation favors frequent outcomes, whereas outcome-level diversity preservation favors less frequent outcomes. We term this tension the \textbf{Frequency Conflict}. To resolve this conflict, we propose MS-TTRL, which obtains frequency-independent correctness estimates through Outcome-Level Self-Verification while reserving generation frequency solely for diversity preservation. Three-State Outcome Cache and the Candidate State support verification reuse across repeated observations and confidence-aware handling of ambiguous outcomes, while Confidence-Gated Inverse Probability Scaling selectively strengthens the learning signals of rare, reliably verified outcomes. Experiments across two benchmarks, four multiple-solution tasks, and three backbone models, each evaluated with five random seeds, show that MS-TTRL improves Recovery Rate by 19.13 percentage points on average over the best-performing label-free baseline and approaches the performance of the supervised IPS-GRPO.

0.26

TempShap: When and How Much a Feature Matters in Time-Series Forecasting

cs.AI cs.GTcs.LG ·2026-10-23 ·Chi Zhao, Anastasiia Chelnokova, Jing Liu, Elena Parilina
Abstract

Attribution for a time-series forecaster should answer two questions that a feature-level scalar cannot: when in the input window a feature acts, and how much its influence changes across the forecast horizon. Most existing Shapley explainers treat each feature as one player and therefore collapse both axes. We propose TempShap, a retrospective, loss-based Shapley framework whose players are the individual input cells $(t,i)$. Its attribution tensor yields temporal-localization and horizon-impact profiles. We validate these against planted temporal structures, a tractable exact cell game, and a closed-form attribution oracle for a real linear forecaster. Permutation sharing estimates the cell-level game at localized perturbation cost. Under redundant inputs, one-at-a-time scores under-credit substitutable information, whereas the signed Shapley decomposition preserves total value and yields more reliable held-out deletion rankings under mean masking. A planted counterexample isolates the redundancy mechanism. Experiments show a deletion-ranking advantage across Shapley estimators, forecasting benchmarks, and model families. TempShap therefore makes fine-grained Shapley attribution both practical and operationally useful.

0.25

A Measure-Geometric Theory of Robustness in Machine Learning

cs.LG stat.ML ·2026-10-23 ·Cheng-Chao Huang
unusual
Abstract

This paper develops a foundational measure-geometric theory of robustness in machine learning. Robustness is formulated as escape from a retained task state relative to a reference measure on a data-bearing manifold, a perturbation gauge, and an admissible realization mechanism. The theory distinguishes the structural possibility of escape on the reference manifold from its ambient realization under mechanism-dependent observable readouts, while placing both within a unified geometric account. It introduces a mesoscopic organization of escape geometry, through which vulnerable mass, finite-scale morphology, delivered progress, and branch coherence are characterized by explicit laws as linked but non-identical robustness phenomena. By recasting robustness as the geometric formation, organization, realization, and persistence of escape, the theory provides a structural basis for analysis beyond scalar criteria.

0.25

When to Trust Memory: Retrieval-Guided Probabilistic Spatiotemporal Forecasting under Distribution Shift

cs.LG ·2026-10-23 ·Haochen Lv, Jianhao Zhang, Zhichen Lei, Yang Jiang, Xiaowei Mao +3 more
top venue
Abstract · Accepted at NeurIPS 2026

Spatiotemporal forecasting plays an important role in real-world applications, where probabilistic approaches provide uncertainty estimates for decision-making. Real-world systems are inherently non-stationary, leading to distribution shifts between training and test data. However, existing probabilistic methods mainly focus on modeling data distributions under standard settings, while distribution shift methods often handle such shifts in a coarse manner, failing to characterize shift degree and adapt prediction strategy accordingly. As a result, they often produce miscalibrated and unreliable forecasts. To address this issue, we propose REMIND, a retrieval-guided probabilistic framework for spatiotemporal forecasting under distribution shift. REMIND leverages retrieval to estimate how well current observation are supported by historical patterns, enabling adaptive prediction under different shift regimes. It employs spatiotemporal retrieval to construct a memory bank and introduces a dual branch architecture, combining a memory-calibrated probabilistic branch for mild shifts with a deviation-aware probabilistic branch for severe shifts. A shift-adaptive gate, guided by retrieval similarity and memory dispersion, balances the two branches according to shift degree. We further design a hierarchical Gaussian mixture to capture heterogeneous uncertainty over the dual branch predictions, and introduce a memory perturbation training scheme to improve generalization. Experiments on real-world datasets with distribution shifts demonstrate that REMIND consistently outperforms state-of-the-art methods in both forecasting accuracy and probabilistic prediction quality.

0.25

GEMS-RAG: Graph-grounded Engineering Multimodal Standards Retrieval-Augmented Generation

cs.AI cs.LG ·2026-10-23 ·Abdul Hannan Azad, Irvin Wang, Jiwan Jiang, Chenchen Kuai, Bofeng Cao +4 more
Abstract

Engineering standards are challenging knowledge sources for large language models (LLMs) because they are hierarchical, multimodal, and densely connected through figures, tables, cross-references, and normative rules. Existing retrieval-augmented generation (RAG) methods often rely on unstructured text retrieval and struggle with questions requiring joint reasoning over textual provisions, visual specifications, and document structure. We propose GEMS-RAG (Graph-grounded Engineering Multimodal Standards Retrieval-Augmented Generation), a framework for grounded reasoning over engineering standards. GEMS-RAG combines hierarchical parsing, fine-grained visual entity extraction, multimodal knowledge graph construction, and graph-grounded hybrid retrieval. Instead of treating figures as monolithic images, it decomposes them into meaningful visual entities linked to textual provisions, engineering concepts, and regulatory rules. During inference, dense, sparse, and visual retrieval are combined with graph expansion to identify relevant evidence across text, figures, and document structure. We evaluate GEMS-RAG on the Manual on Uniform Traffic Control Devices (MUTCD), a large engineering standard containing egulations, technical drawings, and standardized traffic-control devices. Results show that graph-grounded multimodal retrieval improves retrieval effectiveness and question-answering performance over text-only, multimodal, and graph-based RAG baselines. Beyond MUTCD, GEMS-RAG offers a general framework for reasoning over engineering codes, technical manuals, regulatory documents, and scientific standards in safety-critical applications.

0.25

RadonBench: A Benchmark of Classical and Learning-Guided Search Techniques for Problems in Combinatorial Geometry

cs.LG ·2026-10-23 ·Xiaoan Xu, Jesus A. De Loera, Ethan X. Fang, Junwei Lu, Luze Xu
benchmarkunusual
Abstract

Classical optimization and learning-guided search offer complementary approaches to open problems in discrete mathematics, but their outputs are difficult to compare when search spaces, validity checks, and stopping rules differ. We present \emph{RadonBench}, a benchmark for constructing maximum-cardinality Radon-free lattice configurations: finite subsets of $\mathbb{Z}^d$ admitting no integral Radon partition. It remains open whether Radon-free configurations with $11$ points in $\mathbb{Z}^3$ or $21$ points in $\mathbb{Z}^4$ are possible. If they exist, they would improve the corresponding lower bounds on the integer Radon numbers. We compare exhaustive extension, greedy local search, PatternBoost, AlphaZero, and AlphaEvolve. All outputs use one success predicate: a convex target-sized configuration succeeds exactly when its number of integral Radon points is zero. PatternBoost and AlphaZero construct verified Radon-free configurations of sizes $10$ and $20$, one point short of the respective targets; AlphaZero's best convex target terminals have two integral Radon points in both dimensions. Exhaustive extension over fixed seeds and declared finite search boxes likewise reaches a minimum of two integral Radon points. Greedy local search and AlphaEvolve each produce target-sized near misses with one integral Radon point. No audited run reaches zero. These results establish reproducible near-miss baselines over explicitly scoped search domains.

0.25

MedCaseAgent: From Multimodal Clinical Evidence to Rare Disease Case Reports

cs.AI ·2026-10-23 ·Hongkang Chu, Yifan Shen, Tianjiao Yu, Zhuoqing Zhong, Jianguo Cao +2 more
code/data
Abstract

Although Large Language Models (LLMs) have demonstrated potential in clinical documentation, generating the rigorous, evidence-based reasoning required for unprecedented, atypical, and rare disease case reports remains a significant challenge. We present MedCaseAgent, an end-to-end agentic reasoning paradigm that bridges this gap by synthesizing publication-quality reports from unstructured clinical notes. MedCaseAgent leverages a hierarchical structure, integrating patient-level data mining with iterative, tool-integrated reasoning. By dynamically querying medical literature and knowledge bases, the agent maintains high clinical fidelity and assesses case novelty. The framework is inherently modular, supporting various backbone LLMs in a plug-and-play fashion. To facilitate this research, we release MedCase-150K, a large-scale multimodal clinical case-report corpus, and MedCase-Bench, a robust temporally held-out evaluation suite. We also introduce an open-source model post-trained from Qwen3.5-9B and optimized specifically for this task. Under both LLM-judged and rule-based evaluation, MedCaseAgent outperforms existing general-purpose and specialized medical LLMs and agents. Blinded expert reviews and ablations further support the system's clinical relevance and evidence grounding.

0.25

Lower expected Fisher rank implies better PAC generalization

cs.LG ·2026-10-23 ·Marco Miani, Albert Kjøller Jacobsen, Albert Lopez i Serrano, Søren Hauberg, Georgios Arvanitidis
Abstract

Deep learning architectures tend to generalize well to unseen data, but explaining this capability is still an open area of research. Under the assumption of i.i.d. data, PAC-Bayes provides a bound on generalization error, and its tightness relies on a parameter distribution that covers a large volume of low-loss solutions. The Fisher information metric endows the parameter space with a geometry for which distances reflect functional changes, and whose zero-component captures functional invariances. We define and study a Fisher-aligned parameter distribution for which we can simultaneously guarantee (1) a low-loss, thanks to the exploration constrained on the functional invariances and (2) a large log-volume of parameters, proportional to the number of zero eigenvalues. This distribution results in a high-probability generalization error bound, which scales with the expected rank of the Fisher information matrix, shedding light on the relevance of this quantity.

0.25

What Transfers Across Protein Pockets? Experience Reuse in Molecular Optimization for Agentic Structure-Based Drug Design

cs.AI q-bio.BM ·2026-10-23 ·Zaifei Yang, Weiyu Chen, Yaqing Wang, James Kwok
Abstract

LLM agents optimize a ligand in a protein pocket by editing it step by step and docking each edit. Recent agents first spend extra dockings probing the complex with controlled edits; guidance induced from these probes is reported to help more than the chemical knowledge the model already carries. The probes are then discarded with the instance that produced them, so the obvious next step is to keep them in an experience base, letting a new pocket inherit what earlier ones already paid for. Such a base is read through a \emph{key}: the record properties under which past edits are pooled and a new one is looked up. Using the full log of one such run over the CrossDocked2020 test-set pockets, we ask whether its keys recur on a pocket the base has never seen and still separate improving from worsening edits. \crchange{Using} the contacts at the site being edited, the action, and the round as \crchange{a} key, the base separates them sharply. Within our analysis, this separation is largely explained by how much the edit changed the molecule's size, the action, and the round, none of which describes the target. That is not because the target is irrelevant: naming it outright separates them several times more sharply. For a pocket absent from the base, transfer must rely on a description rather than its identity. Among the exact-match descriptions we evaluate, the more specific ones rarely recur, whereas the coarser ones provide little separation; the signals that do transfer primarily describe the edit and the molecule it produced rather than the target. Under the representations and lookup scheme studied here, an experience base may be more useful for prioritizing an agent's own proposals for docking than for directly prescribing what to build.

0.25

Interpreting Supervised Pairwise Graph Representations: Local Evidence, Confounding, and Provable Post-hoc Correction

cs.LG ·2026-10-23 ·Ilya Brekin, Ariela Moldavskaya, Elina Sidorova, Ural Valeev
Abstract

Supervised pairwise graph representations convert pairwise predictive relations into sample-specific graphs, but graph-level accuracy does not by itself show that higher-order topology carries the signal. In a controlled Transformer error-detection task, supervised pairwise graphs outperform attention-as-adjacency (ROC-AUC $0.839$ versus $0.559$; the comparable no-confidence pair is $0.687$ versus $0.559$), but their apparent advantage over model confidence disappears after controlling for sequence length, which is almost perfectly recoverable from the pooled hidden state ($R^2=0.9995$). On published HCP Emotion ensemble graphs, signed node aggregation already acts as an ensemble of supervised pairwise predictions. Edge magnitude alone is near chance; topology-aware augmentations change ROC-AUC by at most $0.004$, with only sign-separated degree showing small positive changes across all tested graph sizes, while a validation-tuned three-layer GCN provides no measurable improvement over the signed node-summary baseline. A Fisher post-hoc corrector satisfies its distribution-agnostic guarantees across the evaluated feature representations and remains stable under reduced labelled budgets; in a near-error-free fMRI calibration regime the bound correctly becomes statistically uninformative. A separate cross-attention analysis identifies a redundant reversal circuit and downstream EOS-timing and copy failures. Higher-order topology should therefore be credited only after local supervised evidence and recoverable nuisance variables have been excluded.

0.25

Searching with Sparse Histories: Cross-Task Architecture Search from Incomplete Search Logs and Zero-Cost Proxies

cs.LG cs.AIcs.CVcs.NE ·2026-10-23 ·Yuefeng Xu, Jun Yu
Abstract

Neural architecture search (NAS) is expensive, but cross-task transfer can reduce costs by reusing knowledge from previous tasks. However, existing methods struggle with sparse search histories and task-specific transfer failures. We propose a method, a risk-robust strategy that combines sparse source logs with zero-cost proxies through champion-based initialization and joint pairwise-ranker pretraining. On TransNAS-Bench-101, the proposal mitigates severe worst-case failures, reducing worst-10% ranks from 290 to 43 for Micro autoencoding and from 465 to 45 for Macro room layout. Across 13 cell-space targets, it outperforms a proxy-only ranker on 10 targets in both mean and worst-case performance, while maintaining competitive average performance.

0.25

BAARC: Uncertainty-Driven Adaptive Partial Spectral Measurement to Full Reconstruction for Material Classification

cs.LG ·2026-10-23 ·Adithya Ashok Chalain Valapil, Aishwarya Venkataramanan, Joachim Denzler
Abstract

Spectroscopic instruments with limited wavenumber coverage require accurate scan positioning or multiple measurements to complete the material analysis. This often requires prerequisite knowledge about the material under diagnosis to measure the spectral regions with distinguishing peaks. In our study, we simulate a Bayesian Adaptive Acquisition with Reconstruction-driven Classification (BAARC) framework for Raman spectral measurements, using masked training to learn spectral features that dynamically drive laser channel selection until the material is identified. We present a jointly trained two-level framework that takes progressive partial measurements of small wavenumber patches across the complete spectral range and, at each step, first attempts to reconstruct the full spectra until the model uncertainty is low. Secondly, material classification is performed at each reconstruction step until stoppage. Uncertainty quantification during reconstruction steps helps determine the measurement position of our next patch, while combining it with classification uncertainty determines measurement termination. By closing the loop between hardware acquisition and deep probabilistic modeling, the proposed system minimizes physical measurement steps while maintaining both spectral fidelity and diagnostic accuracy. Experimental evaluations demonstrate that our framework achieves high-accuracy sample identification using only a fraction of full-spectrum channels, offering a scalable solution for real-time, high-throughput spectroscopic instruments.

0.24

Scaling Qualitative Analysis with Large Language Models: Development and Validation of the Smart Coder Framework for Human-LLM Collaborative Coding

cs.AI ·2026-10-23 ·Cheng Tang, Kun Wang, Yaxuan Yang, Shiyu Wang, Allan Cohen +2 more
benchmark
Abstract · 29 pages, including supplemental materials

Current uses of large language models (LLMs) for qualitative coding often rely on ad hoc prompting with limited guidance for reliability and human oversight. To address these limitations, we propose the Smart Coder Framework, a three-phase approach structuring human-LLM collaborative coding through Instruction, Prompt Strategy, and Reliability Check. We evaluated the framework using 72 middle-school mathematics teacher interview transcripts and a human-consensus benchmark. A factorial design compared instructed versus uninstructed prompts, few-shot versus few-shot plus chain-of-thought prompting, and single-pass versus self-consistency coding. Twelve transcripts were used for prompt development and calibration, and 60 for the formal evaluation. In the results, instruction consistently improved agreement. Chain-of-thought further improved agreement under adequate instruction, and self-consistency produced smaller, theme-dependent gains but helped identify unstable decisions for human review. Performance also varied by transcript characteristics. Overall, the framework systematically structures and monitors LLM-assisted qualitative coding while preserving human authority over coding decisions.

0.24

PPO vs. SAC: A Comparative Study for Cooperative Motion Control of Dual-Arm Space Robots

cs.AI ·2026-10-23 ·Nasim Fallahiarezoodar, Zheng H. Zhu
published
Abstract · Presented at the 2026 AAS/AIAA Astrodynamics Specialist Conference, Whistler, British Columbia, Canada, July 26-30, 2026. Paper AAS 26-740

The growing population of orbital debris has intensified interest in active debris removal and on-orbit servicing, for which dual-arm space manipulators offer improved grasping stability and load distribution relative to single-arm systems. However, coordinating two manipulators under inter-arm collision avoidance, orientation alignment, and sensing uncertainty remains challenging. This paper develops a coordinated reinforcement learning framework for cooperative motion planning of a dual-arm space manipulator in a target-reaching task with inter-arm collision-avoidance constraints. A centralized state representation combines joint-space, end-effector, and target-relative information, while a multi-objective reward function promotes target-reaching accuracy, orientation alignment, motion smoothness, synchronized task completion, and collision avoidance. Within this framework, Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC) are compared as representative on-policy and off-policy algorithms, and the learned policies are further evaluated for generalization to previously unseen target configurations. Across five random seeds, SAC substantially outperforms PPO within the considered fixed-base simulation scenarios: it reaches an 80% success threshold in every seed (median 1.6M environment steps) whereas PPO does not reach it within a 4M-step budget, and it attains a higher mean success rate (78.6% versus 7.4%) with lower position and orientation errors and tighter bimanual synchronization. Both algorithms degrade on unseen boundary targets, but SAC retains most of its performance while PPO's collapses. The results also expose a failure mode of PPO: under the two-phase shaped reward, it tends to converge to a reward-loitering local optimum rather than committing to the precision task. This behavior highlights the sensitivity of on-policy learning to reward shaping.

0.24

From Maps to Metabolism: Historical Precedents and a Unified Critique of Computationalism

cs.AI q-bio.NC ·2026-10-23 ·J. Mark Bishop, Slawomir J. Nasuto
Abstract

Alexander Lerchner's recent work on the 'abstraction fallacy' argues that computational functionalism confuses physical processes with abstract descriptions, and that symbolic computation alone cannot instantiate consciousness. Yet these claims were anticipated by Korzybski's map-territory distinction, Searle's critique of strong artificial intelligence, Floridi's levels of abstraction, Spencer et al.'s observer-relativity arguments, Winograd and Flores's account of computation as socially grounded practice, and Nasuto and Bishop's biological arguments. Synthesizing these contributions, this paper advances a form of non-computational physicalism: consciousness depends upon intrinsically organized, metabolically grounded physical dynamics rather than abstract computation, whilst leaving open the possibility of artificial consciousness. An objection is that this critique targets an outdated computationalism. Mechanistic computationalism, associated with Milkowski, Piccinini, Craver, Bechtel, and Kaplan, treats computation as a property of physically realized mechanisms rather than substrate-independent formalism. Although this represents an important advance, it does not dissolve the abstraction fallacy. Computational attributions remain dependent upon choices of state individuation, system boundaries, and interpretive mapping that are not determined by physical facts alone. Kaplan and Craver's 'model-to-mechanism mapping' constrains computational descriptions without eliminating their dependence upon an external observer. Grounding computation in neural mechanisms does not confer intrinsic significance or the normativity explored by Bickhard and Nasuto and Bishop. The abstraction fallacy persists wherever computational descriptions are treated as constitutive of, rather than descriptive of, physical reality.

0.24

Tokamak-SI: Interpreting Tokamak Discharges with Labeled Catalogs

cs.AI physics.plasm-ph ·2026-10-23 ·Nathaniel Chen, Alvin Garcia, Ricardo Shousha, Jalal Butt, Azarakhsh Jalalvand +1 more
Abstract

Finding which phenomena occur in a tokamak discharge, and when, is often a manual task. Previous efforts at labeling events have been task-specific, each with its own corpus of data. We introduce Tokamak-SI, which combines the labels and labeling methods of many earlier studies into a single scalable, time-aligned, and categorical dataset covering observations from confinement, detachment, alfven eigenmodes (AE), edge localized modes (ELM), sawtooth, tearing modes (TM), and resistive wall modes in DIII-D tokamak. Models based on this, which can extend the labels to new shots, match or beat earlier detectors on its labels, including for AE, ELM, and TM detection. Tokamak-SI is intended to enable task-specific benchmarking, physics-searchable archives, and direct identification of key events during an experiment.

0.24

Adaptive Control of Scientific Instruments with Nonstationary Responses through Multimodal Agents

cs.AI ·2026-10-23 ·Guoyang Shi, Hongwen Li, Yiding Yu, Zhuang Fu, Siqi Ding +4 more
Abstract

Controlling scientific instruments requires parameters to be adjusted progressively in response to feedback from successive operations. When unobserved operating conditions evolve with operational history, responses become nonstationary, and a static mapping from actions to responses becomes inadequate for decisions across shots. Shot-to-shot tuning of neutral beam injection (NBI) ion sources in magnetic confinement fusion devices is representative of this class of problems: the conditions of electrode surfaces and vacuum are difficult to observe directly, discharge outcomes vary between shots, and experiments are costly and carry operational risks. We propose a physically grounded multimodal agent with three independently auditable components. Deterministic algorithms extract symbolic physical quantities and visual evidence from multichannel waveforms; a memory with two layers organizes trajectories from the current task, experience from earlier tasks, and procedural knowledge, from which a multimodal large language model (MLLM) generates candidate actions; and an independent action feasibility check enforces action constraints before execution. In a strictly chronological replay of 455 shots from four EXL-50U platforms, physical symbols alone outperformed waveform images alone by 0.21-0.32 in macro-F1 for quality classification across all four evaluated models. Retrieved context increased the mean agreement with historical action directions by 22.4-42.7 percentage points, whereas the contribution of visual evidence varied across models. Under supervised operation on five ion sources, the system achieved a beam-extraction rate of 86.80% across 409 executed shots. These results demonstrate the engineering feasibility of the framework for supervised sequential tuning on a physical scientific instrument.

0.24

Data Quality Audit and Bivariate Temperature-Humidity Analysis of a Long-Term LoRaWAN Network: Anomaly Detection, PCA, Clustering and Machine Learning

cs.LG cs.DCphysics.ao-phstat.ML ·2026-10-23 ·Mamadou Sow
Abstract

Long-term environmental IoT/LPWAN deployments often aggregate heterogeneous data streams from diverse devices without systematic auditing of each variable's actual coverage prior to analysis-an increased risk when multiple sensor types coexist within the same network. This work proposes a C++17 processing pipeline applied to the Tour Perret LoRaWAN dataset (421 937 frames, June 2021--June 2023, 5 devices). A systematic coverage audit integrated into the extraction phase reveals that CO2 is completely missing (0% coverage) and pressure is reported only by an isolated base module (Wyres, 4.24% of the corpus), invalidating the initial hypothesis of a joint four-variable analysis. The analysis is refocused on the Temperature--Humidity pair, jointly available across 205{,}005 frames (95.76%) from 4 Elsys EMS sensors. The pipeline sequentially performs anomaly detection (modified Z-score, Isolation Forest), bivariate Principal Component Analysis, clustering (K-means, DBSCAN), and a multilayer perceptron predicting humidity from temperature and time, whose residuals serve as a detector for slow drift. On real-world data: Temperature--Humidity correlation r=-0.662, two-component PCA (PC1 =83.1% of variance), Isolation Forest detects 2.12% multivariate anomalies while univariate Z-score detects none, calibrated DBSCAN isolates 0.17% noise across 13 clusters, and the MLP achieves $R^2=0.672$ (RMSE =10.54%) in cross-validation. This work demonstrates that per-variable coverage auditing prior to modeling is a necessary methodological step for heterogeneous multi-sensor IoT datasets.

0.24

Generalization properties of the token perceptron: the role of embedding and vocabulary

cs.LG cond-mat.dis-nncond-mat.stat-mechstat.ML ·2026-10-23 ·Dario Bocchi, Matteo Negri, Flavio Nicoletti, Francesco D'Amico
Abstract · 39 pages, 9 figures

Large language models are trained to predict tokens from a finite vocabulary, each represented by a vector in a $d$-dimensional embedding space. Therefore, token prediction is a classification problem whose classes have a geometry determined by the embedding. We study how this geometry affects learning in a minimal solvable model: a linear predictor that maps a sequence of $N$ embedded tokens to a $d$-dimensional output through one $d\times d$ matrix per position, trained in the teacher--student setting. We call this the \textit{token perceptron}. We compute its generalization error with the replica method, taking $N\to\infty$ at fixed number of examples per position $α=P/N$ and keeping $d$ and the vocabulary size $D$ finite. As a first reference case, we consider continuous embeddings on the sphere, which isolate the role of the embedding dimension. The natural load is then $α/d$, and for $d>1$ a normalized label fixes the direction of the teacher output, so the problem becomes a noiseless regression: the teacher is recovered exactly above $α_{\rm rec}=d^2/(d-1)$, and spherical cross-entropy reaches this threshold at large inverse temperature. In the same setting the storage capacity of random associations scales as $d^2$ for margins $O(1/\sqrt{d})$. We then turn to vocabulary embeddings, where a label only identifies the Voronoi cell of the teacher output, so that each example imposes inequalities rather than equations, as in the classical perceptron. We compute the Bayes-optimal error for a given embedding matrix through an approximate spectral closure and show that the optimal learner first resolves the directions in which the vocabulary has the largest variance, even when the teacher is isotropic.

0.24

Correct Answers, Closed Gates: SolveBind for GUI Agents at the Verification Barrier

cs.AI ·2026-10-23 ·Xinhao Song, Haoxuan Peng, Su Su, Hongliang Wu, Sirui Song +3 more
Abstract · 41 pages, 16 figures

Websites and applications deploy CAPTCHAs and anti-automation mechanisms that use behavioral signals such as interaction trajectories and timing, not outcomes alone, to distinguish humans from automated agents. We call these protected GUI environments, where correctly solving a verification task is insufficient for end-to-end task completion. Agents must turn solver outputs into accepted interactions and adapt execution to runtime feedback under time constraints. We introduce SolveBind, a GUI-agent framework combining hierarchical solving with runtime supervision. A hierarchical solving tree decomposes verification into interdependent Perceive, Ground, Solve, Act, and Verify capabilities. A runtime supervisor grounds action commitments in tool-derived evidence, revises them as evidence arrives, and mediates execution under explicit recovery budgets. Across nine CAPTCHA task families and nine backbone models, SolveBind achieves the highest success rate on eight backbones, outperforming the strongest baseline by 13.3-34.5 percentage points. It achieves the highest success rate on all six AndroidWorld backbones. In a paired GPT-5.5 analysis, MobileAgent produces correct answers but fails dynamic verification on 13 of 90 tasks; SolveBind succeeds on seven of these tasks. With automatic submission, removing binding reduces success from 82.22% to 71.56% and increases latency and token usage. Recovery caps reduce execution cost with a small observed success trade-off: removing them raises success from 82.22% to 83.56%, while increasing latency and token usage to 3.7x and 2.4x their full-supervision values, respectively. These results support evidence-grounded execution control while identifying a success-cost trade-off in recovery and failure boundaries of heuristic submission rules.

0.24

Online Continual Learning in Edge-Cloud Applications

cs.LG cs.CV ·2026-10-23 ·Charalampos Davalas, Dimitrios Michail, Christos Diou, Iraklis Varlamis, Konstantinos Tserpes
Abstract

Deployed machine learning models in real-world IoT and edge/cloud systems face a critical reliability bottleneck: the requirement for online adaptation to non-stationary data streams while operating without access to the full original training dataset. Performance is severely degraded when models encounter Out-of-Distribution (OOD) inputs or suffer from catastrophic forgetting. Addressing these challenges requires adaptation mechanisms capable of learning from evolving data streams while preserving previously acquired knowledge. This paper proposes an edge-cloud methodology for integrating Continual Learning (CL) with Out-of-Distribution (OOD) Detection and Concept Drift Detection in IoT applications. The methodology incorporates ODIN and ECDD as alternative mechanisms for identifying informative samples for annotation and triggering rehearsal-based model updates. Experimental results across automotive, healthcare, and agricultural datasets demonstrate the applicability of the proposed methodology and its potential to support model adaptation while mitigating catastrophic forgetting.

0.24

Model-agnostic predictability quantification of multivariable time series

cs.LG ·2026-10-23 ·Yuxuan Shu, Vasileios Lampos
benchmarkcode/data
Abstract · The code is available at https://github.com/ClaudiaShu/MAPS

Multivariable time series forecasting plays an important role in various domains, where predictions depend on complex interactions among covariates. Forecasting models are commonly evaluated across numerous benchmark datasets. However, forecasting accuracy reflects both model capability and task predictability, making it difficult to characterise the difficulty of a task independently of the model used. Existing predictability measures often fail to capture nonlinear dependencies or are limited to univariate series. In this work, we propose Multivariable Predictability Score (MAPS), a model-agnostic metric for measuring the intrinsic predictability of multivariable time series forecasting tasks. MAPS uses covariate information to assess predictability across different lookback windows and forecasting horizons, and provides a quantitative assessment of the predictive information available for forecasting. We evaluate MAPS across synthetic and real-world forecasting datasets against 6 baseline predictability measures. The former confirms that our method closely aligns with the expected reference predictability (r=0.963). Empirical results further corroborate that MAPS yields more reliable projections in the majority of experimental configurations.

0.24

ZENO: Nonlinear filtering with Zakai Equation Neural Operators

cs.LG q-bio.NC ·2026-10-23 ·Aaditya Prasad, Steven W. Flavell, Brian DePasquale
Abstract

State inference and parameter learning for nonlinear filters is a core computation for scientific machine learning with varied applications in engineering and neuroscience. The unnormalized filtering distribution is governed by the Zakai equation, a stochastic partial differential equation (SPDE) that combines the forward Kolmogorov operator (known as the Fokker-Planck equation) with observed likelihood updates. Although solving these high-dimensional SPDEs is often impossible analytically and intractable numerically, advances in operator learning and physics-informed machine learning have made it feasible. Harnessing these advances, we introduce the Zakai Equation Neural Operator (ZENO), a model that optimizes a neural operator against a Zakai Equation based physics-informed loss to approximate the exact unnormalized filtering distribution for latent stochastic dynamical systems. We use this ZENO-inferred posterior distribution in an expectation-maximization (EM) framework to perform parameter learning and state inference. We demonstrate ZENO's ability to accurately infer the posterior distribution and learn its corresponding SDE parameters in three simulated stochastic dynamical systems (Double-Well bistable system, Van Der Pol (VDP) oscillator, and Lorenz system) and compare its performance against established methods for latent SDE inference and learning. We show that ZENO recovers previously observed limit cycles representing forward and backward locomotion from a whole-brain imaging dataset of immobilized \textit{C. elegans} nematodes. ZENO is a novel approach to the problem of nonlinear filtering and dynamical inference through the lens of partial differential equations and operator theory.

0.23

DealsFlow: An Evidence-Gated Agentic Harness for Venture Capital Deal Triage

cs.AI ·2026-10-23 ·Ming Ze Tang, David Cutting, Boyd Carson, Vasiliki Carson
Abstract · 13 pages, 15 tables, 2 figures. Submitted to the AI4F 2026 workshop (LLMs and Generative AI for Finance) at ICAIF 2026

Large language models (LLMs) can read startup pitch decks, but letting them emit investment scores directly invites score inflation, run-to-run variance and rationales that are hard to audit. We describe DealsFlow, a human-in-the-loop agentic harness deployed at a UK early-stage fund. The LLM reads the deck, cites evidence and rates each category; code combines the ratings, caps them and only ever lowers them. Three mechanisms implement this: a stage-aware corroboration ceiling (78, 70 or 60 out of 100) on factual categories without external evidence; a sceptic call that may lower but never raise a score; and a pre-fetch pattern that retrieves and indexes all evidence before a single generation pass. We evaluate on $N=243$ SEIS/EIS investments with recorded outcomes. Because recent investments have unresolved outcomes, we use the 142 that are at least three years old. There, the full pipeline's score has Spearman $ρ=0.45$ (95% CI 0.32 to 0.57) with the total-value multiple and AUC 0.71 for identifying write-offs, against $ρ=0.38$ and AUC 0.68 for an unconstrained LLM. The gain is small and comes mostly from the sceptic call; we did not test whether its lower-only rule matters. Investments that later exceeded $2\times$ had 64% of factual categories corroborated, against 38% for write-offs (omnibus Kruskal-Wallis test over five outcome groups, $p=0.031$), and our retrieval was not date-restricted, so these figures may overstate screening skill. In a study of 20 inbound applications, analyst time per application fell from 44.7 to 18.9 minutes including verification and correction, and 84% of 240 cited claims were fully supported by their sources, but 70% of reports needed at least one correction.

0.23

CEMIR vs. FISER: a comparative study of reasoning-based and enrichment-based approaches to pragmatic intent recognition

cs.AI cs.RO ·2026-10-23 ·Ziya Hosseynzadeh, Farjad Fallah, Farnoosh Fallah, Soroush Gooran, Mohammad Amanloo +1 more
benchmark
Abstract

Intent recognition in human-robot interaction requires agents to resolve ambiguous natural language instructions by inferring the speaker's underlying goals. Research has increasingly focused on two notable paradigms for addressing this task: structured symbolic reasoning and multimodal input enrichment. However, it remains unclear whether enrichment-based methods provide benefit in settings where the ambiguity concerns unstated human intent rather than missing observable scene information. In this paper, we present a comparative and diagnostic study benchmarking FISER - a framework utilizing structured social and embodied reasoning - against CEMIR (Context-Enriched Multimodal Intent Recognition), a method that performs multimodal enrichment by synthesizing narratives and photorealistic images from textual prompts. By evaluating both approaches across four Large Language Models (LLMs), ranging from Qwen2.5-Omni-3B to GPT-5, on the HandMeThat v2 benchmark, we analyze their capability to resolve pragmatic ambiguities. Our statistical analysis, employing multiple-comparison correction, reveals no statistically significant difference in performance between FISER and CEMIR across any model or difficulty level. This pattern holds across all sixteen model-by-difficulty comparisons, as well as in a targeted ablation isolating CEMIR's narrative and visual components. Instead, performance is primarily associated with backbone capability and the inherent difficulty of the ambiguity. These findings suggest a boundary condition for text-derived multimodal enrichment: when the relevant task constraints are already available in the textual instruction, synthetic multimodal augmentation may not provide measurable benefit over structured reasoning on the same text.

0.23

Do AI-for-Science Agents Preserve Model Safety?

cs.AI ·2026-10-23 ·Michael Yew Chuan Lee, Manya Sriram, N. Benjamin Erichson
Abstract

AI-for-Science agents are increasingly being used for formulating scientific hypotheses, running autonomous experiments, and interpreting evidence. These agents are constructed by integrating a large language model (LLM) into a domain-specific agent harness, which in turn can improve the LLM's scientific capability and enable the model to orchestrate scientific tools. In this work, we ask whether model safety is preserved when an LLM is integrated in an AI-for-Science agent harness. We evaluate five agent harnesses on 50 harmful chemistry requests. Specifically, we study how the safety behavior changes under direct prompting, two simple adversarial prompting strategies, and an adaptive attack. Across four recent GPT variants, our results show that the agent harness can increase harmful compliance even under direct prompting. More critically, we find that adaptive attacks that are largely unsuccessful against the base LLM suddenly become highly successful after the agent harness is introduced. Our results show that the safety behavior of the base LLM is not durable once a harness is introduced. Thus, as AI-for-Science agents become more capable and autonomous, the community needs to evaluate and align the agent using domain-specific safety benchmarks.

0.23

Federated Nonparametric Regression with Heterogeneous Clients

cs.LG stat.ML ·2026-10-23 ·Xiaoran Jiang, Anand N. Vidyashankar
Abstract · 123 pages, 8 figures, 6 tables

Whether a federated predictor benefits an individual client depends on both population heterogeneity and estimation uncertainty. We study this question for Nadaraya--Watson regression in a triangular array with potentially different fitting and evaluation distributions. Each client transmits local kernel summaries, allowing the server to construct a covariate-density-weighted federated estimator without centralizing individual observations. A client combines its fitted local predictor with the federated predictor through a prescribed weight, and gain is the paired reduction in squared prediction loss on an independent audit observation. Exact identities characterize both the gain of the realized fitted pair and its repeated-fitting mean. We also establish pointwise inference for the federated estimator. A two-stage analysis of client gains separates fitting variability from audit variability, retains the covariance created when a client's observations enter both its local estimator and the common server estimator, and yields fixed-block simultaneous inference across clients. Robustness is examined under two perturbation models. A Kullback--Leibler neighborhood provides a variance-sensitive lower certificate in a regular positive-gain-variance regime, while an $L_2$ neighborhood yields an exact least-favorable regression direction and robust value. Exact-model, vanishing-variance, and exact-zero-direction boundaries are kept separate because their scalar inputs are nonregular. The framework thus distinguishes decisions about a realized fitted pair from claims about the learning procedure under repeated fitting.

0.23

Lifelong Context Adaptation through Bidirectional Memory-Skill Learning

cs.AI ·2026-10-23 ·Tianyi Xu, Huazheng Wang
code/data
Abstract · 42 pages, including references and appendices. 18 figures, 13 tables

Lifelong LLM agents must learn from an expanding history of interactions while making each decision within a bounded context window. Memory makes past experience available, but the agent may still need to infer how to solve each new task from the retrieved records. Skills provide reusable procedures, yet may omit the task-specific facts needed to apply them. Because retrieved memory and skills shape the agent's behavior, they also influence the experience available for future learning. We introduce MemCryst, a framework for lifelong context adaptation through bidirectional learning of memory and skills with a frozen language model. Recurring episodes stored in memory support skill creation and revision, and the interactions guided by those skills supply new memory for subsequent learning. Facts retrieved from memory and learned skills jointly form the agent's context, while task outcomes adjust the future retrieval influence of the records actually served without rewriting their content. Across ten settings spanning embodied control, long-document reading, conversation, mathematics, and code, MemCryst leads seven reported comparisons. On ALFWorld, joint serving achieves 63.17% success, compared with 59.83% for memory-only and 50.50% for skill-only serving. Code is available at https://github.com/tianyi0216/Memcryst.git

0.23

Kernelised covariance matrices, dimension reduction and imbalanced spectral clustering

cs.LG stat.MLmath.PRmath.SP ·2026-10-23 ·Xiaoyu Dong, Subhro Ghosh, Soumendu Sundar Mukherjee, Himasish Talukdar
Abstract

The Gaussian Mixture Model (abbrv. GMM) is a classic generative model for clustered data, for which spectral clustering based on dimension reduction via PCA has been an effective framework. In this work, we investigate kernelised covariance matrices as an alternative mechanism for dimension reduction and spectral clustering, focusing on truncation kernels that select pairs of data points with suitable separation constraints. The structure of these matrices is motivated by considerations from statistical mechanics of repulsive particle systems. Our approach is particularly relevant for the setting of imbalanced clusters, where the sizes of the largest and smallest clusters have a diverging ratio with the problem size, and existing PCA-based methods turn out to be of limited effectiveness. We demonstrate efficient dimension reduction and clustering performance of these modified covariance matrices, substantiated by quantitative bounds. In particular, they achieve superior dimension reduction guarantees compared to PCA for a wide range of parameters in the regime of imbalanced clusters. For spectral clustering, our approach exhibits improvements over PCA in the imbalanced setting, in particular guaranteeing accurate clustering in a much larger regime of intra-cluster separations. Our theoretical guarantees reflected in a range of simulation experiments that exhibit these improvements for unbalanced clustering problems. Our analytical approach, based on kernel Laplacians, provide a toolbox as well as a relevant conceptual ingredient for clustering problems at large. We believe that our work will motivate the application of such kernelised covariance matrices for related classes of statistical problems, as well as spur new investigations in the theory of random matrices.

0.22

Flow Mesh: A Velocity-Regularized Flow Matching Framework for Stable Distributional Modeling of Physical Fields on Unstructured Meshes

cs.LG physics.flu-dyn ·2026-10-23 ·Siddharth Verma, Jaewan Park, Jay Phil Yoo, Kazuma Kobayashi, Souvik Chakraborty +1 more
Abstract

Thermal-fluid field distributions underpin the analysis and operation of energy infrastructure, from nuclear-reactor cooling channels to industrial heat exchangers, but resolving these distributions through computational fluid dynamics requires prohibitively large simulation ensembles. Their complex geometries also demand unstructured meshes that are poorly accommodated by generative models developed for regular grids. Here we introduce FlowMesh, a generative framework that combines permutation-invariant mesh encoding with deterministic latent flow matching and an l1 penalty on the learned transport velocity. FlowMesh generates full-field realizations directly on fixed unstructured meshes without auxiliary grid interpolation or a stochastic denoising chain. Across lid-driven-cavity, pressurized-water-reactor subchannel and heat-exchanger datasets, evaluated over six matched training seeds, FlowMesh achieves macro-averaged probability-density-function coverage of 91.7%-97.0%, exceeding an architecture-controlled latent diffusion model by $11.6$--$45.4$ percentage points. Latent-velocity regularization increases coverage by $3.8$ percentage points for reactor flows and 1.1 points for heat-exchanger flows, while reducing temperature-field Kullback--Leibler divergence by 1.7-2.4-fold. Each full-field realization requires 1.23 \pm 0.15 s, providing a 585--1902-fold wall-clock acceleration over computational fluid dynamics. FlowMesh therefore makes large-scale generative characterization of thermal-fluid states computationally feasible, providing a foundation for data-driven analysis and digital-twin development in fixed-geometry energy systems.

0.22

Watermarking Diffusion Language Models with Learned Foresight and Balanced Bands

cs.AI ·2026-10-23 ·Hyojun Kim, Hyeseon An, Yo-Sub Han
code/data
Abstract · 31 pages, 7 figures. Code: https://github.com/hyojuun/lfbb

Green-red watermarks favor tokens from a green list that a secret key derives from the preceding tokens. Diffusion language models, however, often choose a token before its preceding tokens, and the token later shapes its neighbors' green lists. Prior work therefore tilts each choice to raise the expected number of green tokens at its own and its neighbors' positions. Building on this objective, we first learn two components that prior work keeps fixed: vocabulary bands that serve as green lists, separate competing candidates and keep the null distribution exact, and a foresight predictor that estimates how each candidate shifts its neighbors' predictions. Second, we count green tokens as the detector does, once per distinct context-token pair, so repetition earns nothing. On WaterBench with LLaDA-8B-Instruct, each learned component raises the z-score at equal distortion, and their gains are nearly additive. With a strong watermark, ours detects at least 96% of responses at a 1% false-positive rate while matching the LLM-judge score of unwatermarked text. Ours also detects more responses on Dream and as many or more on code and question answering without a significant loss in task accuracy.

0.22

Singular Geometry of Low-Rank Adaptation

cs.LG stat.ML ·2026-10-23 ·Kamillo Ferry, Anthea Monod, Keisuke Yano
Abstract · 40 pages, 8 figures

Low-rank adaptation (LoRA) is now a standard approach to parameter-efficient fine-tuning of large neural networks. Its low-rank factorization is inherently non-identifiable, making LoRA a singular statistical model whose learning complexity cannot be described by parameter dimension alone. We study this complexity using the real log canonical threshold (RLCT) from singular learning theory and how it depends on both the low-rank parameterization and the pretrained network. Under a local linearization, LoRA induces a low-rank trace regression model whose Kullback--Leibler geometry is controlled by a forward-backward covariance, constructed from activations and backpropagated sensitivities at the adapted layer. This covariance determines which low-rank perturbations are statistically observable. When it is nondegenerate along the relevant low-rank directions, we recover the classical learning complexity of reduced-rank regression; when its kernel meets these directions, additional statistically indistinguishable parameters change the singular geometry. We characterize this geometry locally through observable tangent directions and a residual singularity, and globally through affine linear sections of determinantal varieties. The global learning complexity need not be determined at a ground-truth parameter, as singularities on another rank stratum can dominate. We also introduce a curvature-corrected multi-temperature estimator of the RLCT. Experiments on multilayer perceptron and transformer architectures reveal strongly layer-dependent forward-backward geometry and show that ambient covariance degeneracy does not imply degeneracy along the low-rank directions relevant to adaptation.

0.22

TENET: Time-point Encoding Network for Multivariate Time Series Anomaly Detection

cs.LG cs.AI ·2026-10-23 ·Dongchan Cho, Juwon Hwang, Keumyeong Kang, Jiho Han, Namsoon Jung
top venuebenchmark
Abstract · Accepted by NeurIPS 2026

In unsupervised multivariate time series anomaly detection (MTSAD), local variation can inflate reconstruction and forecasting errors, making observation-space discrepancy an unreliable indicator of abnormal system behavior. Since latent-space scoring can mitigate this limitation, we reframe one-class MTSAD as current-point system-state compatibility estimation and propose TENET, the Time-point Encoding Network for Multivariate Time Series Anomaly Detection. We define the current-point system state as a latent representation of the target time point, constructed from variable-level temporal context in the input window. TENET constructs this state through a temporal layer that retrieves window context relative to the current point, encodes time with current-point-centered embeddings, and adaptively mixes the retrieved context with the current representation. With this design, TENET treats the input window not as the object to reconstruct, forecast, summarize, or score, but as context for constructing a latent representation evaluated with a fixed standard-normal negative log-likelihood score. Across benchmark datasets, TENET achieves state-of-the-art performance. Additional experiments further show that TENET remains effective under alternative one-class objectives and outperforms substitutions based on generic sequence encoders, suggesting that its advantage comes from decision-aligned representation construction rather than from a specialized scoring rule. These results indicate that one-class MTSAD can be strong when the latent representation is designed around the current time point.

0.22

LLaCO: Continuous Diffusion for Large Language Models with Competition-Aware Semantic Scheduling

cs.AI ·2026-10-23 ·Zijian Jia, Peixuan Zhang, Genying Li, Di Wu, Yujie Fu +3 more
Abstract

Continuous diffusion models have emerged as a promising non-autoregressive paradigm for text generation over continuous embedding spaces. Along the continuous trajectory, the residual token identity directly shapes the denoising objective. However, the geometric trajectory misaligns with temporal progression, triggering severe temporal compression. In this paper, we show that embedding norm decay causes Euclidean distance to yield spurious overtaking, indicating that token competition must be evaluated directionally. In this space, a candidate overtakes the source token only when the effective noise magnitude surpasses their semantic separation. We formalize these pairwise dynamics as the Competition Threshold Theory (CTT). Crucially, high-dimensional overtaking requires extreme noise levels, causing standard trajectories to concentrate identity decay into a narrow window near the noisy endpoint. To resolve this, we propose Semantic Retention Scheduling (SRS) to align time allocation with the progression of identity decay, and demonstrate that existing token identity metrics are fundamentally statistical aggregations of these pairwise crossings. Building on these theoretical insights, we introduce LLaCO, a generative framework powered by competition-aware semantic scheduling. LLaCO adapts a pretrained backbone to continuous representations via an embedding directional adapter and timestep-conditioned modulation. To operate within the directional space, LLaCO employs a spherical noising process formulated as a geodesic interpolation toward a mask-biased endpoint, eliminating norm decay while preserving accurate token competition. Guided by SRS, LLaCO reallocates the denoising trajectory proportionally to the logarithmic rank decay. Instantiated as an 8B model, LLaCO outperforms previous continuous diffusion baselines and rivals discrete diffusion models of comparable scale.

0.22

Direction-Aware Token Credit Assignment with Reward-Model Gradients

cs.LG ·2026-10-23 ·Yikuan Hu, Jifeng Zhu, Minle Su, Jiaqin Yue, Gujie Shao +4 more
code/data
Abstract · 28 pages, 7 figures. Project page: https://datc-project.github.io/

GRPO shares a response-level advantage across tokens, leaving unresolved how strongly each token should contribute to learning. Updates induced by the shared advantage reinforce or suppress the sampled token at each position, but these updates do not always point toward increasing reward at the token level. This motivates a work-based view of credit: just as work depends on the component of force along a displacement, token credit should reflect the local reward change along the policy update direction. We introduce Direction-Aware Token Credit Assignment (DATC), which maps frozen reward-model gradients into policy logit space and allocates weights by evaluating the alignment between the reward direction and GRPO's signed sampled-action update direction, while preserving the sign and mean of the response-level advantage. DATC outperforms GRPO in all four settings across two model families.

0.22

TGR-EvoDAG: A DAG-based Multi-Agent Code Evolution Framework for Generative Recommendation on Tencent

cs.AI ·2026-10-23 ·Junlin He, Junzhang Zhu, Yudong Li, Yihong Tang, Xuefeng Jiang +3 more
Abstract

Large Language Model (LLM) agents have enabled automated code evolution for algorithm discovery. However, current recursive self-improvement (RSI) research for sequential recommendation faces a dilemma: general-purpose frameworks lack domain-specific feedback, while recommender-specific frameworks rely on linear evolution. Crucially, naively integrating them fails when applied to Generative Recommendation (GR). Unlike end-to-end models, GR relies on a module-cascaded pipeline (tokenization, pre-training, and alignment). In this paradigm, existing evolution methods suffer from: i) cost blindness, squandering budgets by ignoring stage-asymmetric retraining overheads; and ii) short-sighted evaluation, discarding promising upstream innovations that temporarily regress metrics before downstream adaptation. To bridge these gaps, we propose TGR-EvoDAG, a multi-agent code evolution framework tailored for GR. TGR-EvoDAG models code evolution as a Directed Acyclic Graph (DAG) to maintain multiple evolution branches and merge complementary innovations, unifying advanced search with rich GR diagnostic feedback. We further introduce stage-aware cost planning to optimize GPU-hour budgets, and active downstream co-adaptation to navigate temporary regressions and unlock cross-stage gains. On public Amazon datasets, TGR-EvoDAG enables the first autonomous evolution of a complete GR pipeline, improving offline metrics by up to 37.4% and Agent4Rec simulated user satisfaction by 19.7% over the strongest baseline. Furthermore, in industrial deployment on Tencent's large-scale commercial media platform, the evolved GR achieves a 12.16% relative improvement in offline Hit@1 and a statistically significant 0.40% increase in online click-through rate (CTR).

0.21

Isharah-Selfie: Continuous Sign Language Recognition Dataset for One-handed Signing

cs.AI ·2026-10-23 ·Ahmed Abul Hasanaath, Murtadha Aljubran, Sarah Alyami, Muhammad Haris Khan, Hamzah Luqman
datasetcode/data
Abstract

Current sign language recognition and translation benchmarks are commonly collected in controlled settings using fixed or studio-style cameras, particularly limiting their ability to reflect how signers communicate in everyday mobile scenarios. As such, this gap becomes increasingly evident when considering the Arabic Sign Language (ArSL), which is a primary sign language used across Arab countries. In this work, we introduce Isharah-Selfie, a large-scale ArSL dataset collected entirely using front-facing smartphone cameras under unconstrained selfie conditions. The dataset comprises 8,997 video clips performed by 16 signers across 1,493 unique sentences, covering several domains such as healthcare, education, and transportation. Unlike conventional datasets dominated by fixed-camera, two-handed signing, Isharah-Selfie captures realistic mobile communication characterized by one-handed signing, close-range viewpoints, partial body visibility, camera motion, unconstrained environments, and heterogeneous device resolutions. Each video is annotated with both gloss sequences for continuous sign language recognition (CSLR) and Arabic sentence translations for sign language translation (SLT). Moreover, we define signer-independent and unseen-sentences evaluation protocols to assess generalization across unseen signers and novel sentence compositions. Extensive benchmarks using RGB-based and pose-based data show that current CSLR and SLT methods remain challenged by selfie-style signing conditions, particularly under unseen-sentence evaluation, where compositional generalization remains weak. By releasing Isharah-Selfie, we aim to support research toward robust, user-centered sign language technologies that operate reliably in realistic smartphone-based communication settings. The Isharah-Selfie dataset is available on https://isharah-selfie.github.io/.

0.21

A Resource-Efficient and Precision-Scalable FPGA Accelerator for Spiking Neural Networks Using Minifloat Arithmetic

cs.NE cs.AR ·2026-10-23 ·Ali Mehrabi, Philip C. Jose
Abstract

Spiking neural networks (SNNs) are well suited to event-driven edge intelligence, but the arithmetic complexity, memory footprint, and power consumption associated with conventional floating-point implementations can limit their deployment on resource-constrained hardware. This paper presents a precision-scalable and resource-efficient SNN hardware accelerator architecture based on parameterized minifloat arithmetic. The proposed architecture implements recurrent leaky integrate-and-fire (LIF) neurons and leaky-integrator (LI) readout neurons using a unified minifloat datapath whose word length and mantissa width can be configured at synthesis time. A valid-driven arithmetic subsystem employs pipelined minifloat multiplication and addition, while time-multiplexed processing enables arithmetic resources to be shared across synaptic accumulation, neuron-state updates, recurrent connections, and output processing. This organization reduces replicated arithmetic hardware and provides a scalable design approach for different SNN sizes and precision requirements. The architecture supports E8M23, E5M10, E5M6, and E4M3 formats within the same RTL framework, allowing precision, memory capacity, timing, and power to be traded without modifying the underlying neuron architecture. FPGA implementation results demonstrate that reducing minifloat precision substantially decreases arithmetic-resource utilization and estimated dynamic power, while also reducing the theoretical storage requirement of synaptic parameters. A recurrent event-based classification network is used as a hardware validation case, demonstrating that substantial precision reduction can be achieved while retaining competitive inference accuracy. The results establish parameterized minifloat computation as an effective basis for scalable, low-resource, and energy-efficient FPGA SNN accelerators.

0.21

Reinforcement Learning for Controller Parameters in the Auto-Bidding System at Pinterest

cs.LG cs.IR ·2026-10-23 ·Yichu Zhou, Qinyi Chen, Fei Feng, Jacob Loh, Mehdi Ben Ayed +3 more
Abstract

Auto-bidding systems in online advertising must balance competing objectives-maximizing budget delivery while minimizing cost per action-across heterogeneous campaigns under non-stationary market conditions. This trade-off evolves over campaign lifespans: early on, cost optimization dominates when budgets roll over; near deadlines, budget delivery becomes urgent. While PID controllers are widely used in production for their stability and interpretability, fixed tunings cannot adapt to diverse campaigns and shifting traffic patterns. We present RLBP (Reinforcement Learning for Bidding Parameters), a hybrid architecture deployed at Pinterest that uses RL as a meta-controller to select campaign-specific PID parameters daily while retaining PID for high-frequency bid adjustments. RLBP trains a multi-head Q-network using a relative-lift reward formulation normalized against the production baseline, and performs inference-time fusion that adapts the cost-delivery trade-off to campaign states without retraining. Through online A/B testing at Pinterest, RLBP achieves a statistically significant reduction in effective cost-per-click (eCPC) with no revenue degradation, and two months of post-rollout measurements against a holdout also show consistent cost reductions. These results demonstrate that meta-controller RL can successfully augment production control systems at scale.

0.21

Augmentation-Guided Adversarial Training via Local Stability-Based Sample Selection

cs.LG ·2026-10-23 ·Daniel Agyei Asante, Mark Tetteh, Deborah Asamoah, Bernard Asante, Regina Esi Turkson
Abstract

Adversarial training remains one of the most effective defenses against adversarial attacks, but its standard PGD-based training is computationally expensive because adversarial examples are generated for every training sample. Selective adversarial training reduces this cost by perturbing only a subset of samples. However, existing selective techniques typically rely on point-wise difficulty signals such as loss, entropy, margin, or gradients. We show that these point-wise criteria do not capture local instability in a sample's neighborhood. Motivated by this, we propose Augmentation-Guided Adversarial Training (AGAT), a selective adversarial training framework that uses augmentation-induced local variation to allocate samples with unstable neighborhoods for adversarial training. AGAT ranks samples using Augmentation-based Local Stability (ALS), a lightweight score that measures the normalized change in model logits between an input and its label-preserving augmented variant. Experiments across different architectures on MNIST, CIFAR-10, CIFAR-100, and Tiny-ImageNet-200 demonstrate that AGAT achieves comparable or higher robust accuracy while applying adversarial supervision to only 20-30% of the training samples in each minibatch. AGAT also requires less training time than all other adversarial training baselines in the computational cost evaluation.

0.21

Amortized Thompson Sampling and Efficient Information-Directed Sampling for Simulator-Based Design

cs.LG ·2026-10-23 ·Michal Kobiela, Michael U. Gutmann
Abstract

Thompson sampling is a widely used strategy for sequential decision-making under uncertainty, offering a principled way to balance exploration and exploitation. This strategy involves maximizing a target objective under a single sample drawn from the current belief over unknown quantities. Conventionally, this optimization is performed once the belief has been inferred from data, and it becomes computationally burdensome when many Thompson samples are required, especially in continuous action spaces. In this work, we introduce an approach that amortizes this process ahead of time, before any data is collected, so that an approximation of the full Thompson sampling distribution is available immediately once observations arrive. We employ this approach to solve the multi-round setting, involving sequential data acquisition and Thompson-sample recommendation across rounds. Beyond accelerating Thompson sampling, our approach enables efficient estimation of expected information gain about the optimal action. This can be used to select more informative Thompson samples or to obtain maximally informative designs with optimization. Combining this estimate of expected information gain with immediate regret supports information-directed sampling (IDS), which can outperform Thompson sampling by explicitly balancing these two quantities. We demonstrate our approach on two simulator-based design problems in synthetic biology, where the simulators have intractable likelihoods.

0.21

Neural LoFi with Backward Coupling: A Spectral Theory of Cross-Layer Feature Learning

cs.LG cond-mat.dis-nnstat.ML ·2026-10-23 ·Yatin Dandi, Matteo Vilucchio, Hugo Tabanelli, Florent Krzakala
Abstract

During joint training, backpropagation allows later layers to reshape earlier representations, a feedback mechanism absent from purely layer-wise learning. We extend Neural Low-Degree Filtering (Neural LoFi), a recently proposed layer-wise spectral surrogate for feature learning, to capture this interaction. Starting from the exact backpropagation gradients, we derive a local coupling between a neuron's incoming weights and its corresponding next-layer column. In a controlled small-initialization regime with smooth activations and fixed surroundings, these dynamics amplify singular modes of a cross-layer correlation operator. Each growing mode pairs an earlier feature with an upper-layer response through shared neuron coefficients. This motivates a backward-coupling algorithm in which upper-layer sensitivities generate sample-dependent effective labels for earlier spectral steps, while corrected features refine upward transmission. A function-space variational principle characterizes the selected feature-response pairs by maximizing the label correlation of their product under kernel-norm constraints. For three-layer networks, we derive self-consistent continuous-time dynamics for the filtered hidden-layer features, the activation kernels, and the network output under the proposed backward coupling. In a hierarchical multi-index model, we establish a feature-recovery separation: backward coupling recovers lower-layer directions invisible to the original forward label-weighted statistic. The linear specialization recovers residual-dependent singular-mode dynamics, while paired modes provide a dynamical mechanism for linked low-rank covariance components consistent with the rainbow-network picture. Experiments with fully connected and convolutional networks on binary CIFAR-10 show improved accuracy over forward Neural LoFi at matched widths.

0.20

Regularised RL improves sim-to-real transfer

cs.LG cs.RO ·2026-10-23 ·Danilo Vucetic, Mohamad H. Danesh, Florence Clerc, Hsiu-Chin Lin, Jonathan Tremblay
Abstract · 9 pages of main-body content, with supporting details in the appendices. Under submission

Regularised reinforcement learning (RL) has provable robustness guarantees that should ease the transfer of policies from simulation to robots, known as sim-to-real. Yet, weakly regularised algorithms like Proximal Policy Optimisation (PPO) remain the de facto choice in sim-to-real pipelines, while the practical benefits of regularisation for transfer remain poorly characterised. In this work, we study the sim-to-real performance of commonly used RL algorithms, with the aim of characterising the effects of regularisation on robotic performance. To that end, we derive Soft MellowMax Actor-Critic (SMM-AC), a novel regularised RL algorithm based on the soft mellowmax value function, which carries strong robustness guarantees. On MuJoCo continuous control benchmarks, SMM-AC is competitive with baselines like Soft Actor-Critic (SAC). On real robots, SMM-AC achieves higher return, while exhibiting greater robustness to perturbations in observations, the environment, and dynamics than baselines. Together, these results demonstrate that regularised RL improves real-world policy deployment through better robustness to the sim-to-real gap, and changing real-world conditions.

0.20

When to Stop Penalizing Actions? Martingale GRPO for VLA Reinforcement Learning

cs.AI ·2026-10-23 ·Zongyao Zhao, Cheng-Jie Zhao, Ruihang Jiang, Bing Zheng, Zhaolin Wang +1 more
Abstract · 31 pages, 9 figures

A failed vision-language-action (VLA) rollout does not imply that every action contributed to failure. Yet in group-relative policy optimization (GRPO), a single trajectory-level advantage is assigned to all actions in a rollout. Potentially useful recovery attempts can thus be penalized alongside the errors they aim to correct. Ideal stopping conditions for failure feedback are characterized using success-probability and policy-score martingales. At these stops, the removed gradient contribution has zero conditional expectation given the retained information. Truncation therefore preserves the expectation of a reference GRPO group-gradient estimator and can reduce its covariance. One such stop occurs when the conditional probability of success is zero, motivating error and recovery cues for selecting candidate feedback boundaries. Martingale GRPO (MART-GRPO) is introduced with a vision-language model (VLM) for event-guided truncation. The VLM is fine-tuned on our annotated training set using low-rank adaptation (LoRA). The last visible error without observed recovery is localized by the frozen VLM in complete rollout videos. Failure feedback is retained through the selected error interval and removed afterward, while successful-rollout loss terms and original loss normalization are preserved. For retrospective boundaries, changes in expectation and variance are analyzed to assess departures from ideal stopping. Experiments with OpenVLA-OFT on RoboTwin 2.0 show that MART-GRPO improves success rates over full-trajectory GRPO across multiple tasks.

0.20

Spatial Heuristic Strategy based Diffusion Reinforcement Learning for Online 3D Bin Packing

cs.AI ·2026-10-23 ·Tong Li, Qingyang Xu
Abstract

Traditional heuristic methods have accumulated mature experience in core objectives, such as space utilization and layout compactness of 3D bin packing, relying on artificially designed rule-based logic. However, they exhibit obvious limitations in generalization ability and adaptability to dynamic scenarios. Diffusion reinforcement learning, which integrates the spatial search capability of reinforcement learning and the data distribution modeling capability of diffusion learning, has achieved certain progress in action distribution modeling and decision compliance constraints. Nevertheless, there remain issues including insufficient perception of the spatial topological structure in 3D bin packing and limited layout optimization capability. To address these problems, we propose a diffusion reinforcement learning method integrated with spatial heuristic strategies, which combines classical bin packing heuristic rules with a diffusion reinforcement learning framework. Specifically, it compresses the action space through heuristic action pre-screening, introduces a graph neural network to model the spatial structure, and designs a composite reward function incorporating flatness to suppress spatial fragmentation, thereby realizing the organic integration of prior knowledge and adaptive optimization. Experimental results demonstrate that the pro

0.20

Learning Nonlinear Waveform Fusion Across Scales for Integrated Sensing and Communication

cs.AI ·2026-10-23 ·Jiayu Liu, Kai Wan, Shuo Shao, Giuseppe Caire
Abstract

Fixed source superposition and frequency partitioning constrain how communication content interacts with a probing reference in integrated sensing and communication (ISAC). We learn nonlinear waveform fusion across scales for image transmission and localization of multiple targets in range and angle. Reception uses the information available at each end. Adaptive temporal feature weighting and spectral updates combine image features and a deterministic linearly frequency-modulated reference in a shared complex state, generating a waveform with fixed total energy over the array frame. At the user equipment (UE), the reference and channel state information condition a neural demixer for image reconstruction. At the base station, the realized transmit pulse defines a coherent dictionary for sequential localization with learned local correction, physical refinement, and residual updates. We compare the proposed waveforms with superposition in the time domain and learned spectral linear fusion on 512 independent test images, using a shared fixed UE receiver and a baseline sensing receiver using coherent peak search. Averaged across eleven task weights, the proposed waveforms reduce normalized localization error by 29.0\% and 27.0\%, respectively, while increasing mean peak signal-to-noise ratio by 1.40 and 0.82~dB. With the proposed waveforms fixed, learned local correction reduces mean normalized localization error by 12.2\% relative to the same sequential receiver without learned correction.

0.20

Reimagining Enterprise RAG for Multi-Agent Systems: A Knowledge Layer That Organizes Before It Retrieves

cs.AI ·2026-10-23 ·Garima Agrawal, Prasun Das, Akshay Sarvade, Priyanka L, Minisha N +7 more
Abstract

Enterprises are deploying agentic AI across the software development lifecycle, and they need the behavioral logic of the systems they act on, not stakeholder documentation. A seven-agent reverse-engineering pipeline recovered this logic, processing 241 repositories across 18 services into thousands of artifacts describing scope, journeys, rules and acceptance criteria. Recovery does not confer usability. Entities like offer, eligibility, and activation recur across services under a shared vocabulary, related in name but operationally distinct, so one end-to-end journey implicates over a thousand near-identical rules that must not be conflated. Flat vector retrieval fails, returning an identical rule from the wrong service at high similarity with no signal of error. Content-level ontology, the traditional remedy, did not exist, and relating every occurrence of recurring entities would have cost more than the engagement and not survived a release cycle. We place the ontology at the artifact level. In this machine-generated corpus, each artifact retains its producer, component, service, domain and type, provenance that disambiguates before content is read. We present a knowledge layer that organizes before retrieving: a self-describing catalog, a sparse metadata registry, and a navigation hypergraph defined over artifacts rather than extracted entities. Queries resolve to artifact identifiers that constrain vector search, so the competing rule from another service is absent rather than outranked, making disambiguation structural, not statistical. Built in under a week and admitting new services by convention, the layer serves SDLC agents, developers, quality engineers and analysts. It supported root-cause analysis and resolution of 100+ production defects with developer verification and attributed 35 release regression failures to specific rules and services.

0.20

PersonaScienceBench: Evaluating Scientific Agents with Diverse Simulated Researchers

cs.AI ·2026-10-23 ·Siyang Wu, Weizhi Zhang, Chen Yang, Esther Xu Fei, Jorin Overwiening +32 more
Abstract

Scientific agents collaborate with researchers whose expertise and preferences differ, but benchmarks usually hold the collaborator fixed. We introduce PersonaScienceBench: 200 tasks from five benchmarks across four scientific domains. Simulated researchers provide context and feedback according to persona profiles, while task objectives and grading remain fixed. Agents encounter researcher preferences only through conversation. Across eighteen modelconfigurations, task performance, fact recovery, and self-repetition reveal different strengths. Detailed GPT 6 Astra analyses show distinct agent responses across researcher profiles; interaction traces illustrate targeted clar ification and workflow revision. These behaviors can differ even when final scores match. PersonaScienceBench therefore evaluates how agents collaborate with diverse researchers alongside whether they complete the task.

0.20

Internal Discrepancy for Contrastive Learning: Generalization Bounds and Downstream Behaviors

cs.LG cs.AI ·2026-10-23 ·Jialiang Wang, Qing Ma, Xiong Zhou, Gangfeng Hu, Haoliang Li +1 more
Abstract

Contrastive learning has emerged as a popular paradigm for training representation models. While prior studies pursue more practical and less restrictive assumptions to enhance theoretical understanding of contrastive learning, they essentially describe the interconnections among representations within the data or induced by the function class. To encapsulate these assumptions, we propose the measure of internal discrepancy that quantifies the difference between the mean representation of positive samples and latent cluster centers. We then prove an internal discrepancy-aware generalization bound for downstream supervised tasks following contrastive learning, without requiring additional assumptions. This generalization bound not only serves as a unifying framework connecting various assumptions but also motivates us to further reduce the internal discrepancy, thereby introducing the relaxed k-means loss as an approximation. Moreover, we investigate the downstream behaviors of contrastive learning from the perspective of contrastive learning-incurred supervised losses. In the presence of class imbalance, we shed light on why contrastive learning can facilitate imbalanced learning. Finally, we investigate the optimality of these incurred supervised losses, proving several intriguing properties.

0.20

Information-Theoretic Scaling Laws for Generative Diffusion Training

cs.LG ·2026-10-23 ·Druv Pai, Dejan Stancevic, Yi Ma, Luca Ambrogioni
code/data
Abstract · 42 pages, 15 figures

Diffusion models trained on increasingly large datasets sampled from the same data distribution will converge to the same learned distribution. However, in practice, a model trained on finite data depends on the particular training dataset; this dependence is reflected in properties of the model's outputs, such as memorization and sample quality. To understand this dependence through an information-theoretic lens, we study \textit{model information}, a computationally tractable measure of how much samples from the trained model depend on the particular training dataset. We explore its scaling behavior as model size and training dataset size vary. Specifically, we uncover two qualitatively distinct scaling regimes: (1) a nonlinear model information regime marking the transition from pure memorization to imperfect generalization, and (2) a generalization regime characterized by linear scaling of the model information and progressively improved generalization and sample quality. We then posit an information-theoretically principled scaling law to quantify these two regimes, and show that this law yields induced scaling laws for memorization and sample quality. We finally demonstrate that these induced scaling laws extrapolate well outside the fit's domain. Our results demonstrate information-theoretic principles for memorization and sample quality scaling laws in diffusion models. Our code is publicly available on \href{https://github.com/DruvPai/diffusion_info_theory_scaling_law}{GitHub}.

0.20

MOSHO: A Native Discrete Shark-Inspired Metaheuristic for Hardware-Aware Neural Architecture Search

cs.NE cs.AIcs.LG ·2026-10-23 ·Nazim Abdeerhmane Aouni, Zakaria Soualah Mohammed, Maroua Ogab, Nour El Imane Elbar, Rayane Rahmat Errahmane Smara +5 more
benchmark
Abstract

Hardware-aware neural architecture search (NAS) requires jointly optimizing predictive accuracy and device-specific latency over discrete search spaces. While multi-objective swarm optimizers show promise, continuous formulations such as the Multi-Objective White Shark Optimizer (MOWSO) can suffer from quantization noise when rounded to integer-encoded architecture vectors. To address this limitation, we propose MOSHO (Multi-Objective Shark Hunting Optimization), a native discrete metaheuristic that replaces continuous motion equations with discrete architecture-editing operators. MOSHO integrates adaptive operator scheduling, energy-regulated restarts, and deferred crowding-distance truncation to enhance exploration and search efficiency. On the merged NAS-Bench-201/HW-NAS-Bench benchmark across 18 contexts, MOSHO achieves a median normalized hypervolume of 0.9987 and IGD+ of 0.0064, yielding Pareto fronts competitive with NSGA-II while reducing search runtime by 96.4% (0.64 s vs. 17.72 s). These results demonstrate that tailored discrete search mechanisms can enable swarm optimizers to achieve high-quality multi-objective solution quality with substantially lower search cost in hardware-aware NAS.

0.19

Solving and Learning Navier-Stokes Dynamics with the Neural Basis Method

cs.LG physics.flu-dynmath.NA ·2026-10-23 ·Yuhe Wang, Zhu Wang, Min Wang
Abstract · 23 pages, 6 figures

Physics-informed machine learning has become a major direction in predictive modeling of fluid dynamics, combining governing physics with learned representations. Yet prevailing formulations neither guarantee that small PDE-residual losses imply small solution errors nor scale effectively with increasing problem size. Here we present the neural basis method (NBM) as a new approach for Navier--Stokes dynamics. NBM is built on a ``representation-then-projection'' principle: fixed neural bases define the representation space, while projection of the governing equations determines the expansion coefficients. The same bases define common coordinates for the parametric solution manifold, recasting operator learning as a parameter-to-coefficient mapping and enabling scalability by removing explicit spatial dependence. For incompressible flow, NBM yields a velocity-only formulation whose residual provides an a posteriori velocity-error estimate, establishing a certified, data-free training objective for coefficient learning. Theoretical analysis and numerical experiments demonstrate the accuracy, efficiency, generalization, and scalability of NBM, with substantial improvements over baselines.

0.19

POSEIDON: Scaling Training Environments for Scientific Agents from Code Repositories

cs.AI ·2026-10-23 ·Qifan Zhang, Yuhan Li, Nuo Chen, Bowen Liu, Xiao Hu +6 more
Abstract

Training environments remain a major bottleneck for agents in data-driven scientific discovery. Existing collections of executable and verifiable training environments are small, and their construction remains difficult to scale. To address this limitation, we introduce Poseidon, an automated pipeline for generating diverse, verifiable training environments from scientific code repositories at scale. From the same repository budget, Poseidon extracts nearly 10 times as many training environments as a prior approach by decomposing scientific workflows into four complementary task types: Reproduction, Subflow, Completion, and Correction, where each type targets a distinct capability for scientific research. A recursive improvement loop further refines failed candidates and recovers otherwise discarded environments. Using Poseidon, we construct Pearl, a collection of 97,659 executable and verifiable environments from 7,110 public repositories spanning 10 scientific disciplines. To the best of our knowledge, Pearl is the largest and most broadly multidisciplinary collection of scientific-agent training environments to date. Qwen3.5 models (9B, 27B, and 35B-A3B) trained on Pearl achieve consistent gains across six challenging scientific-agent benchmarks. Overall, the scale and diversity of Pearl offer opportunities for training scientific discovery agents.

0.19

Near-Optimal List Replicability from Littlestone Dimension

cs.LG ·2026-10-23 ·Ari Blondal, Pooya Hatami, Chavdar Lalov, Sivan Tretiak
theory
Abstract · 14 pages

We study the list replicability number in the realizable PAC learning framework and prove that every concept class $H$ over a countable domain with Littlestone dimension $d$ and dual VC dimension $v^*$ satisfies \[ \mathrm{LR}(H)\leq 2^{O(d\min\{\log d,v^*\})}. \] This matches the worst-case lower bound of $2^{Ω(d)}$ up to a logarithmic factor in the exponent and exponentially improves the previous best upper bound $\mathrm{LR}(H)\le 2^{2^{O(d)}}$ of Bun, Livni, and Moran [FOCS 2020]. Unlike earlier bounds of $2^{O_ε(d^2)}$ for fixed accuracy $ε>0$ implied by the work of Ghazi, Golowich, Kumar, and Manurangsi [STOC 2021] and Lyu [STOC 2026], our bound is independent of $ε$. The $2^{O(d\log d)}$ bound follows from a new inductive framework for bounding list replicability that builds on the recent notion of irreducibility from differentially private learning. The $2^{O(dv^*)}$ bound follows from a simple globally stable learner that takes the majority vote of multiple empirical risk minimizers, with stability established using tools from stable regularity.

0.19

Retiming Robot Demonstrations with a Learned Speed Prior

cs.AI cs.RO ·2026-10-23 ·Seungyong Moon, Deokjae Lee, Hansang Lee, Munkhtselmeg Battsengel, Youngin Kim +5 more
Abstract

The pace of a robot demonstration is often determined by the teleoperator rather than by the task itself. As a result, the same motion can appear at widely different speeds across demonstrations, and policies trained on such data may inherit operator-specific timing along with task behavior. We propose to learn a speed prior directly from large-scale demonstration data. Given a task instruction, a motion path and the images along it, it predicts a distribution over execution speed along the path, from which a chosen quantile defines a consistent clock for retiming each demonstration. Existing methods derive the acceleration signal from the very data they accelerate, so each new task costs another fit, annotation, or pair of predefined rates. A single trained speed prior instead covers every task in a corpus. We fit a speed prior on both robot teleoperation data and handheld demonstrations in which a human arm, rather than a robot arm, moves the gripper, and use it to retime robot demonstrations to the handheld pace before fine-tuning a vision-language-action policy. For execution, we pair the policy with a real-time optimal-control layer that tracks predicted action chunks under robot constraints using the predicted pace. On 120 RoboLab tasks, policies fine-tuned on our learned clocks succeed more often and move with up to half the jerk compared with policies fine-tuned on the original demonstration timing and uniformly accelerated to the same speed. On a real robot, our policy stacks cups in 3.50s, 1.72$\times$ faster than the default policy's 6.02s while succeeding as often, in 29 of 30 trials. Videos and further material are available at https://mllab.snu.ac.kr/projects/learned_speed_prior.

0.19

Deep Learning-Based Prediction of HER2 Status from Whole-Brain MRI in Breast Cancer Brain Metastases

cs.LG ·2026-10-23 ·Sara Early, Chris Kang, Vibujithan Vigneshwaran, Kimberly Amador, Roger Y. Tsang +1 more
Abstract

Purpose: Human epidermal growth factor receptor 2 (HER2) status is a key biomarker relevant for prognostication and treatment decision making in metastatic breast cancer. Thus, the aim of this work was to explore whether HER2 status can be predicted directly from whole-brain post-contrast T1-weighted (T1w) MRI scans in breast cancer brain metastases (BCBM) using deep learning, and whether providing explicit tumour segmentation information improves prediction. Methods: Using the publicly available BCBM-Radiogenomics dataset, we analysed 135 patients, including 70 HER2+ and 65 HER2- patients. Three approaches were implemented and evaluated: a conventional radiomics model, a simple fully convolutional network (SFCN) using T1w MRI scans alone, and an SFCN using T1w MRI scans with an expert-derived tumour mask as an additional input channel. Results: The radiomics model achieved an area under the receiver operating characteristic curve (ROC-AUC) of 0.602 (95% confidence interval [CI]: 0.528--0.677) while the SFCN achieved a ROC-AUC of 0.667 (95% CI, 0.63--0.71) using T1w MRI scans alone, which increased to 0.692 (95% CI, 0.58--0.81) when an explicit tumour mask was additionally provided. Conclusion: However, the modest performance across all approaches demonstrates the difficulty of predicting HER2 status from structural MRI scans alone. The difference between the segmentation-free and mask-guided SFCN models suggests that much of the predictive signal captured by the deep-learning model may be learned directly from whole-brain T1w MRI without requiring explicit tumour delineation. These findings support further investigation of segmentation-free deep-learning approaches as a potential non-invasive strategy to support HER2 status determination in BCBM.

0.19

DIGOM: Disentangled Interaction Graphs for Outcome prediction in Multimodal healthcare

cs.LG cs.AI ·2026-10-23 ·Hakima Laribi, Olivier Lefebvre, Dan Poenaru, Martin Vallières
Abstract

Multimodal healthcare data contain relational information that extends beyond individual patient observations. Effective clinical outcome prediction therefore requires not only integrating information from multiple modalities, but also modeling the relationships between them. Standard fusion approaches do not explicitly capture these interactions, while existing graph-based approaches typically model only a subset of them. We propose DIGOM (Disentangled Interaction Graphs for Outcome prediction in Multimodal healthcare), a graph-based framework that jointly models three complementary interaction types: Intra-patient cross-modal, Cross-patient same-modal, and Cross-patient cross-modal interactions. DIGOM learns a sparse multiplex graph with disentangled views, each encoding interactions between pairs of modalities across patients. Discrete adjacency matrices with patient-adaptive connectivity are sampled via a differentiable mechanism, enabling end-to-end learning and allowing the model to select informative relationships without imposing a fixed number of neighbors. A regularization objective maintains overall graph sparsity, while message passing integrates this relational context into patient representations for outcome prediction. We evaluate DIGOM on two multimodal healthcare datasets, MIMIC-IV and NIH-TB, across five clinical prediction tasks. DIGOM achieves the highest AUROC on all five tasks while using fewer connections than existing multi-interaction approaches. Ablation studies reveal that each interaction type's contribution varies across tasks, highlighting the need to adapt relational structures per task. DIGOM also demonstrates greater robustness to missing modalities, suggesting learned relationships provide complementary information when modalities are unavailable. Post-hoc analyses further reveal task- and outcome-specific connectivity patterns in the learned graphs.

0.19

Behaviour analysis for large language model-driven agents

cs.AI cs.CL ·2026-10-23 ·Tianqiang Yan, Lizhen Qu
Abstract · Under review by Nature Machine Intelligence

Understanding language-agent behaviour requires learning how actions are organised across a trajectory and how this organisation relates to context and outcomes. We propose a general framework that connects data-driven behavioural discovery with analysis and experimental investigation. Categories developed from recorded activity define what is measured. Analyses preserve the behaviours' positions in the process and express their relationships as interpretable hypotheses. Statistical assessment and tests in held-out contexts examine the strength and scope of the evidence. Experiments, including prompt steering, then test how behaviour and outcomes change under deliberately varied conditions. A study of iterative self-reflection illustrates how discovery, dependency learning and prompt comparisons can inform one another. We develop this approach for trajectories of varying length, event order and agent roles, and identify advances in coding, analysis and experimentation needed to support its wider use.

0.18

CASE: Confusion-Aware Skill Evolution for Frozen Multimodal Emotion Models

cs.AI ·2026-10-23 ·Siyu Wu, Zeyu Liang, Linghao Wang, Qiyun Zheng, Yi Yao +2 more
benchmark
Abstract · 9 pages, 6 figures, 4 tables

Multimodal large language models (MLLMs) have substantially advanced Multimodal Emotion Recognition (MER) by enabling fine-grained emotion understanding through reasoning over multimodal evidence. However, they still exhibit systematic confusion between semantically adjacent emotions with highly overlapping multimodal cues. Existing approaches typically address these errors through global parameter adaptation, making confusion correction costly and inflexible. We instead observe that emotion confusion is highly localized and therefore better addressed through reusable pair-specific expertise. Building on this insight, we propose Confusion-Aware Skill Evolution (CASE), the first agent-skill framework for MER. CASE maintains a sparse, self-evolving library of pair-specific EmotionSkills that externalize affective expertise from model parameters, enabling the learned expertise to be continuously refined and readily transferred across diverse frozen MLLM backbones. During inference, CASE dynamically retrieves the appropriate EmotionSkill to reinforce boundary-aware reasoning while keeping the MLLM backbone frozen. Extensive experiments on CA-MER, DFEW, and EMER demonstrate the effectiveness and transferability of CASE across both general-purpose and emotion-specialized MLLMs. Notably, CASE improves the average accuracy of Qwen3-VL-8B on the CA-MER benchmark from 46.80% to 59.13%, while achieving consistent gains across MLLMs.

0.18

Neyman Orthogonal Activation Steering

cs.LG ·2026-10-23 ·David Arbour, Erfan Jahanparast, Serina Chang, Avi Feller
benchmark
Abstract

Activation steering is a popular approach for shifting LLM output toward a desired behavior, and increasingly for generating realistic synthetic data. The standard recipe estimates a steering direction by contrasting LLM activations on prompts with and without a target concept, such as truthfulness. In this paper, we separate the estimand (the steering direction these methods target) from the estimator used to recover it, two notions the literature often conflates. We frame nearly all existing methods as plug-in estimators of the average gradient of an outcome regression, with the binary contrastive label fixed as the outcome. Using this framing, we can immediately incorporate flexible modeling approaches and extend existing methods to allow for continuous outcomes such as log-probabilities or oracle ratings. Flexible ML models that fit the outcome well, however, might not estimate the gradient well, which motivates estimators that target the gradient directly. Drawing on the debiased machine learning literature, we develop a complementary estimator built on the Stein score, the Riesz representer of the gradient functional. Combining the two routes gives an orthogonalized estimator with classical double-robustness properties. We illustrate the framework on activation steering applied to a number of benchmark datasets.

0.18

DriveMind: Mind-Evolving Belief Tracking for Closed-Loop Autonomous Driving

cs.AI ·2026-10-23 ·Yan Shi, Kehua Chen, Jiarui Qi, Yihong Tang, Haicheng Liao +1 more
published
Abstract · Accepted by Neurips 2026

Recent multimodal large language models (MLLMs) can perceive a traffic scene, reason about it, and propose a driving action within a single model. However, existing MLLM-based drivers reason from scratch at each frame and commit to a single explanation, discarding earlier hypotheses about the scene. When the scene is ambiguous, the agent locks in on the wrong interpretation and acts on it before evidence accumulates. In this paper, we propose DriveMind, an MLLM-based driving framework that keeps several hypotheses about the scene alive at the same time and updates their weights as new frames arrive. Inspired by particle filtering, DriveMind propagates each hypothesis through a VLM, scores it against the current observation, and resamples to maintain diversity. On Bench2Drive, adding belief tracking to the same driving pipeline improves the Driving Score (DS) from 68.50 to 81.37 and the Success Rate (SR) from 41.36% to 54.09%. The largest gains occur on ambiguous scenarios such as unsignalized intersections (+20.0 DS) and oncoming-traffic interactions (+20.5 DS). Beyond the overall metrics, DriveMind also improves fine-grained interactive driving abilities such as merging, emergency braking, and give-way.

0.18

Contextual Decision Making: Predictive or Generative

cs.LG math.OCstat.ML ·2026-10-23 ·L. Jeff Hong, Weihuan Huang, Xinyao Li
theory
Abstract · 63 pages

Contextual decision making uses observed covariates to guide decisions under uncertainty. Two natural approaches are predict-then-optimize (PtO), which learns the conditional expected objective, and generate-then-optimize (GtO), which learns conditional uncertainty and optimizes using generated scenarios. We study when each approach is preferable by examining how the choice of learning target affects statistical accuracy and optimization tractability. Although PtO learns a narrower target, our regret bounds show how differences in smoothness can offset GtO's additional dimensional burden. Exploiting smoothness also raises an optimization tradeoff: within the class of nonnegative-weight PtO methods studied, a lower bound limits the gains from additional smoothness, while more flexible methods may introduce signed weights and lose the guarantee of preserving convexity. GtO preserves scenario-wise convexity, but requires learning a richer object and generating scenarios for optimization. Synthetic experiments and a real case study illustrate how these tradeoffs vary with data availability and problem structure. Together with the scope for reusing a model across tasks, our findings provide practical guidance for choosing between the approaches without identifying a universal winner.

0.18

Sparsity-driven emergence of functional modularity in multitask neural networks

cs.LG cs.AI ·2026-10-23 ·Bowen Shi, Qi Ouyang, Long Qian
Abstract · 34 pages, 6 figures, including Supplementary Information

Neural networks can learn multiple tasks, but how learning organizes shared computation into specialized and reusable components remains poorly understood. In this work, we show that sparsity inducing regularization promotes functional modularity in multitask networks without predefined module boundaries. Using functional profiling in multilayer perceptrons and a lightweight transformer, we identified compact, task-selective subnetworks, termed functional tokens. Across rule-based task families, selectively retaining these tokens preserves their associated functions while reducing performance on other tasks. Such specialization develops preferentially in deeper layers during training, while earlier layers retain shared computation. Regularization strength and form shape module compactness and specialization, while task weighting alters functional allocation. Moreover, extracted functional tokens can be recombined through alternative pathways to reconstruct a learned composite computation without retraining, preserving their connectivity and learned weights. Our results suggest that sparsity-constrained optimization shapes not only parameter use but also the organization and reuse of learned computation.

0.18

Beyond Channel Data: Configuration-Enriched Radio Foundation Models

cs.AI eess.SP ·2026-10-23 ·Shrihari Vasudevan, H. G. Ranjani, Raaghul Ranganathan, Satwik Sahu, Thirumaran Ekambaram
Abstract · Submitted to IEEE ICC 2027 for review

This paper proposes incorporating channel configuration (CFG) information into Radio Foundation Model (RFM) pre-training and demonstrates that the resulting CFG-enriched embeddings serve as effective inductive priors for downstream tasks under label scarcity. We compare the proposed embeddings with raw channel tokens and multiple baseline models, under three different data regimes, for multiple downstream tasks --- Line-of-Sight/Non-Line-of-Sight (LoS/NLoS) classification, timing advance (TA) estimation, and beam selection. Evaluation is done in both in-distribution and out-of-distribution configuration settings. Experimental outcomes show that the proposed embeddings consistently outperform and/or are competitive to the baselines, across experiments in both seen and unseen configuration scenarios.

0.18

Reinforcement Learning with Multi-Step Transition Predictions: A Bellman Theory

cs.LG ·2026-10-23 ·Chenbei Lu, Zaiwei Chen, Tongxin Li, Chenye Wu, Adam Wierman
Abstract

Multi-step predictions of future transition realizations can improve reinforcement learning, but their value remains difficult to quantify. Directly including these predictions in the state can lead to exponential state-space growth, making direct dynamic programming computationally expensive. We study discounted finite MDPs under a block-feedback protocol: at the start of each K-period block, the decision maker observes $K$-step probabilistic transition predictions for a subset of actions and adapts decisions to observed states within the block. We show that the optimal value satisfies a Bayesian Bellman equation on the original state space. This representation reveals that predictive information creates value by recursively changing the local order of maximization and expectation in the Bellman expansion. We formalize this effect through a Bellman--Jensen analysis and quantify the effects of prediction horizon, error, action coverage, and within-block feedback. We propose block-feedback Bayesian offline learning with online adaptation (BF--BOLA), a model-based algorithm for learning near-optimal policies, and establish upper and lower sample-complexity bounds that match up to logarithmic factors in the stated regimes. The results show that predictions can reduce transition learning to uncovered rows and that learning a near-optimal policy can require fewer prediction samples than estimating the Bayesian value function.

0.17

Principled Nonlinear Concept Erasure and Steering

cs.LG cs.CL ·2026-10-23 ·Idan Simai, Shauli Ravfogel, Uri Shaham
Abstract

Learned representations often encode sensitive attributes, such as gender or race, that may be undesirable or unsafe to retain. Concept erasure aims to remove this information while preserving task-relevant content. Existing methods remain incomplete: linear methods provide guarantees only against linear probes, while nonlinear methods achieve stronger empirical erasure but lack formal guarantees and often sacrifice utility. In this work, we introduce a principled conditional autoencoder framework for nonlinear concept erasure. We follow the principle that erasure should remove only the sensitive attribute and nothing else, and we show that, under the stated assumptions, perfect reconstruction together with population-level independence is sufficient for perfect nonlinear guardedness, with preservation guarantees in the imperfect-reconstruction regime. Empirically, we attain a superior utility-erasure trade-off across multiple datasets, sensitive attributes, and language models, surpassing even the supervised variants of competing methods despite using no target-task labels, while being simpler to optimize and converging substantially faster. Beyond erasure, our framework yields concept conversion (representation steering) as a by-product: by changing only the target concept at inference time, it produces representations corresponding to different concept values - at no additional training cost - a capability no prior erasure method provides, and on which we likewise demonstrate strong empirical performance.

0.17

AOS-Bench: Can LLM Agents Manage Dynamic Workloads?

cs.AI ·2026-10-23 ·Erle Zhu, Genrui Zhang, Keyang Hu, Mingyu Su, Hongning Wang
Abstract

Coding agents are moving from handling one request at a time to serving as persistent assistants to which users hand a stream of work: a build, a data-processing job and a bug fix may all be pending on the same machine at once. Such an agent must decide when to take on new tasks, how to delegate and overlap them, and how to stay within the machine's CPU and memory. Existing benchmarks cannot measure this, because they evaluate each task in isolation. We introduce AOS-Bench, which evaluates an agent system as a workload manager: verified tasks arrive over time and must be delivered under shared CPU and memory limits, and the same tasks run one at a time serve as a per-system reference. Across five model--harness pairs, isolated task success does not carry over evenly. On heavy workloads every system loses 13 to 23 points. On medium workloads, systems span 11 points in isolated success but 26 points in stream success, and two systems 0.6 points apart in isolation deliver 54\% and 68\% of the same tasks. Execution-policy interventions trace much of this gap to task intake: serial dispatch leaves 73\% of heavy-workload tasks unclaimed. Yet more elaborate management does not necessarily help. A fixed concurrency of two comes within four points of the best agent-managed policy on both medium and heavy workloads, and under every policy median CPU use stays below 30\% of the provisioned cores while tasks wait. These results identify workload management as a distinct capability for agents, and AOS-Bench as a way to measure it.

0.17

Minimum Description Length Unsupervisedly Separates Law-Grade from Statistical-Grade Knowledge in a Self-Improving Science Agent - and Reveals 86% of Its Own Discoveries Are Statistical

cs.AI ·2026-10-23 ·Wenbin Mao, Pei Zou
Abstract · 6 pages, 2 figures. Companion paper: arXiv:2610.14843

AI systems that discover scientific relations now generate candidate "laws" at scale, but the bottleneck has shifted from generation to vetting: which candidates deserve belief? We show that a deployed self-improving science agent - one that has admitted 252 relations through held-out prediction courts - already computes the quantity that answers this question: the Minimum Description Length compression ratio used for model selection doubles, when read under discrete accounting, as an unsupervised epistemic grader. The compression-ranked top band of the agent's ledger recovers exactly the canonical laws among its discoveries (vis-viva across seven planets, Balmer/Rydberg) plus cross-measurement identities - enrichment $p=3.6\times10^{-12}$ under a hypergeometric test - with zero counterexamples and physically coherent band structure. The same lens retroactively reveals that 86% of the ledger compresses nothing ($\mathrm{cr}\approx 1$): the agent had been recording statistical regularities, including a learned predictor with held-out $R^2=0.541$ admitted through the prediction gate, under a narrative of compression as the unified value of knowledge - a narrative that was never true for them. We formalize the resulting dual-axis bookkeeping - prediction value ($R^2$) and compression value (cr) as nearly decoupled axes of knowledge - demonstrate its robustness to quantization, and register forward falsifiable predictions. The episode is, to our knowledge, the first measured instance of a discovery agent auditing its own knowledge ledger by an internal measure and finding itself over-claiming.

0.17

Accordion: Elastic-Depth Inference via Fixed-Prefix Layer Looping

cs.LG cs.CL ·2026-10-23 ·Yudong Zhang, Wei Ding, Kaili Zheng, Yuchen Ding, Zhenyu Zhang
Abstract

Large language models serve applications with diverse inference budgets, but training and maintaining separate models for each budget incurs substantial costs. Recent Looped Transformers enable flexible inference computation through parameter reuse, but maintaining strong performance across different inference budgets remains an open challenge. We introduce Accordion, a mixture-of-experts-based model that supports five inference budgets through fixed-prefix layer looping. Each budget follows a pre-defined execution path, while the minimum budget retains the unlooped model. Joint pretraining samples one budget per micro-batch and optimizes its next-token prediction loss without additional cross-budget consistency objectives. We further preserve these inference budgets through supervised fine-tuning and reinforcement learning, where parameter-group merging and budget-matched multi-teacher on-policy distillation integrate three domain experts into one model. After 3.6T tokens of pretraining, increasing inference depth improves Accordion-6B-A0.5B's five-category macro-average across 37 tasks by 3.01 percentage points, with the middle budget capturing 89% of this gain using half as many additional layer executions. The final model surpasses each domain expert on its corresponding task-and-budget average, while achieving strong instruction-following performance across all budgets and retaining depth-dependent gains in mathematical reasoning and code generation.

0.17

Beyond Data Schedule: Fast LLM Pretraining on Data Mixtures by Loss-Reweighting

cs.LG cs.AI ·2026-10-23 ·Jiazheng Li, Huaqing Zhang, Jing Xu, Tianle Cai, Jingzhao Zhang
Abstract

Large language models often acquire specialized capabilities through successive pretraining stages, but these gains can come at the cost of forgetting earlier knowledge. In this work, we investigate whether such trade-offs can be improved by changing how strongly each domain contributes to learning while keeping its data exposure fixed. We find that, in controlled knowledge--mathematics experiments, loss reweighting produces a single checkpoint that surpasses the best knowledge and reasoning scores attained along either stage trajectory. This reveals an independent control that data-mixture ratios alone do not capture. Building on this finding, we propose LARE (Loss-Adaptive Reweighting under fixed Exposure), which automatically adjusts domain weights from running loss statistics. Motivated by a theoretical analysis of gradient noise, its normalized inverse-square-root rule balances relative loss scales without changing the sampled mixture or requiring additional runs to search for weights. Across knowledge, mathematics, and code, \Method{} improves all five evaluated benchmarks for a 1.5B model trained on 240B tokens, averaging +7.84 pp over a matched direct mixture, including gains of +17.06 pp on GSM8K and +9.00 pp on MATH. These findings demonstrate that adjusting domain loss scales can improve joint capabilities without additional domain exposure.

