PHYSICAL AI · 2026-10-19

Physical AI Brief

Daily cross-source signals for the Physical AI supply chain — silicon photonics, CPO, VLA models, humanoid hardware, embodied AI. Three streams, one page, zero filler.

401 items today · 338 arxiv · 1 SEC 8-K · 62 humanoid · 0 CN photonics

01 ARXIV · PHYSICAL AI PAPERS

338 items
  1. arxiv:2610.23856 · cs.CV
    Learning to Correct Geometry in Generated Videos
    Weijie Lyu, Tiancheng Shen, Jason Li, Yujing Wang +4

    Generated videos often lose geometric consistency under camera motion: structures warp under viewpoint change, parallax becomes inconsistent, and smooth camera movement may collapse into shot-like transitions. We argue that this failure is largely a data problem. In web-scale video corpora, most visible motion comes from dynamic objects captured by static or weakly moving cameras, leaving only sparse supervision for viewpoint-induced appearance changes. To address this, we propose Video Geometry Corrector (VGC), a post-hoc video-to-video model trained on synthetic paired supervision. Starting from real videos with diverse camera motion and pose annotations, we construct VG-Pairs with a synthetic paired-supervision data generation pipeline. Each video clip is treated as a geometry-consistent target and paired with a distorted counterpart synthesized by a pretrained image-to-video inpainting model using trajectory-adaptive keyframes. VGC learns to map the distorted input to the geometry-consistent target without modifying the source generator, and uses optical flow as an auxiliary cue to help disentangle camera motion from object motion. Experiments on pose-conditioned and prompt-driven benchmarks show that VGC improves geometric consistency across multiple source generators, and scaling to a larger backbone further extends the framework toward more general video refinement.

    benchmark
  2. arxiv:2610.23849 · cs.RO
    Blind Precise Assembly: Proprioceptive Sim-to-Real Peg-in-Hole for Diverse Parts
    Antonia Bronars, Younghyo Park, Kushal Kedia, Pulkit Agrawal

    Every robot already carries high-resolution contact sensors: its own joints. Legged robots and humanoids exploit them to walk, run, and track whole-body motions without vision, but there the contacts are large and forceful. Is joint sensing also enough for precise, contact-rich assembly, where the robot must find a millimeter-scale opening, mesh gear teeth, or catch a thread through light contact? We study this question by training blind assembly policies that observe only a short history of joint positions, velocities, and their own actions, without visual, force, torque, or tactile sensing. The poses of both the fixture and the part are unknown to the policy, and the grasp is randomized substantially. The policy must therefore assemble the parts through contact, sensed exclusively through its own proprioception. Trained with reinforcement learning entirely in simulation over randomized parts, grasps, and contacts, such policies learn peg insertion, gear meshing, and nut threading. The peg insertion policy transfers zero-shot to the real world and inserts pegs of different shapes and sizes in 73% of trials. Controlled comparisons with matched visual policies in simulation show that blind policies generalize far better to unseen part shapes and sizes, though they can be slower. Training such policies needs thousands of different assemblies, each simulated with its clearance intact, so we also release an open-source generator that builds them procedurally from exact convex pieces for MuJoCo; it is more accurate and faster than convex decomposition or MuJoCo's signed-distance collision. Videos and interactive simulations are at https://younghyopark.me/blind-proprioceptive-assembly/.

    humanoidtactilesim-to-realgrasp
  3. arxiv:2610.23845 · cs.AI
    AI-Driven Reproducibility and Analytical Robustness in Scientific Discovery
    Shicheng Liu, Krista Opsahl-Ong, Andrew Drozdov, Julian Killingback +5

    Reproducible and analytically robust research findings are essential to scientific progress, but evaluating these properties typically requires costly manual effort. We investigate whether LLM agents can help automate reproduction and robustness analysis, using a medical case study grounded in the Surveillance, Epidemiology, and End Results (SEER) database, a widely used U.S. cancer registry underlying tens of thousands of published studies. First, we introduce SEER-Repro, a benchmark of 20 published SEER-based papers together with the data and resources required to reproduce their findings. Second, we create an AI agent that successfully reproduces 77.4% of 2,190 atomic claims, while providing executable baselines for further analysis. Third, we develop an agentic analysis pipeline to test the robustness of the medical findings by identifying analytical choices underlying each finding, proposing defensible alternatives, and rerunning the analysis to determine whether those choices affect the conclusion. Across published papers in SEER-Repro, our pipeline identifies potentially consequential analytical choices in 9.5% of fully reproduced findings and flags them for expert review. These results show how LLM agents can support scientific review by reproducing findings, systematically testing their robustness, and directing expert attention toward potentially consequential analytical choices.

    agentai agentllm agentagenticbenchmark
  4. arxiv:2610.23841 · cs.RO
    StopWatch: Goal Grounding for Object-goal Navigation
    Xinhao Liu, Yunsheng Tian, Yuheng Qiu, Pieter Abbeel +3

    Pretrained VLMs bring rich world knowledge and semantic priors to object-goal navigation, and fine-tuning a VLM into a navigation policy has become a common recipe. Despite this progress, success rates remain low, and it is unclear which capability these fine-tuned policies lack. Our diagnosis identifies goal grounding as the dominant failure. These policies explore well, yet often fail to notice that the goal is already in view or to estimate how far away it is. To address this failure, we ground the goal explicitly. Explicit grounding substantially improves success and reaches state-of-the-art results on six benchmarks across three scene collections. We also correct annotation defects in these benchmarks, so that future methods are evaluated more faithfully. Our analysis on the clean benchmarks shows that, with ground-truth grounding, our policy nearly matches the expert that generated its demonstrations, leaving the headroom in perception rather than policy learning. Our work points to goal grounding as a missing piece of object-goal navigation and offers a simple and effective solution to the task.

    benchmark
  5. arxiv:2610.23839 · cs.CL
    DiscoveryRSI: Recursively Learning to Discover Valuable Hypotheses
    Jianyuan Zhong, Chenwei Wu, Zhaochen Yu, Xixian Jiang +20

    Scientific discovery requires developing worthwhile hypotheses into practical investigations, yet learning from this evolving research process remains challenging. We introduce an operational definition of valuable and testable research hypotheses: evidence-grounded claims about consequential uncertainties, coupled with feasible investigations, explicit decision criteria, and alternative explanations. Our DiscoveryRSI framework develops hypothesis trees through multi-agent evolutionary search, compiles selected investigations into executable Harbor tasks, and recursively learns through three mechanisms: hypothesis credit optimization attributes scientific assessments to documented research contributions; harness replay identifies productive strategies and recurring failures to refine research workflows; and rediscovery applies accepted improvements to new investigations, generating further experience for subsequent learning. We also introduce DFM-Bench, which provides comprehensive assessment of hypothesis quality and cumulative research progress at the levels of hypotheses, revisions, development paths and research runs. Across four life-science domains, automated evaluation shows that hypothesis quality rises from 4.86 for the base model to 5.39 after two rounds of self-improvement, with the large gains in falsifiability and data resolvability. We further validate hypothesis quality through expert review, comparing our DiscoveryRSI-27B with other research agents. Preliminary human assessments show that DiscoveryRSI-27B scores higher than GPT-6 Luna on hypothesis importance and whether claims are plausible and evidence-bounded. DiscoveryRSI-27B achieved a 6.3% rate of hypothesis correspondence with research published after the model cutoff, largely outperform GPT-6 with 2.5% and Claude Opus 5 with 2.3%, while delivering greater human-rated improvement along research paths.

    multi-agentself-improvement
  6. arxiv:2610.23837 · cs.RO
    SparseMax-dSTL: Sound Robustness with Horizon-Independent Gradient Allocation for Temporal Logic Trajectory Optimization
    Sai Sandeep Damera, Ryan Matheu, John S. Baras, Calin Belta

    A gradient-based solver that enforces a Signal Temporal Logic (STL) specification uses a smooth approximation of its quantitative semantics, the robustness. In safety-critical trajectory optimization, this approximation must have two properties. First, it must be sound, that is, nonnegative only on trajectories that satisfy the specification. Second, at each temporal operator, the weight of its gradient on the violating samples must stay bounded away from zero, however long the horizon. The approximations in use, including the log-sum-exp, its sound variant, and the generalized-mean robustness, do not have both. We introduce lower approximations of the minimum and the maximum whose gradients are sparsemax projections, and we prove that the robustness built from them has both. It lies below the exact robustness and is therefore sound, and its error has a bound that is independent of the number of samples and known before optimization. The gradient of each approximation is exactly zero on every sample beyond a data-dependent threshold, so the weight on violating samples a gap beyond the others has a positive lower bound at every horizon. On a planar unicycle and on a seven-joint manipulator beside a person in simulation, the log-sum-exp approximation is positive on a violating trajectory. As the horizon of the manipulator's task grows from 10 to 80 s, the gradient weight of the sound variant on the violating samples falls from 0.22 to 0.04, while that of sparsemax stays at 1. In a constrained trajectory optimization of the unicycle, the constraint on the log-sum-exp robustness returns a violating trajectory, whereas the constraint on the sparsemax robustness returns a trajectory that satisfies the specification with the required margin.

    manipulator
  7. arxiv:2610.23831 · cs.LG
    Beyond the Answer: Student-Conditioned Trace Construction for Reasoning Distillation
    Shaoang Li, Yanhang Shi, Yufei Li, Mingfu Liang +5

    A reasoning trace can accompany a correct answer without providing effective supervision for a particular student. We study how to construct traces that improve learning beyond answer-only supervision by considering both their intermediate content and their accessibility to the target learner. We introduce ReDiST, a student-conditioned trace-construction and distillation framework for evidence-rich reasoning. Starting from the student's own attempts, ReDiST revises answer-correct drafts against the supplied evidence and repairs unsuccessful drafts with additional guidance from the verified answer. A frozen student-matched reader screens and ranks original and edited candidates using answer recovery and verified-answer probability, given the question, evidence, and candidate trace. We distill the selected targets while balancing trace and final-answer supervision with an auxiliary answer loss. Construction and selection occur offline, and inference uses only the distilled student. Experiments on question-answering benchmarks show that ReDiST can improve both source-task and cross-benchmark performance over answer-only, original-trace, and the evaluated trace-selection and rewriting baselines. Across supervision recipes, performance rankings can differ between source and transfer benchmarks, highlighting the importance of evaluating trace teaching utility across distributions.

    benchmark
  8. arxiv:2610.23827 · cs.LG
    Quality-of-Transmission Estimation in Quantum Networks: Compute, Simulate, or Learn?
    Anis Bekri, Amar Abane, Van Sy Mai, Abdella Battou +1

    Future quantum networks will rely on control-plane decisions such as routing, scheduling, entanglement provisioning, purification, and memory allocation. Each decision depends on a timely and reliable estimate of the quality of the entanglement that a candidate path can deliver, with end-to-end fidelity as a central metric. Inspired by the optical-networking experience with quality-of-transmission (QoT) estimation, this article asks when a quantum controller should compute fidelity analytically, simulate it, or learn a surrogate model. The optical analogy is useful but incomplete: optical QoT estimation often predicts a scalar transmission margin, whereas quantum QoT estimation must account for probabilistic entanglement generation, random memory waiting times, decoherence, state consumption, and measurement constraints that can make fidelity a stochastic target. We propose a compute/simulate/learn framework for quantum QoT estimation and illustrate it with a simulation-based proof of concept. Our study shows that analytical composition can be sufficient in simple regimes, that simulation becomes necessary when stochastic memory effects shape the delivered fidelity distribution, and that lightweight learned surrogates can approximate simulator labels at much lower runtime cost for online control. We close with open challenges for deployment, including stochastic targets, sim-to-real transfer, data-generation cost, physics-informed learning, and generalization beyond individual paths to full network topologies.

    sim-to-realmemory
  9. arxiv:2610.23826 · cs.RO
    Self-Building Labs
    Jiaru Bai, Yuchi Zhao, Han Hao, Miroslav Bogdanovic +3

    Self-driving laboratories (SDLs) accelerate scientific discovery by automating experimental execution and optimization. AI agents can increasingly propose new experiments, but existing SDLs can only run the procedures they were built for. Adapting them to a new experiment still takes extensive manual engineering, which makes laboratory flexibility the bottleneck for scaling discovery. We introduce the self-building laboratory (SBL), which designs, configures, and operates its own laboratory setup given a high-level scientific goal. SBL starts from whatever already exists, from an empty room to a working SDL, and builds the missing capabilities: the lab layout, its digital twin, robot skills, and experimental workflows. We demonstrate the system working across different robots and setups, scaling from a single workstation to laboratories with several robots working together. We validate SBL's end-to-end capabilities through a real-world quantum-dot synthesis experiment. By shifting the role of human researchers from manually configuring and programming device automation to specifying scientific objectives, SBL makes scientific automation more accessible, flexible, and scalable. Project website: https://self-building-labs.github.io/

    ai agent
  10. arxiv:2610.23825 · cs.LG
    ENFR: Evidence Networks with Flow-Matched References for Amortized Simulation-Based Model Inference
    Kai Lehman, Jameson Dong, Carolina Cuesta-Lazaro, Niall Jeffrey +4

    The evidence, $\mathcal{Z} = p(x | \mathcal{M})$, of a data-generating model $\mathcal{M}$ is a powerful tool in Bayesian frameworks: it enables model inference, model misspecification tests, and goodness-of-fit assessment as an alternative to standard discrepancy statistics, e.g. $χ^2$. While versatile, the evidence is notoriously difficult to compute, even when the likelihood is available. In both likelihood-based and simulation-based inference, estimating $\mathcal{Z}$ provides an immediate solution to high-dimensional goodness-of-fit or model misspecification tests: using $\mathcal{Z}(x)$ as the test statistic induces a minimum-volume level-$α$ test. Such tests are essential to assess model fidelity, which is necessary to trust simulation-based parameter inference. Existing learned approaches largely target Bayes factors, which only measure evidence ratios. Instead, we provide accurate, fast, and amortized estimates of the absolute evidence. We present ENFR: Evidence Networks with Flow-matched References. Building on the Evidence Network framework of Jeffrey & Wandelt (2024), the proposed method first learns Bayes factors with respect to a flow-matched reference, from which the absolute evidence is recovered. We show improvement in estimation of $\mathcal{Z}$ over the flow-based baseline in all applications considered and the best performance of all density estimation methods considered in our benchmarks. We furthermore introduce and provide the proof for an architecture that jointly learns the individual evidences of a set of models $\{\mathcal{M}_i\}$. We demonstrate the performance and scaling behavior of these approaches on data from a variety of scientific applications.

    benchmark
  11. arxiv:2610.23821 · cs.LG
    Learning without relearning physical symmetry: an SO(3)-equivariant convolutional network (CNN) on the sphere
    Joshua Williamson, Niall Jeffrey, Lorne Whiteway

    Many of the most important datasets in the physical sciences live natively on the sphere, and these data often carry no preferred position or orientation. Using methods that do not reflect this symmetry leads to far higher simulation requirements, which can be catastrophic for scientific analyses (forcing reliance on cheap, imperfect models that bias results). We present a novel spherical implementation of a gauge equivariant mesh architecture that is specifically designed for the Hierarchical Equal Area isoLatitude Pixelization (\healpix{}) of the sphere. The $\mathrm{SO}(3)$-equivariant convolution is built from steerable filters, parallel transport and spin-weighted features. The architecture supplies the exact symmetry, so the network does not spend training data learning it. The convolution operations are constructed to give provably exact equivariance, given a band-limit. Our full CNNs can achieve SO(3)-invariance error at per-cent-level and lower, which is controlled through well-understood hyperparameters. On benchmarks, our technique gives the most accurate dense gradient regression and competitive cosmic microwave background foreground removal and denoising. The improved simulation efficiency matters most in simulation-based inference (SBI), where high-fidelity simulations are expensive and neural compressors require large training sets, so a fixed budget forces analyses with less accurate simulators. In cosmology lensing SBI, our technique yields the tightest posteriors at matched simulation count. This approach is a novel method that will enable new convolutional neural network analyses, including generative and inference tasks, for processes and data on the sphere.

    benchmark
  12. arxiv:2610.23820 · cs.LG
    CocoRNA: Collective RNA Design with Cooperative Multi-agent Reinforcement Learning
    Tianmeng Hu, Biao Luo, Ke Li

    Designing RNA sequences that reliably fold into specific secondary structures is essential for understanding their biological functions but remains a challenging computational problem. We propose CocoRNA, a cooperative multi-agent reinforcement learning framework for RNA inverse design. CocoRNA simplifies the design task by decomposing it into smaller sub-problems, each solved collaboratively by multiple agents. This approach reduces the complexity of the problem and improves the exploration of design policies. During training, a centralized critic uses global structural information to guide the agents, enabling them to jointly optimize their design strategies. As a result, CocoRNA learns high-quality RNA design policies that generalize effectively to unseen structures without additional training. Experiments on the Rfam dataset demonstrate that CocoRNA substantially outperforms state-of-the-art methods in both success rate and design speed. Further experiments on other biological sequence design tasks highlight the effectiveness and broad potential of CocoRNA for complex design tasks.

    multi-agent
  13. arxiv:2610.23812 · cs.LG
    Learning Personalized Memory Gating from Delayed One-Sided Feedback
    Kaixuan Ji, Zixiang Chen, Wenting Zhao, Cheng Qian +8

    Long-term memory enables language-model agents to personalize future interactions, but a finite memory bank cannot preserve every extracted fact. Existing memory gating policies are typically shared across users and therefore fail to account for user-specific patterns of future reuse. In this paper, we study memory gating at the fact level. We first propose a retrieval-based labeling strategy that labels each fact by whether it is later retrieved and incorporated into responses. We then formulate online memory gating as contextual apple tasting with delayed one-sided feedback, and propose Warm-Start UCB, which combines offline learning from existing data with online adaptation to the current user. Our theoretical analysis establishes a high-probability regret upper bound of $\tilde{O}(Tε\wedge d\sqrt{T})$ when the delay window is at most $O(\sqrt{T})$, where $T$ is the total number of rounds, $ε$ is the initialization error, $d$ is the dimension of the feature map, and $\tilde{O}(\cdot)$ hides constant and poly-logarithmic factors. The results of our experiment demonstrate that our method outperforms all baselines under different memory budgets and history lengths, establishing it as a promising personalized memory gating policy.

    memory
  14. arxiv:2610.23809 · eess.SY
    Lyapunov Exponents in Power Systems: A Survey of Theory and Applications in Stability, Control, Forecasting, and Protection
    Mohammad Almomani, Venkataramana Ajjarapu

    Lyapunov exponents (LEs) are fundamental metrics from nonlinear dynamical systems theory that quantify the average exponential rate of divergence or convergence of nearby trajectories in phase space. Over the past three decades, LEs have emerged as powerful tools across a wide range of power system problems, including transient and voltage stability assessment, chaos detection and mitigation, short-term load and renewable energy forecasting, and fault diagnosis. This paper presents a comprehensive review of these applications, covering more than 165 publications spanning from the early demonstrations of chaos in simplified machine models (1990s) to the most recent data-driven and machine-learning-augmented approaches applied to inverter-based resource (IBR)-dominated grids (2024--2026). The review is organized into four principal application domains: (i) stability analysis and real-time monitoring, (ii) chaos characterization and control, (iii) forecasting and predictability analysis, and (iv) protection, fault diagnosis, and condition monitoring. For each domain, we trace the evolution of methods, compare model-based and data-driven LE estimation strategies, identify key benchmarks and validation gaps, and outline the most pressing open research problems.

    benchmark
  15. arxiv:2610.23795 · cs.CV
    Persistent Planar Memory for Video World Models
    Yuze He, Bin Tan, Zelin Gao, Yujun Shen +3

    Video world models that generate and explore environments at scale need a persistent 3D memory of what they have produced. Plane primitives offer a natural candidate: a compact set of oriented planes captures the dominant structural elements (e.g., walls, floors, facades) of indoor and outdoor scenes, while requiring only 7 parameters per primitive. We present PPMem, a video world model that adopts plane primitives as its persistent 3D memory. PPMem augments a video diffusion transformer with a planar prediction head and a Depth Token Fusion (DTF) module: each autoregressive chunk jointly produces RGB video and planar geometry, which merges into a persistent plane memory that conditions subsequent chunks via rendered depth. The resulting memory is three orders of magnitude more compact than per-pixel point clouds (~100K primitives, 5MB for a full scene), geometrically explicit (metric depth, normals, exportable surfaces), cross-view consistent (jointly predicted by the shared backbone), and adaptive (refined by the merging procedure as generation proceeds). On DL3DV and RealEstate10K, PPMem outperforms prior 3D-grounded world models in video quality while jointly producing metric-scale depth, surface normals, and exportable planar geometry, sustaining coherence across hundreds of frames under large viewpoint changes. Project page: https://hyzcluster.github.io/ppmem.

    world modelmemory
  16. arxiv:2610.23791 · cs.AI
    DrugHarness: An Evolvable Agent Harness for Computational Drug Discovery
    Yuxing Chen, Zhisheng Zhang, Cheng Tang, Wei Li +19

    Computational drug discovery requires agents to compose scientific operations and adapt to changing evidence and constraints. Here we introduce DrugHarness, an evolvable agent harness coupling composable scientific execution with conditional skill induction. Its environment integrates over 80 scientific operations, biomedical resources and task-specific code, preserving associations among candidates, results and artifacts. From execution trajectories and outcome feedback, DrugHarness induces action hypotheses with observable applicability conditions, recommended actions and supporting evidence. An independent review agent evaluates candidate skills derived from these hypotheses through source verification and, when needed, targeted execution trials. Admitted skills guide subsequent tasks and are revised through feedback. Evaluation on MolBench and SMDD-Bench demonstrates capabilities in molecular screening, optimization and constrained design. Without evolution, DrugHarness exceeds reported MolClaw-CC results by 10.8 percentage points in affinity-comparison accuracy and 55% in mean top-three active hits on MolBench. Retrospective evaluation on skill-development tasks shows that evolution increases Qwen's affinity-comparison accuracy from 62.2% to 81.1% and optimization success from 79.5% to 87.2%. In a prospective HPK1 study, DrugHarness achieves a more favorable median docking score than Biomni among system-specific candidates. A synthesized compound achieves a cellular IC50 of 14.19 nM, a 2.64-fold potency improvement over the best input reference, with activity competitive with two selected compounds designed by experienced medicinal chemists. These results support conditional skill evolution for scientific agents and demonstrate a workflow connecting computational drug design with experimental validation.

    agent
  17. arxiv:2610.23789 · cs.CV
    The Collection of a Real-World, Large-Scale Dataset for Automated Skill Evaluation in Central Line Dressing Change
    Ummey Tanin, Shannan K. Hamlin, Hsin-Mei Chen, Peizhu +1

    Central line-associated bloodstream infections (CLABSIs) remain a major source of preventable morbidity and mortality in healthcare settings. Proper execution of the central line dressing change (CLDC) procedure is essential for infection prevention, yet current approaches to competency assessment rely primarily on manual ob- servation by clinical educators, making evaluation time-intensive, difficult to scale, and limited in consistency and granularity. These challenges highlight the need for AI-assisted skill evaluation methods that can support more objective, reproducible, and scalable assessment of clinical performance. However, progress in this area has been constrained by the lack of broad-scale, clinically grounded datasets for developing and evaluating such methods. In this work, we present the collection process of a large-scale video dataset for CLDC skill evaluation, comprising 365 recordings of registered nurses collected during the 2026 annual competency exam at Houston Methodist Hospital. Each video is paired with subtask-level temporal annotations and clinical performance assessments spanning 17 subtasks of the procedure. We describe the data collection and curation pipeline, including pilot testing of record- ing configurations, annotation protocol design, and post-collection processing. We further establish the feasibility of video-based skill assessment through benchmark experiments with spatiotemporal deep learning models. This dataset provides a foundation for future research on automated clinical skill assessment, feedback generation, and AI-assisted training in nursing education.

    benchmark
  18. arxiv:2610.23787 · cs.CL
    DLite: History-free Speculative Decoding for Agentic AI
    Jiaming Xu, Hanzhen Wang, Tianlang Zhao, Jiewen Xiao +6