0.16

DualGSL: Prior-Guided Graph Structure Learning with Node-Adaptive Prediction Fusion

cs.LG ·2026-10-23 ·Yuxia Sun, Yutszyuk Wong, Huihong Chen, Longhao Lv, Chang-Dong Wang +1 more
Abstract

Graph structure learning (GSL) adapts propagation graphs to prediction tasks, but the relative predictive value of feature-derived priors and learned relations can vary across nodes. We propose DualGSL, a prior-guided framework for transductive node classification that learns node-specific weights for combining predictions from separate prior and task-adaptive propagation paths. The prior branch augments a feature graph with virtual class nodes to incorporate training-label information, while the structural branch combines learned affinity graphs with the feature-graph prior. The two branches share encoder parameters while preserving distinct representations and classification logits. A node-adaptive gate fuses the branch logits based on their projected representations, element-wise absolute difference, and prediction confidence. A joint objective integrates fused and branch-specific classification losses with prior-branch class-anchor alignment, coordinating graph construction, representation learning, and prediction fusion. Experiments against seven baselines on the AMG18 malware attribution dataset and the widely used 20News text classification and USPS handwritten-digit recognition benchmarks show that DualGSL achieves the highest cross-dataset average accuracy and Macro-F1 among the evaluated methods. Ablation studies on AMG18 further demonstrate the benefit of combining prior and learned propagation through logit fusion.

0.16

From Parameter Rebasin to Functional Transfer Across Heterogeneous Transformers

cs.LG ·2026-10-23 ·Filippo Rinaldi, Aniello Panariello, Giacomo Salici, Simone Magistri, Angelo Porrello +2 more
Abstract

Adapting pretrained Transformers to downstream tasks produces task-specific updates that are expensive to reproduce for every model variant. Reusing these updates across different depths and widths is challenging because their parameters have incompatible shapes and do not share a canonical coordinate system. We introduce Ariadne, a training-free method that transfers task-specific behavior across heterogeneous Transformers without resizing the source model or transporting its parameters. Rather than treating the source task vector, i.e. the difference in parameter space between the fine-tuned and the base models, as the object to move, Ariadne characterizes fine-tuning by the activation changes it induces at matched blocks. First, it aligns these effects to a target base through local Procrustes maps. Then, it directly synthesizes a native target update by closed-form ridge regression. Across Transformer expansion, reduction and language-model transfer, Ariadne improves target-model performance while substantially reducing the time and memory required to construct an update relative to previous cross-architecture rebasin approaches. These results show that local functional effects provide an efficient alternative to parameter transport.

0.16

Putting Transformer Theories to the Test

cs.LG cs.CL ·2026-10-23 ·Awni Altabaa, Tobias Schnabel, Kiran Tomlinson
Abstract

Several formal theories have been developed to characterize the capabilities and limitations of the transformer architecture, each employing different abstractions that capture different features of the architecture. Due to the differences in their underlying assumptions and conceptual frameworks, these theories can disagree in their predictions about what tasks transformers find easy or hard. We map out seven theories of transformer capability (Star-free, $\mathsf{AC}^0$, $\mathsf{TC}^0$, C-RASP, MPC, BAPO, and Sensitivity) and translate their results into a common matrix of predictions across a suite of tasks covering different computational structures. We provide new theoretical results to complete this matrix. We then evaluate theoretical predictions through controlled experiments measuring in-distribution sample complexity, length generalization, and pretrained language model performance. We find that no single theory explains the observed pattern of successes and failures. When a theory does not capture the transformer behaviors relevant to a task, its predictions break down; for instance, we show that the $k$-hop induction task is predicted hard by C-RASP because it does not model content-dependent attention, but we train small transformers on this task that length-generalize. Similarly, theories assuming hard attention like $\mathsf{AC}^0$ and BAPO predict binary majority to be hard, for which we also find success in practice. Together, our results point to several specific features that a theory of transformers must capture: soft rather than hard attention, dynamic content-dependent retrieval, bounded computation at each position, an account of learnability and inductive biases, and a combination of upper and lower bounds.

0.15

Sparse Bayesian Neural Networks via Sticky Piecewise Deterministic Markov Processes

cs.LG stat.MLstat.CO ·2026-10-23 ·August Arnstad, Geir Olve Storvik, Filippo Pagani, Leiv Rønneberg
Abstract

There is growing evidence that the performance of a neural network depends largely on a small subset of its weights, which are typically identified by training a dense network and pruning it afterwards. Within a Bayesian framework, pruning can be directly incorporated through a spike-and-slab prior, providing a proper treatment of the uncertainties involved, both with respect to pruning and to parameter uncertainty. In Bayesian neural networks (BNNs), sampling by MCMC is notoriously hard. We demonstrate that modern Monte Carlo methods based on piecewise deterministic Markov processes (PDMPs) can be used to sample posteriors of BNNs asymptotically exact whenever numerically constructed rate bounds hold. Spike-and-slab priors can be incorporated through sticky versions of the PDMPs, setting weights exactly to zero during sampling. We examine the predictive performance and convergence of the samplers, and show that sparse BNNs sampled this way can be more robust to input noise than dense ones.

0.15

Let's Do Turbulence Right: In Space and Time!

cs.LG physics.flu-dyn ·2026-10-23 ·Maximilian Plattner, Fabian Paischer, Johannes Brandstetter
Abstract

The three-dimensional Navier-Stokes equations govern fluid turbulence, a chaotic and non-linear physical phenomenon characterized by complex eddies interacting across vast spatial and temporal scales. A numerically accurate modeling of this multi-scale energy cascade requires fully resolving both the fine spatial structures and the continuous temporal evolution of the flow field. In this paper, we leverage massive, time-resolved direct numerical simulations, which exceed 28 TB per run, to assess how numerical choices govern the resolved (3+1)-dimensional physics of turbulent flows. Building on this analysis, we study how neural networks represent the recorded turbulent motion. Our results highlight that spatial and temporal assessments can disagree for both numerical records and learned representations. We therefore call on the machine learning community to treat (3+1)-dimensional turbulence as a four-dimensional phenomenon.

0.15

All-in-One: A Composable Attack on LLM Transformations

cs.LG ·2026-10-23 ·Xiaohua Zhan, Kazuki Egashira, Robin Staab, Mark Vero, Martin Vechev
Abstract

Quantization, pruning, and model merging are popular transformations applied to deployment-ready LLMs to improve efficiency and downstream performance. Recent work has shown that these transformations can introduce a security threat: an adversary can train a model that passes safety checks in its pre-transformed state, only turning malicious once it is transformed for deployment. However, prior attacks target one specific transformation (e.g., quantization) and generally rely on transformation-specific techniques, making them difficult to combine in practice. In this work, we introduce a simple, composable attack framework that targets multiple transformations simultaneously and triggers when any one of them is applied by a user. Our key insight is that the effect of each transformation can be approximated as a perturbation to the model weights, allowing us to jointly train the model with a benign loss on the original model and a malicious loss on each simulated transformed model. These perturbations are simple enough to simulate at every training iteration across diverse transformations, making them practically combinable within a single training objective, enabling a unified attack. Through extensive experiments across attack scenarios, model families, and a wide range of deployment-time model transformations, we show that our attack achieves high success rates. This demonstrates that model transformations may be utilized to trigger significantly different behavior from the original model and reinforces the associated security concerns by demonstrating that a single attack can jointly target diverse transformations.

0.15

PULSES: multi-agent AI for autonomous excited-state model building and dynamics

cs.MA physics.chem-ph ·2026-10-23 ·Fangning Ren, Lechen Dong, Xu Chen, Weilin Zou, Bushra Alam +1 more
Abstract

Simulating molecular excited-state processes requires resolving electronic and nuclear dynamics, but the cost of direct quantum-chemical calculations limits accessible system sizes and timescales. Machine-learning models can extend these simulations, yet they often need to be tailored to specific molecular systems and scientific objectives. Building and deploying such models requires coordinated decisions across conformational sampling, quantum-chemical labelling, model training, validation and dynamics. Here we introduce Paired agents for Unified Learning and Simulation of Excited States (PULSES), a multi-agent system that automates this workflow within expert-approved boundaries. A human expert and planner define the scientific objective, methods and decision boundaries, while an orchestrator coordinates domain agents and artifact-specific scientific assurance during execution. We evaluate PULSES on 15 computational chemistry tasks and demonstrate end-to-end applications to perylene exciton Hamiltonians, solvent-dependent biphenyl relaxation, and excited-state proton transfer in 10-hydroxybenzo[h]quinoline. These results demonstrate traceable, expert-bounded construction and deployment of specialized excited-state models, while retaining scientific validation as a requirement rather than assuming that successful execution establishes correctness.

0.15

Conformance as a Distribution: Delegation Contracts, Error Budgets, and the Allocation of Evaluation Authority in Use-Based AI Regulation

cs.AI ·2026-10-23 ·John deVadoss
benchmark
Abstract

Use-based AI regulation assigns obligations by deployment risk but lacks an evidentiary object to carry them. A benchmark score is a property of a model; conformance of a deployment is a failure rate under a declared distribution of contexts, bounded by a scope and subject to expiry. We take the delegation contract of earlier work (frozen configuration, scoped claim, executable revision rule) and study it under strategic play between a deployer, an auditor, and an enforcement point. We ask whether a declared context mixture can be made truthful; whether an audit can be unbiased against a system that detects evaluation traffic; whether a certification error budget can sustain a long-lived deployment; whether an aggregate audit can support a per-subgroup guarantee; and whether an incident metric that counts system failures but not fallback costs ranks arrangements correctly. For each we give a mechanism and a proposition: a proper-scoring penalty on the declared mixture bounds the excess failure of a best-responding deployer by $O(λ^{-1/2})$ in the penalty weight; a commit-reveal audit drawn from live traffic is unbiased by construction and grades the evaluator with the same sample; a halving error budget drives certification coverage below $(1-p)^n$ while $α$-investing sustains it under a marginal false-certification guarantee; a per-subgroup guarantee costs audit size scaling as $1/m$ in subgroup mass; and any failure-only incident metric prefers withdrawal at every fallback cost, even when total loss ranks the arrangements oppositely. Five preregistered synthetic studies test each claim; no language model or production system is evaluated. Enforcement sits at distribution, capability-access, monetization and risk-transfer choke points, expiry of the conformance record is a standard of care, and conformance-as-a-rate is set against the EU AI Act's static conformity assessment.

0.14

RSI-Opt: Recursive Self-Improvement Optimizer Agents For Model Training

cs.AI ·2026-10-23 ·Binghang Lu, Yige Sun, Guang Lin
Abstract

Conventional neural-network training typically follows a predefined optimization recipe, motivating a LLM agent that monitors training dynamics and adjusts optimizer settings as training progresses, but a fixed agent policy does not accumulate experience across runs. We introduce RSI-Opt, a plug-and-play Recursive Self-Improvement Optimizer Agents which evolves the knowledge that guides this process: a prompt, a playbook of procedures, a library of tested optimizer code, and a decision cadence. A meta-agent rewrites this policy using evidence from earlier runs, and an archive retains evaluated policies for further evolution. Across 5 generations on three training tasks, the meta-agents develop reusable procedures and code and conduct their own CPU proxy experiments. RSI-Opt substantially improves optimizer-agent performance, with representative raw-space summary $z$-scores increasing from $+10.48$ to $+14.33$ on the evolution suite with Opus 5.5 and from $-0.36$ to $+3.11$ on held-out tasks with GPT-5.5 through cross-model policy transfer. On the Burgers PDE, it reduces mean relative $L_2$ error by approximately 2.7-fold with Opus 5.5 and 8.8-fold with GPT-6.1 over the fixed-prompt agent baseline, although PDE transfer varies across LLMs and tasks. Controlled comparisons show that the evolved policy combines reusable optimization knowledge with adaptation to each task: initial decisions capture most observed gains on seen tasks, while continued adaptation improves outcomes on unseen tasks. These findings demonstrate how procedures and executable tools accumulated across training runs can improve optimizer-agent performance and support transfer to scientific machine learning.

0.14

Acquisition for Spherical-Linear Bayesian Optimization on the Cube

cs.LG stat.MLstat.CO ·2026-10-23 ·Nathan Wycoff
Abstract

Recent work demonstrated that executing Bayesian Optimization (BO) using simply a linear surrogate with a clever transformation is enough to approximate the performance of the standard Gaussian-Process (GP) based version on high-dimensional test functions. When the decision region is a sphere, this strategy admits rapid exact acquisition. However, it is the unit cube we encounter most often in the BO literature, and in this setting a standard approach, namely a multi-start quasi-Newton method, is employed for solving the acquisition subproblem. In this article, we will discuss a series of results, both positive and negative, regarding the complexity of global resolution of various acquisition subproblems over the unit cube. On the basis of this discussion, we will propose a new acquisition function which is globally optimizable using significantly less execution time than the incumbent approach while maintaining or improving performance. The resulting method is competitive with GP-based BO in terms of accuracy per evaluation, and is superior in terms of accuracy per walltime on black-boxes of low to moderate computational intensity.

0.14

LeafSkill: Efficient Leaf-Level Action Grounding for Recursive LLM Agents

cs.AI ·2026-10-23 ·Shen Liu, Ze Zhou, Zhenguo Xu, Shaopu Wang, Yike Gao +1 more
Abstract

Recursive large language model (LLM) agents solve long-horizon tasks through decomposition, execution, and environmental feedback. Yet after producing an atomic subtask, they often invoke the same large model again to translate a familiar instruction into an executable action. We observe that high-level planning is task-specific, whereas leaf-level action grounding recurs across tasks and can be checked against environment-admitted actions. We introduce LeafSkill, an external reuse layer that retrieves semantically related grounding experience, binds action templates to current objects and admissible actions, and uses a lightweight 1.7B selector to choose a bound action or defer the unchanged subtask to the original agent. Skills remain outside the model parameters, allowing a frozen selector to use post-training additions. Under lexical and syntactic paraphrases, exact matching retrieves none of the reference actions, while semantic retrieval preserves 96.0% and 97.0% candidate coverage; the frozen selector adopts 97.34% of unseen skill types. On an independent confirmation set of 200 Robotouille tasks, LeafSkill completes 137 tasks versus 121 for ReCAP (68.5% vs. 60.5%, exact McNemar p=.0259). On the 106 tasks completed by both, it reduces Qwen3.5-27B tokens by 14.42% and calls by 14.08%. A separate held-out repeat yields a consistent 9.0-point success gain and 18.98% token reduction. Reusable leaf grounding can thus improve task completion while reducing host-model inference without replacing high-level planning.

0.14

Empirical concentration for chains with infinite memory

cs.LG stat.ML ·2026-10-23 ·Pierre Alquier, Paul Doukhan
Abstract

Concentration inequalities for time series usually involve unknown quantities measuring dependence. This prevents their direct use as confidence bounds, or as criteria for learning algorithms. Recent works proved empirical concentration inequalities for stationary Markov chains, where the unknown quantities are estimated. We extend this approach beyond the Markov case, by considering stationary chains with infinite memory. We first establish a bounded-differences inequality on bounded Polish metric spaces under a summable Wasserstein contraction condition. We then estimate its dependence coefficients from a single trajectory, using either transition counts when the state space is finite, or derivatives of kernel conditional-density estimates for a continuous state space. We give general error bounds and illustrate their dependence on the memory tail. Finally, a fixed multiplicative safety factor turns these estimators into empirical concentration bounds. While some regularity assumptions are required to make the bounds empirical, our procedure remains fully adaptive in the sense that it does not require the knowledge of the constants involved in these assumptions. The empirical confidence bounds derived from our results are asymptotically as tight as the theoretical ones.

0.14

Beyond Post-Hoc Rewards: Privileged Reasoning Exploration for Multimodal Embeddings

cs.AI cs.CL ·2026-10-23 ·Tianyu Yang, Yijun Tian, Lars Benedikt Kaesberg, Jan Philip Wahle, Bela Gipp +2 more
benchmark
Abstract

Advances in Multimodal Large Language Models (MLLMs) have driven substantial progress in multimodal embeddings. Recent methods exploit generative capabilities by augmenting embedded inputs with self-generated reasoning to produce more informative embeddings. However, conventional retrieval supervision for training embedding models specifies pairwise relevance relations without explicitly guiding the reasoning used for input augmentation. Such reasoning is neither directly observed nor uniquely determined by the input, allowing multiple plausible traces to introduce semantic cues of varying retrieval utility. Existing approaches largely optimize reasoning conditioned on the input alone, limiting the exploration of retrieval-relevant traces. We introduce PILOT, a framework that models reasoning as a latent variable and instantiates its prior and posterior distributions as distinct reasoning modes of a shared MLLM. The prior conditions only on the input, whereas the posterior uses additional training-time relational evidence to guide reasoning exploration. To translate the posterior's broader exploration into inference-time improvements, we derive an importance-weighted variational objective and instantiate it as a prior-posterior co-evolution framework that improves both modes under a shared inference-compatible reward while selectively transferring posterior trajectories to the prior according to their utility and prior reachability. On the 78-task MMEB-V2 benchmark, PILOT outperforms the primary baseline across all 12 task groups at both 2B and 7B parameters, improving overall scores by 3.2 and 3.3 points, respectively, with the largest gains in visual-document retrieval (+6.1 and +6.4 points). Our results show that relational supervision is most valuable not as a post-hoc reward but as a signal for shaping how reasoning is explored in training reasoning-enhanced embedding models.

0.13

DBCM: Dual-Branch Calibration Module for Test-Time Adaptation in Time Series Forecasting

cs.LG stat.ML ·2026-10-23 ·Jaebyeong Jeon, Kunwoo Kang, Sang Yup Lee, Kibok Lee
Abstract

Non-stationarity induces test-time distribution shifts in time series forecasting (TSF), resulting in forecast residuals between predictions and the corresponding ground truth. Recent works on test-time adaptation for TSF address this by introducing lightweight calibration modules around frozen forecasters, updated using observations that become incrementally available at test time. Existing approaches typically employ channel-wise adapters, implicitly assuming that forecast residuals are inherently channel-specific. However, we observe that forecast residuals often exhibit substantial shared structure across channels, suggesting that fully independent channel-wise parameterization may be redundant. Motivated by this observation, we propose the Dual-Branch Calibration Module (DBCM), which shares adaptation capacity across channels through a shared adapter while retaining limited channel-specific flexibility through a low-rank channel-wise adapter, further enhanced by channel mixing and nonlinear gated modulation. Extensive experiments show that DBCM consistently outperforms prior methods across diverse datasets, horizons, and backbones.

0.13

A Foundation Model for Policy In-Context Longitudinal Causal Effect Estimation

cs.AI ·2026-10-23 ·Wenxin Chen, Kyra Gan, Mingyu Lu, Zilin Jing, Fei Wang +3 more
Abstract

Longitudinal causal effect estimation is complicated by time-varying confounding and treatment-confounder feedback loops. Existing methods generally require data-specific nuisance-model fitting, and many are tailored to prespecified treatment sequences or prediction horizons, leading to time-consuming computation at deployment. We introduce *Policy In-Context Longitudinal causal Effect estimation with Prior-data Fitted Networks* (**PICLE-PFN**), the first foundation model for longitudinal causal effect estimation that supports *dynamic* treatment policies. PICLE-PFN represents target policies through in-context history-decision demonstrations, providing a unified interface for a broad class of treatment policies. Pretrained over synthetic data sampled from longitudinal structural causal models, PICLE-PFN amortizes causal effect estimation across datasets, policies, and horizons. Given a new observational dataset and target policy, it directly predicts conditional average potential outcomes (CAPOs) in one forward pass without task-specific training or hyperparameter tuning. Across synthetic and semi-synthetic longitudinal benchmarks, PICLE-PFN achieves competitive RMSE and substantially improved uncertainty calibration with a fraction of deployment runtime, compared to state-of-the-art conventional estimators trained and tuned separately for each dataset.

0.13

Estimation of Propagated Risk in LLM-based Multi-Agent Systems

cs.AI cs.MA ·2026-10-23 ·Shreyan Ganguly, Souradeep Chattopadhyay, Qisai Liu, Soumik Sarkar
Abstract

Large Language Model (LLM) based multi-agent systems introduce a distinctive reliability challenge in which actions taken by upstream agents can affect the correctness of downstream agents. While prior work has investigated error propagation and developed methods for estimating uncertainty inherited from upstream agents, existing approaches do not explicitly model how realized sender-receiver interactions contribute to downstream error risk under different sender correctness states. In this work, we study agent-level error-risk estimation from multi-agent execution traces. We propose Interaction-Effect Propagation (IE-Prop), an interaction-conditioned model that learns separate effects for correct and incorrect upstream agents. For each sender-receiver interaction, the model combines these conditional effects using the sender's estimated error risk, and recursively propagates the resulting risk estimates through the execution graph. We evaluate our approach across three multi-agent architectures and three benchmarks. Across all evaluated architectures, our approach raises the final-node AUROC over the established baseline by approximately 0.05 points on average. Furthermore, tracking risk across multiple agent steps contributes a distinct +0.03 AUROC gain in sequential systems. These results show that realized inter-agent interactions provide predictive information beyond receiver-local evidence, while the value of multi-step risk tracking depends on the system's information-flow structure.

0.13

NgramRL: Accelerating RL Post-Training for Structured Generation via Rollout-Aware N-gram Speculative Decoding

cs.AI ·2026-10-23 ·Nilesh Prasad Pandey, Jangseon Park, Jason Kong, Onat Gungor, Tajana Rosing
Abstract

Reinforcement learning (RL) post-training improves the reasoning capabilities of large language models, but its cost is dominated by autoregressive rollout generation, which can exceed 70% of training time. Existing speculative-decoding methods are poorly suited to this setting: model-based drafters consume scarce GPU memory and may become misaligned as the policy evolves, while existing model-free methods do not explicitly exploit repetition across training epochs. We observe that recent rollouts of the same prompt provide a high-quality, self-updating draft source. Based on this insight, we propose NgramRL, a per-prompt n-gram drafter constructed from bounded recent rollout history. NgramRL requires no auxiliary model, drafter training, or persistent GPU-resident drafter state, while target-model verification preserves the rollout distribution. To hide CPU cache lookup and scheduling, we further introduce an adaptive fan-out prefetcher that predicts likely verification outcomes (j, b), consisting of the accepted draft length j and target-emitted continuation token b. A causal budget allocator concentrates prefetches on high-probability outcomes, while uncovered outcomes fall back to sequential lookup without changing committed tokens. Across two model families and three reasoning and structured-generation workloads, NgramRL commits 3.60-7.77 tokens per target-model forward in the warm-cache regime, up to 1.99x more than SuffixDecoding and 7.0x more than prompt-lookup decoding. On Spider, the prefetcher covers 96.6% of Qwen3-8B outcomes at a candidate budget of 16 and improves measured rollout throughput by 1.38x over synchronous NgramRL, approaching its 1.40x overlap ceiling.

0.12

U-NeXt: General-purpose event detection for neonatal EEG

cs.LG eess.SPphysics.med-ph ·2026-10-23 ·Robert Hogan, Sean R. Mathieson, Aurel Luca, Soraia Ventura, Sean Griffin +2 more
Abstract

We developed U-NeXt, a general framework for neonatal electroencephalography (EEG) event detection based on semantic segmentation. Its convolutional encoder-decoder directly segments individual channels using long input segments to preserve temporal context. A shared architecture and training framework were developed using low-voltage interval, artefact, and seizure datasets containing 51, 98, and 278 neonates, respectively, comprising over 11,000 hours of EEG and 82,000 annotated events. Performance was compared with task-specific ConvNeXt classifiers. For seizure detection, we also evaluated U-NeXt+, trained on an expanded dataset. Both U-NeXt variants and the ConvNeXt baseline were evaluated on two independent held-out seizure datasets, Helsinki and Cork, comprising 130 neonates, with recordings annotated by three experts. U-NeXt improved detection across all three event types, with particular gains in sensitivity for short events. U-NeXt+ achieved Matthews correlation coefficients of 0.862 and 0.891 on Helsinki and Cork, respectively, and improved hourly seizure burden correlations by 30-40% relative to ConvNeXt. U-NeXt+ met the criterion for expert-level performance on both datasets (p < 0.001). In separate computational benchmarks, U-NeXt reduced batch inference time relative to ConvNeXt by 254-fold on a central processing unit and 200-fold on a graphics processing unit. A common segmentation framework improves detection across diverse neonatal EEG event types while substantially reducing inference time. Accurate seizure burden estimates and performance comparable to expert reviewers support its potential use in continuous monitoring and clinical decision support.

0.12

MLIPDiff: Guiding Denoising Diffusion Models for Structure-Based Drug Design via Machine-Learned Interatomic Potentials

cs.LG ·2026-10-23 ·Yiyan Liao, Amin Tavakoli, Yu Zhang, Anima Anandkumar
Abstract

Diffusion models have become a dominant approach to structure-based drug design. However, because they are trained primarily to fit reference structures, they often place ligands in physically implausible poses that clash with the protein. We introduce MLIPDiff, an inference-time guidance method that can be applied to pretrained diffusion models for ligand generation. The method uses physics-informed guidance from UMA (Universal Models for Atoms), a machine-learned interatomic potential (MLIP). The guidance combines four weighted energy terms (protein-ligand interaction, ligand strain, steric clash, and pocket anchoring) and uses the original model checkpoint without retraining. On CrossDocked2020, MLIPDiff completely eliminates protein-ligand collisions while preserving the ligand's local geometry, improving bond-length distributions and physical plausibility relative to the base model. Compared to models such as NucleusDiff that are specifically designed to reduce atomic clashes, MLIPDiff achieves clash-free structures without compromising local geometry. It maintains the base model's PoseBusters pass rate and attains the best median Vina Score, Vina Min, and Vina Dock, surpassing both the base model and NucleusDiff, with a median of 59.6\% of ligands per pocket binding more strongly than the reference. MLIPDiff thus addresses a limitation of prior methods by generating physically plausible structures with favorable predicted binding affinity using a single frozen model and no additional training.

0.12

LSTM-Style Looped Transformers with Shared KV Caches

cs.LG ·2026-10-23 ·Andreas Kirsch
Abstract · 12 pages + 14 pages appendix

Recurrent transformers reuse weights to perform deep computation with fewer parameters. We study an LSTM-style transformer that also shares one key--value cache across its recurrent core. At width 1024, it uses 40.4% fewer parameters than an untied control at nearly matched forward FLOPs, and reduces persistent cache sets from fifteen to four. Parallel training supplies cache entries from a standard-attention pass, whereas continuation reads entries produced recursively using shared attention. We address this mismatch by aligning the two training passes' cache entries with a consistency loss during the final 10% of a fixed training-token budget. Across three model widths on FineWeb, this removes 81--89% of the continuation-loss penalty relative to full-cache evaluation. The improvement replicates on disjoint documents for two training seeds. Across four widths and 1.05--4.19B training tokens, the loss cost of weight tying is smallest at lower token budgets, suggesting an opportunity to use recurrent computation when data and model storage are limited.

0.12

EVOLVE: Uncertainty-Aware Multivariate Time Series Anomaly Prediction with Adaptive Prototype Memories

cs.LG ·2026-10-23 ·Jianxin Jin, Kai Zhao, Shiyan Hu, Yang Shu, Peng Chen +6 more
benchmark
Abstract · 14 pages. An earlier version of this work was submitted to KDD 2026 on February 9, 2026

Unsupervised multivariate time series anomaly prediction (AP) aims to estimate future anomaly risks from current observations, providing early-warning signals for proactive intervention and maintenance in complex systems. Existing unsupervised AP methods follow a perturbation-based paradigm, where predefined or random perturbations are injected into historical windows to simulate abnormal future evolution. However, these methods share perturbations across different variables and contexts, and learn deterministic history-to-future evolution. In reality, anomaly evolution is variable- and context-dependent: the same perturbation may be amplified into abnormal future dynamics in one variable or context, while being absorbed as normal fluctuation in another. In addition, future evolution is not uniquely determined by a historical window, since similar histories may lead to multiple plausible trajectories, such as gradual degradation or abrupt changes. To address these problems, we propose EVOLVE, a unified framework that models unsupervised AP via variable- and context-conditioned perturbations and uncertainty-aware future reasoning. EVOLVE constructs context-specific anomaly perturbations through an adaptive anomaly prototype memory, which can dynamically update and retrieve relevant perturbations at the variable level. EVOLVE models uncertainty by sampling multiple latent historical states and decoding through complementary smooth and deviation-amplifying evolution hypotheses, and the disagreement among decoded future trajectories is used to estimate future anomaly risk. Extensive experiments on six real-world benchmark datasets demonstrate that EVOLVE consistently outperforms state-of-the-art anomaly prediction baselines.

0.12

ActionRAG: Evidence-Refining Graph Diffusion for Long-Term Memory in LLM Agents

cs.AI cs.CL ·2026-10-23 ·Navyansh Mahla, Sashank Vandrangi
Abstract

LLM agents need long-term memory that can recover precise and mutually consistent evidence from growing interaction histories, particularly when answers depend on relations among facts rather than isolated semantic matches. Current memory pipelines remain vulnerable at two stages: indexing errors can fragment repeated entities or merge distinct ones, and online retrieval often treats initial similarity scores as final even when the surrounding graph provides evidence for revising them. We introduce ActionRAG, a graph memory architecture that couples cautious indexing with query-time graph retroevaluation. Its indexing pipeline reconciles type labels and applies tiered entity resolution. Given a query, dual-channel entity and triple retrieval forms an initial evidence set and induces a bounded local graph. Query-conditioned receiver-normalized diffusion refines triple scores using local path structure. Seed protection, hub-penalized flux, and triple materialization preserve direct evidence while recovering missed connections. We evaluate ActionRAG on MemoryAgentBench (MAB), which tests accurate retrieval, conflict resolution, long-range understanding, and test-time learning. Across its eight deterministic-scoring columns, ActionRAG achieves 55.1 average accuracy, improving over the strongest retrieval-based memory baseline by 5.4 points. Under MAB's query-time cost accounting, mean synthesis cost stays below \$0.001 per query on the four reported cost columns, and the sum of separately measured retrieval-side p99 components is 440.7 ms on our consumer Apple Silicon setup.

0.11

Score-Preserving Continual Lasso under Task Shift

cs.LG stat.MLmath.ST ·2026-10-23 ·Jingyi Guo, Runze Li, Lingzhou Xue
Abstract

Continual learning (CL) aims to learn from sequential tasks while mitigating catastrophic forgetting when historical data are no longer accessible. We study continual learning for high-dimensional sparse linear regression, where $\ell_1$ regularization creates a difficulty absent from continual ridge regression: that recursively reusing an $\ell_1$-penalized estimator can propagate earlier shrinkage and accumulate regularization bias across tasks. We propose score-preserving continual Lasso (SP-CL), which separates sparse estimation from state transmission by preserving the historical squared-loss score. Under squared loss, its cumulative estimator exactly recovers a full-history Lasso minimizer. When task coefficients coincide, it inherits the full-history sparse rate. For related task-specific coefficients, SP-CL uses a regularized current-task correction that leaves the transmitted state unchanged and attains the offline source-assisted rate under an oracle related-task condition. We also derive Overall/BWT/FWT decompositions that distinguish historical information loss, task conflict, and current-task specialization, and validate the resulting regimes in simulation.

0.10

DELA: Learning Controllable Edits for Diffusion Language Models

cs.AI ·2026-10-23 ·Xinsong Feng, Chenan Wang, Leshu Tang, Johnny Liu, Dong Li +2 more
Abstract

Diffusion language models (DLMs) support bidirectional denoising and iterative refinement, but generation remains largely controlled by token-level masking schedules. This leaves draft-level refinement implicit and limits explicit control over when to insert missing content, delete redundant spans, revisit uncertain regions, or stop editing. We introduce Diffusion Editing Language Adapter (DELA), a plug-in framework that enables controllable draft refinement with a frozen DLM. DELA operates on clean, variable-length drafts using four explicit editing actions: Insert, Delete, Remask, and Stop. The controller determines when and where to edit the draft, while the frozen DLM fills masked regions through denoising, separating high-level edit control from token-level generation. A naive single-phase editor would be computationally expensive because arbitrary interior edits can disrupt prefix stability and limit reliable KV-cache reuse. To address this, DELA uses a two-phase decoding procedure: a fast, append-only drafting phase that supports KV-cache reuse, followed by a global refinement phase that enables targeted non-monotonic edits to the full draft. We train the controller using a staged curriculum that begins with learning local edit priors from synthetic recovery tasks and mined actions that improve draft quality. Subsequent stages refine edit selection through rollout-based outcome preferences and alignment with downstream task feedback. Experiments on mathematical reasoning and code generation show that DELA achieves gains of up to 5.5 percentage points over direct diffusion decoding while keeping the backbone frozen. DELA thus provides a modular approach to explicit draft-level control in diffusion language models.

0.10

Conditional Flow Models for Offline Maximum Entropy Reinforcement Learning

cs.LG stat.ML ·2026-10-23 ·Hao Cheng, Yuling Jiao, Lican Kang, Jerry Zhijian Yang, Chen Zhong
Abstract · 68 pages, including appendices

Offline maximum entropy reinforcement learning seeks to estimate the optimal soft action-value function \(Q^*\) from a fixed dataset without further interaction with the environment. Although \(Q^*\) is characterized by the soft optimal Bellman equation, the corresponding Bellman operator cannot be evaluated directly because both the reward distribution and the transition kernel are unknown. Consequently, estimating \(Q^*\) requires the joint treatment of environment modeling and value function learning. We propose a framework that addresses these tasks separately. Conditional diffusion models based on probability flow ODEs estimate the conditional reward distribution and transition kernel. The resulting distributional estimators define a data driven approximation of the soft optimal Bellman operator, whose empirical Bellman residual is minimized over a class of deep neural networks to estimate \(Q^*\). We establish nonasymptotic guarantees for the proposed method. For operator learning, we derive total variation error bounds for the conditional flow estimators that account explicitly for score estimation, numerical discretization, and terminal smoothing. For soft value learning, we first use the soft Bellman equation to show that $Q^*$ inherits Hölder regularity from the conditional reward and transition densities. For any Hölder index $β$, the value learning stage achieves an excess Bellman residual risk bound of order $\widetilde{\mathcal O}\big(n^{-\frac{2β}{d+2β}}\big),$ where $d$ is the total dimension of the state and action spaces. In contrast to many existing theoretical analyses of deep reinforcement learning, our results do not require Bellman completeness. Numerical experiments support the theoretical findings and demonstrate the effectiveness of the proposed method.

0.10

Understanding Communication Failures in Multi-Agent LLM Systems

cs.AI ·2026-10-23 ·Priya Pitre, Jun Hou, Yi Fang, Naren Ramakrishnan, Xuan Wang
Abstract

Multi-agent LLM systems aim to improve reasoning through debate, evidence sharing, and mutual correction, but interaction can also introduce failures that final accuracy alone cannot explain. Agents may converge too early, ignore dissent, suppress correct minority answers, or preserve a wrong framing across rounds. We present a corpus-grounded taxonomy of communication failures, built from 3,211 failed multi-round trajectories across medical QA, coding, finance, multi-hop reasoning, ethics, and embodied text tasks. Using this taxonomy, we analyze failure distributions across datasets and system designs, round-level error dynamics, recovery hardness, majority transitions, minority-signal suppression, and self-reported confidence. We find that failures are dominated by recovery/correction and consensus/decision problems rather than isolated first-round mistakes, with many emerging after agents first observe peer responses. Based on these findings, we recommend stronger disagreement handling, evidence verification, minority-signal preservation, and recovery checks before finalization.

0.09

Recommender regret for improving regularised RL estimation with bandits and MCTS

cs.LG ·2026-10-23 ·Danilo Vucetic, Alexandre Larouche, Mohammad Pedramfar, Audrey Durand, Gauthier Gidel
Abstract · 10 pages, plus appendices for proofs and additional information

Regularised reinforcement learning (RL) algorithms rely on estimators, like Monte Carlo tree search or one-step look-ahead, to provide policy, value, or Q-value estimates in training and inference. However, existing bandit-based estimation frameworks evaluate sampling efficiency, not the quality of estimates. We address this by formalising estimate generation as a recommendation process, treating bandit-based sampling as an improvement operator. To evaluate the quality of a recommendation (estimate), we define regulariser-aware recommender regrets, analysing their lower bounds, and proving that simple algorithms like uniform sampling can achieve these bounds. Finally, we unify bandit- and search-based target improvement under recommender MCTS. Across synthetic trees, arcade games, and scientific discovery tasks, recommenders often yield faster target convergence and superior downstream performance. Our experiments characterise for the first time the potential of regularised MCTS in scientific applications.

0.08

Sharpness-aware Orthogonalized Momentum Algorithms

cs.LG math.OC ·2026-10-23 ·Geetika, Shivani Singh Chauhan, Gauranshi Gupta, Anant Jain, Kaushik Kalyanaraman +1 more
Abstract

Adversarial sharpness penalty per iteration of a base optimizer, known as sharpness-aware minimization (SAM), is an accepted technique to improve generalization in deep learning. Recently, Muon has emerged as a dominant optimization algorithm for training large language models (LLMs). It achieves lower sample complexity for heavy-tailed gradient updates - a phenomenon common in LLM training. However, sharpness penalties with Muon updates are yet largely unexplored in the literature. Moreover, the behavior of SAM under heavy-tailed noise is unknown both theoretically and empirically. A potential reason for this gap is the additional cost of extra gradient computation per iteration for penalty perturbations in SAM, which becomes prohibitive for LLMs. This paper introduces Sharpness-aware Orthogonalized Momentum Algorithm (SOMA) to combine a computationally efficient and theoretically grounded perturbation scheme with the Muon algorithm. The key idea is to use stale (purely) stochastic gradient noise for adversarial perturbations. Theoretically, we show that SOMA's sample complexity is invariant to the perturbation direction. We provide explorations and insights into SOMA's robustness to heavy-tailed gradient noise. We extensively explore SOMA's efficacy in (a) mitigating quality degradation due to post-training quantization, and (b) catastrophic forgetting during fine-tuning of LLMs.

0.08

PEAR: Adaptive Advantage Reshaping for Perception and Exploration in Large Vision-Language Models

cs.AI ·2026-10-23 ·Junxin Wang, Yingfeng Luo, Xiangchen Song, Weiqiao Shan, Runsong Zhao +10 more
Abstract

Effective multimodal reasoning requires a model to reliably utilize visual information while sufficiently exploring diverse reasoning paths. To achieve this goal, prior work combines visual dependency and token entropy to reshape sequence-level advantages, but scaling the contribution of token entropy by visual dependency can leave high-entropy tokens with low visual dependency underweighted, despite their important role in reasoning exploration. To address this limitation, we propose PEAR, an adaptive advantage reshaping method for improving perception and exploration in multimodal reasoning. PEAR retains visual dependency as the base signal and adds an entropy-based exploration gain only at high-entropy positions, with lower visual dependency leading to a larger additional gain. The fused signal is then mapped to positive token-level weights that reshape the sequence-level advantage. Experiments on Qwen2.5-VL-3B/7B show that PEAR achieves the highest average performance among evaluated baselines across multimodal reasoning benchmarks, alongside leading results on textual reasoning and general multimodal benchmarks.

0.08

BenchBench-Cloning: Evaluating Molecular Cloning Design and Protocol Development

cs.AI ·2026-10-23 ·Aditya Sivakumar, Peter Ciaccia, Hassaan Qadir, Ian Hsu, Ashu Singhal
benchmark
Abstract

Molecular cloning is one of the most routine wet-lab workflows, yet existing evaluations of language models in biology mostly test knowledge recall or isolated steps rather than end-to-end experimental planning. We introduce BenchBench-Cloning, a benchmark for large language models comprising 65 molecular cloning tasks grounded in sequence records, laboratory notebooks, and published experimental studies. Planning a molecular cloning experiment requires defining a set of constructs needed for a scientific objective and a construction route to produce them. BenchBench-Cloning contains 50 protocol-development tasks, in which agents produce an executable protocol for a specified construct, and 15 construct-design tasks, in which agents determine the constructs a scientific objective calls for. Protocol-development tasks are evaluated by execution in a cloning simulator; construct-design tasks are evaluated with task-specific deterministic checks and, where needed, rubric-based LLM-as-a-judge evaluation. Across eight systems with repeated attempts per task, GPT-6 Astra scores highest at 30.6%, with the others between 8.6% and 22.2%, and the benchmark remains unsaturated under best-of-ten sampling. As integrating individual cloning decisions into complete, valid plans remains challenging even for frontier models, we expect BenchBench-Cloning to serve as a useful benchmark for tracking progress in experimental planning.

0.08

HyScale: Scaling Long-Context Training for Learned Sparse Attention

cs.AI cs.DC ·2026-10-23 ·Yangshen Deng, Xinjing Huang, Yulong Wang, Yunzhuo Liu, Jun Hu +6 more
Abstract

We present HyScale, a long-context training system for learned sparse attention. Learned sparse attention is a promising architecture for million-token contexts, widely adopted in production since DeepSeek-V3.2. Selection, however, grows into the training bottleneck. For each query, it scores every previous token whereas sparse attention reads only $k$; worse, current systems materialize a full score table and repeat selection during recomputation. Our key insight is that the output of selection is not an activation, but rather a set of indices. Building on it, HyScale introduces FlashTopK and selection replay. FlashTopK finds each query's exact top-$k$ on the fly without a score table, while selection replay reuses forward-pass selection in recomputation. Together with holistic system optimizations tackling end-to-end challenges, HyScale speeds up GLM-5.1 training on 1M-token sequences by $2.07\times$ over Megatron-LM on 512 H20 GPUs. It also speeds up the CSA layers of DeepSeek-V4-Flash and the QSA layers of Qwen3.8-Flash-Next by $2.01\times$ and $1.57\times$ at 1M tokens on 32 H800 GPUs.

0.07

DT-JEPA: Learning to Predict Future Representations for Decision-Making

cs.LG ·2026-10-23 ·Changxu Ni, Yiyang Ni, Jun Li, Feiran You, Hongyang Du +2 more
theory
Abstract

Decision Transformer (DT) has demonstrated strong performance in various offline reinforcement learning (RL) tasks by conditioning action generation on expected future returns and historical trajectories. However, treating expected future returns as the condition provides an incomplete description of the future, as trajectories with different state transitions and behaviors may share similar returns. Moreover, expected future returns inherently depend on task-specific reward functions, which makes policy learning reliant on reward-labeled datasets and restricts its scalability to tasks with similar but unseen rewards. Recently, Joint Embedding Predictive Architectures (JEPA) have emerged as a promising paradigm of self-supervised learning-based world models, which learn predictive representations of future dynamics in latent space rather than reconstructing future states in the original space. Inspired by JEPA, we propose DT-JEPA, an offline pretraining and online fine-tuning framework that learns general knowledge from reward-free offline trajectories and achieves excellent adaptation to unseen tasks. Instead of predicting a single deterministic future embedding, we design a novel probabilistic predictor to model the conditional distribution of future representations in the latent space, thereby capturing both predictive information and uncertainty about future dynamics. Theoretically, we show that 1) the JEPA objective seeks to maximize a variational lower bound on the mutual information between historical representations and future trajectories; and 2) predictive future-conditioned policy learning admits a finite-sample performance bound with an O(N-1/4) estimation term. Experiments on D4RL benchmarks demonstrate that DT-JEPA achieves efficient adaptation across varying offline RL tasks, outperforming traditional DT-based and other offline RL methods.

0.07

Head Dimension Matters in Language Models for Long-Context Retrieval

cs.LG cs.AIcs.CL ·2026-10-23 ·Seiya Tokui, Shoichiro Yamaguchi, Sosuke Kobayashi
Abstract

Multi-head attention retrieves and combines context representations through query--key interactions and value aggregation, with head dimension constraining the per-head representation space. In long-context processing, it remains unclear whether allocating a fixed parameter budget to fewer, wider heads improves information retrieval. In this paper, we investigate how head dimension affects the performance of long-context retrieval tasks. Our empirical studies reveal that pretraining loss and short-context non-retrieval task performance are not affected much by head dimension, while long-context retrieval performance improves with wider heads at fixed model sizes. We find that models with wider heads have a higher proportion of retrieval heads, associating more parameters with retrieval. Our ablation studies also show that the retrieval performance improves with wider heads even when the projection weights in attention modules are frozen during long-context continual pretraining, suggesting that wider heads not only benefit updates in attention modules but lead to more effective adaptation of components such as embeddings and MLPs.

0.07

Controlled Evaluation of EEG Foundation Models: Random Initialisation, Band-Power and Validation-Selected Probes

cs.LG cs.AI ·2026-10-23 ·Mingyu Jeon, Suwan Cho, Jae Young Suh
Abstract

EEG foundation models report downstream gains, yet their protocols seldom include three controls. These are the same encoder untrained, spectral features fitted on the same data, and a layer picked without looking at test scores. We therefore evaluate five publicly released EEG foundation models and an in-house JEPA-style model under one protocol with all three controls and several seeds, on five benchmarks and four added task families. The controls change the reading in three ways. First, picking the probed layer by its test score raises the layer-wise peak by an average of +0.034 and reverses the peak-versus-output gap in 9 of 36 cells. Second, with the layer picked on validation, random initialisation matches or beats pretraining for half of the models on sleep and emotion, and no model beats a five-band log-power baseline on emotion or motor imagery. Third, after full fine-tuning only 5 of 30 pairs gain beyond twice their seed spread, and the one clear loss of -0.290 AUROC persists under matched learning rates. From these results we draw four reporting practices, and we release every cell for re-analysis.

0.06

Financial Context-Rich Strategy for Actor-Critic based Agents: A Deep Reinforcement Learning Approach for Stock Trading

cs.AI q-fin.CPq-fin.PM ·2026-10-23 ·Hrishikesh Singh, Aarti Sharma, Millie Pant
Abstract

In modern financial markets, more than 3/4th of the large volume trade is executed by algorithmic trading based on advanced trading strategies to attain higher precision and speed than traditional human traders. The conventional strategies focused on price-based indicators are inadequate in capturing the market dynamism arising from real-world events such as social sentiments, political instability, or sudden economic shifts. With the rise of Deep Reinforcement Learning (DRL) and data availability, there is a requirement for more efficient algorithms that can incorporate these external factors in trading strategies. DRL has shown potential in simulating real-world scenarios and in simplifying complex decision-making since DRL agents can automate the trading process i.e. executing trades in a manner comparable to human traders. They have shown reliable performance since they can update their policies based on risk and quantitative factors. In this paper, we propose a DRL approach based on partial and deep-infusion of sentimental analysis. It features distinct policy designs of actor-critic-based agents: only policy optimization (A2C & PPO) and policy optimization along with Q-learning principles (DDPG, TD3 & SAC). Our findings reveal, that the agents based on policy optimization combined with Q-learning, significantly and consistently demonstrate improved performance upon sentiment deep-infusion in profit maximization as well as risk-averse trading.

0.05

Fully quantum training of parametric circuits with Quantum Hamiltonian Descent

cs.LG quant-ph ·2026-10-23 ·Marco Burato, Leonardo Banchi
benchmarktheory
Abstract

Parametric quantum circuits are usually trained by classical, local optimizers, which use the quantum computer only to evaluate the cost function, or its gradient, one point at a time. Quantum Hamiltonian Descent (QHD) takes a different route: in this quantum algorithm for continuous optimization, the parameters become the position of a quantum particle, with the cost function as its potential, and two time schedules control kinetic and potential energy so that the wavefunction explores the whole landscape and tunnels towards the global minimum. QHD, however, assumes a classically computable cost function, and its digital implementation is poorly understood: standard error bounds grow exponentially with the number of qubits per parameter, and the schedules are borrowed from classical optimization. Here we propose to train parametric circuits with QHD entirely on the quantum computer, through an online step in which the qubits encoding the parameters coherently control the circuit on a second register. We prove that, for cost functions with a smooth periodic extension, including those of parametric circuits, the simulation error admits a bound independent of the grid resolution, and that for circuits the bound grows at most linearly with the number of parameters. We also show that the ideal dynamics depends on the two schedules only through a single curve, which controls the adiabaticity of the search, and that the remaining freedom can be used to minimize the gate count. Simulations on benchmark functions and parametric circuits show that, at equal cost, schedules designed in this way markedly increase the probability of finding the global minimum.

0.05

ReMedI: Inference-Time Refinement and Self-Distillation for Controllable Molecular Optimization

cs.LG cs.AI ·2026-10-23 ·Jia Yi Zhang, Shivanshu Shekhar, Allison McLevige, Tong Zhang
position
Abstract

In drug development, lead optimization must improve competing properties, such as increasing permeability and reducing toxicity, while limiting scaffold changes that could compromise the lead's potency. RL-based post-training, such as C-MORAL, addresses this problem by fine-tuning molecular language models to directly optimize molecular leads for objectives of interest, which relies on expensive task-specific datasets and faces challenges in achieving multi-objective optimization (MOO). We argue a pre-trained model may already contain useful molecular edits that zero-shot generation cannot efficiently reach. Motivated by the limited search capabilities of zero-shot generation, we develop an efficient Metropolis-Hastings search strategy to better explore the proposal distribution. We propose ReMedI, which uses the molecular language model as a base proposer for inference-time alignment. Given a lead and a set of multi-objective tasks, the model proposes edits conditioned on the current molecule, while Metropolis--Hastings refinement guides sequential edits under an objective for property improvement, preservation, and lead similarity. On both C-MuMOInstruct and SMDD benchmarks, refinement improves frozen base models across three backbones, reaching significantly higher multi-objective optimization success rates compared to baselines. These search traces can then distilled into the base model which improves zero-shot generation and subsequent search using approximately 7.8 times less training molecular leads than that of C-MORAL. ReMedI achieves approximately 2.8 times the strict-success coverage of RL baseline C-MORAL, including on out-of-distribution tasks, with matched lead-relative similarity.

0.05

Optimal Multiclass Online Regret with Arbitrary Label Spaces

cs.LG stat.ML ·2026-10-23 ·Sajad Ashkezari, Shai Ben-David
theory
Abstract

We determine the optimal regret in agnostic multiclass online learning with arbitrary, possibly infinite, label spaces. For every hypothesis class of multiclass Littlestone dimension $d<\infty$, we prove a regret upper bound of $O(\min\{T,\sqrt{dT}\})$ over $T$ rounds, matching the known lower bound. This resolves a question of Hanneke et al. (2023) and extends the optimal binary regret bound of Alon et al. (2021). Our proof strengthens the one-sided expert cover of Hanneke et al. (2023) to a fractional cover. We combine this cover with a minimax reduction to a finite expert game, whose best-response rewards provide the disagreement bounds needed for chaining.

0.04

From Visual Affinity to Learning Credit in Multimodal Reinforcement Learning

cs.AI ·2026-10-23 ·Yuhang Han, Dingning Liu, Yujie Chen, Yaojie Zhang, Jingwei Song +5 more
benchmark
Abstract

Reinforcement learning with verifiable rewards (RLVR) improves multimodal reasoning, but uniformly assigning a single response-level advantage to all tokens overlooks the distinct roles of visual observations and subsequent reasoning. Visual affinity alone is insufficient: predictive states may align with both image and prompt-context representations, and upweighting valid visual observations can amplify their negative-advantage updates when later reasoning fails. To address this issue, we propose \textbf{Outcome-Conditioned Visual Credit (OVC)}, a plug-in framework that separates visual-association measurement from token-level credit assignment and is adaptable to policy-optimization methods with uniform response-level advantage allocation. Relative Visual Encoding constructs a context-relative visual coordinate by contrasting each predictive state's affinity to image and prompt-context representations. Outcome-conditioned Evidence Calibration combines this coordinate with uncertainty-scaled predictive features and calibrates their contributions to outcome-discriminative evidence through mean contrasts between positive- and negative-advantage responses, scaled by regularized within-side variation. Visual association thus informs credit through observed outcome differences rather than directly determining update priority. Normalized Credit Allocation conditions the calibrated evidence on the response's advantage sign to produce positive token weights with unit global mean. These detached weights preserve advantage signs while modulating individual token-gradient contributions. OVC reuses existing rollouts without auxiliary reward models, process annotations, or inference-time modules. Instantiated through GSPO-token, OVC retains sequence-level clipping and consistently improves six-benchmark mean accuracy over GSPO across three models, with gains of 1.86-2.39 percentage points.

0.04

Language, Vision & Media 205

The Emergence and Sparsification of a Syntactic Circuit

cs.CL ·2026-10-23 ·Qing Yao, Sasha Boguraev, Tiago Pimentel, Kyle Mahowald
top venue
Abstract · 26 pages, 7 figures, EMNLP 2026

To understand the emergence of grammatical behavior in language models, we track circuits for subject-verb number agreement in pythia-1b and ten seeds of pythia-410m across training. We find that subject-verb agreement is learned early, and is initially handled by a diffuse collection of neurons. However, after agreement behavior has stabilized, the mechanisms responsible for it continue to reorganize. The attention heads which mediate agreement often shift during training: in the case of pythia-1b, behavior is controlled by two heads early in training but by a third (different) head later. Furthermore, the circuit shrinks over training, with pythia-1b's final circuit being strikingly sparse. We take these results to be informative about how linguistic knowledge emerges and is represented in a language model.

0.69

Position, Role or Label? A Controlled Study of Instruction Conflicts in LLM Agents

cs.CL cs.AIcs.SE ·2026-10-23 ·Anish Laddha, Nitesh Pradhan, Gaurav Srivastava
code/data
Abstract · 55 pages, 8 figures, 24 tables. Code and data: https://github.com/anishh15/WhoWins (pip install whowins). Website: https://anishh15.github.i…

LLM agents take instructions from the operator, repository configuration, user, tool results and saved memory, which often conflict. Instruction-hierarchy benchmarks ask whether the higher-ranked source wins, but a source sets position, role and label at once. First, we build WhoWins, 1,080 code-scored scenarios in five task types, each with up to five mutually exclusive instructions, and seven controlled experiments that vary only how instructions are delivered. Second, we study 14 open-weight models (0.6B to 122B, thinking off and on) over 1,887,356 scored replies, checking main findings on held-out scenarios. Inside one message the first of two conflicting lines wins 72.42 to 94.66%, and across messages it keeps 22.95 to 62.65%. A policy sent first in the system (or developer) message loses most conflicts with later lines in every Qwen model but wins most on gpt-oss when it acts. In tool and file-write tasks it wins less often than the same text sent first as a labelled user message, in every model-mode. Third, we test whether pairs forecast larger conflicts; forecasts improve in every model-mode once they know where the task line is. Finally, by our keyword patterns at most 2.92% of decided pair replies flag or hedge a conflict.

0.64

PathJev: Decision-Native Vision-Language Modelling for Histopathology

cs.CV cs.LG ·2026-10-23 ·Cristian Tommasino, Antonio M. Rinaldi
code/data
Abstract · 25 pages, 4 figures. Code: https://github.com/CosmoIknosLab/PathJev ; model and demo: https://huggingface.co/Cristiandieti/PathJev-P

Vision-language models used for decisions in histopathology rarely return calibrated probabilities over the candidates a user cares about. Decision-native models write a question and a candidate set chosen at run time into the input and return a normalised distribution from a small decision head in one forward pass. We build PathJev-P, a decision-native model for histopathology: MedGemma-1.5-4B with LoRA adapters and a decision head, trained on 40,000 patch states from three public collections whose expert labels are turned into typed questions with organ-specific candidate sets and a "none of these" option. In a comparison over 3 seeds with an analysis plan fixed before testing, on test sets without documented overlap with the training data of the backbone, PathJev-P is more accurate than a generative model trained with the same backbone, data and selection rule, the zero-shot pathology model CONCH and a general medical decision model (balanced accuracy 0.541 against 0.492, 0.345 and 0.213); its advantage over the generative model comes from two organs absent from training. The calibration advantage depends on how the generative model's probabilities are read out. With the summed candidate log-likelihood used in previous work, PathJev-P has the lower calibration error (ECE 0.204 against 0.279); with a length-normalised read-out of the same generative model, examined post hoc, the two have similar mean calibration error (0.201) and the generative model is better calibrated on the unseen organs, where PathJev-P remains over-confident. A supervised linear probe on a pathology foundation model is better calibrated in every setting. In a qualitative whole-slide demonstration, PathJev-P labels unsupervised tissue clusters with abstention and separates carcinoma from other tissue while running the model on a fraction of the tiles.

0.63

OlmoEarth v1.3: Embeddings

cs.CV ·2026-10-23 ·Gabriel Tseng, Favyen Bastani, Patrick Alan Johnson, Yawen Zhang, Piper Wolters +11 more
Abstract

We make some OlmoEarth embeddings. To do this, we train a new model: OlmoEarth v1.3. OlmoEarth v1.3 is designed to condense multimodal, multitemporal remote sensing data into a single digestible 64 or 128-dimensional vector. OlmoEarth v1.3 produces state-of-the-art embeddings. In addition, OlmoEarth v1.3 is a strong encoder backbone -- competitive with OlmoEarth v1 and v1.2 -- while requiring 18% and 53% of the GPU hours to train, respectively. We use OlmoEarth v1.3 to compute global, annualized 10m embeddings, which we release under the OlmoEarth artifact license.

0.56

MDMR-Bench: A Multi-Dimensional Benchmark for Multi-Reference Image Generation

cs.CV ·2026-10-23 ·Bowen Zheng, Jintao Lin, Changheng Yi, Haoran Yang, Binxiao Huang +3 more
top venuebenchmarkcode/data
Abstract · Bowen Zheng and Jintao Lin contributed equally to this work. Accepted at NeurIPS 2026, Evaluations & Datasets Track

Image generation models have rapidly advanced, with many models now supporting generation conditioned on multiple reference images and textual instructions. Such models enable a wide range of applications, including entity fusion and attribute transfer across images. However, existing benchmarks for multi-reference image generation rely on limited evaluation dimensions and fail to capture the diverse challenges of the task. To address this limitation, we propose MDMR-Bench, a Multi-Dimensional benchmark for Multi-Reference image generation. MDMR-Bench is organized around representative capability settings and consists of multiple evaluation subsets spanning diverse dimensions of multi-reference image generation, including (1) common entity fusion and attribute transfer, (2) multi-element extraction from a single image, (3) infographic manipulation, and (4) visual reasoning-intensive generation. This design enables systematic analysis of model capabilities and failure modes across different challenge dimensions. Furthermore, to achieve more reliable and interpretable evaluation, we introduce a checklist-based evaluation protocol that decomposes generation quality into fine-grained evaluation aspects. Instead of relying solely on holistic scoring, our framework uses structured checklists composed of both global-level and atomic-level queries, enabling more diagnostic assessment of model performance. By benchmarking both open-source and proprietary models, we demonstrate that MDMR-Bench provides fine-grained insights into model performance across different dimensions. The dataset and code are publicly available.

0.52

Acoustic and Lexical Signals for Native Language Identification of Language Learners

cs.CL cs.LG ·2026-10-23 ·Ahmet Yavuz Uluslu, Mengjie Qian, Kate Knill, Mark Gales
published
Abstract · Accepted for publication at NLP4CALL

Spoken native language (L1) identification for second language learners has traditionally relied on acoustic attributes of speech. Comparatively little attention has been paid to analysing speech transcripts for L1 traces. In this paper, we present a study of L1 identification using both speech and corresponding transcripts from a spoken learner English corpus. We compare a WavLM-based audio L1 classifier with a zero-shot GPT-4o classifier operating on transcripts and analyse their predictions across proficiency levels using a CEFR-balanced evaluation set. Our experiments show that the audio classifier outperforms the transcript-based approach overall when full transcripts are available, although their relative strengths vary with learner proficiency. Audio performs better at A1, whereas text performs better from A2 to C1 with full transcripts. Transcript-based accuracy generally increases with proficiency, while acoustic performance remains comparatively stable. Restricting transcript length substantially reduces text-based accuracy, whereas audio is less sensitive to input length and performs relatively well with limited input. These findings provide insights into L1 identification from acoustic and textual representations of learner speech and how performance varies with proficiency and the amount of available language material.

0.50

ParaTaxe: a Suite of Tools for Processing and Publishing Multi-parallel Corpora

cs.CL ·2026-10-23 ·Olivier Kraif, Cyrille François, Elnaz Jalilian, Natacha Rimasson-Fertin, Pascale Roux
publishedcode/data
Abstract

Over the past fifteen years, digital tools for parallelizing and aligning textual corpora have developed significantly within the field of digital humanities. Several projects and software tools have been created to compare different translations or versions of texts. Despite these technical advances, many parallel corpora produced in academic contexts quickly become inaccessible online due to the lack of maintenance or funding. To address this gap, we present a suite of open-source tools developed within the ACR Grimm Tradalign and ParaTaxe projects. This suite includes the AIlign aligner, based on multilingual embeddings, a Python processing pipeline for converting and aligning corpora, a web interface called ParaTaxe for managing collections of parallel corpora, and a web interface for exploring aligned corpora built with TEI-Publisher.

0.50

DriveOSS: Dataset and Benchmark for 3D Anomaly Segmentation and Open-set Semantic Segmentation in Autonomous Driving

cs.CV ·2026-10-23 ·Zhiran Yan, Savankumar Bhanderi, Sourabh Lolge, Rui Song, Girdharbhai Paneliya +2 more
publishedbenchmarkcode/data
Abstract · Accepted to WACV 2027

Recent advances in 3D semantic segmentation have significantly improved scene understanding in autonomous driving. However, existing methods are typically developed under a closed-set assumption and evaluated under standard driving conditions, where all categories are known in advance. In real-world driving, autonomous vehicles inevitably encounter previously unseen objects, making closed-set perception insufficient for safe deployment. To address this challenge, we introduce DriveOSS, which, to the best of our knowledge, is the first real-world multimodal dataset and benchmark specifically designed for 3D anomaly segmentation and open-set semantic segmentation. It contains 12,661 synchronized and annotated frames captured using six surround-view RGB cameras, a 128-beam LiDAR, and a GPS/IMU unit. The dataset comprises 52 real-world driving scenes collected under clear, nighttime, cloudy, rainy, and snowy conditions and is organized into ID (in-distribution) and OOD (out-of-distribution) subsets. The ID scenes contain only known classes and are used for training, whereas the OOD scenes additionally contain unknown objects annotated as the outlier class and are used for validation and testing. Extensive experiments show that unseen objects remain a major challenge for current 3D perception systems. The dataset, benchmark protocols, and code are available at https://github.com/chrisyan/DriveOSS.

0.49

Sam-Correction: Fully Adaptive Label Noise Reduction for Medical Image Segmentation

cs.CV physics.med-ph ·2026-10-23 ·Takuro Shimaya, Masahiro Saiko
publishedhighly cited
Abstract · 5 pages, accepted to IEEE International Symposium on Biomedical Imaging (ISBI) 2024

The quality of the teacher label has a dramatic impact on the outcome of the machine learning process. Medical image segmentation often suffers from a noisy ground truth (GT) mask due to difficulties in consistent annotation. Although a recent study has proposed a promising method to curate the noisy GT mask, this requires additional well-curated GT masks that are not always accessible without the aid of multiple medical experts. As an alternative, we propose SAM-Correction, which utilizes a foundation model to correct the noisy GT mask in a fully adaptive manner instead of manually preparing well-curated masks. In our approach, we progressively mitigate label noise via the segment anything model (SAM), where prompts are given from another segmentation network that we train concurrently. Experiments with various types of label noise derived from actual human annotations demonstrate that SAM-C achieves precise noise reduction and high-performing segmentation models trained thereby. Our method paves the way for robust model training even when clean labels are entirely unavailable.

0.48

SemGeo-Gen: Unsupervised Generation of Approximate Cross-Instance Semantic-Geometric Correspondences

cs.CV cs.AI ·2026-10-23 ·Roy Amoyal, Shira Ifergane, Oren Freifeld
top venuecode/data
Abstract · Accepted to NeurIPS 2026

We introduce SemGeo-Gen, an unsupervised method for generating approximate cross-instance semantic-geometric supervision from weakly aligned 3D object collections. Our goal is not to recover exact point-level correspondences, which are often ambiguous across different instances and would require expensive, impractical manual supervision at scale. Instead, we automatically produce approximate correspondences that are semantically meaningful, geometrically consistent, and diverse enough to train modern correspondence models. Given only coarse category-level rotation alignment, SemGeo-Gen lifts multi-view DINOv2 features to 3D, aligns object instances using continuous piecewise-affine registration, and prunes candidate matches using semantic consistency. The resulting 3D correspondences can be projected into rendered views, yielding scalable 3D-3D, 3D-2D, and 2D-2D supervision without manual annotation. We validate the generated correspondences against sparse human annotations from KeypointNet, obtaining 93% PCK@0.10 and 97% PCK@0.15. More importantly, we show that using the generated correspondences for synthetic pretraining improves recent state-of-the-art semantic correspondence models on SPair-71k after fine-tuning on real data, with gains of up to 5 PCK@0.10 points. Together, these results indicate that approximate semantic-geometric supervision generated from 3D assets can improve real-image correspondence learning and serve as a scalable alternative to costly manual annotation. Project page: https://bgu-cs-vil.github.io/SemGeo-Gen/.

0.47

How Does Post-Training Shape LLM Value Preferences?

cs.CL ·2026-10-23 ·Yang Liu, Shaoyang Xu, Masahiro Kaneko, Chengzhi Zhong, Wenxuan Zhang +1 more
code/datahot direction
Abstract · 46 pages, 21 figures, 21 tables

During post-training, large language models (LLMs) learn value preferences from corpora that may contain imbalanced proportions of opinions associated with different demographic groups. Amplifying these imbalances may further underrepresent some groups' opinions. However, how opinion mixtures shape value preferences across post-training paths remains unclear. To answer this question, we conduct controlled experiments using post-training corpora with different opinion mixtures and five post-training paths based on SFT, DPO, and GRPO. We find that post-training amplifies opinion imbalances differently across post-training paths. Once value preferences are amplified, whether further post-training can remove them depends on both the post-training method and the opinion mixture in the corpus. More concerningly, a model's stability under interventions depends on whether further post-training changes its value preference. Moreover, these value preferences are transferred to student models even when distillation uses corpora with no explicit opinion questions. These findings suggest that both the opinion mixture in post-training corpora and the choice of post-training method should be considered when developing, evaluating, and reusing LLMs. Our code and data are available at https://github.com/ku-nlp/llm-value-preferences.

0.47

PCFBench: How Far Can Large Vision-Language Models Go in Physics-aware Photonic Inverse Design?

cs.CV ·2026-10-23 ·Shengchao Chen, Ting Shu, Sufen Ren
top venuebenchmark
Abstract · Accepted to NeurIPS 2026, 66 pages, 30 figures, 27 tables

Photonic crystal fibers (PCFs) underpin critical applications in telecommunications, sensing, and high-power laser delivery, yet their design remains a manual, expert-intensive process that couples geometric intuition with deep waveguide physics. Vision-language models (VLMs) offer a promising path toward automating PCF design, but can they truly reason about photonic physics from microscopy images, or do they merely recognize visual patterns? We introduce PCFBench, the first comprehensive benchmark for evaluating VLMs on the perception, reasoning, and decision steps of the photonic inverse-design loop. PCFBench spans 26 tasks organized into six categories of increasing cognitive demand (Geometry Perception, Physics Understanding, Text Generation, Multi-modal Reasoning, Inverse-Design Decision, and Code Generation), built from over 220K FDTD simulation samples across 18 structurally diverse fiber families. We evaluate both individual VLMs on isolated tasks and multi-agent systems (MAS) on the full sequential pipeline from perception to design. Our results reveal a striking perception-design gap: current models can identify fiber geometry with moderate success, yet consistently fail to translate that visual understanding into physics-aware design actions. Moreover, cascading agents through the pipeline exposes significant error propagation, where upstream perception mistakes compound into downstream design failures. These findings suggest that neither model scaling nor naive agent chaining suffices, and that closing the loop demands explicit physical grounding. PCFBench provides the community with a rigorous, multi-dimensional testbed to measure progress toward AI-assisted photonic engineering. The project is available at https://pcfbench.github.io/.

0.46

ARENA: Active Region attENtion Aggregation for interpretable solar flare prediction

cs.CV astro-ph.SRastro-ph.IMcs.LG ·2026-10-23 ·Luis Bauer, Annalena Kofler, Andrés Asensio Ramos, Aparna Venkataramanasastry, Valentín Martínez Pillet +1 more
publishedcode/data
Abstract · 7+12 pages (2+8 figures), accepted at the "Representations for the Physical Sciences" workshop at NeurIPS 2026

Predicting solar flares is a challenging space-weather forecasting problem with potentially significant impacts on Earth's upper atmosphere, including disruptions to radio communications. We propose Active Region attENtion Aggregation (ARENA), a neural network architecture that predicts solar flares from a variable number of active-region cutouts while learning to identify the region most relevant to the prediction. ARENA is trained using multiple-instance learning under weak supervision, requiring only a flare/no-flare label for the observation and no labels specifying the flare origin. The learned attention mechanism therefore provides a means of localizing the active region driving the prediction without explicit origin annotations. ARENA achieves a heidke skill score of 0.45 which is comparable to other flare prediction approaches. These results demonstrate that active-region-based aggregation can provide an efficient and interpretable approach to solar flare prediction. The code is available at https://github.com/LuisBauer22/ARENA.

0.46

Forging Emotions: a deep learning experiment on emotions and art

cs.CV cs.CYcs.HC ·2026-10-23 ·Amalia Foka
published
Abstract · 10 pages. Published in Artnodes no. 31 (2023), Universitat Oberta de Catalunya, under CC BY 4.0. Extended version of a paper presented at IS…

Affective computing is an interdisciplinary field that studies computational methods that relate to or influence emotion. These methods have been applied to interactive media artworks, but they have focused on affect detection rather than affect generation. For affect generation, computationally creative methods need to be explored that have recently been driven by the use of Generative Adversarial Networks (GANs), a deep learning method. The experiment presented in this paper, Forging Emotions, explores the use of visual emotion datasets and the working processes of GANs for visual affect generation, that is, for generating images that can convey or trigger specified emotions. This experiment concludes that the methodology used so far by computer science researchers to build image datasets for describing high-level concepts such as emotions is insufficient and proposes utilizing emotional networks of associations according to psychology research. Forging Emotions also concludes that to generate affect visually, merely corresponding to basic psychology findings, such as bright or dark colours, does not seem adequate. Therefore, research efforts should aim to understand the structure of trained GANs and compositional GANs in order to produce genuinely novel compositions that can convey or trigger emotions through the subject matter of generated images.

0.46

The Paired Thermal Channel Is Two Orthogonal Axes: A Controlled Attribution Study of Calibration Cost and Detection Gain in Infrared Person Detection

cs.CV cs.MM ·2026-10-23 ·Tian Zhang, Yang Chen, Bing Bai, Weikai He, Xiang Gao +1 more
theoryunusual
Abstract

The calibrated thermal channel in infrared person detection comprises two structurally orthogonal axes. True calibrated $ΔT$ outperforms scale-matched grayscale pseudo-$ΔT$ by $+0.0221$ mAP (paired Wilcoxon $p=0.0098$, Cohen's $d=+1.08$), confirming that the calibrated channel carries information the grayscale channel lacks. The calibration axis is driven by the joint pairing-scale condition. Preserving both paired correspondence and absolute calibrated scale raises expected calibration error (ECE) from $\approx 0.05$ to $0.142$. The $2\times 2$ interaction ($p=0.0059$) localizes the rise to this condition. Breaking either factor alone returns ECE to the broken-factor level (true-vs-const $p=0.002$; true-vs-shuffled $p=0.002$). Under post-hoc temperature scaling, the pairing cost survives recalibration ($p=0.002$), while the scale cost is largely absorbed ($p=0.16$). As regards the detection axis, the gain stems from structural de-structuring. Phase randomization, which destroys spatial structure while preserving per-sample variance, yields $+0.0185$ mAP over true $ΔT$ on FLIR ADAS ($p=0.014$, $d=+0.96$). The FLAME~3 gain is $+0.0064$ mAP ($p=0.078$, n=8). On the calibration axis, the same intervention is insignificant on FLIR ADAS ($p=0.47$) but significant on FLAME~3 ($p=0.0078$). These findings rest on a five-arm controlled attribution with temperature-evidential learning (TEL) as the testbed. TEL obtains a single evidence-derived routing scalar together with a complementary diagnostic entropy, and establishes a finite-evidence degeneracy theorem proving that the acquisition product degenerates to this scalar with a closed-form bound. A physical $0.5^\circ$C threshold admits cross-dataset physical readings, whereas a gray threshold of $20$ inverts without physical interpretation.

0.45

A Regularity Condition Under Which a CPAB Transformation is a Diffeomorphism

cs.CV ·2026-10-23 ·Amit Aflalo, Oren Freifeld
unusual
Abstract

Continuous Piecewise-Affine Based (CPAB) transformations, introduced by Freifeld et al. (2015, 2017), are flows obtained by integrating continuous piecewise-affine (CPA) velocity fields. These expressive transformations have proven useful in applications including nonlinear temporal warping and trainable activation functions in 1D, as well as nonrigid spatial deformations in 2D and 3D. While computational costs were not an issue in 1D, they originally limited the practical expressiveness of CPAB transformations in higher dimensions, especially in 3D. Our recent work (Aflalo et al., 2026) removed this bottleneck, enabling efficient optimization with much finer tessellations and substantially greater expressiveness. There is, however, a mathematical subtlety that was overlooked in the original CPAB papers (Freifeld et al., 2015, 2017). CPAB transformations are always bi-Lipschitz homeomorphisms, implying that both the transformations and their inverses are differentiable almost everywhere. They are not, however, necessarily diffeomorphisms. This note remedies the omission by identifying sufficient regularity conditions under which a CPAB transformation is guaranteed to be a diffeomorphism. We explain how differentiability can fail and how the proposed conditions prevent this failure. In 1D and 2D, we show that these conditions can be easily enforced and, for commonly used tessellations, hold with probability one under standard sampling assumptions. We expect analogous behavior in 3D.

0.45

Measure Before You Train: Typed Decision Models vs. a Classical Baseline vs. a Generative LLM for Routing a Personal Multilingual Inbox

cs.CL cs.HCcs.LG ·2026-10-23 ·Federico Lucca
code/data
Abstract · 11 pages, 1 figure, 10 tables. v3: adds Jeva-4B as the eleventh system (local, T4) with the authors' confirmation of the setup; majority cla…

Personal email agents decide, for every message, its type and urgency and whether to draft a reply, create a task or ignore it. In production these decisions are delegated to a generative LLM or to rules, and their error rate is rarely measured: the stored labels were written by the system itself. We describe Foyer, a self-hosted personal inbox agent, and a measurement protocol -- stratified sampling, frozen snapshots, thread-grouped splits, secret redaction, human labels in the review loop -- that produces ground truth before any training. On 231 emails in Italian and English we compare eleven systems: the production rules, a TF-IDF + logistic-regression baseline, local and hosted typed decision models (Tev1, Nimble, Jeva, Jev) and generative LLMs (Qwen2.5, Gemma 4). Four results. The production rules for message type score 28% against a 57% majority class. With about two hundred labels a logistic regression has the highest accuracy on every field (type 0.79, urgency 0.74, needs-reply 0.83, is-noise 0.90) at a latency three orders of magnitude lower, though it never predicts the two rarest types; the best zero-shot system reaches 0.74 on type. The sender's relationship to the owner is recovered from other emails of the same contact with 98.6% accuracy on the 63% of emails whose sender was seen before: it belongs in the contact graph, not in the prompt. Per-email accuracy is not the product metric: composed into actions, the baseline ignores 18% of the emails that needed a reply or a task, against 4% for the best generative model; a cost-chosen threshold cuts that to 6% and makes the classifier the cheapest system at every price. An adversarial third round (722 emails) with eleven pre-registered hypotheses confirms seven and refutes four. Two blind annotators agree with the reference at kappa 0.81 and 0.86 on type. We release the protocol and the harness; the emails stay private.

0.45

RootDown: Deploying and Benchmarking Models for Beet-Root Detection on Edge Devices

cs.CV cs.LG ·2026-10-23 ·Gerardus Croonen, Andreas Trondl, Julia Simon, Daniel Steininger
benchmarkcode/data
Abstract · Accepted at MLCSA Workshop at Asian Conference on Computer Vision (ACCV 2026). Benchmark, code and models available at https://github.com/se…

Quality inspection of sugar beets after harvest and storage is critical to preserve sugar content and ensure optimal processing efficiency. The transfer of harvested beets to field clamps or trucks provides an opportunity to visually inspect individual beets automatically on loader conveyors. However, this is a challenging task as the large number of beets causes partial overlap and occlusion, complicating both annotation and detection. In this work, we propose an end-to-end image-based pipeline for detecting partially occluded sugar beets in loading scenarios as a first step toward automated quality inspection. We introduce a dedicated benchmark and present an automatic annotation strategy that leverages the SAM 3 foundation model to generate bounding boxes. We compare models trained on manually annotated and automatically generated labels. Additionally, we investigate deployment considerations by benchmarking multiple inference configurations, including execution on an edge device, under realistic operational constraints regarding precision, inference time and power consumption. Our results show that, despite substantial occlusion and inherent annotation ambiguity, foundation-model-based labels can substitute for a substantial share of manual annotation, trading labeling effort for data volume at a modest accuracy cost. Furthermore, we demonstrate that a Raspberry Pi 5 equipped with a Hailo accelerator enables high-performance, low-power inference, making it a practical solution for mobile outdoor deployment.

0.45

Flownomaly: Single-Step Conditional-Mean Recovery for Few-Shot Anomaly Detection

cs.CV cs.LG ·2026-10-23 ·Bernardo Perrone Ribeiro, Erkut Akdag, Egor Bondarev
publishedcode/data
Abstract · Accepted to WACV 2027

Few-shot industrial anomaly detection requires modeling category-specific normality from only a handful of defect-free images. Common approaches tie inference to stored exemplars, impose rigid linear structure, or risk reproducing anomalous content. Flownomaly instead casts the problem as normal-counterpart recovery in frozen DINOv2 feature space. Given normal supports, a velocity field is learned through conditional flow matching over frozen patch tokens to transport Gaussian noise toward normal features. Along straight paths, a single recovery step using one evaluation of the optimal field yields the posterior mean of the clean normal token without trajectory integration. The operating time governs how strongly recovery modifies a query token. With no anomalous validation samples available for tuning, a normal-only Jacobian gain-to-noise criterion selects this time from the supports alone. Squared residuals between query tokens and their recovered counterparts yield anomaly signals for both dense localization and image-level detection. Across 1-, 2-, and 4-shot benchmarks on MVTec-AD and VisA, together with sample-shot evaluations on Real-IAD, Flownomaly achieves state-of-the-art performance against recent category-specific methods. In the 1-shot setting, image- and pixel-level AUROC reach 97.8% and 98.0% on MVTec-AD and 95.3% and 98.7% on VisA, with a 2.1-point image-level AUROC improvement over the strongest baseline on VisA. Code is available at https://github.com/b-rbmp/Flownomaly.

0.44

Beyond Pattern Matching in 3D MLLMs: Part-Level Functional Reasoning for Generalizable Recognition

cs.CV cs.AI ·2026-10-23 ·Weihong Li, Hao Jiang, Yaohua Zha, Tao Dai, Shu-Tao Xia
top venuecode/data
Abstract

3D MLLMs recognize objects via pattern matching, which fails when geometrically similar objects require functional reasoning to disambiguate (e.g., dresser vs. wardrobe). Our per-sample Geometric Confusion Score (GCS) exposes this failure: baseline accuracy is 33.7% on ModelNet40's most ambiguous 20% versus 84.0% on its least ambiguous 20%. We identify a function-as-annotation pattern, in which functional language decorates a known identity instead of supporting inference, and hypothesize that reversing this information flow should concentrate improvements on the most ambiguous samples. PFIR (Part-level Functional Inference for Recognition) reverses that flow. Built on PartNeXt part segmentation, its data pipeline produces 507K QA pairs with functional CoTs (geometry $\to$ function $\to$ identity), and its progressive curriculum culminates in dual-objective GRPO: strict functional reasoning and descriptive grounding for Objaverse's diverse distribution. The two RL objectives yield nearly orthogonal task vectors ($\cos = 0.056$), enabling composition via Task Arithmetic. Under a strict dual OOD protocol (Objaverse $\to$ ModelNet40; XYZRGB $\to$ XYZ-only), PFIR raises zero-shot accuracy from 60.5% to 66.7%, with gains concentrated $2.6\times$ more on geometrically ambiguous samples---supporting part-level functional reasoning as an effective route to more generalizable 3D recognition. Code: https://github.com/this233/pfir-3d.

0.44

Semantically Rich, Metrically Affect-Blind: A Geometric Blind Spot in Modern Text Embedders

cs.CL cs.LG ·2026-10-23 ·Mahdi Naser Moghadasi
code/data
Abstract · 6 pages, 2 figures, 3 tables. Code: https://github.com/mahdinaser/sent2vec-sise

We report a systematic dissociation in modern text embedders: sentiment is almost perfectly decodable from their embeddings, yet almost entirely absent from their metric. Across eleven frozen encoders spanning 2021 to 2026, including 2026-generation models and the 4B-parameter LLM embedder Qwen3-Embedding-4B, a linear probe recovers sentence polarity at up to 91.3% accuracy while cosine similarity separates polarity by a margin of only 0.003-0.064. Qwen3-Embedding-4B, the strongest probe in the panel, reaches 91.3% probe accuracy but only 64.3 retrieval MAP. Because retrieval, clustering, deduplication, and retrieval-augmented generation operate on the metric rather than on a trained probe, these systems are effectively blind to information their encoder demonstrably contains. A task-conditioning control suggests this reflects the pretraining objective rather than a prompting artifact: conditioning Jina-v5 on its classification task rather than its retrieval task recovers only about half of the MAP gain that injection achieves on the same weights (63.2 to 73.8, versus 85.1), and leaves the alignment gap essentially unchanged. Sentiment-Injected Sentence Embeddings (SISE) train a small (0.13M-0.69M-parameter) supervised-contrastive head on a frozen encoder in seconds on a CPU, using the labelled training set. The head improves every encoder released since 2022 on full-data probe accuracy (up to +5.8 points), retrieval MAP (up to +26.0) and polarity margin (7-61x). This is a specialization, not a free gain: the compact 128-d head loses 38-41 Spearman points on STS-B, while a fused variant that concatenates it with the base embedding loses 2.4-12.0. With the head, a 0.6B encoder at 128 dimensions matches a 4B encoder given the same head (91.8 vs. 90.3 MAP). We release a CPU-reproducible pipeline and a reusable diagnostic, the Attribute Alignment Gap.

0.44

VTRL-Med: Progressive Visual Refinement with Topology-Guided Regularization for Medical Image Clustering

cs.CV ·2026-10-23 ·Guangyu Meng, Pengfei Gu, Peixian Liang, John P. Lalor, Erin Wolf Chambers +1 more
Abstract

Deep clustering offers a promising approach for analyzing large unlabeled medical image datasets, yet prevailing methods still face two limitations: they may neglect class-agnostic structural priors that capture object boundaries and regions, and they do not explicitly encode topological patterns (e.g., connectivity and holes). These sources of information play distinct and complementary roles: visual-structural features provide the primary discriminative representation for clustering, while topological features provide additional global geometric cues for residual visual ambiguities. We introduce VTRL-Med, a progressive visual-structural learning framework with topology-guided regularization for medical image clustering. VTRL-Med first adapts the Segment Anything Model (SAM) to extract image features that retain structural cues inherited from mask-supervised pretraining. It then generates reliable pseudo-labels through a two-step filtering process that combines prototype-based selection with k-nearest neighbor consistency validation. Finally, it progressively refines the visual branch while an auxiliary topology branch, constructed from persistent homology (PH) descriptors, provides additional geometric regularization during training. Across 13 medical image benchmarks, VTRL-Med achieves state-of-the-art clustering performance, improving average accuracy by 4.48% over existing methods. On three representative datasets, topology-guided regularization further improves accuracy by 1.3--4.3% over visual refinement alone.

0.43

Augmentation Strategy for DINO in Glaucoma Classification

cs.CV ·2026-10-23 ·Ashish Kumar Meena, Chandra Sekhar Seelamantula
publishedbenchmarkcode/data
Abstract · 11 pages, 3 figures. Accepted at the 17th International Workshop on Machine Learning in Medical Imaging (MLMI 2026), held in conjunction wit…

Self-supervised learning (SSL) offers a promising approach to medical image representation learning, yet augmentation strategies of SSL frameworks validated on natural images do not reliably transfer to specialized medical imaging domains. We analyze DINO, a knowledge-distillation based SSL framework, and empirically demonstrate that its default augmentation strategy produces unreliable representations for retinal fundus images, degrading downstream glaucoma classification performance. Guided by the principle that effective cropped views should share minimal redundant information while retaining clinically relevant signal, we redesign DINO's augmentation strategy. Our approach biases the learned representations toward diagnostically critical features rather than low-level image statistics. We evaluate our method for glaucoma classification by training a linear classifier on the frozen representations, achieving strong generalization across seven benchmark datasets. URL for code and weights: https://github.com/ashish-256/task_dino

0.43

JIT-MPE: Just-in-Time Multi-Pitch Estimation under Output Deadlines

cs.SD eess.AS ·2026-10-23 ·Yoshiyuki Kobayashi
code/data
Abstract · 18 pages, 3 figures. Code and trained weights: https://github.com/Frieve-A/jit-mpe

Live pitch displays, score followers and accompaniment systems must know which pitches are sounding within tens of milliseconds, but instrument-agnostic multi-pitch estimation (MPE) has mainly been designed and evaluated offline. We present JIT-MPE, a causal MPE model of 86k parameters for mixed ensembles. It gives an immediate estimate of each 20 ms frame and re-estimates the same frame as 20, 40, 80 and 160 ms of further audio arrive, so that a live display can show pitches at once and then refine them. We score each output only at a deadline by which it can be computed from the audio that has arrived, on five development sets and four held-out sets evaluated only after all design choices were frozen. At a 50 ms deadline, which reads the update made with 40 ms of lookahead, JIT-MPE is significantly more accurate than five of six published offline MPE systems in 38 of 45 comparisons (15 of 20 on the held-out sets) and less accurate in only two, although these systems read the whole recording and all but one have 6 to 170 times as many parameters; the sixth, supervised on URMP, is higher on four of the nine sets. The deadline costs little: on the held-out sets, JIT-MPE at 50 ms is within $-1.27$ to $+0.60$ points of recording-level macro F1 of its own final output with 160 ms of lookahead and within $-2.08$ to $+0.79$ points of a bidirectional model of similar size. Waiting pays off where the immediate estimate is weakest: the update at 50 ms adds 1.7-3.6 points over it on the development data and three of the four held-out sets, and raises the recall of reference frames in the first 20 ms after a note onset from 9.8% to 38.4% on the held-out AAM set. The slowest measured frame takes at most 4.15 ms on one CPU thread. Code and trained weights are available at https://github.com/Frieve-A/jit-mpe.

0.42

The Locality Trade-off in Masked Patch Self-Distillation

cs.CV ·2026-10-23 ·Xi Weng, Zhaoyu Liu, Lianyu Hu, Jin Song Dong
top venuecode/data
Abstract · Accepted at NeurIPS 2026

Masked patch self-distillation is a key component of modern vision foundation models, including DINOv2 and DINOv3, yet its implications for local visual information remain unclear. We show that its semantic benefits can coexist with a scale-dependent locality cost. Using two annotation-free diagnostics, we distinguish recoverable local appearance from spatial correspondence. In smaller models, adding iBOT substantially weakens patch correspondence while largely preserving appearance decodability. In larger models, correspondence improves, but pixel and texture evidence becomes less linearly accessible in deep features. Thus, scaling changes the form of the locality cost rather than uniformly eliminating it. Our analysis of Gram anchoring further shows that preserving patch relations does not guarantee simultaneous improvements in both properties. These findings motivate evaluating and designing visual representations around the complementary demands of semantic abstraction, appearance preservation, and spatial correspondence. Our code is available at https://github.com/winci-ai/jea.

0.42

VERA-Hand: A Verified Multidimensional Benchmark for 2D Hand Pose Estimation in Third-Person Dynamic Gesture Videos

cs.CV ·2026-10-23 ·Gibran Benitez-Garcia, Hiroki Takahashi
publishedbenchmark
Abstract · Accepted to WACV 2027

Hand landmarks serve as input for gesture recognition, sign language understanding, and touchless human-computer interaction. Yet hand pose estimators are typically evaluated with frame-wise localization metrics such as PCK or MPJPE, which are computed over matched predictions and overlook missed or spurious hands, handedness errors, and temporal instability. To address these limitations, we present VERA-Hand, a benchmark that evaluates estimators using manually verified, frame-level 2D annotations for 10,690 hands across four complementary dimensions (detection, localization, identification, and temporal stability), as well as a joint Pose-F1 metric that requires both correct detection and accurate localization. Evaluating six estimators spanning hand-specific, whole-body, reconstruction-based, and foundation-model paradigms, we find that no estimator dominates: rankings reorder sharply across dimensions. VERA-Hand also exposes failures that localization-only and downstream evaluations may not reveal, including cross-hand overlap in over 85% of frames for a recent foundation model. A post-hoc cleanup case study demonstrates how the benchmark can validate corrective interventions, while stratified analyses show that robustness varies across motion-blur classes and hand regions. We release the verified annotations, protocol, and code at https://gibranbenitez.github.io/VERA_Hand/

0.42

Can LLM Annotators Match Human Consistency? Benchmarking Human Agreement and Optimizing Multi-Agent with Contrastive Knowledge

cs.CL ·2026-10-23 ·Xiaopeng Ke, Zhixin Wang, Xuebo Liu, Xinyu Ma, Xing Zheng +3 more
top venuebenchmarkcode/data
Abstract · Accepted to EMNLP 2026 Findings

LLM-based annotation has advanced rapidly, yet existing benchmarks remain misaligned with real-world practice. Actual annotation often involves lengthy, task-specific guidelines, and different annotators may produce divergent but reasonable judgments, especially on boundary cases, where single-labeled dataset fail to reflect this disagreement. This gap also limits existing single-agent and workflow-based annotation systems, which often lack adaptability to complex, evolving guidelines. To bridge this gap, we introduce ANT-A Benchmark, a real-world multi-annotator benchmark spanning 7 datasets across domains and modalities, with task-specific guidelines and 5 independent human labels per instance. ANT-A enables a distribution-aware evaluation, using Fleiss kappa to assess how closely model annotations align with human labeling distributions. We further propose ANT-OR, a multi-agent annotation framework that refines prompts by distilling contrastive decision rules from positive and negative feedback. Experiments show that ANT-OR improves agreement@>=2 accuracy by 4.5 points over the strongest baseline and achieves the highest alignment with human annotator distributions as measured by Fleiss kappa within 2 points. Code and benchmark are available at https://github.com/Krueske/ANT-OR.

0.41

Every Edge Needs a Quote: Building a Quote-Grounded Knowledge Graph of Healthcare Software Adoption with Language-Model Agents

cs.CL cs.AIcs.IR ·2026-10-23 ·Juan Felipe Duran, Arthur Tobler, Martin Vallières
code/dataunusual
Abstract · 19 pages, 3 figures, 9 tables. Code: https://github.com/MEDomicsLab/healthcare-agentic-kg

Hospitals regularly buy software that clinicians and patients depend on. Because these purchases are costly, hard to reverse and can affect patient care, buyers look for evidence of which peer institutions already use a product and which systems it connects to. This evidence is public but scattered across vendor websites, press releases and third-party pages. Language-model agents can now collect it at low cost, but the correctness of the extracted claims is difficult to establish, and a wrongly listed customer can mislead buyers and misrepresent an organisation. We present a pipeline that builds a knowledge graph of customer and integration relations from the open web under one rule: every edge must be supported by a verbatim quote from its source page. Pages come from a deterministic crawl of each vendor's site and a budgeted ReAct agent that searches the rest of the web. A permissive extractor proposes relations with quotes, and a layered grounding stage decides which to keep, using a referent step that reads the whole page when a quote names only one side, deterministic checks, and a panel of five model calls voting on nine criteria. On 61 vendors, the pipeline reaches 0.94 precision and 0.71 recall against a page-level key covering the 6,428 candidate relations proposed by its extractor, whose verdicts were drafted by one model and re-checked with a second. Against the golden reference, the marketplace's own curated list, it scores lower, but 89% of the relations it keeps that the reference lacks are stated on a page. A confidence score built from the panel's votes keeps more correct relations than the fixed rule at the same precision, and reaches 0.90 precision at 0.90 recall. We also discuss how narrow, bounded model calls, string-matched evidence and voting address long-context degradation and run-to-run variance.

0.41

OpenStripe supports open-set identification and spatial stripe profiling in wild Amur tigers

cs.CV ·2026-10-23 ·Yichun Peng, Yansong Li, Jin Hou, Qi Song, Xi Guo +7 more
code/data
Abstract · 14 pages, 4 figures. Code and nonspatial data: https://github.com/Yichun-Peng/OpenStripe

Non-invasive camera-trap monitoring of endangered wildlife requires open-set individual identification---grouping field images into identities without predefined classes or a known population size---yet pose deformation and cluttered backgrounds complicate image comparison. Here, we present OpenStripe, a multi-agent framework that reconstructs canonical two-dimensional (2D) tiger torso patterns and encodes four interpretable local stripe structures: breakpoints, bifurcations, irregular structures, and round structures. We evaluated OpenStripe on 100 camera-trap records from 28 wild Amur tigers (Panthera tigris altaica), representing nearly one-third of the documented population in the Northeast China Tiger and Leopard National Park. Combining all four complementary structures enabled automatic identity grouping with 0.90 matched accuracy and a 0.85 adjusted Rand index (ARI), outperforming 49 baseline configurations under the same backend (up to 0.43 accuracy and 0.14 ARI). In addition, results from OpenStripe indicate that stripe similarity in Amur tigers declines with geographic distance and suggest that traits such as fractal dimension and inter-stripe spacing are associated with geographic proximity. OpenStripe thus provides an interpretable representation that connects automated individual grouping with spatial phenotypic profiling in wild populations.

0.41

Preference Optimization Improves Definition Generation: Semantic Change in EU Institutional Sustainable Finance Disclosures

cs.CL ·2026-10-23 ·Mariia Fedorova, Andrey Kutuzov, Gustavo Cadenas Delascio
benchmarkcode/data
Abstract · the 5th NLP4PI (NLP for Positive Impact) Workshop@EMNLP 2026

Dictionary-like definition generators are known to be useful for lexical semantic change detection (LSCD). However, they suffer from overly granular sense distribution, producing excessive number of definitions for one and the same lexical sense. In this work, we show that fine-tuning with Direct Preference Optimization (DPO) helps to overcome this problem. We test our approach both quantitatively on a curated academic LSCD benchmark and qualitatively on real-world data for a practical task of auditing diachronic changes in sustainability terminology within project documentation of EU development finance. All models, code and predictions produced in this study are publicly available at https://github.com/ltgoslo/dpo-for-definition-generation.

0.40

Tempo-Zero: Recursive Self-Play in Executable Worlds for Video Reasoning

cs.CV ·2026-10-23 ·Jingmin Zhu, Yanyan Zou, Xiuyu Li, James Bailey, Dinh Phung +8 more
code/data
Abstract

Video reasoning requires understanding how events unfold over time, yet videos with accurate temporal annotations are expensive to collect. Self-play offers an alternative by letting a model generate and solve its own tasks without manual annotation. However, existing video self-play methods train on fixed sets of recorded videos, so they cannot alter the underlying event dynamics and rely on noisy pseudo-labels derived from the model's own outputs, such as majority voting, that risk reinforcing shared model errors. To overcome these limitations, we introduce Tempo-Zero, a self-improvement framework that closes a loop of executable environments, execution-verified self-play, and environment evolution, where executing a program renders a video with the ground-truth option and interval of every question. A single policy plays both Questioner and Solver, proposing questions and answering them with timestamped evidence lines, and is trained on execution-grounded rewards instead of model consensus. The Solver's performance steers a curriculum over tasks and difficulty levels, and the policy selects rule changes that enrich the event dynamics of later rounds. Despite being trained solely on simple programs, without supervised fine-tuning or external video data which prior methods typically require, Tempo-Zero outperforms the base models and recent self-play baselines on average across twelve real-world video benchmarks at different model scales, with particularly strong gains on temporal grounding tasks. Project page: https://alchemist-jo.github.io/tempo-zero/

0.40

Hierarchical GTV Estimation: Bridging the Gap between Container-Level Ranking and Item-Level Conversion

cs.IR ·2026-10-23 ·Zerong Lan, Fan Zhang, Chuang Chen, Tianchun Huang, Teng Zhang +1 more
top venue
Abstract · 5 pages; published in the 20th ACM Conference on Recommender Systems (RecSys '26)

Emerging e-commerce platforms, such as live-streaming and Online-to-Offline (O2O) services, inherently operate under a "Container-based" display paradigm. In this setting, a critical granularity mismatch exists: the platform performs ranking and estimation at the container level (e.g., livestream rooms or stores), whereas the actual transactions occur at the item level within these containers. This structural discrepancy complicates Gross Transaction Value (GTV) estimation, introducing unique challenges including high label variance, severe label sparsity, and dynamic item heterogeneity. To address these challenges, we propose a novel framework named Hierarchical GTV Estimation (HGE). HGE employs a Hybrid Set-Aware Encoder (HSAE) to model the dynamic composition of items and feature interactions. To mitigate label variance, we introduce a Proxy-Label Learning (PLL) strategy that decomposes post-click GTV into purchase probability, quantity, and unit price. Furthermore, a Residual Correction and Ensemble Fusion (RCEF) module is designed to handle label sparsity by effectively combining item-level aggregation with container-level inference. Extensive experiments on large-scale industrial datasets demonstrate that HGE significantly outperforms state-of-the-art baselines, achieving a +0.63% lift in XAUC offline and a +1.42% increase in Revenue Per Search (RPS) in online A/B tests. HGE framework has been fully deployed into our main traffic.

0.38

Guided Zero-Shot Anomaly Explanation for Anonymized Crowd Video Monitoring: Context, Categorization, and Spatial Localization

cs.CV cs.AI ·2026-10-23 ·Mulugeta Weldezgina Asres, Lei Jiao, Christian Walter Omlin
benchmarkcode/data
Abstract · 21 pages, 12 figures, 13 tables, preprint

Despite the promising development of deep learning for video anomaly detection (VAD), explainability (XAI) of these models remains a challenge. Explainable and privacy-protecting VAD builds transparency and trust that facilitates adoption of computer vision systems, including safety and security monitoring. However, anomaly explainability in privacy-preserving VAD systems is underexplored. In this study, we present the XAVAD approach, a unified descriptive anomaly XAI for an anonymized VAD, that integrates flexible zero-shot anonymization, anomaly type categorization, visual localization, and scene context description. We provide empirical and qualitative assessments of multifaceted explanations integrated with full-body visual anonymization for privacy protection. We adopt a multimodal CLIP embedding for videos, using improved LLM-assisted prompting and pool-scoring mechanisms for anomaly-type identification. We enhance Grad-ECLIP for spatial anomaly heatmap localization, leveraging computation and accuracy through our novel prompt-guided patching autofocus mechanism. We employ VLMs to generate context and reasoning descriptions of detected anomalies through curated prompt engineering. We have evaluated the accuracy and computational cost of XAVAD on several publicly available CCTV and video datasets. Our experiments demonstrate substantial gains of our XAVAD over state-of-the-art benchmark approaches and provide insights into privacy protection trade-offs in XAI for VAD. We further present ablation studies and assessments across diverse configurations and modeling alternatives to demonstrate the efficacy of our proposed methods. Code: https://github.com/muleina/XAVAD}{https://github.com/muleina/XAVAD .

0.38

LLMs Can Tackle High Energy Theoretical Physics, yet Gaps Remain: Part I, Quantum Field Theory

cs.CL hep-phhep-th ·2026-10-23 ·Tong Liu, Bohan Li, Gongrui Cheng, Zixiong Yu, Yulan He +1 more
benchmarkcode/data
Abstract

Beyond math and code, Large language models (LLMs) are beginning to tackle problems that have traditionally required substantial training for theoretical physicists, but it remains unclear how far this capability extends and where it might break. As a first step toward pushing LLMs in high-energy theoretical physics, we introduce QFT-Bench, a benchmark of 75 graduate- to research-level problems requiring extensive reasoning and derivation across core Quantum Field Theory (QFT) subareas. Each question has a well-defined target and closed-form reference answer. Evaluation combines symbolic verification with LLM-based judge. Across 39 proprietary and open-source models, frontier models already demonstrate substantial capabilities on nontrivial QFT calculations, with the strongest models exceeding 95% accuracy. At the same time, clear gaps remain. Even strong models still make errors in calculations, conventions, and instruction following, while weaker models often fail at the level of physical concepts or derivations. We further monitor models' CoTs and show that current models exhibit nontrivial reasoning unfaithfulness, i.e., they could reach correct answers but the reasoning itself does not justify and logically support them. These results provide a quantitative assessment of current LLM capabilities in QFT and a first step toward broader evaluation in high-energy theoretical physics.

0.38

An Error Budget for Shuffled Multimodal Windows in the 3rd WEAR Challenge: Stacking, Count Constraints and Seam Matching

cs.CV ·2026-10-23 ·Soheil Kooklan
benchmarkcode/dataunusual
Abstract · 12 pages, 10 tables, 1 figure. Technical report on a post-deadline system for the 3rd WEAR Dataset Challenge at HASCA 2026. Code: https://gi…

The 3rd WEAR Dataset Challenge asks for an activity label for every one-second window of four unseen participants. Each window has 15 VideoMAEv2 frame embeddings, the accelerometer signal of one random limb, and the windows are shuffled. This report is an empirical analysis of one system built after the challenge deadline. Window classifiers are stacked with features averaged over per-participant graphs of video neighbours and scored successor candidates, and a mean-field style decoder with a soft count constraint per activity labels all windows of a participant together. The setting is offline and transductive, and the count constraint relies on each activity being performed once. In a participant-wise simulation with participant-level bootstrap intervals, stacking added 0.051 macro F1 over stage-1 fusion and the count constraint 0.011 (95% interval 0.006 to 0.017). A quadratic seam distance ranks the true same-limb successor first among about 720 candidates in 83.9% of cases, but adding it to the link model changed macro F1 by -0.0001 (-0.0006 to 0.0005). With the decoder settings and the mixing weight chosen inside the folds and the mislabelled spans left out of training, the system scores 0.936 in simulation (0.940 for participants who performed each activity once); the submitted configuration scores 0.938 (0.943). Per-participant scores range from 0.80 to 0.97. These results come from a non-nested stacking protocol that uses unlabelled features transductively, so they are not a fully inductive estimate. An oracle that averages the final unary scores over the true segments reaches 0.973, and 79% of the remaining errors involve the null class. On the private leaderboard, which no choice used, the selected Kaggle submission scored 0.9231 (public 0.9254, with bounds partly tuned there), and a later one with bounds chosen in simulation scored 0.9247 (public 0.9253).

0.38

AutoMIL: Automated Multiple Instance Learning for Whole-Slide Image Classification

cs.CV cs.AI ·2026-10-23 ·Jonas Waibel, Frank Kramer, Dominik Müller
benchmarkcode/data
Abstract

Developing multiple instance learning pipelines for the classification of whole-slide images involves many interdependent design decisions and expertise in both pathology and machine learning. We present AutoMIL, an open-source framework that applies principles from the field of automated machine learning to tackle this task by automating integral pipeline steps, such as preprocessing, training and evaluation. In a benchmark evaluation with default settings, AutoMIL reached ROC-AUC scores of 0.94 on UBC-OCEAN, 0.99 on PCam and 0.93-0.96 on a TCGA-NSCLC subset, comparable to specialised pipelines in related work. Ensembling mainly improved performance on imbalanced data, showing that an automated pipeline must adapt to dataset characteristics. AutoMIL thus provides an accessible baseline for whole-slide image classification without manual pipeline configuration.

0.38

Tampered ML: Evaluating LLM Judgment of Machine Learning Research

cs.CL cs.AIcs.DLcs.LG ·2026-10-23 ·Soroush Ebadian, Parand A. Alamdari, Joe Benton, Hailey Joren
top venuebenchmark
Abstract · 28 pages, 10 figures. Accepted at EMNLP 2026 (Main Conference)

One approach to accelerating progress on AI alignment is to automate parts of alignment research itself. Doing so requires systems that can perform reliably on hard-to-supervise research tasks, including identifying flaws in research proposals and evaluating whether a proposed experiment supports its intended claim. To evaluate this capability, we introduce Tampered ML, a benchmark of 1,858 item-pairs, each consisting of an experiment proposal reconstructed from a published machine learning paper and a minimally edited counterfactual that introduces a targeted methodological flaw. Models must identify the more methodologically sound proposal in both presentation orders. A human validation study on a stratified sample estimates that 86% of items are valid (95% CI: 81-91%). Frontier models perform above chance but remain far from reliable (with Opus 5 achieving 77.9% accuracy). They are highly sensitive to the order of presentation and favor flawed designs that appear more specific, transparent, or conventional.

0.37

Efficient Zero-Shot Video Anomaly Detection via Signal-Aware Adaptive Fusion (SAAF)

cs.CV cs.AIcs.LG ·2026-10-23 ·Khaled Waleed Dawoud, Muhammad Haris Khan
publishedcode/data
Abstract · Accepted at WACV 2027

Deploying Vision-Language Models for video anomaly detection confronts a fundamental constraint. Their semantic reasoning demands computational budgets incompatible with the scale of real-world video streams. We present Signal-Aware Adaptive Fusion (SAAF), a framework that achieves efficient zero-shot anomaly detection through three key innovations. First, we introduce a dual-signal adaptive sampler operating in semantic embedding space: a delta signal captures sudden events via frame-to-frame cosine distance, while a drift signal detects static and gradual anomalies through deviation from rolling temporal context. Their OR-gated combination with adaptive EMA-based thresholding provides comprehensive anomaly coverage without training data. Second, we propose a threshold-gated cascaded architecture where lightweight frame-level scoring precedes expensive video-level temporal analysis, invoking it only when confidence warrants additional context. Third, we design a signal-aware weighted fusion mechanism that dynamically adapts score combination based on which signal triggered frame selection. Across four benchmarks, SAAF delivers competitive accuracy at substantially lower cost. Among training-free zero-shot methods, SAAF sets a new state of the art on XD-Violence (92.07%), MSAD (86.70%), and ShanghaiTech (80.54%) AUC-ROC, with XD-Violence and MSAD also surpassing supervised baselines. SAAF sustains 44.95, 48.61, and 41.06 FPS on XD-Violence, UCF-Crime, and MSAD, a 2.43$\times$, 2.63$\times$, and 2.22$\times$ throughput improvement over the prior best zero-shot method. Code: https://github.com/kdawoud91/SAAF

0.37

Video-MemComp: Taylor-Guided Sensitivity-Aware Memory Management for Streaming Video Understanding

cs.CV ·2026-10-23 ·Yonghui Niu, Yuhao Ji, Chang Nie, Jiao Zhang, Yilong Li +4 more
top venuecode/data
Abstract · Accepted by ACM MM 2026. Project Page: https://hhnk666.github.io/Video-MemComp.github.io

The rapid growth of video data has created an urgent need for efficient video understanding systems that can handle long and continuous visual inputs, posing significant challenges for Multimodal Large Language Models (MLLMs) due to severe memory bottlenecks. Moreover, existing compression methods often overlook the explosive accumulation of approximation errors. To address this issue, we propose Video-MemComp, a training-free and sensitivity-aware context management framework for memory-constrained streaming video understanding. Our method theoretically decomposes compression error into orthogonal local approximation error and layer-wise structural sensitivity via a first-order Taylor expansion. To minimize local approximation error, we perform token aggregation in a pre-RoPE safe zone with an attention-conserving logit bias, mitigating geometric distortions caused by RoPE rotations. To manage layer-wise structural sensitivity, we dynamically allocate the KV cache budget based on offline gradient priors and online context entropy, prioritizing high-sensitivity layers. Video-MemComp outperforms or achieves performance comparable to full-cache and state-of-the-art compression methods on benchmarks including MLVU, VideoMME, and EgoSchema. Notably, it improves the overall accuracy of Qwen2.5-VL-3B from 62.3% to 64.1% while reducing the KV cache to an equivalent of 6,000 tokens per layer, saving over 90% of KV memory.

0.37

PersonalQuery-Benchmark:Evaluating Retrieval Robustness to Personalized User Expression Differences

cs.IR ·2026-10-23 ·Wenyu Liang, Hongzhi Yin, Tongtong Wu, Fan Liu, Kexin Dai +1 more
benchmarkcode/data
Abstract

Existing personalized information retrieval methods mainly model differences in user needs through user preferences and query intent, while paying little attention to the personalized expression differences that arise when different users express the same retrieval need. Existing real query data typically lack stable user identities, long-term expression histories, and intent-matched counterparts, making it difficult to systematically evaluate the impact of personalized expression differences on retrieval outcomes. To this end, we propose PersonalQuery-Benchmark (PQB). PQB models personalized syntactic expression and personalized writing errors from users' historical texts, constructs personalized syntactic expression queries and personalized writing-error queries, and validates benchmark quality through LLM assessment and independent human evaluation. PQB covers three Amazon product domains with 6,516 query instances in total, and is evaluated on five retrieval models representing lexical, learned sparse, dense bi-encoder, late-interaction, and hybrid architectures, together with two LLM-based ranking models. Results show that even when the information need is held fixed, personalized expression differences still induce measurable ranking variation; different retrieval models exhibit clearly different sensitivities to syntactic variation and writing-error variation, while higher overall retrieval effectiveness does not necessarily correspond to stronger robustness against expression variation. PQB provides a new benchmark and evaluation framework for systematically studying personalized expression robustness. The associated code and dataset are available at the following address: https://github.com/cpxjazz123/PersonalQuery-Benchmark.

0.37

TileShare: Training-Free Acceleration of Overlapping Tiled ViT Decoding

cs.CV cs.LG ·2026-10-23 ·Xiangzhong Ye, Zhongshen Li, Shen You, Ziyang Meng, Liang Xu +4 more
Abstract · 8 pages main paper + supplementary material

Fixed-window Vision Transformer (ViT) decoders run at deployment resolution as overlapping tiles: every token in an overlap band is processed by several full decoder instances. In the 2.4B-parameter MiniMax-H3 video decoder, about half of all per-token computation is duplicated. Removing the tiles fails because attention, global tokens and coordinates are bound to them. TileShare keeps each released tile's attention and computes every other operation once per token. The released computation changes at one kind of step only: an overlap token's per-tile results are blended into one shared state with the released weights. Interventions show that this blend costs quality where attention reads the shared states, mainly in the first few layers, which TileShare therefore keeps per tile. Without training, keeping six layers per tile decodes 1.38-1.51x faster than batched and compiled tiling on two GPU types (up to 1.96x with FP8) and is equivalent to the released decoder in PSNR, LPIPS and DISTS on DAVIS videos. On unseen UVG videos with the setting frozen beforehand, it matches PSNR and DISTS and improves LPIPS. TileShare carries over to two further decoders and is again confirmed on unseen data. It is equivalent in PSNR and DISTS on MAGI-1. On the ViTok-v2 image autoencoder at 2K-4K, it matches overlapping tiling in PSNR, improves LPIPS and DISTS and decodes 1.5-1.8x faster. How many layers stay per tile is measured once per model (for decoders, without labels); we report it for 14 decoders and encoders. Optionally, two GPU-hours of self-distillation let MiniMax-H3 share every layer without a loss in PSNR, and a coordinate field trained while the network stays frozen brings fully shared CLIP-based open-vocabulary segmentation to parity with sliding windows on seven benchmarks.

0.36

Compression of Vision Foundation Models for Retinal OCT via Growing-Neural-Gas Codebooks and Graph-Guided Masked Image Modeling

cs.CV ·2026-10-23 ·Ali Yavari, Tilman Schmoll, Rainer A. Leitgeb, Wolfgang Drexler, Thomas Schlegl
benchmarkhot direction
Abstract · 31 pages, 2 figures, 8 tables; supplementary material provided as an ancillary file

Vision foundation models for retinal imaging are large and computationally demanding, limiting point-of-care deployment. We ask how much of such a model's downstream utility survives compression into a smaller network. We propose a topology-aware discretization framework that distills a continuous foundation model into a discrete codebook, then uses it as the reconstruction target for masked image modeling (MIM). It combines gradient-free codebook learning via Growing Neural Gas, which yields an adjacency graph over the codes, with a graph-distance soft cross-entropy loss that replaces one-hot MIM targets with topologically graded supervision. With student pretraining on 383 public Duke iAMD optical coherence tomography (OCT) volumes, the pipeline distills an 86M-parameter teacher into a 22M-parameter student with at least 3.6x fewer FLOPs. The student performs close to, and in some settings above, its teacher across three classification and two segmentation benchmarks, with the magnitude and direction of the difference depending on the teacher, task, metric, and student capacity. On the external OLIVES benchmark, it surpasses its MIRAGE teacher on every metric (98.06% vs. 95.22% balanced accuracy). It works with two Vision Transformer (ViT-Base) teachers, retinal (MIRAGE) and natural-image (DINOv3); with the out-of-domain teacher, the student outperforms it on retinal-layer segmentation, indicating that the added value of retina-specific pretraining is task-dependent: largest for segmentation, while a general-domain teacher is competitive for classification. These results yield compact retinal models potentially suitable for deployment, transferring across OCT datasets, devices, and clinical tasks, including three datasets not used in the student's pretraining.

0.36

Efficient Streaming Audio-Visual Target Speaker Extraction for Real-World Acoustic Scenes

cs.SD ·2026-10-23 ·Wendi Sang, Kai Li, Yifan Li, Jianqiang Huang, Xiaolin Hu
benchmarkcode/datacommunity pick
Abstract · 28 pages

Streaming audio-visual target speaker extraction (AVTSE) is essential for latency-sensitive applications. However, current causal systems face two coupled gaps. First, standard benchmarks fix the number of speakers at two and ignore the near-/far-field interference, reverberation, and sudden noise of real scenes. Second, high-performance causal architectures are too heavy for edge deployment. Lightweight alternatives try to reclaim capacity by repeatedly invoking a small separator, which offsets the savings. To address the first gap, we release RealSSA, a realistic AVTSE benchmark consisting of RealSSA-Sim and RealSSA-Real. RealSSA-Sim provides controllable simulated scenes with 1--5 dynamic near-field interfering speakers; scenes with far-field interference additionally include 10--15 far-field interfering speakers. RealSSA-Real provides a held-out real-recorded evaluation set. To address the second gap, we propose Falcon, a causal AVTSE method. Its Separator performs multi-scale time-frequency separation in a single encoder--decoder pass. We train Falcon with an encoder-space speaker contrastive loss. This loss suppresses near-field leakage at zero inference cost and transfers across backbones. Falcon achieves state-of-the-art extraction quality on RealSSA-Sim, LRS2-2Mix, LRS3-2Mix, and VoxCeleb2-2Mix. Compared with the causal AV-TFGridNet baseline, it cuts parameters by 91.9%, computation by 20.2 times, and GPU inference latency by 11.1 times. Code and demo are available at https://cslikai.cn/Falcon/.

0.36

ICASSP 2026 Radar Acoustic Speech Enhancement (RASE) Challenge

eess.AS ·2026-10-23 ·Zhi-Wei Tan, V. G. Reju, Ritesh C. Tewari, Ruotong Ding, Patrick A. Naylor +1 more
publishedbenchmarkdataset
Abstract · Published in IEEE ICASSP 2026. Author-accepted manuscript. Challenge resources and code available via the official project website

This paper presents the ICASSP 2026 Radar Acoustic Speech Enhancement (RASE) Grand Challenge, the first community challenge dedicated to benchmarking speech enhancement from millimeter-wave (mmWave) radar vibrometry. The challenge targets realistic through-glass sensing scenarios where microphones are ineffective or impractical. We introduce a curated dataset of paired radar and clean speech recordings, standardized preprocessing, a baseline system, and a secure Codabench-based evaluation framework with unified objective metrics. The RASE Challenge establishes the first public benchmark for radar-based speech enhancement and aims to foster reproducible research in this emerging field.

0.35

Automated identification of ivory carvings using machine learning: A proof-of-concept approach to detecting the illegal wildlife trade

cs.CV cs.CYcs.LG ·2026-10-23 ·Andrius Tamosiunas, Simone Mozzon, David Bacon, Caroline Cox, Alan Collins
Abstract · 16 pages, 7 figures and 2 tables. Submitted for publication in PLOS One

Despite a ban on the international trading of recently poached ivory, elephants are still killed for their tusks. Ivory and its products are mainly sold in online marketplaces. However, due to the huge volume and variety of adverts and the similarity of ivory with other materials that can be legally sold, it is hard for enforcement officials to identify illicit trading and prosecute sellers. Existing techniques for identifying elephant ivory, such as Fourier-transform infrared spectroscopy and microscopy, can reliably distinguish ivory from other materials, but require physical access to the item and cannot be applied to photographs alone. As a result, these methods are unsuitable for screening the large volume of images posted in online marketplaces, leaving a gap for automated, image-based identification tools. In this proof of concept work, we investigate the potential of machine learning techniques for image analysis to automatically distinguish ivory objects from bone. We employ a semantic segmentation framework using a U-Net architecture with a ResNet34 backbone, leveraging transfer learning on an augmented dataset of 541 original images. Using a dataset of images of ivory and items resembling carved ivory, several models were trialled. On a distinct test set, the classifier achieved an Area Under the Curve (AUC) of 0.81 and an overall classification accuracy of 81% on confident samples. The paper presents the best-performing model as a proof of concept for developing a system that can effectively identify ivory and its products when indicative features of elephant ivory are present in an image. This work provides initial steps towards a machine learning-based framework designed to advise law-enforcement in detecting illegal elephant ivory trade.

0.35

Real-Time Localisation of Dorsal Anatomical Regions on Dairy Cows in Commercial Rotary Parlours from RGB-D

cs.CV ·2026-10-23 ·Vineet Prasad, Shiron Melvin, Sheikh Azid, Hai Wang
publishedunusual
Abstract · v2: camera-ready version accepted to ACRA 2026. Adds confidence intervals over test cows, a control experiment on the region head, an unseen…

Automated monitoring of dairy cows increasingly measures one part of the body, such as the rump for temperature or the back for posture, yet the region itself is usually placed by hand. Existing systems either score the cow as a whole or detect sparse keypoints, often in controlled walkways, and neither yields the named region masks that such measurements need on a crowded commercial rotary platform. We present a perception front end that finds the target cow, ten bony landmarks and four named dorsal regions (withers, back, rump and tail-head) in one forward pass over a whole RGB-D frame. Colour and depth encoders are fused late and share one decoder with three heads. On an anatomical region dataset built on public rotary-parlour RGB-D images (538 annotated frames, cow-disjoint test set), the network (ConvNeXt-T backbone) reaches PCK@10 0.905 for landmarks and region IoU 0.70 to 0.91, and runs end to end in 26.4 ms per frame on a T4 GPU, against an 11 fps camera. Removing the region head costs 0.15 PCK@10 (ResNet-18), but a control shows that it mainly adds training signal: wider landmark heatmap targets recover the gain without it. Agreement between the two heads flags unreliable frames without labels. On our own rig at Western Australia's largest dairy farm, the ConvNeXt-T model transfers with ten labelled frames (PCK@10 0.85), and its regions are carried through depth into a radiometric thermal camera, giving one temperature per region and cow that is stable within a pass.

0.35

System One models Are Not Better Rerankers

cs.CL cs.IR ·2026-10-23 ·Michael Dinzinger, Kanishka Ghosh Dastidar, Michael Granitzer
Abstract

System One models, or decision models, answer a question about an input with a calibrated probability instead of generated text. Within weeks of the first release, in September 2026, they were used as rerankers. We test whether the decision fine-tuning behind them makes a better reranker than plain reranker fine-tuning of the same model. For six Qwen backbones from 0.6B to 27B parameters and ten TREC DL and BEIR collections, we compare the untuned model, asked zero-shot, with its decision fine-tunes and its reranker fine-tunes, two of which we train ourselves. In our work, we draw the conclusion that decision fine-tuning never makes the better reranker. From 0.6B to 9B, reranker fine-tuning wins on every backbone by 5 to 29 points nDCG@10, also on zero-shot collections, and at 27B the two tie. Up to 9B, decision fine-tuning adds nothing over simply asking the untuned model. Moreover, we find that the calibration advantage of decision models depends on the question. Asked whether a passage answers the query, they are better calibrated than the published reranker fine-tunes from 4B on. Asked a slightly different question from prior work, they lose this advantage. Lastly, calibration does not make up for a worse ranking: up to 9B, a reranker fine-tune keeps the better passages at a fixed threshold. If you need a reranker, fine-tune a reranker.

0.35

FlexGlue: Sparse-to-Dense Feature Matching and Visual Disambiguation

cs.CV ·2026-10-23 ·Philipp Lindenberger, Paul-Edouard Sarlin, Ondrej Miksik, Marc Pollefeys
code/data
Abstract

To estimate correspondences across images, existing approaches match either sparse interest points, which is highly efficient but limited by repeatability, or dense pixel grids, which is robust but expensive. Furthermore, these approaches are trained with only matching pairs and thus often fail to reject unrelated pairs that exhibit visual symmetries, which is equally important in applications like Structure-from-Motion (SfM). We bridge these gaps by introducing FlexGlue, a neural network for asymmetric sparse-to-dense matching and disambiguation. Given two images and coordinates of points in the first image, FlexGlue estimates the coordinates of the corresponding points in the second image while simultaneously rejecting the pair if non-matching. By searching for a limited set of correspondences among dense pixels, our approach is both fast and effective at both matching and disambiguation: it combines the sub-pixel accuracy of dense matching with the speed of sparse matching, while rejecting incorrect pairs with high precision. Experiments on multiple datasets show that FlexGlue pushes the Pareto frontier in relative pose estimation and significantly outperforms existing approaches for SfM and visual localization. FlexGlue is publicly available at https://github.com/cvg/flexglue .

0.35

SkillRep: Characterizing and Understanding the Representation Design Space of Agent Skill Retrieval

cs.IR ·2026-10-23 ·Tianyi Sun, Pengfei Yao, Yanze Wang, Hongda Chen, Yunyun Han +3 more
code/datahot direction
Abstract · 12 pages, 3 figures. Code: https://github.com/EverMind-AI/SkillCorpus/tree/main/skillrep

Retrieving the right skill from a large library is essential for LLM agents, yet requests and skills are written for different purposes and in different words. Existing retrievers narrow this request-skill gap with one or two representation choices each, so it is unknown whether their gains hold in combination. No prior work measures the gap itself, so it is also unclear which part of it each choice narrows. We define a representation design space by five choices -- encoder fine-tuning, query-side and document-side expansion, background projection and a lexical channel -- and evaluate them together for the first time, as modules of one retriever, SkillRep. With a 0.6B encoder and LLM expansion, SkillRep improves MRR@10 on SRA-Bench over the strongest same-size retriever by 7.02 points with the official skill pool and 8.79 with UniPool, a 111,247-skill pool we build from public registries, and is best on SkillsBench. The value of each choice depends on the others. Fine-tuning, for instance, reverses the ranking of the four training-free modules and turns them from complementary into overlapping. To understand why, we measure the gap with three quantities: background, the embedding component that does not separate skills; unsaid vocabulary, the gold skill's content a request leaves unsaid; and dilution, the information lost when a skill is encoded into a single vector. They show which part each module narrows and by what mechanism. This reveals that modules narrowing the same part compete for it, while those narrowing different parts complement each other.

0.35

LoCA: Enabling Multi-Modality in Geospatial Location Encoders

cs.CV cs.LG ·2026-10-23 ·Panagiotis Koromilas, Eleanna Vali, Diogenis Antonopoulos, Eleni Saka, Dimitris Christofis +5 more
theory
Abstract

Location encoders map geographic coordinates to high-dimensional descriptors through contrastive alignment with paired Earth Observation imagery. We prove that this construction is structurally limited: any Lipschitz location encoder produces an output of intrinsic dimension at most 2, regardless of architecture or bandwidth, which forces the contrastive loss to strictly exceed its asymptotic optimum. Empirically, joint training also collapses the encoder's embedding covariance to low rank. Although dimension-limited, current location encoders sidestep this spectral collapse by training against a frozen image encoder, which we reinterpret as a covariance regularizer that supplies a high-rank target for the location encoder to fold onto. However, extending this approach to multimodal inputs is challenging, as location encoders should simultaneously map a single location to multiple, potentially incompatible, descriptors. We identify two requirements a regularizer must satisfy for location encoding in multimodal regimes: co-alignment across modalities, since adapters alone cannot align independently pretrained encoders; and high effective rank, since the location encoder inherits its target's spectrum. We propose LoCA (Location Contrastive Alignment), a two-stage pipeline that satisfies both. Stage 1 jointly trains Sentinel-1 and Sentinel-2 encoders with contrastive learning, producing a co-aligned, high-rank multimodal target. Stage 2 freezes the image encoders and trains a location and a tabular metadata encoder as projection adapters with multi-view contrastive learning. Against five state-of-the-art encoders on 17 tasks ranging from land cover and wildfire forecasting to geophysical, socio-economic, biodiversity and climate applications, LoCA achieves the most task wins, is the only model to outperform every other on a majority of tasks, and wins 68.8% of all per-task comparisons.

0.35

Concentrated Causal Influence, Distributed Diagnostic Information: Cognitive Impairment Decisions in Large Audio-Language Models

eess.AS cs.CL ·2026-10-23 ·Xiutian Zhao, Luqi Sun, Björn Schuller, Berrak Sisman
hot direction
Abstract

Large audio-language models (LALMs) can assess cognitive impairment (CI) from speech, yet their answers reveal neither which internal components control the decision nor how much CI information is accessible in their representations. We investigate both questions through CI-contrast neurons (CICNs), selected by class-contrastive activation frequency in decoder MLPs. Across eight open LALMs and two corpora (Pitt and PROCESS-2), we causally test CICNs against size- and layer-matched random masks and activity-matched controls. In three of the five models that meet our eligibility criteria, ablation of small CICN sets causes large, class-directed prediction shifts; speech-identified masks also control transcript-only decisions, and three of four primary Pitt masks retain specificity on PROCESS-2. Position-restricted interventions locate the influence before the final answer position, most strongly after the speech or transcript evidence under letter-labeled options, largely as a decision-direction shift. Beyond causal localization, supervised linear readouts of frozen CICN activations achieve 0.92--0.97 AUC within Pitt, versus 0.67--0.84 for the models' direct answers. Matched random and all-neuron readouts reach comparable performance; top-ranked CICNs show only an exploratory advantage at very small unit budgets. These results distinguish concentrated causal control over output decisions from broadly distributed diagnostic information for CI detection in LALMs.

0.35

Support-guided OOD Removal with Diffusion

cs.CV cs.LGcs.SD ·2026-10-23 ·Alireza Kheirandish, Sara Fridovich-Keil
code/data
Abstract

Many real-world signals are corrupted by localized, structured perturbations whose distribution is unknown: an intermittent background noise in a music recording, a watermark stamped on a photograph, a cast shadow. We study the task of recovering a clean signal $x$ from an observation $y$ containing such a localized out-of-distribution (OOD) corruption. We propose Support-guided OOD Removal with Diffusion, SORD, an iterative algorithm to recover $x$ using only two ingredients: a binary map of the corruption support and a single diffusion model pretrained on clean signals. SORD does not require any paired data or any statistical model of the corruption. We evaluate SORD on natural images and human faces with spatially localized corruptions including handwritten digits, other images, watermarks, and shadows, as well as audio recordings of different musical instruments with temporally localized corruptions including other instruments and human speech. SORD is competitive with, and often outperforms, posterior-sampling inpainting given the same diffusion prior and the same mask, source-separation baselines that require a separate diffusion model for each source, and, on several benchmarks, supervised networks trained on paired clean and corrupted signals. The code is released here: \href{https://github.com/voilalab/SORD}{https://github.com/voilalab/SORD}.

0.34

EDWARD-NMT: Elastic Dimensions With Adaptive Routing and Dual-consistency for Neural Machine Translation

cs.CL ·2026-10-23 ·Edward Nguyen, Thai Tran, Kiet Anh Hoang, Duy Hoang, Duy Vo +2 more
top venuebenchmark
Abstract · Accepted to EMNLP 2026 (Findings)

Sparse Mixture-of-Experts (MoE) is a powerful paradigm for scaling Neural Machine Translation (NMT) and managing language heterogeneity. However, practical limitations hinder its deployment on resource-constrained devices: the entire expert pool must remain in memory, and the per-token compute budget is statically fixed during training by rigid top-k routing. We introduce EDWARD-NMT (Elastic Dimensions With Adaptive Routing and Dual-consistency for NMT), a novel MoE encoder-decoder architecture designed for a train once, deploy anywhere paradigm that exposes a family of deployable sub-networks at inference time. EDWARD-NMT integrates three key innovations: (i) Matryoshka-style elastic dimensions to flexibly reduce the memory footprint; (ii) an adaptive sigmoid router with a learned, task-conditioned threshold, allowing each token to dynamically determine its active expert count; and (iii) a dual-consistency objective to maximize the performance of the extracted sub-networks. Extensive experiments on PhoMT (En-Vi), WMT-14 (En-De), multilingual OPUS-16, and a multi-domain De-En benchmark validate our approach. The full-capacity architecture matches or exceeds strong MoE baselines with improved parameter efficiency. Furthermore, the sub-networks seamlessly extracted from a single checkpoint consistently outperform comparably sized dense Transformers without any retraining.

0.34

How Much Do You Trust A Dog? Quantifying Hint-Following in Reasoning Language Models

cs.CL cs.LG ·2026-10-23 ·Anxin Guo, Chirag Pabbaraju, Sara Ahmadian, Kamesh Munagala, Aravindan Vijayaraghavan
code/data
Abstract

Large language models (LLMs) often exhibit sycophancy, altering their answers to align with user suggestions even on objective factual and reasoning tasks. Yet the quantitative form of this influence--- and whether it is calibrated to source reliability--- remains poorly understood. In this work, we investigate this behavior through hint-induced answer shifts on MMLU questions, evaluating prompts of the form ``[Source] thinks the answer is (X)''. We systematically vary the source to which the hint is attributed (a professor, a friend, or a dog) across three open-source model families, Qwen, Gemma, and SmolLM. We show that the change induced by the hint follows a predictable structure well-characterized by an exponential tilt: the hint induces an approximately constant mean shift in the hinted answer's log-odds, with question--target susceptibility largely shared across sources. To explore how this pattern can arise, we introduce a formal decomposition of the hint response and illustrate in controlled synthetic experiments how KL-regularized reward-based training reshapes both hinted and unhinted answer distributions. Finally, we evaluate whether this shift is calibrated to source reliability, finding that the log-odds shift persists even when the hint-giver is declared inaccurate or accompanied by explicit warnings not to trust the source, highlighting the vulnerability of LLMs to adversarial steering.

0.34

X-Dubbing: Speech-to-Speech Translation under Temporal and Visual Constraints

cs.CL cs.CVeess.AS ·2026-10-23 ·Jeongsoo Choi, Ji-Hoon Kim, Junseok Ahn, Joon Son Chung
Abstract

Cross-lingual dubbing aims to convert source speech into target-language speech that remains synchronized with the original video. The key to this task is to produce a faithful translation that is naturally speakable within the source timing, which is non-trivial because languages differ in duration, rhythm, and articulation. In this paper, we propose X-Dubbing, which formulates the task as a joint problem of time-aware speech-to-text translation and video-aware text-to-speech. Specifically, our method consists of a Dubbing Translator that generates temporally structured scripts and a Voice Actor that renders them in synchrony with the source video, both grounded in a language-agnostic timing budget. We further refine the translator with dubbing-aware feedback from the synthesized speech, encouraging scripts that yield well-synchronized dubbing. This design requires no paired dubbing data, as the two components are trained on speech-text and monolingual audio-visual data respectively. Extensive experiments show that X-Dubbing outperforms recent baselines across diverse metrics, including isochrony and audio-visual correspondence.

0.34

K-CLUE: A Segmentation-to-Clustering Framework for Ki67 Cell Typing

cs.CV ·2026-10-23 ·Ana Beatriz Vieira, Sara P. Oliveira, Diana Montezuma, João Fraga, Jaime S. Cardoso +1 more
published
Abstract · Accepted at the BD-MedVis Workshop, BMVC 2026

Ki67 is a key marker of tumor proliferation in breast cancer, but its quantification from immunohistochemistry images normally depends on supervised classifiers trained per dataset. We examine the role of frozen pretrained visual representations in annotation-free Ki67 cell typing. To this end, we use K-CLUE, a framework that segments individual nuclei, represents each nucleus with a pretrained encoder, reduces the resulting embeddings with principal component analysis, and separates Ki67-positive from Ki67-negative nuclei using k-means clustering, without task-specific training or manual annotations. Clustering the raw components is highly sensitive to the choice of encoder: across seven encoders, the Ki67-positive F1-score on matched MIST nuclei varies by 44 points, although Ki67 status remains linearly decodable from every representation, with supervised probe AUCs above 0.98. Analysis of the embedding spaces shows that clustering succeeds when DAB-related variation occupies the leading directions of variance. Whitening the retained principal components removes this dependence, reducing the spread across encoders to 1.00 F1-score point. Handcrafted color representations are also competitive, confirming that Ki67 status is strongly reflected in chromatic information, but they require an explicit choice of color space. In end-to-end evaluation, K-CLUE reaches F1-scores of 85.00% and 81.91% for Ki67-positive and Ki67-negative nuclei on MIST, and 79.59% and 81.16% on BCData, outperforming DeepLIIF and IHCScoreGAN on both datasets. These results show that pretrained representations can support annotation-free Ki67 typing when their variance structure is accounted for.

0.33

World-in-Context: Interactive Video World Modeling with In-Context Conditioning

cs.CV ·2026-10-23 ·Harold Haodong Chen, Rui Peng, Rongjin Guo, Lingteng Qiu, Kejie Qiu +2 more
community pick
Abstract · Project: https://haroldchen19.github.io/WiC-site/

Interactive video world models require coordinated control over viewing perspectives, character actions, and camera motion. Existing methods typically leave view modes implicit in text captions, conflate action with camera motion, and inject controls through fragmented architectural pathways that capture interdependencies only indirectly. In this work, we present World-in-Context, an interactive world modeling framework powered by In-Context Conditioning (ICC). Rather than relying on specialized conditioning pathways, World-in-Context unifies view, action, and camera controls into explicit tokens contextualized directly within a shared self-attention space. Our formulation reveals that (i) explicit view tokens disentangle conflicting dynamics and unlock emergent view switching from single-view data, while (ii) action and camera inputs play complementary roles in governing local responsiveness and global stability. To preserve temporal continuity across execution intervals, we further introduce Interaction-Transport RoPE (IT-RoPE), parameterizing temporal attention with accumulated control transitions. On WBench, World-in-Context-5B achieves navigation performance on par with 22B EchoWM ($81.5$ vs. $81.6$) and attains a $93.8\%$ perspective-switching success rate, substantially outperforming existing 5B baselines ($18.8\%$). Controlled analyses further demonstrate compounding advantages of World-in-Context under richer conditioning mixtures, favorable data efficiency, and natural applicability to causal rollouts.

0.33

Ampe-DB: A Dataset for Human Activity and Interaction Analysis Using the Ghanaian Game Ampe

cs.CV cs.LG ·2026-10-23 ·Ruth Senior Biney, Isaac Kojo Yedonu Aboah
benchmarkdatasetcode/data
Abstract · 74 pages, 9 figures, 7 numbered tables. Ampe-DB V1 dataset (637 interaction clips) available at https://huggingface.co/datasets/RuthBiney/Am…

Human activity recognition benchmarks underrepresent culturally situated forms of movement and interaction. We introduce Ampe-DB, a multimodal dataset based on Ampe, a traditional Ghanaian game involving coordinated jumping, leg movements, and two-person interaction. It contains 637 canonical clips from 15 participants, with RGB video, two-person pose representations, participant and pair metadata, four-class movement labels, and derived binary movement-relation labels. We evaluate four-class movement recognition and binary movement-relation classification using three protocols: a duplicate-safe fixed benchmark, four-fold pair-held-out evaluation, and participant-held-out diagnostics for participants with complete class coverage. Baselines comprise a frozen Kinetics-400-pretrained R(2+1)D-18 RGB encoder, a two-layer bidirectional GRU for pose sequences, and validation-selected RGB-pose late fusion. For four-class recognition, RGB macro-F1 falls from 0.955 on the fixed benchmark to 0.418 under pair-held-out evaluation and 0.384 under participant-held-out evaluation. Pose is more robust, achieving 0.957, 0.811, and 0.839 across the same protocols. Fusion reaches 0.976 on the fixed benchmark and 0.824 under pair-held-out evaluation, but does not consistently improve on pose under participant shift. Binary movement-relation recognition also becomes substantially harder under held-out evaluation, especially for different-leg configurations. These findings show that strong fixed-benchmark scores can mask failures to generalize across people and pairings. Ampe-DB supports research on fine-grained movement recognition, human-human interaction, multimodal learning, and identity-aware generalization.

0.33

The Answer Knows Where to Look: Zooming before MLLMs Answer under Visual-Token Budgets

cs.CV ·2026-10-23 ·Zhensheng Hu, Guanglong Xu, Guanqiang Hu
Abstract

Multimodal large language models (MLLMs) read images through visual tokens, and under the small per-image budgets that efficient deployment requires, the fine text of documents, charts and infographics becomes illegible. Zooming in recovers the evidence, but the region must be chosen before the model answers. We find that MLLMs know where the evidence is, but not in a form that can be read before they answer: the answer tokens localize it from the middle layers on, whereas the attention available before answering does so only in the top layers, at a depth that varies across models, and diffusely. Supervising the pre-answer attention to point degrades reading, and the benefit of zooming is governed by each image's information gap. Building on these findings, we introduce ours, which lets an MLLM look where its answer would. A dedicated look token, processed in a single prefill and kept out of the answering pass, learns in hindsight to attend as the model's own correct answers do, extracting this knowledge into a concentrated pointer before any answer exists. Its hidden state then selects how to look, from no zoom to crops, square windows, windows that adapt to the pointer and foveated warps, supervised by the counterfactual outcome of every view. ours needs no box annotations, per-model layer search, external localizers or generation before the answer. On Qwen3-VL at three scales and five text-rich benchmarks, it surpasses the strongest baseline by 4.8-9.1 points at 98 visual tokens per view and attention-guided zooming by 11.1-17.0 points, while running faster than attention-guided and grounding-based zooming.

0.33

Interactive image diagnosis using test-time backpropagation from the auxiliary task

cs.CV physics.med-ph ·2026-10-23 ·Takuro Shimaya
published
Abstract · 8 pages, accepted to SPIE Medical Imaging 2025

Multi-task deep learning has been developed to ensure explainability in computer-aided diagnosis (CAD) systems. These systems simultaneously predict interpretable characteristics of findings in auxiliary tasks to help radiologists assess the reliability of the diagnostic prediction. However, the CAD's opinion would not be referred to when the auxiliary task prediction deviates from the radiologist's expectation. To further leverage CAD's knowledge in the above situation, we proposed a multi-task gradient assist (MGA) method that brings interactivity to multi-task models without additional training. MGA allows radiologists to give feedback to the CAD and ask for additional diagnostic suggestions. Using two datasets of mammography and CT, we show that MGA successfully improves diagnostic accuracy upon feedback on various auxiliary tasks. Compared to an existing method using forward propagation only, MGA performs more robustly and can be combined to further improve accuracy. Our method paves a promising way for integrating artificial intelligence into image diagnosis.

0.33

The Punjabi Geet Toxicity Dataset (PGT-FG): Fine-Grained Multimodal Annotations for Toxic Content in Punjabi Songs

cs.CL ·2026-10-23 ·Lakhjeet Kaur, Jaspreet Kaur, Komal Rani, Yougal Attri, Ramakant KumarJaydeep Kishore +1 more
datasetcode/dataunusual
Abstract

Resources for toxicity detection in Punjabi remain limited, particularly for musical content that combines linguistic and multimodal information. We present the Punjabi Geet Toxicity Dataset (PGT-FG), a segment-level resource developed to support research on toxicity detection in Punjabi musical content. The dataset was constructed from publicly available Punjabi songs collected from YouTube. An initial collection of 620 songs resulted in 465 successfully downloaded songs and 3,402 preliminary segments from 450 songs. After quality control and annotation, the final dataset contains 2,396 unique segments from 392 songs. Four native Punjabi annotators produced 2,870 annotation records, including a shared subset of 158 segments annotated by all four. Each annotation includes binary toxicity, eight fine-grained toxicity categories, severity, sentiment, mode, and an uncertainty flag. Inter-annotator agreement was measured with Fleiss' kappa on the shared subset, and majority-based resolution produced the final segment-level labels. Of the 2,396 segments, 705 (29.42%) were labelled toxic and 1,691 (70.58%) non-toxic. Violence was the most frequent category, occurring in 539 toxic segments (76.45%), followed by gang references (227, 32.20%) and drug references (133, 18.87%). The overall toxicity label achieved substantial agreement (kappa = 0.6521). PGT-FG provides a fine-grained resource for Punjabi toxicity detection and multimodal content analysis. The resolved dataset, with one final record per segment, is publicly available on Zenodo (https://doi.org/10.5281/zenodo.23040156). Original lyrics, lyric excerpts, and copyrighted audio or video content are not redistributed.

0.32

An exploration of plug-and-play flow matching for full-waveform inversion on breast imaging

eess.IV math.NA ·2026-10-23 ·Cristina Garcia-Cardona, Siting Liu, Fatma Terzioglu, Shanyin Tong, Lu Zhang
Abstract

Full-waveform inversion (FWI) reconstructs sound-speed fields from measured waveforms, by solving an optimization problem that minimizes the discrepancy between measured waveforms and those predicted by a numerical wave-propagation model. However, choosing suitable regularization for this optimization is challenging because it must stabilize the inversion while preserving relevant structures, prompting us to investigate whether priors learned from representative sound-speed images can improve reconstruction. In this work, we adapt plug-and-play flow matching (PnP-Flow) to FWI by alternating waveform-misfit updates with updates from a pretrained flow prior. We explore the effectiveness of PnP-Flow reconstruction by comparing it with total-variation (TV) regularization using noiseless transmission-only plane-wave data. Separate flow priors are trained on two circular-inclusion model families with constant and spatially varying backgrounds. We evaluate the method both on these in-distribution circle models and under substantial distribution shift by transferring the Gaussian-circle prior and reconstruction settings to breast phantom patches without retraining or retuning. Using the same number of waveform-misfit updates, the proposed PnP-Flow framework generally improves circle reconstruction accuracy and recovers breast lesion contrast more accurately than TV baselines.

0.32

ReLEAP: Upcycling Vision Datasets through Cross-Task Unification and Semi-Supervised Annotation

cs.CV cs.LG ·2026-10-23 ·Julia Simon, Daniel Steininger, Andreas Trondl
publishedcode/data
Abstract · Accepted at Asian Conference on Computer Vision (ACCV) 2026. Code and models available at https://github.com/releap-vision/releap

Developing robust vision models requires large amounts of annotated data, especially for tasks involving fine-grained class distinctions under variable environmental conditions. Even in domains with numerous existing datasets, differences in annotation formats, label definitions and targeted learning tasks hinder data reuse and integration. To address this challenge, we introduce ReLEAP, a scalable Python framework for unifying heterogeneous image datasets through data-source abstraction, systematic reuse of annotations across learning tasks, and mapping of labels to a shared hierarchical taxonomy. We demonstrate its applicability in precision agriculture by integrating fifteen diverse datasets covering a broad range of plant species and annotation types. Beyond unification, ReLEAP enables systematic recycling of existing data and provides a robust foundation for learning-based annotation upcycling. Leveraging this foundation, we conduct extensive cross-dataset and fine-tuning experiments to investigate how domain-specific models trained on unified data and foundation models can be used to automatically generate higher-level annotations under unsupervised, weakly supervised and semi-supervised settings, and how dataset characteristics influence their effectiveness. Our results show that unified data access and annotation upcycling substantially expand the pool of reusable training data and annotations while improving data transparency, enabling more sustainable and resource-efficient use of existing datasets. We release the framework and trained models for academic use to support accessible and collaborative research, particularly in data-scarce and heterogeneous domains.

0.32

Many Perturbations, Shared Failure Dynamics: Tracing Direction-Sensitive Perturbation Influence in Language Models

cs.CL ·2026-10-23 ·Zhibo Hu, Chen Wang, Yanfeng Shu, Liming Dong, Hye-young Paik +1 more
Abstract · 42 pages, 6 figures

Why can a task-preserving rewrite turn a correct answer into a failure while a larger internal change leaves the answer intact? We develop a direction-sensitive, readout-relative account of semantic fragility: heterogeneous language changes can share geometric and readout relationships without sharing an input-independent failure direction. Clean decision reserve, representation transformation, and response alignment provide complementary descriptions of this process. An exact longitudinal-transverse decomposition identifies directional information discarded by scalar transmission. A compact reserve-and-transmission model predicts failures across held-out families and items in Qwen3-32B and Gemma 4 31B, with useful but parameterization-dependent transfer to ReClor. Matched interventions show that orientation carries causal information beyond scalar geometry. Among naturally failed rewrites, attention-component tests on disjoint original items and subsequent answer-order controls favor the natural paired direction over tested random and same-answer-letter donor directions. Yet failures rarely persist across option orders, and current-order corrections use recomputed clean references. Shared predictive regularity thus coexists with context-dependent directional effects. Numerical path integration reduces, but does not eliminate, finite-patch approximation error within a fixed context; a larger discrepancy from the natural rewrite effect remains. A complementary fixed-CoT cue study extends this account to influence unacknowledged in recorded reasoning: readout associations transfer across domains, although the positive-mean causal criterion fails and signed-effect prediction remains post hoc. Together, these results support shared relations among reserve, transformation, and response while distinguishing local causal sensitivity from a complete explanation of natural failure.

0.32

Reuse, Don't Retrain: Efficient, Modality-Versatile Geospatial Foundation Models through Feature Distillation

cs.CV ·2026-10-23 ·Mario Hamon, Pierre Adorni, Minh-Tan Pham
hot direction
Abstract

Geospatial foundation models (GeoFMs) are costly to pretrain, yet model families often repeat this process across backbone scales and sensing modalities. We investigate feature distillation as a more efficient way to reuse existing representations. First, we distill frozen CROMA and TerraMind teachers into smaller Vision Transformers through patch-level feature matching. Across nine PANGAEA tasks, distilled TerraMind Tiny, Small, and Base outperform their scale-matched released counterparts by 2.76, 3.02, and 1.87 mean IoU points, respectively; each also matches or exceeds the next-larger released model. Distilled CROMA-Base gains 2.05 points. TerraMind-Base required only 43 A6000 GPU-hours, approximately 0.5\% of the reported cost of independent pretraining. Second, we introduce TEDI, an unpaired multi-expert approach combining TerraMind for multispectral imagery and DINOv3 for high-resolution RGB. Separate patch embeddings route unpaired inputs into a shared Transformer trunk, removing the need for co-registered data. TEDI-Large achieves the highest single-model mean in our PANGAEA-9 comparison (62.51 mIoU), outperforming both teachers overall while retaining their modality-specific strengths. TEDI-Base reaches 61.75 mIoU with 72\% fewer parameters. These results show that distillation improves the GeoFM accuracy-compute frontier while consolidating complementary experts into a single versatile backbone.

0.32

Past as State, Present as Attention: Persistent-State Blockwise Flow Matching for Long-Horizon Co-Speech Motion Generation

cs.CV cs.GRcs.MM ·2026-10-23 ·Yi Liu, Xiangyue Zhang, Jianfang Li, Jia Ma, Yisheng He +1 more
top venuebenchmark
Abstract · Accepted at NeurIPS 2026. 38 pages, 10 figures, 17 tables

Co-speech motion generation remains challenging over long horizons due to the need for temporal coherence and consistent style. Existing methods typically generate long motions segment by segment, either conditioning on a short seed of recent frames, which gradually drifts, or retaining ever-longer explicit histories, which are costly to maintain and hard to learn from limited data. We trace these limitations to a key asymmetry between past and present: past motion carries slowly varying style that should be compactly summarized, whereas the current segment exhibits fast-changing local dynamics that require fine-grained modeling. This asymmetry motivates a simple Past-As-State, Present-as-Attention principle for long-horizon co-speech motion generation. Following this principle, we propose PASPA, which summarizes past history with an inter-block persistent state and models the current segment with intra-block bidirectional self-attention. We instantiate PASPA within a blockwise flow matching framework, with multi-block supervision for effective training over variable histories, prefill-decode state caching for efficient rollout, and hybrid classifier-free guidance for enhanced history and speech conditioning. Extensive public-benchmark experiments show that PASPA achieves state-of-the-art FGD performance and preserves motion style across long-sequence generation.

0.32

WorldCast: World-Consistent Character Video Generation with Large-Angle Camera Control

cs.CV ·2026-10-23 ·Lei Cao, Yu Liu, Wenhua Liao, Peng Zhang, Fei Wang +5 more
benchmarkdatasetcommunity pick
Abstract · 30 pages, 14 figures, 3 tables

Precise camera control is essential for cinematic character video generation, yet existing methods struggle to follow large-angle trajectories with combined rotation, translation, and radius variation from arbitrary initial azimuths. Such viewpoint changes also reveal character regions unseen in the initial frame, causing geometric distortions, appearance degradation, and identity drift. To systematically study these coupled challenges, we introduce CLACT, a large-scale Character-centric dataset and benchmark covering diverse Large-Angle Camera Trajectories, with accurate camera geometry and multi-view character observations for training and evaluation. We further propose WorldCast, a framework for world-consistent character video generation under diverse large-angle camera motions. Specifically, we back-project the initial frame into a 3D point cloud using estimated depth, align the point-cloud scale with the target camera trajectory, and render it along the trajectory to obtain explicit geometric guidance and coverage masks. To complement the incomplete character information, we introduce Geometry-Aligned Memory Attention (GAMA). Using relative camera poses estimated jointly from the initial frame and multi-view character anchors, we express anchor and target-view geometry in a common coordinate frame. GAMA uses this geometric alignment to guide the retrieval and injection of view-relevant character features during generation. Evaluations across diverse large-angle camera trajectories demonstrate accurate trajectory following while preserving character geometry, appearance, and identity.

0.32

Hey, guess what: CUES lead to inferences about typicality

cs.CL ·2026-10-23 ·Alison Chi, Sanghyeok Choi, Hannah Rohde, Emily Allaway
Abstract

Throughout life, we accumulate knowledge of how typical certain events are. We use this knowledge in daily conversation, often interpreting meaning beyond what has been said explicitly. In this work, we take steps to better understand this meaning recovery, investigating both atypicality and typicality inferences. First, we construct two corpora, EventTyp and the human-validated EventTypGold. We conduct large-scale human and LLM experiments and find that for humans, three CUES predict inferences about typicality: Commonsense Knowledge, Utterance Emphasis, and Snark. However, in most experimental settings, we find that LLMs lack the prerequisite Commonsense Knowledge necessary to accurately assess an event's general typicality, and therefore struggle to follow these CUES.

0.32

Beyond Discrete Tokens: Preserving Numerical Continuity in Medical Vision-Language Models

cs.CV cs.AI ·2026-10-23 ·Dianye Huang, Jiamei Guo, Yue Zhou, Mingde Li, Yi Xiong +3 more
Abstract

Objective: To develop a unified vision-language model for prompt-conditioned quantitative ultrasound measurement that addresses the mismatch between continuous measurement targets and discrete token-based numerical prediction. Methods: This article presents HAD-VLM, a hybrid autoregressive-diffusion framework that combines autoregressive text generation with continuous measurement prediction within a shared backbone. Visual prompt tuning adapts the vision encoder to ultrasound images. A diffusion process predicts a five-component representation comprising caliper endpoint coordinates and measurement length. A post-processing strategy combining the proposed coordinate-length consistency mechanism with ensemble prediction is introduced to refine measurement estimates. We evaluate four echocardiographic measurement tasks on EchoNet-LVH, with additional assessments of medical question answering and X-ray vertebral measurement. Results: HAD-VLM achieves mean absolute errors of 1.36, 2.77, 3.05, and 1.16 mm for IVSd, LVIDd, LVIDs, and LVPWd, respectively, reducing errors by 21.4-37.6\% relative to the Vicuna-based autoregressive baseline. It also outperforms fine-tuned Qwen3-VL and direct MLP regression across all four tasks. Medical question answering performance is preserved on the evaluated dataset. For X-ray vertebral measurement, HAD-VLM reduces mean absolute error from 11.4 to 6.7 pixels. Conclusion: Integrating autoregressive language generation with diffusion-based continuous prediction improves quantitative medical image measurement while retaining prompt-conditioned task flexibility. Significance: HAD-VLM provides a unified approach to combining language interaction with anatomically grounded numerical prediction for quantitative medical image analysis.

0.32

Communicating with the Entire Population: Machine Translation of Emergency Alerts and Warnings

cs.CL ·2026-10-23 ·Jen Wilson, Vipasha Bansal, William D. Lewis
publishedsurvey
Abstract · In Proceedings of the Language Technologies for Crisis Preparedness and Response (LT4CPR Workshop), Hengqin, Zhuhai, China. Association for …

It is essential that we communicate with the entirety of an affected population during a crisis, including those who do not speak a region's dominant language(s). Commercial Translation tools offer a tempting solution to broaden access and can provide a lifeline when no other resources are available. However, Emergency response managers often operate in the dark regarding the quality and adequacy of these automated outputs. In this paper, we conduct a nationwide survey of emergency managers to understand their current workflows, perspectives, and operational challenges. We then evaluate commercial translation tools against these needs, shedding light on their utility and offering practical guidance on how and when these tools can be safely used in emergency response.

0.31

CompMVR: Compression-Aware Multi-View Restoration Using Diffusion Models for Geometrically Consistent 3D Reconstruction

cs.CV ·2026-10-23 ·Dong-hwi Kim, Chaewon Moon, Hojun Song, Dongbeom Kim, Junyeong Jang +11 more
top venuecode/data
Abstract · Accepted to SIGGRAPH Asia 2026 (Technical paper)

Real-world 3D reconstruction pipelines often rely on multi-view images degraded by lossy compression. While compression artifacts have been extensively studied in single-image or video restoration, their impact on multi-view geometry remains largely overlooked. In this work, we identify compression-induced cross-view geometric inconsistency as a critical failure mode, where view-dependent artifacts disrupt feature matching, dense correspondence, camera parameter estimation, and downstream 3D reconstruction tasks such as novel view synthesis. To address this problem, we propose CompMVR, a compression-aware multi-view diffusion restoration framework that jointly refines compressed views while preserving cross-view consistency, and construct compressed datasets for systematic evaluation. Unlike conventional restoration methods that optimize each image independently for 2D perceptual quality, CompMVR leverages learned compression priors and cross-view correspondence to recover visual details while improving geometric reliability across views. Experiments across diverse datasets, codecs, and compression levels demonstrate gains in restoration quality and downstream 3D tasks, including novel-view synthesis, camera pose estimation, and view matching. These results establish compressed multi-view restoration as a distinct problem and highlight its importance for 3D reconstruction under lossy compression. Code and data for this paper are available at https://mchaewon.github.io/CompMVR-project-page/.

0.31

Shorter Is Not Cheaper: A Controlled Comparison of Declarative Animation Specifications and Free-Form Code for LLM-Generated Explanations

cs.CL cs.HCcs.SE ·2026-10-23 ·Jiantang Huang
Abstract · 15 pages, 6 figures, 8 tables

Chat assistants increasingly answer conceptual questions with interactive, animated explanations, generated either as free-form HTML/JavaScript or as declarative specifications that a fixed runtime plays. The declarative route is widely expected to be cheaper, faster and more reliable, because its outputs are shorter and constrained. We test this expectation in controlled comparisons whose hypotheses and analysis plans were frozen before the runs. We designed ExplainSpec, a time-indexed declarative animation format, and an observation contract that makes every generated explanation programmatically inspectable; validated, pacing-agnostic checkers score 20 explanation requests from five domains. Three Claude models (360 jobs) and, in a follow-up, the open-weight Qwen3-14B with thinking enabled and disabled (240 jobs) answered every request in both formats. For every model the specifications were 25-48% shorter in answer tokens (output excluding reasoning), yet with reasoning enabled total output fell significantly only for the largest Claude model and no model generated significantly faster. The reasons differed. The Claude models reasoned far more before writing specifications (exploratory; for the two larger models, from 15-20% to 85-88% of generations). Qwen3-14B reasoned about as much in both formats (a pre-registered test of this shift was not supported), and its shorter answers saved little against a reasoning trace more than twice as long. Nor were specifications more reliable: free-form pages were valid after one repair round in 95-100% of jobs for every model, whereas Qwen3-14B produced valid specifications in only 33% of jobs and passed all semantic tests in 17% (HTML/JavaScript: 62%). For reasoning models, a shorter format alone therefore does not make generation cheaper, faster or more reliable, and a format learned in context can cost reliability.

0.31

Beyond Final Outcomes: Quantifying Claim-Level Information Dynamics in Multi-Agent LLM Interaction

cs.CL ·2026-10-23 ·Haocheng Jiang, Herun Wan, Yongyuan Peng, Minnan Luo
code/data
Abstract · 23 pages, 3 figures. Code and data: https://github.com/fzj-2023/RIVER

When LLM-powered agents interact, information is repeatedly shared, expanded, and forgotten, yet multi-agent evaluation typically focuses on final task outcomes. We study such interaction as claim-level information dynamics and propose RIVER, which decomposes agent outputs into atomic claims and evaluates three complementary dimensions: convergence, exploration, and retention. Convergence captures cross-agent claim correspondence and stance alignment; exploration captures semantic breadth and extension beyond the source; and retention tracks source preservation and the persistence of new claims across stages and lineages. We evaluate five interaction strategies on 40 controlled news-generation cases. Repeatedly pooled interaction promotes convergence but narrows the information space, whereas distributed relay yields broader, less shared information. Further comparisons show that method-specific constraints increase exploration while weakening convergence and retention. Our findings suggest that collective information dynamics are better explained by underlying information-flow mechanisms than by simple downstream performance.

0.31

Learning Pathology-Sensitive Representations for Unsupervised Retinal OCT Anomaly Detection

cs.CV cs.LG ·2026-10-23 ·Tania Haghighi, Sina Gholami, Behafarin Emam, Jennifer I Lim, Theodore Leng +3 more
Abstract

Deep learning approaches for retinal optical coherence tomography (OCT) predominantly rely on predefined disease categories, limiting their ability to detect rare or previously unseen abnormalities. While visual foundation models provide transferable representations, their effectiveness in identifying pathological deviations from normal retinal anatomy remains insufficiently understood. In this work, we introduce an unsupervised framework for retinal OCT anomaly detection and localization that adapts pretrained visual representations to normal retinal anatomy through anatomy-aware reconstruction and contrastive triplet learning guided by synthetic structural perturbations, without requiring disease-specific annotations. We systematically investigate the influence of pretraining strategies, anatomy-aware representation adaptation, and anomaly scoring mechanisms on the detection and localization of unseen retinal abnormalities. Our findings demonstrate that adapting pretrained representations to normal retinal structure improves anomaly sensitivity, while pixel-space reconstruction errors and feature-space deviations provide complementary signals for identifying pathological changes. These findings underscore the importance of anatomy-aware representation learning and offer insights into developing generalizable retinal anomaly detection frameworks beyond predefined disease categories.

0.30

LapaFM: A Laparoscopic Foundation Model via Multi-Stage Continual Learning

cs.CV ·2026-10-23 ·Xiaobo Zhu, Shiwei Wu, Yu An, Yuechen Tao, He Sun +19 more
Abstract

Artificial intelligence-based analysis of intraoperative laparoscopic videos has the potential to improve surgical decision-making and precision. However, effective surgical assistance requires models that can generalize across diverse intraoperative tasks. Here we present LapaFM, a foundation model for laparoscopic surgery trained on more than 20,000 minutes of surgical videos using a multi-stage continual learning strategy that mitigates catastrophic forgetting and inter-task interference. LapaFM learns multimodal visual representations and supports multiple tasks, including smoke removal, surgical phase recognition, instrument and polyp segmentation, and lesion identification. The model was evaluated on 13 retrospective and two prospective cohorts collected from different imaging devices. LapaFM consistently outperformed task-specific models and existing foundation models across datasets, and also generalized to gastrointestinal endoscopy tasks, including polyp detection and classification. Notably, it recovered over 90% of smoke-obscured frames and achieved clinician-level or superior performance in lymph node metastasis prediction and prospective polyp classification.

0.30

The LLENS Platform for On-Premises Large Language Model Agents in a Japanese University Hospital: Development and Evaluation Study

cs.CL cs.AI ·2026-10-23 ·Ken Enda, Jien Saito, Masami Mukai, Feng Han, Emir Turkes +10 more
Abstract · 49 pages, 3 figures, 6 tables, with supplementary material

Large language model agents that use tools could take on part of the electronic health record and desk work that occupies physicians' time, but in many hospitals the medical information systems run inside closed networks, so the model and the agent harness must also be inside the network. We designed, implemented, and deployed LLENS, a clinical support AI agent platform that runs inside the closed network of Hokkaido University Hospital on a single host with 8 NVIDIA H200 GPUs, the open-weight model Kimi K2.6 (~1 trillion parameters), and a web application rebuilt from Open WebUI. The harness has 10 components, including page image reading, full-text attachment input, a path-addressed document tree without a vector index, a network-isolated sandbox, artifact return, self-checking of generated output, subagents, and saved procedures. Kimi K2.6 was chosen from 3 open-weight candidates and scored 480/500 on the 2026 Japanese Medical Licensing Examination; it generated 119 tokens/s for a single user and served 16 concurrent requests at 908 tokens/s in total. On 180 questions over hospital documents (3 seeds each), mean accuracy was 10.7% with no documents, 81.3% with fixed retrieval (top 10 chunks of one lexical search), and 97.2% with interactive access, in which the model reads with document tools and the sandbox (15.9 percentage points above fixed retrieval, 95% CI 10.9-21.3); the difference lay in filtering/counting questions (16.7% vs 87.8%). In an automated test of 13 work scenarios (4 runs each), the expected components were triggered in all runs for 12, and 5 scenarios were run with synthetic data on the deployment host. The platform has run on the hospital network since September 14, 2026.

0.29

SCDM-HTD: Spectral-Wise Conditional Diffusion Model for Hyperspectral Target Detection

cs.CV ·2026-10-23 ·Dunbin Shen, Wenfeng Kong, Jianjun Liu, Xiaorui Ma, Hongyu Wang
benchmarkhot direction
Abstract

Hyperspectral target detection (HTD) aims to locate and recognize objects at a fine-grained scale, playing a pivotal role in remote sensing. Despite significant progress, current HTD methods primarily focus on feature comparison while neglecting distribution learning for background-target separation, which can limit detection robustness in practical applications. This paper addresses this issue by introducing a spectral-wise conditional diffusion model that reformulates HTD as a task of generating a detection map. Specifically, we treat background and target confidence scores as training samples and use the corresponding pixel spectra as guiding conditions to capture the distributional discrepancy between background and target through diffusion and denoising. To enhance practical applicability, we propose a lightweight conditional denoising network based on multilayer perceptron to map noisy confidence scores to their ideal counterparts and adopt a deterministic reverse non-Markov chain to skip sampling steps, thereby accelerating the generation process. Furthermore, we introduce a composite loss to improve background-target separation. This loss integrates constrained energy minimization loss to enhance target saliency and suppress the background by maximizing the target response while minimizing average output energy, alongside binary cross-entropy loss to classify background and target using pixel-wise labels. Experimental results on four benchmark datasets demonstrate the effectiveness of the proposed method, with improvements in both quantitative metrics and qualitative visual effects.

0.29

Disentangling Question, Response, and Verifier Languages in Multilingual P(True) Self-Evaluation

cs.CL ·2026-10-23 ·Jewon Yeom, Jaewon Sok, Seonghyeon Park, Jeongjae Park, Hwiyeong Lee +1 more
benchmarkhot direction
Abstract

When a language model is asked whether its own answer is correct, three languages are involved: that of the question, that requested for the answer, and that of the verifier instruction asking for the judgment. Multilingual confidence evaluations seldom separate the three. We vary them independently in a six-language factorial design with ten open-weight models (3B-14B) on LSQA, MMMLU and TriviaQA, holding each answer fixed across verifier languages and measuring confidence with a first-token P(True) probe. The three languages move different parts of the evaluation: which of them governs accuracy depends on the benchmark, matching the response language to the question or verifier language sharpens discrimination but not calibration, and the verifier language alone shifts calibration error by up to 0.04 without changing accuracy. The scorer matters as well: an unvalidated judge would have shrunk the MMMLU English-Hindi accuracy gap from 0.198 to 0.038. Multilingual confidence results, model rankings included, are therefore comparable only when all three languages are reported and the scorer is validated.

0.29

Consommé: Conditioned Token Merging for Split Inference

cs.CV ·2026-10-23 ·Man Ngo, Jan S. Rellermeyer
published
Abstract · Accepted at WACV'27, please cite the conference version

Split inference partitions a Vision Transformer between a resource-constrained edge device and a server. However, transmitting intermediate tokens across the communication channel is costly. Deploying fine-tuned backbones to every endpoint becomes expensive as well. We study token reduction for split inference under a frozen backbone, where only a small adapter is shipped. We observe that the CLS token plays a dual role: it acts as the global summary the receiver needs for classification, and it serves as a signal on which patches to keep. We propose Consommé, a method utilizing both characteristics. There, an information bottleneck compresses the CLS token into a compact latent which is transmitted in place of the full CLS token and conditions a learned patch gate merging patches into exactly $k$ tokens via soft assignment. It reaches any budget in a single step, thus escaping the per-layer cap of bipartite matching (BPM). On DeiT-S/T, it leads all methods at aggressive token budgets and remains operable at shallow splits where BPM cannot operate, while matching fine-tuned methods at gentle budgets at a fraction of their deployment cost.

0.29

DeformStruct: Deformable Structure Modeling for Point-Supervised Temporal Action Localization

cs.CV ·2026-10-23 ·Mengnan Liu, Sanping Zhou, Kun Xia, Yonghao Dong, Jie Wang +1 more
Abstract

Point-supervised Temporal Action Localization (PTAL) aims to localize action instances using only a single annotated timestamp for each instance. Such sparse supervision offers limited guidance for modeling the structural organization of an action instance. Existing PTAL methods primarily infer action regions from classification responses, leaving the structural organization of multiple action-related parts within each instance insufficiently modeled. To address this problem, we propose DeformStruct, an instance-level deformable structure modeling framework for PTAL. DeformStruct progressively constructs the structure of each action instance through three stages. First, Structural Anchor Mining discovers multiple anchors from annotated points and action-responsive regions, providing structural nodes grounded in the current instance. Second, Deformable Part Alignment associates latent structural parts with these anchors through deformation-aware matching, allowing their temporal supports to adapt to instance-dependent variations. Third, Relational Structural Organization explicitly models the relative temporal relations among the aligned parts, forming a relational representation that characterizes how different action-related parts are organized together. Extensive experiments on THUMOS14, ActivityNet-v1.3, GTEA, and BEOID demonstrate consistent improvements in point-supervised temporal localization and validate the effectiveness of deformable structural modeling.

0.29

Ontology-Guided Knowledge Graph Construction from Structured Metadata and Climate Science Publications

cs.CL cs.AI ·2026-10-23 ·Aayush Acharya, Qi Zhang, Eduard Dragut, Longin Jan Latecki
publishedcode/data
Abstract

The rapid expansion of scientific literature far outpaces the speed of manual curation, creating a fundamental bottleneck in extracting structured knowledge from scientific texts. To address this, we introduce an agentic, ontology-guided framework for knowledge graph (KG) construction. Our framework utilizes specialized agents for document understanding, information extraction, ontology conformance, and entity linking. First, an LLM extracts factual knowledge in the form of triples with explicit textual evidence from research papers. Then the triplets undergo a multistage verification process to ground the triplets in the ontology. By enforcing schema constraints and requiring evidence in the form of text fragments, we ensure the data is both semantically consistent and traceable. In an agreement study over 1,000 extracted triplets, Gemini-2.5-flash achieves Cohen's $κ$ scores of 0.640--0.700 against three human annotators, matching the human--human agreement range of 0.616--0.689. We applied this framework to 2,000 scientific publications and construct a novel, ontology-grounded KG for climate science with 133,990 nodes and 755,308 relations; of the 23,327 triplets that the publications contribute after validation, 98.96\% are judged correct. The pipeline code and KG is publicly available https://github.com/aayushacharya/climatekg.

0.29

From archive to resource: bringing legacy data back to life at scale with large language models

cs.IR ·2026-10-23 ·C. H. Vieira-Vieira, I. Schneider Chernov, A. Serra, M. Correa Marrero, P. L. Bhavani Mohan +3 more
benchmark
Abstract

Pharmaceutical companies maintain vast archives of legacy preclinical study reports that contain study-context details, such as fasting, anesthesia, and laboratory instrumentation, that are rarely captured in structured data formats, yet they can correlate with clinical pathology parameters values and limit the reuse of historical data. This work demonstrates practical, scalable solutions to recover such descriptors from heterogeneous report texts using large language model (LLM)-based extraction pipelines developed independently by two industry teams. The pipelines identify and normalize metadata fields and return verbatim evidence with page and section reference, hasting human verification and iterative refinement. Evaluation of over 400 documents against expert-curated reference values showed high agreement across a broad set of metadata fields. Performance was strongest for consistently reported items and lower for descriptors subject to linguistic or report structure variability. Extraction time has shrunk from 25 minutes in the manual benchmark to 1.6 minutes. As a concrete biological demonstration, we show that fasting status extracted from report text was associated with species- and analyte-specific shifts in clinical chemistry parameters, illustrating how recovered metadata can improve selection of historical controls. Key strengths of the approach are scalability, traceability, and the ability to recover information missing from current industry standards. Main limitations are dependence on source-report quality and initial expert effort to define and refine schemas. These results establish how LLMs can convert heterogeneous, unstructured legacy reports into evidence-grounded, reusable and structured metadata, leading to better experimental comparability, and thus enabling reliable data pooling or informed virtual control selection.

0.28

KINESYS: Test-Time Adaptation for Reference-Based Effect Editing in Images

cs.CV ·2026-10-23 ·Noah Frahm, Luchao Qi, Haoze Li, Dan B Goldman, Sambit Bhattacharya +1 more
code/data
Abstract · Project page: https://noahfrahm.github.io/KINESYS-Project-Page

Visual effects are often easier to demonstrate via reference than to describe in words. We study effect-reference editing, where a reference image shows a desired visual effect and language identifies which properties should guide the edit. Existing single-pass editing techniques often fail to make fine-grained edits, resulting in unintended changes to source content or deviations from the reference effect. We introduce KINESYS, a test-time adaptation framework for effect-reference editing that combines reference-aware and source-aware feedback to transfer the desired effect while preserving the composition and content of the original shot. To ensure KINESYS can produce a diverse set of edits that preserve the source, we develop a robust optimization strategy in which the generative model is fine-tuned across multiple noise samples and optimization runs. Evaluation with VLM judges, source preservation metrics (LPIPS, TPIPS), and a user study shows that KINESYS outperforms state-of-the-art image editors in source preservation; on effect transfer, it remains competitive with Nano Banana 2 and outperforms open-weight editors.

0.28

SCI-AGENT-EXAM: Benchmarking Agents on Scientific Research Tasks

cs.CL ·2026-10-23 ·Tianyun Zhong, Wei Wang, Jialin Li, Jianhong Tu, Liyao Li +15 more
benchmarkcode/data
Abstract

Large language model (LLM) agents are increasingly capable of supporting scientific workflows, but their capabilities across diverse research activities remain difficult to assess systematically. Existing benchmarks often focus on selected scientific capabilities or task types, limiting a comprehensive characterization of scientific agents' strengths and weaknesses in realistic research workflows. We introduce SCI-AGENT-EXAM, a workflow-oriented benchmark comprising 120 expert-curated tasks across 11 scientific fields and 7 research task types defined by the Scientific Workflow Task Taxonomy (SWTT). Each task requires both executable artifacts (e.g., code and scripts) and textual artifacts (e.g., reports and explanations). To evaluate the heterogeneous task outputs, we develop a hybrid evaluation protocol combining deterministic programmatic verification for executable artifacts with a calibrated, expert-validated LLM-based judge for textual artifacts. Beyond outcome assessment, we analyze progress and solution stability along execution trajectories to characterize how agents reach their final results. Experiments with 10 frontier LLM agents reveal substantial performance gaps: GPT-6 Astra achieves the highest overall score of 80.9, yet only 26.0% of its runs exceed the 0.95 success threshold. Further analyses uncover distinct capability profiles across scientific task types and substantial variation in execution efficiency and stability. Our code and data are available at https://github.com/SKYLENAGE-AI/Sci-Agent-Exam.

0.28

Standing on Open Shoulders: Regional Fine-Tuning of Open-Weight Camera-Trap Models for UK Wildlife with Modest Data and Compute

cs.CV ·2026-10-23 ·Paul Fergus
Abstract · 39 pages, 9 Figures, and 8 Tables

Camera-trap species classification is dominated by large models from well-resourced organisations, which often lack the species-level distinctions regional conservation needs, such as native red versus invasive grey squirrels in the UK. We ask whether open weights let a small team with a modest dataset and a single GPU take a real part in AI development. We fine-tuned the open DeepFaune classifier on 28 UK classes, paired it with the unmodified MegaDetector v6, and compared the resulting DeepFaune-UK v1.2 with DeepFaune v1.5.0 and SpeciesNet on 20,212 hand-verified MammalWeb citizen-science images, after removing sequences sharing any of 1,039 images with our training archive. DeepFaune-UK achieved a strict image-level accuracy of 88.8%, including empty frames, against 56.5% and 59.0% for the baselines, but at DeepFaune's coarser label resolution DeepFaune v1.5.0 was more accurate, though not significantly under contributor resampling (91.4% vs. 88.8%). As DeepFaune v1.5.0 uses a newer DINOv3-based classifier, we applied the identical recipe, without new data, to that classifier. The resulting DeepFaune-UK v1.3 reached 90.6% strict accuracy, narrowed the like-for-like gap to 0.8 points and reduced roe deer misclassified as fallow deer by 69%. On 1,850 images from the 26 contributors with no overlapping images, the three DeepFaune-based models did not differ significantly (90.5%, 90.6%, 89.9% like-for-like). Fine-tuning on 10% of the training crops (about 150 per class) gave out-of-distribution accuracy similar to using all of them, and a DINOv2 run on this subset took about 22 minutes with under 4 GB of GPU memory. Open weights thus make regional model development feasible for small teams, but the largest like-for-like gain came from a newer base model trained by others, so this democratisation depends on large actors continuing to release open weights, under licences that may change.

0.28

A Broader Look into MLLM's Retrieval Heads

cs.CL ·2026-10-23 ·Jia Wang, Yuqing Li, Jiangnan Li, Zheng Lin, Mo Yu
community pick
Abstract · work in progress

Locating relevant evidence in long inputs is a core capability of Multimodal Large Language Models. Prior studies attribute this capability to retrieval heads, a sparse set of attention heads that transfer information from the context to the output. However, these analyses are confined to a limited set of modalities and the corresponding models. Moreover, how retrieval heads can be leveraged to enhance the model calls for further exploration. In this paper, we ask whether this mechanism expands together with the input modalities of the model. Taking audio-visual input as a newly added modality, we construct paired inputs in which the same answer is either spoken or shown on screen. With the evidence content fixed, retrieval heads remain sparse in the new modality and are shared to a large extent across modalities. Across 18 language, vision-language, and omni-modal checkpoints, omni-modal models largely preserve the retrieval heads of their backbones. Ablating these heads eliminates nearly all audio-visual retrieval. The heads shared across modalities make up fewer than 2\% of all heads, yet ablating them also degrades document understanding, spatial perception, and visual mathematics. Building on these findings, we turn retrieval heads into a target of optimization by training their attention with a ranking objective. This raises the page Recall@1 of Qwen3-VL-8B on MMDocIR by 6.03 points and improves document retrieval across three model families. Functional attention heads can thus serve both as a lens on multimodal models and as a target for improving them.

0.28

Detecting Temporal Concepts in Video Transformers via a No-False-Positives Test

cs.CV cs.LG ·2026-10-23 ·Jianghuai Li, Andrew Ren, Greg Shakhnarovich
Abstract · 39 pages, 1 figure, 23 tables

Video models have the architecture to support encoding temporal aspects of the input, i.e. concepts that require more than one frame to define, like object movement. However, because motion and appearance change are entangled, a feature that appears to encode temporal concepts like object movement may instead encode non-temporal concepts like spatial presence at different positions whose activations merely correlate with motion. To resolve this, we introduce the No-False-Positives (NFP) test, a statistical test with a formal guarantee of statistical specificity for identifying features encoding temporal concepts. Using specially designed stimuli, we apply this test to representations produced by VideoMAE, V-JEPA2, and TimeSformer, and observe that temporal features exist in each model. We validate the test first by showing it yields no temporal features for the DINOv2 image-encoder. Then, using a synthetic representation space, we show that detection rate increases with a direction's projection onto the temporal subspace, and that Sparse Autoencoder features test-flagged as temporal project much more of their norm onto the temporal subspace than unflagged features. Interventions on temporal SAE features yield mixed results across models, suggesting that intervening on temporal features may not outperform a non-temporal feature baseline.

0.28

Beyond What to Say: Proactive Utterance Timing with Inner Thoughts for Werewolf Agents

cs.CL ·2026-10-23 ·Yoshiki Tanaka, Shuya Uchiyama, Daichi Hamaguchi, Michimasa Inaba
Abstract · Accepted to the 4th International AIWolfDial Workshop at INLG 2026

Participating in a conversation involves not only deciding what to say but also whether to speak at all. This judgment carries even greater significance in environments like the Werewolf game, where participants' interests conflict, because both utterances and silence are subject to interpretation and assessment by others. Therefore, determining when to speak, as well as what to say, is crucial for an LLM agent playing the Werewolf game. In this study, we introduce a Werewolf AI agent developed for the Anytime Utterance Track of the AIWolfDial shared task. Our system builds on the existing Inner Thoughts framework, which determines utterance timing during conversations by generating internal thoughts and evaluating the value of each thought for expression at that moment. Since existing methods were designed for general dialogue, we have redesigned the system specifically for the competitive environment and incomplete information setting of Werewolf games. We call our adapted method Adversarial Inner Thoughts. Through experiments involving analysis of self-play logs and human evaluations, we analyze and discuss the effectiveness of our approach.

0.28

From Promises to Accountability: Comparative AI Modeling and Visualization of Electoral Commitments in Nepali Political Texts

cs.CL ·2026-10-23 ·Prakash Poudyal, Nischal Karki, Amrit Dahal, Rajani Chulyadyo, Sameer Tamrakar +1 more
unusual
Abstract · Under review at Natural Language Processing (Cambridge University Press); revision submitted. 34 pages, 5 figures, 13 tables

Election Commitment Mining refers to the analysis of election commitments made by electoral candidates or political parties during election campaigns. However, many elected candidates and political parties, especially in developing South Asian countries, do not fully deliver on the promises they make before elections. Even after the elections, commitments presented through election manifestos, posters, pamphlets, videos, and other campaign materials are often not systematically or transparently archived. As a result, these promises tend to fade away over time, making it difficult to hold the elected candidates accountable for their commitments. To address this persistent issue, this research aims to develop a system capable of structuring the commitments made by candidates during election periods. In this work, we created the first gold-standard dataset of Nepali political texts (Cohen's kappa = 0.777) and evaluated machine learning, deep learning and large language models to automatically extract and analyze election commitments. LaBSE-LR achieved the best results among machine learning models (0.897 accuracy, 0.896 macro F1-score, 0.947 AUROC), while fastText-LSTM achieved the highest performance among deep learning models (0.914 accuracy, 0.913 macro F1-score, 0.961 AUROC). Among the three evaluated large language models, Qwen3-8B performed best in the few-shot setting (0.848 macro F1-score). The outcome of this work is integrated into a mobile application, where commitments can be visualized with the support of GAI, enhancing transparency, civic engagement, and electoral accountability.

0.28

SceneBloom: Agentic Indoor Environment Generation with Realistic and Diverse Layouts for Embodied Learning

cs.CV cs.AIcs.RO ·2026-10-23 ·Yuhan Liu, Lianhui Qin, Shenji Wan
code/data
Abstract

Training generalizable embodied agents requires realistic, varied, and simulation-ready indoor environments. Language-guided methods support semantic planning and object-level reasoning, but precise spatial grounding remains difficult. Visual generative models offer rich priors over plausible arrangements, yet their outputs require geometric reasoning to become physically valid 3D scenes. We introduce SceneBloom, an agentic framework that combines visual layout sampling with contract-guided 3D compilation. A hierarchical sampler uses diffusion priors to generate furniture arrangements within functional zones. A compiler then uses relations extracted from each proposal to guide geometric fitting and local pose correction. Across five room types, SceneBloom attains the highest visual quality and physical stability among the compared methods and is the only method with no detected mesh collisions and out-of-bounds objects. Its layouts are also closer to the distribution of real rooms than those of all compared baselines. Object navigation policies trained in these realistic scenes achieve a higher success rate in held-out real scans than those trained on an equal number of real training scans. Code is available at https://github.com/Tinaliu0123/scenebloom.

0.28

DiffusedSplitFed: Latent Diffusion-based Obfuscation for Reconstruction-Resilient Split Federated Medical Image Segmentation

cs.CV cs.AI ·2026-10-23 ·Chamani Shiranthika, Parvaneh Saeedi
code/data
Abstract

Federated Learning (FL), Split Learning (SL), and Split Federated Learning (SplitFed) enable collaborative medical image analysis without sharing raw patient data. However, intermediate features and gradients exchanged during SplitFed training may reveal sensitive information through reconstruction attacks. To address this vulnerability, we propose \textbf{DiffusedSplitFed}, a SplitFed framework that integrates Latent Denoising Diffusion Models (LDDMs) at client--server split points to obfuscate transmitted representations while preserving segmentation performance. Three architectural variants (V1--V3) are developed using different combinations of dual conditioning and global feature fusion to investigate trade-offs among reconstruction resistance, segmentation accuracy, and computational complexity. Experiments on three medical imaging datasets (HAM10K, Blastocysts, and FHPsAOPMSB), supported by multiple independent runs and paired statistical analyses, demonstrate improved resistance to the evaluated image reconstruction attacks compared with Baseline SplitFed, while maintaining competitive segmentation performance. Comparisons with existing diffusion-based segmentation methods further demonstrate competitive segmentation accuracy under the evaluated settings. A conditional convergence analysis is also provided to characterize the effects of LDDM-induced gradient perturbations and local optimization on training. These findings highlight the potential of latent diffusion-based representation obfuscation for reconstruction-resilient collaborative medical image segmentation. The implementation is available at https://github.com/ChamaniS/DiffusedSplitFed.

0.27

Hair Segmentation and Skin Characterization on Macroscopic Dermatological Images for Laser Treatment Parameterization

cs.CV ·2026-10-23 ·Alex Navarro Navarro, Marc Arriaga Hidalgo, Carlos Moreno Rubio, Maxime Talarmain, Noelia Vaquero Gallardo
published
Abstract · Published open access in Journal of Imaging Informatics in Medicine (2026)

Current light-based devices, such as laser hair removal, intense pulsed light (IPL), vascular laser therapy, and dermatologic phototherapy, are widely used in clinical dermatology but rely heavily on subjective manual parameterization or uncalibrated point-based colorimeters. To overcome these limitations, this work presents a unified system for hair segmentation and skin characterization in macroscopic dermatological images. Unlike most existing literature that relies on dermoscopic datasets captured under highly controlled conditions, our approach is specifically designed for macroscopic images that replicate real clinical laser treatments. The proposed pipeline integrates a custom-trained variant of the YOLOv9 deep learning architecture for instance segmentation and classical image processing within a single workflow to extract objective skin and hair parameters such as phototype, hair density, color, and thickness. A color calibration stage using spectrophotometric measurements ensures accurate feature estimation regardless of illumination conditions. Experimental results show competitive segmentation metrics (mAP@50 $\approx$ 0.80) on a custom dataset featuring short, shaved hairs, typical of pre-treatment scenarios. These findings demonstrate the feasibility of the proposed pipeline in providing the objective, foundational skin and hair features required for future automated laser parameterization, paving the way for its extension to broader light-based medical therapies.

0.27

Cross-Architectural Feature Fusion for Coronary Artery Stenosis Detection in X-ray Angiography

cs.CV ·2026-10-23 ·Haobo Zhu, Mojtaba Lashgari, Jason Chai, Giovanni Luigi De Maria, Vicente Grau +1 more
Abstract · 12 pages, 3 figures

Coronary artery disease remains a leading cause of mortality, with stenosis detection from invasive X-ray Coronary Angiography (XCA) critical for diagnosis. Manual interpretation is time-consuming, subjective, and variable, especially under low contrast and vessel overlap. Existing deep learning methods often rely solely on detection or segmentation, limiting performance. We propose a hybrid framework, Fusion-YOLOv11-L, combining a YOLOv11-L detector with a pretrained UNet++ via a channel attention gating module. This enables fusion of semantic and vessel-aware structural features, improving localisation of stenotic regions. The UNet++ is pretrained on ARCADE vessel segmentation dataset to provide anatomically informed representations. Experiments on the relabeled ARCADE (ARCADE*) dataset show improved performance over YOLOv11-L, achieving AP@50 of 0.585 and F1-score of 0.596. Cross-domain evaluation on a dataset from the Oxford Acute Myocardial Infarction (OxAMI) Study reveals performance degradation due to domain shift, mitigated using test-time adaptation and augmentation. Test-time augmentation improves AP@50 from 0.434 to 0.480, indicating better generalisability. These results show that cross-architecture feature fusion and lightweight adaptation improve robustness and generalisability of automated stenosis detection in XCA.

0.27

SOPO: Socratic Guided Policy Optimization for Span-Level Hallucination Detection

cs.CL cs.AI ·2026-10-23 ·Jiawei Dong, Zilong Bai, Pengtian zhu, Peng Zhou
code/data
Abstract · 24 pages, 11 figures

Large Language Models (LLMs) have gained widespread adoption in various natural language processing tasks, but suffer from hallucination issues where they generate unfaithful or inconsistent content. While Reinforcement Learning with Verifiable Rewards (RLVR) has shown promising applications in reasoning tasks, its application to span-level hallucination detection is challenged by localization ambiguity, reward sparsity on hard examples, and uncontrolled length inflation. To address these issues, we introduce Socratic Guided Policy Optimization (SOPO), a multi-turn RL framework for robust hallucination detection. The key innovation of SOPO is a Socratic feedback mechanism, where a teacher model guides the reasoning trajectories during rollout by providing indirect hints inferred from the ground truth. This teacher-guided approach enables the model to generate more effective reasoning trajectories for hard examples, thus alleviating reward sparsity. Furthermore, we incorporate Turn-based Advantage Scaling to penalize verbosity bias, encouraging the model to develop efficient and autonomous reasoning capabilities. Empirical results on RAGTruth demonstrate that SOPO (4B) achieves state-of-the-art performance, surpassing leading proprietary models (e.g., GPT-5) and standard RL baselines, while obtaining higher reported mean scores than GRPO on four out-of-domain benchmarks, with mixed evidence against CAPO. We release our training and evaluation code alongside the 1.7B and 4B SOPO models at https://github.com/patsnap/SOPO.

0.27

Hyperparameter-Free Boundary-Interior t-vMF Dice Loss for Medical Image Segmentation

cs.CV ·2026-10-23 ·Hinako Mitsuoka, Kazuhiro Hotta
Abstract

Class imbalance and errors near anatomical boundaries remain challenges in medical image segmentation, motivating loss functions that account for these difficulties. Adaptive t-vMF Dice Loss adjusts the similarity function during training but requires hyperparameter tuning. We propose Self-Tuned Boundary-Aware t-vMF Dice Loss (STB-tvMF Dice), which adapts the function separately for boundary and interior regions without tuning shape-adaptation parameters. Each region's shape parameter of t-vMF function is determined from its mean prediction-target cosine similarity in the preceding epoch and remains fixed throughout the current epoch. This rule maximizes the slope of the similarity function at the observed mean similarity, providing a principled basis for adaptation without validation-based parameter selection. Experiments on three datasets and architectures demonstrated its effectiveness. On Synapse with U-Net, STB-tvMF Dice achieved a mean DSC of 70.75%, compared with 63.03% for Adaptive t-vMF Dice Loss. Our code will be available soon.

0.27

An Evolutionary Hypothesis for White's Illusion: Constant Percentage Correction and Modern Individual Variation

cs.CV q-bio.NC ·2026-10-23 ·Bekir Tevfik Akgün, Ergun Akleman
unusual
Abstract · 12 pages and 7 figures. We may later send a shorter version to a conference

White's illusion is a striking demonstration of context-dependent brightness and lightness perception. In earlier work, we developed a forward translucency-based formula capable of producing approximately constant perceived appearance across varying backgrounds. While validating that formulation, we observed a notable population-level pattern: individuals selected slightly different correction percentages for "constant perception," while the population average remained remarkably stable across experiments. In this paper, we propose an evolutionary hypothesis for this phenomenon. We suggest that a simple constant-percentage correction function could have emerged as an adaptive mechanism for maintaining useful material and reflectance judgments under variable illumination. Under this interpretation, the stability of the population mean may reflect a conserved perceptual mechanism, whereas individual differences may reflect learning, adaptation, or variation around a shared computational response. Using a modification of our previously published forward model, we show that White's illusion can be reproduced by applying a constant-percentage correction to background luminance. The model therefore provides a compact computational account of the observed population-level regularity and motivates a testable evolutionary interpretation of individual variation in contextual lightness judgments.

0.27

Neural Disjunctive Normal Forms for Learning Exact Rank Filters

cs.CV cs.LG ·2026-10-23 ·Diego Marcondes, Wonder A. L. Alves, Ronaldo F. Hashimoto
unusual
Abstract · 12 pages, 4 figures, 2 tables

Gradient-based learning of morphological operators is hindered by locally sparse derivatives with respect to the structuring elements. We develop a method for learning rank filters that retains exact operator evaluation while distributing the derivatives across the window. Using threshold decomposition, we construct a neural disjunctive normal form representation and establish a sufficient condition for exactness with integer-valued images and structuring elements. We then derive a smooth surrogate from the unique root of a smooth threshold equation. Implicit differentiation gives closed-form derivatives with normalised local weights, whose distribution is controlled by a smoothing parameter. These derivatives are used in a straight-through procedure: the forward pass evaluates the exact rank filter, while the backward pass uses the surrogate. For fixed ranks, the procedure learns real-valued structuring elements of any rank filter, including the median, and extends to compositions, infima, and suprema of rank filters. Experiments on image denoising using MNIST and Fashion-MNIST and retinal vessel segmentation show that, starting from a generic random initialisation, the proposed approach consistently improves performance over training with exact subgradients in every setting. In image denoising, our best network also outperforms smooth morphological networks in all settings. On vessel segmentation, the learned erosions recover structuring elements that fit the vessels, as theoretically expected, whereas most of those trained with exact subgradients remain close to their initialisation.

0.27

GeoLEON-CLIP: Graph-Anchored Multimodal Contrastive Learning for Geographic Region Representation

cs.CV cs.LG ·2026-10-23 ·Radoslaw Malek, Szymon Soltysiak, Aleksandra Kawala-Sterniuk, Piotr Szymanski
published
Abstract · Accepted at the 9th International Workshop on AI for Geographic Knowledge Discovery (GeoAI 2026), co-located with the 34th ACM SIGSPATIAL In…

Geographic regions are multimodal: satellite imagery captures visual morphology, text conveys human-interpretable semantics, and map structures describe urban context. Existing geospatial contrastive methods typically align image-text or image-location pairs, underusing structured map data. We introduce GeoLEON-CLIP, a graph-anchored contrastive framework that aligns three views of each H3-indexed region: an OpenStreetMap-derived graph embedding, a satellite image crop, and a textual description. Frozen pretrained LEON graph embeddings provide a semantic anchor between vision and language. Frozen graph, image, and text encoders are connected by trainable residual projection heads and trained with a symmetric tri-modal contrastive objective. Experiments show strong graph-image and graph-text alignment; graph-image retrieval achieves the best cross-modal performance in a mixed-resolution pool. To evaluate the anchoring hypothesis, we propose chained cross-modal recall $\mathrm{R}@A@B$, which routes retrieval through the graph modality and compares it with direct retrieval under a matched candidate budget. At $\mathrm{R}@1@1$, image$\rightarrow$graph$\rightarrow$text and text$\rightarrow$graph$\rightarrow$image outperform direct image-text retrieval by 4.8 and 4.4 points, respectively, despite using a single intermediate cell. Downstream probes show task-dependent gains, with the largest improvements on Airbnb and Beijing housing. These results indicate that OpenStreetMap graph structure can anchor multimodal representations combining maps, imagery, and language.

0.27

TIE: Training-free CLIP Embedding Guidance for Compositional Text-to-Image Alignment

cs.CV ·2026-10-23 ·Sujung Hong, Tae Eun Choi, Youngjun Jun, Chanyong Yoon, Seong Jae Hwang
code/data
Abstract

Text-to-image (T2I) diffusion models have advanced notably, but still fail to satisfy prompt conditions, resulting in attribute mismatches and missing objects. These errors normally fall into two categories: concept loss, when an object or attribute (i.e., concept) goes missing, and concept confusion, when an attribute is assigned to the wrong object or multiple objects blend together. Recent studies suggest that these challenges are caused by the limited capacity of CLIP text encoder to capture fine-grained semantic details. Although several methods have been proposed, concept loss and concept confusion persist in recent T2I models when handling multi-object, multi-attribute prompts. In this paper, we conduct an embedding-level analysis to develop a training-free, inference-time solution for concept loss and concept confusion across CLIP-based T2I models. We find that (1) concepts mentioned later exhibit higher embedding entropy, reflecting less concentrated representations that are more often associated with concept loss, and (2) the first CLIP attention layer captures the strength of binding between each object and its attribute. Guided by our findings, we introduce TIE, a method that improves semantic alignment through a single text-embedding update. TIE addresses concept loss via entropy-aware singular value amplification and resolves concept confusion through interpolation-extrapolation binding based on CLIP attention scores, all in a training-free manner. Extensive experiments demonstrate that TIE enhances semantic fidelity in multi-concept scenarios with minimal sampling overhead. The project page is available at https://hong-sujung.github.io/TIE/.