    Speculative decoding has become a mainstream technique for accelerating large language model (LLM) inference by using a lightweight draft model to generate candidate tokens that are subsequently verified by the target model. However, agentic workloads introduce long and continuously evolving contexts and frequent multi-turn interactions. Under such workloads, we observe that existing draft models suffer from up to an 80% degradation in acceptance length, while maintaining their historical KV cache introduces substantial system overhead. In this paper, we identify the growing historical context as a common source of both problems and introduce DLite, a history-free speculative decoding framework for agentic AI. DLite directly reuses contextual hidden states from the target model and performs local speculative prediction without explicitly maintaining the historical context, transforming an increasingly complex modeling problem into local simple prediction problem. A lightweight sequential head further restores intra-block causal dependencies with low overhead. To effectively train this heterogeneous history-free architecture, we introduce a tailored two-stage training strategy. Experiments on general-purpose, long-context and agentic workloads show that DLite achieves 2-6 acceptance length and outperforms the state-of-the-art speculative decoding method, DSpark, up to 70%.

    long-contextagentic
  19. arxiv:2610.23772 · cs.CV
    Video Diffusion Avatars: Taming Video Diffusion Models into Controllable 3D Avatars
    Simon Giebenhain, Tobias Kirschstein, Liam Schoneveld, Davide Davoli +2

    Video diffusion models (VDMs) trained on internet-scale data encode a rich prior on appearance, motion, and physics that is difficult to replicate with explicit 3D rendering and costly 3D data capture. We introduce Video Diffusion Avatars (VDAs), a method that creates photo-realistic, multi-view consistent avatars from a single image by fine-tuning a VDM for precise control over camera viewpoint, facial expression, and upper-body pose. Departing from 3D morphable models (3DMMs), VDAs use a learned expression space that lifts the expressiveness ceiling of 3DMMs, e.g., for the tongue and mouth interior. However, learned expression codes exhibit heavy viewpoint bias, as our DISTRACT benchmark reveals. We show that this bias can be removed through disentanglement, enabling their use as driving signals for multi-view consistent avatars. Camera and upper-body pose are controlled jointly via proxy-mesh renderings, giving independent control over expression, pose, and viewpoint. A camera-triplet training strategy makes the approach efficient: our models train for 34 hours on a single A100 using only public multi-view data. VDAs outperform state-of-the-art single-image 3D avatars on novel-view reenactment, reducing face LPIPS by 29% and improving perceptual quality (JOD) by 66% on BecomingLit. A static 3DGS reconstruction probe further confirms that our generations are the most multi-view consistent. Built on a video prior, VDAs produce effects beyond the reach of explicit renderers, including physically plausible hair dynamics, shadows, reflections, and refractions.

    benchmark
  20. arxiv:2610.23769 · cs.RO
    Not All Actions Contribute Equally: Learning Precise Progress Estimation from Imprecise Demonstrations
    Huanqian Wang, Boshi Tang, Botong Xiao, Jie Zhang +4

    Progress estimators play an essential role in providing dense reward signals for reinforcement learning in robotic manipulation, but accurately assessing task progress can be much challenging. Existing methods heavily rely on expert data and naively assume that all actions within a demonstration contribute to task success. This assumption often breaks down in real-world scenarios, especially in high-precision manipulation, where subtle geometric errors can lead to task failure. Therefore, these methods can erroneously interpret human mistakes as meaningful progress, further corrupting progress learning. We argue that a key limitation is the lack of a global prior regarding task evaluation criteria, which is crucial for distinguishing meaningful progress from human mistakes. To this end, we propose PRIME, a novel progress estimator capable of learning precise progress estimation from imprecise demonstrations by aggregating experience across trajectories. PRIME constructs a weighted graph over observed states, connecting states across trajectories based on spatial proximity and visual consistency. Edge weights encode spatial distances as transition costs, and shortest-path search approximates the geodesic distance to success. Changes in this distance naturally distinguish advancement, stagnation, and setbacks without assuming monotonic progress over time. Extensive experiments in both simulation and real world on precise manipulation tasks show that PRIME is consistently effective across diverse tasks, and robust to data with varying degrees of suboptimality. Notably, the advantage of PRIME gets more pronounced in higher-precision scenarios, yielding up to a \(5\times\) improvement in policy success rate over the SFT baselines on charger insertion.

    manipulation
  21. arxiv:2610.23766 · cs.LG
    Logic-Guided Graph Denoising Diffusion Models
    Samuel Cognolato, Luciano Serafini, Alessandro Sperduti, Andrea Cavallaro

    Graph diffusion models are competitive approaches for generating complex graph-structured data. However, when such distributions have strict structural constraints, diffusion models may fail to capture them exactly, resulting in the generation of invalid graphs. In this paper, we propose a method for grounding graph diffusion models in domain-specific knowledge via First Order Logic (FOL), leveraging its expressive and interpretable formalism to impose structural constraints on the generative process. We incorporate domain knowledge using classifier guidance, a technique that conditions the diffusion model's distribution on external variables to exert control over the sampled data. During the generative process, classifier guidance steers the denoising trajectory to optimize an objective function, which we define as a fuzzy and differentiable relaxation of the logical rules. Our methodology requires only the specification of constraints in FOL, whose satisfaction is encouraged at inference time to guide a pre-trained generative model. We evaluate our approach by comparing the unconditional generation of knowledge graphs with our logic-guided framework, which incorporates domain ontologies as structural constraints during inference. This approach significantly increases the proportion of generated graphs that satisfy the ontological constraints.

    knowledge graph
  22. arxiv:2610.23751 · cs.CV
    MTEdit: Packed Trajectories Beat Rollouts for Multi-Turn Image Editing
    Houyuan Chen, Fanzhou Wang, Xiangyu Fan, Songlin Yang +6

    Multi-turn image editing aims to modify an image according to a sequence of instructions, making it better aligned with human-AI collaborative creation than single-turn editing. However, constructing high-quality multi-turn training data remains challenging, as conventional sequential rollout accumulates errors and introduces visual-quality drift across editing turns. To address this challenge, we introduce MTEdit, a multi-turn image editing dataset built through a pipeline centered on packed trajectory canvas (PTC) generation. By jointly generating multiple editing states within a unified visual context, this pipeline reduces visual-quality drift across turns. Covering both step-wise and context-aware editing scenarios, MTEdit comprises 500K multi-turn image editing trajectories, with up to 8 turns and an average of 5.38 turns per trajectory, making it one of the most extensive datasets for multi-turn image editing. We further introduce MTEdit-Bench, a benchmark designed to evaluate both scenarios by assessing models' ability to follow editing instructions, maintain content consistency, resist visual-quality drift, and effectively use historical context across turns. Using MTEdit, we fine-tune Qwen-Image-Edit-Causal to obtain Qwen-Image-MTEdit. On MTEdit-Bench, we evaluate a broad range of closed-source and open-source image editing models, including Qwen-Image-MTEdit. Experimental results show that Qwen-Image-MTEdit achieves the best performance among the evaluated open-source models, demonstrating the effectiveness of MTEdit in improving models' multi-turn image editing capabilities.

    benchmark
  23. arxiv:2610.23745 · cs.CL
    VocalGrad: Evaluating Acoustic Perception in Audio Language Models
    Koki Ryu, Hitomi Yanaka

    Recent large audio language models (LALMs) have extended their capabilities beyond lexical speech understanding to tasks requiring paralinguistic and non-linguistic audio understanding, such as emotion recognition. However, these models still struggle to reliably utilize acoustic information, particularly when it co-occurs with spoken linguistic content. Existing benchmarks do not clearly distinguish acoustic perception from higher-level reasoning, making it difficult to identify the source of these failures. To address this issue, we introduce VocalGrad, a benchmark designed to evaluate LALMs' ability to perceive low-level acoustic variations. VocalGrad formulates this problem as a simple binary classification of temporal change direction, enabling evaluation largely independent of higher-level reasoning or semantic understanding. Our benchmark results reveal a substantial gap in acoustic perception between human annotators and state-of-the-art LALMs: while human annotators achieve high accuracy, LALMs perform at near-chance levels. We further conduct linear probing to identify structural bottlenecks in these models. Our analysis reveals two distinct failure modes: for some categories, task-relevant information is preserved in internal representations but not utilized by the language model head; for others, the information becomes progressively less linearly decodable across model layers. Overall, our results reveal substantial limitations of LALMs in modeling acoustic variation under controlled settings and expose representation differences not observable from text outputs.

    benchmark
  24. arxiv:2610.23735 · cs.LG
    LagNet: Lag-Structured Operators for Long-Term Time Series Forecasting
    Zhenan Yu, Guangxin Jiang, Jiuyun Jiang, Jin Yang

    Long-term time series forecasting maps a finite history window to an extended future horizon. Many existing models focus primarily on the design of input representations and model structures, while paying less attention to the structural form of the history-to-future forecasting operator itself. We examine this operator directly and find that, across model families, sensitivity matrices exhibit recurring diagonal patterns whose diagonals correspond to relative temporal lags. Motivated by this observation, we develop a framework for direct multi-step forecasting based on lag-structured temporal operators. Its basic building block, LagLayer, constructs a Toeplitz matrix by tying history-to-future connections with the same relative lag and incorporates horizon-wise and history-wise scaling coefficients to relax exact sharing. This design enables LagLayer to serve as an effective replacement for temporal linear layers in existing forecasting models. Stacking LagLayer modules with activations and multiple feature maps yields LagNet. Experiments on long-term time series forecasting benchmarks show that our method is competitive with strong baselines, while sensitivity matrix diagnostics and limited-data experiments demonstrate that LagLayer provides an effective and sample-efficient alternative to a linear layer.

    benchmark
  25. arxiv:2610.23734 · cs.LG
    Bridging the Language Gap in Scholarly Data II: Enhancing Gender Inference for Chinese Names
    Lisette Espín-Noboa, Mingrong She, Ana Maria Jaramillo, Liuhuaying Yang

    Perceived gender inference from names is important for inclusive scientometrics research, yet existing methods perform poorly on non-Western scholars. We address this gap for Chinese names, which appear in two forms, Hanzi and Pinyin, its Romanized form (e.g., Zhang Wei, Li Na). In international bibliographic records, names are often available only in Pinyin. However, a single pinyin name can correspond to multiple distinct hanzi characters with different gender associations, so gender inference from Pinyin alone is intrinsically ambiguous. To address this, we first develop two hanzi models, a regression model that produces a continuous gender score from 0 (male) to 1 (female), and a classification model that outputs one of three labels (\em{male}, \em{neutral}, \em{female}). We train them on a sample of hanzi names from Chinese-language physics publications, rated by 11 native Chinese speakers. On this sample, the models achieve an RMSE of $0.09$ and an R$^2$ of $0.85$ (regression), and macro-AUC of $0.97$ and a Cohen's $κ$ of $0.81$ (classification), demonstrating high performance that is robust to class imbalance. For pinyin names, we propose a pipeline that aggregates hanzi predictions via transliteration and frequency weighting, together with an ambiguity score that flags names whose gender cannot be resolved from Pinyin alone. Benchmarked against three off-the-shelf tools (Namsor, GPT, and Claude), our models achieve the highest agreement with human annotations for both gender scores and gender labels, in both scripts. Applied to an independent sample of Chinese-named authors in APS journals, our models assign a gender to every name, whereas existing Pinyin-based methods leave between 20\% and 100\% of names unclassified. These findings demonstrate the importance of hanzi-level information for reliable gender inference of Chinese names. Code and data are publicly available.

    benchmark
  26. arxiv:2610.23722 · cs.AI
    Beyond Large Language Models: A Mental World Model Framework for Mental Modeling, Simulation, Evolution, and Interpretability
    Qingshi Wang

    Understanding and simulating how subjective cognition and mental states emerge, evolve, and shape behavior is essential for advancing artificial intelligence toward deeper social cognition and general intelligence. Existing large language models (LLMs) and Theory of Mind (ToM) methods primarily infer mental states from linguistic information, lacking unified mechanisms for subjective world construction, cognitive-psychological interactions, dynamic evolution, and historical dependencies. This paper introduces the Mental World Model (MWM) and its implementation, the Multi-Branch Mental Evolution Architecture (MMEA), which explicitly constructs subjective worlds, models interactions between cognition and mental states, and simulates continuous mental evolution shaped by historical experiences. On DynToM and RecToM, MMEA achieves 96.48% and 94.00% accuracy, respectively, outperforming leading LLMs and specialized ToM methods. Additional experiments demonstrate improvements of 28.7, 39.3, and 16.6 percentage points in mental state and behavior prediction, long-sequence prediction, and implicit mental state inference, respectively. The results demonstrate that MMEA provides an interpretable framework for continuous mental modeling, enabling AI to move beyond surface-level semantic representations to infer individuals' underlying mental states from linguistic information and dynamically simulate their internal cognitive processes and evolution over time. MWM also provides a promising foundation for causal reasoning, personalized AI, sustained human-AI interaction, and future advances toward artificial general intelligence (AGI).

    world model
  27. arxiv:2610.23709 · cs.AI
    GLOBE: Evolving Structured Agentic Memory for E-commerce Product Optimization in Generative Search
    Hyunseo Kim, Suil Choe, Sukkyoung So, Daeyeon Yi

    Optimizing product listings for generative search requires language agents not only to rewrite effectively but also to continually self-improve by retaining successful strategies across diverse categories. However, existing approaches often treat structured listings as free-form text, leaving effective strategies implicit in prompts, search trajectories, or rewritten outputs rather than accumulating them as a reusable agentic memory. To address this limitation, we introduce GLOBE, an agentic framework that drives continual self-improvement by storing optimization knowledge in an evolving handbook of explicit, field-scoped behaviors. In each episode, an Editor applies relevant behaviors and receives rank-based feedback, which a Reflector converts into behavior updates. Crucially for dependable execution in real-world settings, a Moderator removes redundant proposals, a Curator commits accepted changes, and consistently harmful behaviors are deactivated. This cycle transforms episodic feedback into a persistent, field-scoped optimization memory. Experiments show that structured behaviors outperform monolithic prompt optimization, while feedback-driven evolution and moderation further improve effectiveness and reduce downside risk. The learned behaviors also remain effective on unseen product categories, suggesting that agentic memory evolution supports interpretable and transferable product optimization.

    memoryagenticself-improvement
  28. arxiv:2610.23707 · cs.LG
    Tool Recommendation for Multi-Step Agents: Finding Every Tool a Task Needs
    Sam Partee, River Stevens, Thierry Damiba, Valerie Fanelle

    A multi-step agent task fails when a single tool it needs is missing, yet the tools an agent may need are too many to list: the roughly 8,200 tools on the platform we study have definitions that together exceed the context window of any model we evaluate. We present Tool Recommendation, a deployed system that gives an agent all of them, hundreds of times more than a typical MCP server exposes, through a fixed interface of three tools. It runs at the gateway between agent and tools, executing every call under the end user's own credentials and checking it against the user's consent for each service and the tools the agent may use. We evaluate it on both a public retrieval benchmark and billed end-to-end session cost. On ToolRet, Tool Recommendation places every tool a query needs in the top ten 57% more often than BM25, a lexical ranking that commercial tool search offers, and compares favorably with published systems up to nine times its size. In paired multi-step workflows, its first recommendation covered every required tool in 93% of Codex sessions, and against a fixed list of 100 tools it cut input tokens by 17% with Codex and 40% with Claude Code and the total bill in all 14 runs, by up to 40%, with no change in task completion.

    agentbenchmark
  29. arxiv:2610.23705 · cs.CV
    FedRAPP: Federated Test-Time Prompt Tuning for Vision-Language Models with Reliability-Aware Prompt Personalization
    Syed Md. Ahnaf Hasan, Md Akil Raihan Iftee, Sajib Mistry, Amin Ahsan Ali +1

    Federated prompt learning enables vision-language models like CLIP to collaboratively adapt lightweight soft prompts across distributed clients without sharing raw data. However, existing federated prompt learning methods are designed primarily for supervised settings with labeled client data and assumed relatively stable training distributions. At test time, source-free federated prompt tuning remains largely unexplored. Clients may receive unlabeled target samples with heterogeneous, often non-IID distributions that differ substantially across clients and from the source training distribution, causing locally adapted prompts to drift in different directions and reducing the effectiveness of federated aggregation. In this work, we explore Source-Free Federated Test-Time Prompt Tuning (SF-FTTPT), where distributed clients collaboratively adapt soft prompts on unlabeled test streams without access to source data. Our study reveals that conventional federated aggregation methods struggle under heterogeneous client distributions, as locally adapted prompts become increasingly specialized and difficult to combine effectively. To address this challenge, we propose FedRAPP, a personalized aggregation framework that combines reliability-aware client assessment, server-side exponential moving-average prompt memory, and similarity-guided peer aggregation. FedRAPP selectively incorporates information from reliable and compatible clients while preserving local specialization, enabling more robust adaptation under distribution shift. Extensive experiments across nine CLIP benchmark datasets and ImageNet variants demonstrate that FedRAPP consistently outperforms standard federated aggregation baselines under heterogeneous client distributions while maintaining the efficiency advantages of prompt-only adaptation.

    benchmark
  30. arxiv:2610.23703 · cs.LG
    Blitz: Accelerating the Diffusion Modules of AlphaFold3 and Boltz-2
    Vinay Saji Mathew, Soundar Kumara, Gretta Kellogg, William K. M. Lai

    Diffusion-based protein structure prediction models produce all-atom structures in about 200 denoising steps, and that cost has driven a number of few-step samplers. We present Blitz, a few-step distillation framework that uses a geometric noise schedule to distribute steps more evenly across noise scales, then learns those updates through multistep consistency training. Physical steering and target-free candidate selection help reduce the remaining geometric defects at inference. We apply Blitz to AlphaFold 3 and Boltz-2, reducing denoising from 200 steps to eight. Because published few-step results use different benchmarks and scoring procedures, we compare both distilled models with their teachers and other few-step methods under a common evaluation protocol covering structural accuracy, interface quality, ligand placement, and stereochemical validity. Blitz retains teacher-level structural accuracy, with AF3-Blitz outperforming all evaluated few-step baselines and exceeding its teacher on several accuracy metrics.

    benchmarkevaluation protocol
  31. arxiv:2610.23692 · cs.LG
    Expressivity of IQP Feature-Map Quantum Kernels via Reproducing Kernel Hilbert Spaces
    Valeria Repetto, Stefano Melchionna, Romina D'Aurizio

    Quantum kernels are among the most promising methods in quantum machine learning, providing a natural way to embed data into high-dimensional Hilbert spaces via quantum circuits and capturing non-linear relationships in data. Their learning behaviour is governed by the function class they represent, which is formally characterized by the structure of the associated Reproducing Kernel Hilbert Space (RKHS). However, a theoretical characterization of this function class, and of how it depends on the choice of feature map and its hyperparameters, is still lacking. In this work, we focus on the Instantaneous Quantum Polynomial (IQP) feature map and derive a Fourier representation of the associated RKHS generators as finite sums of trigonometric functions evaluated on polynomially transformed input variables, with a finite set of accessible frequencies determined by the circuit construction and scaled by the bandwidth hyperparameter. This representation identifies three bandwidth regimes with qualitatively distinct expressive properties from localised, through multimodal, to periodic. Numerical experiments on synthetic datasets with controlled periodic and multimodal structure, across data dimensions and feature-map configurations, reproduce the three regimes in both supervised classification and unsupervised anomaly detection, benchmarked against a classical Gaussian kernel. These results make the relationship between expressivity and data structure explicit for IQP kernels: expressive power translates into a practical benefit only when the accessible frequencies match with the structure of the data, which provides a criterion for selecting the bandwidth when the periodic or multimodal structure of data is known.

    benchmark
  32. arxiv:2610.23685 · cs.LG
    HardEval: An Execution-Verified Benchmark for AI-Driven Security Hardening and Compliance Remediation
    Manohar Vellala, Soumya Banerjee, Sandip Roy, Ross Gore +1

    Implementing Security Technical Implementation Guide (STIG) controls remains a demanding operating-system administration task. Each natural-language security requirement must be translated into a precise and persistent configuration change. These changes span Secure Shell (SSH) configuration, authentication policy, audit rules, and kernel parameters. Existing benchmarks do not measure whether a model can perform this task on a live system. Code benchmarks evaluate isolated programs without persistent system state, while agent benchmarks largely focus on applications rather than operating-system configuration. In this paper, we introduce \textbf{HardEval}, an execution-verified benchmark for evaluating language-model-assisted security hardening. HardEval is built from Red Hat Enterprise Linux 8 STIG controls that fail on an unhardened baseline system. Expert-written reference remediations and Open Vulnerability and Assessment Language (OVAL) checks serve as hidden ground truth for grading; they are never shown to the model, and each task states only the required security outcome while withholding the implementation. The model produces a Bash script, which is executed on a live virtual machine and graded from fail to pass using the corresponding OpenSCAP/OVAL compliance check. Evaluation of fourteen frontier and open-weight configurations reveals large and uneven capability differences across security domains. A recurring failure mode is plausible security code that does not satisfy the exact compliance condition. For four open-weight configurations, multi-run Wilson confidence intervals preserve the observed ordering. HardEval is publicly available.

    agentagent benchmarkbenchmark
  33. arxiv:2610.23678 · cs.CL
    Beyond the Benchmark: Conditioning, Domains and Failure Modes of Qwen3-TTS
    Séverin Baroudi, Yanis Labrak, Petr Motlicek, Hervé Bredin +1

    Qwen3-TTS reports state-of-the-art zero-shot voice cloning, but its technical report leaves open how the model should be conditioned, how it behaves outside read speech, and how it fails. We characterise its two cloning interfaces, in-context learning (ICL) and X-Vector, against Chatterbox, XTTS-v2 and F5-TTS on read, meeting, emotional and air-traffic-control (ATC) speech, using eighteen objective measures, an LLM judge, a spoofing countermeasure and a listening test. Neither interface dominates: ICL clones timbre more closely from correctly transcribed references, X-Vector clones timing, and ICL degrades sharply when a short reference transcript is wrong. Domain explains four times more variance than system choice. Qwen3-TTS leads on intelligibility and pitch register but trails on prosody, and the objective measures rank systems differently from listeners. An off-the-shelf countermeasure flags its clones no more often than genuine recordings, and on ATC radio it fails for every system. Code and audio are released.

    benchmark
  34. arxiv:2610.23677 · cs.CV
    Agentic Physics-Grounded Video Generation via Memory-Augmented In-Context Reasoning
    Xikang Guan, Shaocong Yang, Yanqi Bao, Can Wang +5

    Physical plausibility is essential for realistic video generation. Existing methods impose physical guidance before synthesis via simulations or refined prompts, but this can limit dynamic diversity, miss errors that emerge only after generation, and require repeated regeneration when inconsistencies remain. We therefore study a generate-diagnose-repair paradigm that preserves the original generation capability and selectively repairs physical inconsistencies afterward, raising two key questions: what to repair and how to repair? To address them, we introduce PhysRepair, an agentic video generation framework that combines causal reasoning over the observed physical failure with prior repair experience to enable localized and adaptive correction. Specifically, Spatiotemporal Causal Constraints identify the first invalid physical state and its causally affected states, producing a repair specification that guides physically plausible local regeneration while preserving unaffected content. Experience-augmented In-context Learning combines the ongoing repair context with relevant experience from previous repairs, allowing the agent to adapt its repair decisions across diverse physical failures and accumulate completed trajectories for future cases. Experiments on Physics-IQ Verified across solid mechanics, fluid dynamics, optics, thermodynamics, and magnetism demonstrate that PhysRepair consistently improves the physical plausibility of generated videos over existing methods. Our project page is available at https://xxkkxxx.github.io/PhysRepair-Website/.

    agentagentic
  35. arxiv:2610.23673 · cs.AI
    On the possibilities and limitations of LLM-based multi-agent systems for multi-criteria decision-making in healthcare operations management
    Marco Boresta, Tommaso Giovannelli, Sepehr Namirad, Giulia Riccardi