0.26

Benchmarking Data-Free Continual Model Merging for Whole-Slide Image Survival Analysis

cs.CV ·2026-10-23 ·Vu Minh Tran, Doanh C. Bui, Maï K. Nguyen, Khang Nguyen
benchmark
Abstract

Continual survival analysis on whole-slide images (WSIs) requires models to integrate knowledge from sequentially arriving cancer cohorts while preserving previously learned information. Model merging offers a data-free alternative by combining independently fine-tuned task experts without retaining previous training data. However, the relative behavior of different merging strategies for WSI-based survival prediction remains insufficiently understood. In this work, we establish an end-to-end benchmark for data-free continual model merging, covering task-specific fine-tuning, sequential merging, and inference under both TASK-IL and CLASS-IL settings. We systematically evaluate six representative merging strategies across four TCGA cancer cohorts and three task orders using C-index, IPCW C-index, Forgetting, and Backward Transfer under a controlled protocol. Results show that no merging strategy consistently dominates across all evaluation criteria, while censoring-aware evaluation can substantially alter the relative ranking of the compared methods. This benchmark provides a unified foundation for developing and evaluating more effective continual model merging methods for WSI-based survival analysis.

0.26

UniWorld: A Unified World Model for Real-Time Visual World Simulation

cs.CV ·2026-10-23 ·Xin Zhang, Yabo Chen, Shuaiwei Wang, Zixuan Duan, Haibin Huang +2 more
code/data
Abstract · Project page: https://zephipage.github.io/UniWorld

Continuous visual world simulation requires coupling evolving scene geometry with temporally persistent appearance. We present UniWorld, a unified framework that integrates depth trajectories and reference appearance within a shared video generation model. The central design maintains visual and geometric histories on a common timeline. Noise-aligned overlapping continuation places generated history at the prescribed sampling noise level and commits only newly generated latents. Rolling depth latent state preserves the corresponding geometric controls, incrementally reads newly available geometry, and keeps the conditioning fixed within each sampling window. Together, these mechanisms ensure exact preservation of latent history while bounding the active depth state by the window size. We evaluate continuous generation, paired history interventions, and long-horizon behavior across diverse scene categories. The same generator supports estimated, native, and on-demand simulator depth, while cached and incremental execution paths produce bitwise-identical outputs. Appearance changes and geometric control branches operate through the same interface without modifying committed history. The measured steady-state DiT execution path achieves real-time generation throughput, demonstrating the efficiency of UniWorld for continuous visual world simulation. Project page: https://zephipage.github.io/UniWorld.

0.26

Distortion-Aware Spatiotemporal Learning for Real-Time Wide-Angle Video Rectification

cs.CV ·2026-10-23 ·Lang Nie, Yihan Yang, Wenbo Nie, Hongguang Zhu, Yang Wei +2 more
benchmark
Abstract · 9 pages, 7 figures

While wide-angle lenses provide an expanded field of view (FoV), they exhibit severe nonlinear distortion near image boundaries, unnaturally stretching people in these regions. Existing methods mostly focus on facial regions, require high computational costs, and lack temporal modeling, resulting in incomplete full-body rectification, impractical real-time deployment, and noticeable inter-frame shakes. To tackle these issues, we propose \textbf{FastVC}, an efficient full-body video rectification framework driven by distortion-aware spatiotemporal learning. Grounded in lens physics, FastVC introduces a distortion-aware thin-plate spline (DA-TPS) module that nonlinearly remaps control points into a peripheral-dense layout, enabling more accurate geometric fitting without additional computational cost. We further develop a dual-level spatiotemporal learning scheme for causal and stable video rectification. At the model level, neighborhood spatiotemporal attention (NST) aggregates local context and selectively fuses historical features with low overhead. At the optimization level, sparse spatiotemporal supervision applies spatial alignment only to anchor frames, decoupling frame-wise accuracy from temporal stability. Besides, we contribute the first benchmark dataset for wide-angle full-body video correction. Extensive experiments show that FastVC achieves state-of-the-art rectification fidelity and visual stability at 57.0 FPS.

0.26

When is more search useful? An Empirical Study of Search Utility in Quality-Aware Decoding for Machine Translation

cs.CL ·2026-10-23 ·Javier Garcia Gilabert, Rico Sennrich, André F. T. Martins
Abstract

Quality-Aware Decoding (QAD) improves machine translation through test-time scaling, generating multiple candidate translations and selecting the final output with a quality-aware decoder. Usually the same candidate budget N is spent on every source sentence, assuming that all inputs require the same amount of search. We show that this assumption does not hold and that the quality gain obtained from additional candidates varies across sentences. We study search utility through three quantities: total scaling gain, how much quality a sentence can gain from a much larger candidate pool; marginal scaling gain, the expected gain from generating one additional batch of candidates; and the sampling saturation point, the candidate budget required to realize most of the total scaling gain. We investigate two sources of information for predicting these quantities: signals available before candidate-set expansion, and statistics computed from the candidate pool accumulated during search. We find that sentence-level signals correlate with total scaling gain but are less informative about the sampling saturation point. In contrast, properties of the current candidate pool provide useful signals of marginal scaling gain. Finally, we show that these signals can identify sentences that obtain larger gains when additional search is selectively applied.

0.26

Knowledge Distillation across Item and Code Spaces for Parallel Generative Recommendation

cs.IR cs.LG ·2026-10-23 ·Yixuan Xu, Hailin Liu, Linxiang Peng, Can Yang
benchmark
Abstract · 12 pages, 2 figures

Generative recommendation with semantic IDs represents each item as a sequence of discrete codes derived from its semantic representation and predicts these codes to generate recommendations. Existing semantic-ID recommenders are typically trained with hard labels for the observed next item. Although such supervision identifies the target, it provides no explicit distributional information about alternative items conditioned on the same user history. Sequential recommenders can supply this item-level predictive knowledge, but transferring it to semantic-ID generation is challenging because the teacher and student operate in item and code spaces, respectively. We propose two complementary knowledge-distillation objectives for parallel semantic-ID recommenders, which predict all code positions simultaneously without autoregressive prefix conditioning. Item Distribution Matching (IDM) reconstructs a catalogue-level item distribution from the student's parallel code predictions and aligns it with the teacher's item distribution. Code Marginal Matching (CMM) projects the teacher's item probabilities onto each code position using fixed item-to-code assignments, then matches the resulting code marginals to the corresponding student heads. Both objectives transfer the teacher's soft distributional knowledge while preserving the student's parallel prediction structure. Experiments on benchmark datasets show that both methods consistently outperform existing state-of-the-art approaches. On average, IDM improves Recall@10 by 15.39%, while CMM improves NDCG@10 by 8.28%. These results demonstrate that distillation in either reconstructed item space or code-marginal space can substantially strengthen hard-label supervision for parallel generative recommendation.

0.26

CounselPlex: A Stressor-Aware Full-Duplex Counseling Model

cs.CL cs.SD ·2026-10-23 ·Yeonju Kim, Bella Godiva, Yong Man Ro
Abstract · 21 pages, 4 figures, 29 tables

Full-duplex speech models can listen and speak at the same time, enabling natural, low-latency interaction. In counseling, however, these models often focus on the latest utterance and lose track of the seeker's underlying problem. We call this failure Stressor Blindness. We propose CounselPlex, a stressor-aware full-duplex framework that uses two operations. CounselPlex updates a short stressor state during seeker silence and injects this state into <PAD> positions while listening. We also introduce RoPE Position Shift Augmentation (RPSA), which broadens positional coverage during training and improves performance at longer turn depths in our evaluation. LLM-judge evaluations show improved counseling quality and stressor grounding over strong baselines.

0.26

Residual Expert Routing for Text-Only Domain Adaptation of Factorized Transducer ASR

cs.CL ·2026-10-23 ·Shashi Kumar, Esaú Villatoro-Tello, Thibault Bañeras-Roux, Sergio Burdisso, Petr Motlicek +2 more
Abstract · Paper submitted to ICASSP 2027 (Under review)

Fine-tuning the non-blank predictor of a factorized transducer on target-domain text improves recognition on that domain but degrades performance on the source domain. A single interpolation weight between base and adapted predictors cannot resolve this: the adapted predictor helps at some prefixes and hurts at others, often within one utterance. We propose conservative residual routing: the adapted predictor becomes a residual correction to the base, and a 16k-parameter gate decides at each prefix how much of it to admit. Closing the gate reproduces the base exactly. On a 40-hour conversational corpus, routing retains essentially all of the text-only gain on two held-out domains (19.89 to 16.74, 20.91 to 16.21 WER) while leaving source-domain WER unchanged (12.90 to 12.93); the adapted predictor alone costs 1.08 WER. Routing needs no per-domain tuning on target-domain audio. Analysis reveals the gate does not behave as a simple domain-level switch: it modulates intervention through the disagreement between the two predictors, suppressing it in-domain and amplifying it out-of-domain.

0.26

Explaining Basketball Shot Classifications with Grad-CAM: An Investigation of Shortcut Reliance with Spatiotemporal Correlations

cs.CV ·2026-10-23 ·Hongyi Tian, Robin Cohen
code/data
Abstract · 2 pages, 2 figures. Code and data: https://github.com/spicybigmac/basketball

Basketball action recognition is used in high-stakes scenarios to inform such tasks as specialized coaching, yet the evidence used by its deep-learning models is poorly understood. This paper investigates whether there is evidence that an R(2+1)D-18 model trained on the Basketball-51 dataset relies on spurious spatiotemporal correlations when classifying basketball shot outcomes. Gradient-weighted Class Activation Mapping (Grad-CAM) was used to generate spatiotemporal saliency maps, whose faithfulness was evaluated by progressively blurring the most salient pixels. Twenty saliency maps were then visually reviewed for recurring salient regions. Grad-CAM highlighted spatially central regions of clips including the shooter, net, and occasionally the courtside crowd. Blurring interventions were conducted to support these patterns. Blurring the central region resulted in an accuracy of 43.7%, compared to the original test accuracy of 69.3% and random-blurred accuracy of 50.9%. The additional 7.2 percentage-point reduction from central blurring compared to random blurring suggests that the model relies on task-relevant visual regions; this highlights the need for continual audits to ensure trustworthy, equitable sports analytics.

0.26

SPOOK: Sparse Pixel Optimization for Off-Manifold Knowledge in Vision Encoders

cs.CV cs.CLcs.HC ·2026-10-23 ·Ankit Gupta, Akshat Pandya, Christoph Adami, Emily Dolson
Abstract · 12 pages, 3 figures, 3 tables, and 14 pages of Supplementary Material with 8 figures

Modern vision encoders are widely used as semantic representations, but the relationship between embedding similarity and human-visible semantics can become unclear far from the natural-image manifold. We probe this regime in a black-box setting by applying sparse pixel edits to blank canvases with a simple pixel-level hill-climbing procedure. Across ImageNet and Flickr30k, we evaluate whether these edited blank canvases can retrieve chosen targets in multimodal contrastive encoders and self-supervised visual encoders. On OpenCLIP, inputs optimized toward class prompts achieve strong top-1 and top-5 targeted retrieval performance; on DINOv3, inputs optimized toward class prototypes achieve high target alignment and near-top prototype ranks. These results show that sparse edits on otherwise blank, visually non-semantic inputs can enter target semantic basins in both language-aligned and self-supervised visual representation spaces. Our findings identify a shared off-manifold failure mode of modern vision encoders and suggest sparse blank-canvas optimization as a simple probe of representation geometry.

0.26

Are Textual Residuals Truly Complementary? Learning Visual Complements under Semantic Superposition for Visible-Infrared Person Re-Identification

cs.CV ·2026-10-23 ·Xiaoyang Li, Mingming Lu, Wei Zhou, Hao Li, Canhui Zhang
Abstract · 24 pages, 4 figures

Text-enhanced visible-infrared person re-identification (VI-ReID) exploits textual descriptions to compensate for appearance cues that are weak or unavailable under infrared imaging. Existing methods construct compensation from differences between modality-oriented textual representations, implicitly assuming that textual differences are visually complementary. We revisit this assumption through semantic superposition and show that dense textual residuals may entangle useful appearance variations with redundant, making direct residual fusion unreliable. We propose Superposition-Aware Residual Calibration, which transforms dense textual differences into selective and visually grounded complements. Low-Interference Sparse Residual Learning represents modality-oriented text embeddings as sparse activation strengths over a shared dictionary, selects positive strength differences, and suppresses both dictionary-level overlap and activation-weighted interference to reconstruct a compact, selectively controlled residual. Visual Complement and Counterfactual Calibration further calibrates the residual against identity-specific RGB-IR visual discrepancies, encouraging retained semantics to provide meaningful compensation while preserving the residual structure. Experiments on Tri-SYSU-MM01, Tri-RegDB, and Tri-LLCM show that SARC consistently improves the VI-ReID backbone and remains competitive with text-assisted methods. Mechanism analyses show that SARC produces more concentrated textual variations, reduces dictionary-level feature interference, and improves residual-visual-complement alignment for most samples. These results demonstrate that effective cross-modal compensation requires textual residuals to be selectively represented and explicitly calibrated, rather than directly treated as visual complements. Code and implementation details will be released upon publication.

0.26

Supervised Source Separation of Electric Vehicle Interior Noise Using a Hybrid Transformer Architecture

cs.SD cs.LGeess.AS ·2026-10-23 ·Philipp Diel, Zhuo Cao, Arne Hildenbrand, Marco Günther, Stefan Pischinger +2 more
Abstract

The interior sound of an electric vehicle is a superposition of several distinct noise contributions - wind noise, tire noise, powertrain (electric drive unit and inverter) noise, and auxiliary-drive noise - whose relative balance governs both the overall loudness and the perceived sound quality in the vehicle interior. Decomposing a recorded vehicle interior signal into these source contributions is a prerequisite for targeted acoustic engineering, but requires effortful measurement campaigns as well as expertise in the field of transfer path analysis. This work formulates the decomposition as a supervised, binaural source-separation problem with four target source contributions and adapts a hybrid time-spectrogram transformer architecture, originally developed for music demixing, to the domain of sound source separation for technical systems. To disambiguate source contributions, conditioning is introduced in the form of operating-state information, including electric drive speed, electric drive load, vehicle speed, and auxiliary-drive speed. Since isolated real-world reference stems are costly to obtain, the model is trained on a physics-based parametric data generator that synthesizes labeled component sounds from randomized virtual-vehicle hardware configurations and driving scenarios, so that the ground-truth contribution of each source is known by construction. Validation on both synthetic and measured vehicle interior sounds probes the achieved separation performance and underscores the feasibility of applying the proposed method for acoustic contribution analysis in technical systems.

0.26

"Not" is Not Lost in Translation: A Causal Analysis of Negation Representations in Multilingual LLMs

cs.CL cs.AI ·2026-10-23 ·Giang Son Nguyen, Wray Buntine, Dung D. Le, Wenya Wang
Abstract · 29 pages, 20 figures

Negation occurs in every language, yet its surface realization differs drastically. Despite this, multilingual large language models (LLMs) translate negation robustly, raising the question of whether they share an underlying representation of negation across languages. We study this question in six open models across English, French, Vietnamese, and Chinese, using causal interventions on a synthetic contrastive dataset and a negation translation dataset derived from FLORES-200. Across models, we find strong evidence that negation representations are causally controllable and largely shared across languages. In particular, we show that a contrastive activation direction extracted from one language can induce or remove negation in the others, and directions extracted from different languages are substantially aligned. We further find that negation is represented in sparse features: a single sparse autoencoder (SAE) feature per model suffices to control polarity across languages, while responding selectively to negated inputs. Finally, both dense and sparse interventions generalize to four additional languages and FLORES-derived sentences. We release data and code to support future work: https://anonymous.4open.science/r/negation-translation/.

0.26

Towards Effective and Transferable Physical Camouflage against Multi-View BEV-based 3D Perception in Autonomous Driving

cs.CV ·2026-10-23 ·Linye Lyu, Jiawei Zhou, Daojing He, Yu Li
code/data
Abstract

Modern autonomous driving systems widely adopt multi-view Bird's-Eye-View (BEV) based 3D perception models due to their superior performance. Despite their success, the robustness of these models against adversarial camouflage attacks remains largely unexplored. Existing camouflage attacks primarily target single-view 2D detectors, limiting their effectiveness in multi-view 3D perception settings. To address this gap, we propose BEVCA, a novel adversarial camouflage generation framework tailored for multi-view BEV-based 3D perception. Our framework integrates a BEV-feature-based adversarial loss with a multi-view neural rendering module, enabling effective and transferable camouflage attacks across different models and tasks. Extensive experiments in both digital and physical settings demonstrate that BEVCA outperforms state-of-the-art baselines and exhibits strong robustness under real-world conditions. Our code is available at https://github.com/WAVE-AT-IDEA/BEVCA.

0.26

UAE GovTrust Scope Filtering Gains and Routing Failures in Bilingual Public Service Retrieval

cs.IR ·2026-10-23 ·Aleksandar Mihajlovic
Abstract · 14 pages, 2 figures, 5 tables. Reproducibility materials included as ancillary files

Public-service queries can mention several locations while only one controls the requested service. We study this distinction in UAE GovTrust, a controlled English-Arabic retrieval diagnostic comprising 22 authored service summaries and 50 scenarios with 110 query variants. Twelve methods separate lexical and dense ranking, rank fusion, scope filtering, a multilingual cross-encoder, and direct service lookup. On 72 matched answerable queries with bilingual evidence, role-filtered reciprocal rank fusion retrieves a relevant first result for 69 queries (95.8%), versus 47 (65.3%) for unfiltered fusion; filtered dense retrieval and direct lookup each achieve 68 (94.4%). With English-only evidence, excluding zero-score lexical candidates raises filtered fusion from 52 to 64 correct first hits. However, the unchanged router selects the correct action and scope for only 2 of 168 post-audit perturbations covering paraphrases, irrelevant facts, negation, and corrections. These perturbations are deliberately authored diagnostics, not held-out evaluation. Fresh-environment execution reproduces all 1,320 original ranking orders and actions. The results show that correct scope metadata can improve ranking while brittle scope inference can remove the evidence needed to answer. The study evaluates retrieval and routing, not generated answers or legal correctness; independent Arabic and domain validation remains necessary.

0.25

Reading Between the Layers: Cross-Modal Alignment and Representational Hierarchy in Vision Transformers

cs.CL ·2026-10-23 ·Sugam Karki, Raúl Vázquez, Yova Kementchedjhieva
published
Abstract · Accepted to AACL 2026 Findings

While large vision-language models excel at optical character recognition, it remains unclear how their underlying vision encoders process text and its meaning, and whether they support anything resembling reading comprehension. We systematically study transformer-based vision encoders, using cross-modal alignment between rendered text and visual concepts together with probing and retrieval-based analyses as a controlled window into these broader capabilities. Our layer-wise analyses span orthographic, grammatical, and semantic organization, and reveal that strong cross-modal mapping emerges only in the deepest layers, after unimodal object and text recognition have already developed, coinciding with a progression from surface-level to semantic representations. Although this hierarchy is strongest in contrastively trained models, larger self-supervised models and text-rich training data partially support similar behaviors without explicit language supervision. Together, our results suggest that increasing model and training-data scale, as well as text-rich training data, can partially compensate for the absence of language supervision.

0.25

SREdit: Scene Response Guided Appearance Editing for Projection Mapping

cs.CV cs.MM ·2026-10-23 ·Yuqiao Xia, Yuchen Deng, Ziheng Chang, Bingyao Huang
Abstract · 28 pages, 65 figures

Projection-based appearance editing changes selected visual attributes of a physical object through projection mapping (PM). To display a desired appearance, PM typically uses projector compensation, which computes a projector input that counteracts the geometric and photometric distortions of the scene. Compensation can only approximate this appearance within the available projection range, and it cannot revise the appearance itself. Pretrained LLM-based image editors provide a direct way to produce the desired appearance from a text prompt or reference image; projector compensation can then attempt to reproduce the edited result through projection. However, these editors do not have an accurate measure of the scene geometry or scene response. Therefore, their outputs may contain geometric misalignment, inconsistent shading, or changes outside the available projection range, which downstream projector compensation cannot correct. To address this problem, we introduce SREdit, an efficient response-guided framework for projection-based appearance editing. SREdit estimates a scene response map from the reconstructed scene that reflects how strongly projected illumination can affect different surface regions. SREdit uses the source appearance to constrain the early sampling process and preserve registration with the fixed physical object, while the scene response map adapts the amount of editing across surface regions. Experiments across two pretrained image editors show that SREdit improves projected appearance quality and reproduction fidelity. It adds less than one second per image with negligible VRAM overhead. Participants also consistently preferred the results produced with SREdit in our user study.

0.25

Polyp-VSSD: A Pre-trained Transformer with Local-Global Joint VSSD Decoding for Polyp Segmentation

cs.CV ·2026-10-23 ·Xiaobo Zhu, Shiwei Wu, Yu An, Yuechen Tao, He Sun +2 more
Abstract

Early detection of colorectal polyps is essential for preventing colorectal cancer, making accurate polyp segmentation a key component in computer-aided diagnosis systems. However, existing endoscopic segmentation methods often exhibit limited representational capacity and insufficient integration of global and local contextual information. To address these limitations, we propose Polyp-VSSD, a novel local-global joint decoding framework built upon a pre-trained Transformer backbone for high-precision polyp segmentation. Polyp-VSSD leverages the strong feature expressiveness of large-scale visual pre-training and unifies global semantic modeling with local fine-grained reconstruction to enhance segmentation fidelity. Extensive experiments on multiple public endoscopic datasets demonstrate that our approach significantly outperforms state-of-the-art methods in both Dice and IoU. On the EndoScene dataset, Polyp-VSSD achieves mIoU and mDice scores of 0.903 and 0.946, respectively, highlighting its robustness and strong generalization capability. This work presents an effective and scalable solution for intelligent endoscopic image analysis.

0.25

Investigating the trackability of landmarks in NBI and WLI

eess.IV cs.CV ·2026-10-23 ·Jonas Below, Piermarco Gerli, Anna Jötten, Dennis Eggert, Axelle Felicio-Briegel +3 more
Abstract · Accecpted at CURAC conference 2026

Reliable visual tracking of anatomical landmarks in endoscopic video is a key prerequisite for computer-assisted interventions in laryngeal surgery. While Narrow Band Imaging (NBI) is clinically established for enhancing vascular contrast, its effect on automated tracking relative to conventional White Light Imaging (WLI) remains unexplored. We present a quantitative comparison of landmark trackability in NBI and WLI using a forward-backward error framework applied to 61 endoscopic videos from 39 patients. Two regions of interest are evaluated: the vocal folds (VF) and mucosa with a prominent vascular pattern (MPVP). We compare a classical feature-based tracker (CSRT) against learning-based point trackers (CoTracker3-Online and TrackOnR) across both modalities. NBI consistently reduced the forward-backward error in both regions, reaching statistical significance for the vocal folds, while failure rates stayed low. These findings indicate that NBI can improve tracking robustness in laryngeal endoscopy, with implications for motion analysis, navigation, and robotic assistance.

0.24

Multimodal Generative AI Assists Comprehensive Echocardiographic Interpretation

cs.CV ·2026-10-23 ·Yi Qin, Marawan Elbatel, Shanshan Song, Taoran Huang, Jiarong Guo +26 more
Abstract

Echocardiography is the cornerstone of cardiovascular diagnostics, yet its global utility is constrained by a critical shortage of experienced interpreters. Current artificial intelligence (AI) solutions remain limited, typically operating as opaque "black boxes" or focusing on narrow tasks that fail to capture the holistic, guideline-driven reasoning of a clinician. Here we introduce EchoChat, a generative multimodal vision-language model that performs comprehensive echocardiographic analysis to support clinician interpretation. By emulating the cognitive workflow of a cardiologist, EchoChat integrates vast medical knowledge with a chain-of-thought reasoning mechanism to explicitly ground its conclusions in visual evidence. Trained on echocardiography-instruction pairs from 46,358 examinations, our model establishes a new state-of-the-art across diverse cohorts from six international medical centers. It supports the broadest diagnostic scope of 33 cardiac conditions and findings, including 15 newly added conditions for which it achieves an average F1 of 73.7% across available external evaluations. It surpasses existing best-performing foundation models in cardiac condition recognition (improving average recall by 13.54% on shared tasks) and quantification (average R2 score increase of 0.275 on shared tasks). Furthermore, in a source-blinded echocardiographer reader study, echocardiography reports co-generated with EchoChat were rated higher in diagnostic accuracy and completeness by two senior echocardiographers, and improved diagnostic performance from 72.78% to 82.75% in terms of F1 score over routine-reader only reports. These results demonstrate that integrating clinical reasoning into multimodal AI can democratize specialist expertise and transform the standard of cardiovascular care.

0.24

Each Insufficient in Its Own Way: Calibrating Language Models Across Varying Evidence States

cs.CL ·2026-10-23 ·Jiayi Li, Dennis L. Shung, Bradly C. Stadie
Abstract

How can a language model know whether it has enough evidence to answer? Language models tend to answer even when the evidence they have is insufficient. Deciding whether to answer or gather more evidence requires a calibrated estimate of how likely an answer is to be correct with the available evidence. However, most calibration work treats the input as fixed, without asking how this estimate should change with the evidence. We introduce evidence-state calibration, which requires a model's confidence before answering to match its probability of answering correctly and to change accordingly whenever the evidence changes this probability. We propose SuffiCE (Sufficiency Confidence Estimation), which fine-tunes the confidence already present in a model's next-token distribution to match how often the model answers correctly from each evidence state. Across clinical diagnosis and multi-hop question answering with four models from two families, SuffiCE reduces the Brier score by up to 60% and increases AUROC by up to 0.16. Its confidence matches the empirical success rate and responds not only to the amount of evidence but also to which evidence is present, showing that models can learn to judge whether their current evidence is enough to answer.

0.24

IAR: An Interaction-Aware Approach to Post-Seek Recovery in Adaptive HLS Streaming

eess.IV cs.MM ·2026-10-23 ·Mahdi Naseri Majd, Amirfarhad Farhadi
unusual
Abstract

After an unbuffered seek, an adaptive bitrate (ABR) player has little forward buffer at the new position, while its controller may retain a bandwidth estimate from earlier downloads. An expensive first request can delay resumption; later requests may interrupt playback while the buffer rebuilds. We introduce Interaction-Aware Recovery (IAR), a temporary layer around the hls.js ABR controller. After an unbuffered seek, IAR discounts the inherited estimate (γ=0.15) to set a representation ceiling. Normal adaptation continues below it until 8 s of forward buffer has been rebuilt. In a 200-run paired study spanning four network profiles, ten seek positions, and five strategies, a conservative first request reduced resume delay by a paired median of 2.008 s relative to standard hls.js. Keeping a low ceiling through recovery reduced later interruption under stronger degradation but lowered early presented bitrate. IAR and Fixed-Low-Recovery both used L0 in this campaign, so the results did not show an additional benefit from deriving the ceiling from inherited bandwidth. In 60 additional runs under a fixed Moderate profile, we reset the estimator to six controlled scalar priors. IAR's ceiling changed from L0 to L1 and L2 at the predicted thresholds. Resume and recovery times increased at those transitions; presented bitrate varied non-monotonically and was highest at L2. No stalls occurred. These reset-created priors do not reproduce complete, naturally evolved exponentially weighted moving average (EWMA) histories. The results show the benefits and presented-quality costs of conservative post-seek recovery, without establishing an advantage for a history-derived ceiling over a fixed low ceiling.

0.24

Separating Angular Alignment from Radial Ordering in Hyperbolic Vision-Language Models

cs.CV ·2026-10-23 ·Mingyang Gao, Yuwei Wu, Shiyu Luo, Wei Wu, Xiaomeng Fan +6 more
Abstract · 34 pages, 12 figures

Hyperbolic vision-language models are well suited to modeling the hierarchical structure of multimodal data, because hyperbolic volume grows exponentially with radius, enabling low-distortion representations of tree-like taxonomies from general to specific concepts. However, existing pretraining objectives for hyperbolic vision-language models entangle semantic alignment with entailment modeling, forcing radial coordinates to simultaneously encode semantic matching and hierarchical structure, leading to conflicting geometric constraints. In this paper, we introduce AURORA, a hyperbolic vision-language framework that decouples the objectives through angular alignment and a novel Radial-Angular Order Cone (RAOC) for entailment modeling. Angular alignment captures semantic similarity while preserving the radial structure for representing hierarchies. Unlike conventional entailment cones, RAOC determines angular allowance from relative parent-child radial gaps. It is incorporated into hierarchical entailment contrastive training to capture partial-order relations. We further introduce a norm regularization term that prevents undesirable radial degeneration for stable training. We train models on millions of image-text pairs and evaluate them across zero-shot classification, image-text retrieval, hierarchical reasoning, and compositional understanding. Our model consistently outperforms strong Euclidean and hyperbolic vision-language baselines, demonstrating the effectiveness of decoupling semantic alignment from hierarchical entailment modeling.

0.24

NUViT: Vision Transformers for Fourier Space Data

cs.CV cs.LG ·2026-10-23 ·Jeffrey Gu, Robert C. Heeter, Ellen D. Zhong
unusual
Abstract

Vision Transformers (ViTs) have demonstrated strong performance in many computer vision tasks but do not naturally extend beyond the natural image domain. One important example is Fourier-space data, which is used to represent a rich and diverse set of modalities, including medical imaging and microscopy. Standard ViT tokenization partitions an image into square patches on a Cartesian grid, which poorly fit the frequency and phase structure of Fourier-space data. We introduce Non-Uniform ViT, a simple, general-purpose vision transformer that redesigns image tokenization, data representation, and positional encoding to respect this structure. Experiments on Fourier-domain image classification, 2D crystal lattice identification, MRI annotation, and cryo-EM protein heterogeneity analysis show that NUViT consistently captures the structure of Fourier data better than standard architectures.

0.24

Easy-3Dfy: Object-Level Geometric Composition for Single-Image 3D Scene Construction

cs.CV cs.AI ·2026-10-23 ·Ziang Zhang, Wenyou Yi, Zehan Wang, Jiachen Xu, Xintong Han +2 more
Abstract

Single-image 3D scene construction is a fundamental problem in 3D vision. Recent methods such as SAM-3Dfy enable end-to-end reconstruction of 3D assets and their scene layouts, but still struggle with fine-grained scene geometry and asset quality. Meanwhile, specialized 3D object generators provide detailed object geometry, while 3D foundation models offer strong and generalizable spatial understanding. We propose Easy-3Dfy, a compositional workflow that decouples object generation from scene layout estimation, allowing specialized models to fully exploit their respective strengths. An RPF-based aligner then bridges these capabilities by establishing geometric correspondence between object-local meshes and partial scene pointmaps, recovering object layout for scene composition. Without requiring complete scene-level supervision, the proposed workflow consistently outperforms SAM-3Dfy across multiple geometric metrics on Scene-Maker and 3D-FUTURE200. Our results demonstrate that explicit geometric composition can effectively combine high-quality object generation with strong spatial understanding, while enabling the workflow to benefit independently from advances in both 3D object generation and foundation models.

0.24

KrishokTech: A Bengali Agricultural Advisory System with Explicit Authorization Controls

cs.CL ·2026-10-23 ·Khan Raiyan Ibne Reza, Sanjana Aktar Maria, Shakil Ahmed, Faiyad Hossain, Sumaiya Tabassum Nimi
unusual
Abstract

For agricultural advisory, generating an answer is only part of the problem: the system must also control unsafe requests, work with limited connectivity, and accommodate changing guidance. KRISHOKTECH is a Bengali agricultural advisory system built around an explicit authorization boundary around generation. Before retrieval, normalization and rule-based admission checks refer policy-covered high-risk requests and ask for a crop when treatment intent is underspecified. The system then binds retrieval to a single crop and asks for confirmation when the text and image indicate different crops. After generation, a narrow verifier removes dosage claims not supported by retrieved BARI/BRRI evidence. Agricultural guidance is kept in an external evidence index that can be updated without retraining the generator, while browser-side vision and local retrieval reduce connectivity requirements. The interface makes these decisions visible as ASK, CONFIRM, REFER, and DROP, together with their evidence. On 200 farmer queries and 400 stress queries, the controls halted underspecified requests and reduced cross-crop retrieval violations from 36.5% to 0% under the tested alias filter. Mismatch detection caught 453 of 454 text-image conflicts, while the verifier caught 114 of 118 inserted dosage errors. The demonstration shows how these controls govern the path from request to delivered advice.

0.24

RVG-Bench: A Benchmark for Reasoning-Driven Audio-Visual Temporal Grounding

cs.CV ·2026-10-23 ·Yue Feng, Boshen Xu, Zihao Yue, Xiaokun Liu, Miao Yang +3 more
benchmarkcode/data
Abstract · 42 pages, including appendices. Project page: https://fyrisk.github.io/RVG-Bench/

Existing temporal video grounding (TVG) benchmarks primarily focus on localizing target video segments according to descriptive text queries of the visual content. However, real-world queries are more diverse and often require the model to first understand the user's intent before localization, such as finding visual evidence, identifying the cause of an event, or recalling a remembered moment, yet such queries remain largely overlooked. To mitigate this gap, we introduce RVG-Bench, a benchmark for reasoning-driven audio-visual temporal grounding, where every query is constructed such that the target segment can only be identified through reasoning about the query's underlying intent. We carefully construct the benchmark and rigorously verify its data quality to ensure models truly comprehend the audio-visual content. RVG-Bench comprises 1,255 videos and 1,890 query-interval pairs across eight reasoning types. We evaluate 22 models and find that the strongest, Qwen3.8-Omni-Flash, achieves only 49.6 mIoU. Through a series of controlled analyses, including oracle-grounding probes and test-time scaling experiments, we systematically examine where and why current models struggle, providing insights for future model development.

0.23

L0-Norm-Based CycleGAN for Unpaired Brain CT-to-MR Synthesis

cs.CV math.OC ·2026-10-23 ·Yunxiao Zhang, Shouqiang Du, Zexian Xu, Chuanlong Xu
Abstract · 13 pages, 2 figures

Computed tomography (CT) and magnetic resonance (MR) imaging provide complementary information for the brain, but acquiring both modalities is not always possible. The cycle-consistent adversarial network (CycleGAN) enables unpaired CT-to-MR synthesis, but because the standard L1-norm cycle?consistency penalty grows linearly with the residual magnitude, the network tends to focus on large residuals. This paper introduces an L0-norm-based cycle?consistency term, in which the L0-norm is approximated by a smooth and bounded function. This approximation not only increases sensitivity to small residuals but also effectively suppresses the influence of large residuals. The proposed model and an otherwise identical L1-norm baseline were trained on 102 images and evaluated on 9 test images. The proposed model achieves a lower mean squared error (0.02746 versus 0.03464, a 20.7% reduction), a higher mean peak signal-to-noise ratio (15.80 versus 14.77 dB), and a higher mean structural similarity index (0.5689 versus 0.5407). It also has a lower maximum MSE and higher minimum PSNR and SSIM than the baseline. These results indicate that modifying the cycle-consistency penalty can improve reconstruction accuracy on this dataset.

0.23

STROKE: Simple Sparse Structural Occlusions for Black-Box Adversarial Attacks

cs.CV cs.CRcs.HC ·2026-10-23 ·Ankit Gupta, Akshat Pandya, Paul Lin, Christoph Adami, Emily Dolson
Abstract · 13 pages, 4 figures, 4 tables, and 16 pages of Supplementary Information with 3 figures and 5 tables

We introduce STROKE, a query-efficient score-based black-box adversarial attack that constructs perturbations from simple oriented rectangular occlusions. STROKE optimizes these perturbations through a lightweight greedy add-and-refine procedure. Across diverse image classifiers, STROKE achieves high attack success with low query counts at nominal 1--2\% area settings, showing favorable performance relative to existing patch-based attacks in several configurations. Matched comparisons further demonstrate substantial reductions in wall-clock attack time, while ablation studies examine the key design choices and effectiveness across structural primitives. These findings suggest that simple structural occlusions offer an effective and computationally efficient search space for black-box adversarial attacks.

0.23

Analysing Austrian Parliamentary Speeches by Predicting Party Affiliations

cs.CL cs.AIcs.CYcs.LG ·2026-10-23 ·Sarah Sulollari, Vanja Karan, Benjamin Roth
Abstract · KONVENS 2026

This paper examines speech characteristics of the Austrian National Council members (CMs). We introduce analysis methodology based on a fine-tuned BERT-Model which assigns to every speech a probability distribution over Austrian parties. These are then aggregated at the speaker and party level. We perform an extensive evaluation to ensure the quality of the generated distributions prior to further analysis. Instead of considering mis-assignments of predicted parties as a performance issue, we leverage them to derive insights about the CMs and political developments. Our findings show that parties that govern together become more similar in their speech characteristics, while, analogously, parties that are in opposition move further away from governing parties. Moreover, CMs who switch party usually adapt their speech characteristics only after the switch occurs. Finally, we publicly release a data set containing party affiliation predictions for the speeches of the Austrian National Council from 1996 to 2018.

0.23

Fully-Automated Multi-Modal Registration of Timelapse Zebrafish OCT+LS Images

eess.IV ·2026-10-23 ·Mahshid Babeli Nia, Leah A. Lewis, Mohammad Mobarak Karim, Manmohan Singh, Arne C. Lekven +3 more
Abstract

Optical coherence tomography (OCT) and light sheet (LS) fluorescence microscopy provide complementary views of tissue morphology and molecular expression, offering a powerful approach for studying embryonic development. However, longitudinal integration of OCT+LS images is challenging because the modalities differ substantially in contrast, spatial resolution, field of view, and visible anatomical features, while embryo growth and movement introduce additional misalignment over time. These challenges stem from both the scale and multimodal nature of the OCT and LS image stacks. We present a fully automated three-dimensional rigid registration framework for aligning OCT and LS volumes of developing zebrafish embryos acquired from 21 to 48 hours post fertilization (HPF). The method does not require pre-existing segmentations or cross-modal feature correspondences, which are often unreliable because OCT and LS capture fundamentally different tissue properties. Registration performance was evaluated across developmental time points using mean target registration error (mTRE) and compared across multiple similarity metrics and optimization methods. Normalized cross-correlation combined with particle swarm optimization produced the most accurate and consistent registration, achieving the lowest mTRE at every evaluated time point compared with the alternative methods. The registration establishes spatial correspondence between OCT tissue morphology and LS fluorescence signals throughout development, enabling integrated longitudinal analysis of structural and molecular changes in living embryos.

0.23

Computational imaging through turbid, bubbly and flowing water near the telescope

cs.CV physics.optics ·2026-10-23 ·Alexander Ingold, Sabina D. Menon, Aadhi Umamageswaran, Alec Ikei, May H. Hlaing +2 more
dataset
Abstract

Cameras in the surf zone or behind ship wakes look through turbid, bubbly, moving water close to the lens, yet restoration networks are usually scored only on displayed images. We introduce a measured dataset of a telescope viewing displayed images and physical objects through a 25~mm water layer 45~mm in front of its objective, under seven combinations of flow, bubbles and turbidity, with clear-water recordings as a second reference. Because the suspension scatters light far outside the 1.3\textdegree{} field of view, turbidity mainly removes contrast, 4.4- to 4.8-fold on a Siemens star. Restoring the mean of eight frames improves structural similarity (SSIM) by 0.032 over a single frame, and learned fusion of the frames adds little. One raw frame supports five-category recognition (96.8\%) as well as restored images. Adding synthetic training pairs leaves display-image scores unchanged but raises the correlation of a restored Siemens star with its clear-water recording from below 0.5 to above 0.9. These networks also improve ordinary underwater photographs without retraining, where training on measured pairs alone fails. Scoring against the digital source or a clear-water recording reverses the ranking of raw and restored images. These results support a simple pipeline: average the frames, classify raw frames directly, and restore with networks trained on measured and synthetic pairs.

0.23

Multi-Agent Retrieval-Augmented Causality Assessment (MARCA) of Individual Case Safety Reports in Pharmacovigilance

cs.CL ·2026-10-23 ·Daisy Rae Lopez, Angela Boccia, Maurizio Sessa
Abstract

Introduction: Increasing volumes of individual case safety reports (ICSRs) challenge expert-led causality assessment in pharmacovigilance. We developed MARCA (Multi-Agent Retrieval-Augmented Causality Assessment), a four-agent framework for Naranjo-based assessment, and evaluated agreement with human experts and exploratory differences between non-RAG and RAG-labelled scores. Methods: MARCA combined scientific literature retrieval, drug-label assessment, objective-evidence classification, and clinical causality reasoning using Gemini 2.5 Pro and DeepSeek-V3.2. Agent-generated rationales informed human-assisted scoring. A consensus expert reference standard covered 215 ICSRs and 1,562 drug-event combinations (DECs) from FAERS (Q1-Q2 2025). Agreement was measured using percentage agreement and Cohen's kappa. Results: Question-level agreement was 96.8% for Naranjo Q2 (kappa = 0.491; 95% CI 0.265-0.663), 85.4% for Q5 (kappa = 0.269; 95% CI 0.180-0.340), and 96.8% for Q10 (kappa = 0.922; 95% CI 0.885-0.951). Final causality classification agreement was 79.4% (95% CI 71.8-85.3; kappa = 0.587, 95% CI 0.450-0.690). In a retrospective subset of 60 DECs, agreement was higher for RAG-labelled than non-RAG scores for Q5 (83.3% vs 65.0%) and Q10 (100.0% vs 18.3%). Discussion: MARCA-assisted assessment showed substantial variation in agreement across Naranjo questions, with lower concordance for alternative causes. Retrospective RAG comparisons are exploratory and do not establish prospective effectiveness.

0.23

FlexA-Mem: Budget-Adaptive Agentic Memory for LLM Agents

cs.CL ·2026-10-23 ·Wujiang Xu, Jiaojiao Han, Xi Zhu, Qizheng Zhang, Kaiqu Liang +3 more
code/data
Abstract

Large language model (LLM) agents work in scenarios with very different memory needs. A question about past conversations needs the exact facts it names, while a multi-step task needs a trace of the steps already taken together with details that the next step cannot name. Existing memories are each designed and tested for one kind of scenario. Retrieval memories such as A-Mem miss steps that the current query does not express, summaries drop exact strings, and pairing the two fixes the split of the budget in advance. We propose FlexA-Mem, a general memory that writes each unit of history once into an immutable store and renders a fresh memory for every call. The rendered memory is a compact timeline of records plus detail that a fixed rule fits into the remaining room without an LLM call. The detail is either relevant units as raw text or excerpts or the verbatim pieces that best match the query. With one store, FlexA-Mem serves both kinds of scenario. On AppWorld challenge tasks at 4k, it outperforms the second-best method by 11.5%, and on 147 unused tasks with predictions fixed in advance it outperforms A-Mem with a summary by 12.2%. On LongMemEval and LoCoMo, it matches a dedicated keyword retriever that writes no memory, while memories built on text written by an LLM fall far behind at tight budgets. Ablations show that the timeline carries AppWorld completion, whereas recovery mainly saves steps there and decides accuracy on exact facts. Code is available at https://github.com/WujiangXu/Flex-Amem

0.22

ACVA-Bench: Beyond Prompt Following in Abstract Concept Visualization Ability for Text-to-Image Models

cs.CV cs.AI ·2026-10-23 ·Mengyan Wang, Jiyuan An, Yining Liu, YaoWen Zhang, Liner Yang ang Erhong Yang
benchmarkhot direction
Abstract

Text-to-image (T2I) models have achieved remarkable progress in image quality and instruction following, prompting growing research interest in their capacity for deep understanding that goes beyond literal prompt matching. However, without an explicit visual scheme provided, it remains unclear and has not been systematically evaluated whether T2I models can autonomously translate abstract concepts lacking a fixed visual referent, such as "loyalty," into visual representations that are accurate, diverse, and distinguishable from closely related concepts. We define this capability as Abstract Concept Visualization Ability (ACVA) and introduce ACVA-Bench, a category-balanced benchmark grounded in psycholinguistic norms of abstract concepts, which examines models' ACVA along three dimensions semantic sufficiency, representational diversity, and conceptual distinctiveness under a low-intervention setting that withholds concrete objects, scenes, or visual metaphors. Since abstract concepts lack a unique reference image, conventional evaluation metrics that rely on reference-image matching are difficult to apply. To address this, we propose a fine-grained visual-semantic evaluation method constrained by concept semantics, which decomposes concept semantics into independently verifiable visual-semantic questions to assess the semantic representation quality of generated images, with its validity further confirmed by human evaluation. Benchmarking results on mainstream T2I models show that these models rely on a small set of prototypical visual patterns, leaving room for improvement in representational diversity and conceptual distinctiveness for abstract concepts, thereby providing an interpretable and reproducible evaluation foundation for future research.

0.22

UniAction: Unified Action Conditioning for World Models

cs.CV ·2026-10-23 ·Bowen Xue, Zizhang Li, Hong-Xing Yu, Jiajun Wu
code/data
Abstract · Project page: https://uniaction.github.io/

Actions that condition world models are highly heterogeneous, and thus existing action-conditioned world models are typically built for individual action types, fragmenting already limited action-annotated data across separate models. We introduce UniAction, a unified action-conditioning framework that represents any action as a sparse set of action elements, each with a discrete semantic identity and a continuous spatiotemporal grounding coordinate. By mapping these coordinates directly into the pretrained video model's native positional space, UniAction allows heterogeneous actions to condition a shared world model within the same spatiotemporal structure. We further introduce cardinality-calibrated attention to address the bias from differing action representation granularities. We evaluate UniAction on four representative action forms: force, camera control, hand pose, and robot pose. UniAction outperforms conditioning alternatives in action following, and a single jointly trained model surpasses the action-specific experts, showing that heterogeneous action data benefit from unified model and joint training.

0.22

MusicTwins: Training Audio-Language Models to Listen like Their Persona

cs.SD cs.AI ·2026-10-23 ·Milan Liessens Dujardin, Kevin Miao
Abstract

People differ in how they perceive music, what associations they make with it, what they like about it, and how they respond to it. Computational models intended to simulate people in environments that include music should account for these differences. Although LLMs have been shown to reproduce aspects of human behavior, whether audio-language models (ALMs) can reproduce a specific person's responses to textual and musical stimuli remains underexplored. We introduce MusicTwins, a framework that combines persona summaries, personal memory, and post-training to generate person-specific responses. We construct persona representations from human writings, capturing background, views, personality, writing style, and a musical profile, and post-train Audio Flamingo Next through supervised fine-tuning and reinforcement learning. Evaluation on ten held-out personas shows notable improvements for Audio Flamingo Next in persona fidelity, including personality, writing style, and responses to music, supporting MusicTwins as a foundation for training audio-language models to listen like their persona.

0.22

Yalda-Embedding: Open Development of a Competitive Persian Retrieval Model

cs.CL cs.AI ·2026-10-23 ·Yalda Taheri, Mohammad Hassan Heydari, Armon Rasooli
Abstract

Despite advances in multilingual text representations, effective dense retrieval for Persian remains constrained by the limited availability of language-specific training resources and compact embedding models. We present Yalda Embedding, a 595.8-million-parameter Persian retrieval model adapted from Qwen3-Embedding-0.6B using more than one million query-document pairs. On eight Persian retrieval tasks from FaMTEB, the model achieves the highest macro-average among the evaluated 0.6B models, scoring 0.5488 and outperforming Qwen3-Embedding-0.6B and F2LLM-v2-0.6B on every task. It also approaches Qwen3-Embedding-4B, winning on four tasks, tying on one, and trailing by 0.0075 in macro-average. An incremental ablation study examines contrastive batch size, optimization stabilization, and relational knowledge distillation from F2LLM-v2-8B. Under the one-epoch training setup, contrastive fine-tuning alone yields a modest mean improvement over initialization and regressions on two tasks. The final distilled configuration improves all eight tasks over the non-distilled model, raising the mean from 0.4813 to 0.5488 without increasing inference-time model size. These gains include hard-negative evaluation tasks despite training without explicitly mined or supplied hard negatives. We release model weights, training and evaluation code, the pairwise dataset, and its teacher-embedded counterpart, enabling inspection of data composition and training procedures, independent evaluation, and reuse without repeating expensive teacher inference.

0.22

PROVE: Probabilistic Visual Grounding via Patch-Level Evidence for Vision-Language Models

cs.CV cs.LG ·2026-10-23 ·Mayank Nautiyal, Tianru Zhang, Li Ju, Ekta Vats, Prashant Singh
Abstract

Visual grounding with vision-language models such as CLIP typically produces a single deterministic relevance map for an image-query pair, without indicating whether the localisation is reliable or whether the referent is present at all. We introduce \textbf{PROVE} (\textbf{Pro}babilistic \textbf{V}isual Grounding via Patch-Level \textbf{E}vidence), a post-hoc evidential adapter that maps frozen CLIP patch features to per-patch Beta distributions without modifying the backbone. PROVE casts grounding as per-patch membership: each patch accumulates positive and negative evidence for whether it belongs to the referent, so a query may match zero, one, or many patches. We train PROVE with an evidential objective on soft patch-coverage targets, and define a prior-pull objective that keeps absent-referent examples near an uninformative prior, so that vacuity serves as an absence signal without a dedicated presence classifier. During inference, a single forward pass yields, in closed form, a posterior-mean grounding map, a spatial vacuity map, and an image-level absence score. We evaluate PROVE on referring-expression benchmarks: it is competitive in localisation with state-of-the-art CLIP grounding methods, its uncertainty tracks localisation error and supports selective prediction under abstention, and its image-level score detects when the referent is absent altogether. Our evidential patch-membership modelling offers a simple, drop-in mechanism for probabilistic verification in frozen vision-language grounding systems.

0.22

CityLift: Compositional 3D Urban Generation from a Single Image

cs.CV ·2026-10-23 ·Haoran Gong, Jie Feng, Chenxu Wang
code/data
Abstract

Generating editable 3D urban scenes from a single image requires detailed building assets with a coherent spatial arrangement. Occlusion and perspective ambiguity make it difficult to infer complete building geometry together with consistent relative scales and positions. We present CityLift, a framework that generates individual textured building meshes from an urban image by coupling scene-level spatial modeling with building-level synthesis. CityLift first learns building shape and placement jointly through coupled scene and instance streams, producing coarse sparse structures in a common coordinate system. We then introduce generative spatial inheritance, which maps each building's coarse occupancy and scene-conditioned 3D feature field into a local volume to guide a flow model that generates finer structure. The same coordinate mapping places the generated meshes back into the scene, preserving their spatial relationship to the scaffold. To address the scarcity of urban data pairing scene images with individual 3D buildings, we construct CityLift-5K, comprising 4,997 synthetic scenes with instance masks, building meshes, and placement annotations. Experiments on CityLift-5K and scene blocks from UrbanScene3D and InstanceBuilding demonstrate improved building geometry and scene composition. The project page is available at https://ranhaogong.github.io/CityLift/.

0.22

Angular Supervision for Visual Object Grounding in Autoregressive Vision-Language Models

cs.CV ·2026-10-23 ·Parsa Esmaeilkhani, Longin Jan Latecki
Abstract

Autoregressive vision-language models (VLMs) struggle to ground generated text in the correct image regions, and existing explainability methods, such as attention maps and Logit Lens, often produce poor spatial explanations due to weak text-to-vision attention and sink tokens. We propose \textbf{Angular Relevance} ($R_L$), a simple visualization measure based on cosine similarity between visual token embeddings and unembedding-matrix rows. Because $R_L$ is invariant to token norm, it produces substantially sharper and more faithful heatmaps than attention maps or Logit Lens on out-of-the-box models. Without any training, $R_L$ yields competitive referring-expression segmentation (RES), outperforming prior diagnostic maps and CLIP-based training-free methods. To preserve localized object information during cross-modal mixing, we introduce \textbf{Angular Object-Aware Loss (AOL)}, an auxiliary loss that directly supervises visual token embeddings to align angularly with the vocabulary directions of their corresponding object concepts. Finetuning with AOL alongside the standard next-token prediction objective significantly sharpens object localization in heatmaps, showing that the VLM's own visual tokens can be used directly for grounding and improving RES, without training any segmentation heads and without degrading general VQA performance. We further demonstrate that the learned grounding also transfers to category-level segmentation of object categories not seen during training.

0.22

Data Quality in RVL-CDIP: A Systematic Audit

cs.CV cs.CL ·2026-10-23 ·Stefan Larson, Kevin Leach
benchmark
Abstract

RVL-CDIP is the de facto standard benchmark for document classification, and reported accuracy on it is now 98.8%. In this paper, we audit RVL-CDIP for data quality issues and find that approximately 10.5% of its 400,000 documents have erroneous labels and that 35.2% of the test set has a duplicate or near-duplicate in the training set. Identification codes on the pages are sufficient to classify at 69% accuracy, and a state-of-the-art model loses 11 points without them. Image-only models trained on RVL-CDIP lose 16 to 28 points on documents from other sources, and the choice of OCR engine changes accuracy by 4.1 to 8.3 points. We find that nearly 16,000 RVL-CDIP documents contain sensitive personal data, and produce de-identified versions of these documents. We re-benchmark 25 models, and with all defects remediated, the 13 that we retrain are 0.9 to 2.9 points more accurate than in the original condition; the most accurate of them, LayoutLMv3, is 96.5% accurate. We release the corrected corpus, the annotations and the out-of-distribution data.

0.22

ChoreoStudio: Planning What, When, and Who in Music-Driven Group Choreography

cs.SD cs.AIcs.CV ·2026-10-23 ·Jing Xu, Cunjian Chen, Qiuhong Ke
Abstract

Music-driven group dance generation must decide not only how each dancer moves, but also what formation the group should form, when it should change, and who should execute each transition. Existing methods can generate orderly and rhythmically synchronized group motion, but remain limited in formation evolution and inter-dancer interaction. In human choreography, formation transitions often align with changes in high-level musical structure, such as the progression from an introduction to a chorus or from a chorus to a climax. We present ChoreoStudio, which makes choreography planning explicit. A music disentangler first decomposes the audio into frame-level rhythmic features and segment-level structural features. Given the segment-level features and sparse labels, a frozen omni-modal model acts as a scriptwriter, producing a global semantic description that guides the Director's choreography planning. Based on this context, the Director produces three coupled interaction decisions: a centroid-relative point set specifies what formation to produce, a transition gate determines when it should change, and a mobility allocation indicates who should execute the transition. The transition gate and mobility allocation are further conditioned on the planned geometric change, coupling formation targets with their temporal and dancer-specific execution. At a finer temporal scale, frame-aligned music features condition the diffusion generator to control detailed body motion and preserve beat-level alignment. The segment-level interaction plan is injected in parallel through frame-aligned instruction tokens and cross-dancer attention biases, coordinating how the generator realizes each transition. Experiments on AIOZ-GDance demonstrate that ChoreoStudio follows the planned interactions while achieving state-of-the-art performance in motion quality, group organization, and formation dynamics.

0.22

AlphaAnything: Extending Pretrained RGB Generators to RGBA

cs.CV ·2026-10-23 ·Qijie Wang, Zhixin Xu, Zhe Li, Haowei Zhu, Honghao Qiao +3 more
Abstract

Existing RGBA generation methods commonly learn a transparency-specific representation through reconstruction, introducing an additional training stage and potentially altering the RGB decoding function inherited from pretrained models. We present AlphaAnything, a lightweight framework that extends pretrained RGB generators to RGBA directly in their original latent space, without RGBA reconstruction pretraining. AlphaAnything inserts LoRA modules into the generative transformer and attaches an alpha branch to the frozen RGB VAE decoder. Unidirectional RGB-to-alpha feature sharing allows opacity supervision to shape the generative latent while guaranteeing that, for any fixed latent, RGB decoding remains identical to that of the pretrained decoder. We further introduce Endpoint-Aware Alpha Loss (EA Loss), which derives opacity supervision from the complementary foreground--background weights used in compositing. It supports exact opacity endpoints and continuous partial transparency while emphasizing localized alpha errors. Experiments on text-to-RGBA and image-to-RGBA tasks across multiple pretrained RGB backbones demonstrate competitive generation quality, and controlled ablations validate the shared-latent training design and EA Loss.

0.22

Real Time Motion Detection in Video Inspired by Photonic Time Stretch

cs.CV eess.IVphysics.optics ·2026-10-23 ·Callen MacPhee, Yiming Zhou, Bahram Jalali
Abstract

Motion perception is fundamental to computer vision and physical AI, and plays a central role in robotics, autonomous vehicles, human-computer interaction, and video compression. Traditional approaches such as background subtraction, frame-differencing and optical flow algorithms often struggle in complex or dynamic scenes, typically sacrificing spatial resolution for computational tractability. Here we introduce MoViD (Motion estimation via Virtual dispersion and phase Detection), a physics-inspired algorithm grounded in the principles that govern photonic time stretch - a highly successful ultrafast realtime data acquisition platform. Drawing inspiration from the optical physics therein, MoViD treats video as a collection of optical time series waveforms associated with pixels. Motion of an object through a pixel broadens the time series spectrum and, after group delay dispersion, induces a temporal phase that exists only if there is motion. Spatiotemporal group delay dispersion templates, corresponding to dielectric functions, classify specific motion trajectories. Unlike conventional empirical phase-based motion-analysis pipelines that require narrow-band spatial channelization and pyramid reconstruction, the physics-based MoViD operates directly on the broadband video signal, vastly increasing its computational speed. The resulting algorithm achieves unprecedented performance in both motion characterization and motion magnification, providing full-resolution motion mapping in real time at nearly 200 frames per second at 4K resolution and 70 frames per second for 8K video - orders of magnitude faster than classical optical flow approaches or learning-based methods. Notably, such extreme performance enables real time motion analysis at ultrahigh resolution for motion detection and compensation in robotics and other autonomous devices.

0.22

Pseudo-Labels as Rewards, Not Targets: Group-Relative Policy Optimization for Semi-Supervised ASR

cs.CL eess.AS ·2026-10-23 ·Shashi Kumar, Yanis Labrak, Sergio Burdisso, Esaú Villatoro-Tello, Petr Motlicek +2 more
Abstract · submitted to ICASSP 2027

Adapting ASR to a domain with only untranscribed audio typically relies on pseudo-labeling, a strong teacher transcribes the audio and a student is fine-tuned on it. Because supervised fine-tuning (SFT) treats every pseudo-label token as a target, the student inherits the teacher's errors. We show that using the same pseudo-labels as a reward reference for Group Relative Policy Optimization (GRPO) is more robust to label noise. In the banking domain, GRPO yields 29.3% to 56.2% relative WER improvements over SFT on Whisper-generated labels, and the advantage holds across four Whisper sizes, four teacher families, three domains and four languages. While SFT reproduces almost 100% of teacher errors verbatim, GRPO passes through only 27%. Rollout statistics explain why, erroneous pseudo-label words rarely appear in on-policy rollouts, so group-relative advantages stay aligned with the clean transcript, whereas SFT forces them into the policy's support and collapses the reward variance GRPO relies on.

0.21

Agentic Memory Management Is the Next Frontier of Information Retrieval

cs.IR ·2026-10-23 ·Qingyao Ai
position
Abstract

Information retrieval has long been focusing on connecting people with information they need to complete their tasks. Recently, the meaning of this has changed dramatically as more and more AI agents now consume, retain, and act on information on a user's behalf. Consequently, information management now also takes place \emph{inside} the agent in the form of memory management. In this perspective paper, we argue that \emph{agentic memory management} (AMM), which studies the structure, analysis, organization, storage, searching, retrieval, and \emph{use} of information in an intelligent agent's memory, is a natural extension of IR research and an area in which the IR community can make distinctive contributions. Specifically, we first present a system-level framework for AMM and introduce a computational taxonomy of memory objects as \emph{constants}, \emph{functions}, and \emph{operators}, each with distinct retrieval semantics. Then, we compare AMM with established IR research and identify their difference and connections. Last but not least, we outline a research agenda covering benchmarks, candidate principles of memory utility, optimization procedures that mix parametric and non-parametric updates, and more. We hope this paper could attract more people, both inside and outside the IR communities, to think more about AMM problems and research from IR perspectives.

0.21

PASTA: Prompt-Aware Annealed Diffusion Sampling

cs.CV cs.AIcs.LG ·2026-10-23 ·Ren Ajiki, Ryugo Morita, Stanislav Frolov, Brian Bernhard Moser, Ko Watanabe +3 more
published
Abstract · 15 pages, 4 figures, 2 tables. Accepted to ICONIP2026

While Diffusion models produce high-quality images, real-world visual design requires intentionally preserving clean regions for headlines, logos, or UI elements. Existing layout-aware text-to-image methods struggle to strictly preserve reserved regions, often introducing unwanted content that complicates downstream editing. Existing approaches either manipulate attention to displace content or suppress conditioning via noise-level interventions. However, these methods sacrifice visual coherence, producing flat regions or noticeable style inconsistencies between reserved and non-reserved areas. To address this limitation, we propose a Prompt-Aware Shared-Latent Training-Free Annealed Diffusion Sampling (PASTA), a training-free framework that enforces reserved-region constraints by selectively applying semantic guidance to object-allowed regions during sampling. PASTA operates on a shared latent trajectory, where full-prompt and context-prompt guidance are fused through mask-guided sampling. An annealed schedule suppresses semantic leakage during early denoising stages while gradually restoring visual consistency in later stages. Extensive experiments demonstrate that PASTA improves reserved-region fidelity while preserving overall image realism and perceptual consistency.

0.21

The Language of Instruction: Genre and Generative AI

cs.CL cs.CY ·2026-10-23 ·James E. Dobson
Abstract

This article argues that instruction-fine-tuned language models should be understood as implementing a genre, what is proposed as the instructional-conversational genre. This genre is an amalgamation of discursive language from prior conversations between generative models and humans recursively run through these massive neural networks. It is instantiated in applications through instruction fine-tuning, the inclusion of a pretext known as a system prompt, and the formalization of inputs in mark-up languages characterized by the use of special tokens and role assignment in turntaking structures.

0.21

Do Halting Heads Really Save Compute in Looped Language Models?

cs.CL cs.LG ·2026-10-23 ·Tsung-Han Wu, Minsuk Heo, Kenny Tan
Abstract · 33 pages, 6 figures

Looped language models run each token through the same layers several times, and each run is one iteration. To save compute, an extra layer, the halting head, gives after each iteration the probability that the token stops there, so that easy tokens can stop early. Pretraining models from scratch, up to 1.3B parameters, we find two problems. First, heads trained on a mix of all iterations predict that tokens need under half the iterations, yet under the standard stopping rule tokens run almost every iteration, because this training never scores the single iteration where a token stops. Scoring every iteration instead fixes this, but in our runs it works only with a strong penalty favoring early stops. Second, savings are commonly reported as an average number of iterations per token, but when sequences are generated in batches that run until their slowest token stops, a model without a head is almost as good at that cost. The reported average can therefore overstate the saving. We recommend reporting the iterations tokens actually run, with the stopping threshold and batch size, and give a test, run before training a head, that bounds its gain.

0.21

Spectrally optimized implicit neural representations mitigate acquisition-related image quality trade-offs in intracranial 4D Flow MRI

eess.IV cs.CV ·2026-10-23 ·Oliver Welin Odeback, Nerea González-Aranceta, Sebastián Jofre, Javier Bisbal, Edward Ferdian +10 more
Abstract · 36 pages, 7 main figures, 2 supplementary figures. Under review at Medical Image Analysis

4D Flow MRI promises opportunities for non-invasive mapping of neurovascular hemodynamics, but its clinical utility faces trade-offs between resolution, acquisition time, and noise. Deep learning-based super-resolution offers a post-scan solution, but supervised approaches require paired or domain-specific data, while patient-specific physics-informed models can be computationally demanding. This creates a need for unsupervised reconstruction models that can adapt to individual acquisitions without paired high-resolution training data, while remaining robust across anatomy, disease severity, and acquisition quality. To address this, we introduce a spectrally tuned implicit neural representation (INR) framework for intracranial 4D Flow MRI, representing the velocity field as a continuous coordinate-based neural field using WIRE activation functions with tunable spectral bias. Data-driven and physics-informed variants were evaluated in nine cerebrovascular cases spanning healthy controls and moderate-to-severe intracranial atherosclerotic disease, using CFD-derived high-resolution reference fields, MR-realistic synthetic low-resolution inputs, and direct fitting to acquired in-vivo data. Cohort-optimal spectral-bias settings generalized across subjects with only small increases in relative error compared with subject-specific optima, reducing the need for case-specific hyperparameter tuning. Physics-informed regularization provided modest additional accuracy but required substantially longer optimization, whereas the data-driven INR achieved near-comparable reconstruction performance within minutes. The framework also remained robust to acquisition-realistic noise and moderate VENC-dependent aliasing. These findings establish spectrally optimized INRs as a flexible unsupervised approach for mitigating acquisition-related trade-offs in intracranial 4D Flow MRI.

0.21

EigenGradCAM with AugSmooth: Enhancing Explainability and Precision in Multi-Stage Alzheimer's Disease Detection

cs.CV ·2026-10-23 ·Arsham Azam, Sagar Ashraf, Tayyaba Farhat, Rasikh Ali, Sheeraz Akram
Abstract

Alzheimer's disease (AD) is a progressive neurological disorder associated with memory loss, cognitive decline, and behavioral disturbances. Early and accurate diagnosis is essential, yet MRI-based deep learning models often lack interpretability, limiting their clinical use. This study presents an explainable deep learning approach for four-class AD staging using 2D axial MRI images. The method combines eigenvector-based Grad-CAM with gradient smoothing to improve saliency localization and heatmap stability. Principal component analysis is applied to gradient-weighted convolutional layers, while Gaussian noise is added to saliency weights before decomposition. To address class imbalance in the Moderate Dementia category, Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) generates synthetic samples to balance the four-stage training distribution. Custom convolutional neural networks are trained on the Best Alzheimer MRI Dataset, achieving 97.2% test accuracy, 97.0% balanced accuracy, and a Matthews correlation coefficient of 0.944. The area under the ROC curve ranges from 0.96 to 0.99 across classes. Generated heatmaps highlight the hippocampus and entorhinal cortex, regions associated with Alzheimer's disease biomarkers. These findings suggest that the proposed approach combines strong classification performance with improved model interpretability, although further clinical validation is needed.

0.21

Concept Acupuncturist: An Efficient and Precise Concept Erasure Method for Diffusion Models via Sparse Autoencoders

cs.CV cs.AIcs.CRcs.LG ·2026-10-23 ·Yiran Xiahou, Siliang Qin, Kai Chen, Shengzhi Zhang, Renyang Liu
code/data
Abstract

Diffusion-based image generation models produce high-quality images but may generate harmful or policy-violating content, motivating machine unlearning (MU) to selectively remove target concepts while preserving general capabilities. However, existing MU approaches often struggle with the distributed and entangled nature of concept representations, leading to inefficient optimization and imprecise edits when applied through large-scale fine-tuning over full diffusion trajectories. To bridge this gap, we propose Concept Acupuncturist (ConceptACP), a novel MU framework that integrates interpretability-guided analysis with inference-free fine-tuning for efficient and precise concept removal. By establishing mappings between semantic concepts and cross-attention blocks in the U-Net denoiser and localizing the most concept-relevant block for targeted optimization, ConceptACP avoids full diffusion trajectory optimization, thereby improving fine-tuning efficiency. To mitigate concept entanglement, we propose a feature-level correction mechanism based on Sparse Autoencoders (SAEs), where SAE-derived sparse, semantically meaningful feature activations are manipulated to construct targeted correction signals, which are then projected back to the output space of the corresponding block and used as supervision for precise concept removal. Consequently, ConceptACP eliminates the need for external inference-time modules while achieving efficient and fine-grained unlearning. Instantiated on SDXL-Turbo, ConceptACP improves the forgetting success rate by 36.34 percentage points in the standard one-step generation setting on average over the best-performing baseline, while preserving image quality and prompt fidelity, and generalizes effectively across other Stable Diffusion backbones. Our code is available at https://github.com/Eran-Xia/ConceptACP.

0.21

Natural Language Autoencoders Identify Without Explaining

cs.CL ·2026-10-23 ·Kristina Šekrst, Barbara Geld
hot direction
Abstract · ISPSM 2026

Natural Language Autoencoders (NLAs) express a model's internal activations in the form of readable text. The verbalizer turns an activation into an English explanation; the reconstructor then converts that explanation back into an activation, and the explanation is considered successful if the round trip brings back the original activation. We claim that this criterion evaluates whether the information is sufficient and does not address descriptive faithfulness. A sentence can contain sufficient information to make it possible to identify an activation even if it wrongly explains what the activation does: in the case of reconstruction, the primary aim is to accurately identify the activations and not to describe them. This issue has the same structure as the chain-of-thought faithfulness problem, in which verbal output that has been trained to go with a process is taken as a report on that process, and it parallels the dissociation between an individual's introspective report and the actual governing process as documented in human psychology. Although proposed as interpretability tools for the purpose of alignment, NLAs reintroduce, through unverified explanations, the epistemic problem that they were intended to resolve.

0.21

MiD-VLM: Detecting and Grounding Global Semantic Misinformation in Image-Caption Pairs

cs.CV ·2026-10-23 ·Tarunya Prasad, Bharatesh Chakravarthi
Abstract

Semantic misinformation can arise when an authentic image is paired with a caption that changes or distorts its narrative. Because the image itself may contain no visible signs of manipulation, methods designed to detect pixel-level artifacts can miss these misleading image-caption pairs. We study this problem using DGM4+, a dataset designed to evaluate the detection of global scene and semantic inconsistencies. We propose a multimodal model that combines image-text representations from SigLIP with visual scene features from DINOv2 through a projection MLP, and adapts Qwen2-VL-7B using LoRA. The model jointly predicts whether an image-caption pair is authentic, identifies the manipulation type, and grounds the misleading caption tokens. On the DGM4+ validation split, the model achieves an F1 score of $0.9679$ for binary authenticity classification. These results suggest that parameter-efficient adaptation of vision-language models is promising for detecting semantic inconsistencies, while also motivating further work on precise token-level grounding.

0.21

Depth-Dependent Safety Geometry in Vision-Language Models: Layer-Wise Localization and Bidirectional Refusal Control

cs.CV cs.CLcs.CRcs.LG ·2026-10-23 ·Muhammad Zeeshan Karamat, Syed Talal Hassan, Christiana Chamon Garcia
published
Abstract · Accepted at the NeurIPS 2026 Workshop

Vision-Language Models (VLMs) are widely deployed, yet the internal mechanisms behind their safety behavior remain poorly understood. We present a lightweight inference-time framework for localizing and manipulating safety-sensitive representations in frozen VLMs, without retraining or modifying parameters. On Qwen3-VL-4B, layer-wise activation divergence analysis reveals non-uniform safety geometry across depth. Causal interventions show that early vision layers support strong bidirectional refusal control through activation steering along a learned safety direction, with unsafe rejection spanning 6%-96%, whereas later layers respond more to activation magnitude scaling and exhibit a substantial safety-selectivity trade-off. Deeper layers remain well separated along the safety direction yet are largely insensitive to steering, distinguishing representational separability from causal controllability. InternVL2.5-4B shows qualitatively similar depth-dependent activation geometry despite differences in layer location and magnitude. Our results expose a white-box, representation-level refusal-bypass attack surface in VLM vision encoders and motivate inference-time stabilization of safety-sensitive representations.

0.20

Representation Learning via Composite Subspace Flow Matching

cs.CV cs.AI ·2026-10-23 ·José Afonso, Mário Figueiredo
Abstract · 27 pages, 4 figures

Distilling semantic meaning from raw image data remains a challenge in machine learning. Current self-supervised paradigms face a trade-off: joint-embedding architectures produce relevant features by aggressively discarding intra-class variance, irreversibly spatial and reconstruction. Conversely, generative models preserve spatial details but struggle to isolate semantics, as their optimization is dominated by pixel variations. To bridge this divide, we propose composite subspace flow matching (CSFM). Rather than corrupting data with unstructured isotropic noise, CSFM formulates representation learning as the variance-minimizing inversion of a deterministic degradation curriculum across spectral, color, and illumination subspaces. By bounding trajectory variance with hard temporal style truncation, a capacity-constrained encoder is driven to ignore transient style artifacts and encode persistent content. Empirical evaluations across CIFAR-10, CIFAR-100, and STL-10 show that CSFM yields competitive discriminative transferability while preserving generative synthesis across both convolutional and vision transformer architectures. By avoiding the collapse pathologies of non-generative methods, CSFM forms a smooth latent space that outperforms standard optimal transport flows on linear probing and few-shot classification, while enabling coherent synthesis and trajectory interpolation. These results confirm that steering the physical dynamics of generative flows offers a scalable paradigm for learning representations that are simultaneously linearly separable and generatively invertible.

0.20

Gaussian Episodic Memory for Long-Term Spatiotemporal Reasoning

cs.CV cs.RO ·2026-10-23 ·Ling-Chia Chen, Abdelrhman Werby, Kaan Aytekin, Kai O. Arras
benchmark
Abstract · Code and data: https://abwerby.github.io/GEMS/

Real-world environments undergo continual semi-static changes, requiring a memory representation that captures both the spatial and temporal evolution of a scene. Existing spatio-temporal memories for robots typically rely either on structured representations, which are limited to the entities and attributes captured during construction, or on image-based memories, which are constrained to previously observed viewpoints. We introduce GEMS, a framework that combines a persistent spatio-temporal 3D Gaussian Splatting (3DGS) map with an agentic reasoning pipeline. The 3DGS map is reconstructed once and subsequently updated only in regions where the scene changes, storing only these modifications while preserving visual detail across time. The agentic pipeline leverages this map to render novel views on demand, iteratively accumulating evidence to answer spatio-temporal questions and ground objects from natural-language queries. To advance future research, we introduce GEMS-Bench, a spatiotemporal question-answering benchmark focused on semi-static environments. It comprises 3 real-world and 12 synthetic environments, spanning 473 question-answer pairs. We evaluate GEMS on novel-view synthesis quality across scene updates, 3D object grounding, spatiotemporal question answering, and storage efficiency. GEMS outperforms recent baselines in 3D object grounding and spatiotemporal question answering, while achieving rendering quality on par with dedicated scene-update methods. Code and data: https://abwerby.github.io/GEMS/

0.20

SEMAC: Spherical Energy Matching Ambisonics Codec for Efficient Higher-Order Ambisonics Compression

eess.AS eess.SP ·2026-10-23 ·Christoph Hold, Jan Buethe, Jean-Marc Valin, Jan Skoglund, Toni Hirvonen +1 more
Abstract · Submitted to IEEE ICASSP 2027

Higher-order Ambisonics (HOA) is an attractive format for immersive spatial audio, but its high channel count and directional properties present significant audio coding challenges. The widely deployed Opus audio codec supports HOA through Channel Mapping Families 2 and 3 (RFC 8486), which code and transmit every Ambisonics channel, so the bitrate grows quadratically with the spherical harmonic order. We introduce SEMAC, which instead represents an Ambisonics signal by a quantized full-order directional energy envelope together with fewer CELT-coded transport channels. CELT separates a per-band energy envelope from the signal shape and restores it after quantization, and SEMAC generalizes this principle to directional signals. Since the reduced transport spans only a lower-dimensional subspace, the decoder re-imposes the envelope observed at the encoder input, restoring the directional energy distribution at full order. A MUSHRA test on fifth-order material compares SEMAC against Opus CMF~3 in matched configurations at 512 and 768 kbit/s. SEMAC with 16 of 36 transport channels is rated above full-order CMF~3 at both rates, and at equal transport channel count above order truncation. The proposed SEMAC strategy of coding fewer transport channels with the directional envelope is found perceptually preferable to coding all HOA channels at low precision.

0.20

SURE-Tagger: Structured Speech Representations for Compositional Evaluation and Targeted Data Selection

eess.AS ·2026-10-23 ·Jing Peng, Huifei Wang, Weihan Chen, Shuo Gao, Bicheng Zhang +9 more
code/data
Abstract · Code is released at https://github.com/PigeonDan1/sure-tagger

Speech conveys linguistic content, speaker characteristics, temporal interactions, paralinguistic cues, and acoustic environments. While recent speech captioning models describe these dimensions in natural language, inconsistent attribute coverage and ambiguous omissions hinder systematic analysis across recordings. We introduce SURE-Tagger, a framework that constructs semi-structured speech representations with explicit attribute values, supplementary descriptions, and abstention under insufficient evidence. By integrating specialist observations through language-model reasoning, SURE-Tagger enables speech conditions to be consistently identified and jointly queried across datasets. We further introduce Tag-Bench, combining real recordings with controlled, agentically synthesized speech scenes and verified attribute annotations. On Tag-Bench, SURE-Tagger achieves the highest scores on 25 of 29 evaluated dimensions for real speech and 24 of 29 for synthetic speech, outperforming caption-based baselines across structural and descriptive attributes. Beyond annotation, we use these representations to construct cross-dataset evaluation subsets and reveal condition-specific ASR errors. Using a separate candidate pool, SURE-Tagger-guided selection reduces the overall ASR error rate of Qwen3-ASR-1.7B by 14.3% relative to the unadapted model under a one-hour fine-tuning budget, while caption-guided and random selection increase error in the same setting. These results demonstrate how structured speech representations can connect rich annotation, compositional evaluation, and targeted model improvement.

0.20

Context or Knowledge: How LLMs Attribute Their Own Answers

cs.CL ·2026-10-23 ·Kirill Afendulev, Alexey Dontsov, Alla Chepurova, Elena Tutubalina, Anton Korznikov
Abstract · 25 pages, 7 figures, 23 tables, including appendices

Language models can answer a question from a supplied document or from their parametric knowledge. When later asked where the answer came from, what determines their source report? We study this by intervening on the model's key-value (KV) cache after the answer is generated and before the source question is asked. On a controlled factual task with four models (7B-32B), changing or removing the document's cache never changes the report. In contrast, transplanting the question-and-answer (Q+A) cache between runs with identical question and answer tokens transfers it: every eligible report follows the donor, while the recipient's document is unchanged. On the two smaller models, steering Q+A values along directions fitted before any source question reverses 98.9-100% of reports across four document conditions at a fixed strength, while norm-matched random directions flip at most 1.1%. The components of keys and values along fitted directions carry the transplant effect: copying them alone nearly reproduces the report change; holding them fixed nearly prevents it. Applied during answer generation, the same value directions also change which source supplies the answer, though more weakly and depending on the prompt. Source reporting in this setting thus depends causally on cached state produced while answering. We do not establish general reporting accuracy or a complete interpretation of this state.

0.20

Organizing Temporal Evidence in Hyperbolic Space for Video Moment Retrieval

cs.CV ·2026-10-23 ·Mengnan Liu, Sanping Zhou, Kun Xia, Jiaxuan Lin, Gang Hua +1 more
Abstract

Video moment retrieval aims to localize a query-relevant temporal segment in an untrimmed video. Existing methods typically perform video-text alignment in Euclidean space, where representations are embedded in a flat geometry. However, accurate moment localization requires organizing query-relevant evidence across multiple grounding states, where target regions, boundary transitions, and contextual distractors exhibit distinct temporal structures and relevance patterns. Such Euclidean representations lack an explicit geometry to jointly capture these relationships, motivating us to explore structured temporal evidence organization from a geometric perspective. Specifically, we propose Hyperbolic Temporal Evidence Organization (HTEO), a geometric framework that organizes query-aware temporal memory in hyperbolic space. HTEO maps temporal representations into a Poincaré ball and organizes temporal evidence through Radial Evidence Organization (REO) and Orthogonal Angular Refinement (OAR). REO adjusts temporal tokens' radial positions to encode the structural differences among temporal evidence, while OAR enhances query-conditioned semantic alignment through orthogonal rotation. The organized representation is then projected back to the Euclidean tangent space and injected into the decoder through a lightweight residual memory update, enabling hyperbolic geometric cues to be incorporated into standard Transformer decoders while preserving the original cross-modal memory. Furthermore, we introduce radial structure regularization and angular relevance supervision to explicitly guide temporal evidence organization and query-conditioned semantic alignment, respectively. Extensive experiments and analyses across multiple benchmarks demonstrate that HTEO improves temporal localization while providing an interpretable geometric organization of temporal evidence.

0.20

Learning Fine-Grained Visual Classification with Text as Privileged Information via Multimodal Distillation

cs.CV ·2026-10-23 ·Aishwarya Venkataramanan, Christian Ickler, Joachim Denzler
Abstract · 16 pages

Fine-grained visual classification (FGVC) requires distinguishing between highly similar categories by focusing on subtle, localized visual differences. Recent multimodal methods have demonstrated that textual attribute descriptions can provide valuable semantic supervision by explicitly describing discriminative object characteristics. However, these descriptions are typically unavailable at test time, limiting their direct use in deployment. We propose FG-LUPI, a Learning Using Privileged Information (LUPI)-based framework that leverages attribute descriptions during training. We train a multimodal teacher that integrates image patches and textual attributes via gated cross-attention, allowing semantic attributes to condition visual representations. We distill this knowledge into an image-only student using multi-level distillation at the prediction, patch, and attention levels, aligning global predictions, patch representations, and attention patterns with the teacher. This encourages the student to inherit attention structure learned from multimodal supervision while maintaining image-only inference. Experiments on multiple FGVC benchmarks show consistent improvements in classification accuracy over strong ViT-based baselines. Furthermore, evaluation using Expected Calibration Error (ECE) shows that FG-LUPI also produces better-calibrated predictions across all six benchmarks. These results suggest that textual attribute supervision can benefit not only fine-grained recognition performance but also the reliability of model confidence, while preserving the efficiency and practicality of image-only inference.

0.20

Adaptive Frequency Threshold Byte Pair Encoding for Mixed Natural Language and Structured Symbolic Corpora

cs.CL cs.LG ·2026-10-23 ·Pietro Di Nicco, Lucio La Cava, Andrea Tagarelli
Abstract

Modern language models rely on subwording tokenization to define the discrete units of input processing, typically using methods like Byte Pair Encoding (BPE) that treat whitespace as a fixed semantic boundary. This design is inherently limited, as many meaningful concepts---including multiword expressions and the structured environments of formal languages like LaTeX---frequently span across whitespaces. While recent superword tokenization methods attempt to address this by relaxing whitespace constraints, they often rely on greedy frequency-driven merging that fails to account for token length or the specific needs of underrepresented domain-specific languages. In this paper, we introduce Adaptive Frequency Threshold Byte Pair Encoding (AdaBPE), a new tokenization method that leverages an adaptive frequent-pattern mining approach. AdaBPE features a twofold dynamicity: it adapts minimum frequency thresholds based on both the length of a candidate token, to prevent favoring frequent short patterns, and its type, to distinguish between general and formal language patterns, using LaTeX as case in point. This allows the model to efficiently allocate vocabulary capacity to domain-specific notations, such as LaTeX, even when they represent a small fraction of the training data. We evaluate AdaBPE across five datasets spanning LaTeX and natural language, showing consistent gains in encoding efficiency and downstream language-model performance on domain-specific content without compromising general-language performance.

0.20

Collection-Date Assignment Redirects Answers in Temporal QA

cs.CL ·2026-10-23 ·Minsu Oh, Seunghui Jwa, Soo Kyun Kim, Yeo-Chan Yoon
Abstract

A collection date records when a source revision was acquired, not when its contents changed. Can assigning this metadata redirect temporal question answering despite unchanged evidence? Swapping the question day and following day between two revisions lowers later-reference exact match (EM) by 37.6-51.4 percentage points across five open-weight LLMs on 595 Wikipedia pages, replicated on 128 disjoint pages. Previous-day swaps retain 29.0-39.6-point losses on 497 pages. Seven prompt variants, including explicit ignore-collection instructions and archived-revision provenance, retain the effect. Fictional answers and English month-name question dates show that remembered answers and identical date strings are unnecessary. Matched thinking reduces Qwen3.5's penalty from 30.1 to 11.1 points, while native reasoners vary. Two API LLMs show small Wikipedia swap effects and rarely shift to earlier answers on fictional records. Large swap-induced losses concentrate around the question date; swaps between distant dates have much smaller effects. Adding fields containing one observed acquisition date lowers EM relative to omission in three of five LLMs after correction. The diagnostic isolates metadata effects from changing evidence or missing the relevant revision, and provides a practical audit for temporal QA inputs.

0.19

RADIANT: Object-Centric Feed-Forward Global Illumination

cs.CV cs.GRcs.LG ·2026-10-23 ·Sarah Zhang, Zhiyang Dou, Minghao Guo, Benjamin Tod Jones, Wojciech Matusik
benchmark
Abstract

Feed-forward neural renderers synthesize global illumination across novel 3D scenes without per-scene optimization or runtime ray tracing. Existing transformer-based renderers tokenize individual triangles, so self-attention scales quadratically with mesh resolution. Yet macroscopic radiative transport is governed by the underlying surface geometry and object arrangements rather than a particular mesh discretization; finer tessellation can thus inflate token counts without changing the underlying transport physics, motivating a discretization-agnostic and object-centric representation. We introduce RADIANT (RADiance-based Inter-Actor Neural Transformer), an object-centric architecture for feed-forward global illumination from point inputs. By representing each object with a compact budget of 32 surface-patch tokens, RADIANT decouples attention sequence length from polygon counts. In a single network, an inter-object scene transformer models 3D radiative exchange and a view decoder renders images at target camera poses. To capture light transport without materializing O(N^2) bias matrices, invariant form-factor attention embeds geometric distances and surface orientations directly into query-key dot products as SE(3)-invariant inner products weighted by learned area and material gains. For primary visibility without mesh connectivity, camera rays cross-attend scene tokens guided by depth-tested point splats in a shared spherical frame. Across a 1.03M-scene benchmark under matched model capacity, RADIANT reaches 25 dB tone-mapped PSNR in 5x fewer training steps than RenderFormer, and achieves 32.2 dB at 600k steps (+3.3 dB over RenderFormer and +7.6 dB over RenderFormer-V2). We further demonstrate the scalability of RADIANT, which completes a 512x512 forward pass of a 500-object scene in 0.83 s, with computational cost scaling with the number of objects rather than polygons.

0.19

Separating Thinking from Spelling for Efficient Generative Recommendation

cs.IR ·2026-10-23 ·Danil Gusak, Anna Volodkevich, Ivan Oseledets, Evgeny Frolov
Abstract

Generative recommenders retrieve items by writing their semantic IDs (SIDs) code by code under trie-constrained beam search, an approach now deployed on large web platforms. Standard autoregressive interfaces pay for the SID length K twice: a history of T items enters the backbone as TK code tokens, and every beam-search step runs through the backbone, so returning B candidates costs O(BK) backbone decode tokens per user. The second cost often dominates, and it grows with the backbone, the SID depth and the candidate budget. We introduce Deco (decoding from concepts), which separates reading the history from writing the identifier. The backbone reads each history item as one fused token and condenses the history once into a compact next-item concept; a light KV-cached decoder conditioned on the concept and the generated prefix runs the entire beam search, while the identifiers, the trie and the autoregressive factorization stay unchanged. The decoder approaches full-backbone accuracy with cross-entropy alone, and distillation improves it further. On seven datasets from e-commerce, movies and music streaming, Deco achieves comparable or better accuracy than both a standard generative recommender and the full-backbone teacher with fused inputs, while delivering 3.2-8.6$\times$ their throughput under a 50 ms p95 latency budget in a production-style serving pipeline. Delegating beam search makes larger candidate pools and models practical: Deco generates 1000 candidates faster than its teacher generates 100, and with a 1.5B-parameter LLM backbone, it matches the 69M-parameter teacher's throughput while achieving higher accuracy.

0.19

MI-MIDI 2: From Observation to Control with Sparse Autoencoders in Text-to-MIDI Generation

cs.SD cs.AI ·2026-10-23 ·Jakub Poćwiardowski, Mateusz Modrzejewski
Abstract · 20 pages, 6 figures, 24 tables. Atlas with audio examples: https://jpocwiar.github.io/MI-MIDIv2-Demo/

Sparse autoencoders (SAEs) can expose features of a generative model without predefined labels, but in language models the evidence used to interpret these features has proven an unreliable guide to which of them steer the output. Whether interpretations of SAE features predict control in music generators has not, to our knowledge, been tested. We develop an end-to-end protocol for discovering and evaluating SAE features in text-to-MIDI generation and apply it to text2midi and MIDI-LLM. The protocol characterizes features through correlations with musical metrics, shared instrument programs, and automated descriptions, each with its own control: note-density adjustment, matched inactive excerpts, and permutation calibration. It then tests candidates by intervention, measured directly in the generated MIDI against unmodified generation and a matched random direction, ranks them in screens, and confirms selected features on new prompts. Evidence that survives these controls points to features that steer the output. Descriptions exceed permutation controls for instrumentation, and a feature whose description names a synth-lead program introduces that program in a prespecified confirmation and in a replication on disjoint prompts and reappears in independently trained dictionaries. Features selected by density-adjusted correlations show a small positive shift of their target metric as a family, whereas features selected by raw correlations show none. An atlas provides MIDI, piano-roll, and audio examples for every feature of the targeted screen.

0.19

Matching Particle-Arrival Rates for Low-Count Image Denoising

eess.IV cs.CV ·2026-10-23 ·Huazhen Liu, Zipei Yan, Jizhou Li
Abstract

Low-count imaging arises when limited illumination or dose yields only a few detected particles in microscopy, such as photons or electrons, causing measurements to be dominated by signal-dependent Poisson shot noise. Supervised denoisers typically require paired clean and noisy observations, whereas many self-supervised approaches exploit spatial or temporal redundancy without directly estimating the underlying particle-arrival process. We introduce PARM, a self-supervised denoising framework based on particle-arrival-rate matching. By modeling particle accumulation over exposure as a Poisson counting process, PARM identifies the conditional particle-arrival-rate field with the posterior mean of the latent clean rate. Poisson thinning then splits each observation into complementary lower-exposure views that provide self-supervised rate targets for one another. In implementation, the network estimates the normalized spatial allocation of the conditional rate, while the observed total count determines its global particle mass. Experiments on synthetic data, natural-video sequences corrupted with Poisson noise, and real low-dose electron microscopy demonstrate leading or competitive performance while effectively preserving fine structures. PARM therefore provides a statistically grounded approach to low-count image denoising when clean references are unavailable.

0.19

4DMemory: Benchmarking Long-Horizon Spatiotemporal Memory with Harness-Augmented Agents

cs.CV cs.AIcs.CLcs.MM ·2026-10-23 ·Yanhao Jia, Yi Zhu, Yiran Zhong, Weigao Sun, Feida Zhu +5 more
benchmarkdataset
Abstract · 22 pages, 5 figures

Humans efficiently process multimodal streams spanning hours to years, building persistent, fine-grained spatiotemporal memories that support reasoning about and predicting long-horizon events. In contrast, existing benchmarks focus on short- and medium-context tasks, and most foundation models struggle to preserve fine-grained information over long horizons. To evaluate and improve long-horizon spatiotemporal reasoning, we introduce the LongVideo dataset and the 4DMemoryBench. LongVideo comprises about 900 hours of video and over 5,000 manually reviewed questions covering visual, textual, and spatiotemporal reasoning across diverse context lengths. We further develop the 4DMemory Harness, which integrates four complementary skills, 22+ specialized tool models, and a model-specific retrieval-augmented generation system. Evaluating 13 leading models, we find that standalone foundation models struggle with long-horizon tasks, while the 4DMemory Harness improves all of them by over 50\%. Compared to previous benchmark, the same models exhibit a performance drop of more than 10\% on 4DMemoryBench. We hope 4DMemory will facilitate the development of spatiotemporal agents capable of addressing complex, long-horizon tasks.

0.18

Sounding Surgical Instrument Localisation through Audio-Visual Correspondence

cs.SD cs.CV ·2026-10-23 ·Dominic Behling, Jonas Hein, Matthias Seibold
Abstract

Surgical scene understanding is a prerequisite for context-aware and intelligent systems in the operating room. Current methods rely mostly on visual data, although sound carries complementary information on surgical activity. This work investigates whether the active instrument can be localised from a single microphone and a camera by learning the correspondence between the appearance and sound of surgical instruments. We adapt weakly supervised audio-visual localisation, which learns from the co-occurrence of image and sound, to the surgical setting. Trained without any spatial annotation, the model predicts a heatmap over the sounding instrument. We extend its training objective by silence and area terms, which suppress the output for silence and background noise, and a composite term, which trains the model to select the sounding one of two visible instruments. For training, we record a new corpus of simulated surgical sawing, drilling, and chiselling. On an independent corpus recorded with a different camera and microphone and with optically tracked instrument positions, the model localises the active instrument, separates instrument activity from silence and background noise, and attributes sound correctly when two instruments are present. Ablations show that the silence and area terms are required for detecting activity, and the composite term for selecting between instruments. Audio-visual correspondence learned from in-domain recordings connects the sound of a surgical action to the instrument that produces it. This proof of concept introduces a novel paradigm that enables the detection and localisation of surgical activity without annotating instrument positions.

0.18

NECTAR: Automatic Negative Clue Estimation via Centroid-Space Mapping for Target Speaker Extraction

eess.AS eess.SP ·2026-10-23 ·Hyungjun Yoo, Yongjoon Lee, Jung-Woo Choi
Abstract

Target speaker extraction (TSE) aims to extract a desired speaker from a mixture using auxiliary clues such as enrollment speech. An interferer-related representation can provide a useful negative clue for TSE, but such information is not directly available from the target enrollment. To derive this representation from the observed mixture, we analyze the relationship between the speaker representations of a speech mixture and its constituent speakers. Our analysis shows that the mixture centroid lies near the midpoint of its constituent speaker centroids, whereas this relationship does not hold for single-utterance embeddings. Based on this observation, we propose NECTAR, NEgative Clue for TARget speaker extraction, which automatically derives an interferer-related negative clue from the mixture using the centroid-space relationship. We further introduce a Centroid-Space Mapper (CSM) to estimate the required centroid representations from a mixture and a target enrollment. NECTAR achieves the highest SI-SDRi on Libri2Mix among the compared systems trained on LibriSpeech train-clean-100, reaching 16.18 dB.

0.18

DS-BiCA-Net: Pair-Preserving Cross-Temporal Modeling for Remote Sensing Change Captioning

cs.CV ·2026-10-23 ·Licheng Zhang, Naveed Akhtar, Siew-Kei Lam
Abstract

Remote sensing image change captioning (RSICC) describes in natural language the semantic changes between two images of the same scene captured at different times. Previous methods commonly merge the pair during repeated processing or combine intra-image and cross-image relations in one operation, weakening source identity and leaving the two temporal directions implicit. Without separate modules for intra-image and cross-temporal processing and a separate path for each direction, the contribution of each part cannot be measured on its own, so it remains unclear which part matters most, and the design of later models has no such evidence to build on. To address these issues, we propose DS-BiCA-Net, which assigns the two kinds of processing to separate modules and keeps the two images as separate streams throughout the encoder. A windowed state-space module models context within each image, and a bidirectional cross-attention module compares the two images along two directed paths that share their projections. The two modules use the same pre-normalisation residual architecture and differ only in their mixers, and the two streams are fused only in the aggregation head, so each module and each direction can be removed and its contribution measured on its own. Controlled experiments on LEVIR-CC and Dubai-CC examine operator interaction, temporal direction and module order, showing that intra-image refinement carries no benefit alone and becomes beneficial only when followed by bidirectional cross-attention, an order that outperforms the reverse at matched parameter count and FLOPs. The bidirectional module also outperforms either temporal direction alone. Our method reaches an S_m^* of 82.65 on LEVIR-CC, the highest among the published results compared, and 66.72 on Dubai-CC without geometric augmentation, the second-highest among the published results compared.

0.18

J-Comm: Jacobian-Guided Latent Communication between LLM Agents

cs.CL cs.MA ·2026-10-23 ·Xiangwei Wang, Wei Wang, Sachith Seneviratne, Ken Chen, Yu Xia +2 more
Abstract

Latent communication requires both an effective continuous reasoning process and an interface through which an external model can leverage its intermediate computation. We present J-Comm, a framework comprising two core components. First, a training-free Geometric Agreement for Latent Reasoning (GEAR) reweights token candidates over existing model embeddings, which monotonically optimizes geometric consensus among candidate tokens in embedding space at each step. Second, we introduce an H+J (Hidden-and-Jacobian) communication interface that gives the Receiver two complementary views of the Sender's reasoning. The H channel supplies continuous inputs derived from feedback distributions, while the J channel conveys intermediate-state cues in vocabulary-indexed Jacobian-lens coordinates. Intuitively, H provides inputs for the Receiver to reason from, while J offers additional guidance within its computation. The Receiver reconstructs these cues using its own dictionary and injects them into intermediate layers during prefill, enabling communication without sharing key-value caches or requiring identical model widths. In Qwen3-8B to Qwen3-4B communication, J-Comm achieves the highest mean score across seven benchmarks, including 94.80% on MATH500, 54.55% on GPQA-Diamond, 73.33% on AIME2024+2025, and 62.93% success on ALFWorld. Compared with TextMAS, it improves performance by 1.29-8.59 percentage points and reduces generated text tokens by 19.2-74.8%.

0.17

Interpretable Singing Emotion Recognition Using Acoustic and Vocal Technique Features

cs.SD cs.AI ·2026-10-23 ·Fotini Kiriakopoulou, Vassilis Lyberatos, Spyridon Kantarelis, Athanasios Voulodimos, Giorgos Stamou
Abstract

Singing emotion recognition relies predominantly on deep learning models that achieve strong performance but offer limited insight into the vocal characteristics driving emotion, of- ten overlooking the singing voice as a distinct signal. We propose an interpretable singing emotion recognition framework using a compact set of singing-specific descriptors across prosody, spec- tral features, phonetics, and vocal technique. To decode model decisions and link predictions to measurable acoustic phenomena, we employ SHAP and apply GRAD-CAM to qualitatively assess the learned representations of vocal techniques. Evaluations show that singing-specific features reduce cross-singer variability, while vocal technique descriptors provide interpretable insights into singing expression. Furthermore, out-of-domain evaluation on VocalSet reveals interesting associations between emotional expression and ground-truth technical execution. The analysis highlights vibrato and pharyngeal techniques as salient cues of emotional delivery, with SHAP indicating that vibrato is context-dependent on articulation rate. These findings emphasize the value of domain-specific features for interpretable singing emotion analysis.

0.17

From Item Scores to Semantic Tokens: Test-Time Collaborative Adaptation for Generative Recommendation

cs.IR ·2026-10-23 ·Xiongfeng Shan, Xing Tang, Hanchong Chen, Fuyuan Lyu, Lingjie Li +3 more
Abstract

Generative recommenders using semantic identifiers predict items through successive token decisions, but beam search may irreversibly prune promising item paths. This limitation is particularly consequential when a few recent interactions signal an emerging interest while earlier behavior favors established preferences. Yet an independently trained discriminative recommender may rank the same items highly from the same interaction history. Such complementary scores are available at test time, offering a way to adapt a frozen generator without retraining either model. The challenge is to use these scores before pruning: combining completed recommendation lists cannot alter earlier beam decisions, while scores over complete items cannot directly inform choices among semantic prefixes shared by multiple items. We propose \textbf{T}est-Time \textbf{C}ollaborative \textbf{B}eam \textbf{S}earch \textbf{A}daptation (\textbf{TCBSA}), a training-free method that incorporates collaborative information into semantic decoding. TCBSA transforms discriminative score margins into potentials over shared semantic edges and uses a confidence gate to regulate their contributions to cumulative path scores before beam selection. We establish a sufficient condition for target prefix survival under global beam competition and bound score perturbations under low discriminative confidence. Full-catalog experiments show consistent improvements over strong baselines across datasets and different generative and discriminative recommenders. Analyses of emerging interests and decoding behavior demonstrate gains in recommendation accuracy and target path survival, while targeted interventions support the contribution of shared prefix corrections. TCBSA also offers a favorable balance between recommendation accuracy and inference latency compared with simply widening the native beam.

0.17

Distill, Debate, Discover: Grounding Multi-Agent Research Ideation in Graph-Distilled Insights

cs.CL ·2026-10-23 ·Qianran Ma, Muxin Liu, Grace Kim, Xiao Liang, Chris Callison-Burch
Abstract

Large language models (LLMs) are increasingly used to support research idea generation and scientific discovery. Multi-agent debate offers a collaborative approach to this task by bringing together diverse perspectives to develop research ideas. However, a key challenge is how to equip agents with distinct and meaningful expertise to guide their contributions. Inspired by how researchers draw on accumulated research experience during collaboration, we introduce InsightAgents, a framework that constructs literature-grounded agent expertise by distilling insights from graphs of related papers. Guided by insights that trace evolving research approaches, reveal underlying assumptions, and identify unresolved challenges, within a line of research, agents jointly identify a research problem and propose, challenge, and refine potential approaches through debate to develop a research proposal. We evaluate InsightAgents against three research idea generation frameworks on 100 NLP research topics through pairwise comparisons by two LLM judges. In both online and offline knowledge settings, InsightAgents achieves the highest average score across the five evaluation dimensions under both judges.

0.17

On-Policy Self-Training: Data-Free Self-Improvement in Large Language Models

cs.CL cs.AIcs.LG ·2026-10-23 ·Leon Sick, Lukas Hoyer, Yannick Strümpler, Christian Reisswig, Timo Ropinski +1 more
community pick
Abstract

On-policy self-distillation enables post-training foundation models without relying on stronger external teachers. However, existing approaches rely on ground-truth question-answer pairs, human-curated problem sets, or other supervised signals. In this work, we reformulate on-policy self-distillation as a self-training framework, driven by self-generated question-answer pairs and privileged reasoning demonstrations. We push supervision to the limit: guided solely by coarse, high-level topic descriptions, the model synthesizes its own curriculum of queries and privileged reasoning traces without access to any human-curated prompts or answers. We find that self-distillation on fully synthetic privileged information enables a model to improve across multiple domains while retaining its capabilities on domains it is not self-trained on. We demonstrate the generality of this framework across text and multimodal domains, achieving consistent self-improvement on standard language benchmarks and Visual Question Answering. Our findings demonstrate that foundation models possess rich latent capabilities that can be amplified without external task supervision or curated problem distributions.

0.17

EditJev: System One Routing for Image Editing Models

cs.CV cs.AI ·2026-10-23 ·Zhile Guan, Xu Jiang, Yule Duan, Gongbo Zhang, Shuyu Wang +2 more
Abstract · 12 pages, 4 figures, 7 tables

Instruction-based image editing models exhibit pronounced sample-specific performance variation: different models can produce substantially different results for the same source image and editing instruction, and the best-performing model often varies across requests. These models also differ substantially in inference latency, making fixed-model deployment inefficient. We therefore study image editing model routing, which dynamically selects an editor from a candidate pool given an input image and editing instruction. We introduce EditJev, a System One routing model that directly predicts sample-level preference scores over candidate image editing models in a single forward pass. A lightweight preference head on a vision-language backbone produces a fixed-dimensional score vector without autoregressive model-name generation or response parsing. To train EditJev, we construct pairwise preferences from offline candidate outputs and optimize a pairwise ranking objective to learn relative model suitability. At inference time, EditJev combines the predicted preference scores with model-level speed scores to balance editing quality and generation latency. Experiments on ImgEdit and GEdit-En show that EditJev improves aggregate quality over the strongest fixed editor, from 4.540 to 4.592 and from 8.194 to 8.255, respectively, while reducing estimated end-to-end latency by 60.8% and 31.7%. These results demonstrate the effectiveness of direct, structured model routing for efficient image editing.

0.17

From Noisy Retrieval to Decision Evidence: Silver Evidence Demonstrations for In-Context Learning

cs.CL ·2026-10-23 ·Masoud Monajatipoor, Fazlolah Mohaghegh, Hamid Reza Hassanzadeh, Akash Uttamrao Bhoite, Robert E. Tillman
Abstract

RAG and retrieval-augmented few-shot ICL can struggle when decision-relevant evidence is buried in long, noisy context. Retrieved passages and demonstrations may contain redundant or distracting information. Although expert-labeled development examples may identify the text supporting each decision, these annotations are typically used only as answer-side rationales rather than as supervision for filtering noisy retrieval. We propose silver-evidence demonstrations, which replace raw retrieved passages in ICL examples with generated, target-conditioned evidence summaries. For each labeled development example, an LLM generates silver evidence without access to the expert label or gold evidence; this evidence is then paired with the expert label and gold evidence where available. At test time, a single structured call generates silver evidence from retrieved passages and then predicts the final label and decision evidence. Across EI 2.0, SciFact, synthetic policy retrieval, and HICRIC/IMR data, our full method improves over equally sized gold-only ICL (label-only for policy) on three of four datasets and improves evidence F1 on all annotated datasets. It also achieves higher mean label accuracy than InstructRAG-style ICL on all four datasets. These results suggest that replacing demonstration-side retrieved passages with compact, target-conditioned silver evidence is particularly useful for decisions based on long and noisy text, where it reduces distracting context while preserving decision-relevant signals and expert supervision.

0.15

The Butterfly Defects: Tracing Cooperation in LLM Agent Networks

cs.CL physics.soc-ph ·2026-10-23 ·Maria Sampani, Maria Lymperaiou, Angeliki Dimitriou, Giorgos Filandrianos, Chrysoula Zerva +1 more
Abstract

As large language models (LLMs) increasingly operate as interconnected agents, a fundamental question emerges: does individual cooperation translate into collective cooperation? We investigate this question through the Iterated Prisoner's Dilemma (IPD), moving from dyadic interactions to homogeneous and heterogeneous networks of open-weight LLMs. Our findings reveal that cooperation is highly sensitive to model composition and seemingly minor linguistic changes. Models that cooperate with identical peers can shift toward defection in mixed populations. A simple uncertainty cue can increase or decrease cooperation, redistribute costs, and affect unexposed agents several hops away. Even reversing action-label order can dramatically alter behavior, with uncertainty framing sometimes amplifying this sensitivity. Together, these findings challenge the notion of cooperation as a stable model-level property, revealing how strategic behavior depends on interaction partners, network positions, and the language used to frame decisions.

0.15

Fast and Useful: Adapting Distilled Generators for Visual Learning

cs.CV ·2026-10-23 ·Patryk Bartkowiak, Dominik Michels, Sören Pirk, Wojciech Pałubicki
Abstract

Distilled text-to-image models generate images quickly, but classifiers trained on those images can perform substantially worse than classifiers trained on images from their slower base models. We adapt released distilled generators to produce better training data while retaining fast generation. We fit a low-rank update to the student's first prediction so that it reaches an intermediate state produced by its unguided base, then disable the update and let the unchanged student finish generation. The base supplies training targets and is absent at inference. Guided targets and shorter jumps give classifiers lower accuracy in our ablations, showing why the state used for supervision matters. Across three generator families, the adapted images train better classifiers than images from the released students. With class-name templates at full ImageNet-1k scale, ResNet-50 top-1 rises from 8.4% to 34.2% for FLUX.2-klein and from 12.7% to 32.1% for SD-Turbo, using two and three denoiser evaluations per image, respectively. Fixed adapters also improve results on additional classes, fine-grained datasets, and other learner architectures and training recipes. Template-based adaptation remains useful against richer-prompt alternatives, while the benefit of combining adaptation with LLM generation prompts depends on the generator. Adapting a single generation step makes fast generators substantially more useful for training visual models.

0.15

PHYRD: A Living Benchmark for Reproducing and Extending Theoretical Physics Research

cs.CL ·2026-10-23 ·Qituan Zhang, Guo Chen, Linfeng Zhang, Wenlong Zhang
benchmark
Abstract

Large language models and agents have shown significant power in advanced physical research. In this paper, We introduce PHYRD, a living benchmark for theoretical-physics reasoning with 63 Reproduce and 43 Discover problems. Reproduce tests derivation of published results; Discover tests consequences of modified theoretical conditions. FormulaForest constructs the tasks with reference solutions and subquestion-level rubrics. Across five models, whole-problem accuracy ranges from 49.21% to 68.25% on Reproduce and from 18.60% to 74.42% on closed-book Discover. With tools, four models complete more Discover problems, but conceptual errors and missing arguments persist. Besides, our experimental results demonstrate the value of FormulaForest as synthetic data. On FrontierScience, a benchmark developed by OpenAI, it raises the accuracy of a Qwen-35B-A3B model from 61% to 72%.

0.15

Profile-Driven User Simulation via Reinforcement Learning with Hierarchical Rewards for Task-Oriented Dialogues

cs.CL ·2026-10-23 ·Juan Li, Fangshu Chang, Xinyi Jiang, Ling Huang, Yunzhang Chen
Abstract

Building Task-Oriented Dialogue (TOD) systems for high-value scenarios such as financial marketing requires large volumes of domain-specific dialogue data, which privacy regulations, annotation costs, and the need for expert labor make hard to obtain. User simulation offers a remedy, but existing methods target general-domain dialogue and exhibit two key limitations: insufficient modeling of domain-specific business processes and compliance constraints, and inadequate capture of the colloquial, spoken-language character of voice interactions. We propose ProfileSim, a profile-driven user simulation framework for financial marketing outbound calls. It jointly models user profiles---demographics and Big Five personality traits---and business profiles---intents reverse-engineered from real call logs---enabling fine-grained control over simulated behavior. The base model is fine-tuned on real voice dialogue data and refined by Group Relative Policy Optimization (GRPO) with hierarchical rewards at the dialogue and utterance levels. Paired with a customer service agent integrating Standard Operating Procedure (SOP) constraints and Retrieval-Augmented Generation (RAG), ProfileSim achieves the highest colloquial naturalness, the strongest user profile adherence and the lowest agent score on long-tail intents among all evaluated user generators, while producing substantially shorter user utterances at comparable business compliance. The generated dialogue data is designed to support intent recognition training, agent evaluation, and rapid cold-start deployment.

0.14

Cycle-of-Thought: Self-Evaluative Iterative Reasoning for Table Visual Question Answering in Document Images

cs.CV cs.AIcs.CL ·2026-10-23 ·Somraj Gautam, Gaurav Harit
Abstract

Recent advances in Vision-Language Models (VLMs) have significantly improved Visual Question Answering (VQA) performance, yet their reasoning over structured content such as tables remains inconsistent, often plagued by spatial misalignment and logical hallucinations. Existing approaches for improving table VQA typically rely on architectural modifications or expensive supervised fine-tuning (SFT), which impose high barriers to entry due to the need for large-scale, expert-annotated datasets and significant computational resources. In this work, we propose Self-RefineVQA, a training-free, inference-time self-correction framework that improves table VQA performance without additional labeled data or model updates. Inspired by human revision behavior, our method enables a single VLM to iteratively act as a generator, critic, and refiner, forming a closed self-feedback loop that corrects common reasoning errors through prompt-guided re-evaluation. Experiments on tabular VQA benchmarks demonstrate that Self-RefineVQA yields consistent accuracy gains in monolingual and cross-lingual understanding, with a particularly significant boost for small models, enabling lightweight models (e.g., 2B-7B parameters) to rival the performance of significantly larger counterparts. Furthermore, we provide a comparative complexity analysis demonstrating that while inference cost increases linearly, our framework offers superior data efficiency and a lower cumulative cost by eliminating the prohibitive annotation bottleneck of human labeling required for fine-tuning. Our results highlight the potential of prompt-based self-improvement as a scalable, accessible paradigm for reliable table reasoning in VLMs.

0.14

Decoupling Patient Personas and Clinical Workflows: A Dual-Agent Approach to Realistic Medical Dialogue Generation

cs.CL cs.AIcs.LG ·2026-10-23 ·Qiyuan Li, Xiaomei Guo, Chao Gao, Haijiang Liu, Deyu Chen +2 more
Abstract

Pre-consultation medical dialogues are essential for collecting comprehensive patient information, but are costly to obtain at scale due to privacy and ethical constraints. Existing LLM-based synthesis methods often rely on prompt engineering or rigid rules, limiting their ability to balance clinical relevance with realistic patient diversity. We propose CoRPS-Dial, a controllable dual-agent framework for pre-consultation medical dialogue synthesis that integrates patient personas with guideline-aligned clinical workflow rules. CoRPS-Dial generates dialogues through interactions between a persona-driven patient agent and a process-guided doctor agent, followed by post-hoc quality control for role consistency, workflow completeness, and textual consistency. Experiments on coronary heart disease demonstrate that CoRPS-Dial improves alignment between synthetic and real dialogues in information coverage, clinical compliance, and dialogue naturalness. Downstream fine-tuning experiments further confirm the utility of the generated synthetic data for medical dialogue tasks. These results highlight the effectiveness of CoRPS-Dial for clinically grounded and controllable medical dialogue synthesis.

0.14

Complementing DINO Features with Image Structure for Part Discovery

cs.CV cs.AIcs.LG ·2026-10-23 ·Samyak Rawlekar, Nikhil Chowdary Paleti, Amey Gupta, Narendra Ahuja
Abstract

Self-supervised Vision Transformers (ViTs) like DINO encode rich semantics, enabling training-free segmentation of objects from the background and discovery of their parts via feature clustering. This paper focuses on part discovery: partitioning a foreground object into semantic parts. Existing approaches produce imprecise part boundaries. We trace this to the rigid ViT patch grid used in training DINO, where patches straddling boundaries contain pixels from multiple parts. While DINO's invariance-based training could absorb this pixel-level mixing into clean features, experiments with over 26K foreground patches show it does not: boundary-overlapping patches produce features with higher entropy and lower assignment confidence, blurring predicted part boundaries. Rather than make the massive effort of retraining DINO over geometrically coherent regions, an easier approach is to take boundaries directly from a boundary detector, which supplies the geometric information DINO's features lack. Motivated by this, we introduce Part-DINO, a training-free framework that combines boundaries from low-level image segmentation with DINO's semantic features. We use the boundaries to partition the image into regions, pool DINO features within each region, and group them to discover parts. Without any training, Part-DINO produces structurally coherent part assignments, outperforming training-free baselines across four standard benchmarks (CUB, CelebA, PartImageNet-OOD, PartImageNet-Seg) and remaining competitive with weakly-supervised and unsupervised methods.

0.13

ImageRSI: Verification-Driven Recursive Self-Improvement for Agentic Image Generation

cs.CV cs.AI ·2026-10-23 ·Shuo Han, Zhilin Wang, Peng Ling, Wenjia Gong, Yang Liu +3 more
benchmark
Abstract

Complex text-to-image requests remain difficult to satisfy reliably. Verifying and repairing an image can fix an individual output, but a successful repair does not necessarily carry over to other requests. We present ImageRSI, a framework that keeps the image generator frozen and combines three components: five-dimensional verification and repair within each task, an experience library shared across tasks, and offline updating of the judge. Candidate experiences enter the library only after paired A/B validation, and the judge is updated with supervised fine-tuning and GRPO on accumulated image-judgment records and evaluated before reuse. We also introduce ImageReq-Bench, a benchmark of 961 requests built around hard cases in real-world image-generation requests, each paired with an evaluation checklist. ImageRSI achieves a Checklist Strict Accuracy (CSA) of 52.6% on MindBench (with Nano Banana Pro as the generator) and 65.66% on ImageReq-Bench, and we compare system outcomes across judge-training configurations.

0.13

SP-Guide: Enhancing Aerial Vision-Language Navigation with Spatial Priors

cs.CV cs.AI ·2026-10-23 ·Haoxiang Tang, Dong Zhou, Shen Tan, Minxuan Jiang, Xitong Li
benchmark
Abstract

Aerial vision-language navigation (AVLN) requires a UAV to reach a language-described target from partial egocentric observations while reasoning over large outdoor scenes. Existing approaches typically fuse view-centered RGB observations and globally anchored history maps as generic visual embeddings, leaving cross-modal spatial alignment to be learned from sparse navigation supervision. We present SP-Guide, a transformer- based goal-prediction framework that addresses this weakness of alignment through a Spatial Prior Cross-Modal Encoder, an Instruction Enhanced Decoder, and a Goal Convergence Filter for terminal target refinement. Experiments implemented on the CityNav benchmark show that SP-Guide improves success rate over the representative baselines, with the largest gains on unseen environments. These results indicate that physically grounded spatial priors provide an effective inductive bias for robust AVLN.

0.13

Why Full-Duplex Speech Models Collapse in Long Conversations

cs.CL cs.SD ·2026-10-23 ·Yeonju Kim, Bella Godiva, Se Jin Park, Yong Man Ro
benchmark
Abstract · Preprint. 31 pages, 16 figures, 18 tables

Full-duplex speech models listen and speak at the same time, which makes them promising for natural, open-ended conversation. But can they sustain a coherent conversation for tens of minutes? Because they generate a token every frame, even while listening, they exhaust their context windows within minutes. Yet existing benchmarks evaluate only single turns or a few conversational rounds, too short to expose the long-horizon problem. We introduce the Full-Duplex Long-Horizon Benchmark (FD-LHB), which tests five recent full-duplex models on long replayed conversations and in live conversation with an LLM partner. Every evaluated model eventually either degrades or exhausts its context. Models that keep the full history often degrade before their context runs out, while sliding-window models still fall into repetition, silence, and gibberish. For sliding-window models, we identify two contributing mechanisms. Gibberish and silence are linked to Initialization-Anchor Eviction, where the sliding window discards the first KV entry. Repetition is also reinforced by the model's own recent outputs. Guided by this diagnosis, we apply two simple fixes: Initialization-Anchor Pinning (IAP) keeps that first entry, and Duplex Repetition Penalty (DRP) discourages echoing while guarding against prolonged silence. Together, they support the diagnosis and, without retraining, substantially reduce conversational collapse in replays of up to an hour while running in real time.

0.12

Search-Based Testing of Modality Sensitivity in Vision Language Models

cs.CV cs.SE ·2026-10-23 ·Oliver Weißl, Eduardo Breitembach, Andrea Stocco
Abstract

Vision-language models (VLMs) process images and text jointly, yet robustness evaluations often perturb these inputs separately or assess only predefined corruption levels. Such evaluations provide limited evidence of which modality is more susceptible to perturbations and whether smaller changes across both modalities can jointly expose failures. We present MoST, a black-box testing framework that characterizes modality sensitivity by comparing image-only, text-only, and joint perturbations. MoST uses multi-objective search to select and combine perturbations and adjust their severity, seeking prediction degradation while limiting changes to images and prompts. We evaluate MoST on four VLMs performing object localization on ImageNet and driving images. The models exhibit distinct sensitivities: textual perturbations dominate for some, whereas others are susceptible across both modalities. In several configurations, joint perturbations match or exceed the effectiveness of single-modality perturbations with smaller modifications to each input. In a human study, 60-76% of assessments across models judge the perturbed inputs sufficiently understandable to perform the task, and human localization is more accurate than that of the corresponding VLMs. These results show that evaluating modalities separately can overlook failure-inducing combinations and that modality sensitivity provides a more detailed account of VLM robustness than aggregate failure rates alone.

0.12

MAGIC: Margin-gated Contrastive Debiasing with Evidence Restoration for VLM Test-time Robustness Scaling

cs.CV cs.LG ·2026-10-23 ·Shunshun Gu, Enze Zhang, Ming Pei, Hang Zhou, Xiao Luo
benchmark
Abstract

This paper studies the problem of test-time robustness scaling for large vision-language models (LVLMs), which aims to mitigate language bias during inference. Existing approaches typically adopt test-time scaling techniques such as contrastive decoding to improve inference performance. Despite the progress, the performance of these approaches is far from satisfactory due to intentional evidence removal during perturbations and redundant inference cost. Towards this end, we propose a novel approach named Margin-gated Contrastive Debiasing with Evidence Restoration (MAGIC). The core of our MAGIC is to evaluate the decision margins of LVLMs to selectively activate textual and visual perturbation branches for efficient debiased inference. In particular, our MAGIC first measures the prediction margins of crucial tokens, and applies a gating function to identify low-confidence responses that are susceptible to language biases. For selected samples, we integrate the token logits from multiple textual and image perturbation branches with the original ones for debiased decision. More importantly, we selectively restore the most informative corrupted views with sharp prediction distributions for ambiguous samples after debiasing, enabling evidence recovery for discriminative decision-making. We further utilize a calibration dataset to actively identify and prune ineffective branches during contrastive debiasing for cost reduction. Extensive experiments on benchmark datasets validate the effectiveness of the proposed MAGIC in comparison with extensive baselines.

0.12

DIAL-S2ST: Learning Latency Control for Simultaneous Speech-to-Speech Translation

cs.CL cs.SDeess.AS ·2026-10-23 ·Jeongsoo Choi, Junseok Ahn, Joon Son Chung
Abstract

We propose DIAL-S2ST, a post-training method for learning inference-time latency control in simultaneous speech-to-speech translation models. Our approach reuses pretrained translation capabilities through self-distillation from lookahead predictions, while limiting the influence of large text distillation losses to account for the imitation gap. This provides supervision for earlier emission without paired target data or reward computation. To make the learned timing adjustable at inference, we jointly train a single LoRA adapter across scales with matched lookahead guidance. Experiments on French-to-English translation with Hibiki demonstrate that only 10 hours of source speech enable this adaptation, providing flexible latency control with a favorable quality-latency trade-off.

0.12

CoupleFlow: Null-Calibrated Auditory Attention Decoding With Time-Locked Brain-Speech Coupling and Flow-Matching Likelihood Ratios

eess.AS ·2026-10-23 ·Ali Alavi, Donald S. Williamson
Abstract

Auditory attention decoding (AAD) seeks to identify the attended talker from brief noninvasive brain recordings by comparing neural responses with competing speech streams. Recent electroencephalography (EEG) decoders report nearly perfect accuracy from one second of data, yet we show that this performance fails to generalize to unseen speech because models exploit trial leakage, eye gaze, and acoustic cues instead of speech evoked neural activity. We introduce CoupleFlow, which decodes attention through temporal coupling between neural and acoustic embeddings. Its predictions remain at chance when the brain signal contains no informative coupling, preventing shortcuts based on speech content or talker identity. A conditional flow matching head implements a likelihood ratio decision rule based on pointwise mutual information between brain and speech. This rule is Bayes optimal when the recording depends only on the attended talker and removes preferences for speech that is likely a priori. We evaluate CoupleFlow with content disjoint splits recovered from the audio and null tests that distinguish genuine temporal coupling from trial recognition. Across nine EEG and magnetoencephalography (MEG) datasets processed through a common pipeline, spanning four languages, high density EEG recordings, listeners with hearing impairment, and a room with multiple talkers, CoupleFlow outperforms linear and deep baselines while all null tests remain at chance. Finally, we derive an evidence accumulation law that predicts accuracy across window lengths from a single parameter per dataset and predicts that parameter from the strength of neural speech tracking. This law identifies recording signal to noise ratio as a limit on decoding speed and reliability and estimates the window length required to reach a target accuracy.

0.11

MG-REPA: Multi-Granularity Representation Alignment for Generative Target Speaker Extraction

eess.AS cs.AI ·2026-10-23 ·Dohwan Kim, Jung-Woo Choi
Abstract

Generative target speaker extraction (TSE) models achieve high-quality extraction, but suffer from considerably slow training convergence. Representation Alignment (REPA) accelerates diffusion transformer training by aligning hidden states with a pre-trained Self-Supervised Learning (SSL) teacher. However, existing alignment in the speech and audio domain operates at a single granularity. For TSE, frame-level alignment leaves the global, time-invariant characteristics of the target speech implicit, whereas utterance-level alignment leaves the time-varying content implicit, even though extracting the target requires both. We propose Multi-Granularity Representation Alignment (MG-REPA), aligning the hidden states with an SSL model at the frame level and a speaker verification model at the utterance level. We further apply temporal normalization to the frame-level target, removing its utterance-level statistics so that each branch concentrates on its granularity. On Libri2Mix, MG-REPA reaches the OVRL and PESQ of a MeanFlow-TSE baseline with $3.0\times$ and $2.1\times$ fewer epochs, respectively. Trained for 750 epochs, it is competitive with the published 2,000-epoch baseline on intrusive metrics while improving non-intrusive metrics. An ablation demonstrates that MG-REPA outperforms alignment at a single granularity.

0.11

Tight Learned Event-Inertial Odometry (TLEIO): a Transformer-Based Architecture for End-to-End Motion from Event Streams

cs.CV ·2026-10-23 ·Matteo Cascone, David Jensen, Alessandro Pirini, Angelo Stefanini, Pietro Venè +2 more
benchmark
Abstract · 14 pages

Event cameras provide microsecond temporal resolution, high dynamic range, and sparse asynchronous measurements, making them compelling for odometry under fast motion and challenging illumination. Yet, accurate event-based state estimation remains difficult: existing systems often use events as a complement to frames or rely on feature tracking, correspondence search, and optimization-heavy backends, limiting the efficiency promised by event sensing. We introduce Tight Learned Event-Inertial Odometry (TLEIO), a compact filter-based pipeline that learns short-window relative motion directly from voxelized event streams and tightly fuses the resulting constraints with IMU propagation. Its learned front-end, EventsFormer, predicts local displacements and per-axis uncertainty from event voxels in a single forward pass, while a stochastic-cloning EKF fuses these overlapping predictions with IMU propagation, producing metric trajectories without explicit feature tracks and pose optimization at inference time. The network is trained directly from pose displacements, without requiring event-flow labels. We evaluate TLEIO on the TartanAir/TartanEvent benchmark and on real DAVIS240c sequences. On TartanAir, TLEIO reduces average ATE from 2.26 m to 1.10 m over the IMU-only TLIO baseline, while achieving up to 48.9% lower runtime than optimization-based event odometry methods.

0.11

HEALPix-Based Implicit Neural Compression of Omnidirectional Images

cs.CV ·2026-10-23 ·Yulong Cheng, Youneng Bao, Junfeng Zhou, Mu Li, Jie Wen
Abstract · [N] pages, 10 figures, 4 tables. Submitted to IEEE Transactions on Image Processing

Real-time decoding of omnidirectional images is essential for immersive virtual reality (VR) experiences. Compared with conventional transform-based learned image compression (LIC), overfitted codecs offer lower decoding complexity by representing each image with a lightweight implicit neural representation (INR). However, existing approaches typically operate on the equirectangular projection (ERP), which introduces nonuniform spherical sampling and spatial distortion, leading to redundant sampling near the poles and latitude-dependent receptive fields. To address these limitations, we propose a spherical overfitted codec based on a multi-scale Hierarchical Equal Area isoLatitude Pixelization (HEALPix) representation, which enables equal-area sampling and efficient hierarchical modeling of spherical signals. Building on this representation, we develop a Spherical Inter-Scale Conditional entropy model that exploits spatial dependencies and cross-scale correlations among HEALPix latents to improve coding efficiency. We further introduce a Multi-Ring Spherical Convolution that aggregates neighboring pixels over a nearly uniform spherical receptive field across latitudes. Experiments on a public omnidirectional image dataset demonstrate BD-rate reductions of 16.3%, 26.5%, and 19.5% over VVC in terms of V-PSNR, V-MS-SSIM, and V-LPIPS, respectively. Our method also achieves a 13.1% BD-rate reduction over ERP-based Cool-Chic 5.0 in terms of V-PSNR, while providing 2.6-3.6x faster decoding than the evaluated learned autoencoder codecs.

0.10

Noisy Target Training Meets Drifting for Weakly-Supervised Speech Enhancement

cs.SD eess.AS ·2026-10-23 ·Yoshiki Masuyama, Dominik Klement, Christoph Boeddeker, Julius Richter, Kohei Saijo +3 more
Abstract · Submitted to ICASSP 2027

In speech enhancement, standard supervised training requires pairs of noisy and clean speech. Such pairs are typically created by artificially degrading clean speech, leading to a domain mismatch between synthetic training data and real recordings. Noisy target training (NyTT) mitigates this limitation by training an enhancement model to predict a noisy recording from a noisier version generated by adding extra noise. However, since the training target is noisy, NyTT might learn to leave residual noise in the enhanced output. In this paper, we address this issue in NyTT by incorporating a Sinkhorn drifting loss that matches the distribution of the enhanced speech to that of unpaired clean speech. The Sinkhorn drifting loss attracts enhanced speech toward high-density regions of the clean-speech distribution while repelling it from the current enhanced-speech distribution in latent space. This loss uses only unpaired clean speech rather than the clean counterpart of each degraded recording, preserving the weakly-supervised nature of NyTT. Experimental results on VoiceBank-DEMAND and CHiME-3 real recordings demonstrate the efficacy of the proposed combination.

0.09

Of Chains and Gains: What Reinforcement Learning changes on Reasoning-based Image Quality Assessment

cs.CV cs.CL ·2026-10-23 ·Yixuan Li, Yuxuan Jiang, Hanwei Zhu, Hancheng Zhu, Fan Zhang +2 more
benchmark
Abstract

Reinforcement Learning with Verifiable Rewards (RLVR) has improved the performance of large Vision-Language Models (VLMs) on image quality assessment (IQA). Existing approaches often attribute these gains to improved reasoning before scoring, supported by increasingly plausible Chain-of-Thought (CoT). However, whether these gains exceed sampling effects and how they depend on explicit reasoning remain unclear. We rigorously investigate these questions through sampling-aware evaluation, controlled reasoning interventions, and shortcut experiments. We find that RL-trained models outperform base models under large-budget $\operatorname{max}_g@k$ evaluation, while improvements beyond averaging are concentrated in-distribution. RLVR produces chains that raise base-model performance when transplanted and improves scoring with the chain held fixed. Yet scoring becomes less dependent on the chain during inference, more clearly at 7B than at 32B. Steering the probed quality direction in thinking tokens also has little effect on scoring. Finally, RL can improve scoring accuracy by exploiting textual cues correlated with mean opinion scores (MOSs), which could inflate benchmark performance. Together, these findings show that RLVR raises quality scoring and produces more useful chains, but that better scoring does not imply a stronger reliance on explicit reasoning, while score-based rewards can also encourage reliance on quality-irrelevant shortcuts. We hope this preliminary study can enlighten more and more concrete reasoning schemes in IQA.

0.08

Image Clustering with Vision-Language Model Embeddings

cs.CV cs.AIcs.LG ·2026-10-23 ·João P. Matos-Carvalho, Nuno Fachada
Abstract

Pretrained vision-language models provide image representations whose suitability for clustering may vary with feature extraction, image domain, and partitioning algorithm. We compare nine representation pipelines across seven image benchmarks and four clustering configurations, yielding 252 baseline combinations. Eight pipelines extract visual features, including two vision-only references; the ninth generates descriptions and embeds the resulting text. With agglomerative clustering (Ward linkage) fixed, Qwen3-VL achieves the highest adjusted Rand index (ARI) on five datasets, including 0.925 on MNIST, while BLIP-2 leads on CIFAR-10 and CIFAR-100 with ARIs of 0.925 and 0.591, respectively. A separate sensitivity analysis compares raw and L2-normalized embeddings at their original dimensionality across all 63 dataset-representation pairs. Normalization preserves the highest-ARI representation in every dataset, increases ARI in 35 pairs, decreases it in 26, and leaves two numerically unchanged. On the SVHN dataset, the highest ARI remains low, rising from 0.097 to 0.098 after normalization. The evaluated caption-mediated pipeline obtains relatively high ARI on isolated digits but does not consistently exceed direct visual features. These findings support selecting and evaluating complete representation-clustering pipelines for the intended domain, with explicit attention to extraction and normalization choices.

0.07

Uni-Stream: A Unified Streaming Speech Language Model for Speech Synthesis and Recognition

eess.AS cs.AI ·2026-10-23 ·Xuan Shi, Chang Zeng, Tiantian Feng, Ye Zhang, Jun Liu +3 more
Abstract

Unified streaming speech language models aim to support speech recognition and synthesis within a single model, enabling real-time human-computer interaction. However, jointly supporting these two directions of speech-text conversion while preserving task-specific performance remains challenging. We propose Uni-Stream, a unified streaming speech language model built on compact pretrained speech models. Uni-Stream processes interleaved text and speech in a streaming format and achieves performance comparable to task-specific models through two-stage training. We also develop a data annotation pipeline for streaming training and will publicly release the generated transcriptions and word-level timestamps. We further examine Uni-Stream's representations across tasks and modalities and find that speech and text residence in distinct latent feature space even when trained in a unified model architecture, while representations within each modality become more task-dependent in later layers.

0.07

ACWM-RM: A Reference-Free Reward Model from Complementary Mismatch Preferences

cs.CV ·2026-10-23 ·Shibo Hong, Siyu Ren, FengJiao Chen, Zhongyuan Peng, Wei Shi +1 more
benchmark
Abstract

Reinforcement learning has gained attention for improving action-conditioned world models (ACWMs), as it enables reward-driven optimization of visual quality (VQ) and action-following fidelity (AF) beyond flow-matching optimization. Realizing this potential requires reward signals that reliably rank videos by both VQ and AF. However, existing rewards capture only one dimension: metric-based rewards assess VQ but require ground-truth videos and do not directly assess AF, while inverse dynamics model (IDM) rewards assess AF without references but do not guarantee VQ. Moreover, IDMs, typically trained on real videos, may provide unreliable rewards for generated videos containing visual artifacts and physical inconsistencies. In this work, we introduce the Action and Video Mismatch Benchmark (AVM-Eval), comprising 67.2k pairs that evaluate how reliably IDMs assess AF on real and generated videos. We then propose ACWM-RM, a reward model that jointly assesses VQ and AF without ground-truth videos during reward computation. To train it, we first construct the Action and Video Mismatch Preference Dataset (AVM-Pref) with 506.3k pairs providing complementary supervision: action-mismatch pairs supervise AF, while video-mismatch pairs supervise VQ and expose the model to generated videos. We then pretrain an action-conditioned predictor on real videos to learn forward dynamics before jointly training it with a patch-weighted score head on AVM-Pref to transfer these dynamics to AF assessment on generated videos. Evaluation on AVM-Eval reveals the limited reliability of IDMs on generated videos, while downstream RL experiments show that ACWM-RM outperforms reward baselines in both criteria.

0.07

Beyond Visible Speakers: Context-Aware Active Speaker Detection with Speaker Registration

cs.SD cs.CVcs.MMeess.AS+1 ·2026-10-23 ·Pu Wang, Hugo Van hamme
Abstract · Submitted to IEEE ICASSP 2027

Active speaker detection (ASD) relies on audiovisual correspondence between speech and visible faces, yet this correspondence becomes unreliable in crowded scenes, or when visual evidence is unavailable. We introduce ROAM-ASD-C and ROAM-ASD-CR to extend ROAM-ASD with contextual and identity-aware reasoning. ROAM-ASD-C jointly reasons across simultaneously visible candidates, allowing other faces to provide complementary evidence when the target observation is ambiguous or incomplete. ROAM-ASD-CR incorporates registered-speaker representations derived from enrollment speech, allowing detection of whether a speaker is active even when that speaker is not visually observed. ROAM-ASD-C achieves state-of-the-art performance across five ASD benchmarks: 96.8% mAP on AVA, 98.9% on WASD, 88.8% on UniTalk, 98.7% on Talkies, and 99.5% on ASW. It also improves ROAM-ASD by up to 13.3 mAP under 75% target-track visual removal.

0.07

SyncPro: Synchronous Reasoning and Responding for Proactive Streaming Video Understanding

cs.CV cs.CLcs.LG ·2026-10-23 ·Xueyun Tian, Bingbing Xu, Jiaming Liu, Jinpeng Yu, Yang Shi +4 more
Abstract

Proactive streaming video understanding requires assistants to track unfolding events and respond when a user's requested condition is met. To make these decisions, assistants must accumulate relevant evidence and reassess response conditions as new observations arrive. Existing work offers complementary capabilities for this process: proactive video assistants learn when to respond, while streaming video reasoning methods generate scheduled intermediate text to organize observations and support subsequent answers. Together, these capabilities provide a foundation for proactive reasoning, but how to coordinate ongoing reasoning with timely responses remains underexplored. In a direct adaptation, fixed reasoning intervals can trigger updates when nothing relevant has changed or postpone useful updates until the next scheduled step. Sharing one generation path also couples two distinct requirements: preserving evidence for future decisions and producing only the response requested by the user. Moreover, responses generated after reasoning must wait for it to finish, even when sufficient evidence is already available. To address these challenges, we introduce SyncPro, a synchronous dual-stream framework with specialized generation paths for evidence tracking and user-facing responses. Its Inner and Response branches share causal history and activate independently, allowing relevant evidence changes to trigger internal updates and responses to proceed concurrently with ongoing reasoning. We construct 189K dual-stream examples and train SyncPro through supervised fine-tuning followed by trajectory-level reinforcement learning. Experiments across three proactive benchmarks show improved response timing and more accurate answers to user requests.

0.07

Frontier-OPD: Discovering the Edge of Agent Capability and Teaching Agents to Go Beyond

cs.CL ·2026-10-23 ·Zheng Jiang, Yiming Chen, Houde Qian, Ling Li, Chaoyang Li +1 more
Abstract

On-policy distillation aligns teacher feedback with a student's current behavior, yet acquiring useful supervision for multi-turn agents requires identifying both where guidance is needed and whether the student can sustain the interaction afterward. We introduce Frontier-OPD, which addresses this problem through the student's learning frontier: decision situations that remain challenging but admit useful autonomous continuation after brief teacher guidance. To operationalize this frontier, Frontier-OPD organizes on-policy experience into a shared-state trajectory graph, where trajectory-balanced aggregation of teacher-student disagreement identifies distinct intervention candidates while preserving the histories needed for replay. Short teacher bridges are then assessed by comparing agreement over the bridge and the ensuing student continuation with an equally long segment of the original trajectory, providing a behavioral proxy for guidance that the current student can build on. Accepted teacher responses augment ordinary distillation data, and fresh rollouts renew the frontier estimate as the policy changes. Experiments on ALFWorld, WebShop, and ScienceWorld report higher mean autonomous performance than the compared distillation baselines across the evaluated student scales and teacher configurations, with ablations indicating complementary effects of candidate selection and guidance filtering. In one ALFWorld timing setting, the additional collection cost is offset by a shorter time to a target success rate.

0.05

Melody Guidance for Audio Diffusion Models Without Retraining

eess.AS ·2026-10-23 ·Cristobal Andrade, Sebastian J. Schlecht, Ben Maman, Meinard Müller
Abstract

Controlling melody in audio generation normally requires training a conditional generative model on audio paired with pitch curves or MIDI notes. We instead guide an existing unconditional diffusion model, originally trained without melody inputs, to follow a target melody without retraining or modifying it. During inference, a differentiable pitch detector steers the sampled audio toward the desired pitches, while the diffusion model keeps it close to the distribution of realistic audio learned during training. Our main contribution is a sampling method that balances these two directions, providing melody control while preserving audio quality. To improve stability, we initialize sampling with a simple synthesized version of the target melody rather than pure noise. Experiments on piano, guitar, and singing voice show melody accuracy comparable to conditional models trained specifically for melody control. Our method therefore adds melody control to existing audio diffusion models without requiring paired melody data or task-specific retraining.

0.05

Two Uncertainties Are Not One: Disagreement versus Decision-Risk in Cheap Structured Content Moderation

cs.CL cs.CEcs.LO ·2026-10-23 ·Karan Yadav
Abstract · Withdrawn by the author to undergo substantial revision and expansion of the experimental evaluation

Content moderation compresses a graded, contested human judgment into a binary act (leave up / take down) over billions of items, at a scale where running a frontier LLM on every post is uneconomical. A new class of frozen "System One" models returns typed, calibrated decisions instead of text at ~USD 0.000025 per call and ~100 ms. We run the first independent audit of one such model on 142,658 human-labeled comments, and benchmark it against 14 frontier reasoning models (the GPT-5.x, Claude, Gemini 3, and Grok families) on a shared probe set. Two findings stand out. First, the cheap model is accuracy-competitive: its zero-shot classification AUROC (0.83) is not statistically separable (paired bootstrap, n=56; a failure to reject parity, not proof of equivalence) from 13 of the 14 frontier systems while costing 1-3 orders of magnitude less, and it is no worse calibrated than several of them. Second, "confidence" conflates two uncertainties that come apart: human disagreement (where annotators split on satire and coded language) and decision risk (where the automatic action is costly-wrong). Under an asymmetric moderation loss these call for opposite escalation signals, so no single confidence number serves both. Because the model prices extra questions at nearly zero, decomposing the judgment into parallel sub-questions and reading their spread recovers a human-disagreement signal that the model's native confidence misses -- but this remedy is specific: it helps the weakest, disagreement-blind models and only when the decomposition aligns with the sources of disagreement; a specialist toxicity model and the strongest LLMs do not need it, and the effect is distribution-sensitive. We will release the instrument, the cached runs, and the analysis.

Security & Privacy 27

Identity Amortization: Shared Resources and the Cost of Sybil Attacks

cs.CR ·2026-10-23 ·Homayoun Maleki, Nekane Sainz, Jon Legarda, Igor Santos-Grueiro
Abstract

Sybil defenses commonly price identities by requiring a resource for each one. When one resource can support several identities, an adversary can pay less per identity than the defense intends. We call this identity amortization: a strict reduction in the minimum cost of provisioning identities when resources may be shared, relative to provisioning each one independently. It is a second route to the gap that a Sybil attack exploits. When influence is assigned per identity, a defense calibrated to the nominal price can overestimate the cost of a Sybil threshold by up to the number of identities one resource supports. For any resource catalogue, we give exact tests for when identity amortization occurs, where it begins, and its largest error; a sharing cap below the onset removes it. Because the tests need only prices and capacities, they apply to any defense that admits identities against a shareable resource, telling designers whether the assumed per-identity cost is paid, how wrong it can be, and which intervention restores it most cheaply for honest participants. On published phone-verification prices, per-account estimates overstate provisioning cost 1.6 to 4.3 times; Tor's raised per-address relay limit cut modeled attack cost by 8 to 72%, depending on the address-to-relay price ratio.

0.51

Developing Security Metrics for Automated Compliance Checking

cs.CR ·2026-10-23 ·Immanuel Kunz, Nico Haas, Angelika Schneider, Christian Banse
publishedcode/data
Abstract · 15 pages, 1 figure, 1 table. Author-accepted manuscript of a paper published in the proceedings of IFIP SEC 2026. The final authenticated ve…

To automatically check for compliance with security-related standards, prose requirements must be translated to technical properties to make them measurable. A multitude of metrics have been proposed in past literature, but they are lacking a systematic methodology for their elicitation. Without such a methodology, metrics remain inconsistent in their data models and abstraction levels, they exhibit considerable overlaps and are difficult to combine to comprehensively cover a standard's requirement. In this paper, we develop a data model for security metrics based on a systematization of compliance requirements, and we propose a methodology for their elicitation. The method is based on an ontology that enables a systematic development of metrics with a comparable degree of abstraction. We also present our initial work on an open-source effort that demonstrates that our approach also supports a collaborative development of metrics.

0.47

A Deterministic Security Runtime for LLM Agents on MPI and Batch-Scheduled HPC Systems: Consensus-Governed, Hardware-Rooted Supervision of Every Rank

cs.CR cs.AIcs.DC ·2026-10-23 ·Alfredo Metere
benchmarkcode/data
Abstract

Large language model agents launch simulations on high-performance computing (HPC) systems, where one sbatch or mpirun can commit thousands of node-hours and a forged result can cost lives. Of 23 surveyed systems, none supervises the running job, and none holds when the control host is compromised. We fix both. Enclawed, our deterministic security runtime for AI agents, has until now served a single user; this work makes it a distributed runtime for parallel HPC and agent swarms. The Enclaweder, our hardware root of trust, keeps a compromised host from becoming a catastrophe. A launch gate decides every launch by policy alone, refusing what it cannot parse. Once the job runs, a swarm chain grants short leases against signed heartbeats and evicts on evidence, and a supervisor around every rank pauses it when its lease lapses and kills it on tampering. Each node signs with its Enclaweder's key, so the host can forge none of it. Enclawed needs no administrator privileges and plugs into existing clusters unchanged, over Secure Shell alone. Supervision leaves the compute loop of a 16-rank LAMMPS job unchanged, on graphics processors too, and adds about 3 s of start-up. Twelve live attacks on Enclaweder-keyed ranks, colluding coordinators included, ended fail-closed. Sixteen agents on a real model attempted 24 launches that planted instructions ordered; the gate denied all 24 and allowed all 108 legitimate ones. Enclawed's core and the HPC Agent Security Benchmark are open source.

0.41

AquaSense-ICS: A Hybrid Real-Simulated Cyber-Physical Dataset and Leakage-Aware Benchmark for Smart Water Quality Monitoring and Attack Detection

cs.CR ·2026-10-23 ·Amal Neji, Mohamed Fourati, Lamia Chaari Fourati
benchmarkdatasetcode/data
Abstract · 24 pages

Machine-learning research on water cyber-physical systems (CPS) is constrained by datasets that are either real but chemically poor, or simulated but free of sensing noise and attacks. We present AquaSense-ICS, an open dataset and benchmark that co-registers (i) a 25.4 h recording window (8.0 h of recorded data, 26% coverage) from a Raspberry Pi 5 edge node (TDS probe and BME680 ambient sensor, 1-2 s sampling, 23,688 records), with (ii) a looped network of 13 nodes (10 junctions, 2 reservoirs, 1 tank) simulated in EPANET-MSX for six water-quality species, (iii) a parameterised library of cyber-attacks (false data injection, replay, denial of service, drift) and unintentional sensor faults with sample-level ground truth, and (iv) augmented and TimeGAN-generated sequences, flagged by an origin column. The attack, fault and water-quality labels are traceable to a seeded generator, a documented rule or a simulation input. We show that the most common way of benchmarking such hybrid data, random row splits and labels computed from the input features, yields near-perfect but uninformative scores (macro-F1 = 0.9998 for Random Forest), and we replace it with a leakage-aware protocol using chronological splits and held-out attack types rather than label reconstruction. Under this protocol, no model exceeds the constant "normal" predictor for attack classification (macro-F1 = 0.465), and the best detectors reach a mean AUROC of 0.64 on attack types unseen during training (0.35-0.93 depending on the type), leaving substantial headroom. Data and simulation files are released under CC BY 4.0 at https://doi.org/10.5281/zenodo.23390541, and attack generators and benchmark code under MIT.

0.39

Post-Quantum Perfect Zero-Knowledge Arguments for NP From Any One-Way Permutation

cs.CR quant-ph ·2026-10-23 ·Frédéric Dupuis, Rémi Ligez, Khaireddine Medhioub, Louis Salvail, Louis-Charles Thibodeau
Abstract · Revised the Introduction and added a Technical Overview section. Corrected typos and minor notation issues. No changes to the results

In this paper, we show for the first time that post-quantum one-way permutations are sufficient for perfect zero-knowledge classical arguments for all of NP with security guaranteed against quantum adversaries. We first establish that Naor, Ostrovsky, Venkatesan, and Yung's bit commitment scheme built from any permutation using interactive hashing, called the NOVY scheme, is sum-binding and collapse-binding against quantum polynomial-time committers provided the permutation is post-quantum one-way. We then show that the collapse-binding property of the NOVY scheme allows for Blum's perfect zero-knowledge argument for HAMCYCLE to remain so exactly as it stands against quantum adversaries.

0.34

Embedding Large PQC KEMs in IKEv2

cs.CR ·2026-10-23 ·Daniel Herzinger, Marco Thomas, Tobias Handirk, Stefan-Lukas Gazdag, Daniel Loebenberger +1 more
code/data
Abstract

Post-quantum key encapsulation mechanisms carry public keys far larger than the established VPN handshake protocol IKEv2 was designed for. At the same time, its initial exchange comprising the public keys cannot be fragmented. In response, the IETF discusses a range of protocol adjustments, yet these proposals have never been compared in terms of their performance and stability under identical conditions. We systematically screened the IETF document corpus for such proposals, added three of our own (one using forward error correction and two using IKEv2 with QUIC, namely QUIKE and QUIKE with a composite key exchange), identified two specification gaps in existing acknowledgment drafts, and implemented all eight on the same open-source codebase. Measurements across different bandwidths, latencies, and packet loss rates, and with ML-KEM, FrodoKEM, and Classic McEliece, total 126 500 handshakes in 1296 different configurations and show that no mechanism fully compensates for the effect of the key size on performance. For Classic McEliece, no approach handles challenging conditions acceptably. The approaches either congest the link or take up to nearly two minutes per handshake on average. For ML-KEM and FrodoKEM, there are two ways forward: building on already standardized extensions or delegating the transport. The former requires the larger implementation change and needs further optimization, such as acknowledgment mechanisms or error-correcting codes, but grants continuity and compatibility. The latter can use a composite key exchange to avoid these drawbacks and achieve a faster, more resilient handshake and leaner implementation, with QUIKE-composite offering the best overall performance.

0.31

The 3DS Oracle: BIN Attack via the Network

cs.CR ·2026-10-23 ·Wasif Faisal, Suprava Saha Dibya
Abstract · Extended preprint

EMV 3-D Secure (3DS) requires a client-visible continuation boundary in browser-mediated flows; in some deployments, that boundary exposes a pre-authorization classification channel. This paper studies 3DS as a protocol, integration surface, and governance problem, combining standards analysis, architectural reconstruction, formal modeling, a documentation audit, disclosure history, and a bounded live evaluation. We show that client-visible integrations can reflect routing and authentication state before financial authorization, allowing 3DS to function as a profiling and screening channel for issuer controls and cardholder/account state rather than only as a defense. In the tested browser-channel Braintree/Cardinal path, the first client-visible result separated researcher-owned anchor cards from bounded adjacent same-range comparison PANs in six of seven distinct ACS/card-range cohorts; separate public evidence, including openly accessible attacker tooling, rotating-proxy support, and illicit paid 3DS card-testing services, shows that comparable signals are already operationalized in the wild. Our live study was deliberately bounded for ethical reasons, so the empirical ceiling reported here reflects researcher restraint rather than an attacker-side scalability limit, although we do not claim network-wide prevalence or candidate issuance. The deeper issue is systemic: decoupled pre-authorization authentication, weak requestor-boundary assumptions, identifier-linked routing invariants, and fragmented accountability let the exposure persist even when individual actors appear locally conformant. Meaningful remediation therefore requires consortium-level reform, including stricter minimization of client-visible pre-authorization responses, more uniform continuation behavior, and stronger client authenticity and session binding.

0.28

Deletion, Substitution, Repair and Transplant: Component Attribution in BARBICAN, a Layered Prompt-Injection Gateway for MCP

cs.CR ·2026-10-23 ·İpek Abasıkeleş Turgut, Edip Gümüş
benchmarkposition
Abstract · 24 pages, 13 tables. Reproduction artifacts (per-request traces, detector scores, verification script): https://doi.org/10.5281/zenodo.23590…

A neural classifier can fail completely inside a layered defense without any end-to-end number moving. We demonstrate this on BARBICAN, a fail-closed gateway for language-model tool mediation whose arbitration stage spent an entire recorded evaluation receiving text mangled by a misconfigured tokenizer. Correcting the misconfiguration lifts that stage's corpus-wide area under the ROC curve (AUC) from 0.392 to 0.855 and pushes system recall down, from 80.67% to 80.34%; a stub that answers "malicious" scores 80.77%, higher than either. The reason emerges once the stage's input distribution is examined: upstream routing hands it a band that is 96.82% malicious, and the enforcement design lets an incriminating verdict block while an exculpating one releases an input only through one narrow branch. Yet deleting it costs 673 detections. Substituting its verdict with a constant instead moves between 12 and 140 of 25,000 decisions. Transplanted into the arbitration stage, an off-the-shelf classifier that separates the band's inputs at an AUC of 0.916 changes 16 of 25,000 decisions. Measured against such detectors with operating points chosen away from the test data, the gateway leads on its own benchmark distribution and falls behind once that distribution changes. In our earlier gateway, the review model's band is 92.94% malicious and a constant verdict is as accurate as the review model. We argue that an ablation figure for any routed stage is unreadable without the class balance of the band that stage receives, and that deletion experiments need substitution experiments beside them.

0.28

Automated Post-Exploitation Attack Chain Exploration Based on a Phase-guided Hybrid Planning-Monte Carlo Tree Search Algorithm

cs.CR cs.CY ·2026-10-23 ·Jie Ying, Mengce Zheng, Ruoxi Chen, Wenrui Cheng, Tiantian Zhu +1 more
benchmark
Abstract

Advanced attackers in post-exploitation scenarios increasingly craft attack chains that combine benign operations, configuration errors, and vulnerability exploitation to achieve objectives like privilege escalation. While such chains pose significant threats, manually discovering them in complex modern host environments is prohibitively labor-intensive. Automating this discovery is therefore essential for proactive defense, yet existing methods still suffer from inefficiency and lack diversity when exploring these attack chains. To address these limitations, we present APEX, an Automated Post-EXploitation attack chain exploration system. This system transforms attack chain discovery into a planning problem and solves it with a novel phase-guided hybrid Planning-Monte Carlo Tree Search (MCTS) algorithm. By organizing attack actions into semantic phases and integrating symbolic planning with MCTS, APEX enables efficient exploration of the state space while generating multiple diverse chains. Evaluations on a benchmark comprising 5 virtual machines, 30 synthetic problems and 291 cloud instances demonstrate that APEX successfully discovers both known and previously unreported attack chains--including 9 novel privilege escalation vectors--providing actionable intelligence for hardening complex production environments.

0.28

Benchmarking Classical and Quantum Machine Learning for Intrusion Detection Across Multiple Datasets

cs.CR cs.ET ·2026-10-23 ·Taha M. Mahmoud, Naima Kaabouch
publishedbenchmark
Abstract · 7 pages, 3 figures. Author's accepted manuscript of the paper published in the 2026 6th International Conference on Electrical, Computer and…

Intrusion detection is examined through a comparative benchmark of classical and quantum machine learning models on three datasets: NSL-KDD, UNSW-NB15, and MQTTEEB-D2025. The analysis focuses on how preprocessing decisions, feature selection strategies, and quantum encoding schemes shape model performance on data with differing levels of noise and complexity. A single unified pipeline is applied, covering normalization, class-imbalance handling, dimensionality reduction, and two feature selection methods: Random Forest importance and a quantum-aware approach that combines Quantum Kernel Alignment with Mutual Information (QKA-MI). Four classifiers are compared: Support Vector Machine, Random Forest, Quantum Support Vector Machine, and Pegasos Quantum SVM. The classical models perform consistently across all three datasets, whereas the quantum models depend more strongly on feature representation and kernel alignment. Quantum-aware feature selection yields marked gains for the quantum models, especially on cleaner datasets, while heterogeneous traffic remains difficult. Pegasos Quantum SVM offers a good trade-off between accuracy and computational cost, underscoring the role of preprocessing alignment in practical quantum intrusion detection.

0.27

AgentTrace: Execution-Level Fingerprinting of Black-Box LLM Agents

cs.CR cs.AI ·2026-10-23 ·Gejian Zhao, Hanzhou Wu, Xinpeng Zhang
Abstract · 10 pages, 4 figures

Verifying the provenance of autonomous LLM agents is a critical security challenge. Existing semantic attribution is easily obfuscated by system prompts, while active watermarking conflicts with schema-constrained tool generation. We propose AgentTrace, an active framework that identifies the foundation model behind a black-box agent via observable execution traces. AgentTrace injects structurally ambiguous probes to trigger tokenization-sensitive serialization failures at rigid tool interfaces. While these initial failures provide coarse-grained separation across model families, the temporal sequence of subsequent repair operations reveals fine-grained, model-specific recovery behaviors. By modeling these multi-step trajectories with lightweight Markov chains, AgentTrace achieves 96.0% attribution accuracy under adversarial identity spoofing. Crucially, sequential repair modeling successfully distinguishes the evaluated intra-family variants sharing identical tokenizers (e.g., Llama-3 8B/70B), substantially outperforming unordered behavioral statistics while remaining robust against prompt obfuscation and temperature scaling.

0.27

Holo-NN: Holographic Nearest Neighbor Search via Compressive Sensing

cs.CR cs.DBcs.IRcs.LG ·2026-10-23 ·Bryant Gipson
Abstract

In the age of AI, privacy-preserving approximate nearest neighbor (ANN) search over high-dimensional embeddings has emerged as a critical yet frustratingly challenging domain, dominated by tradeoffs between speed, privacy, and accuracy. Traditional tree and graph-based indexing structures (e.g., HNSW, ScaNN) rely fundamentally on non-linear operations that are computationally intractable under Fully Homomorphic Encryption (FHE) or blind server-side constraints. In this paper, we introduce Holo-NN (Holographic Nearest Neighbor Search), an algebraic framework that replaces data-dependent branching with a single linear projection. Our key observation is that while distances across a database are dense, neighborhood membership is inherently sparse: only a handful of points ever lie close to a query. By pairing locality-preserving feature transforms with Compressive Sensing, Holo-NN folds the entire database offline into a compact holographic sensing matrix. At query time, encrypted search reduces to a single, non-interactive matrix-vector multiplication---requiring no FHE bootstrapping, no multi-round interaction, and no client-side index state. This turns private retrieval into an embarrassingly parallel, accuracy-tunable computation that runs out of the box on both CPUs and GPUs, scaling efficiently to production databases. We offer a comparison against other state-of-the-art Private ANN algorithms to help situate the method in the broader context of production-grade private retrieval.

0.27

Label-Budgeted Attribute Inference from Protected Features in Edge-Cloud Split Multimodal LLMs

cs.CR ·2026-10-23 ·Junguang Yao, Peichun Hua, Hanxiu Zhang, Yue Zheng
Abstract · 5 pages

Edge-cloud split inference runs the early layers of a multimodal large language model (MLLM) on a device and sends intermediate features to a server to complete the task. While keeping raw images local is often perceived as a privacy benefit, an honest-but-curious server can still infer sensitive attributes, such as age, gender, and race, from the received features. Existing protections intervene at the input, transmitted representation, or model's responses, but their native evaluations rely on final-response refusals or fixed-classifier failures, which underestimate an adaptive server's capabilities. To address this, we propose a label-budgeted attribute-inference framework that measures representation-level privacy through a recovery curve: the share of the unprotected leakage the server recovers as a function of its labeled-sample budget, from zero (a classifier trained on unprotected features) to a full labeled set. Our evaluation across two popular MLLMs reveals two critical findings. First, in unprotected pipelines, intermediate features leak all three attributes regardless of the split depth. Second, even when protection mechanisms are deployed across different stages, sensitive information remains recoverable under limited label budgets. For example, Safe-LLaVA removes biometric content from the model's responses, but from the features it transmits in split inference, 20 labeled samples per class recover 75-85% of the unprotected leakage. This work provides a label-budgeted measure of feature-level privacy for split MLLM inference and motivates defenses that act on the transmitted representation.

0.27

Alignment of Autonomous Cyber Defense Agents via Adaptive Policy Shaping

cs.CR ·2026-10-23 ·Dena Mujtaba, Bharadwaj Ravichandran, Brian Hu, Arslan Basharat
publishedcode/data
Abstract · Accepted to the Conference on Applied Machine Learning in Information Security

Cybersecurity operators increasingly encounter sophisticated and fast-paced attacks, making autonomous cyber defense (ACD) a critical approach for supporting timely and effective decision-making. Reinforcement learning (RL)-based ACD agents have the ability to autonomously observe network states and take actions to defend and recover system assets, making them well suited for ACD. However, these agents often face critical trade-offs, such as improving network security while inadvertently degrading service availability or violating other operational priorities such as confidentiality and integrity, raising the core challenge of aligning agent behavior with human-specified goals and intent. In this work, we study the problem of aligning RL-based ACD agents with human-specified operational objectives without requiring costly retraining. We formalize alignment for ACD as a policy shaping problem, where a pretrained RL decision-making policy can be adjusted at test time using predicted attributes over agent actions. We also propose an adaptive approach that dynamically controls the alignment--performance trade-off as operational priorities shift during an ongoing cyber attack. We evaluate our method in a simulated cyber defense environment, CybORG, analyzing performance across multiple alignment objectives using the Confidentiality, Integrity, and Availability (CIA) information security triad. Our results show that adaptive test-time policy shaping can significantly improve alignment while preserving core cyber defense performance, offering a practical and scalable alternative to training-time alignment methods in RL. Our code is publicly available at https://github.com/ITM-Kitware/alignment-cage-challenge-2.

0.27

From Hype to ROI: An Economic Model for LLM Adoption in Information Systems and Security Operations

cs.CR cs.AIcs.CE ·2026-10-23 ·Pellegrino Casoria, Andrea Araldo, Rosario Patanè, Simon Pietro Romano
Abstract

Large Language Models (LLMs) promise significant efficiency gains across information-intensive operations by automating repetitive language-based tasks. However, adoption decisions remain largely opportunistic, as organizations struggle to move from pilots to sustainable industrialization. Without systematic economic frameworks, practitioners cannot assess whether LLM investments create value in a given context. This paper develops a general economic model to support informed LLM adoption decisions in operational information systems. Facing high uncertainty, we represent parameters as random variables centered on authoritative benchmarks, and evaluate the model by Monte Carlo simulation across pessimistic-to-optimistic assumptions on LLM efficiency. The ROI distributions show the probability of profitability and the Value at Risk for each configuration, together with a sensitivity analysis of its main drivers. As a first instantiation, we apply the model to Security Operations Centers (SOCs), mapping workflows onto established standards to identify tasks suited to LLM augmentation. The model balances productivity gains, driven by alert volume, analyst effort, and LLM efficiency, against operational costs, including inference, infrastructure, and error remediation, covering both cloud API and self-hosted deployments. Results show that adoption economics are context-dependent, and that automation is an exploitable lever, gated by accuracy, while the cost of errors and the model's price are secondary. Across the analyzed models and prices, most configurations are unprofitable, with profitability limited to high-accuracy, low-uncertainty deployments. The model gives organizations a quantitative basis for evidence-based adoption decisions, by locating the boundary between profit and loss, and the loss risked beyond it.

0.26

A-GE: Privacy-Preserving Age Assurance via Fine-Grained Motion

cs.CR ·2026-10-23 ·Mohammed Gharib, Giovanni Cascone, Martin Zizi
Abstract

Age assurance faces a fundamental privacy dilemma: meaningful assurance typically requires revealing identity or other sensitive information, whereas privacy-preserving alternatives often provide little assurance. Existing approaches either disclose sensitive information directly or shift the disclosure and trust requirement to an identity provider, while self-declared age provides little meaningful assurance. Thus, a fundamental gap remains between establishing an age property and revealing information beyond what is necessary to establish that property. We present \textit{A-GE} (Age-Grounded Evidence), a privacy-preserving age assurance system that derives age-related evidence from fine-grained motion captured by commodity inertial measurement units (IMUs). A-GE exploits subtle motion dynamics that we hypothesize reflect underlying neuromuscular and physiological processes during ordinary device interaction. It requires neither identity information nor government identification and operates across available sensor sampling rates, including the approximately 60~Hz limit of web-based sensing, with configurable minimum-age thresholds ranging from 14 to 21 years. We evaluate A-GE on more than 600 participants, with separate participant-disjoint cohorts for model development and final evaluation. At a 16-year minimum-age threshold, A-GE achieves up to 93.8\% accuracy with a comparable F1 score across the evaluated configurations. We further demonstrate client-side deployment on Android, iOS, and the web, and examine unintended attribute inference and residual privacy risks, characterizing both the privacy benefits and limitations of deriving age evidence from commodity motion.

0.25

Data-Efficient Backdoor Attacks on Code Models via Trigger-Neutralized Counterfactual Scoring

cs.CR cs.AI ·2026-10-23 ·Yu-Yu Lee, Ming-Feng Tsai, Tomohiro Morikawa, Chun-I Fan
Abstract · 18 pages, 7 figures, 14 Tables

Pre-trained code models are exposed to backdoor poisoning, in which crafted training samples make the model misbehave on a trigger. A capable attacker minimizes the number of such samples, since each adds cost and exposure and a lower poisoning rate leaves a smaller footprint for defenses. Which samples to poison therefore matters, but existing data-efficient selection methods target image classification, offer no scoring mechanism for autoregressive generation, and have not been studied on code. We propose Trigger-Neutralized Counterfactual Scoring (TNCS), which isolates a trigger's contribution to the target behavior by comparing the scoring model's output on a triggered input with its output on a trigger-neutralized counterfactual, covering classification and generation alike. TNCS attains higher attack success rates than the baselines at very low poisoning ratios, across three CodeXGLUE tasks and five trigger families, without degrading clean-task performance, and its margin grows as the ratio decreases. It is the only method effective on every trigger family, and its selected poison set transfers to mismatched victims.

0.25

SPA: Structure-Preserving Approximation of Gated Activations under FHE

cs.CR ·2026-10-23 ·Rui Li, Wenyuan Wu, Weijie Miao
Abstract

Nonlinear activation evaluation remains a major bottleneck in neural inference under fully homomorphic encryption (FHE). FHE supports efficient addition and multiplication, but lacks native operations for exponential functions, reciprocals, and value-dependent selection. High-degree polynomial approximations on fixed intervals are therefore commonly used to replace nonlinear activations. They can achieve small errors but often incur substantial multiplicative depth and latency. We observe that several widely used nonlinear activations share the algebraic structure $A(x)=xg(x)$ with $g(-x)=1-g(x)$. This class includes GELU, SiLU, fixed-parameter Swish, and the scalar nonlinear branches of GEGLU/SwiGLU. We propose \SPA{} (\emph{Structure-Preserving Approximation}) to exploit this shared structure. \SPA{} retains the exact odd component $x/2$ and approximates only the even residual as a polynomial in $x^2$, yielding a comparison-free arithmetic kernel. This construction removes redundant odd degrees of freedom and enables lower-cost polynomial evaluation at a given error target. Plaintext experiments with Gemma-4-E2B and Llama-3.2-3B show that SPA largely preserves downstream task performance without weight updates. Fixed-interval CKKS benchmarks show favorable error--latency trade-offs. On $[-5,5]$, \SPA{}-balanced achieves a $3.25\times$ speedup over MOAI (ICLR'~26) for erf-GELU with comparable error. On the same interval, \SPA{}-accurate achieves $79.22\%$ lower root mean squared error (RMSE) and a $10.40\times$ speedup over Euston (S\&P'~26) for QuickGELU. For SiLU, \SPA{}-accurate achieves $19.02\%$ lower RMSE and a $2.06\times$ speedup over THOR (CCS'~25).

0.24

EVSEFuzz : A Tool for Automated Testing of Protocol Implementations

cs.CR ·2026-10-23 ·Soham Gumaste, Michael Khan, Roland Varriale, Rigel Gjomemo
Abstract

The rapid adoption and development of electric vehicle (EV) networks has required the expansive deployment of EV charging infrastructure (EVCI) which relies on communication protocols such as the open charge point protocol (OCPP) to coordinate data transfer between charging network operators and the EV supply equipment. In this paper, we present EVSEFuzz, a tool for automatically building fuzzers to test the security of protocol implementations. To achieve this we make a novel use of Python datatype validation packages and tools to ``understand'' datatypes and message structures and generate both compliant and non-compliant messages. To date, EVSEFuzz has been tested successfully with several open and closed-source implementations and has found several vulnerabilities leading to the publication of two CVEs.

0.23

Spoofing the Toolbox: Optimizing Tool Descriptions to Deceive LLM Agents

cs.CR ·2026-10-23 ·Shih-Chieh Dai, Xinyu Sun, Guanhong Tao
Abstract

Large language model (LLM) agents are widely used in many applications and often rely on external tools. Tool information, such as function descriptions, helps agents determine which tools best suit a given context. However, attackers can manipulate this information to mislead LLM agents into using malicious tools. Prior attacks mainly rely on keyword-based techniques, which are easily detected, or rule-based methods, which have limited effectiveness. In this work, we introduce TSA, a stealthy tool-spoofing attack that crafts descriptions for malicious tools that are difficult to distinguish from those of existing benign tools. To achieve this, we propose an LLM-powered optimization method that iteratively refines malicious tool descriptions. Specifically, TSA begins with a variant of a benign tool description and updates it based on feedback from a surrogate LLM agent. The optimization method incorporates a feedback function and contrastive in-context learning to guide the search for descriptions that are both stealthy and effective. Experiments on both single- and multi-tool use benchmarks demonstrate that TSA achieves an attack success rate of up to 92.13\%, outperforming four state-of-the-art approaches while remaining highly stealthy. Our user study further shows that even human participants cannot distinguish descriptions generated by TSA from benign ones. Moreover, TSA remains effective against five tool-malware detectors.

0.20

Privacy-Preserving Federated Loss Exploration for Vision-Based Perception under Non-IID Data

cs.CR ·2026-10-23 ·Nergiz Yuca, Christian Internò, Nikolay Matyunin, Markus Olhofer, Barbara Hammer +1 more
Abstract

Vision-based image classification models benefit from Federated Learning (FL) across vehicles, however learning accurate models from distributed vehicular cameras is challenging due to data heterogeneity and privacy concerns. To improve performance on non-IID data, advanced FL optimization approaches have emerged that share auxiliary insights, such as client gradient behaviors. While these gradient-guided optimized FL methods improve the model accuracy, their reliance on sharing additional gradient information introduces a previously unexplored privacy vulnerability. In our empirical privacy-leakage experiment, we show that a malicious server can infer clients' label sets from shared guidance matrices under heterogeneous client label distributions (simulated via Dirichlet-based label partitions), between 50% and 84% accuracy. To address this privacy leakage, we propose PPFLex that replaces the central server with three Secure Multi-Party Computation (MPC) servers to securely compute the global guidance matrix and perform federated averaging under semi-honest and malicious MPC server assumptions. We evaluate PPFLex on widely used traffic-sign classification datasets (e.g., GTSRB and BTSC) as representative autonomous vehicle perception benchmarks, comparing against baselines. Our results show that PPFLex achieves accuracy comparable to the plain-text FedLEx pipeline while preventing privacy leakage, incurring acceptable MPC overhead.

0.20

On Task, Off Track: Benchmarking LLM and Agent Robustness to Near-Target Prompt Injection

cs.CR ·2026-10-23 ·Yuyang Gong, Zihao Wang, Jiawei Liu, XiaoFeng Wang
benchmark
Abstract

Prompt injection benchmarks assess whether large language models (LLMs) and agents resist malicious instructions embedded in external content. However, a system could perform well by rejecting attacks that obviously abandon the user's request, while still accepting changes that retain most of a legitimate answer but make it incorrect or harmful. Testing more instructions for unrelated outcomes would not necessarily reveal this weakness. We call attacks seeking responses designed to resemble legitimate answers while introducing incorrect or harmful changes near-target prompt injection. We introduce NTBench, the Near-Target Benchmark, which builds on legitimate tasks from existing datasets. For each task, a legitimate reference outcome guides the generation of target--instruction pairs, with separate tracks targeting incorrect and harmful outcomes in responses or agent tool invocations. Harmful targets can also accompany otherwise correct task completion. We combine LLM-based checks with human validation of the intended deviations, then assess whether injected instructions induce them using response judgments or tool-call checks, excluding deviations already observed without injection. We evaluate seven LLMs on ten datasets for language tasks and two agent benchmarks. Under a common budget of three attack executions per sample, both tracks achieve higher overall attack success rates than the fixed-target baseline for every evaluated model in each setting. On language tasks, the increases range from 9.3 to 35.8 percentage points for incorrect targets and 1.7 to 17.3 percentage points for harmful targets. These results show that resistance to the tested fixed-target attacks does not imply resistance to related incorrect or harmful outcomes, and that reference-guided targets broaden the failures revealed by robustness evaluation.

0.19

Beyond Copying: Lifelong Smart Contract Security Task Construction via Protocol Mutation

cs.CR cs.AI ·2026-10-23 ·Zifan Peng, Peng Huang, Jingyi Zheng, Yingjie Xue, Xuechao Wang +3 more
Abstract

AI agents can inspect code and run tests to support smart contract security. SlowMist reports about $38.6 billion in losses across the broader blockchain ecosystem. Existing evaluations can reward memorized or copied answers, including attack code recovered from public blockchain transactions. Reconstructing contracts and validating tasks also require costly manual effort. To address these issues, we propose CONSTRAINTDIAL, a framework for lifelong construction of smart contract security tasks via protocol mutation. It automates local contract reconstruction and uses attack execution to select changes to the required calls or inputs. Execution checks guide revisions that make the original attack fail while preserving exploitability, the security objective, and normal functionality. New incidents and audit findings enter the same procedure. With CONSTRAINTDIAL, we construct 3,045 task inputs from 203 public cases, covering detect, exploit, and patch. Compared with reconstruction alone, protocol mutation reduces success with supplied public answers by 10.9 percentage points on average across detect, exploit, and patch. The fraction of tasks solved by copying online answers falls by 9.3 points.

0.15

SynMark: Semantic-Coordinate Multi-Bit Watermarking during LLM Code Generation

cs.CR ·2026-10-23 ·Rui Xu, Weizhi Liu, Zhaoxia Yin
Abstract

LLM-generated code is increasingly reused across downstream applications and often undergoes identifier renaming or structural refactoring, making robustness to functionality-preserving edits essential for generative multi-bit code watermarking. Existing methods commonly derive message routes from surface-token prefixes, so such edits can desynchronize routing and degrade watermark recovery. We propose SynMark, which replaces surface-dependent routing with learned semantic coordinates and further calibrates residual recovery errors after transformation. For watermark embedding, Scope learns semantic coordinates from aligned prefixes of functionally equivalent variants for keyed routing and applies a syntax-aware adaptive bias during generation. For watermark extraction, Traced recomputes semantic routes from transformed code and uses reliability-weighted aggregation to suppress unreliable token contributions. Under 100% identifier renaming, SynMark retains 82.27% and 83.77% BitAcc on CodeLlama and DeepSeek-Coder, respectively. Across three code LMs, it also improves clean BitAcc and Pass@1 by up to 4.53 and 11.18 percentage points over the strongest watermarked baselines.

0.14

Hidden Backdoor Attack on Deep Contrastive Clustering

cs.CR ·2026-10-23 ·Zifeng Zhou, Shengda Zhuo, Jiahui Huang, Zeyan Li, Jinchun He +4 more
Abstract

Deep contrastive clustering has become a representative paradigm for unsupervised image clustering, yet its security under training-time poisoning remains largely unexplored. This paper investigates a hidden backdoor threat against deep contrastive clustering and proposes HBDCC, a Hidden Backdoor attack for Deep Contrastive Clustering. Here, hidden does not indicate that the trigger is visually imperceptible, but rather that the malicious trigger-cluster association remains behaviorally dormant on clean inputs and is activated only when the trigger appears at inference time. HBDCC poisons only a small fraction of unlabeled training samples and exploits an inherent structural weakness of contrastive clustering: augmentation-persistent triggers can be repeatedly aligned across different views at the instance level and further reinforced by cluster-level consistency objectives. As a result, the poisoned model preserves normal clustering behavior on clean data, while producing severely distorted cluster assignments for triggered inputs. We provide a dual-level analysis of the backdoor implantation mechanism and evaluate HBDCC on ImageNet-10, ImageNet-Dogs, and STL-10 with representative contrastive clustering frameworks. Extensive experiments demonstrate that HBDCC can substantially degrade triggered clustering performance in a dataset-dependent manner while maintaining sufficient clean-data utility to evade standard evaluation protocols. We further conduct comprehensive studies on poisoning ratio, attack scope, trigger type, trigger size, and a lightweight knowledge-distillation-based mitigation strategy.

0.12

Tracing Money Laundering Across DeFi, Bridges, and Mixers on Ethereum

cs.CR ·2026-10-23 ·Huan Xing, Gaohe Jin, Bang Fu, Yue Li, Baoning Niu +1 more
Abstract

Illicit actors exploit Ethereum's deep liquidity and diverse on-chain services to launder criminal proceeds by combining decentralized finance (DeFi) protocols, cross-chain bridges, and mixers in multi-stage processes. Existing anti-money laundering tracing methods fail to disentangle the underlying asset-flow paths in complex transactions, potentially misdirecting subsequent tracing; cannot identify cross-chain intent without pre-collected bridge information; and treat mixers as generic exit services, overlooking post-mixer laundering flows. We present FPTracer, a framework for tracing illicit assets across Ethereum services. First, we propose a method based on the Context-Aware Transfer Tree (CATT) to recover intra-transaction asset-flow paths, which guide accurate tracing and service analysis. Second, we extract discriminative features from asset-flow and execution-evidence perspectives and train a classifier to identify transactions with cross-chain intent. Third, we leverage evidence from the upstream laundering network to analyze potential post-mixer laundering activities. Evaluation on 20 real-world laundering cases demonstrates FPTracer's effectiveness. It achieves the highest recall for both known laundering addresses (98.6%) and destinations (94.2%). It identifies all known uses of bridges in these cases and uncovers 30 previously unreported ones. Among 18 cases with identified Tornado Cash usage, FPTracer can identify potential withdrawal-side activities, which are corroborated by public reports where available.

0.11

Detecting Insecure Compositions of Clean Agent Skills via Capability-Flow Graph Analysis

cs.CR ·2026-10-23 ·Yihao Li, Chuanpu Fu, Yizhou Zhao, Miao Chen, Qi Li +1 more
Abstract · 11 pages, 5 figures, 5 tables

AI agents increasingly extend their capabilities by invoking third-party skills, where each skill encapsulates instructions, scripts, and resources for agents to execute on demand. This paper reveals that deploying many benign skills together introduces compositional security risks: multiple individually clean skills can compose into safety-violation chains, because the agent passes an upstream skill's output to a downstream skill's input across invocations. The risk arises not from any single skill, but from the agent's runtime orchestration, and thus cannot be captured by existing malicious skill detection methods. To this end, we develop KnotScryer to detect insecure skill compositions by treating skills as typed capability flows. In particular, we abstract each skill into a capability summary, and construct the Skill Capability-Flow Graph (SCFG), where edges denote cross-skill information flow. Moreover, we design a search policy that tracks protected or untrusted data along handoff edges to locate paths reaching insecure operations. Finally, a two-layer verification method replays the insecure information flow for reachability analysis and applies a task-anchored LLM chain audit to confirm the semantic plausibility of the identified violation chains. We deploy KnotScryer to analyze 13,598 clean skills from ClawHub, and validate the detected chains across three agent frameworks and three LLMs. KnotScryer identifies 612 insecure compositions, which are consistently exploited across agent frameworks. Meanwhile, under GLM-5.1, it outperforms tag-match and graph-only baselines by improving the exploitation rate from 16.5%-18.2% to 85.6%, while reducing the search space by 85.3%. Furthermore, 13.7% of the detected chains consist entirely of skills labeled safe by a single-skill scanner, showing that KnotScryer effectively identifies insecure compositions that per-skill auditing cannot.

0.09

Systems & Networking 31

Quantum Cadence: Post-Quantum Consensus with Multiple Concurrent Proposers

cs.DC cs.CR ·2026-10-23 ·Mohammad Mussadiq Jalalzai, Tobias Klenze, Jovan Komatovic
Abstract · 35 pages, 3 figures

The consensus protocols of most high-performance blockchains use quorum certificates, which bundle the signatures of a quorum of validators. Signature schemes such as BLS can aggregate the signatures of a whole quorum into a single short signature, which keeps these certificates small. A large-scale quantum computer would break these schemes, and no post-quantum scheme with comparable aggregate size and aggregation time is known. Without aggregation, a certificate holds one large post-quantum signature per member of the quorum, so it becomes infeasible for each validator to broadcast certificates, as modern consensus protocols often require. We present Quantum Cadence, a post-quantum consensus protocol with multiple concurrent proposers that avoids certificates altogether: every validator counts quorums among the messages it has received itself. Its safety and liveness do not even depend on signatures, although we add signatures to obtain accountability. Quantum Cadence builds on Cadence and replaces its slot consensus, Chorus, with Quantum Chorus, which keeps the optimal fast-path latency of Chorus: dissemination followed by two rounds of votes. When its fast path fails, Quantum Chorus runs Coda, our signature-free Byzantine agreement for multiple proposers. We also outline how the Conductor, which schedules the slots of Cadence, can be adapted to Quantum Cadence. In simulations with 200 validators, Quantum Chorus has a latency comparable to that of Chorus and scales better with the number of proposers.

0.50

Measuring the Energy Cost of Completing Tasks with Quantized Language-Model Agents

cs.PF ·2026-10-23 ·Stephen Abkin, Aldric Yang, Alex Flesher, Nitheesh Kodarapu, Deepak Panigrahy +2 more
Abstract · 26 pages, 7 figures. Alex Flesher and Nitheesh Kodarapu contributed equally

Quantization reduces the memory and energy needed to serve large language models (LLMs) and AI agents for coding and data analysis can now run on locally hosted LLMs. In practice, this benefit is commonly benchmarked per output token. Yet an agent calls its model repeatedly, runs tools, and revises its work, and quantization can change both how much the agent generates and whether it succeeds. Existing energy studies either compare quantized models on single requests or measure agents at one fixed precision, overlooking how a change in precision plays out across a complete agent task. To address this gap, this paper presents the first characterization of the task-level energy trade-offs of model quantization in locally hosted agents. We develop 29 agent tasks with automatic pass/fail checks in six families, ranging from single-answer SQL analytics to multi-step repository bug fixing, and measure CPU and GPU energy per output token and per completed task for 4-bit and 8-bit versions of two open-weight models on bare-metal NVIDIA A100 nodes with llama$.$cpp. Our findings reveal that 4-bit precision generally saves energy per token, though whether it saves energy per completed task is task dependent and changes with the model and the execution budget (i.e., limits on turns, tokens, and time). We trace this discrepancy to the extra tokens needed to complete a task in full, which translates into using up to 3.5 times as much energy per completed task. These findings offer actionable guidance for developers and operators choosing model precision for agents, showing how tokens per completed task, set against the per-token saving, reveal whether lower precision completes the work with less energy.

0.48

TranSAGE: Automating Spatial Accelerator Generation for Large-Scale Transformer Models

cs.AR ·2026-10-23 ·Chi Zhang, Yi Sun, Weichuang Zhang, Quan Chen, Jieru Zhao
top venuecode/data
Abstract · Accepted to the 32nd ACM International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS 2027)

Large-scale Transformers have achieved remarkable success across NLP, computer vision, and robotics, but their growing computational and memory demands challenge efficient inference. Spatial accelerators offer advantages in on-chip bandwidth and energy efficiency, yet designing them for large models remains costly and time-consuming. Existing accelerator generators typically focus on single-operation support, overlooking the benefits of multi-operation pipelining, which limits throughput and energy efficiency. We propose TranSAGE, a compiler-based framework that jointly optimizes scheduling for large-scale Transformers under multi-operation pipelining and generates corresponding accelerator implementations. Our approach integrates a design space exploration (DSE) engine with multi-operation dataflow optimization and a dynamic programming-based graph partitioning algorithm to efficiently identify pipeline-friendly operation clusters and their optimal dataflow parameters. We further design spatial architecture primitives and lowering passes to automatically generate efficient RTL implementations. Evaluations on four representative LLMs demonstrate that TranSAGE achieves up to 2.41x higher throughput and 13.04x higher energy efficiency than commercial GPU/NPU baselines. Compared with state-of-the-art accelerator generators, TranSAGE achieves up to 11.33x higher throughput and 6.07x higher energy efficiency. Our code is available at https://github.com/sjtu-zhao-lab/TranSAGE.

0.40

Migration as a Steady-State Execution Primitive: A Runtime for Virtual Stationarity in Low-Earth Orbit

cs.DC cs.NI ·2026-10-23 ·Daniel Kleebinder, Alireza Furutanpey, Stefan Nastic
published
Abstract · Accepted for publication at the 16th International Conference on the Internet of Things (IoT 2026)

Low Earth Orbit (LEO) satellite mega-constellations enable orbital edge computing for latency-critical geocentric applications, but satellite motion compresses line-of-sight communication windows to mere minutes. To maintain logical geographic proximity, a paradigm termed Virtual Stationarity (VS), applications must continuously migrate across passing satellites in the constellation. Traditional live migration mechanisms optimize individually initiated relocations or replication, but not the continuous handover between successive hosts that orbital motion requires. We reformulate live migration as a steady-state execution model and introduce a novel virtual stationarity runtime paradigm. We formalize the viability of continuous migration through a lightweight model capturing latency and availability service-level objectives (SLOs), continuous migration overhead, and the convergence bound of state replay. To satisfy these constraints, we derive a Virtual Stationarity Runtime (VSR) framework whose Migration Membrane combines Capability Virtualization and Continuous State Projection to rebind host resources and reconstruct evolving execution on a successor. Evaluation on physical heterogeneous hosts shows 95th-percentile downtime below 6 milliseconds across 300 recurrent handovers, migration time effectively invariant to memory-write rate, and no cumulative growth in migration time or continuation state. A 1,600-satellite simulation further confirms completion within orbital handover windows.

0.39

Nemo: Efficient Caching with Index Pushdown for Disaggregated OLTP Databases

cs.DB cs.DC ·2026-10-23 ·Wenjie Hu, Xinjing Zhou, Zhihan Guo, Xiangpeng Hao, Xiangyao Yu
published
Abstract

Storage-disaggregated databases decouple compute from storage, improving elasticity but suffering from network bottlenecks. Most cloud databases use page-based B+trees and buffer pools. Under disaggregation, each index traversal incurs serial network round-trips. Page-granularity transfers and caching also amplify network traffic and lower cache efficiency. Pushing down the full index to storage and caching only hot records at compute is appealing, but an efficient design must (1) provide a holistic record cache that accelerates both point and range queries, (2) mitigate the negative search penalty of confirming non-existence via remote probes, and (3) enforce phantom protection despite the index being encapsulated by the storage service. We introduce Nemo, a holistic index-and-buffer design optimized for storage-disaggregated OLTP with index pushdown. Nemo employs a tree-structured record buffer that unifies caching and buffering and accelerates both point and range queries. It further reduces the negative search penalty by caching negative search results. Nemo provides phantom protection that requires no knowledge of the storage-side data layout and is robust against large key domains, varying data skew, and partition schemes. Experiments show that Nemo outperforms page-based B+Tree and the state-of-the-art 2-Tree by up to 11.42x and 2.97x on positive point queries, and achieves up to 4.76x speedup on negative searches over a hash-indexed record cache. For phantom protection, Nemo outperforms PCL by up to 1.55x even under a dense key domain. Moreover, Nemo maintains stable performance while PCL degrades sharply as the key domain grows large and sparse.

0.39

Peregrine: A Language-Agnostic Packet Processing Intermediate Representation with Formal Guarantees

cs.NI ·2026-10-23 ·Sean Rhee, Xiangyu Gao, Srinivas Narayana, Nate Foster, Joseph Tassarotti +1 more
code/data
Abstract · 16 pages, 7 figures, 4 tables

Many different programmable network devices and domain-specific languages (DSLs) for programming them have been developed in recent years. However, the affiliated compiler infrastructure for each DSL has not kept up, leading to unreliable compilation results and ultimately a poor experience for developers. In response to this fragmented ecosystem, we propose an intermediate representation, Peregrine, to consolidate the common compiler functionalities within this domain into a shared, open-source infrastructure. We designed Peregrine to provide formal guarantees from the beginning, rather than adding them in a later step. The language comes with a formal semantics, along with an equivalence checker written and proven correct in Rocq. Experimental results show that Peregrine is general enough to express multiple applications written in various state-of-the-art network programming languages. The equivalence checker designed in Peregrine can be used for validating the correctness of a range of untrusted compiler passes, as well as for other surprising use cases such as certifying the correctness of packet classification algorithms within a reasonable amount of time.

0.38

Bounding Queue Waits for Fair Scheduling of Bursty HPC Workloads via Hierarchical Algorithms

cs.DC ·2026-10-23 ·Karaen Senthilkumar, Suman Raj, Maxime Gonthier, Kyle Chard, Donpaul Stephens +1 more
hot direction
Abstract · 12 pages, 9 figures, 4 tables

HPC clusters are shared among many projects, each holding an allocation of node-hours, and ensuring fairness is becoming harder. Foundation models workflows and automated pipelines have made job submissions bursty and concentrated among a few heavy users, so occasional users absorb queueing delays created by frequent ones, with no indication of how long their jobs will wait. Existing schedulers either forget past usage, protect only the job at the head of the queue, or assume that running jobs can be preempted. We propose HyDRA, a Hybrid Deadline and Ranking-Aware scheduler that enforces fairness among users while bounding the queue wait of every job. HyDRA assigns each job a deadline on arrival based on its size, the priorities of its project and user, the current backlog, and how far its project and user are from their entitled share. It ranks waiting jobs by a priority score that grows with waiting time, serves jobs close to their deadlines first, and backfills a job only when no deadline, or fairness limit is violated. We evaluate HyDRA on workloads derived from a year of ALCF Polaris job traces, against conventional schedulers and variants inspired by PCS, Themis, and the PBS job-sort formula used at ALCF. HyDRA achieves $2\times$ the project-level Jain's fairness of the best baseline, starts 47\% of jobs by their deadlines compared with at most 36\% for the baselines even for largest workloads, and lowers the 95th-\%ile bounded slowdown by 4.2$\times$ relative to the next best baseline, at a moderate cost in utilization.

0.38

MetaDB: Putting File-System Structure Back into Distributed Databases

cs.DB cs.DC ·2026-10-23 ·Zebin Ren, Biao Cao, Kang Chen, Zhihui Yin, Qiushi Chen +8 more
published
Abstract · 16 pages, accepted by NSDI'27

Storage systems in production increasingly rely on distributed databases for metadata management, yet our experience at Baidu shows that treating a general-purpose database as a black box leaves substantial performance and efficiency untapped. File-system metadata exhibits directory-tree locality, differentiated consistency requirements, and clustered deletions that general-purpose databases largely ignore. We thus propose MetaDB, a metadata-aware distributed transactional database that serves as the shared metadata layer for Baidu's large-scale storage systems. MetaDB addresses three key mismatches between general-purpose databases and metadata workloads: Data Model Mismatch, Transaction Model Mismatch, and Storage Implementation Mismatch. To bridge these gaps, MetaDB combines (1) tree-topology-induced sharding and execution, (2) semantic-aware off-critical-path processing, and (3) a lifecycle-aware storage engine with in-memory MVCC and locality-aware compaction. MetaDB achieves 2.2x-4.1x the throughput of our previous general-purpose database and halves the number of metadata machines. Deployed in production at Baidu for over five years, MetaDB unifies metadata management across the company's storage fleet and scales to over 1,000 metadata nodes, tens of trillions of metadata records, and exabyte-scale storage in a single cluster.

0.37

Concurrent Task Orchestration: Hierarchical Trusted Collaborator Group Formation for Constrained Resource Allocation

cs.NI ·2026-10-23 ·Jiazhi Chen, Xianbin Wang, Xuemin "Sherman" Shen
Abstract · 13 pages, 7 figures. Submitted to IEEE Transactions on Cognitive Communications and Networking (TCCN)

Effective collaboration among resource-constrained mobile intelligent machines, such as Unmanned Aerial Vehicles (UAVs), is essential for the completion of complex tasks. However, concurrent task demands, limited distributed resources, and dynamic collaborator availabilities make it challenging to satisfy task-specific completion goals in diverse collaborative systems. To address these challenges simultaneously, we propose a new framework of trusted collaborator group formation and the supporting resource allocation strategy for concurrent task orchestration. Specifically, a collaborative task is decomposed into subtasks subject to dynamic resource availability and inter-subtask dependencies. A task utility model is developed for collaborator group evaluation support subtask allocation according to task completion goals and unified value-risk analysis. Based on this model, concurrent task orchestration is formulated as a long-term utility maximization problem under $α$-fairness. To solve the problem efficiently, we develop a hierarchical matching strategy that first clusters tasks and collaborators into coarse-grained sets and subsequently performs fine-grained collaborator group formation and task allocation within matched sets. Task-specific trust is predicted through spatial-temporal Graph Neural Networks (GNNs) and embedded into a fairness-aware Deep Reinforcement Learning (DRL) framework to continuously update the matching policies. Simulation results demonstrate that the proposed strategy improves concurrent task completion performance by 22% over representative baselines under resource constraints.

0.35

Hardware-Friendly Estimation of LCU-Based Expectation Values via Quantum-to-Classical Workload Shifting

cs.ET quant-ph ·2026-10-23 ·Hari Hara Suthan Chittoor, Thirumalai Manimaran
Abstract · 12 pages, 3 figures

Estimating the expectation value of an observable $O$ for a quantum state evolved under a general, possibly non-unitary, operator $A$ is a fundamental task in quantum simulation, chemistry, and variational algorithms. Existing linear-combination-of-unitaries (LCU) approaches implement $A$ on hardware, requiring ancilla qubits, controlled unitaries and, in randomized variants, fresh circuit executions for every Monte Carlo sample. These costs are prohibitive on NISQ and early fault-tolerant devices. We propose HERMES (Hardware-Efficient Reusable Measurements for Estimating LCU-based expectation valueS), which moves the processing of $A$ entirely to the classical domain. Expanding $A$ and $O$ in the Pauli basis and sampling pairs of Pauli terms classically, HERMES rewrites $μ= \mathrm{Tr}(OAρA^{\dagger})/\mathrm{Tr}(AρA^{\dagger})$ as a ratio of weighted sums of Pauli expectation values on the unmodified state $ρ$, with no ancillas, controlled unitaries, or quantum implementation of $A$. A two-level measurement reduction scheme (Pauli deduplication and commuting-group partitioning) bounds the number of distinct measurement settings independently of the sample count, unlike randomized LCU methods whose circuit executions grow linearly with it. On ground-state energy estimation of $\mathrm{H}_2$ and $\mathrm{LiH}$ via imaginary-time evolution, HERMES matches or beats Pauli single-ancilla LCU baseline using shallow circuits with no two-qubit gates beyond state preparation under qubit-wise commuting grouping. It is also more robust to noise and completes a noisy $\mathrm{LiH}$ simulation infeasible for the baseline, at the cost of classical overhead growing with the number of Pauli terms in $A$.