    Multi-criteria models for healthcare operations need weights that express how stakeholders trade off patient waiting time, provider workload, and operating cost, but these weights are difficult to quantify. Teams of large language model (LLM) agents, each prompted to represent a stakeholder, have been used to supply them. We propose a framework for assessing the reliability of LLM-generated preferences as inputs to multi-objective optimization and quantifying how their uncertainty affects decisions. We study explicit preferences by asking agents to rate the importance of the criteria in an outpatient appointment scheduling problem, and implicit preferences by inferring trade-offs from the agents' pairwise choices between emergency department configurations. The computational experiments show that the explicit preferences respond to the clinic's operating conditions and differ across roles, but deliberation among the agents can erase these differences, whereas simple averaging preserves them. The pairwise choices used to reveal the agents' implicit preferences depend largely on which option is presented first, and we find limited evidence that a statistical model using contextual information alongside presentation order predicts the agents' choices on new comparison pairs better than a model using presentation order alone. We then formulate robust weighted-sum optimization models to account for preference uncertainty. In most cases, these models select different schedules or configurations from those obtained using average preference weights. These results indicate that LLM-generated preferences need an audit of this kind before they are used as optimization weights in healthcare decision support.

    multi-agentagent system
  36. arxiv:2610.23670 · cs.AI
    OwlSight: A Panoramic Monitor for Chain-of-Thought Reasoning
    Lewen Yan, Yanxu Zhu, Jie Zhang, Jing Shao

    Chain-of-Thought (CoT) monitoring offers actionable evidence for oversight AI systems, with the potential to enable early detection and intervention. However, existing approaches are focused on specific tasks, whereas the regulation of AI, particularly reasoning models, requires monitoring a broader range of tasks. We introduce Factor for CoT monitoring, a strategy that supports customizing monitoring requirements in the prompt, which enables a single monitor to cover different tasks by relying on different factors. Furthermore, we introduce two types of factors, fine-grained and coarse-grained (e.g., monitoring whether a specific email address has been leaked versus monitoring privacy leakage), to accommodate diverse user monitoring needs. Based on this, we develop a CoT monitor named OwlSight that relies solely on factor to monitor behaviors related to multiple dimensions, including capability, safety, and trustworthiness. Experiments across 15,971 reasoning trajectories show that OwlSight surpasses general-purpose LLMs by 1.32-30.57 and 11.71-27.81 points at fine and coarse granularity, respectively, and achieves state-of-the-art results on OpenAI's CoT monitorability benchmark. It also generalizes to unseen models' CoTs and remains robust to Factor rephrasings, CoT compression, and obfuscation, offering a promising paradigm for CoT monitoring that can contribute to the oversight of frontier AI.

    benchmark
  37. arxiv:2610.23668 · cs.AI
    TUPLE-Bench: Evaluating and Mitigating Privacy-Utility Trade-offs in Typed Decision Models
    Yiqu Yang, Qiaoyuan Zheng, Yang Tian, Yu Fan +1

    Typed decision models often return more than a selected category. They also expose probabilities for each candidate and, sometimes, a separate confidence score. These fields support downstream decisions but may reveal information beyond a recipient's authorization. We introduce TUPLE-Bench (Typed decision models' Utility and Privacy Leakage Evaluation Benchmark), a benchmark of paired synthetic cases in which recipients may learn a coarse operational category but not its fine-grained cause. Across nine models, instructions to withhold the cause fail to protect it: 91.88%-99.93% of responses reveal the cause through direct selection or residual scores. Even when causes are excluded from the candidate set, full outputs enable 62.85%-75.35% recovery on held-out cases, compared with 51.04%-57.64% using only the selected category (50% chance). Full outputs also reduce mean downstream cost in all 27 model-task comparisons. However, confidence fields increase leakage without consistently reducing cost. We further introduce workflow-constrained privacy quantization (WCPQ), which reduces secret recovery while preserving the behavior and cost of specified downstream workflows. Our findings highlight the need to audit decision interfaces field by field and workflow by workflow.

    benchmark
  38. arxiv:2610.23663 · cs.CV
    PCDFusion: Proposal-Context-Detail Bayesian Rendering for Infrared-Visible Image Fusion
    Rui Liu, Lin Gu, Hesong Li, Ying Fu

    Infrared-visible image fusion (IVIF) aims to combine visible and thermal infrared observations for human perception and downstream vision tasks. Visible and thermal infrared observations are physically complementary but unreliable in different ways. Existing deep IVIF methods mainly improve feature interaction and reconstruction, but often lack an explicit local rule for deciding when visible appearance should dominate and when thermal responses should contribute. To address this issue, we reformulate IVIF as local Bayesian posterior rendering, where fusion is controlled by posterior action weights rather than direct modality mixing. In this paper, we propose Proposal-Context-Detail Fusion (PCDFusion), a visible-anchored framework that implements this formulation through three rendering stages. PCDFusion preserves potentially useful thermal responses, updates exposure-degraded visible regions, and renders reliable band-limited infrared structures into the final luminance. Extensive experiments on public IVIF benchmarks show that PCDFusion improves fusion quality, target recovery, and downstream detection and segmentation performance under challenging illumination. The code is available at https://github.com/BIT-RuiLiu/PCDFusion.git.

    benchmark
  39. arxiv:2610.23662 · cs.RO
    Learning Robot Metareasoning: Adaptive Computation Budgets for Real-Time Control
    Lorenzo Magnino, Antonio Marino, Michael Amir, Daigo Shishika +1

    Planning while moving faces an historical trade-off between deliberation and reactivity. A planner running concurrently with execution needs time to compute, so the robot must commit to an open-loop action chunk for that duration: a longer commitment buys a better plan at the cost of feedback frequency. Existing controllers fix this action commitment, and consequently the computation budget, before deployment. However, the optimal computation time is highly state-dependent (due to factors such as unmodeled dynamics, safety constraints, and marginal optimisation gains), i.e. dynamic, therefore offline static tuning risks leaving systems blind during critical maneuvers or unable to find global effective solutions. We cast the identification of the optimal computation budget as a semi-Markov decision process (SMDP) in which a single decision defines both the robot action commitments and the time allocated to the solver. To solve this, we train a reinforcement-learning policy conditioned on the current state, the executing plan, and inexpensive planner statistics. Across ground-robot navigation, agile quadrotor flight, and contact-rich manipulation, the learned policy changes its budget with the demands of the scenario, discovering efficient reasoning patterns that outperform fixed-budget baselines. https://lorenzomagnino.github.io/learning-robot-metareasoning-review/

    manipulation
  40. arxiv:2610.23650 · cs.LG
    JAXtari: High-Throughput and Easy-to-Modify Arcade Learning Environment
    Quentin Delfosse, Raban Emunds, Paul Seitz, Sebastian Wette +4

    The Arcade Learning Environment (ALE) has been a cornerstone of Deep Reinforcement Learning research, providing a standardized benchmark for evaluating general agents. However, its lack of native support for object-extraction and environment modification restricts research into evaluating whether agents learn robust causal models or rely on brittle visual heuristics. Furthermore, the ALE's dependence on CPU-based emulation creates a significant bottleneck in modern GPU-accelerated training pipelines, limiting both experimental scale and iteration frequency. In this work, we introduce JAXtari, an open-source, high-performance reimplementation of Atari 2600 environments. By leveraging massive GPU parallelization, JAXtari allows PPO agents to finish training on 100M environment steps on a single A100-GPU in under an hour. Beyond speed, it provides object-centric observations and a flexible modification system, allowing researchers to, e.g. develop neuro-symbolic methods or discover novel shortcuts and subsequently train and adapt their agents rapidly, sometimes in only minutes.

    benchmark
  41. arxiv:2610.23642 · cs.CV
    Diversity Without Labels: Task-Agnostic Coreset Selection Across Perception and Action
    Julia Hindel, Benjamin Ramtoula, Kyuhwan Yeon, Nick Heppert +3

    While data collection for model training is relatively easy, curating high-quality datasets from large collections is inherently challenging. Coreset estimation methods address this problem by identifying the best data subsets to train models on. However, existing approaches often require proxy model training, extensive annotations, or task-specific tailoring. Unsupervised alternatives typically exploit geometric properties in foundation model feature spaces, but are limited to pairwise image-to-image comparisons or comparisons against a fixed subset. We introduce CoreDNA, an unsupervised, task-agnostic coreset estimation approach that pools neuron activations from vision foundation models into per-image and per-dataset distributions. By comparing each image's distribution to the dataset-level distribution, CoreDNA can capture both low-level visual structure and high-level semantic information across layers. We find that high-quality samples concentrate in the tails of the last layer's activation statistics, which we validate by relating them to training-based coreset methods. Building on these findings, we develop a computationally efficient selection mechanism to construct task-agnostic, high-quality coresets for selective annotation and model training. Across image classification, semantic segmentation, depth estimation, and imitation learning for end-to-end driving and mobile manipulation, CoreDNA matches or outperforms the state of the art, demonstrating unprecedented generalization across perception and action tasks. These results highlight the effectiveness of compact activation-based representations for data selection.

    manipulation
  42. arxiv:2610.23622 · cs.RO
    Foundation-Model-Driven 3D Scene Graphs for Mobile Manipulation
    Dominic Maggio, Lorenzo Shaikewitz, Aryan Naveen, Kris Hauser +1

    Geometric and Semantic Foundation models are becoming increasingly popular tools for generating spatial maps for robots. Recent work (e.g., FOUND-IT) has leveraged these tools to construct real-time, task-driven, and open-set 3D scene graphs from a single uncalibrated camera. However, while classical robot mapping techniques have undergone extensive system demonstrations, limited work has been done to develop both the fundamental and the systems wide advancements needed to integrate these new types of robot spatial memory into a full, useable autonomy pipeline. In this paper, we bridge this gap, and in doing so we also evaluate the effectiveness and limitations of this type of 3D memory. Towards this end, we integrate FOUND-IT into a mobile manipulator platform and demonstrate it is well-suited to support multiple objectives such as object change detection, support for multiple mapping sessions, determination of navigable paths, and directly support object grasping for the task of scene re-arrangement. Furthermore, we provide nuanced insights into the effectiveness and limitations of a monocular, foundation-model-driven 3D scene graph construction system using observations from rigorous on-robot experiments. Code is available at https://github. com/MIT-SPARK/FOUND-IT.

    manipulationmanipulatorgraspmemoryscene graph
  43. arxiv:2610.23618 · cs.LG
    ThetreStageAI: Multi-View Retrieval-Augmented Generation for Structured Bengali Theatre Production Planning
    Md. Mehedi Hasan Naeem, Mst. Kamrunnahar Ruma, Shakila Sultana, Nafiza Anjum +2

    Staging a theatre performance requires coordinating dialogue, actor movement, stage directions, lighting focus, cue timing, and equipment constraints within the same production plan. This work presents ThetreStageAI, a retrieval augmented framework for structured Bengali theatre production planning. Rather than retrieving from one mixed corpus, the system maintains separate scene, actor-blocking, and lighting views. Bengali text is embedded with BGE-M3 and indexed in view-specific Qdrant collections. Retrieved references are combined with a production brief and passed to Gemini, which generates schema-constrained JSON containing characters, scenes, dialogue, stage directions, blocking, lighting, and optional sound cues. Structural and semantic checks are applied before acceptance, with at most one constrained repair attempt. We also construct a human-guided, LLM-assisted synthetic dataset of 500 short Bengali theatre records that expands to 1,500 retrieval documents. Dataset diagnostics confirm consistent identifiers and valid lighting fields while also revealing substantial dialogue repetition. A paired camera-ready evaluation compares single-combined and full multi-view retrieval under matched settings. The source code, dataset, and evaluation materials are publicly available.

    retrieval-augmentedretrieval augmented
  44. arxiv:2610.23615 · cs.RO
    Beyond Action Prediction: Learning Search-Free Planning in Latent Space
    Guanchen Du, Jinhao Du, Yanpeng Sun, Shan Zhang +2

    Latent world models learn action-conditioned state transitions in a compact representation space, providing a predictive model for visual planning and control. However, planning with these models typically requires costly test-time optimization over action sequences. We propose to replace this online search with a goal-conditioned inverse model that directly predicts actions from current and goal observations. A key challenge is that action prediction alone neither ensures that goal representations encode action-relevant information nor that predicted actions are consistent with the learned latent prediction model. We introduce Hindsight Inverse JEPA Planning (HIJP), a search-free framework that addresses both aspects during training. HIJP relabels future observations along offline trajectories as goals, providing action supervision across multiple temporal horizons. By retaining gradients through the goal representations, this supervision directly shapes the visual representation. HIJP further feeds predicted actions into the latent predictor model and aligns the resulting next-state representations with encoded successor observations, enforcing action-transition consistency. At inference time, the inverse dynamics model and latent predictor models are applied alternately to generate multi-step action sequences without online search. Across four visual control tasks, HIJP achieves 86.8% overall success and a 1.9-16.5x planning speedup over search-based planning.

    world modelaction-conditioned
  45. arxiv:2610.23612 · cs.AI
    SIT-Eval: Situation-Based Interactive Turn-by-Turn Evaluation of Conversational Agents
    Ravil Mussabayev, Saveliy Zavertyaev, Dmitriy Parpulov, Andrey Savchenko

    The rapid proliferation of conversational AI agents necessitates evaluation methods that assess performance across diverse dialogue scenarios, utilizing multiple situational criteria, as appropriate behavior depends on context and encompasses various aspects. Existing dialogue benchmarks typically use static natural conversations, which underrepresent rare situations---a critical safety gap, as low-probability events become frequent failures at production scale. However, limited sample sizes force a choice: either maintain natural dialogues for realism or turn to adaptive situation sampling for comprehensive yet statistically reliable coverage. We introduce SIT-Eval, an adjustable framework for evaluating situation-based conversational agents. We formalize an agent's desired behavior as situational rules: if situation is P, follow rules Q. At each conversation turn, an LLM-interlocutor enacts a uniformly sampled situation. A context-aware LLM-judge leverages all conditionally applicable rules as criteria to evaluate the agent's reply for that situation. The judge is validated using human-verified synthetic golden dialogues. Benchmarking 8 LLMs on 1292 conversation turns reveals failure modes hidden by aggregate metrics: models with similar overall scores vary widely in their adherence to individual rules. All models struggle most with ambiguity clarification and Socratic questioning.

    ai agentbenchmark
  46. arxiv:2610.23610 · cs.CV
    Not Every Mention Is a Claim: Rethinking Object Hallucination Detection in Vision-Language Models
    Hsin-Ling Hsu, Neeraja Kirtane, Kuan-Hao Huang

    Object hallucination in vision-language models (VLMs) is usually measured by checking each object a caption names against the image. Most benchmarks and detectors alike treat every object mention as a claim that the object is present. This assumption fails when a caption names an object without asserting it, as in "a bench for people to sit on", which mentions people without claiming that any are in the image. We call such mentions non-assertive. Treating them as hallucinations inflates the error rates attributed to models that use such language. To separate them from real hallucinations, we introduce NAME-Bench (Non-Assertive Mention Evaluation Benchmark), 25,497 object mentions from captions that five VLMs generate for COCO images. Each mention is labelled concrete (asserted and present), hallucinated (asserted and absent), or non-assertive, and non-assertive mentions are further divided into four subtypes: depictive, purpose, negation and hedged. All the hallucination detectors that we evaluate flag non-assertive mentions as hallucinations, at rates between 14% and 80%. Adding non-assertive mentions to detector training as negatives largely corrects this on LLaVA-1.5. On the four newer VLMs, a comparable correction takes five to ten times more data, and the detectors then catch fewer real hallucinations. These findings show that current detectors conflate mentioning an object with asserting it, and that this conflation is harder to undo on the four newer VLMs than on LLaVA-1.5. Hallucination evaluation should therefore separate what a caption asserts from what it merely mentions.

    benchmark
  47. arxiv:2610.23599 · cs.AI
    From Trajectories to Differentiated Skills: Cognitive Differentiation for Self-Improving Agents
    Yewei Zhao, Derong Xu, Wei Chen, Yin Xu +5

    Large language model (LLM) agents tackle complex tasks through reasoning and interaction with external environments. Agent skills provide persistent, transferable capabilities that guide task planning, inform contextual decisions, support tool use, and enable executable procedures. Yet existing skill-evolution methods largely organize knowledge around skill artifacts, overlooking the functional heterogeneity within individual skills and its implications for knowledge maintenance and use. Inspired by cognitive science, we propose Cognitive Differentiation in Skill Evolution (CDSE), which evolves a unified skill through four functionally distinct yet coordinated components: Meta, Experience, Tool, and Script. CDSE extracts evidence-grounded semantic gradients from successful and failed trajectories, consolidates them by functional role, and applies component-specific updates and validation with cross-component coordination and adaptive invocation. An analysis of 500 top-ranked skill packages shows that 96.8% contain knowledge spanning at least two functional roles. Across six benchmarks, three agent models, and two execution harnesses, CDSE achieves the highest scores in all 36 settings, improving over the Base agent by 7.6 percentage points and the strongest competing method in each setting by 2.6 points on average. Ablation studies further support the importance of functional differentiation and adaptive invocation.

    agenttool useself-improvingbenchmark
  48. arxiv:2610.23598 · cs.AI
    RL-Guided Temporal Localization for Dual-Channel Retrieval in Long-Horizon Agent Memory
    Dabin Sheng, Zhe Wu, Jinming Zhao, Junliang Xing +1

    Time-aware retrieval is crucial for long-horizon LLM agents to effectively leverage ever-growing long-term memory. Existing methods focus on lexical or semantic similarity, leading to target memories being crowded out by temporally mismatched yet semantically similar candidates. Even time-aware approaches inject temporal cues only during memory organization or answer generation, overlooking temporal constraints at the retrieval stage, where candidate selection occurs - the root cause of the displacement. Inspired by the temporal contiguity effect in human episodic memory, we propose TIDER (Temporal Interval-driven Dual-channel Evidence Retrieval), a time-aware agentic retrieval framework. Central to TIDER is a reinforcement learning (RL)-trained temporal interval locator that infers the temporal windows in which the evidence lies. RL optimizes window localization as a discrete, non-differentiable retrieval decision, using composite rewards to convert sparse answer supervision into fine-grained retrieval feedback. These temporal windows first define the temporal retrieval channel; the retrieved candidates are then merged with results from the semantic channel and reranked by proximity to the windows. This temporal grounding enables TIDER to retrieve temporally aligned evidence that would otherwise be drowned out by the volume of semantically similar yet temporally mismatched memories. With a lightweight 1.7B locator, TIDER achieves state-of-the-art accuracy on both long-horizon memory benchmarks - 61.52% on LifeBench (+8.82%) and 78.96% on LoCoMo (+3.66%) - which indicates that temporal cues are crucial for pinpointing target evidence amid a vast pool of semantically similar historical events in long-term daily-life memory.

    memoryepisodic memoryagent memoryagentllm agentagentic
  49. arxiv:2610.23588 · cs.CV
    Enhancing RAG-based Medical VQA with Gated Multimodal Representations
    Daniela Jimenez-Pano, Hugo Jair Escalante, Delia Irazú Hernández-Farías

    Medical Vision-Language Models (Med-LVLMs) have shown strong performance in Medical Visual Question Answering (MVQA), yet they may still produce factually inconsistent responses. Retrieval-Augmented Generation (RAG) can mitigate this issue by incorporating external clinical evidence; however, retrieval is often guided by a single modality, potentially overlooking the joint semantics of medical images and clinical questions. We introduce a multimodal RAG framework that learns a joint query representation through a Gated Multimodal Unit, enabling evidence retrieval to leverage information from both modalities while keeping the underlying Med-LVLM frozen. We evaluate the proposed approach on benchmark datasets. Results show that the multimodal query representation consistently outperforms a reference Med-LVLM and remains competitive with state-of-the-art methods. Our findings demonstrate the potential of multimodal query representations for more effective evidence retrieval in Medical VQA.

    retrieval-augmentedragbenchmark
  50. arxiv:2610.23587 · cs.CL
    Analyzing Occupational Gender Bias across the LLM Training Lifecycle
    Manthan Gadhia, Alexander Hoyle, Karolina Stańczak

    Post-training is widely assumed to reduce the social biases language models inherit from pretraining, yet direct evidence for this assumption remains mixed. In this paper, we examine how gender disparities evolve across the language modeling pipeline. In particular, we study occupational gender bias in OLMo 3, a fully open language model family, across its post-training stages by prompting each checkpoint to generate open-ended narratives about synthetic personas given their occupation and education. We consider two settings, a gender-given setting in which the profile states the persona's gender explicitly and a gender-assumed setting in which it must be inferred by the model. We find that when gender is stated, post-training increases semantic diversity for male-coded profiles, whereas for female-coded profiles the increase is smaller after SFT and not significant after DPO and RLVR. When gender is assumed by the model, post-training affects narratives about female and male personas similarly. The model's own gender assumptions, however, show a consistent female-preferential skew relative to United States labor-force demographics, except for heavily male-dominated occupations. Every successive post-training stage monotonically increases the odds of attributing a female label over the base model. A provenance analysis of the supervised fine-tuning corpus finds only a modest 7.7% female-token surplus, too small to explain the size of these downstream effects. The project code is available under: https://github.com/manthangadhia/llm-lifecycle-bias.

    post-training
  51. arxiv:2610.23583 · cs.CL
    MARS: Memory-Aware Routing for Session-Level LLM Assistants
    Minghan Wang, Jinhang Zuo, Zhongxiang Dai, Zhiyong Wang +1

    Long-term LLM assistants must repeatedly decide what prior context to expose and which model should answer as an interaction evolves. Existing memory systems and model routers usually make these decisions separately, although memory construction changes the evidence, distractors, and difficulty of the resulting input. We introduce MARS, a memory-aware routing framework that uses session state as a shared control interface between memory construction and model selection. A Query and Topic Manager (QTM) estimates compact decision-relevant state that guides a fixed state-conditioned memory policy and an online router, enabling each turn to receive an appropriate memory view and a suitable model under a quality-memory-cost trade-off. We evaluate MARS on RealMem and MultiChallenge with 16 system configurations each, covering long-history memory-driven interaction and bounded sessions with structured context interference. On RealMem, MARS preserves 98.2% of full-context GPT-4o's QA while using only 14.5% of its generation API cost, including both generation and controller calls for MARS, and achieves the strongest measured memory-recall and retrieval scores among compared systems. On MultiChallenge, MARS trails the strongest static system overall but approaches its performance at lower recorded cost on a mechanism-focused subset. Together, these results support a favorable quality-memory-cost trade-off for long-history, memory-dependent interaction.

    memory
  52. arxiv:2610.23574 · cs.AI
    Darwin-Claw: Environment-Adaptive Experience Consolidation for Recursively Self-Evolving Agents
    Mingze Yin, Xiaohan Wang, Dian Li, Longfei Luo +3

    While Large Language Model (LLM) agents have demonstrated remarkable capabilities across complex tasks, they predominantly operate in isolated sessions, lacking the ability to continually learn from historical interactions or adapt to user-specific environments. Existing self-evolution paradigms often indiscriminately inject raw interaction trajectories into memory or skill modules, leading to obscured behavioral patterns and compromised update controllability. Furthermore, as foundational LLM reasoning improves, the bottleneck for agentic evolution shifts from accumulating generic knowledge to effectively managing situational experience. To address this, we propose Darwin-Claw, an environment-adaptive self-evolving framework that systematically distills episodic experiences to drive the comprehensive evolution of operational skills and durable memory. Theoretically grounded in an iterative agent-actor and evolver-critic co-adaptation, Darwin-Claw introduces a Hierarchical Experience Consolidation System that extracts fine-grained error attributions and generalizes them into dynamically updated prototype vectors. This structured situational experience serves as interaction patches for context-sensitive perception and explicit optimization signals for skill refinement. Practically, Darwin-Claw is a readily deployable, automated framework compatible with mainstream agent systems without requiring manual curation. Extensive experiments on WildClawBench demonstrate that Darwin-Claw yields substantial performance improvements, establishing a controllable and stable pathway for the continual, user-centric adaptation of autonomous agents in real-world ecosystems.

    memoryagentautonomous agentagenticagent systemself-evolving
  53. arxiv:2610.23571 · cs.LG
    BMuon: Budgeting Muon's Spectral Geometry for Structured Compressibility
    Rida Assalouh, Tom Jacobs, Rebekka Burkholz, Yani Ioannou +1

    Muon has emerged as a strong optimizer for neural network training and admits an interpretation as steepest descent under a spectral-norm geometry. Yet we find that Muon-trained Transformers can degrade sharply when their MLP hidden channels are pruned after training. To reconcile Muon's optimization performance with structured compressibility, we introduce Budget Muon (BMuon), which modifies Muon's spectral geometry so that hidden channels across blocks compete for a shared allocation budget. The resulting oracle is exactly normalized steepest descent under a family of norms that interpolate between Muon's geometry and structured sparse geometries. Our principal sparsity control is an interpretable global channel-allocation budget, rather than a setting-dependent sparsity-penalty coefficient. For linear factorization models, we show that Muon admits minimum-cost exact representations with redundant hidden width, whereas under the BMuon oracle geometry and a sufficiently scarce budget, every minimum-cost representation is width-minimal. For practical training, we derive a minorization-maximization surrogate for the exact allocation problem. On ImageNet-1K with ViT-S/16, BMuon improves over Muon under gradual structured MLP pruning, and its learned allocations provide a more effective pruning criterion than weight magnitude. On a 1B-parameter LLaMA-style model trained on FineWeb, BMuon improves over Muon under post-training MLP channel pruning.

    post-training
  54. arxiv:2610.23566 · cs.RO
    Enhancing Visual Task-Redirection Attacks on Vision-Language-Action Models via Closed-Loop Optimization
    Chunjin Ye, Guanying Chen, Jianjie Huang, Jianhao Chen +5

    Vision-language-action (VLA) models enable language-guided robot manipulation, but remain vulnerable to visual attacks that undermine user control over robot behavior. Prior studies optimize visual perturbations using fixed target supervision. However, the perturbation can change earlier actions and, in turn, the physical states reached later. As the attacked trajectory evolves, this supervision may no longer match the behavior needed at the states actually reached. Accurately matching these targets can therefore fail to advance the attacker-specified task. Even with suitable supervision, reducing average matching error does not ensure sustained task progress. The robot may still fail at a critical stage, and further updates can disrupt previously successful behavior. To address these limitations, we propose TRACO, a framework that enhances existing targeted visual attacks through closed-loop optimization. Its Execution-Driven Supervision rebuilds supervision for the states reached under attack to guide subsequent updates. Within this framework, Progress-Guided Selection focuses updates where progress stalls while maintaining supervision on the earlier behavior needed to reach that point. Experiments across VLA models and simulation environments show improved task-redirection success with TRACO, highlighting the security implications of closed-loop attack optimization.

    vision-language-actionvlavla modelmanipulation
  55. arxiv:2610.23565 · cs.LG
    Beyond Expected Values: Risk-Sensitive Planning with Distributional Monte-Carlo Tree Search
    Quynh Trinh, Trung Nguyen, Nam Nguyen, Tuan Dam

    Monte-Carlo Tree Search (MCTS) typically stores only scalar mean estimates, which makes it poorly suited for tail-risk objectives such as conditional value-at-risk (CVaR). We introduce Categorical Thompson Sampling with Optimism (CATSO) and Particle Thompson Sampling with Optimism (PATSO), two distributional MCTS algorithms that maintain finite-support empirical backup laws at Q-edges. CATSO represents Q-edge laws with categorical atoms and Dirichlet Thompson sampling, while PATSO represents them with adaptive particles. Both algorithms select actions using the CVaR of a Thompson-sampled Q-edge law plus a polynomial optimism bonus, and propagate scalar continuation values through visit-weighted averages of child Q-edge CVaR scores. For the finite-depth nested CVaR planning target, we prove an $O(n^{-1/2})$ bound on the root value error after $n$ rollouts, which upper-bounds the risk-sensitive simple regret. CATSO adds a fixed-grid discretization bias, while capped PATSO adds a tunable Wasserstein compression bias. Experiments on stochastic planning benchmarks show that distributional Q-edges improve lower-tail return and reduce catastrophic outcomes when mean-optimal and risk-sensitive behavior differ.

    benchmark
  56. arxiv:2610.23554 · cs.AI
    SkillBrowser: Selective Skill Exposure for Scalable Skill-Using Agents
    Jin Ye, Yifan Jiang, Lan-zhe Guo

    Large language model agents increasingly rely on skills to extend their capabilities in complex tasks. However, scalable skill use requires Selective Skill Exposure (SSE): providing relevant procedural content when needed while avoiding unnecessary exposure that may disrupt reasoning. Existing paradigms fall short in complementary ways. Progressive disclosure exposes metadata for a growing skill library, introducing context pollution; retrieval limits catalog exposure, but static retrieval from the initial task query may be misaligned with capability needs that emerge during reasoning and may still inject unnecessary skills. We propose SkillBrowser, a Meta-Skill-managed framework that enables an agent to identify current capability needs, formulate targeted queries, and selectively load skills on demand. Across six SRA-Bench subsets and five LLMs, SkillBrowser improves aggregate accuracy over static retrieval and LLM-based selection, including multi-step tool use, and mitigates accuracy degradation on already-solvable instances. Benefits remain model- and task-dependent, with under-invocation of useful skills and additional latency as key limitations. These results demonstrate the viability of Meta-Skill-managed SSE for scalable skill repositories.

    agenttool use
  57. arxiv:2610.23551 · cs.RO
    Embodiment-Aware Planning for Heterogeneous Bimanual Manipulation
    Sukhvansh Jain, Anurag Maurya, Shivam Vats, Narendra Kumar Dhar +1

    Many manipulation tasks that are effortless for dexterous human hands remain challenging for robotic systems equipped with simple, task-specialized grippers. Industrial platforms commonly rely on end-effectors such as suction cups or parallel-jaw grippers, yet no single mode can reliably satisfy the diverse constraints encountered across different phases of complex manipulation tasks, including pickup, reorientation, insertion, and constrained placement. In this work, we investigate whether complementary simple grippers within a heterogeneous bimanual setup can collectively realize tasks that exceed the capability of any individual gripper alone. Given a single human demonstration specifying the desired object motion, we formulate embodiment-aware execution as a finite-horizon optimization over arm-trajectory assignments. At each segment, the planner determines whether single-arm manipulation or a coordinated dual arm handover is required, jointly reasoning over heterogeneous grasp feasibility, motion constraints, manipulability, and cross-gripper compatibility. Exploiting this formulation, we compute globally consistent assignments with explicit switching regularization. Experiments on a dual Franka FR3 platform equipped with suction and parallel-jaw grippers demonstrate improved feasibility and reliability in tasks that are infeasible under single-gripper execution.

    manipulationdexterousfrankagrippergrasp
  58. arxiv:2610.23550 · cs.RO
    Reason-guided Human Reward Modeling for Preference-based Reinforcement Learning
    Anshul Nayak, Shahil Shaik, Jonathan Smereka, Yue Wang

    Reinforcement learning (RL) policies optimized for task rewards may produce behaviors that are efficient but misaligned with human expectations, safety preferences, or intent. This paper presents a reason-guided preference-based reinforcement learning (PbRL) framework that learns a human reward model from trajectory-level feedback. In addition to preference labels, the framework incorporates reason tags explaining why a behavior is preferred, enabling the learned reward to capture both human choices and their underlying rationale. Compared with preference-only reward learning, the reason-guided formulation provides additional supervision for learning human-aligned rewards and improves the interpretability of the resulting policy. The learned human reward is combined with the environment reward during policy optimization, allowing the agent to improve task performance while producing more interpretable, human- aligned, and preference-sensitive behavior

    agent
  59. arxiv:2610.23548 · cs.RO
    Reinforcement Learning Based Multi-Terrain Adaptive Whole-Body Control for a Quadruped Manipulator
    Yanzhou Jin, Fei Suo, Chang Liu, Yasuhisa Hasegawa +4

    Legged robots equipped with manipulators offer promising capabilities for loco-manipulation tasks in unstructured environments. However, achieving task-consistent end-effector behavior during locomotion remains challenging due to terrain-induced disturbances and the strong coupling between locomotion and manipulation, particularly for small-scale quadruped platforms. A reinforcement-learning based whole-body control framework is proposed for quadruped loco-manipulation on complex terrain. The proposed framework combines a hybrid task representation with task-prioritized whole-body learning to coordinate locomotion and manipulation while preserving task-consistent end-effector behavior. A lightweight quadruped manipulator platform is further developed to validate the proposed framework on a small-scale quadruped system. Real-world experiments demonstrate successful deployment on complex terrains such as slopes up to 30 degrees, stairs, and deformable terrain. These results demonstrate the feasibility of deploying learned whole-body control policy for quadruped loco-manipulation in real-world environments and highlight the importance of task-consistent whole-body coordination for terrain-adaptive loco-manipulation.

    manipulationmanipulatorquadrupedwhole-body control
  60. arxiv:2610.23542 · cs.CV
    Execution Before Memory: Diagnosing Memory Failures in Video World Models
    Weiyang Jin, Jiwen Yu, Sihan Xu, Yuwei Niu +1

    A memory test for a video world model should say whether the scene was set, whether the asked change was carried out, and whether the last frame shows the result. Current tests fold these into one number, so a low score is easy to read as forgetting. When the change is not carried out, the hidden state transition cannot appear, so the test is not really a test of memory. We introduce MIST, an evaluation that separates the change from the memory it should leave behind. The model first sets the scene while the change stays hidden, and only later is asked to show it, with leakage before that point kept under control. This separation locates the bottleneck: the model sets the scene but does not carry out the instructed change. The same gap appears in general memory evaluations. On that basis we propose M3-Agent (Make the Missing Move), which recursively revises the details of the input instruction so that the model successfully carries out the requested instruction. Relative to the baseline, it more than triples the MIST reveal score. On the benchmarks, WRBench's score T rises by 26.1% and MBench-T's score U by 52.7%. The method meets the problem from the execution side: it shows what the frozen model can do once the move is stated, and it sets a stricter baseline.

    world modelmemorybenchmark
  61. arxiv:2610.23540 · eess.SY
    Net-Zero Power UAVs: Facilitating Quantum HRLLC in Multi-Hop UAV Repeaters
    Ali Ranjha, Georges Kaddoum, Halim Yanikomeroglu

    In this paper, we propose a novel joint optimization framework for autonomous, quantum-enabled UAV repeater networks designed to support hyper-reliable low-latency communications (HRLLC) in challenging environments such as post-explosion chemical plant operations. The proposed system uses a symmetric dual-UAV architecture with three sequential quantum channels operating under strict energy harvesting (EH) constraints. We formulate a highly challenging non-convex mixed-integer non-linear programming (MINLP) problem. This model jointly determines qubit allocation, 2-D uncrewed aerial vehicle (UAV) spatial positioning (distance and altitude), and transmit powers to minimize the total decoding error. Crucially, our framework explicitly models real-world quantum physics, including decoherence, gate imperfections, and entanglement degradation. Since the objective function is demonstrably non-convex, to efficiently solve the MINLP, we propose a quantum-resource-optimized GA (QRO-GA). The results of our numerical simulations show that the QRO-GA yields better results than benchmark methods and it is highly efficient, with the convergence speed significantly increasing with the total qubit allocation. Convergence is achieved in as few as 5 iterations for higher qubit allocation. Moreover, the results reveal a critical U-shaped tradeoff, proving that an optimal latency distribution within a strict $80\ μ$s transmission window is strictly required to maintain HRLLC reliability. In addition, we find that the QRO-GA successfully optimizes the 2-D UAV positioning to minimize total decoding error. Finally, it balances consumed and harvested power to achieve net-zero power UAVs system.

    benchmark
  62. arxiv:2610.23537 · cs.CV
    PanoBench: A Panoptic Occupancy Prediction Benchmark for Autonomous Driving
    Yi Feng, Zizhan Guo, Junwu E, Yu Ma +7

    Panoptic occupancy unifies 3D geometry, semantics, and instance identities, but its development is limited by scarce instance-aware annotations. Existing benchmarks mainly provide semantic labels reconstructed from incomplete sensor observations, while panoptic datasets have limited domain coverage or annotation completeness. We introduce PanoBench, a camera-centric panoptic occupancy prediction benchmark spanning synthetic and real-world driving scenes. It combines PanoBench-CarlaOcc, PanoBench-nuScenes, and PanoBench-KITTI360 through canonical data interfaces and unified evaluation protocols, providing multi-resolution annotations with persistent instance identities. The synthetic branch derives complete panoptic geometry from simulated meshes. For real-world data, a new pipeline reconstructs backgrounds using geometric priors and topology constraints, generates foreground mesh assets from multi-frame multi-view images, and fuses them into sparse panoptic voxels. A unified valid-mask protocol evaluates complete thing instances supported by temporal observations while scoring stuff regions at the voxel level. Analyses of observation fidelity, geometric completeness, spatial continuity, and temporal consistency demonstrate the benefits of the resulting annotations over existing occupancy benchmarks. We benchmark representative methods on semantic occupancy prediction, panoptic occupancy prediction, and 4D occupancy forecasting, and study voxel resolution, synthetic-to-real transfer, and storage efficiency. These results reveal complementary strengths of volumetric and ray-based metrics and persistent challenges in real-world instance-aware prediction. PanoBench is available at https://mias.group/PanoBench.

    4d occupancybenchmarkevaluation protocol
  63. arxiv:2610.23533 · cs.CV
    ReVision: Reprogramming Image Generators for Dense Visual Perception
    Yitong Yang, Yinglin Wang, Xuexin Liu, Wei Chen +5

    Recent image generators exhibit rich visual priors that can support dense visual perception. However, existing approaches typically encode multiple perception tasks into a shared parameterization, making joint training sensitive to the mixture and weighting of heterogeneous data and prone to cross-task interference. To address this issue, we introduce ReVision, an instruction-conditioned parameter generation framework built on a frozen multimodal understanding model and a frozen image generator. Given an image and a free-form instruction, the understanding model jointly encodes them into a multimodal image--instruction representation, which a trainable parameter generator maps to transient low-rank updates for the generator. Each image--instruction pair induces a distinct set of transient prediction parameters, enabling instance-specific adaptation while both pretrained backbones remain unchanged. ReVision requires no task-specific heads, explicit task identifiers, or stored adapters. A shared RGB output interface unifies segmentation, depth estimation, and surface-normal prediction. Across diverse dense prediction benchmarks, ReVision achieves competitive performance with state-of-the-art methods while supporting flexible task specification through natural-language instructions. Our analysis further reveals substantial gradient disagreement across tasks, highlighting the conflict introduced when heterogeneous objectives share a single parameterization. Meanwhile, the generated updates exhibit structured task- and input-dependent organization in weight space, further validating instance-level parameter generation for multi-task perception.

    benchmark
  64. arxiv:2610.23525 · cs.AI
    Trapped in Your Story: Evaluating Narrative-Induced Sycophancy in Large Language Models
    Runtai Yang, Dandan Zhu, Kaiwei Zhang, Guangtao Zhai

    Large language models (LLMs) are known to exhibit sycophancy by adjusting their responses toward users' expressed views. However, real users often justify their opinions through reasons, personal experiences, causal explanations, and broader narratives, raising the question of whether such narratives can further steer model judgments. We construct a controlled benchmark of 500 debatable questions across ten domains, where the user stance is held fixed while its justification is systematically varied without introducing new task-relevant evidence. Each question is instantiated under 11 matched conditions: one neutral baseline and five mirrored narrative levels for each of two opposing positions, yielding 5,500 evaluation instances. We evaluate seven LLMs and observe substantial differences in narrative-induced agreement. The effect is not monotonic with narrative elaboration: personal experience can reduce agreement, whereas causal framing and worldview-based synthesis increase it. Controlled GPT-5.1 ablations further show that causal framing produces the strongest positive effect, while claim-preserving bullet presentation also increases agreement over paragraph form. These findings suggest that sycophancy is shaped not only by user stance, but also by how supporting explanations are structured and presented.

    benchmark
  65. arxiv:2610.23524 · eess.SY
    QIEN: Quantum-Inspired Energy-Neutral UAV Trajectory for WPT-Enabled Industrial HRLLC
    Ali Ranjha, Georges Kaddoum, Halim Yanikomeroglu

    This paper introduces Quantum-Inspired Energy-Neutral (QIEN) UAV Trajectory for WPT-Enabled Industrial HRLLC, a novel framework for mobility-integrated trajectory and resource optimization in solar-powered uncrewed aerial vehicle (UAV) relay systems supporting autonomous mining robots in post-disaster environments. We formulate a mixed-integer nonlinear program (MINLP) that jointly optimizes UAV trajectory, finite blocklength allocation, transmit power, and wireless power transfer (WPT) over discrete time horizons to maximize aggregate data throughput. The system operates under strict extreme hyper-reliable low-latency communication (HRLLC) constraints, energy sustainability requirements via solar harvesting, and continuous robot power delivery. To solve this large-scale optimization problem, QIEN employs a hybrid Cooperative Quantum-behaved Particle Swarm Optimization (Co-QPSO) algorithm using quantum tunneling mechanisms and a heterogeneous swarm architecture. Our simulation results demonstrate that the proposed QIEN framework significantly outperforms benchmark algorithms, achieving successful convergence while maintaining energy-neutral operation and satisfying all reliability, power, and mobility constraints.

    benchmark
  66. arxiv:2610.23521 · cs.CL
    Perceive, Solve, Reason: Progressive Capability Formation in Time Series-LLMs
    Xin Qiu, Junlong Tong, Wei Zhang, Xiaoyu Shen

    Large language models (LLMs) can now perceive, describe, and reason about images and audio, yet time series understanding remains underestimated: most work uses LLMs as forecasters and classifiers, reasoning studies are few and narrow, and no suite trains and evaluates the full range of capabilities. In this paper, we build the first such suite, with 337K training and 44K evaluation samples across three progressive capabilities, Perceive, Solve, and Reason, and use it to ask how time series should enter an LLM that must master all three, and in what order. We find that a dedicated encoder is essential: with one, a 4B model surpasses GPT-4o on every perception attribute we test, whereas textualizing the series costs 8--10 points in reasoning accuracy. The capabilities then build on one another: learning them in order beats mixing the same data and rivals reasoning-only training with over three times the reasoning data, but only when an encoder is used. Analyses further reveal that capabilities are transferred through distinct network depths rather than simply stacked, and that a stage converged on its own benchmark is not necessarily ready for the next. Distilled into a simple recipe, these findings yield 4B TS-LLMs that, without any target-domain data, outperform four recent 7B TS-LLMs on seven of eight out-of-domain metrics, by up to 11 points. We therefore advocate treating time series as a first-class modality for LLMs, read through dedicated encoders and taught to perceive before they reason.

    benchmark
  67. arxiv:2610.23508 · cs.LG
    Adaptive conformal prediction on the Grassmannian
    Dimitrios Loukrezis, Zachary J. Grey, Dimitris G. Giovanis

    We introduce a conformal prediction framework for regression with Grassmann-valued responses. The proposed method constructs prediction regions in the form of metric balls directly on the Grassmann manifold, using basis-invariant geometric distances as non-conformity measures and scaling the corresponding non-conformity scores through a learned prediction difficulty estimator. This yields locally adaptive, input-dependent prediction regions that respect the underlying manifold geometry, while retaining the finite-sample marginal coverage guarantee of split conformal prediction under exchangeability. We evaluate several Grassmannian distance metrics and compare their prediction-region efficiency on synthetic benchmarks and an airfoil inverse shape-design problem. The results highlight the advantages of geometry-aware non-conformity scores, while local adaptation improves empirical coverage uniformity across prediction difficulty levels and, in strongly heteroscedastic settings, can substantially reduce prediction-region size while maintaining the target coverage.

    benchmark
  68. arxiv:2610.23480 · cs.CL
    MuninAI: A Self-Hosted Agentic Framework for Independent Research Groups
    Markus Fischer, Jens Meiler, Matthias Elgeti

    Research groups that want a language model grounded in their own literature face a choice between hosted services that receive their unpublished work and self-hosted deployments that keep the data local but give the model no access to the corpus. Munin is an open-source framework for a self-hosted scientific assistant that indexes a group's corpus and its citation graph, and wraps whichever open-weight model the group can serve in an agent harness with specialised reading, search, and computation tools. We claim that the harness, rather than the specific model, is what makes the system useful on domain literature, and we ship the evaluation suite that tests this claim with the software. On 199 paired literature questions, the harness more than doubles the accuracy of the base model, from 0.387 to 0.874 (+0.487, 95% CI [0.407, 0.568]), while naive retrieve-then-answer scores below the unaided model on every backbone tested. Under the harness, roughly twice as many answer claims trace to a passage the model read itself as with fixed top-5 retrieval. When the source paper for a question is removed from the corpus, the system abstains on about nine in ten of the questions it had answered correctly on the Qwen backbones and eight in ten on gpt-oss-20b, and it produces no fabricated local citation across 300 fabricated references. The pattern holds on three backbones from two developers, including a 20-billion-parameter model that runs on a single consumer-grade GPU. With all outbound network access disabled, both Qwen backbones still exceed the published PaperQA2 accuracy. The evaluation suite runs as a single command, so an adopting group can re-run it on its own corpus and model.

    agentagentic
  69. arxiv:2610.23479 · cs.RO
    HexaShell: A Thin Harness for Exploring the Limits of VLM Agents as Robot Coding Agents
    Hongcheng Gao, Alex Zhou, Jay Zhu, Zelin Zheng +12

    Agents built on vision language models (VLMs) now compete with learned policies on robot manipulation leaderboards. We ask how far a coding agent can go when it controls a robot instead of a computer, and where it stops. Our system, HexaShell, keeps the harness thin: it only turns the agent's commands into robot motion. The agent drives a robot with two arms through one command line. It sees three RGB cameras and knows their calibration. After each failed attempt, the agent reads the log, finds the cause, and writes or fixes small programs called tools. In this way it keeps what it learns as code, as a coding agent does when it fixes a failing test. The tools may use only what the robot observes. They use triangulation to find where objects are, relative to each other and to the gripper. If two cameras see the same point, the point lies where the two camera rays meet. If a point lies on a known plane, such as the table or the top of a block, it lies where its ray meets that plane. When the head camera is not enough, the robot moves a wrist camera to get a second view. We build the tools on only ten layouts of one seed per task, then freeze them and test them on ten new layouts of the same seed. We use Codex with GPT-6-Astra at medium reasoning effort. On all 42 RoboDojo tasks, HexaShell reaches a success rate of 39.7% and a progress score of 48.0, averaged in the same way as the leaderboard. This is a new state of the art on RoboDojo. The limit is contact, not reasoning: pushing reaches 92.5% and memory tasks 65.0%, but tight insertion only 8.3%. When the agent fails, it usually sees the failure but cannot make the motion work. We open source the tools evolved so far at https://github.com/HexaFuture/HexaShell.

    manipulationgrippermemoryagentleaderboard
  70. arxiv:2610.23470 · cs.LG
    QD-Rec: Open-Ended Evolution of Recommendations via Quality-Diversity
    Takashi Ishida, Taro Makino, Issa Sugiura

    Open-ended search combined with foundation models has produced advances such as faster algorithms, new results on open mathematical problems, and self-improving agents. We extend this paradigm to recommendation for decision-making, focusing on decisions where the set of possible choices is fixed but the reasons for choosing them are not. In high-stakes decisions, e.g., which company to acquire or which customer to target, comparing diverse, well-reasoned options can be more useful than a single ranked list. We formalize open-ended recommendation as building a diverse, high-quality set of recommendations, each an item together with a hypothesis, a natural-language reason why the item fits the needs of a decision maker. We propose QD-Rec, a quality--diversity method that selects items with a contextual bandit and generates hypotheses with a foundation model from earlier ones kept in an archive as stepping stones. A new recommendation enters the archive if it is novel or better than its nearest neighbor. Across M&A, customer targeting, and paper and book recommendation, QD-Rec gives the highest mean quality and builds the largest archive in most settings.

    self-improving
  71. arxiv:2610.23465 · cs.AI
    S-GAP: Stealthy Graph Attack Primitives for Graph Large Language Models
    Wenda Li, Tongya Zheng, Shunyu Liu, Hanyang Yuan +7

    Graph Large Language Models (GraphLLMs) enable powerful semantic understanding and reasoning over Text-Attributed Graphs (TAGs), while their semantic reasoning may exhibit vulnerabilities to perturbations in node text and graph structure under adversarial attack evaluation. Existing attacks typically construct textual and structural perturbations in separate stages guided by distinct realization intents. However, this separation results in attack-intent fragmentation, causing semantic incoherence between the two perturbation types, thereby undermining attack stealth and making them easier for GraphLLMs to identify and disregard, ultimately weakening attack effectiveness. To avoid attack-intent fragmentation, we propose the Graph Attack Primitive (GAP), an atomic semantic representation that unifies textual and structural perturbations under a coherent attack intent. By carefully calibrating the semantic composition of a GAP, we can achieve strong attack effectiveness while maintaining sufficient attack stealth. Furthermore, we introduce S-GAP (Stealthy GAP), a multi-agent framework that iteratively refines the semantic content of the GAP. The framework unifies GAP planning and realization and continuously refines this process through surrogate evaluation and textual feedback. Extensive experiments across four datasets, three GraphLLMs, and two attack settings show that S-GAP improves ASR over the strongest baselines by 7.02% on average while producing more relationally plausible perturbations, exposing critical vulnerabilities in current GraphLLMs and motivating more robust defenses.