0.35

Tile-Based Decompression in Compiled GPU Engines

cs.DB cs.DC ·2026-10-23 ·Antonio Boffa, Ananya Gupta, Hamish Nicholson, Anastasia Ailamaki
top venue
Abstract · 15 pages, 6 figures, 3 tables. Accepted to PVLDB Vol. 20

Modern GPUs reach record analytical performance only when a query's working set fits within its GPU memory budget, and that budget is under growing pressure: GPUs are partitioned and shared, working sets are outgrowing high-bandwidth memory (HBM), and HBM is a fixed on-device resource whose per-dollar capacity has stagnated across generations. Lightweight compression can relieve this pressure by shrinking both footprint and bandwidth demand, but existing GPU compressed-execution techniques either bind decompression to a specific execution model or materialize decompressed columns in HBM before query execution. We present a compression subsystem for compiled GPU query engines that treats decompression as a first-class physical operator. Compression-specific logic is isolated behind a narrow backend interface, so new GPU-oriented encodings can be added without rewriting downstream operators. We implement the design in Proteus and study where decompressed values should be materialized in the GPU memory hierarchy. Our compiler-integrated shared-memory strategy fuses cooperative tile decompression with tuple-at-a-time query execution inside a single generated GPU kernel, avoiding decompressed-column materialization in HBM. Compared with HeavyDB and Sirius on GPU-resident data, fused on-chip decompression achieves competitive query times while reducing the resident memory footprint by up to 4x and memory-bandwidth consumption by up to 6x.

0.33

CCAaaS: An Orchestration Platform for High-Velocity Design, Development, and Deployment of Data Center Transports at Meta

cs.NI ·2026-10-23 ·Prankur Gupta, Balasubramanian Madhavan, Miao Xu, Prashanth Kannan, Jie Meng +5 more
published
Abstract · 19 pages, 11 figures, 9 tables. To appear in NSDI '27 (24th USENIX Symposium on Networked Systems Design and Implementation)

Today, data center networks run multiple congestion control algorithms (CCAs), each tailored to distinct traffic characteristics, from latency-sensitive intra-region RPCs to high-throughput cross-region AI workloads. The rapid evolution of AI training and inference further accelerates this trend by continually introducing new traffic characteristics that demand new CCAs. This paper introduces CCA-as-a-Service (CCAaaS), our eBPF-based ecosystem for the rapid design, deployment, and operation of production-grade CCAs at hyperscale. We present (1) a modular BPF-TCP framework and general lessons for designing modular CCAs; (2) a deployment and orchestration layer that provides disruption-free CCA updates, along with best practices for managing a zoo of CCAs at scale; and (3) a lightweight observability tool for CCAs and a blueprint for scaling CCA observability. CCAaaS manages over 97% of our server-to-server traffic using three active eBPF-based CCAs; two of the five CCAs we have built have since been deprecated. We show that our eBPF-centric approach to CCAs development delivers the same application-level performance (retransmissions, bandwidth, and tail latency) as their native in-kernel counterparts, and that our custom CCAs add negligible CPU and memory cost over their eBPF baselines. In terms of realized benefits, the CCA built with CCAaaS significantly improved network, and application metrics, for example, one of our CCAs reduced P99 latency for cross-region services by up to 20X.

0.32

PARIS: Parallel Arithmetic Reduction and Integrated Selection for MSDF Inner Products

cs.AR ·2026-10-23 ·Malik Zohaib Nisar, Ic-Pyo Hong, Dorit Merhof, Muhammad Usman
Abstract

Existing most-significant-digit-first (MSDF) accelerators serialize input-term processing, causing inner-product latency to increase with fan-in. This paper presents PARIS, an MSDF architecture that eliminates input-term serialization through parallel bit-plane reduction and shared carry-save residual selection. For $B$-bit operands, PARIS processes $B^2$ bit-position pairs, independent of fan-in within the supported tile, while generating exact MSDF outputs. Synthesized in 45\,nm at 100\,MHz, the 784-lane INT8 core occupies 0.008701\,mm$^2$ and completes an inner product in 88 cycles. Evaluation on two MNIST multilayer perceptrons demonstrates arithmetic-workload speedups of 240.44$\times$ and 180.47$\times$ over a matched reused-MSDF implementation, reducing arithmetic energy by 98.89\% and 98.58\%, respectively. A synthesized 16-engine accelerator occupies 0.90273\,mm$^2$ and meets the 100\,MHz timing constraint. Assuming immediate weight availability, it delivers a projected throughput of 24.33\,GOPS and an estimated energy efficiency of 1,581.14\,GOPS/W at unchanged mapped power. Bias-aware ReLU detection further reduces modeled arithmetic cycles by 5.52\% and 4.86\%. A separate 256-engine RTL implementation completes Model A in 259 cycles, demonstrating output-parallel scalability.

0.30

LooseLeaf: Enabling Serverless GPUs for Computational Notebooks

cs.DC ·2026-10-23 ·Yusen Wu, Tingfeng Lan, Benjamin Carver, Haoliang Wang, Geoffrey Fox +1 more
Abstract · 15 pages, 13 figures, 3 tables

Computational notebooks reserve a GPU for the whole session, yet the GPU computes only while a cell executes and sits idle otherwise. Serverless execution would release the GPU between cells, but reconstructing the accumulated notebook state on a fresh worker imposes substantial cost on every GPU cell. This paper presents LooseLeaf, a serverless GPU platform for interactive notebooks. LooseLeaf derives the state and import closures of each cell, the minimal variables and packages that the cell needs. Our analysis on 5,000 real-world notebooks shows that these closures are small, and user-cell editing leaves seconds to prepare them. LooseLeaf prepares a GPU worker with these closures while the user edits the cell. After execution, the worker saves the updated state within seconds, so the GPU is reclaimable after every cell. LooseLeaf starts 95-98% of GPU cells with no visible delay, matches the interactivity of a pinned session, and provisions 32-55% fewer GPU-hours than state-of-the-art notebook platforms.

0.29

Multi-Timescale Context-Aware Radio Resource Optimization for Harmonious NTN-TN Coexistence: A Digital Twin and Open RAN Enabled Framework

cs.NI eess.SP ·2026-10-23 ·Ecem Izde Kaya, Md. Zoheb Hassan, Georges Kaddoum
benchmark
Abstract · Submitted to IEEE for possible publication

Coexistence between non-terrestrial networks (NTNs) and terrestrial networks (TNs) is essential for future wireless systems, but terrestrial downlink transmissions can generate harmful aggregate interference at low-Earth orbit (LEO) satellite receivers. We present DT-STABLE, an Open Radio Access Network (O-RAN)-enabled, digital twin (DT)-assisted framework for context-aware NTN-TN coexistence. Using site-specific information, the DT estimates interference at LEO satellites, enabling proactive management with reduced real-time measurement overhead. We formulate joint beam activation and user association under RF-chain limits, load balancing, and aggregate satellite interference constraints; prove NP-hardness; and develop an efficient DT-guided scheduling algorithm. DT-STABLE couples short-timescale radio resource adaptation with periodic mid- and long-timescale DT updates through standardized O-RAN interfaces. Site-specific 28 GHz simulations under worst-case satellite geometry and low elevation angles show that DT-STABLE balances rate and interference utility, achieving 89.63% interference-to-noise ratio (INR) satisfaction and a mean sum rate of 210.96 bps/Hz, with nearly 100% INR satisfaction when satellite protection is prioritized. The framework reduces beam measurement overhead by 99.71% relative to exhaustive sweeping and achieves up to 33.1 times lower runtime than the Gurobi-based exact optimization benchmark.

0.29

Scaling Graph Query Processing: Focused Exploration on Partitions Using Load Balancing

cs.DB cs.DCcs.IR ·2026-10-23 ·Harshit Modi, Abhishek Santra, Sharma Chakravarthy
Abstract

The focus of this paper is on query processing and not search (approximate or exact). The goal is to find all or k answers and the expressiveness of a query (as opposed to a search expression which is typically simpler) is considered important. Although SQL query processing has matured over the last several decades, query processing over graph data has not received the same attention. Only recently, systems, such as Neo4J, and others have started using graphs for data representation on a large scale. Their approaches and algorithms remain proprietary. In this paper, we address the problem of answering graph queries and their efficient processing on arbitrary-sized graphs. We propose the partitioning of graphs and the use of the distributed paradigm to achieve efficiency and scalability. Our approach takes a subgraph as query input including wild cards and comparison operations. Our partitioned approach leverages focused exploration and distributed processing to achieve the desired parallelism and improve the response time. We also present heuristics to reduce the number of partitions loaded for query processing. We present our partitioning algorithm, distributed processing architecture used, and load balancing for computing query answers. We establish the correctness of the approach in the presence of partitions, compare heuristics for load balancing, and extensive experimental results over large synthetic and real-world graph data sets.

0.28

Embodied Emissions of 2.5D and 3D Scaling in Chipletized GPUs for Low-Carbon AI

cs.AR ·2026-10-23 ·Hannah Watson, Arjun Singh, Jennifer Stephenson, Job Soethoudt, Jonas Svedas +2 more
Abstract

GPU compute resources are scaling to support growing AI workloads, leading to further rises in the climate change impact of hyperscale data centers that rely on high-performance GPUs. To model how packaging decisions impact the carbon footprint we have estimated the embodied emissions (EE) of a chiplet-based GPU that has been scaled in 2.5D and 3D using the imec.netzero modelling platform. We doubled the number of compute units and high bandwidth memory (HBM) capacity in the 2.5D and 3D scaled systems and found that 3D scaling results in an 89% increase in EE, as compared to an increase of 141% for 2.5D scaling, positioning 3D scaling as an attractive solution to limit the embodied emissions due to GPU scaling.

0.27

Quantum-Classical Link Prediction on Dark Web Hyperlink Graphs

cs.NI cs.LG ·2026-10-23 ·Vojtěch Novák, Radovan Rečka, Ondřej Knebl, Aktham Youssef, Ivan Zelinka
Abstract

We analyze HTML content and hyperlinks from an offline Tor crawl containing 383 onion hosts and 2,067 directed cross-host links. The study combines semantic analysis with degree-controlled quantum--classical link prediction. Two directory groups emit 94.8\% of the observed host links. Directed preferential attachment reaches ROC-AUC 0.976 with random non-edges, but falls to $0.550\pm0.014$ under source-conditioned target-degree matching. With disjoint training and test pairs, multimodal logistic regression reaches $0.672\pm0.021$, GAE $0.899\pm0.016$, and VGAE $0.906\pm0.016$. A four-qubit variational quantum classifier (VQC) comparison tests two encodings across 20 disjoint-pair graph splits under exact and 1,024-shot simulation, with classical controls using the same inputs and training examples. Its best exact mean ROC-AUC is $0.549\pm0.023$, compared with $0.551\pm0.016$ for RBF SVC and $0.573\pm0.029$ for a 16-parameter MLP on the same inputs and training examples. These results support predictive signal in combined graph/content features and show why degree control, feature compression, and optimization conditions matter in quantum--classical comparisons.

0.26

GLP Networking API Specification

cs.NI cs.DCcs.PL ·2026-10-23 ·Dan Bachar, Ehud Shapiro
Abstract

Multiagent deterministic GLP (madGLP) runs Grassroots Logic Programs (GLP) as agents that communicate solely by asynchronous message passing, and assumes fair delivery: every message sent is eventually delivered. Each agent runs on a person's smartphone, named by the person's public key. A grassroots platform depends on no global resource but the network, so agents must reach one another over Bluetooth Low Energy, on a local network, and across the Internet from behind NAT, via rendezvous servers anyone can run. What madGLP requires of such a network, and what the network requires of the platform, has not been specified. We specify the networking layer beneath madGLP. Anything that can be done in GLP is done in GLP: the layer provides only what GLP cannot reach - the network stack, the location service, application attestation, and cryptographic primitives. Three transport underlays share one interface. Over PAN and LAN, an agent advertises a rotating recognition token derived from its key and time slot, so a peer who knows the key recognises it before contact and nobody else can track it; whether it answers unrecognised contacts is a per-underlay trust level set by GLP. Over WAN, a rendezvous server is a static-address host running the headless layer and a GLP rendezvous agent that coordinates hole punching, relays no payload, and belongs to no social graph. On every underlay, peers establish a mutually authenticated Noise session from Ed25519 keys, then attest, bound to the handshake hash, that each runs the application published under its developer's key, verified against Apple's and Google's roots with no third party contacted. One message protocol, over UDP datagrams and GATT writes, provides the fair, unordered, atomic and authenticated delivery madGLP assumes.

0.26

Experimental Classification of xApp Conflicts in O-RAN: Slicing vs. Per-UE Scheduling

cs.NI ·2026-10-23 ·Rashmi Sharan Sinha, Merim Dzaferagic, Marco Ruffini, Nima Afraz
Abstract · Accepted for 2026 IEEE Globecom Workshops (GC Wkshps): 7th Workshop on Emerging Topics in 6G Communications (GC Wkshps 2026-Topics-in-6GComm…

Concurrent applications (xApps) in the Near-Real-Time RAN Intelligent Controller (Near-RT RIC) of the Open Radio Access Network (O-RAN) can issue control intents that are jointly infeasible when they act on hierarchically coupled radio resources. This paper experimentally studies an interaction between slice-level physical resource block (PRB) allocation and per-user-equipment (UE) scheduling. We implement a utilisation-aware slicing control (the spectral-efficiency xApp) that reclaims a fixed fraction of the baseline allocation from under-utilised slices and redistributes pooled capacity to saturated slices. The control logic is exercised in simulation and on a 5G testbed comprising an OCUDU-based next-generation NodeB (gNB), an Open5GS core, a Universal Software Radio Peripheral (USRP) X310, and five commercial off-the-shelf (COTS) UEs. Because the gNB counters available in this testbed are UE-keyed but not slice-labelled, the slicing control derives a per-slice throughput proxy from core-side traffic measurements. A command-bus detector evaluates explicitly defined control-intent conditions: a per-UE floor above its parent-slice cap, aggregate UE floors above the cell budget, and repeated slice-target changes within a specified window. Over a 276-cycle continuous run, the detector logged 147 per-slice floor--cap violations, 49 shared-budget over-subscription episodes, and 11 allocation-oscillation episodes. Against the conventional static slice split replayed on the same load trace, the slicing control reduces unmet demand by 12--33\,\% and idle allocation by 21--23\,\%. The results demonstrate reproducible cross-granularity control-intent interactions on real hardware and provide an experimental basis for analysing conflict detection in O-RAN.

0.25

How AI Load Swings Reduce Datacenter Power Conversion Efficiency

cs.AR eess.SY ·2026-10-23 ·Heetarth Joshi, Ali Alenezi, Fang Chen Lin, Prasad Enjeti, Xiaoyang Wang +1 more
Abstract

AI training and inference workloads make GPU power swing repeatedly between compute and communication phases. As a result, the cascaded converters in a datacenter rack keep moving between operating points with different efficiencies. This paper uses an energy based, cycle averaged method to quantify that effect along a 240 V AC to 400 V DC to 51 V DC rack power path, using measured efficiency curves of commercially available GaN based PFC and LLC converters. For a 1 Hz, 50% duty-cycle load swinging between 5 kW and 510 W, conversion efficiency falls to 95.97%, compared with 96.82% at the equivalent constant average load, a loss of 0.88%. When 51/48V to 1V point of load conversion to the chip rail is included, it is shown that the loss grows to 1.66% (89.45% vs. 90.96%). These results show that efficiency rated at average load overstates real world performance, and that the penalty grows with each load dependent conversion stage between the grid and the GPU.

0.24

Toward Open ISA Supercomputing: A Comparative Study of RISC-V, Arm, and x86 Architectures on HPC Workloads

cs.DC cs.ARcs.PF ·2026-10-23 ·Emanuele Venieri, Giacomo Madella, Gabriele Ceccolini, Simone Manoni, Federico Ficarelli +7 more
code/data
Abstract · Preprint submitted to Future Generation Computer Systems

RISC-V is emerging as a candidate architecture for high-performance computing (HPC), but its competitiveness against established x86 and Arm processors remains difficult to assess because existing studies use different platforms, workloads, and methodologies, and rarely consider energy efficiency. Our work provides an evaluation of contemporary HPC RISC-V, Arm, and x86 processors using a common software stack and methodology. We characterize the Sophgo SG2044 and compare it with the NVIDIA Grace CPU Superchip and Intel Xeon Platinum 8480+ across four HPC benchmarks (STREAM, HPL, HPL-MxP, and HPCG) and four representative applications (iPIC3D, PLUTO, OpenFOAM, and llama$.$cpp). The version of HPL-MxP linked to OpenBLAS built and used for this work is made available as open-source software. The SG2044 reaches 8.02 GFLOP/s of FP64 performance and 7.8 GB/s of memory bandwidth per core, increasing to 294 GFLOP/s and 107 GB/s at 64 cores. L2-cache-aware process placement enables near-linear scaling up to 16 cores, beyond which memory-bandwidth saturation limits further gains. Across generations of RISC-V hardware, in the context of the Monte Cimone (MC) project, the SG2044-based MCv3 doubles both single-core performance and peak memory bandwidth relative to MCv2, while improving memory bandwidth, node-level FP64 performance, and energy efficiency by approximately 100x, 158x, and 10x, over MCv1. Against the Arm and x86 systems, the SG2044 shows 8-15x less absolute performance, reflecting its narrower vector resources and less advanced technology node, but delivers competitive scaling and energy efficiency on several workloads. Its most notable result is LLM decode, where 64-core execution provides the highest relative scaling and energy efficiency among the evaluated platforms while exceeding Intel in throughput. Our results show that RISC-V processors are closing the gap with the competition.

0.23

When Simple Is Enough: Resource-Aware Heuristic versus DDQN for O-RAN Mobility Robustness

cs.NI ·2026-10-23 ·Ali Deylamy, Nima Afraz
Abstract · accepted for 2026 IEEE Globecom Workshops (GC Wkshps): Workshop on Edge Intelligence meets AI-Native RAN: Theory, Algorithm and Implementati…

Open Radio Access Network (O-RAN) enables mobility-management xApps, but a learned controller is not necessarily the most effective or efficient handover policy. This paper compares standard Event A3, a resource- and mobility-aware heuristic, and a Double Deep Q-Network (DDQN) through a common xApp-oriented interface and paired system-level simulation. The heuristic combines radio quality, target-cell resources, UE mobility, and service class, with weights optimized offline using NSGA-II, while DDQN-Multi uses temporal radio and resource observations. All controllers are evaluated on identical UE trajectories, traffic assignments, and ray-traced radio inputs under Base, High-Mobility, and High-UE scenarios. Performance is assessed using service, handover, resource, energy, and fairness KPIs, complemented by CRITIC--TOPSIS ranking and complexity analysis. Results show that the heuristic achieves the lowest outage in all three scenarios and ranks first in the CRITIC--TOPSIS comparison, while retaining linear online complexity and interpretable decisions. DDQN-Multi improves target feasibility under congestion but requires offline training and produces more handovers and ping-pong events. The results indicate that lightweight resource-aware control can provide a favourable performance-complexity trade-off relative to DDQN under the considered scenarios.

0.20

Automated Multi-Line Bug Repair Using Model Checker Aided LLMs: A Case Study on RISC-V Processors

cs.AR cs.SE ·2026-10-23 ·Debleena Das, Chandan Kumar Jha, Muhammad Hassan, Rolf Drechsler, Ansuman Banerjee +2 more
Abstract

Functional verification of Register-Transfer Level(RTL) designs has become a critical bottleneck in modern semiconductor development, often consuming the majority of project resources. Debugging, in particular, remains an especially time-consuming task, requiring careful analysis of simulation traces or counterexamples. Recent work has explored applying Large Language Models (LLMs) to RTL debug and repair with great promise. However, the existing approaches typically are for single-line bugs, operate in open-loop, rely heavily on prompts or retrieval, lack scalability, and often produce incorrect fixes. In this paper, we propose a novel automated RTL bug repair methodology that combines LLMs, assertions, and a model checker to provide scalable, trustworthy semantic grounding and formal guarantees. Concretely, the methodology leverages failing assertions to guide the LLM in generating patches for multi-line bugs. Each candidate patch is formally verified using a model checker (JasperGold), ensuring only valid repairs are accepted. We evaluate the efficacy and scalability of our methodology on the BRISC-V RISC-V core using mutation testing, introducing several mutation types, and demonstrate that our methodology can repair up to 98% of single-line RTL bugs and 93% of the multi-line bugs. To the best of our knowledge, this is the first work to show automated, LLM-assisted RTL multi-line bug repair results on a complex RISC-V processor design in a scalable manner.

0.19

Beyond Hotness: High-Resolution Memory Characterization of Cloud and LLM Workloads for Tiered Memory with ARM SPE

cs.DC ·2026-10-23 ·Jie Ren, Ivy Peng
Abstract

Memory demands from data analytics and LLM workloads grow rapidly, driving up memory costs in cloud and data centers. Multi-tier memory has emerged as a promising solution to balance performance and cost. However, effectively leveraging it requires a deep understanding of workload memory behaviors. This work introduces a high-resolution memory characterization methodology for analyzing workload-specific opportunities for multi-tier memory usage. In response to the increasing deployment of ARM-based processors in data centers, we implement our methodology using the recent Statistical Profiling Extension (SPE) for ARM. By characterizing eight memory-intensive workloads from CloudSuite and LLM applications, we identify region-specific access patterns, evaluate the effectiveness of cloud-oriented cache hierarchies, and propose new metrics based on memory idle periods and latency profiles to determine suitable data placement for tiered memory.

0.19

InferSlice: A Reference Architecture for SLO-driven LLM Inference over a Telco Wide-Area Network

cs.DC cs.AI ·2026-10-23 ·Juan Parra-Ullauri, Anthony Chiu, Hari Madhukumar, Zhaozhou Wu, Anderson Bravalheri +4 more
Abstract

LLM inference is increasingly deployed as a core service for enterprise and interactive applications, but sovereignty, compliance, latency and resource-locality requirements can limit the suitability of centralised cloud deployments. Telco operators offer geographically distributed compute and network infrastructure for addressing these constraints. However, existing LLM serving frameworks are designed for tightly coupled data centres and provide limited support for coordinating geographically distributed and heterogeneous resources. We present InferSlice, a reference architecture for Service-Level Objective (SLO)-driven LLM inference over a telco infrastructure, integrating network slicing, GPU partitioning, elastic serving and observability. Led by BT in collaboration with the University of Bristol and NVIDIA, and realised on the JOINER UK national-scale wide-area network (WAN), InferSlice is evaluated on service differentiation across tiers, elastic scale-out, client-edge bandwidth as a bottleneck, and the reach of tensor, pipeline, data parallelism, and prefill/decode disaggregation over the WAN. Our results show that, on a general-purpose telco network, tiers yield distinct latency-throughput regimes, and weight streaming shortens scale-out time. Replication is placement-insensitive but memory-bound; pipeline parallelism tolerates near- and mid-cross-site hops, while tensor parallelism and disaggregation fail beyond a single subnet hop. We share practical insights from operating SLO-driven inference on a telco network, lessons learned and open questions, and hope to encourage further research on distributed and heterogeneous LLM inference systems.

0.19

On Latency-Optimal Implementations of the Timestamp Object

cs.DC ·2026-10-23 ·Aurojit Panda, Scott Shenker, Sam Toueg, Wen Zhang
Abstract

Timestamps are widely used in distributed algorithms to order operations. In this paper, we study the timestamp object (TS), which provides a single operation that returns strictly increasing integers. We ask two questions: What is the maximum resilience and the minimum latency of TS implementations in asynchronous message-passing systems, where latency is measured by worst-case message delay? We answer these questions under various types of failure: crash failures, and Byzantine failures with and without digital signatures. To do so, we give lower bounds and algorithms that determine (1) the optimal resilience of TS implementations under each type of failure, and (2) the optimal latency in all cases but one. Along the way, we introduce a new lower-bound proof technique called a cyclic chain argument, which constructs a cycle of indistinguishable schedules along which the returned timestamps must strictly increase, yielding a contradiction.

0.19

Clean-label Poisoning Attack and Defense Game for Federated Learning in Vehicular Networks

cs.NI cs.AIcs.DCcs.LG+1 ·2026-10-23 ·Yalin E. Sagduyu, Tugba Erpek, Nathan Stephenson, Mengran Xue
Abstract

This paper studies interactions between clean-label poisoning attack and defense in vehicular networks using federated learning (FL). Basic Safety Message (BSM) classification for anomaly detection is modeled as a distributed learning problem over a vehicular network with dynamic topology. As a critical threat to FL, a feature-space clean-label attack is introduced, wherein benign samples are perturbed toward malicious feature extremes, degrading class separability while preserving labels. To counter this, a feature-consistency relabeling defense is proposed to detect and relabel suspicious samples. A game-theoretic framework is developed to capture attack-defense interactions and characterize equilibrium strategies based on attack effectiveness and cost parameters. Results show that the attack scales effectively with compromised nodes, while the proposed defense keeps attack success low without degrading clean accuracy and shifts equilibrium regions toward attack-averse regimes.

0.15

KnitTrans: Coordinated Trajectory Transfer for Effective and Efficient LLM Reinforcement Learning

cs.DC cs.LG ·2026-10-23 ·Bangjun Xiao, Chiheng Lou, Xiaodong Ji, Xiangwei Deng, Yikai Zhao +2 more
Abstract

Trajectory transfer from rollout to training affects both learning quality and system efficiency in large language model reinforcement learning (RL). Enforcing batch-wide training requirements can delay the admission of completed trajectories, resulting in higher architectural staleness: the additional parameter-version gap from disaggregated pipelining relative to colocated execution. Beyond admission, misalignment between data access and training consumption causes redundant movement and unnecessary waiting when transferring large trajectory payloads. We present KnitTrans, a system that coordinates the rollout-to-training path through three designs. First, constraint-preserving mini-batch admission allows flexibility in early mini-batch composition, while checking that enough capacity remains to satisfy the remaining quotas. Second, on-demand trajectory data access matches storage units to consumer assignments and overlaps micro-batch data preparation with computation. Third, multimodal encoding deduplication supports reuse across a mini-batch while preserving gradient contributions from all backbone micro-batches. Experiments show up to 93% lower architectural staleness than with train-batch admission, narrowing the learning-quality gap between disaggregated execution and colocated execution. On multimodal workloads, KnitTrans reduces transfer latency by two orders of magnitude and achieves a speedup of up to 5.8 times in transfer plus training over the baseline.

0.12

OCTOPUS: Towards A Tag Model for Cross-Query Reuse in Semantic Query Processing

cs.DB ·2026-10-23 ·GuangXue Zhang, Nada Lahjouji, Yiming Lin, Ashwin Colaco, Sharad Mehrotra
Abstract

Large language models make it possible to pose semantic questions over diverse, multi-modal data. Recent systems give such workloads the declarative feel of a database through semantic operators: relational operators whose conditions a model judges. Most of them now run on a relational engine, yet almost none keeps what the model says past the query: every judgment is discarded on completion and paid for again the next time a query needs it. We present the tag model and tag operators, the right unit for semantic processing, and OCTOPUS, a prototype built as a middleware over DBMS. Two mechanisms are central. First, the tag: each semantic answer a model produces is written back to the relation as data, which the database engine filters, indexes, and stores as it would any other. Second, the tag operator: the unit into which common semantic operators are translated, which fixes the granularity of reuse. A model judges a tuple once, and later queries reuse that judgment as data: through relationships between predicates, a stored judgment can accept a tuple for a later predicate, eliminate one that cannot satisfy it, simplify its test, or order the search of a query that stops early. On the movie scenario of SemBench, our prototype sends the model 10 to 108x fewer rows than iPDB, ThalamusDB, Palimpzest, and LOTUS at broadly competitive quality, and three of its eight queries issue no model call at all. More broadly, a database that keeps its judgments lets its schema grow from the questions its users ask, so that each semantic answer is paid for once, stays consistent across queries, and can be audited through its provenance.

0.09

Democratizing LLM Training with the Edge: Challenges, Systems, and the Road Ahead

cs.DC cs.LG ·2026-10-23 ·Ziyue Luo, Jiaxuan Cai, Cedric LeDenmat, Srijith Nair, Fatemeh Nourzad +25 more
Abstract

The growing computational power of modern GPU data centers has let a few industrial giants train and deploy large language models (LLMs) on massive amounts of data. Yet the sheer scale of these models and their training data makes centralized development increasingly hard to sustain, and at the same time shuts out smaller institutions that hold relevant private data but lack comparable infrastructure. In this paper, we organize our positions on distributed LLM training around the pretraining and fine-tuning paradigms, covering federated learning, split learning, and decentralized approaches, and we assess their readiness for geo-distributed settings with heterogeneous hardware and unreliable connectivity. We then present a {\em unified} system architecture built around a cloud-hosted parameter server that coordinates heterogeneous clients over outbound gRPC connections, reusing one communication, compression, and fault-tolerance stack for both pretraining and a family of fine-tuning methods. We validate the design with real geo-distributed experiments on GPT-2 Medium and Llama3-1B pretraining across multiple university sites, showing stable convergence under realistic conditions. Finally, we chart the road ahead, identifying open problems that span frontier model support, elastic fault tolerance, adaptive compression, federated alignment, privacy and trust, incentive design, and system--algorithm co-design, all of which must be addressed to make democratized LLM training a reality.

0.04

Software & Programming 33

Trusting the Tool: Asymmetric Deference of LLM Vulnerability Auditors to Static-Analysis Evidence

cs.SE cs.CRcs.LG ·2026-10-23 ·ZhiHong Liu, Shuyao Gao
code/data
Abstract · 33 pages, 9 figures, 5 tables. Submitted to IEEE Access. Code: https://github.com/lzh0082-cloud/CODE-DETECTION

Pipelines that combine large language models (LLMs) with static analysis let the model consume the analyser's output as evidence, but whether the model distinguishes true from false evidence, and whether its scrutiny depends on what the evidence claims, has not been measured. We built an intraprocedural analyser for C/C++ that produces proof obligations and audited vulnerable and patched functions from PrimeVul under six evidence conditions that differ only in the notes, including fabricated assurances that a missing check exists and fabricated alarms that an added check is missing. The study covers 916 pairs in four disjoint sets, eleven model configurations from four vendors and 16,572 audits; no model is trained or fine-tuned. True evidence barely improved accuracy (+0.02). False evidence acted asymmetrically: relative to true notes about the same operations, false assurances lowered the detection of real vulnerabilities by up to 19 points, whereas false alarms never significantly lowered accuracy and were disputed far more often. An internally registered replication on seven new models confirmed that false assurances move verdicts towards their claim more than false alarms do (mean asymmetry +0.16, 95% CI +0.10 to +0.24). Instructions in the audit prompt did not reliably reduce deference; the best one failed to replicate. Verifying each claim in a separate call and showing the refuted claims raised accuracy under false assurances by 12 points in a pre-specified test (p=0.002) without loss under true evidence; for a second model the gain was smaller and not significant. Pipelines should therefore verify a tool's guarantees before the model relies on them, and studies should measure deference in a direction-aware way.

0.70

Tool-Calling LLM Agents Under Realistic MCP Faults: Scale Does Not Purchase Fault Tolerance

cs.SE cs.AI ·2026-10-23 ·Antonio Ken Iannillo, Joshua Segun Owotogbe, Roberto Natella, Francesco Avallone, Kristina Kudryavtseva +1 more
publishedcode/data
Abstract · Accepted at NeurIPS 2026 (poster). 10 pages of main text plus appendices and the NeurIPS checklist. Code and data: https://gitlab.com/unilux…

LLM agents increasingly rely on external tools exposed via the Model Context Protocol (MCP), yet current benchmarks assume tools respond correctly and on time, leaving agent resilience to realistic tool failures untested. We introduce MCP-Injector, a drop-in, protocol-aware fault-injection middleware that interposes on the MCP tool-call path with controlled persistence (permanent, transient, intermittent) and seedable schedules. Its fault model is calibrated to the observed incident distribution (failure mode x persistence) of the public MCP ecosystem, mined from 57,240 issue and pull-request threads across 1,774 open-source MCP repositories. In a 1,350-cell experiment crossing two backbone sizes (20B, 120B), three agent stacks (custom multi-round plan-and-execute, OpenAI Agents SDK, LangChain), and three fault injection conditions, we find that faults drawn from this permanent-heavy stress profile cause severe quality degradation (task-completion score falls by 60%, from 8.08 to 3.23 out of 10), and that scale does not purchase fault tolerance for the two gpt-oss backbones we test: the 120B retains no larger share of its fault-free score under faults (resilience-ratio model, $p = 0.96$), and its completion rate under faults is, if anything, lower (91.1% vs. 95.6%; difference 4.4 points, 95% CI $[-0.2, 9.3]$). Agent stacks differ in how well they absorb faults, which suggests that framework-level error handling matters more than model scale. These results demonstrate that capability benchmarks alone can mask operational brittleness, motivating protocol-aware fault injection as a complementary evaluation dimension for tool-calling agents.

0.55

Complexity concentration as a god-class and god-method detector: a pre-registered test of a scale-free outlier rule against tool definitions and human labels

cs.SE ·2026-10-23 ·Daniel Sagenschneider
unusual
Abstract · 16 pages, 1 figure, 6 tables. Pre-registered study; ancillary files contain the run data, analysis outputs and readings. Archived at doi:10.…

God-class and god-method detectors mark concentrated complexity with a raw-scale threshold, a constant fitted on a corpus and re-fitted when the population moves. This paper pre-registers a test of a rule with no raw-scale constant: a unit is a god unit when its complexity is a robust outlier for its own codebase, a modified z-score of 3.5 or more. On 445 repositories in ten languages, 171 of them Java, and the 492 Java repositories of MLCQ, the rule is scored against PMD's GodClass, MLCQ's reviewers and two further definitions, with signal and cutoff chosen on half the repositories and read on the rest. LLM-written code may gather its complexity differently, the case for a detector that needs no re-fitting; the first question is whether one can match the scale of existing human-written code. The verdict is inconclusive on both populations, by different arms. The rule ranks PMD's god classes from the per-file maximum of cognitive complexity at per-repository AUCs of 0.916 and 0.908. On the main run it is, at its default, non-inferior within the margin to a cutoff tuned on the same population on recovery, and flags about twice as many non-god classes; on MLCQ it flags about as many and falls short of the margin on pooled recovery. Pooled, the raw value ranks god classes better than the per-repository score, and a cutoff carried between the populations lost almost nothing: both are human-written code on one scale, so the test is of matching that scale, not of surviving a shift. The rule's extra flags are small classes, just over half holding a method at PMD's own method-level threshold; its misses are large classes whose complexity is spread over many methods. Two readers of the source did not agree on what the extra flags are. Concentration finds where the complexity is; whether that place is a god unit is a question about coupling, which a registered follow-on study takes up.

0.54

Beyond Correctness: How Novice Software Engineering Students Exercise Oversight of Generative AI

cs.SE cs.HC ·2026-10-23 ·Ali Shakiba, Thomas Chaffey
Abstract · 10 pages, 3 figures, 2 tables

Generative AI (GenAI) is shifting software engineering students' work from producing artefacts towards evaluating and supervising AI-generated outputs. However, little is known about what characteristics students identify in outputs when reporting interventions, or how they exercise oversight during authentic software engineering work. We analysed 962 student-AI interaction turns documented by 145 first-year software engineering students through structured AI Journals, including 172 student-declared interventions. Using a multi-dimensional coding framework, we characterised the output characteristics identified in these reported interventions, students' reasons for intervening, the actions they took, and how those actions were enacted. Fitness problems were identified in 74.4% of interventions, compared with 16.9% involving correctness problems, indicating that oversight more often concerned whether plausible AI outputs were suitable for the engineering task than whether they contained explicit errors. Among classifiable interventions, students most frequently modified generated outputs (51.0%) or refined their prompts (28.2%), with both response types commonly associated with engineering fit. Within-student analysis further showed that outputs perceived as containing minor flaws were associated with lower evaluation confidence than outputs perceived as perfect. Taken together, these findings characterise novice AI oversight as a process that extends beyond error detection to encompass both judgement of fitness for purpose and active intervention in AI-assisted engineering work.

0.49

Speculative Forking: Turning Idle Wait into Parallel Exploration for Repository Environment Setup

cs.SE cs.CLcs.LG ·2026-10-23 ·Yuling Shi, Mingchu Zhong, Zhensu Sun, Weihan Peng, Zhi Chen +6 more
Abstract

Before developers can use a software project, they must first get it to run. This often requires installing the right language runtime, system packages, build tools, and test commands. Recent LLM-based agents automate environment setup by inspecting a repository, writing setup scripts, running them, and repairing the resulting errors. However, these agents still follow a sequential trial-and-error loop: they try one setup route, wait for slow commands to finish, observe the failure, and only then decide what to try next. Because environment commands often take long to run, this sequential workflow makes setup slow and limits exploration to one route at a time. We present SpecBuild, an LLM-based environment setup system designed to be both faster and more successful through speculative forking. SpecBuild explores candidate setup branches in parallel, returns the first committed environment, and discards the rest. It applies this design at two granularities: strategy-level forking explores alternative setup routes, such as different base images and dependency policies, while command-level forking predicts failures of running commands and prepares repair branches before those failures are observed. We evaluate SpecBuild on Multi-Docker-Eval, covering 334 tasks from 40 repositories and 9 programming languages. Across three backbone models, SpecBuild improves the success rate by 6.4 to 12.0 percentage points over the strongest baseline, while reducing average setup time by 22.4% to 41.6%. These results show that LLM-based environment setup need not trade speed for reliability: parallel speculative exploration can make agents both faster and more successful.

0.47

Green Code or Costly Code? Evaluating the Energy Costs of Iterative LLM-Based Code Optimizations

cs.SE cs.AI ·2026-10-23 ·Sadra Emami, Benjamin Weigell, Mohammad Reza Mousavi, Gunel Jahangirova, Hugo Araujo +1 more
top venuebenchmarkcode/data
Abstract · Accepted at the 49th IEEE/ACM International Conference on Software Engineering (ICSE 2027)

Large Language Models (LLMs) are increasingly used to optimize code for performance. This practice often assumes that improvements in execution speed naturally translate into overall lower energy consumption. We challenge this assumption empirically by measuring whether performance-optimized code produced by LLM self-refinement pipelines is also energy-efficient relative to canonical human reference solutions, and whether the energy cost of the optimization process can be offset by subsequent execution-time savings. We evaluate 11 open-source models across four decoding configurations on all 1,000 problems of EffiBench, a state-of-the-art performance benchmark, while collecting per-process, hardware-level energy measurements throughout. Inference energy spans 0.79 to 8.68 MJ across model-variant pairs, which is roughly equivalent to the energy of 250 to 2800 GPT-4.1 queries. Limiting the maximum number of output tokens has the largest impact, reducing energy use by up to 69% compared to the default configuration. On the execution side, 42 of 44 configurations produce code that consumes more energy per problem than the canonical human reference, with overheads ranging from +4% to +886%. The two exceptions achieve only marginal gains, with break-even points on the order of 10^6 to 10^7 executions. Mixture-of-Experts architectures consume substantially less inference energy than equivalently-sized dense models without sacrificing correctness, making architectural choice a more consequential sustainability lever than decoding parameters. Performance optimization and energy efficiency are not always aligned goals: LLM self-refinement consistently produces faster-seeking but energy-heavier code relative to what a human engineer would write. All artifacts are available at https://doi.org/10.5281/zenodo.23598170.

0.41

Navigating Epistemic Parity in LLM Agents: A Benchmark for Cross-Source Conflict Resolution Between Memory and Procedural Skills

cs.SE ·2026-10-23 ·Prakhar Pandey
publishedbenchmarkcode/data
Abstract · 17 pages, 2 figures, 6 tables. Accepted at the NeurIPS 2026 Verify-Agents Workshop. Code and data: https://github.com/prakhar1605/epistemic-…

Production LLM agents increasingly run episodic memory pipelines and procedural skill modules (e.g., SKILL.md files) in tandem. Both are injected into the same context region with no defined precedence, so they can issue contradictory instructions that no instruction hierarchy resolves, a condition we term epistemic parity. We introduce a benchmark for measuring how LLM agents adjudicate such cross-source conflicts. We (i) develop a four-way taxonomy of memory-skill contradictions (factual, format, procedural, permission); (ii) instantiate 52 scenarios under a tripartite test-case anatomy (memory injection, skill injection, ambiguous user task), scored almost entirely by deterministic checkers over tool-call payloads and output structure; and (iii) run each scenario under three context conditions to isolate positional bias, measure the efficacy of an explicit precedence directive, and classify disclosure behavior, quantifying silent adjudication: resolving a conflict without telling the user. Across 1,383 runs on three open-weight models, neither source wins uniformly: resolution depends jointly on conflict type, model, and injection order. Moving the skill block nearer the task raises its win rate in all six factual and procedural cells (sign test, p = 0.031) but reverses for format conflicts, indicating that adjudication tracks token placement rather than source semantics. An explicit "skill takes absolute precedence" directive raises skill compliance on the two stronger models yet leaves 3-10% residual non-compliance, and moves the smallest model's factual and procedural resolution in the wrong direction. Most consequentially, in 98.8% of runs where the agent acted through a tool call it emitted no accompanying text at all, and 89% of all runs disclosed nothing about the contradiction: silent adjudication is the default. Code, scenarios, and raw outputs are public.

0.40

OPTARI: An open-source software framework for clinical translation of optoacoustic imaging

cs.SE cs.CVphysics.med-ph ·2026-10-23 ·Moritz Schillinger, Maximilian Bader, Adrian Buehler, Felix Wachter, Ferdinand Knieling +2 more
code/data
Abstract · 15 pages, 4 figures, 2 tables. Submitted to the Journal of Biomedical Optics. Code and documentation: https://mo-sc.github.io/OPTARI/

Combined optoacoustic (OptA) and ultrasound imaging is a promising modality across a wide range of pathologies. Its use in clinical research, however, is limited by a lack of analysis software accessible to users without programming expertise. Existing tools are either script-based engineering frameworks or proprietary vendor applications. We present OPTARI, an open-source graphical software application to facilitate large-scale clinical optoacoustic-ultrasound (OPUS) studies. OPTARI builds on the napari image viewer and the PATATO toolkit for OptA data processing. It supports native vendor and open HDF5 data formats, aligns with the IPASC standard, and is distributed as a standalone executable. Features include region-of-interest annotation, reconstruction, chromophore unmixing, deep-learning-based tissue segmentation, and temporal and spectral analysis modules. OPTARI offers a combination of features for clinical study analysis that none of the compared tools provide. The automated tissue segmentation produced results comparable to expert labeling for muscle tissue, and batch processing reduced analysis time from minutes to under one second per scan. The tool is used in ongoing clinical OPUS studies. OPTARI connects engineering frameworks with medical research and provides an open-source basis for reproducible and standardized analysis in clinical OptA imaging studies.

0.40

WebUniverse: Scaling Verified Training Data for Agentic Web Development

cs.SE cs.CV ·2026-10-23 ·Yaorui Ding, Fengqing Jiang, Yichen Feng, Yuetai Li, Zhangchen Xu +6 more
code/data
Abstract · Project Page: https://webuniverse-nsl.github.io/

Training agents for front-end web development requires demonstrations of code editing, execution, browser inspection and revision, yet public trajectories and reusable task environments remain limited. We introduce WebUniverse, a pipeline that constructs verifiable tasks with verified development trajectories from open-source repositories and real web pages. The tasks cover text-grounded application development in standalone and iterative settings and vision-grounded reproduction across pages and viewports. We collect agent trajectories for these tasks and verify the resulting applications against task-specific functional or visual requirements. We release WebUniverse-Env, containing 15K executable task environments, and WebUniverse-Mix, containing the corresponding 15K development trajectories with verification records and scores. Fine-tuning Qwen3.5-35B-A3B on WebUniverse-Mix raises the Vision2Web average score from 51.7 to 68.3, WebGen-Bench accuracy from 34.5% to 63.9%, and Web-Bench pass@2 from 12.2% to 23.1%. The resulting model also outperforms a public-data baseline trained with the same configuration and sample budget across all reported metrics. Ablations show complementary benefits from text and visual trajectories and continued gains as training data scales from 5K to 15K trajectories.

0.35

On the Expressiveness of Asynchronous Processes with Nested Session Types (Extended Version)

cs.PL cs.DC ·2026-10-23 ·Sjors Wortelboer, Dan Frumin, Jorge A. Pérez
theory
Abstract · Extended version of a paper submitted to EXPRESS/SOS 2026

Session types are a type-based approach to correct message-passing programs, where types specify protocols for channels and type checking guarantees communication correctness. Traditional session types express regular protocols only, which leaves out many protocols found in practice. Aiming at overcoming this limitation, Nested Session Types (NSTs) can specify nested recursive protocols. NSTs are known to be characterised by one-counter machines and deterministic context-free languages. In this paper, we study the expressiveness of ASP (Asynchronous Session Processes), the process model associated with NSTs as introduced by Das et al. We prove that AST is Turing complete by giving a correct encoding of two-counter machines into typed processes. Our encoding uses a non-trivial state-aware representation of counters and instructions that is neatly governed by NSTs.

0.34

From Cosmetic to Structural: A Paired Experiment on Metric-Guided Agentic Refactoring

cs.SE ·2026-10-23 ·Nadim Hagatulah, Markus Borg, Emma Söderberg
code/data
Abstract · 41 pages, 5 figures, 14 tables. Under Review

The role of coding agents is expanding beyond writing new code toward maintaining existing systems as AI teammates. Yet recent studies show that agent refactorings favor superficial, consistency-oriented edits over structural transformations needed to resolve Technical Debt (TD). We investigate if explicit static-analysis feedback closes this gap. Through the CodeScene MCP server, we give an open-source agent (gptme) access to code smell reports and CodeHealth (CH) scores (1--10, higher is better), and compare guided against unguided refactorings in a paired experiment covering $24{,}620$ single-file competitive programming solutions in C++, Java, and Python. We find that guidance shifts the agent from performing cosmetic edits toward structural operations, raising the rate of effective refactorings by $21$--$26$ percentage points. The benefit depends on initial code quality, with language-dependent break-even points at CH $4.9$--$7.1$, below where the structural transformations pushed by guidance increasingly risk regression, indicating where the agent needs more support to refactor. Guided runs also shift the failure profile toward logical test failures, introduce new coupling smells, and consume $\sim$$3$--$5\times$ more tokens over $\sim$$2\times$ as many interaction turns. Our results indicate that AI teammates can benefit from similar static-analysis feedback that human developers already use, enabling structural transformations toward more effective TD remediation, at a higher token cost and more interaction turns.

0.33

Cross-Platform Application Migration to HarmonyOS: An Agentic Pipeline with Artifact-Derived Verification

cs.SE cs.AI ·2026-10-23 ·Ergun Biçici, Mehmet Kaygusuz, Fatih Turan Gündoğdu, Emre Filiz, Çağatay Kızıldağ +6 more
code/data
Abstract · 6 pages, 2 figures

The migration of applications written for a different platforms like Android or iOS to HarmonyOS NEXT requires a complete rewrite in ArkTS. Although LLMs automate the translation of source constructs into ArkTS, syntactic compilation is not enough and functional assurance is a critical open challenge in HarmonyOS NEXT migration. We present HMOS~Migrator, an agentic pipeline that reads an application's source code, generates an ArkTS project, and validates the result by replaying recorded interactions on a real device. We cast the scoring problem as quality estimation without a reference implementation; we score a migration using only the source application's recorded behaviour and the artifacts produced by the pipeline. We define seven weighted metrics, each constructed so that reduced measurement cannot artificially raise the score. Over twelve runs across six open-source applications, the pipeline's structural reproduction is high (completeness \pct{83.3}--\pct{100.0}), while on-device behaviour (\pct{4.3}--\pct{63.1}) and UI fidelity (\pct{21.1}--\pct{63.2}) remain the primary limits. Scores improve steadily over time: $+0.45$~points per day overall ($R^2=0.44$, $p=0.019$) and $+0.47$ over the model-free subset ($R^2=0.39$, $p=0.029$), with corresponding to fit angles of $+8.4^{\circ}$ and $+11.7^{\circ}$ respectively.

0.33

Pairing Matters: Performance, Failures, and Efficiency in Lightweight Coding Agents

cs.SE ·2026-10-23 ·Lu Zhao, Christopher M. Poskitt, David Lo
code/data
Abstract

Self-hosting coding agents is often necessary for software-engineering teams that must keep proprietary code and development traces within controlled infrastructure. Lightweight, open-weight models make such deployment more feasible by reducing hardware requirements, while strong results on repository-level benchmarks demonstrate their potential for demanding coding tasks. Deploying these models as coding agents requires pairing them with an agent harness, making their joint performance a key deployment consideration. Yet existing leaderboards primarily report fixed harness-model configurations rather than controlled comparisons across alternative pairings. To address this gap, we investigate harness-model interactions through a controlled full-factorial evaluation of six open-source harnesses and three lightweight coding models across 100 SWE-bench Verified tasks, producing 1,800 graded runs under common serving, metering, and grading conditions. For the same model, resolution rates differ by up to 44 percentage points across harnesses, while harness rankings change across models. We apply an evidence-based failure taxonomy to 1,100 unsuccessful runs and find substantial differences in failure patterns across harness-model pairings. A model's dominant failure modes under one harness need not dominate under another. Efficiency is similarly dependent on the pairing: even among model-directed loop harnesses, charged tokens per resolved task vary by more than 20-fold. These findings make harness-model selection a central deployment decision for lightweight coding agents. Strong performance with one harness or model does not guarantee similar results with another. We translate these findings into practitioner guidance for pairing selection, token-efficiency assessment, and trajectory-informed harness engineering, and release the complete experimental records and analysis artifacts.

0.31

Choosing What to Observe: Higher-Order Obstructions in Horn Abduction

cs.LO ·2026-10-23 ·Yukun Yang
Abstract · 9 pages, 1 figure

An explanation may exist in every complete context yet be impossible to select from the information observed. We study this distinction in finite propositional Horn abduction with integrity constraints. Observations are sufficient when each report admits one hypothesis set that is consistent and entails the manifestation in every compatible context. We characterize sufficiency by minimal context families with no common explanation. Their variable spans form a canonical hypergraph whose transversals are exactly the sufficient observation sets. The characterization recovers classical discernibility for unique explanations and supports composition across independent components. Minimal obstructions have unbounded cardinality even for acyclic rules, while rectangular explanation families admit a tight bound given by their largest choice domain. With circuit-defined contexts, sufficiency is $Π_3^P$-complete and selecting at most $k$ observable variables is $Σ_4^P$-complete, already for unrestricted contexts and acyclic rules. The selection reduction makes observation choices enforce globally fixed existential commitments. Explicit contexts reduce both tasks to NP-complete problems; tabulated explanation families make sufficiency polynomial. These results connect explanatory choice, information requirements, and representation cost.

0.31

Lessons Outlive Their Failures: A Case Study of Rules for AI Coding Agents

cs.SE ·2026-10-23 ·Hassan Hassanzadeh Aliabadi
benchmark
Abstract · 16 pages, 4 figures, 4 tables

AI coding agents increasingly work inside a harness: rule files such as AGENTS.md, on-demand skills, runtime hooks and automated checks. Most studies of harnesses look at a single point in time or at short benchmark runs. I report a six-week case study of my personal system of record, a plain-markdown repository that three AI agent tools edited daily under one shared rule set. I built a ledger of 118 agent failures, their fixes and whether they came back. I also tested the system's monthly LLM documentation audit against 47 known stale statements and 20 planted errors. Finally, I replayed five past failures in isolated git clones, with and without their fixes. All results come from one repository with three runs per condition, so I report them as exploratory. Stale documentation (drift) was the largest failure class (48 of 118), although 35 of those were found on one audit day. A single audit run found about half of the known drift (Claude Sonnet 5.5 mean: 50%), and at least 93% of its findings were correct. On Sonnet 5.5, the agent finished four of the five replayed tasks. In three of those four, the failure did not happen, with or without its fix. Only the stale-count failure came back without its check, and the check stopped it in every run. On older models, an agent drafted an em-dash once while the em-dash lint check was active, and the check rejected it. A written rule on run-time estimates did not change Sonnet 4.5's overestimates. These results suggest that harness lessons can outlive the failures they were written to fix, while drift persists and still needs mechanical checks. Tools that evolve a harness add lessons faster than they retire them, so they should re-test their rules from time to time and retire the obsolete ones.

0.30

EVM-Anchored Provenance for AI-Assisted Code Evolution: A Smoke-Scale Deployment Report

cs.SE ·2026-10-23 ·Einar César Santos
code/data
Abstract · 7 pages, 3 tables and 1 figure

We report a smoke-scale deployment of an EVM-anchored provenance framework for the state stream produced by AI-assisted code generation. The system comprises codebase-state-manager, an open-source Model Context Protocol (MCP) server (MIT, v0.2.1) that exposes codebase state, transition, and reward operations to any MCP-compatible assistant, and CodebaseStateRegistry, a Solidity contract deployed on Base Sepolia that emits per-state commitment events. Payloads are pinned to the InterPlanetary File System (IPFS) and indexed through a rebuildable SQLite mirror that resynchronizes by replaying contract events. We delimit precisely what anchoring does and does not attest (existence, ordering, and non-equivocation by the committer key; never truthfulness or fabrication-prevention) and outline the orthogonal supply-chain controls required to close that gap. Under an author-produced synthetic workload on a local Anvil simulator, the median commit costs 103,282 gas, modeled at US$0.0018-0.0035 in rollup fees per commit (US$17.9-34.8 per 10,000 states), and simulator latency for the create operation is 539.8 ms (p50) and 725.0 ms (p95) under loopback storage and instant mining. The live deployment holds 14 StateCommitted events verifiable on-chain; the last consumed 98,977 gas at 1 gwei (about US$0.30). All empirical figures are descriptive quantiles from a single author run; no testnet latency is reported, and no claim of superiority is made against any baseline. A receipts-trie consumption mechanism, a BountyEscrow use case, and a FAIR-aligned dataset publication pattern are specified as outlook and are not implemented. A property-by-property comparison against OpenTimestamps, Sigstore Rekor, Centrifuge AnchorRegistry, in-toto/SLSA, Radicle, and Lalli et al. places this work relative to the surrounding provenance landscape.

0.30

Runtime BiDi Structural Verification Catches Right-to-Left Layout Defects that Accessibility, Actionability, and Logical-Property Linting Miss

cs.SE cs.HC ·2026-10-23 ·Karan Yadav
benchmarktheoryunusual
Abstract · 7 pages, 2 figures, 4 tables

Bidirectional (right-to-left, RTL) layout defects arise when interfaces authored with physical directional CSS (margin-left, border-left, text-align:left) are rendered under dir="rtl" without being mirrored. Accessibility engines (axe-core), actionability engines (Playwright), and even Stylelint's current core logical-property rule fail to localise these defects. We present runtime structural invariants -- a zero-golden mode over computed style and geometry, and a bidi-conservative differential mode -- that catch them in a headless browser. On a held-out benchmark of 14 real RTL bugs reconstructed from dated fix commits in Bootstrap and Material Components Web, the detector recovers 13/14 (2 via zero-golden, 11 via the differential mode), with no false alarm on any patched version; on the actual committed historical bootstrap.rtl.css it catches 1 of 2 attempted. It reports 0 false positives on 16 negative LTR-island cases and, in both modes, on 43 unmodified real components across four design systems and three RTL architectures (rule-of-three 95% upper bound <= 7%), and it surfaced a real directional defect in the current Bulma 1.0.4. By contrast, Stylelint's core rule emits an identical 1744 warnings on Bootstrap's LTR and correct-RTL builds, pseudo-localisation detects none of the mirroring bugs, and a naive pixel mirror-and-diff reaches only F1 0.87 with a test-set-tuned threshold. The two modes are complementary: zero-golden is the sole detector of the decoration/styling-asymmetry class (5/6) but false-flags the correct logical border-inline-end, while the differential detects none of that class yet never false-alarms and is empirically exact on the block-level inset/margin class. We scope this paper to detection; repair is future work.

0.28

Locally eventually onto cellular automata

cs.FL cs.DMmath.COmath.DS+1 ·2026-10-23 ·Jan Kristian Haugland, Nadia S. Larsen, Tron Omland, Henrikke G. Rustad
Abstract · 29 pages

We give a new checkable criterion to ensure that a one-dimensional cellular automaton on a finite alphabet is locally eventually onto. In this setting, openness is equivalent to being n-to-1 for some n and therefore implies surjectivity. Bipermutive cellular automata are known to be locally eventually onto, while open cellular automata that are neither bipermutive nor bijective appear to have received less attention. Building on Sutner's work with de Bruijn graphs, we describe a method for deciding openness and at the same time determining n. We then introduce the notion of block-regularity of local rules as a generalization of bipermutivity and prove that it implies openness. We further identify two conditions that, together with block-regularity, ensure that every nonempty cylinder set is mapped onto the whole configuration space after a finite number of iterations, and hence that the cellular automaton is locally eventually onto. The appendix provides the algorithms that check these finite criteria and then presents examples of locally eventually onto, non-bipermutive cellular automata for small diameters.

0.27

OpenDraco: A Platform for Engineering and Experimenting with Configurable Multi-Agent Architectures for Software Engineering

cs.SE ·2026-10-23 ·Xiang Feng Ye Pan, Vincenzo De Martino, Matias Martinez
code/data
Abstract

Experimenting with LLM-based multi-agent systems (MAS) involves an iterative cycle of modeling configurations, executing and evaluating them, comparing results, and refining system designs. Existing approaches provide limited end-to-end support for this experimentation lifecycle. We present OpenDraco, an open-source platform that integrates these activities into a single workflow. OpenDraco represents agents, prompts, tools, models, parameters, and communication topology in exportable, versioned JSON specifications, translated into executable LangGraph workflows supporting different interaction structures. Users can modify and re-evaluate configurations using local or cloud-based LLMs. We instantiate OpenDraco for automated program repair (APR) and evaluate a sequential baseline and seven reconstructed configurations on synthetic and real-world SWE-bench Lite issues. The results reveal substantial cost differences among configurations with similar repair outcomes, while the baseline resolves 12 of 77 SWE-bench issues using open-source LLMs. Additional proof-of-concept tasks illustrate applicability beyond APR. OpenDraco supports iterative experimentation with heterogeneous MAS configurations, enabling researchers to examine their outcomes and costs within a common environment. Demonstration video: https://youtu.be/278sOR4lBEk.

0.26

On the Expressive Power of Hook-Indexed State: Comprehending State in the React UI Library (Extended Version)

cs.PL ·2026-10-23 ·Toby Ueno, Ezra Cooper, Simon Fowler, Sam Lindley
Abstract

React is currently the most popular JavaScript library for front-end web development. It endows UI components with state via the useState construct, through a mechanism that we name hook-indexed state. However, React's implementation of hook-indexed state is both fragile and poorly understood. Programmers frequently experience unexpected behavior, and the benefits of programming in this paradigm are wholly unclear. We define $λ^\mathsf{react}_\mathsf{hook}$, a core calculus that captures the essence of hook-indexed state, and give it a type system to statically rule out unexpected behavior. We then translate $λ^\mathsf{react}_\mathsf{hook}$ into a calculus $λ^\mathsf{react}_\mathsf{ref}$ based on ML-style references, and prove that this translation preserves typability and semantics, thus showing that hook-indexed state is expressible in terms of standard references. The translation is local and admits a compositional definition, but involves significant code duplication. As well as illuminating the essence of state in React, our observations can inform the design of future reactive programming systems.

0.25

CrowdStrike Channel File 291: An Engineering Analysis of a Global Endpoint Outage

cs.SE ·2026-10-23 ·Luiz Parente
Abstract · Keywords: Configuration management, dependability, endpoint security, incident analysis, software deployment

On July 19, 2024, a CrowdStrike Falcon content update caused Windows systems to crash, affecting about 8.5 million devices according to a Microsoft estimate. The immediate defect was an inconsistency between a detection template that declared 21 inputs and sensor code that supplied 20. A new matching criterion exposed an unchecked array access in the kernel-mode content interpreter. This article reconstructs the incident from vendor technical reports, Microsoft's crash analysis and recovery guidance, congressional records, and operational impact reports. It distinguishes documented findings from engineering inferences and examines how interface validation, test selection, deployment policy, and recovery dependencies interacted. Four assurance gaps were consequential: validation against a definition the consumer did not meet, tests that avoided the missing input, distribution without health-gated promotion, and recovery paths requiring connectivity that some crashing hosts could not maintain. Recommended controls separate defect prevention, exposure containment, and service restoration, with verification criteria, release-cadence constraints, and operational tradeoffs. The case shows why configuration consumed by privileged software requires lifecycle assurance comparable to code, and why release reversion, endpoint recovery, and service restoration must be assessed separately.

0.25

Large-Scale Study of GitHub Copilot Trajectories

cs.SE ·2026-10-23 ·Izah Sohail, Yerdanat Kurmantayev, May Mahmoud, Sarah Nadi
Abstract

AI coding agents are increasingly integrated into software development workflows. However, most coding agent studies rely on benchmarks similar to SWE-bench, which assume autonomous agents operating in a single trajectory. This assumption does not mirror real-world usage, where developers interactively collaborate with agents. We address this gap by studying human-in-the-loop agent trajectories in the wild, focusing on 7,581 Copilot-authored PRs with publicly accessible session data. We find that 58% of Copilot PRs have no human intervention, and that when developers do intervene, collaboration is brief (median of 2 Human-AI loops). Merged PRs also tend to have fewer initial trajectory iterations than unmerged ones. We also find that the trajectory following the first human intervention exhibits a significant drop of 20.8 iterations and 8.4 min in duration. However, trajectories following later interventions show minimal iteration drops and no significant duration change, signifying that most of the work happens in the initial trajectory. Our qualitative analysis of illustrative PRs explains this pattern: many developer interventions are small, focused requests that Copilot completes in shorter later trajectories. We also illustrate that developers collaborate with Copilot in multiple ways, including delegated work through feedback as well as adding their own commits. However, we find that developers may need to guide Copilot through multiple (sometimes repeated) interventions in the PR, or even across multiple editions of the PR.

0.24

Non-Expansive Fuzzy Modal Logic Has Uniform Interpolation

cs.LO ·2026-10-23 ·Stefan Gebhart, Lutz Schröder, Paul Wild
Abstract

Fuzzy logics with truth values in the unit interval vary in their choice of propositional connectives and their semantics, ranging from just minimum and maximum as in Zadeh logics to truncated addition and subtraction as in Łukasiewicz logics or their extensions with truth constants, known as Pavelka logics. In the more specific world of fuzzy modal logics, non-expansive fuzzy modal logic has recently been proposed as a sweet spot between Łukasiewicz (more precisely, Pavelka) and Zadeh fuzzy modal logic; it restricts Łukasiewicz propositional connectives to have one argument constant, equivalently extends the Zadeh propositional base with constant shift operators. It retains reasonable expressiveness both in practical terms and abstractly; in particular, it characterizes the standard notion of behavioural distance on fuzzy Kripke frames. At the same time, it allows for satisfiability checking in the same complexity as for its two-valued correspondent (the modal logic K, or the description logic ALC), in contrast to the notorious computational hardness of Łukasiewicz modal logic. In the present work, we show that non-expansive fuzzy modal logic admits uniform interpolation, a key property in modularization, information hiding, summarization, and other applications that fails for both Łukasiewicz (or Pavelka) and Zadeh modal logic.

0.24

DeepAVR: Deep Analysis via Multi-Agent Collaboration for Automated Vulnerability Repair

cs.SE cs.CR ·2026-10-23 ·Woorim Han, Hyunjun Kim, Brian Junekyu Lee, Kyeongmin Kim, Younghan Lee +2 more
Abstract

Automated Vulnerability Repair (AVR) has emerged as a critical defense against software security flaws, aiming to triage and fix bugs autonomously. Early LLM-based approaches focused on patch generation under the assumption of perfect fault localization, while recent end-to-end methods attempt to repair vulnerabilities directly from sanitizer logs. However, these methods often fail when the root cause is deeply hidden--either physically distant from the crash site or logically masked by interdependent errors--making the initial report insufficient. We propose DeepAVR, a multi-agent framework for end-to-end AVR that performs deep analysis by transforming intermediate repair outcomes into structured diagnostic feedback. DeepAVR decomposes repair into specialized roles--Fault Localization, Patch Generation, Compile Validation, Security Validation, Functionality Validation, and Patch Review--and routes feedback across these roles to iteratively refine both localization and patch hypotheses. In particular, DeepAVR introduces patch-conditioned re-localization, which treats execution results from intermediate patches as diagnostic signals to uncover hidden root causes, and a root-cause-oriented patch review stage that detects symptom-level fixes and incomplete repairs beyond execution-based validation. Evaluation on the real-world C/C++ vulnerability repair dataset VulnLoc shows that DeepAVR significantly outperforms prior methods, generating correct patches for 92.3% of vulnerabilities. Furthermore, results on SAN2VULN demonstrate strong generalization, with DeepAVR effectively repairing vulnerabilities unseen during training.

0.22

Operationalizing Governed Agentic Workflows in Regulated Finance: A Customer Zero Case Study using Zafin AIOS

cs.SE cs.AI ·2026-10-23 ·Shahir Daya, Rob Dunlap, Alexandre Boulenger, Ahmed Abdelmoneim, Salvatore Vella
Abstract · 19 pages

Regulated industries such as finance and healthcare face critical governance challenges that hamper the adoption of AI-assisted software development and hinder the transition from experimental pilots to production work. This paper presents Zafin AIOS, an end-to-end orchestration platform and control plane designed to deploy AI within regulated industries such as finance. We provide the architectural principles and an empirical comparison between AI-assisted and human-led workloads. To validate this framework empirically, we conduct a case study analysis of Zafin's internal enterprise workflows, which we refer to as Customer Zero. This paper contributes a production case study of AI-assisted software engineering in regulated industries. We evaluate the results of using Zafin AIOS with two different production codebases, one for the pricing execution platform and the other for the client deal negotiation platform. Our contribution is a customer case study for regulated industries with production codebases and governance, which includes proof of work. The results demonstrate that strictly governed agentic orchestration increases enterprise capacity without compromising control. In the pricing execution platform workflows, the platform provided complete observability and reduced effort by 38\%. In the client deal negotiation platform, it reduced reporting work by 76\% while maintaining all predefined data privacy, role-based access, and audit guardrails. This paper bridges the gap between AI-assisted software engineering theory and the operational reality of compliance. Closing this gap is important to help enterprises take advantage of AI technologies. These results are descriptive of the benefits of using governed workflows in regulated industries.

0.22

WaitLens: Temporal VLM-Based Adaptive Waiting for Industrial GUI Testing Agents

cs.SE ·2026-10-23 ·Shika Ikeura, Xuan Wang, Ziyi Kuang, Jianping Xu, Chao Kang +4 more
Abstract · 11 pages, 7 figures, and 2 tables

Improving the efficiency of GUI testing agents is important for industrial deployment. Beyond LLM inference, fixed waiting between consecutive interactions introduces substantial overhead. Our industrial study reveals that fixed waiting accounts for 42.3% of the time within the sampled interaction intervals, comparable to the 54.8% spent on model calls. Moreover, 62.9% of the waiting time is spent after the GUI has already finished loading. Although prior studies have proposed adaptive waiting strategies, their reliance on app-specific data or platform-specific information limits their applicability across diverse industrial environments. To this end, we propose WaitLens, which fine-tunes a lightweight VLM on GUI loading trajectories annotated with waiting end times. We introduce temporal resampling to construct diverse training examples, enabling the model to learn varying GUI loading behaviors from temporal observations. Experiments show that WaitLens achieves 86.08% accuracy in deciding whether to keep waiting, compared with 54.13% for the backbone VLM. Our evaluation using recordings from Meituan's industrial testing pipeline shows that WaitLens reduces mean extra waiting after the interface is ready by 94.98% compared with a 20-second fixed wait.

0.20

CC4Q Workbench: From Quantum Repositories to Analyzable Code Comments

cs.SE ·2026-10-23 ·Zenghui Zhou, Yuechen Li, Zheng Zheng
code/data
Abstract

Quantum software development kits (SDKs) expose high-level APIs whose use requires an understanding of underlying mathematical principles, circuit semantics, and hardware properties. Code comments can provide natural-language or structural explanations that are not explicit in API calls. However, studying comments in a new quantum SDK still requires researchers to coordinate separate steps for comment extraction, quantum-specific classification, and manual verification, resulting in repeated integration effort. To address this problem, we present CC4Q Workbench, an open-source workbench that integrates these steps into a reusable workflow. Given a Python repository at a specified revision, CC4Q Workbench automatically extracts code-comment pairs, segments comments into sentence-level units, and applies two trained classifiers to identify quantum-specific content and assign quantum-knowledge categories. Its web interface supports exploration of bundled datasets, inspection of code snippets, corpus construction from new repositories, manual annotation with model predictions hidden during binary labeling, and classifier fine-tuning using exported annotations. The release provides aggregate statistics for six SDK datasets comprising 39,593 sentence-level code comment units (SCCUs), of which 17,517 are browsable in the web interface. Applied to three additional application-layer Qiskit repositories outside the original CC4Q dataset, the pipeline ran without code changes and produced 11,517 SCCUs. Manual verification of 379 SCCUs showed 85.0% agreement between human annotations and the binary classifier's predictions.

0.20

Are LLM-Generated Formal Specifications Faithful to Their Requirements? A Study on Bimodal Formal Specification Generation

cs.SE cs.PL ·2026-10-23 ·Wenhan Wang, Shiming Chen, Zeyu Sun, Shenghua Feng
benchmark
Abstract

Formal specification generation is a critical step towards formal verification and correctness assurance of software. Recently, large language models have been widely adopted in automatic generation of formal specifications. However, existing LLM-based specification generation approaches often assume that the input program is correct, which deviates from the real purpose of formal verification---to figure out whether the program meets its requirements when its correctness is unknown. Therefore, we conduct an empirical study of LLM-based specification generation under the bimodal setting: given both the natural language requirement and its corresponding code implementation, whose correctness is unknown. We construct a benchmark of 1,403 C programs and evaluate ACSL specification generation with Frama-C verification. Our study reaches several interesting findings: first, when the code implementation is correct, providing the natural language requirement does not further improve the pass rate of generated specifications. Second, when the code implementation is incorrect, LLM-based specification generation may erroneously pass verification by producing weak or implementation-misled specifications, and this cannot be fully alleviated by advanced prompting approaches. Compared with the non-formal HoarePrompt approach, LLM-generated formal specifications are less effective in predicting requirement-code consistency. Our study takes a first look at applying LLM-based formal specification generation to verifying program correctness against requirements, and points to requirement-faithful specification generation as a new direction for future LLM-based formal specification research.

0.20

StateMirror: Pre-Execution Detection of Hallucinated Tool Calls in Coding Agents

cs.SE ·2026-10-23 ·Shiwen Shan, Yintong Huo, Zhiying Wu, Yuxin Su
Abstract

Large language model (LLM)-based coding agents may propose invalid actions that are not applicable in the execution environment, such as accessing non-existent paths or missing packages. We term this mismatch hallucinated tool use. A preliminary study of 4,398 agent trajectories reveals that these hallucinations are pervasive and costly, requiring a median of 3 additional interaction steps and 22.4K tokens to recover. Existing frameworks rely on reactive, post-execution feedback, wasting runtime resources. Nevertheless, since an agent's execution environment is grounded in concrete codebase and harness components, many action prerequisites can be extracted and validated prior to execution. To this end, we introduce StateMirror, a lightweight pre-execution detector. StateMirror intercepts actions at the harness boundary, extracts state predicates via AST and command analysis, and deterministically validates them against the live repository. By pinpointing exact environment mismatches before execution, StateMirror preempts invalid actions and subsequent recovery cycles. Evaluation on 2,508 actions shows StateMirror achieves 0.940 precision and 0.817 F1 in detecting such hallucinations, surpassing the strongest LLM judge by 35.3% in F1 while reducing median detection latency from 17.73 s to 1.50 ms with zero token consumption. For practical mitigation, StateMirror's pre-execution feedback enables agents to rectify over 95% of hallucinated tool calls. During live mitigation on 79 tasks, StateMirror reduces agent execution time by 15.8% and token consumption by 11.8% while maintaining task resolution.

0.19

From Faulty Claims to Incorrect Patches: An Empirical Study of Task-Grounded Reasoning in Coding Agents

cs.SE ·2026-10-23 ·You Wang, Jinjun Huang, Zhonghao Jiang, Jialun Cao, Jiachi Chen +1 more
Abstract

Large language model (LLM)-based coding agents increasingly resolve real-world software engineering tasks, yet still fail on many of them. Prior trajectory studies characterize these failures through task-agnostic patterns of agent behavior, such as repeated actions. However, an agent may follow a sound workflow yet act on a faulty understanding of the existing code, the task requirements, or a candidate fix. Reasoning models, which power most state-of-the-art coding agents, reveal this understanding in their native chain-of-thought (CoT), whose claims about the task can be checked against the facts. Yet it remains unclear how many such checkable claims the CoT contains, how many are faulty, and whether faulty claims actually cause incorrect patches. We present the first empirical study of such claims, which we term task-grounded reasoning, analyzing 400 trajectories of four open-weight reasoning models on SWE-bench Pro. We extract atomic claims from the CoT with an LLM, assess them against evidence gathered by an LLM agent, and manually trace the defects of incorrect patches back to faulty claims, confirming each link with a claim-corrected replay. Task-grounded reasoning covers only 18.5% of the CoT tokens, yet yields 191 atomic claims per trajectory, at least 9.6% of which are faulty. Faulty claims also occur in 96.7% of the resolved trajectories, so their mere presence is a weak signal of failure. Yet two thirds of the 188 incorrect patches trace back to faulty claims. Correcting such a claim makes the agent fix the linked defect in 89.0% of the cases, and correcting all such claims resolves 11.7% of the failed tasks. Our findings show that the CoT carries task-specific signals for diagnosing and repairing agent failures, but exploiting them requires pinpointing the few faulty claims that actually cause patch defects.

0.15

Responsibility Boundaries for Agentic Systems: A Comparative Evaluation of Probabilistic Agentic Programming

cs.SE cs.AIcs.PL ·2026-10-23 ·Khadim Lo, Soukaina Alaoui Ismaili, Jean-Sébastien Dessureault
Abstract · 43 pages, 8 figures, 9 tables

Language-model agents combine reasoning, tools, memory, delegation, and external effects. Authority is often distributed across prompts, host code, configuration, and logs. Probabilistic Agentic Programming (PAP) makes responsibility boundaries explicit programming abstractions connecting mandate, authority, effects, and evidence. We compare PAP with programming paradigms, agent-oriented and BDI systems, workflow and policy languages, and LLM tool-use frameworks along seven dimensions. A mediation kernel formalizes checks before tool execution, operation-bound approval, resource limits, and evidence. Inductive proofs establish conditional safety for arbitrary model responses, assuming correct guards and complete mediation. A backward simulation matches graph executions to source executions in a defined sequential Touba subset. Its human checkpoints release program continuations; operation-bound approval remains a separate kernel contract. Finite-state exploration checks the model. Scripted enforcement tests exercise selected refusal paths and permitted controls. An exploratory study combines seven business cases, including two three-agent workflows, seven models from three providers, five methods, and five repetitions per case and method. Shared English instructions and exact checks of stored outcomes yield 1,225 assessed observations. Native Touba completes all 35 observations on six models and 30 on the remaining model; structured Touba ranges from 21 to 35. Native matches the highest observed count for each model, including ties with the frameworks. All paired case intervals include zero. These observed counts describe the evaluated tasks; the runtime authority boundary remains fixed. The results connect responsibility contracts, execution rules, and evidence.

0.15

Can AI Agents Help Robotic Engineers? Lessons from Two Focus Days in Industry

cs.SE cs.AI ·2026-10-23 ·Salim Fares, Steffen Herbold
Abstract

Software developers increasingly use AI agents that execute multi-step programming tasks on their own. However, it is unclear if and how this transfers to robotic engineering, where engineers program industrial robots in niche, vendor-specific languages and rely on visual feedback from simulation tools. Within this paper, we report a case study on the use of AI agents by an SME that programs industrial robots. Our results show that the agent did not make the engineers more productive, with the exception of a task that required the generation of highly repetitive code. However, the results with the agent improved over time, as the engineers learned how to use it and the agent was adapted to the domain. Moreover, the experience with AI mattered more for the success with the agent than the experience with robotic engineering. This shows that AI agents are not a plug-and-play solution. Instead, companies must train their engineers on the usage of agents, plan time for a learning curve, and ensure that agent configurations match the domain and developer requirements. The strong learning curve we observed further implies that studies on productivity gains with AI should be longitudinal, as they otherwise miss how engineers adopt agents in practice.

0.14

Beyond Coverage and Mutation: A Blind-Spot Analysis of LLM-Generated Unit Tests

cs.SE ·2026-10-23 ·Xiaoyu Shi, Wenqiang Luo, Boyang Yang, Haoye Tian
benchmark
Abstract

Existing evaluation protocols for unit tests rely on aggregate metrics such as branch coverage and mutation score. These metrics quantify the extent to which a test suite exercises program behavior, but do not identify the behavioral obligations it systematically leaves unchecked. We propose BlindSpotEval, an obligation-based evaluation framework aligned with the Arrange-Act-Assert pattern. The framework comprises three metrics, Boundary-Value Adequacy, Control-Flow Adequacy, and Oracle Strength, complemented by an assessment of execution reliability. We also introduce ObligBench, a complexity-stratified Java benchmark comprising 895 complex and 700 low-complexity functions with source-side obligation annotations. Evaluating 14,320 tests from four LLMs under four prompting strategies reveals systematic blind spots. Incorrect assertions account for 15.6% of test failures, illustrating that coverage alone does not establish oracle correctness. Among executable tests, coverage of null, negative, and empty arguments reaches only 34.8%, 23.8%, and 38.8%, compared with average coverage above 93% for booleans. Catch blocks and default groups have Control-Flow Adequacy scores of only 0.100 and 0.312. Across both dimensions, LLM-generated tests exhibit the happy-path bias observed in human-written tests, concentrating on normal inputs and expected flows while leaving unusual values and exceptional paths unchecked. Oracle strength is high on average but uneven, with weaker assertions for object return property values and exceptions. These blind spots also occur in low-complexity functions, suggesting that code complexity alone does not explain them. BlindSpotEval provides construct-level adequacy profiles that identify under-tested obligation types and can guide targeted improvements beyond those supported by aggregate scores alone.

0.12

Theory & Algorithms 44

StarRTT: An End-to-End Real-Time Framework for Blind Zone-Free RTT Prediction in Starlink LEO Satellite Networks

cs.IT cs.AI ·2026-10-23 ·Jaeyong Lee, Seongjin Jeong, Yusang Park, Hyunho Hwang, Sangmin Han +1 more
top venue
Abstract · 15 pages, 5 figures

Low Earth Orbit (LEO) satellite mega-constellations such as Starlink require frequent handovers due to high orbital velocities, resulting in substantial Round-Trip Time (RTT) fluctuations. Accurate RTT prediction can mitigate these fluctuations, yet it fundamentally relies on continuously updated serving- satellite information. Existing serving-satellite identification methods typically complete identification only at the end of each scheduling cycle, creating an identification blind zone that limits real-time prediction. To overcome this limitation, we develop StarRTT, an end-to-end real-time framework that eliminates the blind zone by integrating per-second serving-satellite identification with RTT prediction. Within this framework, we propose Real-Time Long Short-Term Memory (RT-LSTM), an LSTM-based RTT prediction model designed to capture both non-stationary RTT evolution and time-varying geometric relationships among the serving satellite, user terminal, and gateways. Specifically, a historical LSTM encodes historical RTT observations and real-time serving-satellite and gateway-relative features, while a future LSTM processes Two-Line Element (TLE)-propagated future features of the identified serving satellite up to the next scheduling instant. Under strict real-time back-testing, RT-LSTM achieves a Mean Absolute Error (MAE) of 1.405 ms averaged over all 15 prediction horizons and 15 reference time-slot indices, reducing the error by 4.3 %, 11.6 %, and 22.1 % compared with the existing baseline models, LSTM, StarNet, and Attention-based Spatiotemporal Model (ASTM), respectively. Notably, the reduction over LSTM reaches 12.5 % at the scheduling instant, highlighting the benefit of real-time serving-satellite information when recent RTT observations are limited.

0.52

Compressed Boolean Matrix Multiplication using Bicliques

cs.DS ·2026-10-23 ·Nicolás Araya, José Zagal, Cecilia Hernández, Gonzalo Navarro
Abstract

Boolean matrix multiplication is a fundamental operation in many graph-based applications, including web and social network analysis, information retrieval, and graph query evaluation. However, conventional sparse representations such as CSR}/CSC still incur significant space and time costs on large graphs. In this work, we introduce a method for multiplying Boolean matrices based on a biclique-based compressed representation that decomposes them into dense bipartite subgraphs, thereby capturing large sets of implicit edges. We first improve the iterative biclique extraction algorithm of Hernández and Navarro by introducing percentile-based parameter adaptation that dynamically adjusts its parameters according to the degree distribution, enhancing both extraction efficiency and compression ratio. We then extend the Schoor multiplication algorithm to operate directly on biclique-encoded matrices, thereby enabling a compositional form of multiplication where the output can also be represented as a collection of bicliques. Our experimental results on large real-world web and social graphs show that our approach achieves substantial reductions in space and computation time with respect to CSR/CSC-based representations. For example, using biclique extraction just as a tool to speed up multiplication (but retaining the CSR/CSC format) yields up to 34-fold speedups. Further, using the biclique-based representation for input and output yields up to a 10-fold reduction in space and up to a 290-fold speedup for matrix multiplication. These results demonstrate that our representation based on bicliques is both space- and time-effective to support an algebra of Boolean matrices.

0.52

A probabilistic construction of maximum $h$-scattered subspaces

cs.IT math.CO ·2026-10-23 ·Alessandro Giannoni, Alessandro Neri
theory
Abstract

Let $1\le h\le n-2$ and $g=\gcd(n,h+1)$. We prove that, for every prime power $q$, a maximum $h$-scattered $\mathbb{F}_q$-subspace of $V(r,q^n)$ exists whenever $(h+1)\mid rn$ and $$r\ge\frac{h+1}{g}\bigl(h(n-h-2)+1\bigr).$$ In particular, for fixed $n$ and minimum distance $δ\in\{2,\ldots,n-1\}$, $\mathbb{F}_{q^n}$-linear MRD codes of length $N$ exist for every $q$ and every sufficiently large $N$ such that $n$ divides $(n-δ+1)N$, with a bound on $N$ depending only on $n$ and $δ$. We take the $\mathbb{F}_{q^s}$-span of the columns of a random matrix $A$ over the algebra $R=\mathbb{F}_{q^n}\otimes_{\mathbb{F}_q}\mathbb{F}_{q^s}$, for a suitable $s$. Such a subspace is maximum $h$-scattered if and only if the kernel of $A$ contains no nonzero vector whose expansion with respect to an $\mathbb{F}_q$-basis of $\mathbb{F}_{q^s}$ has rank weight at most $h+1$. When $\gcd(n,s)=1$, $R$ is a field, and a first-moment argument, counting these vectors up to multiplication by $\mathbb{F}_{q^n}^*\mathbb{F}_{q^s}^*$, shows that with positive probability the kernel contains none of them if $s>h(n-h-2)$. When $d=\gcd(n,s)>1$, $R$ is a product of $d$ fields, and vectors vanishing in some factors lie in the kernel with larger probability. We control them by proving that, for an $(h+1)$-dimensional $\mathbb{F}_q$-subspace $D$ of $\mathbb{F}_{q^n}$, the elements of $D\otimes_{\mathbb{F}_q}\mathbb{F}_{q^s}$ vanishing outside $k$ given factors, $0<k<d$, form an $\mathbb{F}_{q^s}$-space of dimension at most $(h+1)k/d$, with equality only if $D$ is linear over a subfield of $\mathbb{F}_{q^n}$ larger than $\mathbb{F}_q$. Taking $s=gr/(h+1)$ gives the bound on $r$. Finally, if $(n-h-1)\mid(h+1)$, then for every $r$ satisfying the above inequality there is an $h$-scattered $\mathbb{F}_q$-subspace of $V(r,q^n)$ of dimension $\lfloor rn/(h+1)\rfloor$, without a divisibility condition on $rn$.

0.51

Combinatorial Minimum Cost Flow in Almost-Linear Time on Dense Graphs

cs.DS ·2026-10-23 ·Bernhard Haeupler, Yonggang Jiang, Thatchaphol Saranurak
Abstract · 101 pages

We give a combinatorial algorithm that computes an exact minimum-cost flow in $n^{2+o(1)}$ time. This is the first combinatorial improvement since Goldberg-Tarjan [GT87], and is almost optimal in dense graphs. All previous faster algorithms, including the almost-linear-time algorithm of Chen et al. [CKL+22], rely on interior-point methods. Our main conceptual contribution is a black-box reduction from exact minimum-cost flow on general graphs to its approximation on DAGs, with only an $n^{o(1)}$ loss on the running time. This reduction allows future work on combinatorial minimum-cost flow to focus on DAGs, typically much easier to handle than general graphs. Our $n^{2+o(1)}$-time algorithm follows by combining this reduction with a simple augmenting-path-based approximation algorithm on DAGs. The reduction is based on a new structural result, called a flow-cost-preserving DAG projection. Given an arbitrary directed graph, this construction produces a DAG that approximately preserves both the congestion of flows and their cost up to a small loss.

0.49

A counterexample to Tokareva's conjecture on sums of two bent functions

cs.IT math.CO ·2026-10-23 ·Yansong Feng, Hao Jiang, Guancheng Li, Tianyi Li, Lili Tang
Abstract · The counterexamples and proofs in this paper were obtained through large-scale automated reasoning with LLM agents while solving Problem~10 …

It is well known that the sum of two bent functions on $\mathbb{F}^n$ has algebraic degree at most $n/2$ for even $n\ge 4$. Tokareva conjectured that the converse holds: every Boolean function with degree at most $n/2$ can be written as a sum of two bent functions. The conjecture has been verified for even $n\leq 6$ and for several special classes of Boolean functions. We disprove this conjecture by showing that $x_1x_2(x_3x_4+x_5x_6+x_7x_8)$ is not a sum of two bent functions on $\mathbb{F}^8$. We extend this construction to obtain counterexamples in every even dimension $n\ge 8$. Using Shaporenko's reduction, we also obtain explicit counterexamples to Tokareva's derivative conjecture in every even dimension $n\ge 10$.

0.47

AFRA: Age-Threshold Frame-Based Random Access for Timely Communication in Multi-Agent Systems

cs.IT ·2026-10-23 ·Jiedan Tan, Fang Liu
publishedtheory
Abstract · Accepted for presentation at IEEE ROBIO 2026

Timely state updates are essential for coordination and monitoring in multi-agent systems, such as teams of mobile robots exchanging state information. When agents share a wireless channel, collisions and slow convergence of distributed access schemes can degrade information freshness over finite operating periods. This paper proposes age-threshold frame-based random access (AFRA), a distributed medium access control protocol for minimizing age of information (AoI). Once its age reaches a threshold, an agent starts a local frame with a mandatory transmission and, if unsuccessful, makes one further attempt at a uniformly selected remaining slot. Each attempt carries a freshly generated update. AFRA requires only local state and acknowledgment feedback, with shared slot boundaries but independently timed local frames. We prove almost-sure convergence from any legal initial state to collision-free periodic operation when the frame length $L$ and age threshold $T$ satisfy $L\ge T>N\ge2$, where $N$ is the number of agents. We also derive closed-form expressions for steady-state throughput, average AoI, and average peak AoI. Simulations demonstrate that AFRA combines rapid convergence at small age thresholds with lower window-averaged AoI than the existing age-aware random access schemes evaluated in this study.

0.42

Tightening the Online Bin Packing Gap from Both Sides, and the Limits of the Current Methods

cs.DS ·2026-10-23 ·Karyamsetty Martin
theorycode/data
Abstract · 15 pages, 3 tables. Code, exact certificates and a one-command reproduction harness: https://github.com/ksnmartin/online-bin-packing-bounds …

The asymptotic competitive ratio of online bin packing has been bracketed since 2018 by the lower bound 1.54278 of Balogh, Bekesi, Dosa, Epstein and Levin and the upper bound 1.57829 of their Advanced Harmonic algorithm. We revisit both sides with exact rational computation and prove what the computations say. On the upper bound, we rebuild the certification of Advanced Harmonic as an exact optimization over its dual parameters; with the published parameters this already certifies a ratio strictly below the published one, and re-tuning the algorithm's proportion parameters gives a certified ratio of 1.5782147437 (an exact rational), independently re-verified on all 402 scenarios. To our knowledge this is the first movement of the upper bound since 2018. On the lower bound, we prove that the 2018 adversary construction is rigid: extending it by any finite number of new batches of ordinary size (outside the window of sizes s with 2s > 1 and rational part at most 0.64819), placed anywhere, with counts depending on the observed statistics, adaptive or not, never improves the bound at any truncation level. The load-bearing fact is that the construction is volume-tight at every stop. A single window-sized batch added to the truncation-3 tree is shown by machine-verified certificates to leave its value below the published limit. A computational map covers the remaining natural extensions: unions of structural variants, two-statistic multi-batch adversaries, and three families of adaptive conditions that the paper's separation theorem allows but never uses; none improves the bound. On the algorithm side we show that the marking, bonus-item and red-sand mechanisms of Extreme Harmonic bind nowhere in Advanced Harmonic, whose bottleneck is a two-medium-item bin that no known mechanism targets. We record two corrections to commonly cited numbers.

0.40

Beyond 3SUM-Hardness for Geometric Problems

cs.DS ·2026-10-23 ·Nick Fischer
Abstract

Since the early days of fine-grained complexity, many geometric problems have been shown to require quadratic time under the 3SUM Hypothesis. But as the 3SUM Hypothesis was recently refuted, these lower bounds have lost their foundation. In this note we point out that many of them can be recovered under an alternative hardness assumption, the $k$-AP Hypothesis. The required changes to the proofs are trivial, so the lower bounds remain just as easy to teach.

0.40

Optimal Worst-Case Open Addressing: Braided Cuckoo Hashing

cs.DS ·2026-10-23 ·Andrew Krapivin, William Kuszmaul
top venue
Abstract · 60 pages, 1 figure, to appear in SODA 2027

In this paper, we introduce braided cuckoo hashing, an open-addressed hash table that supports load factor up to $1 - \varepsilon$ (for some parameter $\varepsilon \in (0, 1)$), while supporting queries, insertions, and deletions in worst-case time $O(\log \varepsilon^{-1})$, with high probability. This result represents an exponential improvement on insertion/deletion time over previous constructions, and disproves a conjecture of Bender, Kuszmaul, and Zhou that the product of worst-case query time and expected insertion/deletion time must be $Ω(\varepsilon^{-1})$.

0.35

The Thermodynamic Complexity of Circuits

cs.CC cond-mat.stat-mech ·2026-10-23 ·Anne Condon, David Doty, Niels Kornerup, David Soloveichik, David Wolpert
Abstract · 68 pages, 5 figures

When any physical system executes a logically irreversible (non-injective) step of a computation, the excess heat released over the Landauer limit is an energetic cost that cannot be recovered. In modern computers this cost vastly dwarfs the Landauer cost. Modern stochastic thermodynamics refers to this cost as the ``entropy production'' and has derived fundamental lower bounds on its value. Here we introduce a thermodynamic model of circuits using heat functions and investigate its entropy production, deriving new results on the thermodynamic complexity of circuits. We first show that computing the optimal heat functions for gates in a given circuit (minimizing the entropy production) is PP-complete and multiplicative approximation of those heat functions is NP-hard, but additive approximation admits a polynomial-time randomized algorithm. We give tight bounds for entropy production for certain specific families of circuits (on uniformly distributed inputs): an $n$-ary AND tree of depth $O(\log n)$ requires $Ω(n)$ entropy production, while a serial chain achieves $O(1)$ in an idealized model, showing an energy-depth tradeoff. The Batcher odd-even mergesort network on $n$ bits has $Θ(n)$ entropy production, asymptotically less than its $Θ(n \log^2 n)$ gate count. On random permutations of distinct integers its entropy production is $Θ(n \log^2 n)$, asymptotically exceeding the Landauer cost of sorting; but preceding it with Leighton and Plaxton's almost-sorting network reduces the total heat released to $O(n \log n)$. Finally, we show how to sort $n$ values with zero circuit entropy production over any exchangeable input distribution. Our framework also yields streamlined proofs of known results in stochastic thermodynamics on mismatch cost and modularity cost.

0.34

Identifying codes of Archimedean grids

cs.DM math.CO ·2026-10-23 ·Rudini Sampaio
theory
Abstract

An identifying code of a graph is a dominating set $C$ such that the sets $N[v]\cap C$ are pairwise distinct. The minimum density of an identifying code has been studied extensively for several infinite grids. It is known to be $1/4$ for the triangular grid and $7/20$ for the square grid, whereas for the hexagonal grid the problem is still open, between $5/12$ and $53/126$. These three grids are the regular members of the family of the eleven Archimedean grids. In this paper, we start a systematic study of identifying codes in this family. We determine the minimum density for five of the nine Archimedean grids for which it was not known: it is $2/3$ for the truncated hexagonal grid, $3/7$ for the truncated square grid, and $1/3$ for the trihexagonal (kagome) grid, the elongated triangular grid and the snub hexagonal grid. All lower bounds are proved by the discharging method. For the three remaining non-regular Archimedean grids, we also obtain some bounds: the minimum density lies between $3/7$ and $13/30$ for the truncated trihexagonal grid, between $1/3$ and $11/30$ for the rhombitrihexagonal grid, and between $2/7$ and $9/28$ for the snub square grid. As a by-product, we prove that every identifying code of a cubic graph with the slow growth property in which every vertex lies on exactly one triangle (resp. $4$-cycle) has density at least $2/3$ (resp. $3/7$).

0.34

Subgraph Isomorphism, Integer Linear Programming, and Lattices: A Linear Algebraic Framework

cs.DM cs.DSmath.OC ·2026-10-23 ·Bryant Gipson
Abstract · 24 pages, 8 figures

The subgraph isomorphism problem (SGI) asks a deceptively simple question: is a graph B found inside a parent graph A? Famously NP-complete and the subject of volumes of study, attempts to practically solve subgraph problems typically require specialized software or the focused application of problem-specific theoretical knowledge. Here we present a straightforward linear framework that neatly joins SGI to lattice theory and constrained satisfiability and, importantly, yields an easily implemented algorithm that naturally leverages consumer-grade hardware and off-the-shelf libraries to compete with state-of-the-art specialized solvers. We additionally show practical realizations in approachable settings, along with some benchmarks that reduce 6+ hour computations to seconds. Aimed at readability, we develop a series of computable algebraic invariants and demonstrate their application to familiar problems such as the Knight's Tour, with a focus on providing an accessible entry point for researchers in graph theory, abstract algebra, and students in general.

0.34

The Exact Robustness of Smith's Rule with Predicted Processing Times

cs.DS ·2026-10-23 ·Ohad Ayalon, Ilan Reuven Cohen
Abstract · 39 pages, 3 figures. Includes full proofs and exact-arithmetic ancillary scripts

Smith's rule orders jobs by their weight-to-processing-time ratios. We study the robustness of its list schedule on identical parallel machines when job $j$ is ordered using a nominal processing time $μ_j$ and its true processing time satisfies $μ_j\leq p_j\leqαμ_j$, where $α\geq1$. We determine the exact worst-case approximation ratio against the clairvoyant optimum, taking the supremum over machine counts. The ratio is an explicit continuous piecewise-algebraic function of $α$. It equals the Kawaguchi--Kyan constant $(1+\sqrt{2})/2$ at $α=1$ and is exactly $\sqrtα$ for $α\geq4$. On one machine the exact ratio is $\sqrtα$. For every fixed machine count, no deterministic nonpreemptive algorithm can improve on this guarantee, even with knowledge of $α$. The proof reduces to a common nominal Smith ratio and endpoint errors. Cutting the list schedule at its minimum final machine load yields a prefix--suffix certificate; two packings of matching lower-bound instances explain the branches below $α=4$. We also prove a nominal-to-robust transfer: any fixed schedule ordered by nominal Smith ratios on each machine, with nominal approximation ratio $1+δ$, has realized ratio at most $\sqrtα+αδ$. A nominal approximation scheme then gives, for every fixed $\varepsilon>0$, a deterministic polynomial-time algorithm independent of $α$ with ratio $(1+\varepsilon)\sqrtα$ simultaneously for $1\leqα\leq4$.

0.32

On the Propeller Conjecture

cs.CC math.FAmath.PR ·2026-10-23 ·Steven Heilman
unusual
Abstract · 16 pages, 1 figure

We simplify and improve OpenAI's proof of the propeller conjecture of Khot and Naor from 2008: the sum of the squared lengths of the Gaussian moments of any partition of Euclidean space is at most $9/(8π)$. We show equality only occurs for three 120 degree sectors in the plane (up to symmetries, measure zero changes, and taking orthogonal cross products with Euclidean space). We also remove computer assistance from the three-dimensional case of the conjecture, which previously used an epsilon net traversal.

0.32

Variance of relative entropy over random states

cs.IT math-ph ·2026-10-23 ·Linfeng Wei, Lu Wei
Abstract · 23 pages, 2 figures

Average relative entropy over major generic random states is recently studied. In this work, the investigation is extended by deriving the variances of relative entropy over the Hilbert--Schmidt and Bures--Hall ensembles, which characterize its fluctuations around the average value. Exact results for the variances of relative entropy between independent states drawn from the same ensemble and from two different ensembles are obtained. These results further indicate that the average relative entropy is typical in the considered limiting regimes. In addition to the matrix-level calculations similar to the average relative entropy, eigenvalue correlation functions of the ensembles are involved in the variance computations through the use of zonal polynomials.