    multi-agentagent framework
  72. arxiv:2610.23449 · cs.AI
    GatewayBench: Unified Capability and Security Evaluation Platform for MCP Gateways
    Yitian Zheng, Zhaorun Chen, Haoyu Li, Bo Li

    AI agents have increasingly interacted with external environments by invoking tools, with Model Context Protocol (MCP) emerging as a standard interface for connecting agents to heterogeneous tools and services. As organizations deploy agents at scale, MCP gateways have become a critical middleware layer for aggregating, governing, and shaping the tool surface exposed to agents. Despite the growing importance of MCP gateways in agentic systems, no prior work has systematically studied how gateway design affects agent capability, efficiency, and security. To address this gap, we introduce GatewayBench, the first unified evaluation framework and benchmark for MCP gateways. We characterize gateway architectures into two major paradigms: flattened gateways and hierarchical gateways. GatewayBench comprises 524 tasks across six domains, together with a pool of 11,118 real-world tools from 251 authentic MCP servers. To emulate large-scale deployments, we systematically inject task-irrelevant tools sampled from this pool into each agent environment, enabling controlled evaluation as the accessible tool surface scales. Based on GatewayBench, we conduct extensive experiments across seven widely deployed MCP gateways and diverse agent configurations, covering both benign and adversarial tasks. Our results reveal substantial differences across gateway designs in task completion, tool-use efficiency, and resilience to attacks, and further demonstrate that these effects interact strongly with the underlying model and agent harness. These findings highlight the MCP gateway as a consequential yet underexplored component of the agentic AI stack. By enabling controlled evaluation of gateway architectures under realistic tool ecosystems, GatewayBench provides both practical guidance for MCP gateway design and a rigorous testbed for developing more capable, efficient, and secure agentic systems.

    agentai agentagentictool-usebenchmarkevaluation framework
  73. arxiv:2610.23443 · cs.AI
    ScoutRoute: Efficient Mixture-of-Experts Routing Using Internal Language Model Signals
    Rajdeep Pandhare, Delzaan Bharucha

    Efficient expert routing is a fundamental challenge in Mixture-of-Experts (MoE) systems, particularly for heterogeneous collections of independently trained Large Language Mod els (LLMs). Existing routing approaches typically rely on jointly trained gating networks, prompt-based predictors, or computationally expensive multi-model inference, limiting their flexibility and efficiency. This paper presents ScoutRoute, a lightweight post-hoc routing framework that predicts expert competence using internal language model signals extracted from a single forward pass through an inexpensive scout model. Hidden-state representations are used to estimate cross-model competence through lightweight logistic probes, while logit-derived confidence measures are incorporated within a domain-aware routing strategy for efficient expert selec tion without modifying the underlying language models. ScoutRoute isevaluatedon fiveopen-weightlanguagemodels across fourbenchmark datasets covering mathematical reasoning, code generation, commonsense reasoning, and multidis ciplinary knowledge. Experimental results show that ScoutRoute consistently outperforms individual routing signals while requiring only a single scout-model inference. The proposed competence probes also achieve superior calibration compared with conventional confidence measures, demonstrating that internal language model representations provide transferable information for accurate and efficient expert routing across heterogeneous language models

    benchmark
  74. arxiv:2610.23440 · cs.CV
    ShadowCast: On-the-Fly Context-Aware Shadow Augmentation for Remote Sensing Image Segmentation
    Yechan Kim, Dongho Yoon, Jongmin Joo, JongHyun Park +1

    Shadows cast by elevated objects alter the spectral appearance and obscure the boundaries of objects in remote sensing images, thereby degrading semantic segmentation performance. Most existing approaches handle shadows as degradations to be detected and removed, requiring additional costs from dedicated networks and shadow-specific annotations. We propose ShadowCast, which takes the opposite approach: it synthesizes shadows and injects them during training for remote sensing segmentation. It extracts masks of elevated objects, such as buildings and trees, directly from the segmentation labels and estimates an image-specific cast-shadow direction cue by probing the brightness around them. It then projects synthetic shadows along the inferred cast-shadow direction, requiring no learned network and no acquisition metadata. Experiments on three widely used remote sensing benchmarks, Vaihingen, Potsdam, and LoveDA, show that ShadowCast consistently outperforms the baseline across diverse state-of-the-art segmentation architectures. Moreover, it brings additional performance gains when combined with standard augmentation and regularization techniques, indicating that stochastic shadow injection serves as a simple and complementary augmentation strategy for remote sensing segmentation.

    benchmark
  75. arxiv:2610.23438 · cs.RO
    No Eyes, No Problem: What Does Touch Contribute to Blind Baoding Ball Manipulation?
    Ayush Deshmukh, Nalin Agarwal, Elle Miller, Sethu Vijayakumar

    Humans manipulate objects within their hands through touch alone with remarkable ease, but for robots, achieving sustained real-world manipulation without vision remains an open challenge. Although reinforcement learning (RL) can generate sophisticated blind manipulation policies in simulation, transferring them to a real dexterous hand remains hindered by continuously changing contacts, motion-induced tactile information drift, and the gap between simulated and physical dynamics. We study this problem through Baoding ball rotation on the Shadow Dexterous Hand Lite, using only proprioception and binary tactile feedback as observations. We achieve successful zero-shot transfer through a combination of a sensor-aligned, coupling-aware embodiment in simulation, tactile calibration, and two novel episodic domain randomisations: slew-rate action limits and STuck-AT Taxels. On the physical hand, the policy achieves sustained blind Baoding rotation, completing 225 consecutive $\mathbf{180^\circ}$ rotations (112.5 full rotations) at $\sim0.36 \;\text{rot/s}$ in a single continuous trial without any vision or object-state information. Furthermore, we evaluate the contribution of tactile sensing relative to proprioception alone and identify the hand regions where tactile information is most important. This is the strongest reported hardware result for Baoding ball manipulation in both rotation count and speed, irrespective of the sensing modality used. By effectively bridging the sim-to-real gap, our framework sets a new benchmark for sustained tactile-driven dexterous manipulation on physical hardware.

    manipulationdexteroustactilesim-to-realbenchmark
  76. arxiv:2610.23436 · cs.LG
    Generative Policy Improvement from Observational Data
    Jonas Schweisthal, Dennis Frauen, Stefan Feuerriegel, Emma Brunskill

    We study how to improve decision policies from observational data to generate better actions when actions are complex, structured objects, such as text messages or images. This task is subject to two major challenges: (i) action-level overlap is often too weak for reliable off-policy evaluation and improvement, and (ii) generating new structured actions often requires expensive training or fine-tuning (e.g., of an LLM) and can be ill-defined due to information loss in action representations. To address both, we propose PolicyLift, a three-stage framework for generative policy improvement with structured actions from observational data. In Stage 1, PolicyLift learns a small set of discrete action codes that capture high-level strategies in the observed action space and are tailored to support policy improvement under limited overlap. In Stage 2, it learns a context-dependent policy over these action codes. At deployment time, Stage 3 "lifts" the learned code-level policy back to the original action space by generating actions from the historical behavior system in a code-conditioned way. We show theoretically that improving the policy over action codes guarantees improvement of the induced policy in the original action space. We demonstrate PolicyLift across experiments in synthetic and real-world text generation tasks with consistent policy improvements over the behavior policy.

    policy evaluation
  77. arxiv:2610.23434 · cs.LG
    Bridging Compressed Communication and Local Steps via Spectral Method in Federated Learning
    Dmitriy Bystrov, George Chirkov, Daniil Medyakov, Dmitry Bylinkin +1

    Federated Learning (FL) enables collaborative training of complex models across large-scale networks of edge devices while keeping data localized. However, data heterogeneity and high communication overhead remain fundamental bottlenecks. Recently, the spectral optimizer Muon has demonstrated remarkable advantages over Adam-like methods in centralized training. While recent efforts have attempted to transition Muon to FL, naive integrations into traditional architectures fail short. Crucially, existing approaches do not support the the communication compression and local steps --- the two foundational techniques of efficient FL. To address this limitation, we introduce a novel spectral FL method that utilizes both local updates and compression. We establish rigorous convergence guarantees under mild assumptions, notably eliminating the restrictive bounded heterogeneity condition. Empirical evaluation on standard image classification benchmarks validates the efficiency and robustness of our approach.

    benchmark
  78. arxiv:2610.23433 · physics.app-ph
    Hydrogen Ion-Cut Single-Crystal Diamond Membranes for Wide and Ultrawide Bandgap Semiconductor Heterointegration
    Xuanyu Zhou, Chenyu Wang, Dmitry Shinyavskiy, Lu Chen +9

    Single-crystal diamond membranes (SCDms) were created using a hydrogen ion-cut process combined with a two-step post-implantation annealing, electrochemical etching, and micro-transfer printing. Monte Carlo simulations comparing H, He, C, and O implantation predicted a narrower vacancy-distribution profile for hydrogen under the implantation conditions examined. Raman spectroscopy was used to evaluate implantation-induced lattice damage and its recovery during annealing and membrane release, while optical bandgap analysis provided further evidence of restoration of the diamond lattice structure. After delamination and removal of residual graphitic material, freestanding SCDms with improved crystalline quality were obtained. The SCDms were subsequently transferred to GaN, SiC, AlGaN, and AlN substrates without an adhesive layer. The successful formation of diamond/WBG and diamond/UWBG heterostructures demonstrates the feasibility of integrating the SCDms with a range of semiconductor platforms. Overall, these results establish hydrogen implantation as an effective route for producing high-quality SCDms for future diamond-based heterogeneous integration applications.

    heterogeneous integration
  79. arxiv:2610.23423 · cs.CV
    Momentum: Persistent Dynamics for Video World Models
    Nedko Savov, Stefan Maria Ailuro, Danda Pani Paudel, Yilun Du +2

    Video world models must capture how a scene evolves beyond the current camera view to generate consistent observations when its content reappears. Existing approaches rely on observation history, static point clouds, or image caches constructed from past frames. These representations are often expensive and, more importantly, cannot update dynamic scene content outside the field of view: generated vehicles and pedestrians may be forgotten after leaving the field of view or reappear in physically inaccurate states. We represent an evolving scene through synchronized videos from reference-camera rigs, each comprising eight cameras that jointly cover the surrounding scene. We connect these rigs in a graph and propagate video generation between neighboring rigs to coordinate scene dynamics across locations and time. We introduce Momentum, a two-stage pipeline: WorldUpdate jointly generates reference-camera videos as a shared memory of scene dynamics, and WorldInteract grounds camera-controlled video generation in this evolving memory, allowing moving content to remain represented outside the target view and reappear consistently. To support training and evaluation, we develop a CARLA pipeline that captures synchronized videos from reference camera rigs and a moving target camera, together with an evaluation framework measuring consistency within the generated world state and the target observations.Experiments on generated street scenes show that grounding target videos in scene dynamics generated by Momentum improves dynamic scene consistency by 126.2% in unseen scenes over static image conditioning, while also improving visual fidelity. Our models and source code will be made publicly available.

    world modelmemoryevaluation framework
  80. arxiv:2610.23421 · cs.LG
    Residual Stochastic Concept Bottleneck Models
    Stephen Marcou, Mateo Espinosa Zarlenga, Sonia Laguna, Moritz Vandenhirtz +2

    Concept-based models (CMs) learn human-interpretable intermediate representations that enable users to understand the model's reasoning and steer its predictions through test-time concept interventions. However, in realistic settings, their performance can be limited by concept incompleteness, where the training concepts fail to capture all task-relevant information. Adding an unconstrained residual pathway to the model's predictive process can overcome the consequences of incompleteness. However, this introduces an inscrutable, leak-prone prediction pathway that may weaken concept interventions, as these often leave residual variables unchanged. To address this limitation, we introduce Residual Stochastic Concept Bottleneck Models (Res-SCBMs), a family of concept-based models that jointly model concept and residual variables through a multivariate Gaussian distribution. Res-SCBMs learn a covariance structure that explicitly captures dependencies between the known concepts and learnt residual variables, allowing interventions to propagate to statistically dependent residual variables through Gaussian conditioning. This extends the intervention capabilities of previous CMs without requiring concept completeness during training. Our results show that, across multiple concept-incomplete benchmarks, Res-SCBMs improve predictive performance over competing baselines while retaining effective test-time concept interventions.

    benchmark
  81. arxiv:2610.23416 · cs.CV
    GeoReason: Bridging Logical Reasoning and Spatial Fidelity in Remote Sensing Segmentation
    Cheng Fan, Han Wu, Tingzhang Luo, Minjing Dong +2

    Multimodal Large Language Models (MLLMs) have significantly advanced instruction-driven perception through the "Embedding-as-Mask" paradigm. However, extending this capability to Remote Sensing (RS) reasoning segmentation remains exceptionally challenging under cluttered geographical contexts and extreme scale variations, often leading to Spatial Fidelity Degradation and Target Misidentification. We identify a critical yet underexplored bottleneck: the sequential alienation of the segmentation token [SEG] from its initial visual grounding within the autoregressive generation paradigm. Produced only at the end of long reasoning chains, [SEG] relies on visual evidence that has been progressively diluted by linguistic abstraction and autoregressive noise. This information decay is particularly catastrophic for RS targets with minuscule pixel footprints, where even marginal fidelity loss leads to localization failure. To address these bottlenecks, we propose GeoReason, a novel framework designed to fortify logical deduction and preserve spatial integrity. First, we introduce an Anticipatory Prior mechanism that injects the [SEG] token directly into the initial user query, shifting it from a terminal output to a primary condition. This paradigm shift transforms the token into a persistent spatial anchor, preventing information decay during long-form linguistic inference. Second, we enhance visual reasoning via a Location-aware Chain-of-Thought (CoT) that enforces coarse-to-fine localization using bounding boxes, coupled with a Latent Patch-to-Token Alignment (PTA) Loss that explicitly aligns the latent segmentation token with physical RS textures. Extensive experiments demonstrate that GeoReason consistently outperforms state-of-the-art methods on the Earthreason and RRSIS-D benchmarks. Code is available at https://github.com/Fanc0126/GeoReason.

    benchmark
  82. arxiv:2610.23415 · eess.SY
    ConQUESTA: Topology-Aware Post-Training Structural Pruning of Quantum Neural Networks
    Saim Rehman, Muhammad Shafique

    Deploying quantum neural networks (QNNs) on sparse hardware requires preserving task accuracy while reducing compiler-induced two-qubit gates, SWAPs, and depth. We present ConQUESTA (\underline{C}onstrained \underline{Qu}antum \underline{E}ntangler \underline{S}election with \underline{T}opology \underline{A}wareness), a post training pruning framework that combines task saliency with compiler cost, screens candidate entanglers, and searches complete pruning masks under an explicit accuracy constraint. Unlike one-shot ranking, each mask is transpiled as a whole across multiple coupling topologies. Across 4-, 6-, and 8-qubit QNNs and eight model seeds at 55% pruning, ConQUESTA achieves mean resource reductions of 61.6% on Breast Cancer, 72.0% on Digits01, and 82.2% on Iris, exceeding the strongest matched simple baseline by 3.5, 1.7, and 8.9 percentage points, respectively. Fixed-mask recovery raises feasibility from 16/24 to 22/24 runs. In a fixed-k=4 scaling study on a scaling probe of 4 to 12-qubits, the raw mask space grew from 495 to 12.1M while the evaluated masks increase only from 189 to 305. Resource-weight changes alter depth/routing priorities with at most a 0.44-point mean validation shift.

    post-trainingpost training
  83. arxiv:2610.23411 · cs.CV
    Can We Use Saliency Maps for the Evaluation of Generative Image Models?
    Eleonora Ristori, Luca Bindini, Marco Lippi, Paolo Frasconi

    Evaluating generative image models remains challenging, as commonly used metrics depend on fixed pretrained feature extractors optimized for different tasks that do not always align with human realism perception across datasets and model families. To address this limitation, we investigate a human-inspired evaluation framework that avoids representations learned for unrelated tasks. Specifically, we train a generator-specific classifier to distinguish real images from samples produced by a given model and analyze its decisions using saliency maps. Across diverse benchmarks, we observe that these classifiers learn generator-specific artifacts rather than a universal notion of "fakeness", exhibiting strong accuracy and limited cross-generator transfer. This makes saliency analysis particularly informative: the resulting attribution maps highlight visual cues and structural irregularities characteristic of each model's outputs. By combining classification accuracy and saliency information, we obtain a composite score aligned with human judgments, without relying on large pretrained backbones and using only real and generated samples from the evaluated model.

    benchmarkevaluation framework
  84. arxiv:2610.23409 · cs.LG
    Flash RNN-T: Memory-Efficient Joint Network and Transducer Loss Computation
    Mahmoud Ashraf

    The recurrent neural network transducer (RNN-T) remains a practical choice for streaming speech recognition, but training requires large intermediate tensors. We present Flash RNN-T, a joint and loss implementation combining specialized kernels with optional tiled recomputation over the full alignment space and vocabulary. Tiled and untiled Flash form the entire measured throughput-memory frontier on medium and long workloads. On H100, untiled Flash achieves 24-39% higher throughput with 26-56% less memory than the fastest baseline on these workloads wherever it fits. Tiling reduces memory further through recomputation. On medium, tiled Flash uses ninefold less memory than that baseline at 16% lower throughput when both use one backward call per batch. On long, it fits where all tested alternatives with that backward schedule run out of memory with 80 GB. These comparisons use each implementation's lowest-memory configuration reaching at least 95% of its peak throughput.

    memory
  85. arxiv:2610.23408 · cs.CV
    LithoWorld: An Endpoint World Model for Lithography Process-Window Recovery
    Botong Zhao, Fang Yu, Xinyuan Chen, Yue Lu

    Lithography process drift can move a fabricated pattern outside its process window, producing unqualified scanning electron microscopy (SEM) morphology. We present LithoWorld, an endpoint world model for SEM-guided parameter recovery. Interior representative targets jointly determine qualified reference appearance and correction supervision. A causal predictive branch aligns currentimage features with endpoint representations, while a conditional flow policy generates a two-dimensional correction as an action chunk. The model learns from 173 groups of measured SEMs organized as focus-exposure maps, preserving unmeasured cells as unknown. On 200 test pairs from the 73-group subset, LithoWorld achieves 88.5% qualified-region success and 46.0% interior success, exceeding the strongest comparison method by 2.5 and 6.0 percentage points, respectively. Ablations separate the effects of target selection, endpoint alignment, and output parameterization. Qualitative SEM retrieval illustrates recovery to qualified interiors without requiring exact reference-coordinate matching.

    world model
  86. arxiv:2610.23402 · cs.LG
    Contracting with an Agent under Model Mismatch
    Elif Yılmaz, Jakub Tłuczek, Victor Villin, Christos Dimitrakakis

    In the classical principal-agent setting, the principal delegates some task to the agent, who receives a payment conditioned on the outcome of the task, which in turn benefits the principal. To align their objectives, the principal designs a contract that incentivizes the agent to take actions beneficial to the principal. In practice, however, the agent may act under misspecified beliefs about the underlying environment, leading to systematically suboptimal decisions. In this work, we study this problem through a principal-agent Markov Decision Process (PAMDP), in which the principal repeatedly interacts with an agent whose transition model may differ from the true environment dynamics. The principal has access to the true transition model and designs contracts that account for the agent's belief and incentives. We characterize contract design under model mismatch and develop a framework that enables the principal to induce desirable long-term benefits even though the agent has a potentially inaccurate understanding of the environment.

    agent
  87. arxiv:2610.23400 · cs.CL
    On the Tip of the Tongue: First-Token Cueing as a Behavioral Probe of Recall Failures in LLMs
    Jaewon Sok, Jewon Yeom, Seonghyeon Park, Jeongjae Park +2

    Large language models often answer factual questions incorrectly in closed-book settings, and such errors are usually attributed to missing knowledge. Yet models often encode answers they fail to produce. We show that many closed-book errors are first-token recall failures: the model commits to a wrong first token, but once given the right one it completes the answer and often recognizes it as correct. We study this with First-Token Cueing, a behavioral probe that prefills the response with the first token of the gold answer. Model uncertainty peaks at the first answer token, where it already predicts correctness, and among gold prefixes of any length the first token alone accounts for the largest share of recovered errors. Across 16 open models, this one token corrects between a fifth and a half of the errors on multi-token answers in English QA and improves accuracy in all seven languages of a multilingual benchmark. Most of the errors it corrects are also solved by the model's own sampled reasoning, a criterion for encoded knowledge, and models recognize correct starts they do not produce: they continue a correct start more confidently than an incorrect start of the same answer type and prior probability, yet rarely find it without the gold answer. These results locate a recoverable class of factual errors at the first decoding step, where inference-time methods may gain the most.

    benchmark
  88. arxiv:2610.23396 · cs.RO
    AWAM: Active Trajectory-Conditioned View Imagination for Robust Robot Manipulation
    Jiawei Geng, Zhenyang Liu, Yuzheng Wu, Yanwei Fu

    Rapid progress in embodied intelligence has enabled increasingly capable robots to follow language instructions and manipulate diverse objects. However, strong results in structured environments do not automatically transfer to open-world conditions, where clutter, occlusion, and changing viewpoints can obscure the spatial relationships required for action. To address this perceptual gap, a virtual active-perception paradigm is introduced. AWAM, an Active World-Action Model actively plans where to observe and imagines what will be seen along the planned trajectory before acting. Task-related virtual camera motion is predicted and used as a structured query into a geometry-conditioned world model, producing a reusable spatial prior for manipulation, which is later compressed into a fixed memory and fused with live observations through cross-attention for diffusion-based action prediction. AWAM achieves competitive success rates with 75.9% on LIBERO-Occ, 62.4% on LIBERO-Plus, 76.5% on RLBench-OG Occlusion1, and 94.4% in real-robot trials. These results support improved robustness to the evaluated occlusions and distribution shifts, addressing part of the gap between structured manipulation and less controlled environments.

    embodiedmanipulationliberoworld modelmemory
  89. arxiv:2610.23394 · cs.LG
    Forecasting Voltage Time Series under Anomalies Using a WGAN Model With Lipschitz Control and Progressive Kernel Critic Strategy
    Najmieh Sadat Safarabadi

    In time-series forecasting tasks, the use of recursive forecasting on high-dimensional data substantially increases the complexity of the prediction process. The main challenge is that long-term dynamics becomes harder to model, since the sequences show greater temporal variability. To counteract this limitation and improve the modeling of long-term dynamics, and more importantly, address high-dimensional recursive forecasting, this paper uses a Wasserstein GAN equipped with a progressive Lipschitz penalty and an adaptive kernel. This model is capable of taking a window of the data stream and forecasting the next values, subsequently reconstructing the time-series stream and dynamically adapting its kernel to each new window. The proopsed model is a deep learning framework based on a Wasserstein GAN (WGAN-CLP) with an adaptive kernel to model alterations in streaming voltage time series. To establish the methodological foundation, we review the WGAN-LP formulation another variant model in this class and extend its framework through an adaptive kernel critic model. We then evaluate this extended model using power ratios, correlation coefficient (CC), and relative root mean squared error (RRMSE). The experimental results demonstrate that the proposed model outperforms benchmark methods in both accuracy and robustness. The evaluation shows promising performance of the WGAN-CLP with an adaptive kernel framework in forecasting sequence alterations across the voltage and energy time-series datasets.

    benchmark
  90. arxiv:2610.23378 · cs.RO
    Look Before You Act: Bridging Active Perception and Vision-Language-Action Models via Observation-Aligned Data Adaptation
    Jiawei Geng, Zhenyang Liu, Yuzheng Wu, Yongchong Gu +1

    Robots manipulating objects in clutter must decide not only how to act, but also where to look. Although active camera motion can expose an occluded target, it may move the robot into an initial observation-state distribution that differs from the manipulation policy's training distribution. We identify this problem as first-frame distribution sensitivity: under perturbed starts in LIBERO, naively prepending camera repositioning to pi_0.5 reduces success from 46.3% to 11.5%. We introduce ActiveAlign, a data-adaptation framework that trains a Vision-Language-Action (VLA) policy from the states at which active perception terminates. A learned viewpoint module first generates a camera trajectory. Offline, ActiveAlign uses closed-loop alignment to connect the trajectory endpoint to an existing demonstration, replays the remaining manipulation trajectory, and retains successful continuations for fine-tuning. At deployment, the same viewpoint module precedes the adapted VLA without modifying its architecture. Across 40 LIBERO tasks, ActiveAlign achieves 70.7% success with pi_0 and 72.7% with pi_0.5, outperforming their standard-training baselines by 40.0% and 26.4% respectively. Matched-state and data-matched controls support gains beyond prefix shortening and generic viewpoint augmentation; OpenVLA and GR00T N1 also show the same trend. Success rate under physical occlusion in real world increases from 35/90 to 83/90. These results show that active perception benefits manipulation when policy training is aligned with the observation and robot-state distribution induced by viewpoint selection.

    vision-language-actionvlamanipulationopenvlagr00tlibero
  91. arxiv:2610.23377 · cs.CV
    FloodVLM-Pix: Language-Guided Pixel-Level Flood Understanding
    Chengyang Zhang, Zhihao Wang, Zhiyu Long, Xueqian Wang +3

    As climate change intensifies, floods are becoming more frequent and destructive, making accurate flood understanding an urgent priority. However, existing vision-language models perform poorly on flood-related tasks, suffering from attentional bias and inaccurate descriptions of flood magnitude. To bridge this gap, we construct the first large-scale remote sensing flood multi-modal benchmark that supports language-level flood understanding and pixel-level flood segmentation. The benchmark contains 131,200 flood image patches and corresponding 443,100 textual descriptions for flood severity and evolution. Based on this benchmark, we propose FloodVLM-Pix, a language-guided pixel-level flood understanding model that includes a flood prior module and a multi-temporal flood pixel decoder to achieve accurate flood understanding. The flood prior module automatically distills sub-region flood spatial priors and inputs priors into FloodVLM-Pix, solving the attentional bias problem in flood language understanding. The flood pixel decoder introduces a novel general-specific feature fusion mechanism to address performance degradation caused by interference (e.g., clouds or mountain shadows) during flood segmentation. Experiments on FloodVL-Pix demonstrate that FloodVLM-Pix improves both flood description quality and pixel-level flood segmentation performance compared with state-of-the-art VLMs and flood segmentation models. Our code and data will be released at https://github.com/Haomiaoyanmeng/FloodPix.

    benchmark
  92. arxiv:2610.23374 · cs.CV
    MOSS-Image Technical Report
    MOSI. AI

    Recent closed-source image generation systems can produce diverse visual content from simple compositions to information-dense graphics, yet open-weight text-to-image models still struggle with dense text and intricate layouts. To narrow this gap, we introduce MOSS-Image, an open generation system that combines high-fidelity visual synthesis with intent-aware composition. First, we pretrain a 5B single-stream diffusion transformer (DiT) from scratch with a carefully designed recipe, enabling bilingual long-form text rendering and precise layout control. Second, we develop an intention-aware agentic prompt enhancement (PE) module that translates diverse user requests into detailed content and layout specifications. Together, they enable MOSS-Image to generate text-rich and layout-coherent images. We further introduce LongLongTextBench (LLTB) for evaluating long-form English and Chinese text rendering, and RealLayoutArena (RLA) for end-to-end layout-rich generation. MOSS-Image achieves state-of-the-art performance among open-weight models on LLTB, with an average text coverage of 96.3%. On RLA, MOSS-Image ranks second in the end-to-end system comparison, while its backbone ranks first among generators evaluated with shared enhanced prompts. We will release the model, training recipe, and evaluation framework to support further research and promote the text-to-image community.

    agenticevaluation framework
  93. arxiv:2610.23372 · cs.RO
    Beyond Success Rate: Interventional Evaluation of Memory in Vision-Language-Action Systems
    Keyuan Luo, Yi Xie, Wanyun Ling, Yuanzhe Li +2

    When we say that a robot uses memory, does that memory actually play a role? Success-rate ablations alone cannot establish the use of task-relevant facts: they change information and computation together, while aggregate rates can hide offsetting gains and losses. We propose an Interventional Memory Evaluation Protocol (IMEP) for vision-language-action (VLA) systems to address four questions: (1) Does memory provide a net benefit? (2) Under what conditions does memory help? (3) Do changes to memory alter task outcomes? (4) Are these effects specific to task-relevant facts? At the retrieval interface, we compare target-fact substitutions with fact-preserving replacements, minimal numerical perturbations, and exact replicas, while assessing decision correctness separately from motor execution success. We evaluate within-episode, cross-episode, and agentic memory. In cross-episode learned-bank experiments, a one-unit-in-the-last-place (ULP) intervention, including its installation procedure, changes 11.9% of seed-paired outcomes, while changing key facts produces no clear additional effect over fact-preserving replacements. To examine when memory helps, we introduce ChainBench, where correct decisions require historical information unavailable in current observations. The full memory configuration raises task success from 39.1% to 74.3%, matching the oracle-decision condition under shared recorded executions. These results show that correct history can support robot decisions, but establishing whether a memory system realizes this benefit requires fact-specific evidence beyond aggregate success rates.

    vision-language-actionmemoryagenticevaluation protocol
  94. arxiv:2610.23362 · cs.AI
    Monitor-Validity Failures: LLM Safety Evidence Paths at Evolving Framework Boundaries
    Junchi Liao

    Modern LLM applications rely on runtime monitors that inspect tool calls, traces, streams, detector outputs, and verifier inputs. This creates an interface-reliability problem: a framework path may remain operational while policy-relevant evidence never reaches its decision point. In one representative case, a regex detector returned a positive match, but the framework stranded the result before the block decision. This differs from jailbreaks and detector errors because the evidence is already recognized yet disconnected from enforcement. We study this repair-boundary signature with canary contracts: parent/fixed confirmation units that place a known payload only on the changed evidence surface and record monitor visibility together with the downstream decision for action-boundary cases. Applying the protocol to public framework repairs, we confirm 18 evidence-path failures across 11 LLM or agent safety project groups and four implementation ecosystems. Nine are established-boundary repairs, eight are new-surface policy-coverage gaps, and one is a boundary-only evidence case. Fourteen reach E1 enforcement boundaries: seven native action outcomes, six configured enforcement boundaries, and one synthetic-consumer consequence. The remaining cases affect verifier, safety-prompt, or observability channels. The cases form carrier, adapter, authority, and surface families and show how ordinary application behavior can remain intact while a safety-specific evidence path misses its decision. Exact revisions, deterministic outcomes, negative controls, and a frozen prospective audit document the confirmation rule. Monitor assumptions therefore need regression tests at the schemas, streams, tools, and adapters that frameworks evolve.

    agent
  95. arxiv:2610.23360 · cs.CV
    UniPolar: Joint Image and Polarimetric Restoration under Compound Degradations via Physical Consistency Learning
    Yang Jin, Guangyu Guo, Linghao Shen, Binglu Wang

    Polarimetric image restoration aims to recover polarization angle images and reliable polarimetric quantities from degraded observations. Beyond visual fidelity, it also requires preserving the physical relationships underlying reliable polarimetric estimation. However, existing methods typically address either device- or environment-induced degradations in isolation, limiting their applicability to color polarization filter array (CPFA) imaging, where both often coexist. Such compound degradations jointly perturb image intensities and cross angle polarization differences. Consequently, improved image fidelity alone does not necessarily ensure accurate Stokes, DoLP, and AoLP recovery. To address this challenge, we propose UniPolar, a coupled Image Stokes framework for joint image polarimetric restoration from compound-degraded CPFA observations. Specifically, UniPolar employs two complementary streams to reconstruct four polarization-angle images and directly predict Stokes difference components. To coordinate these streams, Degradation Conditioned Cross Domain Interaction combines Dual Domain Frequency Interaction for frequency selective evidence exchange with Degradation Conditioned Modulation for condition adaptive feature correction. Furthermore, at the output level, Image Stokes Physical Consistency Learning aligns image derived and directly predicted Stokes components while regularizing angular fidelity and the physical feasibility of DoLP. On the compound-degradation benchmark, UniPolar achieves average image and DoLP PSNR of 31.522 dB and 24.097 dB, respectively, with an average AoLP error of 16.833 degree. These results underscore the importance of jointly considering visual fidelity and polarimetric reliability, offering a physically grounded perspective on learning based restoration under compound degradations.

    benchmark
  96. arxiv:2610.23345 · cs.AI
    OliO: ODE-based Linear Transition Operator for Self-Supervised Time Series Forecasting
    Kwangryeol Park, Seulki Lee

    Self-supervised representation learning has advanced time series forecasting by capturing robust features from unlabeled data. However, existing contrastive and masking-based methods often struggle to explicitly model the underlying temporal flow, either by treating time lag as noise or by disrupting the inherent temporal dynamics essential for forecasting. To address these limitations, we propose OliO (**O**DE-based **Li**near Transition **O**perator), a plug-and-play self-supervised learning method that learns bidirectional temporal flows as a continuous and structurally consistent dynamical system. OliO introduces a transition operator derived from linear Ordinary Differential Equations (ODEs) to align latent representations across arbitrary time shifts. By imposing a strictly upper triangular constraint on the transition matrix, we mathematically ensure numerical stability and enforce a robust polynomial boundary that effectively captures long-term dependencies while preventing the exponential divergence typically found in unconstrained ODE-based models. This structural constraint induces a temporally coherent representation space that preserves the underlying flow essential for accurate forecasting. Extensive experiments across various backbone architectures and benchmarks demonstrate that OliO achieves significant performance gains, providing Mean Squared Error (MSE) reductions of up to 5.75% over the most competitive grid-searched SOTA baselines. The code is available at *https://github.com/eai-lab/OliO*

    benchmark
  97. arxiv:2610.23344 · cs.AI
    Language Models Learn to Improve Language Models
    Minhua Lin, Zijun Wang, Zewen Liu, Zhaotian Weng +6

    Large language models (LLMs) are increasingly used to automate AI research, including language-model (LM) development, offering a potential path toward recursive self-improvement. However, turning experimental feedback into a learned ability to propose effective improvements across models faces two challenges. First, training such an ability with real feedback is expensive: reinforcement learning (RL) requires evaluating tens of thousands of proposals, each potentially requiring a full model-training run. Even at ten GPU-hours per run, reward collection alone would consume hundreds of thousands of GPU-hours, motivating a cheaper surrogate. Second, surrogate feedback can become unreliable as policy optimization favors overestimated proposals when the proposal distribution shifts. To address these challenges, we propose LM2LM, which trains LLMs with RL to generate effective LM improvement proposals. To provide scalable feedback, we use a research world model (RWM), a frozen LLM that predicts proposal gains from real experimental records provided in context, as the reward model for RL. To maintain feedback reliability, we periodically execute current-policy proposals ranked highly by the RWM and add their measured outcomes to its context. Across eight autoregressive and diffusion LM development environments, real-execution evaluations on held-out tasks show that LM2LM enables Qwen3.8-27B to outperform GPT-6-sol, even when GPT-6-sol is given 300 real experimental records in context, while requiring 33% lower data-generation cost. LM2LM also generalizes to a model family unseen during training.

    world modelself-improvement
  98. arxiv:2610.23342 · cs.AI
    Robust Function Calling for LLM Agents Under Interface Drift
    Aniruddha Chattopadhyay, Kanishka Halder, Partha Sarathi Roy

    Tool-augmented language-model agents rely on interface text: tool descriptions, argument schemas, and the candidate tool list shown at inference time. These interfaces often change after deployment even when the underlying tool semantics remain fixed, creating a practical reliability problem for function-calling systems. We study this failure mode as interface drift and evaluate whether a lightweight, training-free repair layer can recover robustness without retraining the base model. Our method, SchemaShield-Lite, combines schema validation, candidate-set retry for missing tool names, and one-shot constrained repair. On Qwen3.5-9B, it raises DICE-300 accuracy from 0.717 to 0.813 and BFCL-200 accuracy from 0.580 to 0.805, recovering 19/22 and 33/45 drift-induced failures with no observed repair harms. Naive retry does not explain the gains. Ablations show that candidate-set drift, especially distractor tools, accounts for most observed damage, while description and schema edits alone are less harmful. The results position validation-and-repair as a practical robustness layer for deployed LLM-agent tool use.

    llm agenttool use
  99. arxiv:2610.23334 · cs.RO
    LAW: Unifying Language, Action, and World Modeling for Agentic Physical Intelligence
    Jiajun Cao, Kai Tang, Bohua Zhang, Zezhong Qian +16

    World-action models jointly predict actions and future world states, but long-horizon manipulation further requires language reasoning to plan subtask objectives, assess execution progress, and guide action selection. We present LAW, a closed-loop VLM--WAM framework that unifies language reasoning, action generation, and world modeling through semantic subtask progress. Within the LAW agent, a hybrid-expert vision-language model plans the remaining subtasks, estimates the values of and fuses actions proposed by a world-action model, and asynchronously assesses execution status to support continued execution and subtask transitions. Within the world-action model, Mixture-of-Transformer Experts integrates modality-specific and shared computation for joint video--action modeling. Dual-stream hierarchical visual memory maintains temporal context for both models, while subtask-local value learning aligns action evaluation with semantic objectives and preserves supervision from successful prefixes of failed trajectories. LAW achieves 94.21% average success across 50 RoboTwin tasks. Replacing the WAM backbone with OpenWAM-$alpha$ further raises the average success rate to 95.56%, establishing a new open-source state of the art on RoboTwin. Despite having only 2B parameters, our VLM also outperforms variants using Embodied-R1.5 and GPT-5.5 High. On three long-horizon real-robot tasks, LAW achieves 65% average success, compared with 28.33% for $pi_{0.5}$ and 25% for Fast-WAM.

    embodiedmanipulationrobotwinworld modelmemoryagentic
  100. arxiv:2610.23328 · cs.LG
    Peak-Memory Schedules You Can Trust: Exact Realization, Certified Optimality, and the Limits of Recomputation for Inference
    Ivan Fomin, Ivan Koryabkin, Boris Sheludko

    Real-time inference on small devices needs low activation memory without extra latency, which makes operator order the most valuable memory lever: it is free in time. We show that the usual way of deploying a better order, handing it back to the compiler and regenerating code, is unreliable. In PyTorch Inductor the regenerated program misses its predicted peak in 22 of 54 CPU configurations. It is worse than the stock program in 9 of them and in 19 of 50 on a GPU, and on the GPU it also changes numerical results. We instead profile the generated program, model its memory exactly from its own statements and their read/write effects, and permute those statements. We prove that this realizes the modeled peak whenever allocation does not depend on history. It does so to the byte, with bitwise-identical results, in all our CPU runs. On the GPU we characterize why the caching allocator breaks the condition. To know when to stop, we give forward and reverse "trap" lower bounds, each computable with $O(n^2)$ minimum cuts. They certify the found order optimal for 94% of held-out wrapper graphs and all 16 microcontroller-class models we study, in under a second. So most inference graphs are already scheduled optimally. Branching networks are the exception, where reordering is free and pays, e.g. 10.0% on DeepLabV3. In compiled training we save up to 7.6%. Beyond the certified optimum we study selective recomputation with per-operator caps. It is strongly NP-hard for every fixed cap even at treewidth two, encodes knapsack exactly on series-parallel graphs, and is solvable exactly on bounded treewidth. It is useless on public tiny models whose peak is a single layer, but on architectures that hold a cheap value for a late reuse it pays: on the public ReDimNet speaker networks, whose order is already optimal, it lowers the peak by 5.2-6.1% at under 0.4% extra MACs, with bitwise-identical outputs.

    memory
  101. arxiv:2610.23324 · cs.CV
    CatchUp: Depth-Aware Re-Admission of Visual Tokens for Efficient Large Vision-Language Models
    Eunju Park, Uiwon Hwang

    Large Vision-Language Models (LVLMs) repeatedly process hundreds of visual tokens across decoder layers, incurring substantial computation and memory costs. Existing inference-time visual token pruning methods often irreversibly remove low-ranked tokens, although the set of visual tokens receiving high query-conditioned attention can change across decoder depth. Recoverable routing preserves skipped tokens for later reconsideration, but re-admits them with stale representations after bypassing intermediate transformations. We further find that replacing stale re-admitted states with their current-depth oracle states from a full-token reference trajectory substantially improves grounding performance. We therefore propose CatchUp, a training-free framework for depth-aware visual-token re-admission. CatchUp repeatedly re-evaluates query-conditioned visual relevance, processes only selected visual tokens through each decoder interval, and preserves deferred tokens for later re-admission. Upon re-admission, CatchUp approximates missed representation evolution using reference-to-current transition signals from semantically and spatially related visual tokens. Across three referring-expression grounding datasets and six multimodal benchmarks, CatchUp achieves the highest average relative retention in both benchmark groups at 64-, 128-, and 192-token budgets. At a 64-token budget with 88.9% fewer active visual tokens, CatchUp retains 92.17% of the full model's performance on average across six multimodal benchmarks while reducing end-to-end latency by 8.0%. Code is available at https://github.com/pej0918/CatchUp

    memorybenchmark
  102. arxiv:2610.23322 · cs.LG
    The Loss or the Architecture? What Matters in Survival Analysis with Competing Risks
    Julie Alberge, Gaël Varoquaux, Judith Abécassis

    Often, \emph{eg} in healthcare and predictive maintenance, what matters is which of several competing causes will strike first, and with what probability: survival analysis with competing risks. Corresponding deep learning methods are developed as bundles: a new architecture together with a new training loss, tested on one to five datasets against a handful of baselines. Thus, a state-of-the-art result cannot be attributed to the architecture, the loss, or the evaluation protocol. The coupling is not accidental: with one incidence curve per cause, a network's head is tied to the losses expressible on it. We vary the two axes independently, three losses, two of them strictly proper, on four backbones, across two modalities that tease the axes apart: 20 tabular and 3 pathology-image datasets. One loss wins in both: modelling cumulative incidence directly under censoring weighting. It brings twice as much as benefits a good choice of backbone on competing-risks data, and no more than it on single-event data: the loss is decisive for competing risks. The backbone matters too, and must match the modality: tabular networks lead on tables, pretrained vision encoders on images. The two choices compound rather than substitute. Proper scoring rules reveal these dynamics; the concordance index (C-index), which is not proper, does not separate models: neither axis moves it, not even tuning on the loss built to improve it. This is an avenue of progress: composed, the two give a model nobody has tried, RealMLP or TabM trained with the cumulative-incidence loss, which takes the best Brier score in our benchmark.

    benchmarkevaluation protocol
  103. arxiv:2610.23319 · cs.RO
    From Sliding Windows to Self-evolving Lifelong Memory: A Dual-perspective Survey of Memory Paradigms for Long-Horizon Robotics VLA Models
    Yong Wang, Boyu Shen, Xinwei Zhang, Yizheng Li +1

    Long-horizon robotic tasks expose a fundamental limitation of current Vision-Language-Action (VLA) models because their restricted temporal context pre-vents effective accumulation and reuse of past experiences. As a result, memory has become a fundamental component for enabling embodied agents to perform long-horizon reasoning and adaptive decision-making. This survey presents a systematic taxonomy of VLA memory mechanisms based on memory scope and persistence, categorizing existing approaches into five representative paradigms and one emerging frontier. Based on this taxonomy, we compare representative VLA memory systems in terms of memory characteristics, architectural com-plexity, and open-source maturity, revealing a clear trade-off in which increasing architectural complexity is generally accompanied by decreasing open-source ma-turity. Finally, we discuss the key challenges in building full-lifecycle memory systems, including efficient memory management and lifelong knowledge evolu-tion. Overall, this survey provides a unified perspective on memory architectures for VLA agents and offers insights into the development of lifelong embodied in-telligence.

    vision-language-actionvlavla modelembodiedmemorymemory architecture
  104. arxiv:2610.23317 · cs.AI
    AgentOnto: Ontology-Guided LLM Agents for Enterprise Data Analytics
    Dayan Pan, Bincheng Zuo, Xiaojian Wang, Daogang Zhang +1

    Enterprise data analysis requires agents to connect analytical requests with knowledge distributed across databases, business documents, and existing example tasks. Without an explicit representation of this knowledge, an agent may produce incorrect analytical results and cannot reliably incorporate feedback. We present AgentOnto, an ontology-guided LLM agent for enterprise data analytics. AgentOnto organizes heterogeneous enterprise knowledge into an ontology that directly guides analysis and incorporates confirmed feedback. Iterative Ontology Preparation further improves the ontology through self-generated analysis tasks. We evaluate AgentOnto on two public data sources. AgentOnto substantially outperforms representative methods, while module-level evaluations confirm the effectiveness of ontology construction, analysis, and updating. A real-world deployment further demonstrates its practical value in enterprise analytics. These results demonstrate that explicitly constructing, using, and updating enterprise knowledge supports reliable and adaptable data analysis.

    agentllm agent
  105. arxiv:2610.23313 · cs.AI
    From Memory Hits to Answerable Evidence: Query-Conditioned Memory Compilation for Long-Term Agent Memory
    Shangze Li, Yanyu Yao, Derong Xu, Zhi Zheng +1

    Long-term agent memory requires recovering answer-relevant evidence dispersed across turns and sessions. Existing systems increasingly use structured memories and adaptive retrieval, yet often pass retrieved or expanded units directly to the reader, coupling memory addressing with evidence presentation and leaving the final context redundant or incomplete. To address this gap, we propose MemCompile, a query-conditioned evidence compilation framework for long-term conversational memory. Given a query, the Address Plane locates relevant support through compact source-linked descriptors and capacity-aware Cells, while a normalization boundary resolves heterogeneous retrieval hits to stable SourcePacket identities. Its Evidence Plane then allocates a bounded evidence budget across complementary packets, admits limited additional support, and renders the retained claims associated with each selected packet into an answer-ready context. MemCompile achieves 80.52% Judge Accuracy on LoCoMo and extends the empirical quality-cost Pareto frontier to a higher-quality regime. Its strong performance on LongMemEval-S further suggests that the benefits of query-conditioned compilation generalize across long-term conversational-memory benchmarks.

    memoryagent memoryagentbenchmark
  106. arxiv:2610.23311 · cs.AI
    Self-Reflective Chain-of-Thought (SRCoT): Evidence-Based Text Generation via a Closed-Loop Retrieve--Reflect--Revise Pipeline on Edge LLMs
    Fabian Broszio, Michael Hoffmann, Adrian Paschke

    Large language models (LLMs) can generate fluent outputs that lack source support or conflict with local operational constraints. We introduce Self-Reflective Chain-of-Thought (SRCoT), a modular architecture that organizes generation as a closed Retrieve--Reflect--Revise loop. SRCoT separates environment state, persistent knowledge, accumulated experience, and short-term memory into provenance-labeled stores. BM25 retrieval supplies local context; subsequent passes expose intermediate drafts, missing-information markers, follow-up queries, and optional claim--evidence mappings. These artifacts inform retrieval, revision, clarification, or abstention decisions. We examine the architecture through qualitative, trace-based case studies in two fully synthetic settings: emergency medical services and Munich-Sim, an emergency-operations simulation. Local prototype runs on an Apple M4-family laptop with 128 GB memory illustrate source-labeled documentation, explicit gap marking, and structured dispatch-style outputs. The traces also reveal provenance-assignment errors, incomplete evidence coverage, and unsafe assumptions that remain possible despite prompting constraints. These cases demonstrate report-level inspectability and local execution, not comparative superiority, hallucination reduction, clinical validity, or operational safety. The architecture separates retrieval and control from the model interface, but behavior across different models remains to be evaluated. SRCoT is a research proof of concept requiring human review, with its principal contribution lying in an inspectable control procedure rather than a new model.

    memory
  107. arxiv:2610.23307 · cs.AI
    MobileInject: Evaluating Reasoning-to-Action Security of MLLM-Based Mobile Agents under Cross-Modal Consistency Injection
    Songping Wang, Peiyang Li, Yi Huang, Fengting Li +4