0.32

Targeting Interventions for Volatility Reduction in Network Systems - Part I: Idiosyncratic Shocks

cs.GT ·2026-10-23 ·Luca Damonte, Leonardo Cianfanelli, Giacomo Como, Fabio Fagnani
Abstract · 13 pages, 7 figures

We study optimal intervention design problems aimed at minimizing the aggregate volatility in network systems. We consider linear network equilibrium models arising in network games, production networks, and opinion dynamics, where exogenous shocks hitting a few nodes can propagate through the network and cause detrimental systemic effects. Focusing on idiosyncratic shocks, we first identify a general network centrality measure that determines the amplification factor of the variance of each node's shock to the aggregate system volatility, and show that, in relevant special cases, it reduces to Katz-Bonacich, cycle, and concentration centralities. We then consider a system planner who can attenuate the shocks on the nodes, subject to a budget constraint, so as to minimize the aggregate volatility. We solve this problem analytically and show that its solution has a water-filling-type structure: for low budget, the intervention is concentrated on the node(s) maximizing a quantity ω_i that combines the shock's variance, the node's centrality, and the intervention cost; as the budget increases, more and more nodes are targeted, and for sufficiently large budget every node with ω_i > 0 receives an intervention proportional to ω_i. Finally, we study a robust version of the problem, where the planner minimizes the volatility against the worst-case vector of shock variances chosen by an adversary within a given budget. We solve this min-max problem analytically and show that its solution also has a water-filling-type structure, with the ratio between the node's centrality and the square root of the adversarial cost playing the role of ω_i.

0.32

Settling Nondeterministic Communication Complexity of Graph Connectivity

cs.CC ·2026-10-23 ·Zhijun Zhang
publishedtheory
Abstract

Determining the communication complexity of graph connectivity has been one of the longest standing open problems in communication complexity since the seminal work of Babai, Frankl, and Simon [FOCS'86]. A simple protocol transmits a spanning forest using $O(N\log N)$ communication on $N$-vertex graphs. While this was shown optimal for deterministic protocols by Hajnal, Maass, and Turán [STOC'88], the optimal nondeterministic and randomized complexities have remained unresolved for decades. We settle the nondeterministic communication complexity of graph connectivity by proving an optimal $Ω(N\log N)$ lower bound. The best prior bound was $Ω(N\log\log N)$ due to Raz and Spieker [Combinatorica'95]. Our bound is shown by upper bounding the maximum size of $1$-monochromatic rectangles. Generalizing our techniques to weighted, almost $1$-monochromatic rectangles yields an $Ω(N \cdot \log(\min(1/\varepsilon,N)))$ lower bound for randomized protocols with error probability $\varepsilon$. In particular, this bound is optimal for any polynomially small $\varepsilon$. For constant error probability, we also prove that any $r$-round protocol requires $Ω(N/r \cdot \log N)$ communication, which is optimal for every constant $r$. Previously, optimal $Ω(N\log N)$ lower bounds for graph connectivity were known only for $1$-round [Sun, Woodruff; APPROX/RANDOM'15] and $2$-round protocols [Radhakrishnan, Reddy, Venkat; ITCS'26]. Our proof introduces a new communication problem which can be viewed as a composition of inner product and pointer chasing problems. Our lower bounds can be extended to other fundamental graph properties, and also imply optimality of known query and streaming algorithms for related problems.

0.31

On the complexity of Medvedev reducibility

cs.CC math.LO ·2026-10-23 ·Remi Pallen
theory
Abstract

The Medvedev lattice is a tool used to classify the complexity of effective closed sets. In this article, we are interested in the complexity of using this tool. In particular, we study the complexity of deciding whether two effective closed sets have the same degree, when one is fixed and the other is given as input. This problem is naturally~$Σ^0_3$, and among other results, we show that for all degrees except the maximal one, the problem reaches its maximal complexity. It turns out that the argument is not uniform and depends on the self-similarity of the fixed degree, a notion that we introduce in this article. The complexity of having maximal degree is particularly difficult to establish, but we obtain a lower bound that almost matches the upper bound. We also show that deciding whether an effective closed set has a degree incomparable with a fixed degree $d$ is $Π^0_3$-complete whenever $\mathbf{0}_s <_s d <_s \mathbf{1}_s$.

0.30

The Minimum Number of Samples in the Labeled Coupon Collector Problem

cs.DM ·2026-10-23 ·Andrew Tan, Daniella Bar-Lev, Ryan Gabrys, Zohar Yakhini, Paul H. Siegel
Abstract · 26 pages + 2 pages Bibliography + 3 pages Appendix, 4 figures, Submitted to SIAM Journal on Discrete Mathematics

The Labeled Coupon Collector Problem (LCCP) is an extension of the classical Coupon Collector Problem (CCP) in which one must deduce an unknown mapping between coupons and labels by collecting unordered subsets of k coupons and their associated labels, where deduction relies on analyzing the intersections and differences of the unordered subsets. In this work, we establish a fundamental connection between the LCCP and the combinatorial theory of separating systems. Using this framework, we completely characterize the minimum required number of samples to deduce all labels in two settings: Type-I, where the label set is known in advance, and Type-II, where it is unknown. We provide deterministic constructions of minimal separating and covering systems, yielding explicit bounds that are tighter than those found in existing literature. Furthermore, we extend our analysis to the partial recovery problem, characterizing the minimum number of samples required to identify a specified number of target coupons.

0.30

A Linear Lower Bound for Convex Body Chasing in Every Norm

cs.DS cs.CGmath.MG ·2026-10-23 ·Nicolas Christianson
theory
Abstract · Updates the previous version to extend the lower bound to all norms

We establish an $Ω(d)$ lower bound on the competitive ratio of convex body chasing in $\mathbb{R}^d$ with movement costs measured in any norm. Previously, the best known lower bound valid across all norms was $Ω(d^{1/6})$, established by Papazov (2023); stronger lower bounds were known for $\ell_p$ norms, where Bubeck, Klartag, Lee, Li, and Sellke (2019) had established a lower bound of $Ω(\max\{\sqrt{d}, d^{1-1/p}\})$. In particular, a linear lower bound was known only for the $\ell_\infty$ norm, realized by the standard construction of chasing faces of a hypercube. Our lower bound strictly improves on the prior universal and $\ell_p$ lower bounds for $p < \infty$ and shows that the functional Steiner point algorithm of Sellke (2023) - which has competitive ratio $O(d)$ - is asymptotically optimal for every norm. Our approach is based on the recursive construction of Bubeck, Coester, and Rabani (2023) for metrical service systems, which gives a lower bound of $Ω(r^2)$ for requests of size $O(r)$. We show that requests in their construction, each comprising a set of discrete points in a graph metric, can be mapped to the vertices of convex requests in a $(2r+1)$-dimensional $\ell_p$ vector space while approximately preserving movement costs up to a factor of $O(r)$; the general result then follows via Milman's (1985) quotient-of-a-subspace theorem.

0.29

Proximity Theorem and FPT Algorithm for Minimum Group-labeled Matroid Bases

cs.DS math.CO ·2026-10-23 ·Siyue Liu, Taihei Oki, Chao Xu
theory
Abstract

Let $M$ be a matroid, let $\ell:E(M)\to G$ be a labeling of its elements by an abelian group, and let $F\subseteq G$ be finite. A base is $F$-avoiding if the label sum of its elements is not in $F$. We prove that, for arbitrary costs, a min-cost $F$-avoiding base can be obtained from any min-cost base using at most $|F|$ exchanges. This resolves the weighted proximity conjecture by Hörsch, Imolay, Mizutani, Oki, and Schwarcz. Our main technical tool is to replace the assumption of Schrijver--Seymour Conjecture in Liu and Xu's earlier work with Kemperman--Scherk Theorem. The proximity theorem combined with approaches in earlier work gives an FPT algorithm that finds a min-cost $F$-avoiding base parameterized by $|G|$. We additionally give an FPT algorithm parameterized by $|F|$, answering a question of Hörsch et al.

0.29

Polarization on Bounded Euclidean Spaces and More General Compact Groups

cs.IT ·2026-10-23 ·Sharang M. Sriramu, Lampros Gavalakis, Ioannis Kontoyiannis, Aaron B. Wagner
unusual
Abstract

We establish polarization for memoryless sources taking values on the unit-perimeter circle, or on more general compact groups, under Arıkan's polar transform. For sources with finite differential entropy, all but an $O((\log N)^{-1/12})$ fraction of the conditional differential entropies in successive cancellation order lie within any fixed positive distance of zero, where $N$ is the block length. Examples indicate that the true polarization rate is indeed very slow: an inverse-logarithmic rate does not underestimate the polarization speed at moderate block lengths.

0.28

Stable Set Reconfiguration in Dense Digraphs

cs.DM cs.CCmath.CO ·2026-10-23 ·Sarah Houdaigoui, Nacim Oijid, Matthieu Petiteau, Alexandre Vigny
Abstract

In a directed graph, a stable (or acyclic) set is a set which does not induce a directed cycle. Given a directed graph and two acyclic sets of the same size, the Acyclic-Set-Reconfiguration with Directed Token Sliding (ASR) problem asks whether one can, by iteratively replacing a vertex by an adjacent one, transform the first set into the second one, while maintaining acyclicity during the transformation. In this work, we present a study of the complexity of ASR and provide an FPT algorithm for semi-complete digraphs, when parametrized by the size of the sets being reconfigured, the directed pathwidth (or equivalently, the directed treewidth) of the instance, and a newly introduced parameter called the iteration of the reconfiguration sequence. To the best of our knowledge, the reconfiguration of stable sets has only been studied on undirected graphs, or (more recently) on sparse directed graphs.

0.28

Single-Machine Guarantees on Unrelated Machines and Beyond: Proportional Fairness for Nonclairvoyant Polytope Scheduling

cs.DS ·2026-10-23 ·Sven Jäger, Alexander Lindermayr, Bart Zondervan
theory
Abstract · Shortened abstract

The Polytope Scheduling Problem (PSP) is a general framework for preemptive scheduling in which feasible instantaneous job allocations form a packing polytope. In the nonclairvoyant online setting, Proportional Fairness (PF), a practical allocation rule rooted in market theory, is the leading general-purpose algorithm. On a single machine, PF coincides with weighted Round-Robin, which is $2$-competitive for weighted completion time with uniform release dates (optimal for nonclairvoyant algorithms) and $(2+\varepsilon)$-speed $O(1/\varepsilon)$-competitive for weighted flow time. Known guarantees beyond a single machine are weaker. This raises a central question: Are these heterogeneous environments intrinsically harder than a single machine under nonclairvoyance? We show that they are not: PF retains both single-machine guarantees on the novel class of support-submodular polytopes, which includes unrelated machines and polymatroids. On unrelated machines, this refutes a belief (JACM'18, SODA'25) that PF is not $O(1)$-speed $O(1)$-competitive for weighted flow time. Our main technical contribution is an equivalent characterization: a polytope is support submodular if and only if, for all job weights, the optimal weighted completion time is concave in the processing requirements. This structural insight underlies both guarantees. For general PSP, the picture changes: we give the first lower bound above $2$, showing that every randomized nonclairvoyant algorithm has a competitive ratio of at least $2+10^{-5}$ for weighted completion time, even with uniform release dates. On the positive side, we substantially improve the best known guarantee for weighted completion time from $27$ (SICOMP'26) to $4$ via a new dual-fitting analysis of PF. For weighted flow time, we show that PF achieves the first $O(1)$-speed $O(1)$-competitive nonclairvoyant guarantee for graph scheduling.

0.27

Fisher Information and the Minimum Mean-Square Error for Locating Closely Spaced Point Sources

cs.IT ·2026-10-23 ·Chuling Wen, Jian Lu, Qingtang Jiang, Charles K. Chui
theory
Abstract

We determine the optimal average error of locating closely spaced sources from noisy superpositions of a known smooth response with unknown complex coefficients. Relative phases control how changes in one location can be offset by the other unknowns, so local Fisher information alone does not determine the attainable error. For every fixed number $K\ge2$ of sources with independent uniform phases, we prove matching Bayes risk bounds jointly in separation and noise, for all sufficiently small separations and all noise levels. Odd source counts have two risk regimes; even counts have an intermediate regime in which mean squared error is linear in the noise standard deviation. A projected least-squares estimator attains these orders without knowing the phases or the prior density. Bounded nonuniform phase laws preserve the orders under stated support conditions, whereas priors concentrated near poorly identified configurations can change the exponents. A common finite comparison yields both a global inverse bound and paths controlled by the original prior. These give sufficient conditions under which the risk has the order of the pointwise capped Cramér--Rao bound, and a lower-tail calculation explains the power and logarithmic regimes. We also give conditional guarantees when the cluster center, separation, and source count are unknown, with an additional low-noise requirement for count selection. A calibrated covariance factor model illustrates the same criterion beyond source localization.

0.26

Polyhedral RIS Architecture with Quantum-Assisted Facet Selection for Uniform Coverage

cs.IT eess.SY ·2026-10-23 ·Noha Hassan, Halim Yanikomeroglu
Abstract · This work has been submitted to the IEEE WCL for possible publication. Copyright may be transferred without notice, after which this version…

Reconfigurable intelligent surfaces (RISs) direct coverage toward moving user equipment (UE), but practical deployments involve discrete phase control, uncertain facet positions, and UE mobility, which require the problem to be re-solved continually. We consider these challenges for the chamfered hexagonal polyhedral RIS, which provides nearly uniform coverage. With both incidence and departure angles, we bound in closed form the gain loss of its best facet relative to the best possible reflecting facet, include the known b-bit quantization loss, and extend the bound to joint random mounting errors of all facets. For a precoded multi-user downlink, each facet-selection subproblem of the max-min rate design is shown to be an exact quadratic unconstrained binary optimization (QUBO) problem, suited to repeated time-constrained solving. The deployment of the base station in relation to the RIS allows the polyhedral surface to bring down the percentage of dead zones in vertical planar panels from 77% to 0.3% and increase the minimum rate by 34% relative to the STAR panel. The simulated annealing algorithm finds the optimum solution in 95% of the cases. The quantum approximation algorithm fails to do so.

0.26

Distortion-Rate Outage Probability for Non-ergodic Classes of Sources

cs.IT ·2026-10-23 ·Dongmin Park, Ibrahim Abou-Faycal, Razane Tajeddine
Abstract

The concept of outage has been commonly used in the context of transmission rates over non-ergodic channels. For those channels, one channel state or realization may be encountered for the whole transmission duration of a codeword. In the dual setting of source coding and whenever source statistics are not ergodic, we define in this work the notion of outage; we discuss related properties, consider the special case of compact source classes and provide examples.

0.26

Beam Search for Error-Correcting Codes via Conditional Diffusion

cs.IT ·2026-10-23 ·Ziv Aharoni, Henry D. Pfister
Abstract

We introduce a diffusion framework for neural decoding of error-correcting codes (ECCs). We interpret decoding as a conditional diffusion process that progressively predicts codeword bits while treating the physical channel observation as fixed side information. At each stage, a neural model estimates bitwise posteriors conditioned on the physical observation and the current noisy state of the codeword. We focus particularly on masked diffusion, for which the reverse process naturally induces a Maxwell beam search (MBS) procedure inspired by the Maxwell construction of Measson et al. Rather than committing to a single uncertain decision, the decoder maintains multiple candidate trajectories, removes parity-infeasible candidates, and ranks the remaining candidates using accumulated reverse-transition log-probabilities. We evaluate the framework across multiple code families on binary phase-shift keying (BPSK) over additive white Gaussian noise (AWGN). Experiments show competitive decoding performance across Polar, low-density parity-check (LDPC), and Bose-Chaudhuri-Hocquenghem (BCH) codes. The MBS decoder consistently improves upon the corresponding single-trajectory diffusion decoder and, in all evaluated settings, improves upon available neural decoding results.

0.26

Optimal Voting: Beyond a Utilitarian View

cs.GT ·2026-10-23 ·Ben Armstrong, Nicholas Mattei
Abstract · 19 pages, 12 figures, 1 table

A common measure of the quality of a voting rule is the social welfare of the alternatives it elects. Prior work shows that, when voter utilities are drawn from a neutral distribution, expected utilitarian welfare is maximized by a positional scoring rule. We extend this approach to the power mean family of welfare functions, which includes utilitarian, Nash, and egalitarian welfare as special cases. We show that, across the power mean family, a deterministic positional scoring rule (PSR) is asymptotically optimal for expected social welfare among all social choice functions and we characterize this rule for a broad class of utility distributions. We further show that, among randomized PSRs, plurality uniquely maximizes expected social welfare for all power means. Experimentally, we find that optimal score vectors often yield substantially higher welfare than canonical score vectors. On highly structured utility distributions, optimal score vectors take on novel shapes unlike existing canonical score vectors, highlighting open questions for future analysis.

0.26

Joint Service Depth and Priority Design in Queueing Systems

cs.GT econ.THmath.OC ·2026-10-23 ·Adam N. Elmachtoub, Yash Kanoria, Jonathan Y. Tan
Abstract

In AI-assisted service systems, the server can choose how much processing effort to spend on each job, which we call the service depth. Deeper reasoning improves output quality but occupies the server longer and increases congestion, and the realized processing time remains random because AI models are probabilistic and tasks vary in complexity. We study the joint design of service depth and priority in a single-server queue whose customers differ in delay sensitivity, with the goal of maximizing social welfare. For each customer type, the decision maker chooses a priority level and a service depth, and service times are exponential with mean equal to the chosen depth. With two customer types and linear service values, we characterize the centralized optimum. Counterintuitively, it prioritizes the patient type and gives it a smaller service depth when patient customers are sufficiently prevalent. When delay sensitivity is private, however, incentive compatibility requires the impatient type to receive a weakly shorter sojourn time, so the optimal price menu never prioritizes the patient type. Consequently, there can be a welfare gap between the centralized decision maker's optimal control policy and the optimal decentralized pricing mechanism. With a continuum of types, the centralized optimum is monotone: more delay-sensitive customers receive weakly smaller depth and higher priority, so the optimum can be implemented by a price menu with no welfare loss. Allowing randomized depths similarly closes the gap in the two-type model. Finally, we calibrate an agentic AI service system using published coding-agent trials from DeepSWE and numerically examine priority inversion, the value of joint design, and the welfare gain from personalization.

0.26

Low-Crossing Spanning Paths via Quasi-Uniform Sampling and their Applications

cs.CG cs.DS ·2026-10-23 ·Mónika Csikós, Nabil H Mustafa
Abstract

Given a set system (X, S), we consider the problems of constructing perfect matchings and spanning paths (also known as Welzl orders) of X with low crossing numbers with respect to S. Matchings with low crossing numbers were originally introduced in the late 1980s in the seminal works of Welzl [29, 30] and Chazelle-Welzl [7], who showed that any set system (X, S) with bounded dual VC-dimension d has a matching with crossing number O |X| 1-1/d , which is optimal. Low-crossing matchings and spanning paths have found several applications outside of their initial realm of computational geometry, including algorithmic graph theory, model checking, and game theory. In many of these applications, |S| is near-linear in |X|, which motivated the search for an improved construction for this regime. However, the proposed new algorithms provided only average-case bounds or sub-optimal crossing number guarantees for d $\ge$ 2 [5, 12]. We propose a new quasi-uniform sampling based framework. This allows us to obtain a simple algorithm with expected running time of __ |X| 2 + |S| $\bullet$ |X| , which improves upon the previous-best Our result has several immediate consequences, including new faster algorithms for constructing low-crossing spanning paths for set systems induced by balls, as well as for neighborhood set systems of graphs. Moreover, we get the first quadratic-time algorithm for combinatorial discrepancy, with nearly optimal discrepancy (up to a log factor), for bounded dual VC-dimension set systems. Beyond these applications, we believe that our idea could be applied to other algorithms that use the multiplicative weight update method.

0.25

Capacity Modification for Matching with Couples: An Integer Programming and Computational Study

cs.GT ·2026-10-23 ·Zexiao Wu, Genjiu Xu, Péter Biró
Abstract · 31 pages, 4 figures

We study capacity modification in the Hospitals/Residents problem with Couples, in which hospital capacities and a stable matching are jointly determined. Exact and Cutoff integer programming formulations are developed for the resulting problem. We assume that, for each couple, either both members are matched or neither is. For arbitrary strict preferences, every instance admits a stable matching after changing each hospital capacity by at most two. Total capacity increases by at most one, and this bound is tight. Under sub-responsive and sub-complete couple preferences, a stable matching exists after adding at most one position to each hospital. The total number of additions is at most the number of couples, and this bound is also tight. Computational experiments show that Exact formulations are modestly faster at fixed capacities, whereas Cutoff formulations perform substantially better in the minimum repair experiments and attain higher completion rates in markets with up to 30,000 doctors. Across 96,000 synthetic and reconstructed markets, small capacity changes restore existence whenever no stable matching exists at the original capacities.

0.25

Logarithmic Separation between LZ77 and LZ-End and Grammar Compression

cs.IT cs.DS ·2026-10-23 ·Philip Bille, Inge Li Gørtz, Rikke Jessen, Jingxun Li, Simon Rumle Tarnow +1 more
Abstract

We show that there is a logarithmic separation between LZ77 and LZ-End, and between LZ77 and grammar compression.

0.24

Acyclic 2-edge-coloured graphs: polynomial-time linkages and colouring hardness

cs.DS math.CO ·2026-10-23 ·Jørgen Bang-Jensen, Stéphane Bessy, Amadeus Reinald
Abstract · 21 pages, 4 figures

A path or cycle in a $2$-edge-coloured graph is alternating if consecutive edges have different colours. These objects are natural analogues of paths and cycles in graphs, and even more so in digraphs. A well-studied class is that of acyclic $2$-edge-coloured graphs, that is, those forbidding alternating cycles. In particular, (acyclic) $2$-edge-coloured graphs generalize directed (acyclic) graphs (DAGs), but the complexity of problems on DAGs is much better understood than that of the corresponding problems on acyclic $2$-edge-coloured graphs. We investigate the complexity of two such problems, whose directed counterparts are well-known. Our first result is an $n^{O(k)}$-time algorithm for the alternating $k$-linkage problem on acyclic $2$-edge-coloured graphs. This problem asks for $k$ vertex-disjoint alternating paths joining $k$ prescribed pairs of vertices. This generalizes the directed $k$-linkage algorithm for DAGs. For every fixed $k$, this also yields a polynomial-time algorithm for finding $k$ internally vertex-disjoint alternating $(s,t)$-paths, solving an open problem by Abouelaoualim et al. from 2008. We then consider the alternating chromatic number of $2$-edge-coloured graphs, defined as the minimum number of parts in a vertex partition such that each part induces an acyclic subgraph. This notion is analogous to the dichromatic number of digraphs, where it is known that deciding dichromatic number $2$ is NP-complete even on tournaments. Our second result is that deciding alternating chromatic number $2$ is NP-complete on complete $2$-edge-coloured graphs. Equivalently, deciding whether a graph can be partitioned into two induced threshold graphs is NP-complete.

0.24

Encoding-Sensitive Nullstellensatz Certificates for Syndrome Exclusion in LDPC Codes: Sharp Local Bounds and a Linear Separation

cs.IT ·2026-10-23 ·Malihe Aliasgari, Hassan Khodaiemehr, Khadijeh Bagheri
theory
Abstract · 48 pages, 4 tables. Extended journal version of a paper submitted to IEEE ITW 2027

Certifying that a parity-check matrix has no nonempty set of at most t variable nodes leaving at most b checks unsatisfied arises in minimum-distance and trapping-set analysis. Such absence is the infeasibility of a polynomial system, provable by an independently checkable Nullstellensatz certificate. The degree of these certificates depends strongly on how the two cardinality constraints are encoded, although the encodings have identical solutions. For a local encoding with inclusion-exclusion and cardinality obstructions, we construct explicit certificates of degree at most t + b + K + 1, with K logarithmic in the maximum check degree; this bound is independent of the block length and cannot be improved in general. For a standard counting encoding, a generalized planted-knapsack theorem, which pulls back a knapsack functional with an arbitrary target, gives lower bounds linear in the block length for cycle codes, regular codes of girth at least six near the feasibility boundary, array-based quasi-cyclic codes and minimum-distance certificates for arrays of circulant permutation matrices, where the local degree stays bounded. On array codes and regular codes of girth at least eight, planted column pairs extend these bounds to odd column weight with no unsatisfied checks. The planting argument is optimal among such pullbacks. All computed degrees and certificates are verified in exact rational arithmetic; bounds are distinguished from exact minimum degrees. Local certificates for quasi-cyclic matrices, including a lifted fifth-generation New Radio base graph, pass two independent verifiers.

0.23

Deterministic Root-Finding over Finite Fields with Elliptic Curves

cs.SC cs.CCmath.NT ·2026-10-23 ·Pierre-Jean Spaenlehauer
Abstract

Finding the roots of a polynomial over a finite field $\mathbb F_q$ in deterministic polynomial time is a long-standing open problem. Classical deterministic methods work in the multiplicative group $\mathbb F_q^\times$ and are efficient only when $q-1$ is smooth, which is rarely the case. We propose a method that replaces $\mathbb F_q^\times$ by the groups of rational points of an elliptic curve $E$ and of its quadratic twist $\widetilde E$. The computation splits into an offline phase, which depends only on $\mathbb F_q$ and outputs a short \emph{elliptic preprocessing datum} $\mathcal D$, and an online phase, which processes the input polynomial. Writing $Ξ(\mathcal D)$ for the largest prime factor of $\lvert E(\mathbb F_q)\rvert\cdot\lvert\widetilde E(\mathbb F_q)\rvert$, the online phase is deterministic and finds the roots of any $f\in\mathbb F_q[X]$ of degree $n$ within $\widetilde O(Ξ(\mathcal D)^2\,n\log^2 q+n^ω\log q+n\log^3 q)$ operations in $\mathbb F_q$. A precursor of our method is an algorithm by Rónyai, which recovers the roots through Galois computations over number fields, relies on GRH, and comes with a polynomial bound of unspecified degree; in contrast, our online phase works entirely over $\mathbb F_q$ and is unconditional. Moreover, $\mathcal D$ can be certified deterministically and unconditionally in time polynomial in $\log q$ and $Ξ(\mathcal D)$. We conjecture that every $\mathbb F_q$ of large enough characteristic admits a datum with $Ξ(\mathcal D)\leq(\log q)^{O(1)}$, and support this with a heuristic and with experiments. Constructing such a datum in polynomial time remains open.

0.22

Deterministic Incremental Triconnectivity in $O(\log^2 n)$ Worst-Case Time

cs.DS ·2026-10-23 ·Niels Inselmann Christiansen, Eva Rotenberg
Abstract

We consider the problem of maintaining the triconnectivity in a graph $G$ in the incremental dynamic setting. We consider updates in the form of edge insertions and edge subdivisions, and triconnectivity queries, i.e. queries that take as input variables the vertices $a$ and $b$ and are of the form ''are there at least three vertex disjoint paths between $a$ and $b$?''. All algorithms presented in this paper are deterministic. When starting from a biconnected graph, all updates take worst-case $O(\log n)$ time, and all queries are answered in $O(\log n)$ time. By combining this result for biconnected graphs with a result of Holm and Rotenberg [SODA'20] we get a deterministic algorithm for general graphs with a worst-case update and query time of $O(\log^2 n)$. Both algorithms represent an improvement of the state-of-the-art by a factor of $O(\log n)$.

0.22

Exact Algorithms and Complexity of the Smallest Grammar Problem

cs.DS ·2026-10-23 ·Xiangle Chen
theory
Abstract · Revised version with a uniform alphabet-sensitive subexponential algorithm and quantitative transfers to SAT. Source package includes verifi…

The Smallest Grammar Problem (SGP) asks for a smallest straight-line grammar generating a given string, with size measured by the total right-hand-side length. For an input of length $n\ge2$ over at most $σ$ symbols, where $2\leσ\le n$, let $R=\log_σn$. We give a deterministic exact algorithm running in $n^{O(1)}2^{O(n\log\log n/(R\sqrt{\log n}))}$ time and polynomial space, with constants independent of $σ$. This gives time $2^{O(n\log\log n/(\log n)^{3/2})}$ for fixed alphabets and $2^{O(n\log\log n/\sqrt{\log n})}$ for arbitrary alphabets, up to polynomial factors. The algorithm combines pairwise disjoint representative occurrences of dictionary strings, a sparse interval certificate, and the Boolean tensor contraction theorem of OpenAI. It explicitly enforces dictionary membership, rule boundaries, and the grammar-size budget. Independently of that contraction theorem, disjoint representatives yield exact algorithms in $O^*(1.872258^n)$ time and $2^{O((n/R)\log(R+1))}$ time using $O(n)$ words of space. Anchor-based grammar projections establish binary NP-completeness and reductions with tunable alphabet size. We state their consequences as unconditional quantitative transfers from SGP algorithms to sparse SAT algorithms, and quantify the additional loss from sparsifying general formulas.

0.21

Selection under partial information

cs.DS ·2026-10-23 ·Benjamin Aram Berendsohn
Abstract

The selection problem asks to find the $k$-th smallest element in a given set $X$. Its complexity is well-known to be $Θ(|X|)$ for every fixed $k$. We consider the poset selection problem where a partial order $P$ on $X$ is given, and the underlying order of $X$ is guaranteed to be a linear extension of $P$. This generalizes several variants of the selection problem studied before. Our main result is an algorithm that performs $O( k + \operatorname{OPT}(P,k) )$ comparisons, where $\operatorname{OPT}(P,k)$ is the minimum number of comparisons required for the input $(P,k)$. If $P$ is a rooted tree, we obtain an $O( \operatorname{OPT} \log (k / \operatorname{OPT}) )$ algorithm, which can be further improved to $O( \operatorname{OPT} \log \log (k/\operatorname{OPT}) )$ or $O( \operatorname{OPT} \log \operatorname{OPT})$ under mild assumptions about $P$. Our algorithms are based on the work of Kaplan, Kozma, Zamir, and Zwick [SOSA 2019], who used Chazelle's soft heaps [J. ACM 2000] to simplify several older results. We introduce a new soft heap variant, the weighted soft heap, which allows greater control over which elements get corrupted.

0.20

Pareto-optimal Non-Clairvoyant Scheduling with Predictions

cs.DS cs.GT ·2026-10-23 ·Ziyad Benomar, Vianney Perchet
theory
Abstract

We study preemptive single-machine non-clairvoyant scheduling with predictions, where a prediction of the order of the jobs (or of their sizes) is available, and the objective is to minimize the total completion time. An algorithm is $c$-consistent if it is $c$-competitive when the prediction is correct, and $R$-robust if it is $R$-competitive whatever the prediction. Since the introduction of this problem as one of the first examples of learning-augmented algorithms, the best possible trade-off between consistency and robustness has remained open. We determine it exactly for deterministic algorithms. For every $R>2$, we give an $R$-robust algorithm whose consistency is \[ c(R)=R\Big(1-\Big(1-\frac{R+2}{R^2}\Big)^{\frac R{R+2}}\Big), \] and we prove that no deterministic $R$-robust algorithm, with either predicted order or predicted job sizes, has better consistency. In particular, consistency $1+\varepsilon$ is achievable with robustness $\varepsilon^{-1/2}+\frac16+O(\sqrt\varepsilon)$, whereas all previous algorithms required robustness $Θ(1/\varepsilon)$. Our algorithm is greedy: it follows the predicted order as fast as possible, subject to the invariant that the cost paid so far is at most $R$ times the optimum of the instance in which every unfinished job would end immediately. The matching lower bound rests on the observation that every robust deterministic algorithm must satisfy the same invariant along its run on correct predictions. Furthermore, we give a simple randomized algorithm, using predicted sizes, with consistency $1+\varepsilon$ and robustness $(3\varepsilon)^{-1/2}+O(1)$, strictly below the deterministic frontier. The question of determining the Pareto-optimal frontier among randomized algorithms remains open.

0.19

Learning Positive Linear Combinations of ReLUs in Fully Polynomial Time

cs.DS cs.LG ·2026-10-23 ·Ilias Diakonikolas, Jelena Diakonikolas, Puqian Wang, Nikos Zarifis
published
Abstract

We study the fundamental problem of PAC learning positive linear combinations of $k$ arbitrary ReLU activations on $\mathbb{R}^d$ with respect to the Gaussian distribution, in $L_2$ loss. We give the first fully-polynomial time algorithm for this learning task. Specifically, our algorithm draws $\mathrm{poly}(d, k, 1/ε)$ samples, runs in $\mathrm{poly}(d, k, 1/ε)$ time, and outputs a hypothesis that is $ε$-close to the target concept with high probability. All prior algorithms for this problem require super-polynomial time in $k$ or $1/ε$, or leverage strong well-conditioned assumptions on the weight matrix. On a technical level, our approach develops a novel boosting procedure of a ``weak'' learner that recovers the hidden weight vectors to constant error, in a list-decodable manner. To develop this ``weak'' learning procedure, we leverage implicit tensor computation techniques to construct a low-degree filtering procedure based on Hermite tensors of the target function. This procedure amplifies separation among (clusters of) weight vectors to learn each of them to constant error. We then argue that the same procedure can be used for improving the estimation of hidden weight vectors, using gradient descent-like updates on the Euclidean sphere. A crucial ingredient of our algorithm is a pruning approach that ensures only relevant vectors that approximate the hidden weight vectors at target accuracy are maintained by the algorithm.

0.19

Pinching Antenna and Relay Assisted Joint Indoor Outdoor NOMA Communications with SWIPT

cs.IT ·2026-10-23 ·Amit Agarwal, Susmita Sain, Keshav Singh, Sudip Biswas
Abstract

This paper presents a unified downlink architec- ture for serving indoor and outdoor users from a shared multi-antenna rooftop base station (BS). To overcome severe building-penetration loss, the BS transfers information and radio- frequency (RF) energy to an indoor access point (AP) using simultaneous wireless information and power transfer (SWIPT). The AP harvests this energy and performs amplify-and-forward (AF) relaying, enabling self-sustained indoor communication. A movable waveguide-mounted pinching antenna (PA) further improves indoor coverage by adapting its position to the prop- agation environment. For outdoor transmission, zero-forcing (ZF) beamforming suppresses inter-group interference, while power-domain non-orthogonal multiple access (NOMA) supports multiple users within each group. We maximize the outdoor weighted sum rate (WSR) while satisfying stringent indoor and outdoor quality-of-service (QoS) requirements. The resulting nonconvex problem jointly optimizes the BS power fraction, SWIPT power-splitting ratio, PA position, and NOMA power allocation. To obtain an efficient solution, we develop a two-stage method based on alternating optimization (AO) and sequential quadratic programming (SQP). Numerical results show that the proposed architecture improves outdoor spectral efficiency while maintaining reliable, energy-sustainable indoor coverage. Increasing the BS array size further raises the indoor QoS- satisfaction probability toward unity, even under demanding rate targets.

0.17

Structure and Exact Certification of Nonnegative Tensor Decompositions

cs.DS cs.CCmath.AGmath.CO+1 ·2026-10-23 ·Haoming Wang, Ming Yuan
theory
Abstract

We study the structure and exact certification of nonnegative tensor decompositions supplied through rational factor matrices. Every failure of minimality or essential uniqueness contains an irreducible positive exchange between $p$ supplied terms and $q$ competing terms. A dimension count based on the Lovitz--Petrov splitting theorem bounds the effective order of such an exchange, the sum over the modes of the joint factor-span dimension minus one, by $p+q-2$, and this bound is attained at every supplied length $R\ge2$. After eliminating ambient-row variables by the Farkas lemma, the resulting semialgebraic systems have fewer than $2R^2$ real variables. For input bit length $L$, this gives a deterministic $G(R)L^{O(R^2)}$-time algorithm that computes the nonnegative rank and an exact minimum-length decomposition, decides minimality and essential uniqueness, and returns exact witnesses when either property fails. The exponent is independent of the tensor order and the mode dimensions. For Lebesgue moment tensors, positive quadrature determines the nonnegative rank and the uniqueness behavior, and some rational length-$R$ inputs force a common field of degree at least $\lfloor R/2\rfloor!$ for any exact minimum-length output. As an illustration, an implementation resolves all 114 instances left inconclusive by earlier positive-scattering criteria, of which 95 are minimal and essentially unique and 19 are minimal but nonunique.

0.12

A proof of Wang's planar $D$-coloring conjecture

cs.DM math.CO ·2026-10-23 ·Harshit Verma
theory
Abstract · 22 pages, 3 figures. Ancillary files contain the certificates and verification scripts for the computer-assisted case of maximum degree 6

We prove Wang's planar $D$-coloring conjecture. A $D$-coloring of a graph is a proper edge coloring in which every subgraph isomorphic to $K_4$ minus an edge is rainbow. Wang conjectured that every planar graph of maximum degree $Δ$ has a $D$-coloring with $9$ colors if $Δ=4$, $10$ colors if $Δ=5$, and $2Δ-1$ colors if $Δ\ge6$. Hu, Kong and Wang proved the cases $Δ\le5$ and $Δ\ge33$. We prove the sharp bound $2Δ-1$ for every $Δ\ge6$, settling the remaining cases. The proof uses discharging; the case $Δ=6$ is computer-assisted.

0.10

Robotics, Control & Signals 87

Automated Fetal Magnetocardiography Reconstruction Using a Biophysical Forward Model

eess.SP ·2026-10-23 ·Jonas Emrich, Karin Narushima, Lena Wunderl, Philipp Wunderl, Tobias Jensch +5 more
top venuecode/data
Abstract · 29 pages, 16 figures. Includes supplementary material. Published in IEEE Transactions on Biomedical Engineering. Code: https://github.com/ML…

Objective: Fetal magnetocardiography (fMCG) enables early detection of cardiac disease by non-invasively measuring fetal electrophysiological activity. However, current signal estimation methods for fMCG require complex, labor-intensive human processing because they rely on blind source separation or template-based projection techniques. Methods: We propose an automated signal reconstruction approach that recovers fetal and maternal cardiac activity from fMCG measurements without manual annotation. Our approach leverages the relationship between measurements and cardiac signals via a biophysically grounded forward model that represents cardiac sources as time-varying magnetic dipoles. This enables reconstruction of physiologically interpretable three-dimensional vectorcardiographic signals along with their spatial position. Results: Validation on synthetic data with known ground truth confirms accurate recovery of cardiac source parameters. On real fMCG recordings, the reconstruction shows strong agreement with established reference methods. Conclusion: The proposed forward-model approach automatically reconstructs fetal and maternal cardiac activity from fMCG recordings and produces signals in an interpretable vectorcardiographic source domain. Significance: Automated, interpretable fMCG reconstruction could make fetal cardiac assessment more reliable and support potential translation into prenatal care.

0.50

Causal-based approaches to explain and learn from self-extension - A review

cs.RO ·2026-10-23 ·Rebeca Marfil, Pablo Bustos, Antonio Bandera
published
Abstract · 20 pages

The last decades have seen a revolution in autonomous robotics. Deep learning approaches and their hardware implementations have made it possible to endow robots with extraordinary perceptual capabilities. In addition, they can also benefit from advances in Automated Planning, allowing them to autonomously solve complex tasks. However, on many occasions, the robot still acts without internalising and understanding the reasons behind a perception or an action, beyond an immediate response to a current state of the context. This gap results in limitations that affect its performance, reliability and trustworthiness. Deep learning alone cannot bridge this gap, because the reasons behind behaviour, when it emanates from a model in which the world is a black-box, are not accessible. What is really needed is an underlying architecture based on deeper reasoning. Among other issues, this architecture should enable the robot to generate explanations, allowing people to know why the robot is performing, or has performed, a certain action, or the reasons that may have caused a certain plan failure or perceptual anomaly. Furthermore, when these explanations arise from a cognitive process and are shared, and thus validated, with people, the robot should be able to incorporate these explanations into its knowledge base, and thus use this understanding to improve future behaviour. Our article looks at recent advances in the development of self-aware, self-evolving robots. These robots are designed to provide the necessary explanations to their human counterparts, thereby enhancing their functional capabilities, in the quest to gain their trust.

0.50

TravSUITE: Traversability via Self-Supervised, Uncertainty-Aware IRL and Terrain Estimation

cs.RO ·2026-10-23 ·Samuel Triest, Amirreza Shaban, David D. Fan, Wenshan Wang, Sebastian Scherer
publishedhighly cited
Abstract

Traversability analysis in off-road settings remains a fundamental challenge for mobile robots. Key difficulties include constructing an accurate and expressive local map from multimodal sensor data, and using the map to design traversability rules that yield desirable navigation behavior. Importantly, this system must be resilient to the limited sensing regime brought about by complex environments and high speeds. In this paper, we present TravSUITE, a traversability system suitable for high-speed navigation in off-road environments. TravSUITE consists of two major components: 1) an efficient voxel mapper that leverages visual foundation models (VFMs) to build a rich geometric-semantic local map from streams of on-board sensor data, and 2) a unified neural network that jointly predicts traversability-relevant quantities in bird's eye view (BEV), including geometry, semantics, speed and cost. Our training strategy is entirely annotation-free and self-supervised, leveraging tasks such as map inpainting and inverse reinforcement learning (IRL) to learn both map representations and traversability. We perform a thorough ablation study and comparison to state-of-the-art approaches, and the results indicate that cost learning and auxiliary inpainting each contribute significantly to planning quality, and their combination is critical for achieving state-of-the-art performance in off-road path planning. We also design a simple risk adaptation mechanism to leverage our method's uncertainty estimates at deploy-time, and demonstrate that a combination of inpainting and risk estimation can result in 80% fewer navigation errors and 5% faster autonomous traversal speeds in real-world hardware experiments.

0.49

Bracketing Shows the Way: Adaptive Multi-Exposure Policy as Input for Robust Visual SLAM

cs.RO ·2026-10-23 ·Olivier Gamache, Anthony Lavertu, David-Alexandre Duclos, Jean-Michel Fortin, François Pomerleau +1 more
code/data
Abstract · 9 pages, 6 figures, submitted to 2027 IEEE International Conference on Robotics and Automation (ICRA 2027)

Rapid illumination transitions can disturb Visual SLAM (VSLAM) by saturating images before the Automatic Exposure (AE) policy adapts, removing the visual correspondences needed to maintain tracking. Our proposed methods mitigate this disturbance at two levels: B-Adapt acquires adaptive exposure brackets to cover the scene dynamic range while promoting photometric overlap, and BVSLAM directly uses the multi-exposure sequence as constraints in a factor graph state estimator without relying on High Dynamic Range (HDR) merging. We evaluate both methods against state-of-the-art approaches in nominal environments and during a repeated tunnel-exit transition at an emulated vehicle speed of 32 m s-1. In the tunnel, combining our B-Adapt with BVSLAM reduces the state estimation error by 78 % compared with AE images. These results suggest that bracketing efficiently improves visual pose estimation robustness to rapid illumination disturbances while retaining low error under nominal conditions. Our code is available at https://github.com/norlab-ulaval/bracketing-shows-the-way.

0.44

Stability Enhancement of Cascaded DC Systems via Source Output-Impedance Reshaping Using Model-Free Control

eess.SY ·2026-10-23 ·Wanrong Li, Shuhan Hu, Hengping Xiong, Runnan Wang, Huawei Yuan +2 more
published
Abstract · Accepted at the 2026 CPSS & IEEE International Symposium on Energy Storage and Conversion (ISESC)

Cascaded DC systems may become unstable through source-load impedance interactions even when their constituent converters are locally stable. This paper proposes a source-side model-free control (MFC) method that reuses bus-voltage feedback to reshape the source output impedance while preserving nominal reference-to-output dynamics. The controller uses only nominal source-loop information and requires neither a load-converter model nor load-side measurements. A current-controlled buck source supplying a regulated buck-resistor load that exhibits constant-power load (CPL) behaviour at low frequencies is used as the case study. Impedance analysis predicts that the proposed MFC increases the maximum stable load power by approximately 50%, thereby improving the stability margin of the cascaded system. Switched-model simulations confirm preservation of the nominal reference-to-output response and show that the MFC-controlled system remains stable at load levels for which the baseline system becomes unstable.

0.40

DexPort: A Hand-Held Skill Interface for Dexterous Hand Data Collection

cs.RO cs.AIcs.LG ·2026-10-23 ·Amber Xie, Jiawei Fu, Xiaofeng Guo, Carmelo Sferrazza, Guanya Shi +1 more
Abstract · Website: https://amberxie88.github.io/dexport/

Collecting high-quality demonstrations remains a central bottleneck for learning dexterous manipulation. Teleoperation records executable robot actions, but it requires a complete robot setup and places the burden of coordinating many finger joints on the operator. Human demonstrations are easier to collect, but their motions must be estimated and retargeted to the robot, and they do not directly provide robot actions. In this paper, we introduce DexPort, a portable data-collection interface for high-DoF robot hands. It uses the same hand hardware for collection and deployment, avoiding a hand embodiment gap, and records both commanded joint targets and measured joint states. Our key observation is that a small set of skills can support a wide range of dexterous tasks. During collection, the operator moves the hand-held interface to specify the robot wrist motion and selects or adjusts skills that coordinate the finger motion. This removes the need to control each joint independently while retaining behaviors such as precise grasping and in-hand manipulation. We collect demonstrations across several real-world tasks and in unstructured environments without a robot arm. We then use behavior cloning to train visuomotor policies directly on the recorded wrist motions and hand commands, without skill labels. We evaluate across 4 tasks: pick-and-place with diverse objects, screwing with in-hand tool use, drawer stowage which requires simultaneous grasp and pulling, and a bimanual scanning task. We collect over 2,500 demonstrations with our portable device, in both laboratory and in-the-wild settings. Across our tasks, we see an improvement from 11% to 64% when using recorded robot commands instead of states, and an improvement from 36% to 78% for task success when using in-the-wild data.

0.40

Iterative Behavioral Identification for Covert Misappropriation Attacks

eess.SY cs.CR ·2026-10-23 ·Victor Mattos, Amit Bhaya, Daniel Sadoc Menasche, Alan O. de Sá
published
Abstract · Presented at the 17th APCA International Conference on Automatic Control and Soft Computing (CONTROLO 2026), Coimbra, Portugal, September 9-…

We present a data-driven, three-phase framework for covert misappropriation attacks on closed-loop networked control systems. A man-in-the-middle (MITM) adversary, with access only to intercepted input-output data, incrementally identifies the plant's behavioral subspace while maintaining stealth under an infinity-norm intrusion-detection system (IDS). Phase 1 collects an initial rank-deficient Hankel matrix using a low-amplitude probing signal that satisfies an analytical stealth bound. Phase 2 adapts online rank-increasing experiment design to the closed-loop setting, using causal predictor-based masking to limit IDS residuals while the Hankel rank grows from 1 to the full behavioral rank n + mL in exactly n + mL - 1 steps. We show that the identified subspaces are strictly nested and that the projection error of any desired trajectory decreases monotonically, strictly so almost surely at each rank increase. Phase 3 employs the final Hankel matrix in a DeePC-style optimization to synthesize a covert attack that tracks an arbitrary reference while respecting the IDS constraint. Simulations on a DC motor with PI control demonstrate full-rank identification and stealthy tracking. Stealth is guaranteed analytically in Phases 1 and 3 and is numerically supported in Phase 2. The method is model-free and assumes noise-free LTI dynamics.

0.38

RAIWAY: A Large-Wheeled Bipedal Robot with Contact-Informed Blind Locomotion

cs.RO ·2026-10-23 ·Kyungmin Nam, Jiseong Lee, Donghyuk Choi, Jeongsoo Park, Woojin Kim +1 more
top venue
Abstract · 8 pages, 7 figures. Accepted to IEEE Robotics and Automation Letters (RA-L)

We present RAIWAY, an in-house wheeled bipedal robot with 20-inch wheels, together with a contact-informed control approach for rolling locomotion across irregular and discontinuous terrain. Such locomotion is fundamentally contact-dominated, governed by local wheel--terrain interaction that is impractical to instrument directly and depends on more than terrain geometry alone. The proposed framework exploits the contact information available in simulation, distilling multi-contact events into a latent representation. At deployment, the controller infers this representation online from proprioception, requiring no contact sensing. In real-world experiments, the proposed system traverses diverse environments including steep slopes, gravel beds, studded surfaces, and rocky terrain, and demonstrates continuous ascent on the tested outdoor staircases with riser heights up to 15~cm. On our platform, the proposed controller further outperforms baselines without the contact-informed mechanism across challenging indoor and outdoor settings. The video can be found at https://youtu.be/Ch0divLg64w.

0.38

Regret-Based Max-Pressure Traffic Signal Control under Uncertain Traffic States

eess.SY math.OC ·2026-10-23 ·Chaopeng Tan, Qiongdan Hu, Meng Wang
Abstract · 29 pages, 7 figures

Max-pressure (MP) traffic signal control selects phases using pressure scores computed from estimated traffic states, but these states are uncertain and their consequences depend on whether they alter phase ranking and how much pressure is lost. This paper develops a decision-oriented Regret-Based Max-Pressure framework that represents uncertainty through phase-selection regret, defined as the pressure loss of a candidate phase relative to the pressure-maximizing phase under each plausible latent state. We propose Stochastic Regret MP, which minimizes expected regret under a calibrated residual distribution, and Distributionally Robust (DR) Regret MP, which minimizes worst-case expected regret over a Wasserstein ambiguity set. Robustifying regret rather than absolute pressure makes the formulation invariant to state-dependent pressure components shared by all candidate phases, thereby focusing robustness on relative phase quality. We further derive an explicit queue-linear pressure-loss bound and establish strong queue stability when the net-service reserve exceeds the resulting loss coefficient. Numerical experiments show that both methods reduce normalized true-state pressure regret by about 70% relative to Point-estimate MP, while DR Regret MP yields lower network queues than direct robustification of absolute pressure. In microscopic SUMO experiments on a 28-intersection real-world network, Stochastic Regret MP and DR Regret MP reduce mean delay by 18.9% and 14.9%, respectively, and retain benefits under detector undercount and failure. The results show that representing uncertainty through its decision consequences can improve reliable traffic control under imperfect state information.

0.37

RAISE-Vision: An Automated Contact Angle Measurement Platform for Contact Lens Wettability Characterization

cs.RO ·2026-10-23 ·Erin Ng, Mohammad Nazeri, Julie S Liu, Tithi Choksi, Lyndon Jones +1 more
code/dataunusual
Abstract · 27 pages, 7 figures

Contact angle (CA) is a common parameter measured in the development of novel contact lens materials and surface coating technologies. Existing literature on in vitro CA of contact lenses via the sessile drop method is hindered in repeatability, reliability, and comparability due to inter-laboratory and inter-operator protocol variation. In this work, the Robotic Autonomous Imaging Surface Evaluator for vision applications (RAISE-Vision) is introduced as an automated platform for CA measurement on curved hydrogel surfaces, with specific application to in vitro contact lens wettability analysis throughout air exposure. Customized 3D printed hardware, such as lens mounts, camera holders and drying apparatuses, was designed to facilitate an automated repeated sessile drop workflow within a liquid handler. Collected images were directly fed to a custom open-source image processing script for CA determination on a curved surface that combines the elliptical fit and tangent methods to allow for variations in droplet shape. The CA of senofilcon A contact lenses measured using the RAISE-Vision platform increased from $62.6 \pm 5.5^\circ$ to $83.4 \pm 2.1^\circ$ over 10 sequential measurements in air and was able to replicate results obtained using an analogous manual protocol ($p = 0.470$). The custom image processing script was also separately demonstrated to generate comparable results to a commercially available CA determination software ($p = 0.343$). Aside from minimizing operator variability for CA measurement, RAISE-Vision provides a platform for standardized in vitro contact lens wettability analysis and understanding of surface wetting behaviour during air exposure.

0.36

Robust Offset-Free Lane-Keeping Control Framework for Vision-Only Automated Vehicles

cs.RO cs.CVeess.SY ·2026-10-23 ·Der-Hau Lee
Abstract · 13 pages, 10 figures

Achieving accurate, real-time lane-keeping control during high-speed cornering on large-curvature roads while ensuring robust vehicle stability is a critical challenge in autonomous driving. To address this challenge, this study developed a robust offset-free model predictive control (OF-MPC) framework for vision-based autonomous vehicle lane-keeping. The regulation problem in this framework is solved using several algorithms. Numerical simulation results indicated that the interpolation-tube-based constrained iterative linear quadratic regulator (itube-CILQR) algorithm outperformed solvers based on the standard Interior Point Optimizer (IPOPT) and Operator Splitting Quadratic Program (OSQP), achieving a mean computation time of less than 5 ms. The itube-CILQR algorithm was then used for vehicle control in vision-based lane-keeping experiments in a simulation environment to quantify the perceptual uncertainty of a visual perception model (multitask UNet). This quantification was performed using the OF-MPC framework for traffic scenes that were not encountered during model training. Overall, the developed OF-MPC framework and the proposed method for characterizing the perceptual uncertainty of a visual perception system in unseen environments can contribute to the future development of advanced autonomous driving systems.

0.36

HoloBoxing: A Simple Recipe for Humanoid Boxing with 7 Minutes of Single-GPU Training

cs.RO ·2026-10-23 ·Xihan Ma, Bo Zhang, Boxiong Li, Yuhan Bao, Ziyue Liu +1 more
Abstract

HoloBoxing presents a simple recipe for learning humanoid boxing by building on a pretrained whole-body controller. The recipe separates reusable motor skills from task decisions: a compact planner requests future body-point targets, while the frozen controller handles joint-level execution. A learned decoder maps 32-dimensional actions to structured target windows, and measured-state imitation initializes the planner before reinforcement learning. Task learning then refines interaction behavior while reusing the decoder and controller, avoiding the need to relearn whole-body control. This combination keeps task optimization focused on a compact upper policy and supplies an initial behavior before reward-driven exploration. Starting from that available controller, the recipe accounts for about seven minutes of additional training, including reused representation fitting and a single-GPU task run. This excludes controller pretraining, data preparation, evaluation, and search. Simulated boxing experiments and component controls evaluate the resulting behavior and the roles of imitation initialization and the target representation.

0.34

DiscoDemo: Discovering Efficient and Diverse Robot Demonstrations for Imitation Learning

cs.RO ·2026-10-23 ·Minho Park, Kinam Kim, Donghu Kim, Byungkun Lee, Dongyoon Hwang +4 more
code/data
Abstract

Behavior cloning for robotic manipulation is highly sensitive to the coverage of its training data, yet collecting diverse successful demonstrations remains costly. We study reinforcement learning (RL) not as a deployable controller, but as a data generator for downstream visuomotor imitation learning. Our key premise is that multiple successful solution modes cover a broader range of successful trajectories than repeatedly collecting a single narrow solution. We introduce DiscoDemo, which trains a privileged latent-conditioned RL generator with a parallel reverse curriculum and a success-gated diversity objective to discover diverse solutions while preserving task success. The trained generator produces multiple successful closed-loop behaviors, which are safety-filtered and exported as a static dataset for downstream imitation learning. Across four manipulation tasks, DiscoDemo produces broader successful-state coverage and improves downstream visuomotor policies in both simulation and the real world. Averaged across tasks, downstream success increases to 62% from 40% for the strongest baseline in simulation, and to 78% from 39% on the real robot, respectively. These results suggest that successful solution diversity improves imitation learning for a pretrained flow-matching VLA and remains complementary to corrective data collection. Project page: https://davian-robotics.github.io/DiscoDemo

0.33

Strengthening AC optimal power flow relaxation through offline bound tightening and linear approximations

eess.SY math.OC ·2026-10-23 ·Xiangxin An, Constance Crozier, Santanu S. Dey
Abstract · 15 pages, 4 figures

The alternating-current optimal power flow (ACOPF) problem is fundamental to power system operation and planning, but its nonconvexity makes globally optimal solutions of large-scale instances computationally challenging. Convex relaxations provide tractable lower bounds for assessing solution quality and supporting global optimization, although their effectiveness depends strongly on relaxation strength. This paper proposes an offline convexification and bound tightening framework for constructing strong linear programming (LP) relaxations for families of related ACOPF instances. Rather than repeating the tightening procedure for each instance, we compute variable bounds and second-order cone outer-approximation (SOC-OA) cuts offline using a demand-slack model and reuse them for subsequent instances with the same topology and admissible demand variations. Computational results on PGLib instances show that the fully offline approach achieves an average gap of $4.32\%$, compared with $3.80\%$ for the fully online approach, while achieving an average speedup of approximately $272\times$. The results demonstrate that offline convexification and bound tightening can recover most of the relaxation improvement of online tightening while significantly reducing the computational burden for repeated ACOPF evaluations.

0.32

Pinching Antenna Systems (PASS): Principles, Positioning and Opportunities

eess.SP ·2026-10-23 ·Tianwei Hou, Yaoyu Zhang, Jinhua Wang, Guangyu Li, Zongyi Li +3 more
survey
Abstract · 30 pages, 21 figures

Pinching-antenna systems (PASS) have garnered significant attention owing to their ability to enhance the capacity and coverage of wireless networks by guiding electromagnetic waves through waveguides before radiation. As such, PASS are regarded as promising enabling technologies for sixth-generation (6G) communication networks. In this paper, we present a comprehensive survey of the state of the art in PASS, covering operating principles, positioning methodologies, performance evaluation, resource allocation, and the application of artificial intelligence (AI) to PASS-enhanced wireless networks. We first introduce the fundamental principles of PASS from both materials and physics perspectives. Building on this foundation, we present positioning methodologies for both the user side and the base station side. We then evaluate the communication performance of PASS in both uplink and downlink scenarios under different blockage models. Furthermore, we systematically review existing designs for PASS-enhanced wireless networks from a performance optimization perspective. Finally, we survey research efforts that leverage AI to address challenges in dynamic and complex scenarios.

0.32

OWLS: Open-world Language-conditioned Speed Fields for Zero-Shot Rover Navigation

cs.RO ·2026-10-23 ·Chahyon Ku, Erica Tevere, Georgios Georgakis, Rudra Mukherjee, Bernadette Bucher
code/data
Abstract · 8 pages. 6 figures

Semantics can provide a proxy for physical traversability. Terrain that looks like loose sand or a rock field is likely to have similar physical properties. Open-vocabulary vision-language models (VLMs) enable this physical reasoning over terrain outside any fixed label set. We present Open-world Language-conditioned Speed Fields (OWLS), which plans directly over the dense feature map of a VLM. Prompted with a description of traversable terrain, the VLM scores every pixel of a single camera image; OWLS projects these scores onto the ground, adapts them to the platform with two morphological operations, and uses the result as the speed field of a minimum-time planner, so terrain appearance and obstacle clearance shape one trajectory. OWLS needs no task-specific training, no terrain annotation, no persistent map, and no exteroceptive sensor beyond the camera. We evaluate it on a physical rover indoors and off-road, on simulated Mars and Moon rovers over 200 procedurally generated terrains, and on real planetary imagery. Project page: https://chahyon-ku.github.io/owls/

0.31

Modeling Liquid-Cooled Data Centers in EnergyPlus

eess.SY cs.CEcs.DCcs.LG ·2026-10-23 ·Yun-Dam Ko, Dipashreya A. Sur, Cade Napier, Carlos D. Díaz-Marín, Rishee K. Jain
code/data
Abstract · 14 pages, 7 figures, 6 tables. Code: https://github.com/Urban-Informatics-Lab/energyplus-liquid-cooling

As AI workloads accelerate data center transition toward high-density racks with direct-to-chip (DTC) liquid cooling, existing simulation tools remain limited in evaluating whole-building and multi-year efficiency. To bridge this gap, we develop a native EnergyPlus object, ElectricEquipment:ITE:LiquidCooled, that models DTC liquid cooling for chilled-water and warm-water architectures as plant-loop components. It couples facility-water loops, rack-side secondary loops, coolant distribution units, and chip-junction temperatures within a single solver. Validation against an independent Modelica implementation across 21 operating conditions, including two full-year runs compared at one-minute resolution, yields loop-temperature RMSE at or below 0.021 K, well within the 0.1 K acceptance threshold. Our case study demonstrates that warm-water liquid cooling reduces annual PUE from 1.43 to 1.25, eliminates on-site water use, and cuts total facility energy by 37.9% relative to an air-cooled baseline. By embedding liquid-cooling logic into EnergyPlus, our work provides whole-building, long-term energy modeling for high-density AI infrastructure within an open-source framework.

0.31

Graph Spectral Series and Its Signatures

eess.SP ·2026-10-23 ·Feng Ji, Jingyang Dai, Apostolos Evangelidis, Xingchao Jian, Marco Rauscher +1 more
Abstract

Bandlimitedness is a central modeling assumption in graph signal processing (GSP), but many graph signals exhibit substantial energy throughout the graph-frequency domain. We introduce a new framework for analyzing such non-bandlimited signals by treating their graph Fourier spectra as ordered sequences and representing them with path signatures. Specifically, motivated by the observation that non-bandlimited graph spectra often display rough, time-series like behavior, we introduce the notion of graph spectral series and develop a signature-based representation that captures global spectral structure without frequency truncation. We establish conditions under which the proposed representation is continuous and characterize its information-preserving properties. We further provide an asymptotic interpretation of graph spectral series for sequences of expanding graphs. Experiments on real-world trend-prediction and change-detection tasks show that signature features can improve performance over representations based directly on the original graph signals. The proposed approach provides a novel principled signature-based tool for analyzing non-bandlimited graph signals.

0.31

QoE-Driven Hybrid Resource Allocation for Multi-User OFDMA Semantic Communications

eess.SP ·2026-10-23 ·Xiaotian Xu, Zhifeng Tang, Guangchen Wang, Nan Yang, Xiangyun Zhou +2 more
Abstract · 6 pages, 8 figures

In this paper, we develop a quality-of-experience (QoE)-driven hybrid resource allocation framework for multi-user downlink orthogonal frequency-division multiple access semantic communication systems, where the hybrid framework jointly handles continuous and discrete decision variables. Specifically, we formulate a mixed continuous-discrete optimization problem to jointly optimize transmit power, subcarrier allocation, and semantic decisions under coupled power and spectrum constraints. To better characterize semantic service quality, we propose a preference-inverse multiplicative exponential QoE that is bounded and non-compensatory, capturing the intrinsic ``both-required'' property of semantic rate and similarity. To solve the optimization problem, we develop a projection-based multi-agent deep deterministic policy gradient framework that enforces feasibility for continuous and discrete actions. Furthermore, we integrate Gumbel-softmax relaxation with an adaptive performance-driven temperature annealing scheme to enhance the learning of discrete semantic decisions. Simulation results demonstrate that our proposed framework achieves superior convergence, greater stability, and higher QoE compared to the conventional QoE design and penalty-based or fixed-temperature baselines, highlighting the benefits of our QoE design and the effectiveness of hybrid decision learning in semantic communications.

0.30

Decision Timescales in Max-Pressure Traffic Signal Control: Theory, Methodology, and Validation

eess.SY ·2026-10-23 ·Chaopeng Tan
benchmark
Abstract · 31 pages, 11 figures

Max-pressure (MP) control adapts signal phases to prevailing traffic states, but its decision timescale is typically fixed a priori. This paper treats the decision timescale as an explicit dimension of MP control. We first characterize the trade-off between responsiveness and switching efficiency by quantifying how an equivalent switching loss reduces service capacity over a finite decision interval. This yields a load-dependent critical timescale determined by switching loss and normalized traffic load, together with a performance-oriented operating range above the stability boundary. Building on this result, we propose Variable-Timescale Max-Pressure (VT-MP), in which each intersection selects its decision interval according to local operating conditions, allowing the control timescale to vary across time and space. We establish strong stability of the resulting asynchronous heterogeneous controller under uniformly admissible local loads and bounded decision intervals, and introduce a downstream receiving-supply correction for finite-storage implementation. Three experiments provide mechanism validation and network-level evaluation. A single-intersection experiment confirms that the preferred decision interval increases systematically with traffic load and follows the derived characteristic timescales. In a store-and-forward network, heterogeneous VT-MP reduces total queueing time by 31.0% relative to the best fixed-interval benchmark. In a microscopic 5x5 SUMO network, it reduces mean trip delay by 24.3%, from 213.3 s to 161.5 s, and peak network accumulation by 12.3%. These findings show that the MP decision timescale should be treated as a traffic-state-dependent control variable rather than a fixed implementation parameter.

0.30

EMG Tokenization Preserves Movement Information at a Fraction of the Bitrate

eess.SP cs.HCcs.LG ·2026-10-23 ·Raul C. Sîmpetru, Amin Olamazadeh, Devon Rohlf, Vlad Cnejevici, Alessandro Del Vecchio
benchmark
Abstract · 11 pages, 7 figures; 12-page supplement included. Supplementary Data 1-9 (CSV) are provided as ancillary files

Wireless electromyography (EMG) interfaces must balance electrode coverage against transmission bandwidth. Here, we show that compact EMG tokens substantially reduce payload while retaining information for three-dimensional hand-pose decoding. We evaluate two causal tokenizers fitted without pose labels: an encoder pretrained to reconstruct EMG spectra and one based on SPAR, an engineered spectral-spatial representation of neighbouring electrodes. In personalized, within-session experiments with 320-channel EMG from nine participants, we compare tokenized and continuous features using matched decoder architectures and training procedures. Averaged across decoders, tokenization changed mean error by +0.01 mm for SPAR (95% confidence interval: -0.12 to 0.14) and +0.45 mm for the learned representation (0.10 to 0.82). Both increases met a 1 mm non-inferiority criterion. Stored token arrays reduced payloads 41-fold relative to raw EMG and 36-fold relative to continuous SPAR features. On sampled windows of the public emg2pose benchmark's test partition, absolute differences in mean error between learned tokens and continuous features were below 0.05 mm across 176 participants and in the subset of 20 participants unseen during training. Electrode coverage, calibration, and unseen recordings affected SPAR decoding more than quantization did. In a computational benchmark, optimized SPAR generated tokens for 63 channels within 20 ms on a Cortex-M55 microcontroller. These findings support tokenization as a compact interface between wearable sensing and movement decoding, allowing decoders to be updated while the wearable encoder and transmitted representation remain fixed.

0.29

Delay Randomization Widens the Delay Margin of Learned Humanoid Motion Trackers

cs.RO ·2026-10-23 ·Shaoqin Li, Yi Sun
Abstract · 8 pages, 1 figure

Learned humanoid controllers are usually trained in one simulator and then checked in a second one before they go near hardware. We ask what that check says about command delay. We train motion-tracking policies for a 24-joint humanoid on five motions taken from monocular video and test them in an independent CPU MuJoCo harness, which we first verify against the training model. Every policy completes its motion in at least 26 of 30 episodes under the training randomization. With 40 ms of constant command delay, however, survival drops to 15% on average (0% - 45%), and the delay margin is only 20 - 30 ms. Randomizing the delay during training (0 - 50 ms) lifts survival at 40 ms to 95% on average (75% - 100%) and moves the margin to 50 - 70 ms. A control arm trained for the same number of extra iterations without delay does not improve, so the gain comes from the randomization itself. This holds for all five motions with three training seeds each, and a policy trained in PhysX behaves the same way. The price is slightly looser tracking (0.02 - 0.06 rad on four motions) and, for the most dynamic motion, less robustness to other perturbations. We also show that the first sim-to-sim gap we saw came from the harness, a missing floor and a wrong timestep, and not from the physics engines. The 0 - 20 ms delay ranges common in locomotion training lie inside the margin an unrandomized tracker already has. The benefit comes from randomizing beyond it, so the range should be set from the robot's measured latency, and delay margin belongs next to other sim-to-sim metrics.

0.29

Trust-Region Benders for Scalable Multiperiod Expansion Planning for Transmission and Storage

eess.SY ·2026-10-23 ·Kevin Wu, Rabab Haider, Pascal Van Hentenryck
published
Abstract · Presented at PowerUp 2026, September 9-11, 2026, Boulder, Colorado

Transmission expansion planning (TEP) seeks to identify cost-effective investment strategies that enable power systems to reliably meet future demand, while accounting for the location, timing, and operational impacts of infrastructure upgrades. As demand grows and renewable penetration increases, there is growing interest in jointly modeling energy storage investments to enhance operational flexibility. This work presents a multiperiod formulation that co-optimizes transmission reconductoring and energy storage installation across multiple planning periods. To solve the resulting large-scale problem efficiently, a Benders decomposition with per-period trust-region stabilization and a relaxation-based warm-start procedure is proposed. The approach is demonstrated on a 2,000-bus Texas system, producing an investment plan spanning 2030-2045 with optimality gap below 1.0%, and is compared against a rolling-horizon approach to highlight the benefits of multiperiod coordination.

0.29

Conditional Neural Processes for Self-Supervised Goal Discovery in Action Imitation

cs.RO cs.AI ·2026-10-23 ·Marco Gabriele Fedozzi, Yukie Nagai, Francesco Rea, Alessandra Sciutti
published
Abstract · Presented at the poster session of the Workshop on World Models and Predictive Coding in Cognitive Robotics @ IROS2023 (https://world-model.…

Inspired by infants' ability to learn novel tasks by imitating others, in this work we evaluate to what extent Conditional Neural Processes can autonomously learn representations of actions that are informative about their underlying goal. We show that the addition of an attention mechanism in the hidden space fosters the emergence of robust goal representations early on in the observation of an action sequence, making it a promising model for a robotic self-supervised general learner.

0.29

Monte Carlo-Based Indoor UAV Localization in Dynamic Environments Using Only Four Onboard Range Sensors

cs.RO ·2026-10-23 ·Mohammad Reza Ranjbar Divkoti, A. Pedro Aguiar
Abstract · 8 pages, 8 figures

Indoor UAV localization with minimal onboard sensing remains challenging, particularly in dynamic environments. We propose augmented filtered Monte Carlo localization (AF-MCL) for 2D localization using only four single-point range sensors and measured heading. AF-MCL rejects range measurements inconsistent with the known static map and adaptively injects particles to recover from localization divergence and kidnapping. Beam classification is performed only at the previous position estimate, making its ray-casting cost independent of the particle count. The particle distribution additionally provides an online measure of localization uncertainty. Real-world experiments in a 2.10x1.63 m environment yield mean position errors of 6.4-7.7 cm under static and dynamic conditions. Simulations in 6x6 m and 10x10 m environments yield 4.5-5.2 cm mean errors; under dynamic obstacles, standard MCL reaches 12.3 cm and reliable MCL 7.7 cm. The results show that robust indoor localization under dynamic obstacles is achievable with only four single-point range measurements and heading.

0.29

Do LLM-derived text features improve electricity price forecasts beyond document activity? Evidence from Australia's National Electricity Market

eess.SY ·2026-10-23 ·Fanxing Ning, Jianguo Zhu
code/data
Abstract · 27 pages, 4 figures, 12 tables. Code and data: https://github.com/ningfanxing-sys/nem-llm-text-activity

Large language models (LLMs) can turn market notices and news into features for electricity price forecasting. Such features also record when documents arrive and whether their extraction failed, so a gain measured against a model without text can reflect document activity as well as content. We add the counts of the same documents and of their failed extractions to a no-text LightGBM model and measure each LLM representation against this activity reference, which needs no additional LLM requests. With gpt-6-luna as the extractor, we forecast Australian electricity prices from operator notices (5-30 minutes ahead) and WattClarity news (day-ahead) after the extractor's documented knowledge cutoff, averaging forecasts over three training seeds. For news, the counts alone lower mean absolute error (MAE) by 3.0%, most of the 3.6% reduction that the better LLM representation achieves relative to the no-text model; its increment beyond the counts is not significant. For notices, two of the three LLM representations lower MAE beyond the counts, by 0.14-0.21% and 0.45-0.52%, whether the counts are cumulative or cover recent windows, whereas the third shows no significant reduction. Time-shifted placebos give mixed evidence on whether these gains require text aligned with the target. Benchmarks using the operator's pre-dispatch forecasts (notices) or price history alone (news) are more accurate than all these text models. Gains against the no-text model and the activity reference differ in sign at individual seeds and, for notices with cumulative counts, in seed-averaged forecasts. Text gains should be reported against both references.

0.29

Facultative Swarming: A Capability, Not a Category, for Multi-Robot Systems

cs.RO ·2026-10-23 ·Ning Zhou, Edmund R. Hunt
Abstract

Collective robotics is split by a long-standing but increasingly artificial boundary between centralised multi-robot systems that prioritise optimality and decentralised swarms that prioritise robustness and adaptivity. We propose a unified control framework, shifting from `obligate' swarms---systems constrained to pure self-organisation---to facultatively swarming systems: hybrid architectures that operate as centrally planned systems under permissive conditions yet degrade gracefully into decentralised swarming under environmental stress. By reframing swarming as a dynamic capability, rather than a rigid system category, we outline a pathway towards collective robotic systems that benefit from centralised optimality when conditions permit and from the flexibility of self-organisation when they do not.

0.28

Model-Free Output-Admittance Reshaping for Stability Enhancement of Grid-Connected Inverters

eess.SY ·2026-10-23 ·Runnan Wang, Wanrong Li, Huawei Yuan, Sinan Li
published
Abstract · Accepted at the 2026 IEEE PES Australasian Universities Power Engineering Conference (AUPEC)

Grid-connected inverters can be unstable when their current-control loops interact with an uncertain external network. Returning the existing controller to improve damping also alters current tracking because tracking dynamics and output admittance are coupled. This paper adds a local model-free control (MFC) loop around the existing current controller to reduce output admittance while closely preserving nominal current-tracking dynamics. The added loop uses local inverter parameters without an external grid or neighbouring converter model. A Norton formulation separates reference tracking from the current response to terminal-voltage disturbances and establishes a tracking-preservation condition for admittance reshaping. Reducing the disturbance response makes the inverter behave more like an ideal controlled current source. Averaged-model simulations show reductions exceeding 65% in the magnitudes of all four dq output-admittance elements over 10-200 Hz, with closely preserved current-tracking dynamics. Generalised Nyquist analysis predicts small-signal stability for all sampled grid-impedance cases with MFC, compared with 90.0% without MFC.

0.28

Terra-Fi: Decoupling Sensing and Wireless Depths for Sustainable Deep Underground Monitoring

eess.SY cs.ETcs.NI ·2026-10-23 ·Kaiqiang Lin, Mohamed-Slim Alouini
Abstract · 8 pages, 4 figures, 2 tables, submitted to IEEE IoTM

Wireless underground sensor networks (WUSNs) enable in-situ monitoring of subsurface processes and buried assets to deliver substantial social and economic benefits across diverse verticals. However, meter-scale underground monitoring remains challenging since conventional WUSNs place the entire underground device at the target sensing depth, where severe soil-induced attenuation significantly degrades communication reliability and makes wireless energy transfer (WET) inefficient and even impractical. To address this limitation, we propose Terra-Fi, a novel shallow-hub and deep-sensing architecture that decouples the sensing depth from the wireless communication and energy-replenishment depth. In Terra-Fi, a shallow wireless hub is deployed at a depth favorable for communication, WET, and maintenance, while application-specific sensing probes are extended to meter-scale target layers through power-and-signal feeders. On this basis, we first examine the depth limitations of underground communication and WET and compare representative monitoring architectures. Then, we present its system architecture, operation flow, and potential benefits. Through a precision agriculture case study, our numerical results reveal that Terra-Fi establishes more robust wireless connectivity and a longer operational lifetime than conventional fully buried WUSN architectures, particularly at greater sensing depths and higher volumetric water content. Moreover, its performance gain can be enhanced through time allocation and transmission-configuration optimization. Finally, we pinpoint practical challenges and future research directions for deploying Terra-Fi into WUSN systems.

0.28

RynnUnify: Towards Unified Embodied Brain

cs.RO ·2026-10-23 ·Minghao Zhu, Bohan Hou, Kun Xiang, Zixuan Wang, Yinzhou Tang +18 more
community pick
Abstract

We introduce RynnUnify, a 16B unified embodied model that combines strong embodied understanding with long-horizon visual planning through interleaved video--language reasoning. With capabilities spanning spatial reasoning, physical prediction, and grounding, RynnUnify connects an understanding of the physical world with visual anticipation of how it evolves across successive subtasks. Building on these capabilities, we further develop an agentic extension through reinforcement learning to selectively use video predictions for reasoning about unobserved states and interaction outcomes. We also develop a world-action extension of the full RynnUnify model, translating its understanding and prediction capabilities into robot control. To evaluate these capabilities, we introduce Rynn-VP and Rynn-GEU for long-horizon visual planning and anticipatory reasoning. Experiments show that RynnUnify achieves the highest average embodied understanding score among the evaluated unified multimodal models and improves the Rynn-VP overall score by 9.75 points over Cosmos 3 Nano. Its agentic extension improves the Rynn-GEU average by 6.0 points over direct reasoning, while the world-action extension achieves 51.5\% success on 20 selected composite RoboCasa365 tasks and 46.7\% average success across three real-robot tasks. Together, these results establish RynnUnify as a shared foundation for embodied understanding, visual planning, and robot control, taking a step toward a unified embodied brain.

0.27

Low-Complexity Wideband Channel Estimation for ISAC in Analog Millimeter-Wave Systems

eess.SP ·2026-10-23 ·Hamza Touati, Raffaele Viterbo, Mattia Brambilla, Sandra Roger, Carmen Botella-Mascarell
published
Abstract · Accepted for publication in the 2026 IEEE 1st Annual Integrated Sensing and Communication Conference (ISAC)

Accurate wideband channel estimation is a key enabler for millimeter-wave (mmWave) integrated sensing and communication (ISAC) systems. In the narrowband scenario, the Transformed Spatial Domain Channel Estimation (TSDCE) method was previously shown to provide fast and accurate angle estimation with substantially lower complexity than conventional compressed-sensing approaches based on orthogonal matching pursuit (OMP). However, wideband ISAC scenarios additionally require delay estimation, making joint angle-delay sparse recovery computationally expensive due to the multidimensional dictionary search. In this paper, we propose a wideband extension of TSDCE, namely TSDCE-TD, which incorporates time-domain support and decouples angle and delay estimation. Two variations are proposed for delay estimation: TSDCE-TD (Dict), which uses a one-dimensional delay dictionary, and TSDCE-TD (FFT), which applies Fast Fourier Transform (FFT) zero-padding interpolation. Path estimates are iteratively refined through successive interference cancellation and alternating angle-delay updates. Simulation results show that the proposed methods outperform OMP-based baselines under practical dictionary sizes, while achieving comparable performance at high-resolution settings with substantially lower computational complexity. Moreover, the angle-delay estimates obtained using the proposed approach enable accurate single-anchor indoor localization in a ray-traced environment, achieving sub-20 cm positioning errors and satisfying practical indoor positioning requirements.

0.27

DynaSlots: Slot-Centric 3D Dynamics Pretraining for Robot Manipulation

cs.RO ·2026-10-23 ·Meng Li, Lijiang Chen, Qi Zhao, Jun Jiang, Longhao Zou +2 more
code/data
Abstract

Robot manipulation involves localized changes: picking up a cup moves the cup while leaving the surrounding scene largely unchanged. Yet existing 3D visual pretraining methods, including recent dynamics-aware variants, model state transitions at the scene level, compressing the motion of everything in a scene into a single global latent action. This conflates semantically distinct transitions in scenes with multiple moving parts, obscures dependencies between changing regions, and leaves the learned representation vulnerable to task-irrelevant scene content. We introduce DynaSlots, an action-free, label-free pretraining framework that decomposes each point cloud frame into a small set of temporally consistent slots and models dynamics at this per-slot level rather than the scene level. A temporal slot propagation module promotes consistent slot assignments across frames; an inverse dynamics model then extracts one latent action per slot from feature differences, and a diffusion-based forward dynamics model, paired with a self-supervised interaction head, predicts each slot's future state while learning pairwise proximity-and-motion relations. Experiments on Meta-World, Adroit, and a RealMan RM65-B robot demonstrate that DynaSlots outperforms prior scene-level baselines. It achieves superior success rates in simulation and exhibits strong robustness to scene clutter and unseen objects in real-world manipulation. Code is available at \href{https://zhucezhanghao12345-source.github.io/DynaSlots-Anonymous/}{https://zhucezhanghao12345-source.github.io/DynaSlots-Anonymous/}.

0.27

Solid State Transformer - Next Generation MV Interface for Native DC loads

eess.SY ·2026-10-23 ·Steffan Hansen, Marco Liserre, Mohamed Moharram
Abstract · 8 pages, 5 figures

About PowerNexa PowerNexa is the project name for a start- up initiative driven by Marco Liserre and Steffan Hansen with the ambition to develop the next generation power grid interface. Battery integration and DC loads Ongoing electrification is fundamentally changing how electrical energy is produced and consumed. DC loads are becoming an integral part of the infrastructure while battery integration is increasingly needed to manage peak demand and address grid capacity constraints. The next generation grid interface PowerNexa aims to combine a fully controllable Medium Voltage grid interface with native battery integration and multiple DC voltage levels, based on an architecture designed for high efficiency, modularity and redundancy.

0.27

A SysML v2 Domain Library for ROS2 Robot Architectures

eess.SY ·2026-10-23 ·Sai Sandeep Damera, John S. Baras
Abstract

Systems engineers who model a ROS2 robot in SysML v2 must first model ROS2 itself. Its nodes, communication endpoints, message types, lifecycle states, quality-of-service policies, and coordinate frames have no ready definitions in the language, so each project reconstructs them. We present ros2-sysmlv2, a native SysML v2 domain library that supplies these definitions for reuse. A robot model imports the library and specializes its definitions to declare its own nodes, ports, connections, frames, and parameters. The library maps ROS2 concepts to SysML v2 definitions and usages and organizes them in Primitives, Infrastructure, and Application layers. We check its declarations for parity with the ROS2 Jazzy sources they model. Because every ROS2 element of a model specializes a library definition, a tool can select those elements by type. We state what this selection is and use it for a round trip. A ROS2 package is generated from the selection, and a monitor generated with the package compares the running graph with the same selection and reports every departure against the model. The generation-gap workflow derives topic endpoints and parameter declarations while preserving developer-owned node algorithms. In a case study on a generated ground robot, the generated conformance monitor reported each of the three departures that we introduced in the implementation. The library is distributed as a versioned package that any SysML v2 model can import.

0.27

Scalable Reservoir Computing for Disturbance Localization and Identification in Power Grids

eess.SY ·2026-10-23 ·Erin Obermayer, Amir Sajadi, Bri-Mathias Hodge, Juan G. Restrepo
Abstract · 11 pages, 9 figures, submitted to IEEE: transactions on power systems

In this paper, we present a novel method for the localization and temporal waveform reconstruction of disturbances in complex power grids using observations of system state variables. Although various methods exist for detecting these events, accurately identifying and responding to them remains a significant challenge due to the grid's size, speed, complexity, and inherently dynamic behavior. We formulate a method for disturbance localization and reconstruction based on reservoir computing, a recurrent neural network architecture. We show that our method is able to recover the time dependence of previously unobserved disturbances, is robust to missing information, and can operate online. The approach requires only observed system data and knowledge of the network topology, without requiring a physical model of the system dynamics, and can recover simultaneous disturbances occurring at multiple locations. We implement a scalable parallelized reservoir architecture and demonstrate our algorithm on test IEEE cases and a reduced WECC system.

0.26

Optimization, Optimal Control, and Reinforcement Learning for Robotics

cs.RO ·2026-10-23 ·Yonghyeon Lee
Abstract · Preliminary lecture notes, version 1. Compiled and edited by Yonghyeon Lee through a human and AI collaborative writing process. No figures;…

These lecture notes provide an introduction to optimization, optimal control, and reinforcement learning for robotics, emphasizing their mathematical connections and practical relevance. They present established material compiled and edited by Yonghyeon Lee, rather than new research results. The notes were developed through a human and AI collaborative writing process, with AI assistance in drafting, organizing, and refining the exposition. This first version is a preliminary collection of notes, not yet a fully developed textbook, and currently contains no figures or graphs. The notes may be revised, corrected, and expanded in future versions.

0.26

MAST-VLA: Skill-Conditioned Adaptation with Trajectory-Affordance Priors for Multi-Skill Robot Manipulation

cs.RO ·2026-10-23 ·Jiachi Yang, Zhiyi Xu, Hongquan Zhang, Xin Tan, Zhizhong Zhang +2 more
published
Abstract · Accepted to CoRL 2026

Generalist robot manipulation requires a single vision-language-action (VLA) policy to execute diverse skills across scenes and instructions. However, current multi-skill VLA policies often leave two aspects of an instruction insufficiently modeled: which primitive skill to execute and where/how to contact and move objects. We propose MAST-VLA (Multi-skill Adaptation with Skill-conditioned experts and Trajectory-affordance priors), a dual-conditioning framework for robot manipulation. MAST-VLA introduces skill-conditioned LoRA Mixture-of-Experts (LoRA-MoE) adapters in the LLM backbone, where skill-query tokens and routing biases guide low-rank expert specialization, and converts a predicted affordance heatmap into a patch-level trajectory-affordance spatial prior for the action head. On the CALVIN ABC$\rightarrow$D long-horizon multi-skill cross-scene generalization setting, MAST-VLA achieves state-of-the-art performance, reaching an Avg. Len of 4.65 over five-subtask evaluation sequences. Real-robot multi-skill sequence experiments further validate its deployability under real perception and control noise.

0.26

An Evidence-Guided Framework for Tissue-Barrier Impedance Review: Deletion and Resistance Sensitivity

eess.SP ·2026-10-23 ·Leo Yang, Yingtong Zhou
Abstract · 5 pages, 3 figures, 3 tables

Reviewing tissue-barrier impedance requires connecting spectral anomalies with resistance estimates and experimental context. We present an evidence-guided framework that coordinates numerical checks and traceable actions for human review. In-house simulations assess the effect of removing flagged frequencies, public datasets test model adaptation, and local measurements demonstrate the review workflow. Removing flagged frequencies reduced resistance error under quadratic objectives, whereas robust fitting showed little average benefit from deletion and a mean error increase for endpoint-pair perturbations. High errors clustered around a few fitted circuit parameter sets, highlighting the value of examining parameter identifiability. Under the tested noise and perturbation conditions, these findings support robust full-spectrum estimation, flag-guided inspection, and matched sensitivity checks.

0.26

Resilience in Wireless: A Lyapunov-Constrained Integrated Sensing, Communication, and Control Approach

eess.SP ·2026-10-23 ·Yifan Ge, Ningyan Guo, Yuanhao Cui, Dingyou Ma, Zhiyong Feng +1 more
Abstract · 13 pages

Traditional wireless networks mainly use reliability metrics to evaluate network stability. However, reliable data delivery alone cannot fully characterize network stability under variations in wireless environments and physical tasks. Wireless network resilience provides a broader perspective by describing the ability to maintain acceptable service and recover from disruptions. We develop an integrated sensing, communication, and control (ISCC) framework that incorporates communication reliability and sensing error into a state-dependent Lyapunov constraint for adaptive BS beamforming. The resulting transmit-power minimization problem is non-convex and solved via semidefinite relaxation and successive convex approximation with rank-one recovery. We establish conditional mean-square ultimate boundedness, while simulations demonstrate improved resilience and faster recovery over a static baseline.

0.25

ViT-H: Generative Video-To-Humanoid Control with Motion Tracking Feedback

cs.RO cs.CV ·2026-10-23 ·Zimo He, Tianyi Bian, Zizhong Yang, Hongjie Li, Nan Jiang +5 more
Abstract · The first four authors contributed equally. Project website: https://awfuact.github.io/vit-h/

Learning humanoid control from monocular human videos offers a scalable way to acquire diverse whole-body behaviors. Yet, existing reconstruction-retargeting-tracking pipelines are limited in system-level scalability by separately optimized intermediate stages and cannot directly optimize reference motion executability using downstream tracking feedback. We present ViT-H, a video-to-humanoid control framework following a generation-and-tracking paradigm. ViT-H directly generates humanoid reference motions from videos with a diffusion model, bypassing human motion reconstruction and retargeting, and uses motion-tracking outcomes to post-train the generator with direct preference optimization. We further introduce an autoregressive variant that enables real-time humanoid video teleoperation with sub-400 ms latency. Extensive experiments on the Unitree G1 demonstrate that ViT-H outperforms state-of-the-art reconstruction-retargeting-tracking pipelines in both motion quality and executability. We further show that ViT-H scales favorably with model size and with both pre-training and post-training data scale, while motion-tracking feedback consistently improves executability across different motion trackers.

0.25

Reducing 6G Energy Consumption: TDD-Aligned Drain Voltage Modulation for Power Amplifiers

eess.SP ·2026-10-23 ·Meizhen Zhuo, Gunter Wolff, Stephan ten Brink, Silvio Mandelli
Abstract · 6 pages, 5 figures. This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after wh…

The Power Amplifier (PA) of a base station operates inefficiently at partial load, where back-off from the peak output power wastes supply headroom. Adapting the PA drain voltage to the instantaneous signal recovers this loss but requires frequent switching, increasing hardware cost. We propose TDD-Aligned Drain Voltage Modulation (TDD-DVM), a drain voltage modulation scheme that instead selects the drain voltage level once per Time-Division Duplex (TDD) period based on the downlink resource load and switches the voltage during the uplink window. This is enabled by the authors' previous work on POLITE link adaptation, which spreads the transmit power over resources left unused at partial load and moves the PA into a regime where a lower drain voltage suffices. In a 3GPP Urban Macro system-level simulation across traffic loads, we compare TDD-DVM against baseline approaches under conventional and POLITE link adaptation. At equal delivered throughput, TDD-DVM with POLITE keeps the packet failure ratio of the full drain voltage while consuming up to 35% less power than it under conventional link adaptation and up to 7% less than it under POLITE, capturing most of the gain of an idealized reference at moderate loads and roughly half at the highest load.

0.25

Quadrotor Guidance for Tasks along Elliptical Boundary

eess.SY ·2026-10-23 ·Ravi Khandelwal, Ashwini Ratnoo
Abstract

This paper presents a guidance method that enables a quadrotor uncrewed aerial vehicle to perform inspection tasks along an elliptical boundary enclosing a structure of interest. The proposed guidance approach leverages a bifurcation theory-based framework that supports multiple inspection tasks, including convergence to a desired viewpoint, switching between viewpoints along the boundary, constant-speed elliptical circumnavigation, and circumnavigation with sectorial dwell-time constraint. Stability analysis of the proposed guidance dynamics establishes convergence of the generated reference trajectories to the desired behavior. The control parameters in the proposed method further allow the settling time of the reference trajectory to be specified for each task. The resulting reference velocity expressions are closed-form and algebraically simple, facilitating practical implementation. A qualitative comparison with existing guidance methods highlights the flexibility of the proposed approach in achieving different inspection tasks through a unified framework, while relying on simpler guidance computations. The proposed guidance method is validated through numerical simulations using a six-degree-of-freedom quadrotor model with model uncertainties, and a cascaded sliding-mode controller that tracks the generated reference trajectories under bounded disturbances.

0.24

Heterogeneous Space Data Centers: A Key Paradigm for Future Non-Terrestrial and In-Orbit Services

eess.SY cs.AI ·2026-10-23 ·Peng Hu, Mustafa Alhassan, Amjad Iqbal
hot direction
Abstract

Lower space-launch cost and advances in AI are driving space-based data centers (SDCs) that provide AI-centric computing and communication services to in-orbit and non-terrestrial assets. However, architectural principles, service models, and energy provisioning for such SDCs, particularly in low-Earth orbit (LEO), remain largely unexplored. We introduce heterogeneous SDCs composed of large and small nodes, differentiated by computing capability, size, and power budget, to support applications spanning ground, aerial, and in-orbit users. We define use cases, node sizing and placement strategies, and interconnection topologies integrating space-to-space and space-to-ground links. Focusing on dawn-dusk LEO configurations, we analyze solar power availability, visibility windows, and latency-aware coverage. Using baseline workloads for typical scenarios, we evaluate latency and service reliability. We quantify energy requirements of different SDC roles and deployment scales, highlighting trade-offs among node size, orbit selection, and service portfolio, and provide design guidance and future research directions for AI infrastructure in space.

0.24

HGWM: Hierarchical Graph-Guided World Model for Zero-Shot Object Navigation via Graph Matching

cs.RO cs.MA ·2026-10-23 ·I-Tak Ieong, Hao Tang
Abstract

Zero-shot object navigation requires an agent to reason about where a target may be found while retaining evidence from previous observations. We present HGWM, a hierarchical graph-guided world model that combines a category-conditioned room--anchor prior with persistent spatial-semantic memory. A coarse stage uses a vision-language model to score candidate headings, while a fine stage explicitly compares observed objects and relations with the goal prior. Phase-dependent fusion integrates these signals with a language-model recommendation to guide local navigation. HGWM achieves 62.6\% success rate (SR) and 31.5\% success weighted by path length (SPL) on HM3D v0.1, improving over WMNav by 4.5 and 0.3 percentage points, respectively. On MP3D, it achieves 46.8\% SR and 17.9\% SPL, with gains of 1.4 and 0.7 points. On HM3D v0.2, the full model reaches 74.4\% SR and 31.9\% SPL; removing individual components reduces SR by 0.9--2.3 points and SPL by 0.6--1.5 points. These results support the use of relational priors and persistent graph memory for target-directed search.

0.24

A Robust Control Architecture for Multirate Systems and their Interconnections

eess.SY math.OC ·2026-10-23 ·Julian Berberich
unusual
Abstract

We introduce a framework for analyzing and designing linear multirate control systems and their interconnections through the lens of robust control. The proposed approach models multirate systems as impulsive systems with multiple independent jump maps and corresponding jump-time sequences. We tackle analysis and design problems with stability, dissipativity, and robustness objectives, relying on a multi-clock-dependent Lyapunov function parameterization. Further, we develop a compositional approach for analyzing interconnections of multirate systems based on individual dissipativity properties. The theoretical conditions can be verified via linear matrix inequalities and they require weak assumptions on the sampling rates, allowing for asynchronous and time-varying sampling rates with or without separated time scales. The practicality of the framework is demonstrated with numerical examples.

0.24

Natural-Language to Nav2: Operational Policies for Costmap-Based Navigation

cs.RO ·2026-10-23 ·Jiyeong Chae, Hyunkyo Seo, Heonjae Lee, Kyung-Joon Park
code/data
Abstract · 23 pages, 16 figures, including supplementary material. Jiyeong Chae and Hyunkyo Seo contributed equally. Project page: https://a56239637-do…

Autonomous mobile robots support material transport in warehouses and industrial facilities, where navigation must adapt to changing operating conditions. Sensor observations reveal physical obstacles, but do not by themselves specify temporary access rules, route preferences, or exceptions known to operators. This paper presents a natural-language policy interface that brings this knowledge into ROS 2 Nav2. A local large language model converts instructions into JSON policies specifying destinations, spatial constraints, and temporal or event conditions. Deterministic checks govern mission acceptance, and three costmap adapters project accepted policies into the existing navigation stack without changing its planner or controller. On 150 researcher-authored English commands, four local models produce 107-142 correct accepted policies, compared with 44 for a rule-based parser, with 3-20 incorrect acceptances per model. Fault injection records rejection of eight endpoint/JSON faults and 29 validator-stage faults. Navigation experiments use correct task policies to evaluate keepouts, zone costs, person-dependent regions, and event-conditioned passage in simulation and on a TurtleBot3. For Tasks 1-3, mean travel times are 18.4-24.5% lower in simulation and 23.1-35.0% lower on the physical testbed than with the baseline lacking those policies. A native KeepoutFilter comparison and two-robot demonstrations examine basic keepout behavior and shared policy projection. The interface connects operator instructions to explicit navigation policies while retaining Nav2's planning and control components.

0.24

Incremental-Smoothing-Based ARAIM Protection Levels for Tightly Coupled GNSS/IMU Integration in Urban Canyons

cs.RO eess.SP ·2026-10-23 ·Jiahao Hu, Xiangru Wang, Weisong Wen, Qian Meng, Jiale Wang +1 more
Abstract · 9 figures;removed journal-template header

Protection levels (PLs) bound position errors under specified fault hypotheses and risk budgets, but advanced receiver autonomous integrity monitoring (ARAIM) is formulated for snapshot GNSS estimation rather than the correlated posterior of tightly coupled GNSS/IMU factor graph optimization (FGO). This paper formulates ARAIM solution separation and PL computation on the FGO posterior. Because current-epoch pseudorange factors enter only the current-state block of the fixed-lag information matrix, the fault-excluded position covariance, solution separation, and nominal-bias propagation follow exactly from the linearized current-state marginal, which re-tains correlations with velocity, IMU-bias, and clock states and with historical information. Woodbury updates with selected-output evaluation compute these quantities for fault-free, single-satellite, constellation-wide, and second-order hypotheses with-out repeated full-state inversion or graph reoptimization. Surrounding monitoring layers screen measurements, handle post-fit outliers, and determine PL validity and availability. In two road experiments, both nominal sequences achieve 100% horizontal and vertical empirical bounding over PL-valid epochs, with a maximum error-to-PL ratio of 0.96. Normalized errors show that the posterior covariance alone does not capture slowly varying ranging biases: a covariance-only bound computed from the same posteriors retains no resolvable margin, and strong temporal correlation limits the empirical evidence to consistency checks rather than validation of the integrity budget. On identical Experiment A posteriors, the proposed method reduces fault-hypothesis evaluation and PL solving time by 65.2% relative to direct full-state inversion with matching integrity decisions, and controlled fault injections demonstrate fail-safe PL invalidation while navigation continues.

0.24

Automated screening of Parkinson's Disease using Smooth Pursuit Eye Movements

eess.SP cs.LGq-bio.QM ·2026-10-23 ·J. Camilo Puerta-Acevedo, Maria F. Alcala-Durand, Julián D. Arias-Londoño, Elisa Luque-Buzo, Francisco Grandas-Pérez +1 more
Abstract · 23 pages, 8 figures, including supplementary material

Smooth pursuit eye movements are frequently altered in Parkinson's disease, but previous studies have relied mainly on conventional parameters that describe the relationship between gaze and target. This study investigates whether complexity and nonlinear dynamics measures of the gaze position sequences can support the automatic screening of Parkinson's disease. Gaze position sequences were recorded with a video eye tracker in 41 participants with early Parkinson's disease and 47 healthy controls during 12 smooth pursuit tasks. Four experimental frameworks and four embedding delay strategies were compared. The selected experimental setting combined baseline wander removal with a fixed embedding delay. After feature selection, a support vector machine was trained for each task and axis and their outputs were combined in a mixture of experts. The pipeline was developed in 70 participants and evaluated on an external test set of 18 participants. On the external test set, the pipeline obtained an area under the receiver operating characteristic curve of 0.86. The most informative features described less persistent and more irregular gaze position sequences in Parkinson's disease, mainly along the horizontal axis. These results support complexity and nonlinear dynamics measures of smooth pursuit as candidate markers for Parkinson's disease screening, pending validation in larger independent cohorts.

0.24

Foundation Models Meet Power Grid Topology for Cascade Prediction

eess.SY ·2026-10-23 ·Soumav Biswas, Mia Naeini
Abstract · 10 pages

Accurate early-stage prediction of cascading failure severity is essential for assessing the vulnerability of power transmission networks. However, existing data-driven approaches often rely on network-specific representations, limiting their generalizability across power system topologies. This work investigates the complementary capabilities of Large Language Models (LLMs) and Graph Neural Networks (GNNs) for cascading failure severity prediction. First, standalone LLM predictors are developed using a LoRA-fine-tuned LLaMA-3-8B-Instruct model with varying degrees of topological information. A hybrid Graph-LLM fusion framework is then proposed, integrating a Topology-Aware Graph Attention Transformer (TAGAT) with the LLM through confidence-gated fusion. TAGAT captures structural dependencies among transmission lines, while the LLM extracts representations from early-stage power-flow observations and aggregated cascade descriptors. The frameworks are evaluated on 30,000 simulated cascading failure scenarios for each of the IEEE 118-bus and IEEE 300-bus systems. Using only the first four cascade steps, the hybrid framework achieves weighted F1 scores of 0.94 and 0.92, respectively. In cross-system transfer from IEEE 118-bus to IEEE 300-bus, it achieves 0.72 zero-shot accuracy, improving to 0.83 with limited target-system calibration, outperforming standalone models and conventional baselines. These results show the potential of combining graph-based structural learning with foundation models to improve cascading failure severity prediction and cross-system transferability.

0.24

Active Reachability Learning for Humanoid Recoverability Monitoring

cs.RO eess.SY ·2026-10-23 ·Zeyuan Feng, Baljeet Singh, Sarthak Ranjeet Kaingade, Somil Bansal, Bryan Trinh
Abstract

Deploying humanoid task policies requires reliable runtime monitors that anticipate unrecoverable falls in time to trigger a fallback. We formulate this task as learning the stochastic, finite-horizon recoverability envelope of a given fallback policy. Synthesizing this monitor is challenging: existing methods assume training data naturally covers the recoverability envelope, yet acquiring such data remains a major bottleneck for contact-rich systems. Consequently, they recover only a fraction of the true envelope, leaving them prone to optimistic hallucinations for out-of-distribution (OOD) inputs. We resolve this via a unified framework that iterates between value learning and active exploration. Data exploration is reframed as a reach-avoid problem, where an uncertainty-guided explorer systematically uncovers unseen states while avoiding confidently unrecoverable regions. This loop is further supported by a combined Temporal Difference and Monte Carlo objective for accurate in-distribution value prediction and reliable uncertainty estimation. By explicitly treating the remaining high-uncertainty regions as unrecoverable, the monitor mitigates optimistic OOD hallucinations. Evaluated on stochastic benchmarks and a 29-DoF Unitree G1 humanoid, our approach drastically improves monitor coverage and accuracy and successfully transfers zero-shot to real-world hardware.

0.23

Do VLAs Know They Will Fail? Interpretable Success and Failure Patterns in Physical AI Systems

cs.RO cs.AIcs.LG ·2026-10-23 ·Christian A. Rogers, Qian Lin, Daniel S. Brown
Abstract · 12 pages (4 main pages + 8 pages appendix), 4 figures. Accepted at the Science of Physical AI Safety Workshop at the 2026 Conference on Robo…

Vision-Language-Action (VLA) models have emerged as powerful general-purpose robot policies, but their successes and failures remain difficult to diagnose and understand. In this work, we show that sparse autoencoder (SAE) representations of OpenVLA activations contain outcome-predictive information and can be used to predict task failure before the robot reaches an unrecoverable state. Moreover, to provide insights into what common behaviors may contribute to these outcomes, we identify individual SAE features whose strong activations are associated with specific stages of success or failure, and uncover recurring behavioral patterns by localizing temporal intervals where these features exhibit significant activation changes. We further develop a VLM-based interpretation pipeline to generate behavioral descriptions of these patterns and validate their correspondence to feature activations on held-out intervals.

0.23

Physical Layer Security in Chirp Permuted AFBM

eess.SP ·2026-10-23 ·Pedro Schuves Marodin, Henrique L. Senger, Gustavo P. Gonçalves, Hyeon Seok Rou, Bruno S. Chang +2 more
Abstract · Currently 6 pages, 6 figures

We propose a physical-layer security scheme for affine filter bank modulation (AFBM), a waveform proposed for sixth-generation (6G) integrated sensing and communications (ISAC). AFBM contains two discrete affine Fourier transforms, one for precoding and one for spreading, and the scheme permutes, with one secret key per transform, the chirp sequence that each transform places between two Fourier stages. A wrong key is then conjugated by a discrete Fourier transform into a dense matrix, the restoration of complex orthogonality fails, and the intrinsic filter-bank interference reaches the detector. Over a doubly-dispersive channel, an eavesdropper with perfect channel state information, co-located with the legitimate receiver and holding either key wrong, keeps a bit error rate near 0.5 and a rate below 0.09 bit/symbol, also under a blind carrier-phase attack. Compared with a single-permutation variant, the double scheme costs about 2 dB of peak-to-average power ratio and enlarges the search space from 10^61 to 10^93.

0.23

Impact of Synchronization Offsets and CSI Feedback Delays on Coherent Joint Transmission in Distributed MIMO-OFDM

eess.SP ·2026-10-23 ·Kumar Sai Bondada, Daniel J. Jakubisin, R. Michael Buehrer
Abstract

This paper studies the impact of synchronization offsets and CSI feedback delay on coherent joint transmission (CJT) in distributed MIMO-OFDM systems. Unlike conventional co-located MIMO, distributed CJT is highly sensitive to transmitter-dependent synchronization impairments because coherent combining requires phase, frequency, and timing alignment across geographically separated transmitters. Received signal models are derived for residual timing offsets, carrier frequency offsets (CFOs), sampling clock offsets (SCOs), phase noise, and channel variations accumulated during channel state information (CSI) feedback delay. The corresponding spectral efficiency (SE) expressions quantify the degradation in capacity due to loss in coherent-combining gain caused by residual phase rotations, inter-carrier interference (ICI), and precoder misalignment. The SE expressions are numerically evaluated to determine synchronization requirements for maintaining negligible SE degradation, yielding the following findings: a Gaussian-distributed start-time jitter with a standard deviation below 10\% of the sampling period results in negligible SE degradation, with the SE loss decreasing from 0.68\% at an SNR of 0 dB to 0.17\% at 40 dB. A normalized CFO standard deviation of 0.001 also results in negligible ICI-induced degradation, with the SE loss increasing from 0.05\% at an SNR of 0 dB to 2\% at 40 dB. At an SNR of 26 dB, limiting the SE loss to 5\% yields feasible operating points with clock-error standard deviations up to 50 ppb, Doppler-delay products up to 0.1, and phase-tracking error standard deviations up to $40^{\circ}$ at carrier frequency of 3.5 GHz with 30 kHz sub-carrier spacing with 15.36 MHz sampling rate.

0.23

Coupling Execution Horizon and Sampling Depth in Flow-Matching VLA Policies

cs.RO ·2026-10-23 ·Jiabing Yang, Yuan Xu, Yixiang Chen, Qisen Ma, Peiyan Li +11 more
code/data
Abstract

Flow-matching Vision-Language-Action (VLA) models generate expressive continuous trajectories for robotic manipulation, yet balancing real-time responsiveness with execution stability remains a persistent challenge. Standard deployment workflows typically treat numerical solver depth and execution horizon as decoupled choices. The horizon amortizes inference latency, while solver depth is kept high to preserve integration fidelity. In this paper, we reveal a fundamental compute-execution coupling in flow-matching VLAs governed by a systematic sign reversal. While coarse numerical integration degrades precision under rapid replanning, it markedly improves task performance under extended open-loop commitment. Geometrically, numerical truncation contracts lateral action dispersion around an invariant trajectory centroid, serving as an implicit regularizer that suppresses compounding open-loop drift. Across LIBERO, RoboTwin 2.0, and real-world Franka manipulation on representative VLA backbones including $π_0$, $π_{0.5}$, and GR00T, exploiting this joint coupling consistently achieves superior task success while drastically lowering inference latency, entirely training-free and without architectural modifications. These findings establish numerical sampling depth as an operational variance regulator, shifting its role from a static precision budget to an active deployment knob in robot learning. Project page: https://jiabingyang01.github.io/Horizon-Depth-VLA/

0.22

Channel orthogonalization via movable antennas

eess.SP ·2026-10-23 ·Heedong Do, Angel Lozano
Abstract

This paper investigates the orthogonalization of far-field multiuser channels through the optimal positioning of the antennas constituting an array. To solve the ensuing continuous geometry optimization problem, a graph-theoretic framework is introduced that maps pairwise user channel correlations to graph edges, translating channel orthogonalization into a discrete graph-covering problem on a complete graph. The minimum number of array antennas required for orthogonalization is established for 1D, 2D, and 3D array geometries, and a binary linear program is formulated that identifies the most compact array attaining such orthogonality. The graph framework is also shown to accommodate null-steering scenarios and multipath channels. Furthermore, a fundamental tradeoff is revealed whereby relaxing exact orthogonality to near-orthogonality reduces the antenna count at the expense of a less compact array, and a method is presented to enforce minimum antenna spacings without sacrificing orthogonality.

0.22

Classical, Machine Learning, and Hybrid Solvers for Partial Differential Equations: An Evidence-Weighted Review and Decision Framework

eess.SY ·2026-10-23 ·Yingchao Huang, Xin Wang, Shanshan Yao, Fanhua Zeng, Wei Peng
surveybenchmark
Abstract

Machine learning for partial differential equations raises a practical question: when are classical, learned, or hybrid solvers justified? This survey compares the three families in a decision framework organized by data availability, query volume, and certification needs. An amortization law, tested on parametric Darcy flow, shows that surrogates trained on solver-generated data break even only after many queries, beyond thirty thousand for errors below one percent. An evidence ladder, a benchmark reanalysis, and three case studies show that claims of speedup and generalization carry the weakest evidence. Solver choice should follow regime, query volume, and evidence quality, not method novelty.

0.22

Adaptive World-Models for Anomaly Recognition (AWARE): A Framework for Robot Dynamics Prediction and Introspection

cs.RO cs.AI ·2026-10-23 ·Luke Beddow, George Mavroghenis, Rodrigo Chacon Quesada, Kamil Dreczkowski, Cong Sun +3 more
Abstract

Unrecognized anomalies in critical infrastructure can rapidly cascade into catastrophic physical or operational failures. Current anomaly detection approaches often rely on direct access to the monitored system by utilizing onboard sensors, such as motor encoders, to detect anomalous states. However, legacy hardware often lacks these sensors, or strict security protocols may prohibit access altogether. We therefore present AWARE, a world model framework that is capable of anomaly detection using only external camera monitoring via system dynamics estimation and state prediction. Evaluated across 240 real-world robotic crane trajectories, including payload collisions and motor degradations, AWARE demonstrates a 26% improvement in anomaly classification over our strongest baseline under noisy, camera-only state estimation. Furthermore, the inferred dynamics parameter can be decoded to identify the source of faults, localizing the degrading motor with 88% accuracy.

0.22

Concurrent Space-Based Observer Placement and Scheduling for Cislunar Space Domain Awareness

eess.SY math.OC ·2026-10-23 ·Dominic K. Amato, Hang Woon Lee
Abstract · 41 pages

As interest in cislunar space grows, so does the need for space domain awareness (SDA) capabilities to safely support missions. To enhance cislunar SDA capabilities, the use of space-based observers has been proposed. Prior research has explored optimizing observer placement but has not considered scheduling limitations. We propose a mixed-integer linear programming formulation that concurrently optimizes the placement and scheduling of space-based observers for cislunar SDA. Our work considers three aspects of observer scheduling: sensor tasking, communication, and charging. We utilize a scenario-based approach that generalizes observer placement for unknown targets while applying scheduling constraints to each scenario. However, this problem quickly becomes difficult for commercial solvers to handle as the number of scenarios grows. As an alternative, a column generation approach is employed to introduce well-performing schedules into the problem for each orbital slot. A comparative analysis is performed across multiple instances to evaluate the performance of column generation against that of a commercial solver. A case study is then conducted in which the performance of the observers' placement from the concurrent formulation is tested against a set of validation target sets and compared with the performance of two baselines in which placement and scheduling are performed sequentially.

0.22

Enabling 3GPP-Driven Services Over Optical Transport Network

eess.SY ·2026-10-23 ·Renato Ambrosone, Ramon Casellas, Alessio Giorgetti, Riccardo Schips, Stefano Straullu +8 more
Abstract · 4 pages, 3 figures. ECOC 2025 demonstration paper

We demonstrate a multi-layer slicing architecture enabling deployment of 3GPP network slices as self-contained, connected functions atop optical transport slices. The system provides Optical Spectrum as a Service (OSaaS) by integrating disaggregated optical control, multi-tenant orchestration, and telemetry-based optimization via standard interfaces.

0.21

Type-based Over-the-Air Computation

eess.SP ·2026-10-23 ·Marc Martinez-Gost, Ana Pérez-Neira
publishedtheory
Abstract · Manuscript submitted for publication in IEEE Transactions on Wireless Communications

Over-the-air computation (OAC) exploits the superposition property of wireless channels to compute functions of distributed data, reducing the communication overhead associated with its multiple access. However, existing OAC frameworks are often designed around specific classes of functions, limiting their applicability to more general nonlinear computations. This paper investigates type-based multiple access (TBMA) as a general framework for OAC, enabling the receiver to recover a histogram of the distributed observations and compute a broad class of functions without modifying the underlying communication scheme. We formulate TBMA as a histogram estimation problem and propose two digital detectors. We prove that a low-complexity successive cancellation algorithm achieves the optimal maximum-likelihood (ML) performance. We derive analytical expressions for the mean squared error (MSE) and compare TBMA with conventional schemes in terms of computation accuracy, transmission duration, bandwidth and energy consumption. Numerical experiments validate the analytical results and demonstrate that, unlike existing OAC techniques, TBMA exhibits an exponential MSE decay with increasing signal-to-noise ratio (SNR) and constant transmit energy, independently of the function being computed. These findings position TBMA as a step toward more versatile OAC architectures, compatible with existing transmission schemes, and broadening the range of functions that can be computed directly over the wireless channel and enabling new possibilities for distributed computation and decision-making.

0.21

CoCarry: Contact-Support-Structured Retargeting for Dual-Humanoid Cooperative Carrying

cs.RO ·2026-10-23 ·Ziyu Lin, Shihang Du, Tianhang Wang, Xiaodong Liu, Runze Zhang +4 more
dataset
Abstract · 8 pages, 6 figures, 2 tables

Recent advances in large-scale human motion data, motion retargeting, and motion imitation have enabled increasingly capable whole-body control for individual humanoids. Cooperative carrying is more challenging because successful coordination requires not only plausible individual motions, but also consistent cross-agent and agent-object relations, including shared grasp, relative configuration, and whole-body support. We introduce CoCarry, a cooperative-carrying dataset containing 220 synchronized two-person sequences across 21 task schemes, five shared-object geometries, four terrain types, and approximately 61 minutes of motion. We further propose Contact-Support-Structured Retargeting, which starts from nominal individual motions and jointly refines both humanoids around the same prescribed object trajectory. Cooperation-aware object-local hand anchors preserve compatible four-hand grasp relations, while complementary upper object-interaction and lower support structures retain object-relative geometry and coordinated support. Across all 220 CoCarry sequences retargeted to two A3 humanoids, our method achieves 0.9995 contact preservation, reduces penetration duration by 67.0% and 71.0% relative to GMR and OmniRetarget, respectively, and maintains a 100% retargeting success rate. On CORE4D with Unitree G1 humanoids, it achieves 0.9906 contact preservation with zero measured sliding. We further validate the generated references through physics simulation and multiple real-world cooperative carrying experiments on two physical A3 humanoids, demonstrating their usability for downstream dual-humanoid control.

0.21

LogLook: Execution-Grounded Memory for Frontier-Model Robot Planners

cs.RO ·2026-10-23 ·Tianyi Jiang, Tianyao Tong, Zhenshan Bing
benchmarkposition
Abstract · 19 pages, 11 figures, 10 tables

Many manipulation tasks require a robot to remember what it has done and what it has seen. Frontier vision-language models can plan such tasks without task-specific training, but how such a planner should remember an episode is an open question. We argue that its memory must keep track of what the planner has done and learned, and must let the planner choose what to look at, since every frame costs tokens; a note rewritten at every step meets both poorly. We propose LogLook, whose execution-grounded memory lets the system keep the record, an execution log and the model's earlier notes, while the model decides where to look: frames are returned only where it expects a change, and demonstrations are skimmed before it inspects the parts it chooses. Measured with the benchmark's motion planner, LogLook matches the success of a note-only memory with 43% fewer tokens and less than half the images, and fails far fewer episodes through what it remembered or saw. With direct control, LogLook solves 93.5% of the RoboMME test set.

0.21

ORBIT: A Modular Multi-finger Robotic Hand Platform with Rotational Finger Reconfiguration

cs.RO ·2026-10-23 ·Benito Ribadeneira, Siqi Shang, Lillian Chin
Abstract · NA

Despite the popularity of multi-fingered robotic hands for dexterous manipulation, there is no systematic platform to compare different hand morphologies against one another. We present ORBIT (Orbital Reconfigurable Base-actuated Integrated Theta-motion Hand), a modular platform designed to enable principled exploration of multi-fingered hand morphology. The core contribution of ORBIT is a rotating knuckle mechanism, where independent fingers travel around a cylindrical palm via a belt drive. This enables ORBIT to perform quick recon- figuration (< 1 s) of finger layout without hand disassembly. ORBITs a robust system, with finger average angular velocity of 1.57 rad / s1 and a root finger force resistance of 50 N. Using a three fingers instantiation of ORBIT, we demonstrate an initial pilot grasping study, comparing anthropomorphic, isometric, and Barrett-style finger layouts across eleven objects of varied size, shape, and texture. We further demonstrate how ORBITs nonanthropomorphic orbital DOF enables the achievement of tasks that would be difficult to perform with fixed-knuckle designs, such as cube face reorientation, lid threading, and precision pinching of sub-centimeter objects. These results establish ORBIT as a viable platform for direct apples to apples comparison of multi-fingered hand morphologies for specific manipulation tasks

0.21

When Human Corrections Travel: Safety in Federated VLA Adaptation

cs.RO cs.LG ·2026-10-23 ·Vigneshwaran Subbaraju, Obaidullah Zaland, Monowar Bhuyan
position
Abstract · Accepted for presentation at SPAIS workshop in CoRL 2026

Vision-language-action (VLA) policies can adapt to local human intervention, while federated learning can combine distributed robot experience without pooling raw records. We examine their prospective intersection: sharing adaptations derived from diverse human corrections across tasks, environments, people, and robot embodiments. Corrections differ in modality, safety semantics(urgency, behavioral scope, etc.), and privacy sensitivity. A shared update that improves mean task success may suppress a rare safety intervention or produce an unsafe response on another robot. We argue for three requirements: preserve the context and uncertainty of multimodal corrections when deciding what to share; assess candidate updates in each receiving robot's closed loop before deployment; and evaluate leakage alongside retention of safety-critical evidence. Immediate, legible defensive behavior remains necessary during human interaction. We offer counterexamples and a falsifiable evaluation agenda, not experimental results.

0.21

On the Discernibility of Topology Changes in Networked LTI Systems

eess.SY ·2026-10-23 ·Giulio Fattore, Maria Elena Valcher
Abstract · Submitted as an invited paper to the Bicentennial Special Issue of the Journal of the Franklin Institute. Under review

This article investigates the discernibility of topology changes in networked linear time-invariant systems, with particular focus on edge disconnection. We introduce a new notion of discernibility motivated by a limitation of the definition commonly adopted in the literature. In consensus networks, topology changes are inherently indiscernible for all initial conditions belonging to the synchronization subspace. To overcome this issue, we first introduce the notion of dynamical discernibility of a topology change for elementary consensus networks. This notion explicitly disregards the intrinsic indiscernibility associated with states in the consensus subspace. We characterize this property for directed communication graphs and then specialize the resulting conditions to topology changes arising from either the disconnection of an edge or a variation in its weight. We then extend the analysis to networks of homogeneous agents by introducing discernibility modulo synchronization, which factors out the unavoidable indiscernibility associated with synchronized trajectories. We derive a general characterization showing how discernibility depends jointly on the topology change and on the dynamics of the individual agents. Finally, we specialize the analysis to undirected communication graphs and show that, under mild additional assumptions, discernibility modulo synchronization is equivalent to the dynamical discernibility of the corresponding topology change. This establishes a direct connection between the two notions and extends existing results on the discernibility of topology variations in dynamical networks. The resulting characterizations are then further specialized to the case of edge disconnection.

0.20

Neuromorphic Decision Dynamics for Deadlock-free Safe Navigation in Congested Environments

cs.RO cs.MA ·2026-10-23 ·Himani Sinhmar, Vaibhav Srivastava, Naomi Ehrich Leonard
Abstract

Safe navigation in congested environments requires both collision avoidance and coordinated passage. We introduce neuromorphic decision dynamics (NDD), a decentralized, event-based controller, that provides navigation in congested spaces with guarantees on safety, liveness, robustness, and scalability. NDD makes safe-and-live deconfliction a controllable, locally tunable emergent property of local agent interactions, without requiring communication, preassigned priorities, models of neighbors' objectives or future actions, or estimates of group size or global network properties. By design, NDD couples the dynamics of internal decision and attention states with changing physical states and locally observed relative motion. Each agent continuously modulates its speed along its prescribed path, responding as soon and as much as conflict resolution requires. Agents self-organize into disjoint, locally engaged groups that resolve conflicts concurrently. When any group's conflict becomes safety-critical, decision-action feedback destabilizes collective indecision, and coordinated conflict-resolution emerges that stabilizes a safe traversal order within the group. Order persists as conflicts clear, but groups may split or merge as agents autonomously adapt to new conflicts and those with resolved conflicts resume their nominal motion. We derive explicit conditions on sensing range, feedback gains, response time scales, and actuation limits that guarantee collision avoidance and traversal-order formation in finite time with no deadlock or livelock for finite, heterogeneous agent populations of arbitrary, changing size. Computational cost scales with the number of local interactions sensed by each agent. Large-scale simulations, comparisons with existing approaches, and hardware experiments demonstrate NDD in complex congested scenarios and validate its scalability.

0.20

State Estimation as In-context Tabular Prediction

eess.SY cs.LG ·2026-10-23 ·Ayushi Jolotia, Parikshit Pareek
Abstract · 3 pages, 5 figures

This letter poses state estimation (SE) as an in-context tabular prediction (ICTP) problem. The key idea is a tabular arrangement in which historical measurement vectors and corresponding state estimates serve as the context, and a new measurement vector forms the query for which we predict a state vector. We propose to use tabular foundation models (TFMs), pretrained on synthetic data, whose weights are never updated, so power system data is needed only as context. With frozen TFM weights we estimate the state of 14-, 30-, 33-, 70-, and 118-bus transmission and distribution networks, and do so where weighted least squares (WLS) fails or degrades, under low observability and with bad data. ICTP errors are 11 to 57$\times$ below the per-bus average of the context at full metering, and 1.3 to 4.0$\times$ below it with only 5% of measurements available. With measurements at six buses biased by 8$σ$, the ICTP angle error rises 1.8 to 5.0$\times$ over its bias-free value against 2.5 to 12.7$\times$ for WLS with bad-data detection. Without retraining, ICTP has lower voltage error than DSS$^2$ on both of its feeders, and lower than a retrained TAPINE on two of three reconfiguration sets.

0.20

From Tabletop to Walking Bot: Deploying a VLA Model on a Legged Manipulator Robot

cs.RO ·2026-10-23 ·Giovanni Minelli, Antonio Langella, Giovanni B. Dessy, Lorenzo Amatucci, Giulio Turrisi +2 more
Abstract

Vision-language-action (VLA) models are typically pretrained on broad, diverse datasets and subsequently fine-tuned on a comparatively small dataset tailored to a specific target robot. Such adaptation recipe is now widely established, achieving its most mature results on fixed-base arms due to their static base dynamics and consistent reference frame. Mounting the same arm on a quadruped extends the workspace to the entire floor, but fundamentally changes the embodiment: rather than controlling the arm in isolation, the policy must jointly coordinate base navigation, body posture, and end-effector motion. In this work, we study the practical demands of this embodiment change across data collection, model adaptation, and inference design to achieve closed-loop loco-manipulation on a Unitree Go2 with an AgileX Piper arm. We report the choices that made this real-robot deployment feasible, namely model variant, action horizon, and demonstration conventions, together with a study tailored to a pick-and-place task, separating data coverage from model capacity. We show that expanding demonstration coverage reduces held-out end-effector error by 30%, and incorporating secondary skills (such as pushing rather than grasping) introduces new behavioral modes without degrading baseline performance. Finally, closed-loop real-world experiments show how dataset composition shapes success and failure, validating the pipeline for more complex loco-manipulation.

0.19

RoboCode: Learning Reusable Action Primitives for Robot Manipulation

cs.RO cs.LG ·2026-10-23 ·Xianyue Peng, Devi Amarsaikhan, Iman Soltani
Abstract · 12 pages, 5 figures

Recent robot manipulation policies commonly predict continuous action chunks directly from observations, without explicitly modeling reusable structure in the action space. However, manipulation demonstrations often contain recurring short-horizon motion patterns that are shared across tasks. We introduce \textsc{RoboCode}, which learns these patterns through a shared soft-gated primitive action interface. Rather than assigning each action chunk to a single discrete code, \textsc{RoboCode} represents it as a soft composition of multiple learned primitives, providing a flexible intermediate representation of short-horizon actions. The primitive interface is learned from action chunks pooled across multiple tasks and then frozen while task-specific state encoders are trained to map current state inputs into the shared latent space. Experiments on five ManiSkill manipulation tasks address three questions. First, the learned primitive interface reconstructs action chunks with an average reconstruction MSE of 0.0187. Second, analysis of code activations reveals structured code reuse across tasks. Third, \textsc{RoboCode} achieves the highest or tied closed-loop success rate on all evaluated tasks, with an average success rate of 75.6%, compared with 68.8% for ACT and 41.6% for the larger Diffusion Policy baseline. These results suggest that soft-gated primitive codebooks provide an effective and interpretable intermediate action representation for robot manipulation.

0.19

TTT in VLA Models: Exploring Training with Human Demonstrations

cs.RO cs.AIcs.CV ·2026-10-23 ·Qiangong Zhou, Jiafeng Liang, Yimin Dong, Yang He, Hongxiang Li
Abstract · 22 pages, 6 figures, a techreport form standard-robots company

Long-horizon robot tasks require retaining and using past information. We integrate Test-Time Training (TTT), which stores context in fast weights, into InternVLA-A1 and Pi0.5 to study action generation conditioned on human demonstrations through training on paired demonstrations and robot trajectories. We implement memory transfer between the two phases and truncated backpropagation, and examine training objectives, gradient coverage, and contrastive regularization across video, short-video, and single-frame cue tasks. In our experiments, adding TTT and training on paired data at the current scale did not lead to consistent target selection based on demonstrations. Under the tested settings, shorter sequences, full trajectory gradient connectivity, and contrastive regularization did not clearly improve responsiveness to demonstrations. When demonstrations were reduced to single-frame cues, real-robot execution still showed a preference for cups at fixed positions. We document the implementation, training configurations, and observations to support further inspection and replication.

0.19

Semantic Representation for Agentic Self-Play in Automated Laboratories

cs.RO cs.AI ·2026-10-23 ·Yves Delley, John Coller, Zachary Cutts, Alexander Daryin, Mathias Bellaiche +10 more
position
Abstract · 16 pages, 4 figures

Navigating complex, high-dimensional biological search spaces necessitates scalable physical and computational intelligence to explore beyond local optima through iterative self-play across ideation, experimentation, and automation. Translating experimental designs into reliable physical execution remains bottlenecked by static laboratory automation and unmodeled physical variability. To address this, we introduce VERified Infrastructure for Discovery Laboratories, or VERID, a foundational infrastructure designed to bridge the gap between the high-level reasoning of artificial intelligence (AI) agents and complex, error-prone physical realities of wet labs. VERID is a persistent, high-fidelity semantic digital twin of the laboratory workcell that provides four core services: queryable introspection for agentic reasoning, time-compressed simulation for protocol validation, comprehensive trace persistence for a complete scientific record, and a framework for safe execution. We present three case studies using aqueous dye and deionized water mixtures to validate the optical and fluidic foundations of VERID. These studies demonstrate how VERID enables pre-emptive detection of critical errors in experimental protocols, stateful mid-run error recovery and post-hoc identification of subtle physical failures. The simulation functions as a discrete accounting engine, allowing agents to safely and inexpensively explore experimental designs in silico and refine experimental strategies through iterative self-play. VERID also captures structured multimodal experimental data for downstream analysis, knowledge mining and agent training. We argue that this semantic layer is an essential prerequisite for deploying autonomous agents in the biochemical and biomedical sciences, ensuring data integrity, operational safety, and paving the way for tackling previously intractable grand challenges in biology.

0.18

$R^2E:$ A Role-driven Reward Evolutionary Framework for Automated Reward Function Design

cs.RO ·2026-10-23 ·Shouhao Chang, Xuan Liu, Hongye Zhu, Xinning Chen, Shigeng Zhang
Abstract

Designing effective reward functions is a critical challenge in Reinforcement Learning (RL), which traditionally requires costly manual trial-and-error. While Large Language Model (LLM)-based methods have shown promise in reward design, undifferentiated sampling and limited use of historical feedback often lead to insufficient exploration of the reward space and optimization instability. To address these challenges, we propose $R^2E$, a Role-driven Reward Evolutionary framework for automated reward function design that structures the process through explicit role specialization. $R^2E$ employs multiple LLM roles with complementary objectives: an Explorer that promotes novelty to expand the search space, a Guardian that performs conservative refinements to improve training stability, and an Evolver that recombines high-performing rewards via crossover and mutation to integrate effective structures. A dynamic scheduling strategy coordinates these roles, progressively shifting the search from broad exploration to focused exploitation. To mitigate optimization instability and provide a medium for cross-role collaboration, $R^2E$ incorporates a Global Elite Pool that retains the best-performing rewards to guide subsequent generations, thereby enabling coordinated interaction among roles and ensuring a reliable and consistent refinement process. Extensive experiments on multiple robotic tasks demonstrate the effectiveness of our method.

0.18

TinyFlow: Breaking the Warping Bottleneck for Edge Optical Flow

cs.RO cs.CV ·2026-10-23 ·Xiao'ao Song, Chahat Deep Singh
Abstract · 9 Pages, 5 Figures, 4 Tables

High-performance optical flow networks are difficult to deploy on resource-constrained edge accelerators because their iterative refinement warps features by the current flow, which relies on flow-dependent resampling and data-dependent memory access. We present TinyFlow, a 1.26M-parameter warp-free optical flow network that preserves iterative coarse-to-fine refinement by replacing the warp with a fixed-window correlation over constant offsets, so the whole network compiles as a single static INT8 graph. Quantization-aware fine-tuning, warp-teacher distillation, and model souping reduce the Float32-to-INT8 penalty to below 0.2 end-point error (EPE) on MPI-Sintel. TinyFlow outperforms EdgeFlowNet and NanoFlowNet on Sintel Clean, Sintel Final, and FlyingChairs in both Float32 and INT8, while using 54% fewer parameters than EdgeFlowNet, and our INT8 model achieves 6.29/7.13 EPE on Sintel Clean/Final. In flow-only monocular visual odometry on TartanAir and UZH-FPV, the same INT8 model yields the lowest trajectory error on seven of eleven sequences.

0.17

Density control of large-scale multi-agent systems via tuning of agent interactions

eess.SY ·2026-10-23 ·Mena Zkhary, Alain Boldini
Abstract

Collective behavior in biological systems emerges through adaptive interactions among units, enabling large populations to coordinate without centralized control. This mechanism raises the question of whether collective behavior in engineered multi-agent systems can likewise be regulated by modifying the interaction law between agents, rather than by directly controlling individual units. To answer this question, we consider large-scale multi-agent systems evolving on a ring and treat the interaction kernel as a time-varying control variable whose evolution regulates the collective dynamics. Using a mean-field formulation, we develop a framework for the analysis and design of kernel-based controllers. We derive explicit conditions for controllability in application-relevant scenarios, and synthesize a state-feedback controller to regulate the kernel dynamics, validated on the microscopic, discrete dynamics. Numerical simulations demonstrate the approach's effectiveness and robustness to unmodeled perturbations. This work constitutes a first step in the design of bio-inspired control strategies for large-scale multi-agent systems by modulation of interactions between agents.

0.17

Graphs of Trajectory Bundles for Nonlinear Motion Planning

cs.RO ·2026-10-23 ·Joris Verhagen, Jon Arrizabalaga, Jana Tumova, Zachary Manchester
Abstract · 9 pages, 4 figures

We present the Graphs of Trajectory Bundles (GTB) method for motion planning of systems with nonlinear and black-box dynamics, combining sampling-based motion planning with trajectory optimization and graph search. Based on a graph-of-convex-sets decomposition of the workspace, GTB iteratively constructs a graph of trajectory bundles that captures the local dynamics of the system at any vertex. By solving a shortest-path problem on this graph, we find maximally dynamically feasible iterates that satisfy the system's dynamics in the limit and any constraints such as collision avoidance or additional safety or specification constraints by construction of the underlying graph. As a result, GTB can handle highly nonlinear, unstable, and even black-box dynamics with complex collision and specification constraints, making it a powerful tool for motion planning of complex systems.

0.17

If at First You Don't Succeed, Backtrack: Lightweight Recovery for Frozen VLAs

cs.RO cs.LG ·2026-10-23 ·Thomas Landeg, Thierry Tambe
Abstract · 17 pages, 1 figure, 14 tables

Vision-Language-Action (VLA) models perform well on robotic manipulation, yet remain brittle once their own actions carry them into states their training data barely covers. Recovering from such failures usually requires recovery demonstrations, policy retraining, or a separate model at inference time. We introduce Backtrack, a lightweight failure detection and recovery mechanism for frozen VLAs that needs none of these. A small failure prediction head, trained only on outcome-labeled rollouts of the frozen policy, reads the VLA's existing internal representations to estimate whether the current episode will fail. When the predicted failure probability exceeds a threshold, a controller pauses the policy, retraces the arm's recently achieved motion, and hands control back to the unchanged policy so it can retry from an earlier configuration. The head adds only 0.48-0.88% to the policy's parameter count. We evaluate Backtrack on LIBERO with three VLAs spanning two action-generation families: SmolVLA and $π$0.5, which use flow-matching action experts, and the autoregressive $π$0-FAST. On LIBERO-Long, the suite of long-horizon multi-step tasks and the one on which all three policies fail most often, Backtrack reduces each policy's failure rate by 19-48%, raising success by up to 7.4 percentage points; on the three shorter suites, where control success is already 85-99%, its effect is smaller. On MetaWorld, all twelve controller configurations tested with SmolVLA improve on the unassisted policy.

0.16

Predictive Long-Horizon Progress Modeling with Action-Aware Quantification

cs.RO ·2026-10-23 ·Tianchi Xu, Yalin Zheng, Yangxin Zhu, Ruonan Ren, Zheng Fu +2 more
benchmarkdataset
Abstract

Quantitative and dense task progress feedback is essential for robotic manipulation across data curation, policy evaluation, and reinforcement learning. Existing progress annotations often rely on sparse keyframe markings derived from subjective manual judgments, limiting their scalability and cross-task consistency, particularly in long-horizon manipulation. To overcome these challenges, we propose action-aware quantification, a unified progress specification that decomposes long-horizon trajectories into object-level subtasks via a vision-language model (VLM) and assigns weighted action credits to capture fine-grained execution dynamics. Using VLM auto-labeling and human calibration, we construct an 800+ hour training video dataset and LOOM, a long-horizon progress benchmark containing 19,618 human-calibrated evaluation frames. To model these dense progress dynamics, we introduce NimCritic, a predictive progress critic trained via Joint Video-Progress Learning. We further incorporate Historical Progress Perturbation to mitigate sequential drift, and design a contrastive Video-Progress Alignment objective that bridges asymmetric expert streams by mutually grounding intermediate visual dynamics and denoised progress trajectories. NimCritic matches or outperforms state-of-the-art process reward models (PRMs) on the established Video Progress Benchmark (VPB) and LOOM, surpassing GVL (GPT-6-Sol) on VPB by about 65% in terminal success accuracy (TSA). On the large-scale RoboArena policy evaluation benchmark, it also retrieves human-selected successful rollouts with about 97% overall recall. These results establish action-aware quantification as a scalable foundation for dense progress feedback in robotic manipulation.

0.16

Tiny Recursive Policy: Recursive Latent Refinement for Visuomotor Control

cs.RO ·2026-10-23 ·Hakan Aktas, Cengiz Oztireli
Abstract

Behaviour cloning has become a powerful approach for robotic manipulation, but many modern policies reconstruct their internal representation from a fixed observation window at each prediction and do not maintain persistent latent context throughout execution. We introduce Tiny Recursive Policy (TRP), a parameter-efficient behaviour-cloning policy that combines persistent temporal state with recursive latent refinement. At each control step, TRP repeatedly updates high- and low-level latent states using a shared computation module before decoding an action sequence, and propagates the resulting state to subsequent predictions. We evaluate TRP across LIBERO-100, MetaWorld ML45, and the LIBERO-Mem benchmarks. TRP performs competitively across multi-task and few-shot settings and shows substantial gains on long-horizon and context-dependent tasks, while using significantly fewer parameters than large generative baselines. Execution-time perturbations show that reliance on the persistent carry increases with temporal-context demand, while training with carry perturbations substantially improves robustness to both carry and visual corruptions. Ablations reveal a strong interaction between persistent state and recursive refinement: the benefit of carry depends strongly on the amount of within-step computation, and the resulting performance cannot be explained by parameter count alone.

0.15

Contract-Based Agentic Systems Engineering: Delegation as Quotient for Assured System Design

eess.SY cs.RO ·2026-10-23 ·Sai Sandeep Damera, Ryan Matheu, John S. Baras
Abstract

A component an agent builds from a prompt can satisfy it and still break the system it joins, because the prompt omits what its neighbours assume of it. The artifact to hand over is computed, not written. A SysML v2 requirement with assumed and required constraints is an assume/guarantee contract, so when a system requirement is decomposed over fixed neighbours, the open component's requirement is the contract quotient, the weakest sound one, and again a SysML v2 requirement. Its required constraints carry the neighbours' assumptions, which a requirement for the mission alone drops and unit tests with mocked neighbours cannot reach. Signal Temporal Logic is closed under the quotient. We build the requirements, with derivation, allocation, satisfaction and verification, as a SysML~v2 model and compute the quotient from it. A case study delegates the local controller of a simulated warehouse robot, whose route crosses a worker's path and ends at a dock, to one language model under the quotient requirement, a mission-only requirement and a natural-language task, and scores the deliveries on live episodes against the system requirement from ground truth.

0.14

Geometry-Aware Reinforcement Learning for Close-Fitting Wearable Insertion

cs.RO ·2026-10-23 ·Takuma Tsukakoshi, Elle Miller, Tamon Miyake, Namiko Saito, Yushi Wang +3 more
Abstract

Robot-assisted dressing is particularly challenging due to narrow openings and articulated body parts. Even small errors in the relative pose between the body part and garment opening can cause snagging, missed insertion, or incomplete dressing. To address these challenges, we formulate dressing as a geometry-aware insertion problem that represents task-relevant spatial relationships between the body part and object opening, rather than relying on the full deformable-object state. Based on this formulation, we propose a Reinforcement Learning (RL) framework combining a staged curriculum, which progressively guides the policy from coarse alignment to insertion and final placement, with a finger-wise insertion soft gate that provides continuous feedback on individual finger insertion progress and helps suppress premature insertion. We evaluate the framework on rigid and deformable bracelet dressing tasks through component ablations, systematic variations in the initial hand pose, and closed-loop execution of the learned policies on a real robotic system. On the post-training deformable evaluation, the full method reaches 84.7 +- 1.2% success. The insertion gate accounts for the largest gain, and adding the curriculum on top of that gate reaches the top score. On a mannequin hand the same full policy succeeds on 15/20 real-robot trials.

0.11

Cross-Scenario Controller Parameter Transformation from Real-World Driving Data through Flow Matching

eess.SY cs.LG ·2026-10-23 ·Christian Zink, Maik Pfefferkorn, Eric Lenz, Felix Kallmeyer, Rolf Findeisen
Abstract

Advanced Driver Assistance Systems have become increasingly widespread, aiming to support drivers, reduce workload, and improve traffic safety. However, many systems are rarely used because their behavior feels unnatural and does not match individual driving behavior. The growing availability of large-scale, real-world human driving data enables more human-like and driver-individual assistance functions. This paper formulates the adaptation of Adaptive Cruise Control parameters as a driver-individual scenario-to-scenario transport problem, transforming controller parameters from one driving context to another. Using flow matching, we train a generative model to predict the target-scenario behavior of previously unseen drivers from only a few demonstrations in a source scenario. We compare the approach with a linear parameter-shift baseline and evaluate flow-matching variants incorporating driver-consistent coupling, driver embeddings, and classifier-free guidance based on driving profiles extracted from trajectory data using a variational autoencoder.

0.09

Interpretable Collision Avoidance and Feasibility Guarantees through an Uncertainty Aware Pipeline

cs.RO ·2026-10-23 ·Doris Xu, Jin Hyun Park, Guadalupe Alvarado Fernandez, Adhyan Prasad, Mark Campbell
Abstract

Autonomous vehicles (AVs) operating in dynamic, challenging conditions such as intersections and roundabouts must guarantee collision avoidance under uncertainty while maintaining real-time feasibility. Most existing planners struggle to simultaneously guarantee both requirements. Chance-constrained formulations embed safety constraints directly in the optimization but become infeasible as uncertainty compounds; deterministic planners provide feasible solutions but fail to address the uncertainty that causes collisions; end-to-end learned policies provide neither guarantee in an interpretable form. We present a modular perception-prediction-planning pipeline that closes this gap by propagating uncertainty as structured, interpretable quantities across module boundaries. Rather than embedding probabilistic risk directly in the optimization, our pipeline resolves collision probabilities upstream and constructs convex safe sets within which the planner provides a guaranteed feasible solution at every cycle -- decoupling the safety guarantee from solver feasibility. Across Monte Carlo trials in a four-way unsignalized intersection, our pipeline achieves 0% collisions and 0% solver infeasibility at 12.7ms per cycle, compared to 36% and 10% collision rates for deterministic and chance-constrained baselines respectively. The ablation study confirms the necessity of each uncertainty source and the hardware deployment on Clearpath Jackal UGVs validates the pipeline's real-time viability on physical platforms.

0.09

UAV-SwarmAgent: Scalable Multi-UAV Reconnaissance and Task Allocation via Code Generation and Vision-Language Reasoning

cs.RO ·2026-10-23 ·Oleg Sautenkov, Yasheerah Yaqoot, Muhammad Ahsan Mustafa, Grigorii Kalinin, Chih-Yung Wen +1 more
Abstract

Supervising a multi-UAV reconnaissance mission usually means plotting waypoints and watching every drone. We present UAV SwarmAgent, a hierarchical multi-agent system in which an operator states the mission in one natural-language sentence and a ground-station agent writes and executes code that turns it into per-UAV routes. The agent grounds inspection targets in an aerial map with a vision-language model, splits them among N UAVs by balancing route length, and dispatches worker agents that verify incidents with a compact onboard decision model. In simulation over 4 incident types and 3 foundation models, the swarm located the incident in 244 of 255 missions (95.7%). A paired scaling study with one to 5 UAVs shows that flight time follows an Amdahl-type law. The second UAV delivers half of the attainable saving and the speedup is bounded at 1.52x. We discuss what this predictability, and the false alarms we observed, imply for an operator who supervises the swarm, and outline the user study needed to test it.

0.08

On Exact Critical-Region Boundaries in Continuous-Time Multiparametric Optimal Control Problems

eess.SY math.OC ·2026-10-23 ·Lida Lamakani, Efstratios N. Pistikopoulos
Abstract

Continuous-time multiparametric model predictive control solves the optimal control problem once, offline, for every initial state, and divides the initial-state space into critical regions, each with its own control law and sequence of active constraints. A boundary between two regions is either a hyperplane, given by one linear equation, or curved. The curved boundaries have not been computed to date; a classifier has been trained instead to distinguish the regions, so the partition has been known only approximately. In this paper we show that a curved boundary can be computed exactly, without solving any optimal control problem. The method rests on one property. Once the sequence of active constraints is fixed, together with the instants inside it at which a constraint becomes active or inactive (the switching times) and the instant at which the change of the neighboring region first appears (the event time), every condition on a boundary is a linear equation in the initial state, so the computation of a boundary, curved or not, is exact. We fix one of these instants, a switching time or the event time, and solve the linear equations that correspond to it for the corresponding initial state on the boundary; stepping the fixed instant along the horizon yields the whole boundary. A hyperplane is the case in which no instant is left to fix. We illustrate the method on the Newell-Lee evaporator, which has seven critical regions and eight boundaries, four hyperplanes and four curved, computing one boundary of each type in full with the adjacent regions at its ends.

0.08

From Freehand Demonstrations to Autonomous Robotic Ultrasound: An Anatomy-Anchored Learning Framework

cs.RO ·2026-10-23 ·Yuan Bi, Fangxu Han, Ye Su, Yiping Zhou, Zhongliang Jiang +1 more
Abstract

Ultrasound imaging is highly operator-dependent, motivating robotic systems that reproduce expert scanning skills consistently. These skills involve coordinated interpretation of anatomical landmarks, ultrasound images, and probe-tissue interactions that are difficult to encode in hand-designed control strategies. Imitation learning offers an alternative, but its observation and action representations should reflect the anatomically structured nature of scanning. We propose an anatomy-anchored imitation-learning framework that combines anatomy-guided probe placement with learned ultrasound navigation. The placement module identifies a task-specific anatomical starting point from external visual observations and establishes the initial probe pose as the reference frame for scanning. A flow-based policy then generates scanning motions from anchor-relative probe poses, ultrasound images, probe-mounted camera views, and contact forces, while a termination head predicts task completion. This avoids continuous reliance on third-person vision and reduces dependence on the robot-patient configuration. A handheld acquisition platform enables multimodal demonstrations to be collected during natural freehand scanning without a robot. On 16 unseen participants, task success reaches 96%, 94%, 88%, and 94% for kidney, gallbladder, and liver standard-plane acquisition and radial artery sweeping, respectively. Compared with a fine-tuned pi_0 model, success on the abdominal tasks improves by 10-42 percentage points, with lower time-to-target, peak contact force, and trajectory jerk. Ablations support the complementary roles of anatomical anchoring, egocentric vision, and force feedback. End-to-end trials with automatic placement achieve 90-100% success, and the policies transfer without retraining to a second robot with 83-100% success.

0.07

Robotic Optical Coherence Tomography-Guided Geometric Modeling of Soft Tissue: 3D Representations for Data-Driven Closed-Loop Shape Control

cs.RO ·2026-10-23 ·Chaoyu Zhang, Guangshen Ma, Mark Draelos
Abstract · 30 pages, 18 figures

In medical robotics, data-driven soft-tissue shape control requires deforming tissue toward desired morphologies through sequential physical interactions. This challenge is particularly important in microscale manipulation, where localized deformation is difficult to observe, represent, and control, while physics-based models can be computationally expensive and difficult to adapt to realistic environments. Optical coherence tomography (OCT) provides high-resolution volumetric perception of tissue deformation, but target-conditioned control requires effective representations of both observed and desired tissue states. In this work, we introduce an OCT-guided data-driven framework for tissue shape control, learned from user-guided demonstrations and executed with closed-loop OCT feedback. We study full-volume, tissue-masked sparse-volume, and voxelized point-cloud representations, with or without residual conditioning, using deterministic (Explicit) and generative diffusion (Diffusion) policies. Across 12 configurations and 576 phantom rollouts on 3-DoF depth and 6-DoF directional tasks, OCT-derived 3D representations support both Explicit and Diffusion policies, with performance depending on representation and residual conditioning. Residual conditioning reduces directional Shape RMSE from 0.748 to 0.555 mm for full-volume Diffusion, while voxelized point-cloud Explicit with residuals achieves 0.534 mm for depth and 0.443 mm for directional control, with comparable accuracy at intermediate unseen targets and consistent performance across deployments. We further evaluate the framework in 60 ex vivo porcine rollouts, extending to biological tissue with final geometric errors comparable to human demonstrations. These results establish volumetric OCT as current and goal representations for learned closed-loop tissue shape control across deterministic and generative action models.

0.07

Resilient Decentralized Radar-LiDAR-Inertial Odometry in Challenging Subterranean Environments

cs.RO ·2026-10-23 ·Moumita Mukherjee, Avijit Banerjee, Anton Koval, George Nikolakopoulos
Abstract

Accurate localization in GPS-denied and visually degraded environments affected by smoke, dust, or fog remains a critical challenge for autonomous systems. While LiDAR provides high-resolution spatial perception, its performance deteriorates under particulate occlusion. In contrast, frequency-modulated continuous-wave (FMCW) radar offers robustness under such conditions but suffers from sparse measurements and outliers. This article presents a resilient decentralized LiDAR Radar Inertial Odometry (LRIO) framework that fuses the complementary strengths of LiDAR, radar, and inertial sensing to achieve robust pose estimation in challenging subterranean environments. At the core of the framework is an Adaptive Residual Thresholding (ART) mechanism that autonomously detects temporal sensor degradation and selectively adjusts the fusion process to minimize unreliable measurements. Building on this, a Robust Error-Tolerant Optimal Information Filter (RETOIF) is employed within a decentralized fusion architecture to integrate LiDAR-inertial and multiple radar-inertial odometry estimates in a fault-aware manner, enhancing resilience to environmental disturbances and intermittent sensor failures. The proposed framework is evaluated using real-world sensor data collected in GPS-denied subterranean tunnels under smoke-free and smoke-degraded conditions. In the smoke-free experiment, the proposed RETOIF framework achieves a root mean square localization error (RMSE) of 2.1733, compared with 2.9929 for the conventional centralized LRIO approach, corresponding to a 27.4\% reduction in RMSE. Experimental results further demonstrate the framework's ability to maintain pose estimation under smoke-induced sensor degradation, highlighting its potential for resilient autonomous navigation and mapping in challenging subterranean environments.

0.05

People & Society 33

Democratic Accountability for Autonomous Systems: A Legislative Notification Framework and Comparative Institutional Design

cs.CY cs.CR ·2026-10-23 ·Samarjeet Malik, Mansi Ganga Nayak, Mugdha Raje Ashiya, Hritika Adhikary, Vaibhav Sharma
published
Abstract · 15 pages, 3 figures. Presented at the 2026 IDeaS Conference (Carnegie Mellon University)

No domestic legislature has ever reviewed a state's decision to authorise an autonomous cyber weapon before it acted. Closing the international attribution gap is necessary but not sufficient for democratic accountability. We formalise the Tripartite Attribution Framework's three attribution levels and add a fourth, non-legal Democratic Accountability level ($A_{dem}$): a 72-hour legislative notification requirement. Scoring six democratic oversight bodies against five adoptability dimensions yields a counter-intuitive result: the statutes granting these bodies privileged access to classified material are often the same statutes withholding particular ongoing operations from them, so at most one could lawfully receive such a notification today, and the carve-out Level 4 needs is required in nearly every jurisdiction rather than only the least-equipped. Retrospective analysis of Stuxnet, NotPetya, and Volt Typhoon shows one pattern: legal attribution resolvable, domestic democratic review absent.

0.43

Playing Bias: Epistemic Games for Interpreting Text-to-Image Models

cs.CY cs.CVcs.HC ·2026-10-23 ·Amalia Foka
publishedposition
Abstract · 11 pages. Accepted at iLRN 2026, the 12th International Conference of the Immersive Learning Research Network, Athens, Greece, June 2026. Vi…

Text-to-image (T2I) systems do not merely render prompts; they circulate defaults for "what X looks like," shaping recognition through pose, camera, costume, props, and setting. While technical audits quantify disparities, they rarely cultivate the literacy needed to argue how images assemble stereotypes and confer authority. This paper translates a three-stage, humanities-informed framework into a set of epistemic games that make this cultural work visible and debatable: Interpret (art-historical close looking with AI outputs), Interrogate (counter-intent prompting as artistic exploration), and Expose (creative misuse through paired prompts with a constant motif and a BiasSlot rule, in which only the bias cue varies). In a pilot with undergraduate fine-arts students, the games shifted discussion from impressionistic reactions to concise, warranted hypotheses grounded in visual evidence, supported by lightweight classroom instruments that enabled reproducibility and structured collaboration. We argue that such game-based approaches foster critical literacy across educational contexts, equipping learners to recognize, discuss, and contest how T2I models encode bias.

0.43

Present Drawing: Exploring Avatar-Mediated Drawing Events with Physical Traces

cs.HC ·2026-10-23 ·Kojiro Tanaka, Takuya Iwamoto, Yuki Okafuji
top venue
Abstract · Accepted (ACM UIST 2026)

We introduce Present Drawing, an approach for staging an avatar-mediated drawing event that leaves physical traces in the real world. Present Drawing aims to reconfigure conventional drawing events by relaxing their physical constraints. We describe an organizer-oriented design space for such an event along four elements: Time, Site, Control Source, and Avatar Embodiment. To ground this concept, we developed a proof-of-concept prototype in which an observer wears a head-mounted display to view an avatar. The avatar's motion is synchronized with a customized pen plotter whose mechanical parts are hidden from the observer. We also developed five application examples spanning live, playback, multi-site, mascot, and system-controlled configurations. We conducted an initial evaluation of one representative playback autograph configuration, showing the basic plausibility of the core Present Drawing experience as an avatar-mediated drawing event and highlighting factors associated with perceiving the avatar as the drawer.

0.42

Autonomous Cyber Weapons as AI-Enabled Digital Harms: A Platform Accountability Framework

cs.CY cs.CR ·2026-10-23 ·Mansi Ganga Nayak, Samarjeet Malik, Shaurya Pratap Singh, Vaibhav Sharma, Hritika Adhikary
published
Abstract · 15 pages, 2 figures. Presented at the 2026 IDeaS Conference (Carnegie Mellon University)

Governments govern AI-generated online harms from commercial platforms (EU DSA, AI Act, UK Online Safety Act), yet the most severe class remains ungoverned: autonomous cyber weapons (ACWs) deployed by nation-states against civilian infrastructure. We address this gap with three contributions: a digital-harm taxonomy (six ACW modes $\times$ five harm categories); the Meaningful Human Control (MHC) Model, a governance-compliance function with a closed-form IHL threshold and a proved weight-sensitivity bound; and a platform accountability architecture with treaty language grounded in ARSIWA state-responsibility doctrine, not Rome Statute individual responsibility. Applied to eleven documented state-attributed operations and one hypothetical defensive AI system, every documented case falls in an accountability vacuum; the threshold is target-dependent, and against wholly civilian targets no system studied clears it, including the hypothetical. Three further incidents are correctly excluded on autonomy/attribution grounds, a discriminant-validity check. The weights and threshold parameters are illustrative, not fitted.

0.41

Cutting out the middleman? Elon Musk and the rhetoric of citizen journalism in platform-mediated communication

cs.SI cs.CY ·2026-10-23 ·Urko Peña-Alonso, Koldobika Meso-Ayerdi, Simón Peña-Fernández, Jordi Morales-i-Gras
published
Abstract · 13 pages, 5 figures, 4 tables

This article examines how Elon Musk constructed discourse about journalism and the media on X between the 2024 United States presidential election and Donald Trump's presidential inauguration. It investigates how references to journalism, media organisations, and platform-based communication were articulated within Musk's posts, and how users responded to these narratives during a period of intense political polarisation. The study combines computational classification, qualitative thematic coding, and inferential statistical analysis. A corpus of 2,906 posts published by Musk between 5 November 2024 and 20 January 2025 was compiled. From this corpus, 500 posts referring to journalism, media organisations, platforms, or media-related discourse were identified through AI-assisted filtering and subsequent manual review. In addition, 42,066 user responses were analysed from a stratified sample drawn from 3,233,112 interactions. The analysis examined patterns of agreement and disagreement with Musk's stance and their relationship with thematic, gender, and professional variables inferred from publicly available profile information. The findings show that Musk's discourse is organised around three recurrent dynamics: criticism of institutional journalism, the association of media organisations with political and ideological agendas, and the presentation of X as an alternative communicative space. Journalism is predominantly framed through accusations of bias, disinformation, propaganda, and electoral manipulation, whereas X is more often linked to participation, innovation, and direct communication. User responses largely converged with Musk's position, although disagreement remained visible. While thematic orientation and professional background showed statistically significant associations with agreement patterns, effect sizes were limited.

0.40

CRM-Agent-Bench: Benchmarking Tool-Using Agents in Stateful Enterprise Environments

cs.HC cs.CL ·2026-10-23 ·Monojit Banerjee, Zhujun Lan, Defu Cao
benchmarkpositioncode/data
Abstract · 52 pages

Deploying language agents in real applications demands more than the right answer: an agent must execute a multi-step business process correctly, follow domain-specific rules, and interact with a user across a dialogue, where the gap between saying and doing is the gap between a satisfied customer and a real loss. We argue that a deployable-agent benchmark should jointly assess four facets: the outcome an agent reaches, the trajectory it takes, its operational viability, and its policy compliance. Customer relationship management (CRM), where mutable records meet operational rules and consequential effects, exercises all four. We introduce CRM-Agent-Bench, an execution-grounded benchmark of 555 stateful, multi-turn tasks across nine business domains and 36 capability tags. Each task supplies a task-specific executable contract drawn from checks over actions, arguments, order, response, target state, and protected state; success is the strict conjunction of the checks declared for that task. We evaluate the complete agent system through these contracts and report scored coverage, latency, and cost. Across sixteen systems evaluated on the audited final release, strict success spans 64.32% to 83.60%; the closely grouped leading systems nonetheless diverge sharply in cost and per-domain reliability. CRM-Agent-Bench offers an auditable framework for measuring contract-grounded business-transaction integrity and adherence to task-declared benchmark policies, with a portable structure designed for extension as models improve. Github repository: https://github.com/salesforce-misc/crm-agent-bench

0.38

A Highly Efficient Depth-Averaged Material Point Method for Geophysical Mass Flows

cs.CE ·2026-10-23 ·Johann Kammholz, Yidong Zhao, Lars Blatny, Thomas Pauze, Johan Gaume
code/data
Abstract

Probabilistic hazard assessment of geophysical mass flows requires large simulation ensembles and thus fast and robust numerical models. We present an open-source depth-averaged material point method formulated in global Cartesian coordinates, using Lagrangian particles to track the flow margin. A separate basal friction step based on the Coulomb, Voellmy, or $μ(I)$-rheology brings the flow to rest without artificial stopping criteria. Our model reproduces analytical solutions for a sliding block and dam-break problems and matches the deposit of a laboratory granular column collapse. A back-calculation of the 2023 Brienz/Brinzauls rockslide (Switzerland) at 1m resolution (190'000 particles) runs faster than real time on a single CPU core. Calibrated to the observed runout, the three friction laws yield very similar results, but Coulomb friction shows a slightly wider affected area than Voellmy and $μ(I)$. A probabilistic 1632-simulation ensemble at 5m resolution was computed in 76s. The model is one to two orders of magnitude faster than established depth-averaged models used in engineering practice, making catchment-style ensemble simulations feasible on a desktop computer.

0.35

Thumb on the scale: Using traditional wargaming adjudication practices to improve AI-judged evaluations and grading

cs.CY ·2026-10-23 ·Daniel S. Epstein
Abstract · 35 pages, 2 figures. Preprint, version 1. Not peer reviewed

Open-ended AI work has no single correct answer to check, so a judge, often another AI, decides each result. Research on these judges asks which AI to use and how to word the rubric. It rarely tests the procedure that the judge follows. Wargames have no correct answer either, and for two centuries the field has designed that procedure: named methods that set how much freedom the referee has, their known failures, and fixes. We map eleven wargaming practices onto AI evaluation failures. We test one, the adjudication method, on 378 player actions from nine games of an AI-refereed exercise. A player states its odds, the adjudicator turns them into a target number, and a 20-sided die decides pass or fail. With rolls fixed, the target came from the rulebook (Rigid), a judge AI adjusting it (Semi-Rigid), a judge AI setting it alone (Free) or a panel of four judge AIs. In evaluation terms these are a fixed scoring rule, a rubric the judge may adjust, holistic grading and a jury. The judge AIs were Opus 5 and Gemini 3.1 Pro, each re-run. An outcome changes when an action on the same roll passes under one ruling and fails under another. Re-run, a judge changed 3 to 6 per cent of outcomes under Semi-Rigid and 14 to 16 per cent under Free. Swapping Opus for Gemini changed about 3 percentage points more than a re-run did, a gap nine games cannot confirm. The panel changed 40 of the rulebook's 378 outcomes, but the changes cancelled and the pass count moved by 8. A pass rate hides them. One published finding rests on 15 rulings, and most of our AI adjudicators and every re-run changed how often it appeared. We measure how much rulings move, not which ruling is right. The method set how often a judge disagreed with itself. Choose it before the judge AI, report how often the judge disagrees with itself, and count the rulings behind each conclusion. We propose an adjudication statement.

0.35

Eye Tracking Signal Quality Evaluation: Meta Quest Pro vs HTC VIVE Pro Eye vs HTC VIVE Focus Vision

cs.HC ·2026-10-23 ·Aleksandr Chernysh, Mehedi Hasan Raju, Kamrul Hasan, Md Arif Shahriar, Proloy Karmakar +2 more
Abstract · 14 pages, 6 figures; supplementary material (16 pages) included. Submitted to ETRA 2027. Data: https://hdl.handle.net/10877/25342 ; code: ht…

Gaze research in virtual reality needs eye trackers of known error. Thirty participants completed the same protocol on the Meta Quest Pro (QP), HTC VIVE Pro Eye (PE) and HTC VIVE Focus Vision (FV); to our knowledge, this is the first evaluation of the FV eye tracker. In a random-saccade task, median participant mean error was 1.04° on QP, 1.50° on PE and 2.41° on FV, which remained least accurate after linear recalibration. In a later block with a darker background, PE error rose by 0.70°. FV samples were positively correlated: dispersion was 9.8 times the sample-to-sample root mean square. The device clocks showed losses that mean error hides. QP paused for up to 6.04 s, and FV skipped single samples. During instructed eye closures, QP flagged every sample valid while gaze moved 9.74°. Evaluations should report repeated timestamps, loss composition on the device clock and validity during eye closures.

0.34

Asymmetric Decline Rather Than Backfire: A Repost-Network Analysis of Japanese Vaccine-Shedding Misinformation on X

cs.SI cs.CY ·2026-10-23 ·Akinobu Seki, Erina Murata, Masaki Chujyo, Fujio Toriumi
published
Abstract · Accepted at WI-IAT 2026

This study investigates the diffusion of vaccine-shedding discourse on X in 2024, focusing on polarization in repost networks and reactions to corrective information. Here, "shedding" refers to the scientifically unsupported claim that harmful substances are released from individuals vaccinated against COVID-19 and cause adverse health effects in nearby unvaccinated people. Using 109,170 posts containing the Japanese term for "shedding" and 478,089 reposts collected from January 1 to December 31, 2024, we constructed a post-level repost network, identified discourse clusters, and examined changes in repost participation before and after major events. The results show that shedding discourse formed a two-pole structure consisting of pro-vaccine and anti-vaccine clusters. After news coverage denying shedding, we found no evidence of diffusion-level backfire in anti-vaccine repost activity. Instead, although the overall topic declined, repost participation in the anti-vaccine cluster decreased more sharply than that in the pro-vaccine cluster. User transition analysis further suggests that this change was not driven by the conversion of strongly anti-vaccine users, but by the entry of previously non-expressive users into pro-vaccine diffusion and the withdrawal of anti-vaccine supporters from repost activity. These findings suggest that corrective information may reshape the composition of misinformation diffusion participants without necessarily converting core opponents.

0.34

Gaze Under Strain: How Oculomotor Behavior Tracks Cybersickness Progression in VR

cs.HC ·2026-10-23 ·Nevzat Umut Demirseren, Mehedi Hasan Raju, Oleg Komogortsev, Isayas Berhe Adhanom
published
Abstract · Accepted for publication at ACM VRST 2026. 10 pages, 4 figures

Cybersickness remains a major barrier to comfortable, accessible virtual reality (VR). Arising from mismatches between visual and vestibular signals, it places sustained demands on gaze stabilization and visual processing. Because gaze is the mechanism through which users actively sample the virtual environment, cybersickness-induced disruptions may alter both what users perceive and how discomfort itself emerges. Yet, how eye-movement behavior evolves alongside discomfort during prolonged VR exposure remains poorly understood. We present a behavioral analysis of eye-tracking data from 19 participants navigating a realistic virtual environment over a 20-minute session, with discomfort rated repeatedly throughout the experiment. Linear mixed-effects models examined gaze constructs spanning fixation stability, visual exploration, and gaze allocation as functions of discomfort, exposure duration, and their interaction. Our results show that fixation stability degraded progressively with increasing discomfort, with the effect strengthening over time. Visual exploration showed no marginal discomfort effect but exhibited significant discomfort-by-time interactions, indicating that saccadic behavior adapts dynamically rather than responding uniformly to discomfort. Gaze allocation was not associated with discomfort but shifted systematically with time. These findings show that cybersickness does not uniformly disrupt gaze. Instead, different aspects of eye-movement behavior respond differently and change in ways that only become visible when discomfort is tracked repeatedly over time. Understanding how gaze evolves during cybersickness progression provides the behavioral grounding needed to better interpret how discomfort shapes the user's visual experience of immersive environments and to design more sensitive detection approaches.

0.31

Beyond pass or fail: an information-processing account of the Turing Test

cs.HC ·2026-10-23 ·Qiwei Zhao, Oriana Pansardi, James W. Strachan, Eugenio Scaliti, Guido Manzi +7 more
Abstract

The Turing Test is passed when a human interrogator cannot distinguish machine-generated responses from human ones. Yet behavioral indistinguishability can reflect two fundamentally different failures: a failure of encoding, in which machine and human output distributions overlap so closely that source information is effectively absent; or a failure of readout, in which source information is present but the interrogator fails to extract it. To dissociate these possibilities, we inverted the Turing Test, placing humans and AI systems in symmetrical interrogator roles. Using a quantitative information-processing framework, we separated encoding, the information contained in the text; readout, the extraction of that information by an interrogator; and meta-readout, the interrogator's explicit access to the features guiding its own judgments. Across two series of experiments involving more than 3800 human interrogators and two generations of GPT models (GPT-4 and GPT-5), we found a striking dissociation across processing levels and interrogator types. As AI-generated responses became more human-like from GPT-4 to GPT-5, source information became more difficult for human interrogators to extract. GPT-5 interrogators, by contrast, read out source-discriminating information more accurately than GPT-4 interrogators. This improvement, however, came with a specific limitation: GPT-5 explanations only partially aligned with the feature weightings that actually predicted their judgments. Human interrogators showed the complementary pattern, with modest readout alignment but comparatively higher meta-readout alignment. Together, these findings suggest a double dissociation between readout and meta-readout across human and GPT interrogators.

0.30

DotSpace: A New Kind of Network for People and Their Agents

cs.SI ·2026-10-23 ·Qingyun Wu, Chi Wang
Abstract

We introduce DotSpace, a network paradigm in which people and their agents participate as first-class members. Participants interact in persistent shared spaces, forming relationships, exchanging knowledge and taking action across ownership boundaries, while each owner retains authority over their own agents. In such a network, relationships can become channels for coordinated action: people and their agents share context, make decisions, create things and maintain ongoing commitments together, while their agents carry the work forward between people's visits. Our central proposition is that persistent networks of people and their agents can widen access to distributed capabilities and support increasingly complex collective activity. Over time, such networks may become environments in which knowledge, practices and capabilities develop, shaping both what participants can accomplish and how they organize their activity together.

0.29

A unified and composable framework for modeling packed-bed liquid chromatography

cs.CE q-bio.OT ·2026-10-23 ·Jan Michael Breuer, Samuel Leweke, Johannes Schmölder, Antonia Berger, Gregor J. Gassner +1 more
unusual
Abstract

Packed-bed chromatography is a cornerstone of bioseparation, yet its modeling landscape remains fragmented by numerous model variants, inconsistent terminology, and limited structural clarity. While prior work has categorized and contrasted individual variants, our approach shifts the focus to unification. We present a unified, modular, and composable framework that integrates diverse mechanistic packed-bed chromatography models into a coherent structure. Starting from a high-resolution three-dimensional model, the framework systematically derives a hierarchy of reduced-order models through formal model order reduction. It encompasses a broad range of bulk transport phenomena, intraparticle mass transfer processes, binding mechanisms, and chemical reactions. Formulated orthogonally, these modules are independently meaningful and can be systematically combined into mathematically consistent models tailored to specific applications. The framework recovers and generalizes existing models, establishes a consistent taxonomy, and enables novel model combinations and extensions, including alternative column and particle geometries, particle-size and particle-type distributions, component-specific pore accessibility, rapid-equilibrium binding and reaction formulations, and size-exclusion with simultaneous binding. The framework is accessible through CADET-Equations, an interactive web application for model assembly and automatic generation of governing equations in LaTeX, promoting standardization and transparency. Its design principles are increasingly reflected throughout the CADET ecosystem, whose modular architecture supports the majority of the proposed model family. The generality and composable design of the framework provide a foundation for extending chromatography modeling software toward integrated bioseparation processes and advanced computational biotechnology applications.

0.27

Whale You Go Green? Immediate and Longitudinal Effects of Embodied Conversational AI Avatars on Sustainable Behavioral Tendencies and Pro-Environmental Attitudes

cs.HC cs.AI ·2026-10-23 ·Rachel Poonsiriwong, Pat Pataranutaporn
Abstract

Embodied conversational AI can be persuasive, but can it motivate people to develop sustainable behavioral tendencies and increase pro-environmental attitudes? We conducted a randomized experiment (N = 576) comparing four embodied avatars--a beluga whale, Greta Thunberg, Emma Watson, and a generic human--with a text chatbot and no-intervention control. All embodied conditions increased immediate sustainable product selection relative to the control (beta = .401-.609, all q <= .0126) and chatbot (beta = .398-.604, all q <= .007), without differing significantly from one another. Only the Beluga condition increased immediate pro-environmental attitudes relative to the control after correction. Among 287 returners assessed two months later, following an intermediate assigned-condition session, Beluga also showed significantly higher pre-booster sustainable product choice compared to the control (beta = .657, q = .0176). Although it did not significantly outperform other embodied agents, these findings establish the immediate potential of nonhuman embodied avatar interfaces as a promising direction for memorable environmental interventions.

0.26

Translations in library catalogues: how to find them?

cs.DL ·2026-10-23 ·Péter Király
Abstract

In library catalogues based on MARC21 bibliographic metadata schema there are about a dozen data elements that catalogers can utilize to record information about the different aspects of translations: whether or not the edition is a translation, source and target language, title of the original work, translator and other contributors, publication data of the original work, literary genre. Libraries however rarely explore the full arsenal, thus to find translation is not a trivial task. The paper describes these important MARC21 data elements, and shows a use case of finding translations in five catalogues (of the Norwegian, Croatian, Polish, Belgian and Dutch national libraries) explaining the differences among the practices, and the results.

0.26

The Effects of Generative AI on Collaboration and Social Networks Among Knowledge Workers

cs.CY physics.soc-ph ·2026-10-23 ·Giuseppe Soda, Alessandro Iorio, Leonardo Rizzo
Abstract

The rapid diffusion of generative artificial intelligence has raised concerns that knowledge work may become less dependent on human-to-human interaction. Yet because AI-generated outputs require interpretation, verification, and social legitimation, generative AI may instead increase collaboration among workers. We test these competing predictions using passively collected behavioral data from more than 12,000 knowledge workers observed over 13 months during the staged organizational rollout of an AI-powered assistant (Microsoft Copilot). Using two complementary quasi-experimental designs -- a staggered difference-in-differences model and a fixed-effects instrumental-variable approach -- we estimate the effects of generative AI on workers' collaboration behavior and social network structure. We estimate that generative AI increased collaboration time across communication channels and reshaped individual networks. Those using generative AI interacted with more colleagues and their networks became less clustered, indicating broadening in collaboration structures. At the same time, they also concentrated more of their collaboration on ties with stronger prior histories, deepening reliance on already trusted colleagues. Together, these findings suggest that, while generative AI changes how people work, it may not displace human-to-human interactions.

0.26

A Gaussian Process State Space Formulation for Materials Constitutive Modeling: Application to Quasi-brittle Materials

cs.CE cs.LGmath.NAmath.PR ·2026-10-23 ·Houlin Xu, Chenyang Li, K. T. Ramesh, Michael D. Shields
Abstract

Constitutive models for history-dependent, path-dependent materials typically encode this dependence in prescribed functional forms through internal variables such as damage or plastic strain that are phenomenological, model-dependent, and not directly observable. These rigid forms lack flexibility to match observations, make calibration potentially non-unique, and complicate uncertainty quantification. This paper develops a Gaussian-process state-space model (GP-SSM) that replaces the fixed form of the constitutive model with a flexible GP regression, and the prescribed internal variables with latent state variables learned directly from observable stress-strain histories. As a GP, the model is naturally endowed with a measure of uncertainty and can be constrained to satisfy physical principles. We demonstrate this framework for quasi-brittle materials under triaxial compression, where the latent variable evolution is driven by the current-state stress invariants and the increment of equivalent strain to reflect irreversible, pressure-sensitive history. A sparse multi-output GP then maps the latent state and current strain tensor to the six-component stress response, trained by variational sparse inference. The trained model is exported into Abaqus/Explicit through a user material subroutine. Four verification cases are reported: (1) offline rollout on validation and test histories; (2) a single element under changing confinement; (3) a frozen-memory unloading-reloading extension; and (4) a multi-element structural model compared against the Karagozian and Case Concrete (KCC) model. These results show that the learned state-space update reproduces held-out RVE-scale histories in closed-loop rollout, tracks evolving confinement, supports unloading without violating the irreversibility of the learned history state, and reproduces structural-level behavior consistent with KCC.

0.24

When disclosure rules are not enough: authorial warrant and judgment in ai-supported scholarship

cs.CY ·2026-10-23 ·Marcin Mirończuk
published
Abstract

Editorial policies increasingly require authors to disclose substantive uses of generative AI, but such requirements commonly focus on whether, where, and how a model was used. Even function-specific disclosure may locate a model in the research or writing process without showing how its outputs entered scholarly judgment, how they were evaluated, or whether contributors can reconstruct the relevant reasoning and defend resulting claims without treating those outputs as sufficient justification. Drawing on research ethics, AI disclosure policy, human-model interaction, and researcher formation, this article argues that disclosure documents model use, whereas responsible use requires claim-level authorial warrant: a contributor's capacity to understand, justify, and defend claims within their contribution. The article distinguishes cognitive delegation, in which model outputs shape substantive choices, interpretations, or justifications, from routine delegation, which does not ordinarily bear on the substance of judgment. Its conceptual contribution is to link the epistemic role of outputs that shape judgment to the function-appropriate evaluation required to justify decisions and retain warrant for resulting claims. It then repurposes disclosure categories as a claim-sensitive ex ante heuristic, mapping six recurrent research functions to risks, forms of oversight, and formative questions for recognising cognitive delegation and selecting function-appropriate evaluation. The article contributes to the ethics of information technology by clarifying when AI-supported scholarship remains under responsible human judgment and when cognitive delegation threatens the epistemic basis of authorship.

0.24

Freeform Programmable Liquid Crystal Elastomers via Monolithic Integration of Co-Optimization and 4D-Printing

cs.CE math.OC ·2026-10-23 ·Yibo Wang, Asad Farooq Ahmed, Weichen Li, Weiyun Xu, Xiaojia Shelly Zhang
Abstract

Liquid crystal elastomers (LCEs) are active materials capable of large, reversible thermally induced actuation, offering significant potential for applications such as artificial muscles and soft robotics. However, their broader deployment is hindered by the lack of a unified design-to-fabrication framework, the strong coupling among geometry, director fields, and manufacturing constraints, and the difficulty of reliably translating complex designs into physically realizable architectures. Here, we present an autonomous and monolithic integration of co-optimization and 4D printing that enables the end-to-end design-to-realization of versatile programmable LCE architectures with superior precision, complexity, and versatility. The proposed framework establishes a unified design strategy that co-optimizes the freeform multilayer geometry and fully continuous layer-specific director distributions, and subsequently translates the optimized fields into fabrication-ready DIW toolpaths through an automated manufacturing-aware pipeline. To bridge computation and realization, we develop an optimization-centered 4D printing approach with automated, fabrication-aware toolpath generation tailored for direct ink writing of LCEs. With the proposed design-to-fabrication paradigm, we experimentally realize architectures that exhibit complex, non-intuitive thermo-mechanical behaviors. Experiments achieve highly challenging prescribed functionalities beyond the reach of intuition-based approaches and show excellent agreement with numerical simulations across all samples. This work establishes a generalizable inverse design and physical realization of freeform programmable LCE systems, enabling function-oriented and optimization-driven intelligent LCE applications for next-generation soft robotic and responsive material engineering.

0.23

National-Scale Multimodal Accessibility Fields from Open Geospatial Data

cs.CY ·2026-10-23 ·Till Zacher, Louis Stille-Hönig, Markus Lienkamp
Abstract

Large-area accessibility analysis remains difficult to reproduce because multimodal routing, heterogeneous opportunity data, and reusable spatial outputs are rarely integrated in a single transparent workflow. We present MAPPER (Multimodal Accessibility and Place Profiling Engine for Arbitrary Regions), an open-data pipeline for multimodal accessibility analytics. The workflow builds walk, bike, car, and public transport travel-time graphs from OpenStreetMap and GTFS data, harmonizes mapped POI- and proxy layers representing the land-use side of the accessibility relation, and computes continuous accessibility fields using network kernel density estimation. A low-overhead native compute core (contraction-hierarchy routing, C++/OpenMP accessibility accumulation, DAG-style orchestration eliminating redundant computation) is the enabler that makes many graphs, large extents, and a broad layer set jointly tractable on commodity hardware. Outputs are sampled to a uniform H3 grid for downstream urban and regional analysis. A Germany-wide production run yields an H3-indexed accessibility dataset with nearly 40 million cells and 972 mode-layer fields across nine network and intermodal modes. Routing comparisons against an empirical mobility panel show low mean bias for walking, cycling, and car, while public transport results are interpreted more cautiously as schedule-based supply estimates sensitive to waiting and transfer assumptions. A Bavaria case study operationalizes the travel-time targets of the state's central-place planning system on the resulting fields and reveals strong modal differences in accessibility to basic services. The contribution is a reproducible framework for large-area multimodal accessibility analysis, with an aligned output structure designed for reuse in urban and regional analysis, planning support, and scenario modeling.

0.22

Universal neural network approximation of symmetric polyconvex energies and envelopes

cs.CE math.NAcs.LG ·2026-10-23 ·Florian Gammer
theory
Abstract

Symmetric polyconvexity provides a sufficient condition for weak lower semicontinuity in small-strain elasticity and a tractable lower bound for the relaxation of non-(quasi)convex energies. Since classical algorithms for computing such functions can be computationally expensive, neural network approximations have recently emerged as an alternative. A central difficulty with establishing such neural network based approaches for symmetric polyconvexity is the additional monotonicity property. As we show, the corresponding weight restrictions on ReLU-based input-convex networks do not yield universal approximation of the lifted representers. Here, we introduce symmetric polyconvex neural networks (SPCNNs) for constitutive modeling and approximation of symmetric polyconvex envelopes. The deep architecture combines convexity with the monotonicity required of a representer in the space of minors. We prove a universal approximation theorem for symmetric polyconvex energies, without requiring isotropy, and quantify the number of affine pieces attainable by the lifted network and derive reduced architectures for isotropic energies. Numerical examples illustrate stress-based regression and approximation of the symmetric polyconvex envelope of a double-well potential. Symmetric polyconvexity is guaranteed by construction, while the accuracy of the fitted energies and envelopes is assessed numerically.

0.21

Political sense-making in "individual" human-technology-world relations: the case of LLM-generated autobiographies

cs.HC ·2026-10-23 ·Anthony Longo, Olya Kudina
Abstract · Reference: Longo, A., and Kudina, O. (in press). "Political sense-making in 'individual' human-technology-world relations: the case of LLM-g…

As AI systems increasingly mediate how people describe, recognize, and present themselves, their politics cannot be reduced to what remains hidden inside the "black box." This chapter argues that AI's political significance emerges within technologically, intersubjectively, and temporally mediated interpretive practices. Bringing Arendt's political philosophy into conversation with Gadamerian hermeneutics, the chapter develops a postphenomenological relational-temporal account of political sense-making across Human-Technology-World relations. Against the charge that postphenomenology remains individualistic and using the case of LLM-generated biographies, the chapter shows how seemingly private human-LLM encounters are worldly and collective. When users ask systems such as ChatGPT to describe them, they do not simply receive algorithmic representations. They interpret, contest, edit, share, and publicly test these descriptions. In doing so, people move interpretively among affective resonance, friction, public re-articulation, and the sedimentation of meaning. This means that the so-called "micro-level" of postphenomenological analysis is already saturated with publicity: affective uptake becomes communicable, reflection becomes accountable, and uses sediment into shared standards of appearance. Simultaneously, we show that once mediation is treated as relational and temporal, the political is not an add-on at the "macro" level; it is intrinsic to the very movement through which technologies amplify and reduce, stabilize and unsettle, and thereby compose a shared world in which judgment can take place. The chapter finishes by discussing several design and policy implications for AI of such political foregrounding in postphenomenology.

0.21

Virtual Placebo: a scoping review of control conditions in randomized controlled trials of psychological immersive virtual reality interventions

cs.HC q-bio.NC ·2026-10-23 ·Madis Vasser, Mary-Ann Kubre, Kadi Tulver, Jaan Aru
Abstract

Research on immersive virtual reality (IVR) as a therapeutic tool is growing rapidly, with randomized controlled trials (RCTs) now commonplace. Yet the control conditions in many of these studies that claim strong specific IVR effects vary from usual care to active immersive experiences, making it difficult to draw clinically meaningful conclusions. We conducted a scoping review of recent literature on psychological IVR interventions to assess the main forms of control conditions used and their relations to study outcomes. In this paper, we make the case that most of the recent RCTs studying the effects of specific IVR interventions compare active immersive VR to usual care with no multimedia component. This can lead to strong yet potentially misleading study outcomes and can potentially inflate the IVR interventions' effect. Namely, the lack of an appropriate immersive control condition may risk misattributing general effects of VR devices to the specific content of the intervention. In this review, we discuss the implications of these design choices and propose strategies for including a suitable "virtual placebo" condition to improve the methodological rigor and interpretability of results.

0.19

Learning the Past from the Future: Retrospective Reconstruction in Temporal Interaction Networks

cs.SI cs.LG ·2026-10-23 ·Amarnath R
Abstract · 12 pages, 5 figures, 3 tables

Temporal link prediction uses past interactions to predict future links. In practice, interaction records from web and mobile applications can contain gaps caused by logging failures, API limits or lost records, even when later events remain available. To our knowledge, the extent to which later events help recover interactions within such gaps has not been measured directly for this ranking task. We study retrospective reconstruction using the tgbl-wiki and tgbl-comment datasets. We hide a two-hour window, use its interactions as ground truth, and rank the directed pairs active within it. We compare count-based heuristics and logistic regression models using past information alone or together with 1, 6 or 24 hours of later events. Target pairs and candidate lists remain fixed across conditions, so differences in ranking performance cannot be attributed to changes in candidate selection. With 24 hours of later events, the 26-feature logistic regression improves mean reciprocal rank (MRR) by 0.039 on tgbl-wiki under exhaustive ranking, 0.046 under historical-random ranking, and 0.108 on tgbl-comment under historical-random ranking. The 26-feature logistic regression benefits from adding later features to past features, whereas replacing the past pair count with the later pair count reduces MRR by 0.157 on tgbl-wiki under exhaustive ranking. A control removing later interactions from query sources to historically observed destinations reduces the MRR gain by 59% on tgbl-wiki and 10% on tgbl-comment. Because this control removes more than repeated target pairs, these reductions cannot be attributed to pair recurrence alone. Later events improve the ranking of hidden interactions on both datasets, including tgbl-comment, where only 6.4% of target pairs reappeared within 24 hours.

0.18

The As-If Machine: Personalized AI counterfactuals make distant policy risks feel close

cs.HC cs.AI ·2026-10-23 ·Joshua Ashkinaze, Yubo Zhou, Pengyu Chen, Avi Bhattacharya, Hyebin Park +4 more
Abstract

Life experiences can shape policy attitudes, but we only live one life. Citizens may therefore fail to support policies addressing risks they have never personally experienced. Narratives can make such risks immediate and personal, but personalized stories remain difficult to produce at scale. Here we ask whether AI can make distant scenarios feel concrete by simulating a participant's life as if that scenario occurred. We present the As-If Machine (AIM), a multi-agent system through which people co-write personalized counterfactual narratives---stories about their own lives under different policy futures. In two pre-registered and quota-matched experiments on science funding (N=1,007) and satellite cybersecurity (N=894), we compared AIM with two controls: reading an expert-written vignette on the same topic and co-writing with AIM about an unrelated topic. AIM increased psychological closeness---how close a risk feels to the self---by nearly identical margins in the two domains (0.36 and 0.38 standard deviations (SDs) against the vignette, 0.80 and 0.90 against unrelated writing). AIM increased psychological closeness even among participants with unfavorable views of AI. For policy attitudes, AIM's effects were 0.16 SDs against the vignette and 0.10 SDs (n.s.) against unrelated writing for science funding; the corresponding effects for satellite cybersecurity were 0.54 and 0.67 SDs. Across 23,496 human-AI autocomplete interactions, participants and the system played distinct roles: AIM supplied concrete detail that grounded the scenario, and participants supplied first-person perspective and emotion. Co-writing with the As-If Machine makes distant policy risks feel close.

0.17

From Perception to Pixels: Human-Derived Visually Distinct Colors for Improved Dominant Color Extraction

cs.HC cs.CV ·2026-10-23 ·Uldana Shyndali, Nurdaulet Toleugaliyev, Kuat Zharykbayev, Zarina Abildasheva, Pakizar Shamoi
Abstract

Dominant color extraction is a fundamental task in computer vision, with applications in design, image retrieval, and content recommendation. However, conventional algorithms based on device-oriented color spaces (e.g., RGB) and fixed quantization often fail to capture how humans perceive color differences. In this study, we propose a human-perception-based framework for dominant color extraction, built on a novel Human Color Dictionary (HCD) derived from a perceptual experiment. Participants (n=110) were sequentially shown 120 colors and asked to identify those they perceived as unique. Based on their responses, we constructed a 120 by 120 co-occurrence matrix and applied a greedy selection algorithm to derive 64 perceptually distinct colors, forming the HCD. This dictionary was integrated into RGB color histogram and K-means color extraction algorithms by initializing centroids and mapping pixel values to their nearest perceptual color using the deltaE 00 distance. Evaluation on a dataset of 4,562 images paired with expert-designed palettes demonstrated that our method improves perceptual alignment, as measured by Earth Mover's Distance (EMD) and best-match distance. Our findings suggest no meaningful gender differences in the number of uniquely perceived colors, indicating that sex does not significantly influence color perception in this task. Another key finding is that highly saturated colors are perceived as the most distinguishable, whereas dark and deep colors tend to perceptually collapse and are less reliably differentiated. The results confirm that incorporating experimentally derived human color representations ensures the generated color clusters align more closely with human perception, thus improving the usability of color palette generation in practical applications.

0.16

AI Alignment for Digital Democracy: Fairer Voters Representation against LLM Biases

cs.CY cs.AI ·2026-10-23 ·Yi Qi, Srijoni Majumdar, Evangelos Pournaras
code/data
Abstract

AI representation is emerging in digital democracy, assisting in voter emulation and decision-making. Large language models (LLMs) behave differently depending on the model, the prompt context, and the voter information they receive, representing some groups more accurately than others and potentially making the emulation unfair to certain groups. We investigate demographic biases in LLM-emulated electorates: how accurately LLMs represent individual votes, whether some voter groups are represented more accurately than others, and whether AI alignment can reduce these disparities without sacrificing overall accuracy. Using three U.S. presidential elections, we compare a closed-source and an open-source LLM across voters' demographic features and test AI alignment strategies toward fairer voter representation. Removing gender from the voter profile reduces the largest group error of the models by 22-42%, and removing education reduces it by 64%, whereas removing self-reported race increases the mean group error by over 40%, all without lowering accuracy. Common alignment strategies, such as repeated queries or fairness instructions, do not reduce these errors and can double them.

0.16

Fast and Scalable Influence Estimation in the Independent Cascade Model: a unified framework

cs.SI ·2026-10-23 ·Matic Požar, Miklós Krész
benchmark
Abstract

Computing diffusion processes on networks is fundamental to applications including influence maximization, epidemic modeling, information propagation, and diffusion inference. The Independent Cascade (IC) model is one of the most widely used models of progressive diffusion, yet exact computation of its activation probabilities is #P-hard, while Monte Carlo simulation becomes costly when accurate estimates are required repeatedly. Despite its central role, analytical approximation of the IC diffusion function has received comparatively little systematic attention relative to the extensive literature on influence-maximization algorithms. We present a comprehensive study of analytical approximations for the IC model. We review a wide range of existing approaches and unify them through a common framework based on temporal representation, retained propagation information, and aggregation rule. This perspective exposes connections between previously distinct approximation families and motivates several new methods. We further conduct, to our knowledge, the first large-scale cross-family comparison of analytical IC approximations for node-level marginal activation probabilities. Our benchmark spans real-world and synthetic networks with systematically varied topology, seed-set size, edge-probability model, and diffusion strength. The proposed second-order dynamic message-passing method achieves the lowest pooled median error and the lowest dataset- or graph-family-level median error on every network considered. The results reveal systematic differences between approximation families, identify regimes in which higher-order corrections are most beneficial, and establish analytical IC approximation as a rich accuracy-efficiency design space rather than a collection of isolated methods.

0.15

GlanceAR: Repurposing Phone Sensors via Low-Cost Reflectors for Always-On Obstacle Awareness

cs.HC ·2026-10-23 ·Boyang Zhou, Xizhe Hao, Minbeom Kim, Jaewook Lee, Jon E. Froehlich
Abstract

Smartphone use while walking reduces environmental awareness and can lead to accidents and injury, yet existing obstacle-awareness systems remain unadopted because they require users to hold the phone upright and aimed forward, conflicting with how people actually hold their devices. We introduce GlanceAR, an always-on mobile obstacle-awareness system designed for natural phone-holding postures. GlanceAR utilizes a low-cost passive mirror attachment to redirect the phone's rear LiDAR and RGB camera forward without additional electronic components. In a within-subjects study with N=11 participants navigating indoor and outdoor obstacle courses while engaged in a primary on-phone task, we evaluated three sensing configurations: GlanceAR, a LiDAR-only variant for privacy, and an attachment-free ultra-wide camera baseline. Our findings reveal trade-offs among sensing fidelity, obstacle cue design, and social acceptability, and surface insights around AI complacency and the tension between on-screen task focus and physical safety.

0.13

From Regulation to Engineering: Organizational Privacy Climate as Mediator

cs.CY cs.HCcs.SE ·2026-10-23 ·Oshrat Ayalon, Eran Toch, Irit Hadar, Michael Birnhack
survey
Abstract

Data protection legislation aims to protect privacy rights by regulating how information systems store and process personal information. While such legislation defines legal obligations, translating regulatory requirements into engineering practice is far from straightforward. As new frameworks such as the EU AI Act emerge, understanding these mechanisms becomes increasingly relevant. To investigate them, we conducted an online survey of 156 U.S. information system developers, examining the role of organizational privacy climate in the association between developers' perceptions of the GDPR and their privacy engineering practices. Using structural equation modeling, we found that organizational privacy climate mediates the relation between developers' perceived GDPR importance and their privacy attitudes, which in turn are associated with their privacy engineering practices. The mediation chain - regulation, climate, attitudes, practices - complements the direct effect of regulation on practice, revealing the organizational and attitudinal pathway through which regulatory intent reaches engineering behavior. We discuss implications for the next generation of privacy and AI regulation.

0.12

Dual-Corrected Stochastic Block Model ($\text{DC}^2\text{SBM}$): A Generalized Structure Detection Method for Attributed Networks

cs.SI cs.LGstat.ML ·2026-10-23 ·Zhenhai Chang, Shuo Wang, Yumeng Liu, Wei Liu
Abstract · 24 pages, 8 figures

Community detection in attributed networks is one of the core tasks of complex network analysis. The stochastic block model (SBM) and its variants are able to detect generalized network structures such as assortative, disassortative and mixed structures. However, existing generative methods either ignore node heterogeneity, or characterize node activity in the topological and attribute dimensions by one and the same factor, making it difficult to correctly model node activity when the two kinds of activity are decoupled. In this paper, we propose the Dual-Corrected Stochastic Block Model ($\text{DC}^2\text{SBM}$), which introduces independent correction factors---a topological degree factor and an attribute abundance factor---into the link-generation and attribute-generation channels, respectively, thereby effectively decoupling node activity from membership inference, and which further introduces a global scaling parameter to resolve the mismatch between the normalization constraints and the network scale. The model is inferred by the expectation--maximization (EM) algorithm, and the newly introduced factors cancel out completely in the posterior computation, so that no additional computational cost is incurred. Experiments on synthetic and real attributed networks show that DC^2SBM outperforms state-of-the-art methods in community detection accuracy, with an especially pronounced advantage when the topological signal is weak. The two correction factors can be recovered independently from their respective channels without cross-talk. Analytical as well as controlled experiments further demonstrate that modeling with a single shared factor fails when the two kinds of activity are decoupled---a regime that is prevalent in real networks---thereby validating the necessity of dual independent correction.

0.09

When Should I Trust This? Designing and Evaluating Factuality Indicators for LLM-based Chatbots

cs.HC ·2026-10-23 ·Greta Warren, Christine M. Lehane, Jessica Heal, Sila Eser, Denise M. Leto +1 more
survey
Abstract

Generative AI systems increasingly mediate how people seek information, yet communicating AI factuality to laypeople remains challenging. Through a mixed-methods investigation involving interviews (N=8), a survey (N=1,000), and an experiment (N=5,151), we examine how people evaluate AI-generated information for gardening, mathematical, and medical queries, and how factuality indicators (confidence and sources) shape perceptions. Our interviews and survey show people selectively verify AI responses using external and AI-provided evidence, moderated by contextual factors. Although confidence-indicators affected trust and accuracy perceptions, people hold diverse, ambiguous interpretations of AI confidence. Sources were viewed as crucial, yet users raised concerns about source reliability and verification effort. Our experiment showed that confidence and source-based indicators influenced initial perceptions, but effects disappeared once participants learned the response's correctness. Medical queries elicited skepticism regardless of indicator. Findings motivate helping users interpret model confidence and integrating source information into the interface to support calibrated trust in AI-supported information-seeking.

0.07