    Multimodal large language model (MLLM)-based mobile agents are rapidly evolving from conversational assistants into autonomous systems that perceive, reason over, and act on smartphone environments. This capability also exposes a new security risk: adversaries can embed malicious objectives in cross-application mobile content and steer agent behavior through indirect multimodal injection. Although prompt injection has been studied in desktop and web settings, its cross-modal properties in mobile ecosystems remain poorly understood, and existing security analyses are still limited. To close this gap, we introduce MobileInject, the first benchmark for evaluating mobile agents against multimodal injection attacks and reasoning-to-action security risks. MobileInject provides a high-fidelity interactive mobile evaluation framework covering 11 scenarios, 9 security-risk categories, and 108 executable risk instances. It further introduces Cross-Modal Consistency Injection to exploit corroboration across modalities, two enhanced attack methods, Multi-Role Collaborative Deception and Reasoning-Camouflage Injection, to intensify trust manipulation and reasoning hijacking, and a Reasoning Security Diagnosis Framework for fine-grained failure analysis. Systematic experiments show that mobile agents are broadly vulnerable to multimodal injection, and that many failures stem from reasoning hijacking at early stages. These findings establish MobileInject as a systematic cross-modal benchmark for studying and improving the security of mobile agents. Code release details are provided in the Appendix Open Science section.

    manipulationagentbenchmarkevaluation framework
  108. arxiv:2610.23306 · cs.AI
    SimOPD: Relational Knowledge Transfer for On-Policy Distillation
    Kai Tang, Xing Zheng, Haobo Wang, Guangcheng Zhu +5

    A key challenge in on-policy distillation (OPD) is extracting effective supervision from teacher knowledge. Supervision from token probabilities may underutilize internal knowledge, while direct hidden-state matching forces the student to shift to the teacher's feature space and complicates cross-architecture transfer. Inspired by contrastive learning's use of relationships between instances, we propose \textbf{SimOPD} to transfer teacher internal knowledge through response similarities. Specifically, SimOPD uses paired memory banks of student-generated responses to extract supervision from the teacher's internal representations as similarity distributions. The student then matches these distributions within its own representation space, retaining representational freedom and removing the need for cross-model projection. Extensive experiments show that SimOPD improves the overall average over the strongest baselines by up to 2.6 and 3.1 percentage points in same- and cross-architecture settings.

    memory
  109. arxiv:2610.23304 · cs.AI
    Designing for Imperfect Prompts: A Simulation-Based Evaluation of an LLM-Based Enterprise Data Assistant Using Knowledge Graphs
    Niklas Wagner, Alexander Mädche

    Enterprise data assistants using large language models (LLMs) promise democratized data access in organizations. Yet in our real-world deployment within the CFO unit's data lake at an industry partner, we observed that business users frequently pose imperfect prompts that include domain-specific vocabulary and lack schema references or structural information. This Design Science Research study examines automatic resolution using a semantic mediation layer. We design and evaluate two design variants: schema-based mediation via semantic similarity, and knowledge graph-based mediation leveraging an automatically constructed knowledge graph from the data lake. Our simulation-based evaluation using real-world data, supported by LLM-as-a-Judge assessments and human validation, shows that knowledge graph-based mediation outperforms the schema-based approach in Precision, Recall, F1, and preserving user intent. We propose a design principle for knowledge graph-based mediation, demonstrating that structured knowledge representation bridges user intent and database structure and guides the design of enterprise data assistants across skill levels.

    knowledge graph
  110. arxiv:2610.23302 · cs.LG
    Look, Reason, Draw a Region: Recursive Refinement for Spatial Context Prediction
    Wajeeha Tahir, Anna Tegon, Lionnus Kesting, Gamze İslamoğlu +3

    Spatial context prediction requires visual grounding and spatial reasoning to identify vacant regions relative to objects in a scene. Spatial relationships and physical constraints remain challenging for large vision-language models (VLMs). Recursive reasoning has demonstrated strong performance on discrete and abstract reasoning tasks, motivating its use for improving spatial context prediction in natural images. Weight sharing enables higher accuracy with a small trainable parameter count. We introduce Spatial-TRM, a compact recursive model with frozen vision-language features, a convex region decoder, and just 11.2M trainable parameters. This weight-tied reasoner supports six spatial relationships, three reference frames, and open-vocabulary anchor objects. On the RoboSpatial-Home benchmark, Spatial-TRM achieves 29.51% context accuracy, a gain of 9.02 percentage points (44.0% relative) over a separately trained single-step model with the same parameter count. Including the frozen encoders, the model has 877.8M total parameters, approximately $15\times$ fewer than the reported 13B size of RoboPoint. These results show that repeated refinement improves spatial context prediction from pretrained vision-language features at a fixed trainable parameter count on this benchmark.

    benchmark
  111. arxiv:2610.23299 · cs.AI
    Optimist Deep Research: An Iterative, State-Centric Agent for Autonomous Literature Research
    Gihan Panapitiya

    Most retrieval-augmented systems treat literature search as a single retrieval step followed by summarization. Real literature research works differently: the investigator's understanding of the question changes as they read, new subquestions appear, and evidence is weighed, not just collected. We present Optimist Deep Research, an open-source agent that runs literature research as an iterative evidence-acquisition process over an explicit, mutable research state. A research question is broken into a tree of subquestions, each served by search intents that are translated into source-specific queries over arXiv, OpenAlex, Crossref, PubMed, Semantic Scholar and the web. Results are merged into a canonical, provenance-preserving document registry through a conservative four-tier deduplication hierarchy. The registry is expanded by citation snowballing until a saturation criterion is met, and then screened, ranked and allocated to subquestions round-robin. Documents are read in parallel into a structured evidence store of claims with support, location, type and strength. A critic reviews the state along eight dimensions and returns research actions rather than a binary verdict: it can add, split, rephrase, deprioritize or close subquestions, so the research plan is an output of the investigation as well as an input. The final report is written section by section from the evidence store, and every citation resolves to a retrieved document. We describe the design, the deterministic machinery that keeps a language-model pipeline auditable, and a case study that read 28 of 237 retrieved documents in about three and a half minutes. We also report weaknesses the case study exposed.

    retrieval-augmentedagent
  112. arxiv:2610.23297 · cs.CV
    FlashReFocus: Interactive Image Refocusing in Seconds
    Ching-Heng Cheng, Chia-Ming Lee, Ming-Ching Chang, Xin Li +2

    Single-image refocusing changes the focal plane and depth of field of a photograph after capture. It is inherently interactive: users try many focal points and blur strengths on one image, so each edit should be fast and change only the blur. Recent generative methods, however, rerun large diffusion models for every edit. We observe that the two sub-problems differ in kind: recovering the all-in-focus image is ill-posed and needs a generative prior, whereas rendering blur is largely determined by depth and the focal plane. We therefore propose FlashReFocus, which runs diffusion once per photograph and renders every edit with a lightweight network. To address the ill-posed restoration, a one-step diffusion deblurrer with a decoder refiner recovers fine detail efficiently while balancing fidelity and perceptual quality. For precise control, the renderer is conditioned on an explicit defocus map and uses defocus-bias attention, which lets pixels at similar defocus share information without interference from pixels at other depths; in-focus content stays unchanged, and editing the map yields tilt-shift and two focal planes. To further test generalization to unseen cameras, we collect 3CReal, a three-camera benchmark of 89 pairs. FlashReFocus achieves the best perceptual fidelity for zero-shot bokeh rendering on EBB! and 3CReal, refocuses 148.3x faster than GenRefocus, and takes about half a second per further edit, 462x faster.

    benchmark
  113. arxiv:2610.23293 · cs.RO
    SkyAgent: Revisable Target Memory for Continuous Aerial Search
    Kaibo Yang, Xingyu Wu, Enze Zhu, Yajie Chai +5

    Continuous aerial search builds target hypotheses from sparse, viewpoint-dependent observations. When later evidence contradicts a stored hypothesis, however, the candidate can remain available to the planner and continue triggering approach or verification actions. We call this trajectory-level failure false candidate persistence (FCP). We equip SkyAgent with Revisable Target Memory (RTM), which associates evidence with persistent candidates, discounts correlated views, and revises each candidate through Tentative, Confirmed, and Rejected states. The key interface binds every generated waypoint to its source candidate: rejection therefore invalidates stale goals, while conservative reactivation protects temporarily occluded targets. Across four splits from the CityAVOS, UAV-ON, and OpenUAV benchmarks, our SkyAgent equipped with RTM outperforms all competitors, improving SR by 2.90--6.31 points while substantially reducing Behavior-FCP and post-contradiction travel with a low false-rejection rate. Across 35 matched real-world DJI scenarios per method, RTM consistently improves task success and behavioral reliability, further validating its state-to-action correction under physical deployment conditions.

    memorybenchmark
  114. arxiv:2610.23290 · cs.AI
    CoTA++: Understanding the Repeat Curse in dMLLMs from an Information Flow Perspective
    Qiyan Zhao, Xiaofeng Zhang, Xu-Yao Zhang, Shijie Han +3

    Diffusion-based multimodal large language models (dMLLMs) decode by iterative denoising with bidirectional attention, which precludes the standard autoregressive key--value cache. Practical deployments therefore rely on approximate caching that reuses intermediate states across denoising steps. Such caching, however, often makes a response repeat one token at adjacent positions in a run that breaks the sentence, a failure we term the \emph{Repeat Curse}. Tracing the context-token information flow reveals three disruptions. Within a step, caching disperses the attention that anchors a position on nearby context tokens. Across steps, similarity-based refresh is not staleness-aware, so some positions are decoded against stale key--value states. Across layers, the deep-layer entropy of the context fails to converge where repetition arises. We accordingly propose CoTA++, a training-free, plug-and-play framework that counteracts each disruption where it originates: Context-Token Attention Re-anchoring re-forms the attention routing against the current context, Decode-Aware Refresh reserves part of the cache's refresh budget for the positions about to be committed, and Context-Token Entropy-Guided Voting discounts the confidence of a position by the deep-layer entropy of its context, which reorders the unmasking. The three components act at non-overlapping stages of a decoding step and compose freely with existing dMLLMs and caches. Across three dMLLMs and two caching backends, CoTA++ reduces the sample repetition rate of long responses from up to 99.6\% to at most 17.0\%, recovers most of the accuracy lost to caching on nine benchmarks, and keeps the speedup of caching at under 15\% overhead, making approximate caching reliable for dMLLMs.

    benchmark
  115. arxiv:2610.23289 · cs.LG
    Data-Efficient Classifier Pretraining Through Diffusion Models
    Qinyu Li, Yee Whye Teh, Chris Holmes, Joel Jaskari

    Machine learning with limited data is susceptible to overfitting, especially with modern deep neural networks. Prior to the introduction of AlexNet, reconstruction-based pretraining with auto-encoder and denoising auto-encoder were found to be beneficial, partially due to the limited size of the training sets at the time. Motivated by these traditional approaches, we propose DiffPT, a simple diffusion modeling approach to pretrain a given classifier backbone for limited domain-specific training data. We experiment on time-series data, standard natural-image benchmarks, and medical imaging datasets, and contrast the diffusion approach to a multitude of modern and traditional un/self-supervised methods. While contrastive pretraining performs better on CIFAR-10 and Tiny ImageNet in low-data regimes, DiffPT is particularly effective on medical imaging and time-series tasks, where it outperforms self-supervised baselines across a range of low-data regimes. Our results suggest that DiffPT is a simple and effective approach for data-efficient classifier pretraining on different modalities.

    benchmark
  116. arxiv:2610.23288 · cs.RO
    CoEvolve: Self-Evolving Code Policies from Skills to Strategies for Humanoid Robots
    Zicheng Liu, Haochen Wu, Jiawei He, Suyun Zhao +3

    Recent advances in Vision-Language-Action (VLA) models and World Action Models (WAMs) have expanded the manipulation and whole body capabilities of individual robots. Yet a single robot remains limited by its embodiment, and many real world tasks require capabilities that can only emerge through cooperation. Existing multi robot frameworks largely emphasize task decomposition, allocation, and replanning, while providing limited support for adapting executable coordination when physical execution deviates from the intended behavior. We present CoEvolve, a framework that represents team coordination as executable code over fixed robot motor skills. CoEvolve introduces Verifier Guided Coordination Evolution, where a multimodal foundation model instantiates coordination strategies through motion primitives, semantic objectives, temporary robot roles, and bounded execution parameters. Physical verification evaluates each executed transition, rejected behaviors trigger structured coordination revision from an accepted checkpoint, and verified experience is retained to guide subsequent decisions. We evaluate CoEvolve with two simulated Unitree G1 humanoids transporting a freely supported rigid payload through constrained corridors. Across 15 base task runs with three multimodal backbone families, CoEvolve completes 8 tasks, with every backbone achieving end to end success. Additional evaluations yield verified placement on two additional corridor layouts and after both same-backbone and cross-backbone experience initialization, while exposing persistent limitations in motion precision and contact stability. These results support executable coordination as a practical interface for adapting long horizon multi robot collaboration under shared physical constraints.

    vision-language-actionmanipulationhumanoidself-evolving
  117. arxiv:2610.23287 · cs.CV
    I2V-DETACH: Source Grounding Detachment for Unauthorized Image-to-Video Generation
    Chanhui Lee, Yeonghwan Song, Yewon Kang, Jeany Son

    Recent diffusion models have advanced image-conditioned video generation, but their accessibility raises risks of unauthorized manipulation. Image immunization has emerged as a promising defense for Image-to-Video (I2V) protection, yet existing methods often rely on degrading visual quality or temporal coherence. In this work, we revisit I2V immunization as reducing source-content preservation, shifting the focus from output degradation to weakening source-conditioned content propagation. To this end, we propose I2V-DETACH, an image immunization method that prevents unauthorized I2V generation through source grounding detachment. I2V models preserve source content by propagating source-image cues into video latents during denoising. Accordingly, I2V-DETACH disrupts this process in two complementary ways: it suppresses source-video grounding to reduce source cue propagation, and strengthens intra-video interactions to shift denoising toward video-side dynamics with reduced reliance on the source image. To obtain informative optimization signals, we initialize auxiliary video latents with source-oriented states that remain within the source-conditioned generation regime while deviating from source-preserving trajectories. Extensive experiments across diverse diffusion architectures show that I2V-DETACH substantially reduces source-content preservation, improves over prior immunization methods, and exhibits strong black-box transferability.

    manipulation
  118. arxiv:2610.23284 · cs.AI
    Annotrace: A Workbench for Step-Level Annotation of Agent Traces
    Arun Josephraj Arokiaraj, Chenyi Huang, Sahan Bulathwela, Zekun Wu +1

    Every study of why LLM agents fail begins by labelling traces by hand, and every such study builds its own throwaway tool. Observability platforms record agent traces but offer no structured labelling; annotation tools label text but have no notion of a step, a delegation or a tool call, leaving it unclear what the smallest unit of execution to label should be. Annotrace reads six trace formats, including OpenTelemetry GenAI, LangSmith and Langfuse, and turns each run into an ordered, nested list of steps. The annotator writes their own labelling codebook; a frontier model proposes a starting set from a sample of their own traces, answers first, and is asked exactly the questions the human is asked, making annotation a joint human-AI process. Annotrace measures agreement between any two annotators, human or AI. On a multi-agent question-answering corpus it reproduces a previous manual effort's 235 annotation units exactly, and two models labelling the same traces agree 91% of the time at $κ= 0.82$.

    agentllm agentmulti-agent
  119. arxiv:2610.23279 · cs.CV
    Xiaomi-Lens-0 Technical Report
    Pei Fu, Yike Liu, Yue Sun, Hanmo Wang +11

    Multimodal large language models (MLLMs) answer questions about whole images well, yet they still struggle with fine-grained region understanding. A key reason is that text-only supervision never specifies which regions support an answer. To close this gap, we present Xiaomi-Lens-0, a 4B MLLM built on Qwen3.5 that predicts masks from its own visual representations. Specifically, a lightweight decoder combines vision-encoder patch features with visual latents, the language model's mask-token states, so mask supervision reaches the shared backbone without an additional segmentation encoder. Because each latent decodes to a mask, the model can interleave reasoning with region-specific latents and thus think with visual latents, while grounded reasoning traces supervise the regions it visits during chain-of-thought. As a result, across all eight RefCOCO, RefCOCO+, and RefCOCOg splits, Xiaomi-Lens-0 gains 5.6 points on average over Qwen3.5 in referring comprehension and achieves the highest segmentation cIoU among compared models. Moreover, it surpasses Qwen3.5 across all perception benchmarks. Homepage: \href{https://seerray-lab.github.io/Xiaomi-Lens-0}{this https URL}

    benchmark
  120. arxiv:2610.23275 · cs.CL
    Verbalising Model's Computation for Chain-of-Thought Faithfulness Detection
    Yujing Wang, Yulan He, Wenya Wang, Hanqi Yan

    Current Chain-of-Thought (CoT) verification methods detect reasoning faithfulness from model outputs or static activations, but are limited in explaining how unfaithfulness develops during generation. We introduce VERITAS to verbalise the model's computation: how different reasoning anchors emerge, transfer across layers, and surface in the output. VERITAS adopts the layer-wise Jacobian projection to produce a verbalised account of the model's internal reasoning evolution at each generated CoT token. By training a classifier on the extracted features along the traces, we show that these structured features provide insightful perspectives on how and where reasoning diverges toward unfaithfulness. Experiments on four datasets from FaithCoT-Bench and BonaFide show that VERITAS achieves a significant performance gain over LLM-as-Judge, with a 28.4% gain in balanced accuracy, and an average improvement of 16.0% over activation probing. Further analysis shows VERITAS can be generalised to different datasets and model architectures. Our work shows that verbalising model computation provides a richer and more interpretable lens on CoT faithfulness, moving beyond detection toward a mechanistic understanding of how reasoning fails.

    llm-as-judge
  121. arxiv:2610.23274 · cs.RO
    LY-GWM-RoboCasa-Human300: Prediction-Guided Action Selection with a Graph Dynamics Model
    TianLiang Ma, Ting Ma, ZhongYuan Ma

    We present LY-GWM-RoboCasa-Human300, a policy that combines a Human300-fine-tuned GR00T-N1.5 action generator, an action-conditioned graph dynamics model, and a learned binary state-reward scorer. At inference, the policy samples four action chunks, predicts the future latent features and robot state for each, and executes the candidate with the highest predicted-state reward logit. Predictions therefore influence the actions sent to the environment. The evaluated system uses a 16-step prediction horizon and keeps all three components frozen during evaluation. In the local evaluation on 50 RoboCasa 1.0.1 tasks in the pretrain kitchen split, with 50 episodes per task, the system records 207/2,500 successes (8.28%), compared with 181/2,500 (7.24%) in the historical self-trained GR00T-N1.5 baseline. The observed difference is 26 successes, or 1.04 percentage points. The atomic group improves from 141/900 to 150/900. Separately, the RoboCasa maintainers re-evaluated the submission and reported 354/1,500 successes (23.6%) across 50 tasks, with leaderboard rates of 48.3%, 16.3%, and 3.1% for Atomic-Seen, Composite-Seen, and Composite-Unseen. The local and official results are reported separately; their difference is not attributed to a model improvement. Recorded local audits identify 212,700 changed environment-input chunks, including 28,223 in atomic tasks. We describe the actual training stages, implemented inference mechanism, evaluation provenance, and public reproduction entrypoints. These results document an active prediction-and-selection system and a positive historical difference.

    gr00taction-conditionedleaderboard
  122. arxiv:2610.23271 · cs.RO
    NestedVLA: Learning to Consolidate and Generate Skills for Vision-Language-Action Models
    Guodong Du, Wanyu Lin

    Vision--Language--Action (VLA) models promise generalist robotic agents, yet consolidating skills learned across heterogeneous tasks and environments remains difficult: training data is fragmented across platforms, joint multi-task training suffers from interference, and model merging tends to introduce parameter conflicts or require architectural changes. We instead advocate a paradigm in which skills are first learned independently and then consolidated post hoc. To this end, we propose NestedVLA, which replaces parameter aggregation with parameter generation: a nested hypernetwork, conditioned on the target task, synthesizes task-adaptive action-expert parameters that compose with a frozen base VLA. Task-specific compressed VLM residuals are retained separately. This formulation enables flexible task-conditioned skill reuse while mitigating interference. Across three RL benchmarks (MetaWorld, ManiSkill, CALVIN), NestedVLA outperforms the strongest model-merging baseline by 41 pp on average, matches per-task fine-tuning within 1 pp on ManiSkill, and substantially closes the gap on CALVIN, pointing toward a storage-efficient route to skill consolidation. Code is available at https://github.com/WanyuGroup/NeurIPS2026-NestedVLA.

    vision-language-actionbenchmark
  123. arxiv:2610.23269 · cs.CV
    Agentic Image Generation in the Wild
    Yifan Li, Bowen Qin, Kun Zhou, Haochen Zhao +7

    Agentic image generation pairs an image generator with a search agent that retrieves references from the web, extending generation beyond the generator's training data. Real-world requests, however, often concern entities that appeared only days ago. We call this setting Image Generation in the Wild. In this setting, the web offers limited visual evidence, search results are crowded with easily confused entities, and the target is rarely shown on its own. To study this setting, we build an automated pipeline that synthesizes prompts from recent headlines and keeps only those whose real search results are highly confusing, yielding requests that are hard to solve yet natural to ask. With it, we construct Savanna, a benchmark of 500 prompts across 16 subcategories. On Savanna, even harness frameworks built on the frontier models retrieve a correct reference for only about half of the prompts. We then propose Locate-Corroborate-Isolate, a paradigm in which the agent first pins down the target's identity through text search, then verifies candidate images against multiple sources, and finally crops a clean reference for the generator. Trained on 15K trajectories that follow this paradigm, Ranger-8B achieves a search score of 67.2, outperforming harnesses equipped with frontier models. It also achieves the best generation quality with every generator we test, and surpasses prior agentic methods on other image generation benchmarks. We release the pipeline, benchmark, training data and model at https://github.com/RUCAIBox/Agentic-ImageGen-Wild.

    agentagenticbenchmark
  124. arxiv:2610.23266 · cs.LG
    Beyond Marginal Influence: Interaction-Aware Memory Selection for LLM Agents
    BHVSP Subrahmanyam, Neeraj Choudhary

    LLM-agent memory selection often uses retrieval rankings based on record relevance or individual influence. Such rankings cannot identify sufficient subsets under redundant or jointly necessary evidence. We study per-decision memory sufficiency: selecting a small subset that preserves full-memory behavior. To separate search error from model error, we introduce JointCoreBench, a benchmark with nine interaction families, six workflow domains, natural-language decisions, and exact sufficient-set oracles, and JointCore, a relevance cost-ordered group-deletion baseline that tests blocks before individual records. On 6,912 exact conditions, cost-ordered JointCore recovers a minimum-cost sufficient set in 100.0% of cases, reduces subset evaluations from 30.00 to 11.13 (62.9%), and achieves 87.8% active-memory savings. Across four 12B-70B model families and 432 tasks per model, its unweighted mean is 10.28 versus 16.00 model queries and 50.6% versus 43.1% minimum-cost recovery relative to cost aware singleton deletion. Both methods preserve full-memory behavior under the deterministic primary interface. Tests with physical deletion, UNKNOWN masks, PENDING sentinels, counterfactual values, and alternative labels identify interface dependent behavior. JointCore is an efficient benchmark baseline; the study does not establish deployment-ready memory compression.

    memoryagent memoryllm agentbenchmark
  125. arxiv:2610.23262 · cs.RO
    Expand-and-Prune: Test-Time Candidate-Space Adaptation for End-to-End Autonomous Driving
    Fabian Schmidt, Karol Fedurko, Markus Enzweiler, Abhinav Valada

    Candidate-based end-to-end driving planners reason over a finite set of trajectory hypotheses. However, for sparse planners, a limited candidate set may provide insufficient support for scene-relevant maneuvers, while fixed downstream candidate budgets do not account for prediction uncertainty. To address these challenges, we introduce Expand-and-Prune, a test-time framework that jointly adapts candidate support and capacity without updating learned parameters. The framework expands the native candidate set using semantic driving intent and temporally propagated predictions and then uses uncertainty in intermediate trajectory scores to adaptively prune the resulting candidates. Across four heterogeneous candidate-based planners and four benchmarks, Expand-and-Prune consistently improves planning performance. It achieves state-of-the-art results on LangAuto, increasing the Driving Score from 68.2 to 75.9, while providing further improvement on Bench2Drive, Fail2Drive, and NAVSIMv2, with particularly strong improvements in out-of-distribution scenarios. Moreover, in its full instantiation, 72.5% of selected trajectories originate from the newly introduced intent-guided or temporal candidates, confirming that candidate-space expansion provides useful hypotheses beyond those generated natively by the planner. The code is available at https://github.com/iis-esslingen/Expand-and-Prune.

    benchmark
  126. arxiv:2610.23256 · cs.AI
    Beyond Reward Scoring: Online Multi-Agent Preference Learning for Multimodal Response Generation
    Ziyun Zou, Jiwoo Hong, Yifan Lu, Shao Tang +4

    Multimodal preference alignment often follows a score-based RLHF pipeline: a reward model is first trained to learn human preferences, then used to score on-policy responses and optimize the response model's generation. This turns preference supervision into a scalar signal for generation: a model may learn to produce answers with higher preference scores, but is rarely trained explicitly to learn the reasons underlying human preferences. We therefore ask whether making preference understanding an internal capability of the response model can improve its generation. To investigate this question, we propose \textbf{PrefMind}, an online dual-agent reinforcement learning framework with two trainable agents: a response Generator and a preference Critic. Given a pair of candidate responses, the agents analyze and discuss which response better aligns with human preference. Human labels supervise their judgments and corrections. For pairs on which both the Generator and revised Critic correctly identify the human-preferred response, the Generator produces an improved answer, and the Critic compares it against the original preferred response to provide a generation reward. Both agents are jointly updated from interactions sampled with their current policies. Through this process, \textbf{PrefMind} trains the Generator to understand human preferences via explicit judgment and correction, and to translate this understanding into better responses. Experiments across three model scales show improvements in both standalone preference judgment and downstream response generation.

    multi-agentrlhf
  127. arxiv:2610.23255 · cs.LG
    Sample, Mix, or Verify? Allocating Test-Time Compute under Imperfect Verifiers
    Srinivasan Iyengar

    A common way to make language models and agents more accurate is to spend more compute at inference time: generate many candidate solutions and submit the one a verifier prefers. Success then requires that some candidate is correct and that the verifier picks it. The first grows predictably with the number of samples; the second does not, because real verifiers (unit tests, majority votes, LLM judges, reward models) accept wrong answers, reject right ones, and make correlated mistakes. With a fixed budget, a practitioner must decide how many samples to draw, which models to draw them from, and how much to spend on verification, and today these choices are made by rules of thumb. We introduce SPLIT, which makes them from a small labelled pilot. It rests on an exact formula for the probability of submitting a correct answer under accept/reject and top-score selection that accounts for varying task difficulty, blind spots shared across models, and correlated repeated verifier calls. SPLIT fits the formula by empirical Bayes and uses it to plan budgets on unseen tasks. Across seven benchmarks spanning math, science QA, code and software-engineering agents, 26 models, 175 agent systems and six kinds of real verifiers, all from public releases, SPLIT lands within 0.02-0.8 percentage points of the best allocation in hindsight, while every fixed rule we test loses at least 2.5 points in some setting. Regret depends only on how well SPLIT ranks configurations, and it stays near-optimal under compute, price and latency costs and online. The formula also explains when asking more LLM judges lowers precision. As a forecaster far beyond the pilot, SPLIT beats power-law fitting when the verifier caps success, and we report where it loses.

    agentagent systembenchmark
  128. arxiv:2610.23254 · cs.CL
    Ignored, Not Refused: What Low Prompt-Injection Success Rates Hide in Small Tool-Using Agents
    Farhan Zarif, Kazi Shahid Ahmed Galib, Synthia Akter Promi, Shabnam Sharmin Tarin +3

    Tool-using agents read untrusted external content, which exposes them to indirect prompt injection (IPI), where hidden instructions hijack the agent's tool calls. Benchmarks measure IPI security by the attack success rate (ASR), which treats every failed attack as a successful defense. To test this assumption, we audit 33,000 trajectories from eleven open-weight agents (2-14B parameters) on InjecAgent and AgentDojo. A label-blind LLM judge shows that most failed attacks are ignored (76-80%) rather than refused (17-18%). With identical model weights, changing only the system prompt shifts InjecAgent ASR from 0.2% to 43%. Our Attack-Task Outcome Matrix (ATOM) further shows that in 21.7% of AgentDojo episodes, the agent stayed safe but did not complete the user's task. A low ASR therefore mostly reflects agents silently ignoring injections, not refusing them. Evaluations should report how attacks fail and whether the user's task is completed, alongside ASR. We release the labeled corpus, judge prompt, and analysis code.

    agentbenchmark
  129. arxiv:2610.23252 · cs.AI
    Preventing Data Hallucinations in Agentic AI for Optical Network Troubleshooting
    Jeremias Dötterl, Martin Kuipers

    An agentic AI system for automated troubleshooting in optical communication networks is demonstrated that integrates data references into the agent architecture to prevent data hallucinations, improving the interpretability and trustworthiness of root-cause identification and reasoning.

    agentagentic
  130. arxiv:2610.23251 · cs.LG
    FLARE: Physics-Grounded Evaluation of Language-Model Agents for Feeder Restoration with Degraded Observations
    BHVSP Subrahmanyam, Neeraj Choudhary

    Restoring service after distribution-grid faults requires decisions from incomplete and conflicting information while respecting electrical constraints. Addressing this need, we introduce FLARE (Feeder-Level Agent Restoration Evaluation), a physics-grounded benchmark for post-isolation feeder restoration. FLARE con- tains 800 AC-feasible cases across five observation settings and allows agents to gather evidence, evaluate switches, and perform restoration actions under fixed operational limits. We evaluate five instruction-tuned models over 78,750 episodes, including topology, distributed-generation, and loading shifts. Mean normalized regret ranges from 0.082 to 0.148. Although every model has zero median regret, P95 regret ranges from 0.814 to 1.000, revealing an important failure tail. Shadow replay shows that 96.4% of 15,706 guard-rejected actions would have caused a physical violation. FLARE enables reproducible evaluation of language-model based restoration support while emphasizing the need for external safety checks.

    agentbenchmark
  131. arxiv:2610.23250 · cs.CV
    LonGaussian: Tracking Foundation Priors for Long-Context Feedforward Gaussian Reconstruction
    Zhoufeng Xie, Guanjun Wu, Shuai Zhang, Xinggang Wang +1

    Reconstructing high-quality, renderable 3D scenes from long videos is essential for robotic perception and embodied intelligence. Many feed-forward Gaussian reconstruction methods rapidly predict 3D scenes from a small set of images, but extending them to longer sequences presents two bottlenecks: jointly processing all views increases GPU memory requirements, while accumulating pixel-aligned predictions redundantly represents the same surfaces and increases storage and rendering costs. We present LonGaussian, a streaming feed-forward framework that reconstructs compact, globally renderable Gaussian scenes from long image sequences with known intrinsics and unknown camera poses. Built on a pretrained streaming 3D reconstruction backbone, LonGaussian decodes long-context geometric features into dense Gaussians. Within each chunk, multi-frame point tracks link repeated observations of the same surface and guide a learned merger to consolidate the corresponding Gaussian predictions. Rendered-depth consistency filtering further suppresses Gaussians from other chunks that incorrectly occlude locally reconstructed surfaces. On DL3DV and RealEstate10K benchmarks, LonGaussian improves novel-view rendering quality while using substantially fewer Gaussians than the evaluated feed-forward baselines.

    embodiedmemorylong-contextbenchmark
  132. arxiv:2610.23248 · cs.LG
    Can the Parts Fool the Test? Counterfactual Pairing Cycles for Relational OOD
    Yibo Zhou, Yirong Xiang

    Relational out-of-distribution errors are easy to describe and hard to evaluate cleanly. An image may be natural, a caption may be fluent, and all mentioned objects may be common, yet the caption may describe the wrong relation in the image. Many evaluations intended to test this behavior can be solved without checking the pairing at all, because negatives differ in caption style, template, length, object frequency, or image statistics. We propose counterfactual pairing cycles, a blocked evaluation primitive that reuses the same two images and the same two captions on both sides of the comparison, changing only which image is paired with which caption. This makes image-only, text-only, and additive component shortcuts cancel exactly in finite samples. The same contrast also gives a training loss for relational outlier exposure and a diagnostic for bilinear vision-language scores. Across synthetic controls, GQA, and COCO, cycles separate coarse image-caption matching from directional role-binding failures, while also identifying geometric size relations that remain strongly visible in global features. In a shortcut-contaminated COCO outlier-exposure stress test, a model selected by a near-perfect shortcut validation AUROC retains substantial relational signal but remains below cycle training on exact cycles; cycle training improves CycleAcc from 0.882 to 0.927. The main lesson is simple: to test whether a model understands a pairing, the benchmark should keep the parts fixed, change the pairing, and ask only about coupling. The official implementation is available online.

    benchmark
  133. arxiv:2610.23246 · cs.LG
    Think More or Read More? A Token-Budget Theory of Reasoning with Shared Context
    Manh Nguyen, Sunil Gupta, Hung Le

    Reasoning systems increasingly improve answers by incorporating additional context into the model's input, from multi-agent debate, where agents read their peers' responses, to self-refinement, where a model revisits its own drafts. In these systems, communication consumes context as one agent's output becomes another's input. We study the regime where token reading and generation draw on one fixed token window, so every token spent reading is a token that cannot be spent reasoning. We develop a \textbf{token-budget theory} that formalizes this trade-off, modeling accuracy as a function of the reasoning budget remaining after reading and aggregating votes with a Condorcet jury model that permits correlated errors. For multi-agent debate, the theory gives a closed-form prediction for the optimal number of agents. This quantity is estimated from single-agent measurements, allowing the optimal group size to be determined without running debate, and the same accounting extends to a heterogeneous set of different models. We empirically verify our theory and show that, at the predicted group size, our method outperforms strong baselines under the tightest calibration spend, where debate-calibrated rules can afford only one or two questions. Self-refinement obeys the same accounting, with its own draft in place of a peer message. The theory further predicts that communication pays only when the gain in the accuracy ceiling exceeds the reasoning room spent on reading, with this condition tightening as the window shrinks. Across five open models and four reasoning tasks, self-consistency matches or beats deployed debate at $93$\% of $980$ matched window-and-group-size settings, with the gap being largest under tight windows and nearly disappearing at a loose one, consistent with the theory. Under a shared window, how much room a model has left to \textit{think matters more than what it reads}.

    multi-agentself-refinement
  134. arxiv:2610.23245 · cs.AI
    MGE-PO: Entropy-Aware Credit Assignment for Mathematical Reasoning
    Jihao Huang, Guangjia Chai, Zhirui Chen, Lei Wang +8

    GRPO assigns the same group-relative advantage to equally rewarded responses within a question, leaving their generation uncertainty unused. We introduce MGE-PO, centered on four-quadrant (4Q) advantage shaping that combines verified correctness with rollout generation entropy. Correctness determines the sign of an additional correction, while entropy controls its magnitude: lower-entropy successes receive larger bonuses and higher-entropy failures receive larger penalties. An online entropy-based question curriculum supports this credit rule using cached rollout measurements. At the final checkpoint after eight training epochs on Qwen3-8B-Base, adding 4Q across three policy optimization settings yields absolute gains of 3.23-3.69% in six-benchmark macro avg@32 and 4.24-11.06% in macro pass@32. On Qwen2.5-7B-Base, the complete configuration raises four-benchmark macro avg@32 from 41.5% to 45.4%.

    benchmark
  135. arxiv:2610.23239 · cs.LG
    Response-Informed Value Estimation for Cost-Aware Multi-Turn LLM Routing
    Haotian Yan, Jiaqi Liu, Bingyi Jing, Tianjun Yao

    Multi-turn LLM routing selects a model at every turn before its response is available, yet task success is revealed only when the interaction ends, leaving each intermediate choice without a label of its own. Logged trajectories, however, retain the response each call produced, which exposes what each turn required and how well the chosen model served it, evidence that outcomes alone hide. We introduce RIVET (Response-Informed Value Estimation for Turn-Level Routing), which exploits this post-call evidence during training while keeping routing strictly pre-call at deployment. RIVET jointly trains two value estimators over a shared state-model representation: a deployable Routing Value Estimator that scores candidate models from pre-call information, and a training-only Response-Informed Value Estimator that additionally observes the realized response. Conditioning on the response turns response-induced return variation from residual noise into an observed covariate, and we show that this reduces the leading asymptotic estimation error of the shared representation. On the MINT benchmark, RIVET solves more held-out tasks than the strongest multi-turn baseline while using 74.7% less total call cost.

    benchmark
  136. arxiv:2610.23238 · cs.CV
    Kingfisher: Fast Personalized Visual Autoregressive Generation via Scale-Aware Prompt-Conditioned Scale Decomposition
    Dinh-Khoi Vo, Akihiro Sugimoto, Minh-Triet Tran, Trung-Nghia Le

    Personalized image generation has been extensively studied in diffusion models, often relying on subject-specific optimization of model weights, textual embeddings, LoRA modules, or adapters. In this work, we explore a different direction for fast personalization in visual autoregressive models. Our key observation is that a short prefix of early autoregressive scale codes is sufficient to strongly determine both object structure and global appearance, while later scales primarily refine the established generation trajectory. Based on this insight, we introduce Kingfisher, a model-tuning-free framework that decomposes early residual scale codes into content- and style-related components through prompt-conditioned scale-code subspaces. Given two real images, Kingfisher first decomposes their early scale codes, then recomposes the content components of one image with the style components of the other into a compact semantic prefix. The frozen generator subsequently predicts the remaining scales through bitwise self-correction for coherent refinement. Experiments on CSD-100 show that Kingfisher achieves strong content preservation and the highest style and text alignment among the evaluated methods, while reducing pair-specific preparation time by approximately 54x compared with diffusion-based methods and 7.5x compared with VAR-based alternatives. These results demonstrate that effective personalization can be achieved by directly manipulating the native scale-space representation of a frozen visual autoregressive generator.

    self-correction
  137. arxiv:2610.23234 · cs.CL
    MultiAgent-OPD: End-to-End Training of Hierarchical Multi-Agent Systems with On-Policy Distillation
    Yuhan Sun, Jiaxin Zhao, Yongdong Luo, Yibo Hu +5

    In a hierarchical multi-agent system, a Leader assigns tasks and subagents carry them out. Jointly training the Leader and the subagents is difficult because a final answer gives only indirect feedback to the subagents. An LLM judge can score the performance of the subagents, but these scores depend on the judge's own assessment of the quality of subagents' responses. We present MultiAgent-OPD, which combines on-policy distillation with reinforcement learning to jointly train a shared-parameter Leader-Searcher-Answerer team end to end. We first train a Teacher in a Searcher-Answerer workflow using answer correctness. During team training, the frozen Teacher scores student-generated search and summary tokens to provide local process supervision for the Searchers, while the Leader learns from team success and the Answerer from candidate correctness. Across seven datasets, the trained system reaches 38.68% macro avg@4, exceeding Prompted Critic by 4.06 percentage points and C-MAS + RL, the trained cascaded multiagent system, by 2.61 percentage points. The team also solves more questions within four attempts.

    multi-agentagent system
  138. arxiv:2610.23230 · cs.CV
    RAP: A Dataset for Paper-to-Poster Generation via Declarative Markup from Human-Crafted Exemplars
    Haonan Han, Yukang Lin, Minliang Lin, Haoxian Tan +6

    Making a conference poster is visual coding in miniature: an entire paper must be compressed onto a single page, which exercises summarization, visual design, and code implementation at once, using nothing beyond the paper itself. Yet benchmarks for this task in the context of visual coding remain scarce: existing resources are small and annotated only with layout boxes, so evaluation must fall back on VLM-as-judge or human checklists rather than objective render-and-compare, and the generation methods they support design layouts from scratch, leaving human-crafted posters unusable as either a retrievable design library or as verified supervision. We introduce RAP (Reconstruct A Poster), a dataset and benchmark for paper-to-poster generation as a visual coding task. RAP contains 25K+ verified paper-poster pairs from various machine learning conferences held between 2022 and 2026, annotated with a three-level content taxonomy, plus a held-out benchmark of 150 test papers each paired with its human poster. A fully automated derendering pipeline turns each poster into styled HTML/CSS of typed, refillable slots, so every sample is a paper-poster-markup triplet verifiable by rendering against the source poster. The corpus serves both inference and training: PosterRAG, a training-free multi-agent pipeline, retrieves an exemplar and refills its slots without touching layout or styling, while an SFT+RL recipe trains a dedicated model from the same triplets. An online demo is available at https://reconstruct-a-poster.github.io/.

    multi-agentbenchmark
  139. arxiv:2610.23226 · cs.LG
    Characterising and Analysing Retrieval-Augmented Generation Techniques
    Emmanuel Igbekele, Olawande Daramola

    The advent of Large Language Models (LLMs) across many application domains has increased the need for Retrieval-Augmented Generation (RAG) to reduce hallucinations and improve LLM performance. Thus far, several RAG techniques with impressive features and capabilities have been proposed in the literature, making it a herculean task to select the most appropriate RAG techniques for specific problem scenarios. Thus, we propose a conceptual framework comprising main dimensions and sub-dimensions for characterising (defining "what it is") and analysing (defining "how well it does") existing RAG techniques. The conceptual framework was developed through a qualitative analysis of the attributes of RAG techniques reported in the literature. It can serve as a pre-empirical filtering mechanism for researchers and practitioners in selecting the highest-fit RAG candidates that meet specific user and task requirements before empirical benchmarking of RAG techniques, which can be computationally and financially expensive.

    retrieval-augmentedragbenchmark
  140. arxiv:2610.23225 · cs.CL
    ReLaDrift: Recurrent Language Generation via Drifting for Test-Time Scaling
    Yuki Otsuka, Daisuke Oba, Naoaki Okazaki

    Drifting Models generate a sample with a single function evaluation, making them attractive for fast parallel language generation, but their one-step formulation does not support iterative refinement. We propose Recurrent Language Generation via Drifting (ReLaDrift), which uses a shared Drifting Model to generate an initial draft from noise and iteratively refine it in a continuous latent space. We train this recurrence by sampling the number of rounds and applying the drifting loss only at the final round, stopping gradients through preceding rounds so that only one round requires backpropagation. Unlike the one-step Drifting Model, whose accuracy does not improve when iterated, ReLaDrift enables a trade-off between test-time computation and generation quality, as diffusion and flow language models do. On Sudoku Hard and GSM8K, it achieves competitive or superior performance compared with these iterative-generation approaches across a range of inference budgets. Our code is available at https://github.com/Yuki-Ots/ReLaDrift

    iterative refinement
  141. arxiv:2610.23223 · cs.LG
    Process Reward Modeling with Reliability- and Conservatism-Aware Bayesian Belief
    Zhongjian Qiao, Zhixiang Zhou, Jiafei Lyu, Peisong Wang +3

    Process Reward Models (PRMs) play a central role in test-time scaling for large language model (LLM) reasoning. However, PRMs may provide inaccurate process rewards for test-time scaling, which we identify as arising from two sources of uncertainty: (1) \textit{PRM uncertainty} that arises from insufficient supervision data and imperfect PRM training; and (2) \textit{policy uncertainty} that results from the mismatch between the policy to collect supervision data and the policy for test-time scaling, leading to systematic reward overestimation. Previous studies fail to jointly consider both uncertainties. To bridge this gap, we propose BayesianPRM, which treats the process reward as a random variable within a reward hypothesis space and maintains a Bayesian belief over it. Since exact Bayesian inference over the continuous hypothesis space is generally intractable, we discretize the reward hypothesis space with a reward ensemble and introduce two complementary belief updates, each targeting one source of uncertainty: (1) to address PRM uncertainty, we introduce \textit{reliability-aware belief updating} that measures how well each ensemble member explains the supervision data and updates the belief with an amortized belief network optimized by the evidence lower bound (ELBO); and (2) to tackle policy uncertainty, we propose \textit{conservatism-aware belief calibration} which further reweights the belief toward lower rewards to suppress overestimation. These two updates yield a reliability- and conservatism-aware Bayesian belief, whose expectation determines the process reward. Experiments on multimodal reasoning benchmarks indicate that BayesianPRM consistently yields more accurate process reward estimates while enhancing test-time scaling performance.

    benchmark
  142. arxiv:2610.23213 · cs.AI
    Not All Test-Time Compute Is Evidence: Model-Specific Bias in LLM Reasoning
    Ming Shen, Tianjun Yao

    Test-time scaling methods implicitly assume that additional computation from the same model provides independent evidence. We show that this assumption fails within a single reasoning model: errors are correlated across repeated samples, across reasoning depth, and in the model's own self-evaluation, so more compute from the same model is not more evidence. We introduce DIA (Disagreement-triggered Independent Arbitration), which treats disagreement between an intermediate-stage and the final-stage prediction as a per-question signal of reasoning instability and resolves it by asking a small panel of heterogeneous models to solve the question independently, without seeing either candidate, and aggregating their answers directly. Across seven benchmarks and two base models, DIA outperforms full chain-of-thought in 12 of 14 settings, including a 7.00% gain on MuSiQue, while activating on only $\sim 17\%$ of examples and using up to 47.2% fewer tokens than a multi-agent baseline. Ablations show that both independent solving and solver heterogeneity are necessary for these gains.

    multi-agentbenchmark
  143. arxiv:2610.23205 · cs.CL
    MAS-Bai: Linguistically Informed Adaptive Multi-Agent Translation for Bowen-Script Bai
    Jingting Zheng, Hong Ren, Yuehan Cui, Shaolin Zhu +1

    Large Language Models (LLMs) remain unreliable for extremely low-resource languages whose writing systems conflict with high-resource pretraining priors. We study Bai-Chinese translation in the Bowen script, where Chinese characters are frequently borrowed phonetically rather than semantically, causing severe literal-translation hallucinations. We propose MAS-Bai, a linguistically informed adaptive multi-agent framework for Bowen-script Bai-Chinese translation. MAS-Bai routes each input through either a lightweight retrieval-translation path or a deeper philological path with etymological and syntactic agents, and uses cycle-consistency with a lightweight reverse translator as an inference-time reliability signal. We construct a 6,536-sentence Bai-Chinese corpus and evaluate on a 1,536-sentence test set. MAS-Bai achieves 16.23 BLEU and 62.15 COMET, improving over Standard RAG by 8.11 BLEU and 6.85 COMET. Forward LoRA, expert ablations and controlled routing experiments distinguish the roles of supervised adaptation, Bai-specific expertise and conditional routing. The largest gains occur on Bowen phonetic loans and divergent Bai word order, while human evaluation and reverse-verifier analyses support the automatic results.

    ragmulti-agentagent framework
  144. arxiv:2610.23203 · cs.RO
    SimpleVLN: Towards Simple and Frontier Vision-Language Navigation
    Lingyi Hong, Weiren Cai, Jiayu Chen, Xinyu Zhou +6

    Vision-and-Language Navigation (VLN) requires embodied agents to ground language in visual observations and make long-horizon decisions. Existing methods often address these demands through task-specific architectures. As multimodal large language models (MLLMs) become increasingly capable, an important question remains: Have we truly unlocked the potential of native MLLMs for VLN before introducing task-specific architecture? We present SimpleVLN, an architecture-preserving framework developed through studies of navigation formulation, data construction, and optimization. At the formulation level, SimpleVLN combines single-turn decision inputs with fixed-budget global-local memory to efficiently use navigation history without unbounded context growth, while action chunks enable efficient output of multiple primitive actions per model invocation. For data construction, we combine expert demonstrations with policy-dependent DAgger experience and develop a scalable synthesis pipeline that generates large-scale tasks for reinforcement learning. At the optimization level, a progressive IL-DAgger-RL pipeline advances from action imitation to task-level outcome optimization. Comprehensive experiments validate the effectiveness of these three aspects. We instantiate SimpleVLN across multiple MLLM backbones and model sizes, demonstrating its generality. SimpleVLN achieves 76.7% and 80.1% success rates on the R2R-CE and RxR-CE splits, respectively, substantially outperforming existing methods. These results show that the navigation capabilities of native MLLMs can be effectively unlocked without task-specific architectural designs.

    embodiedmemoryembodied agent
  145. arxiv:2610.23199 · cs.AI
    Validating LLM-generated political networks with survey-reported partisan exposure
    Marco Tonin, Michele Tizzoni, Bruno Lepri, Luca Maria Aiello

    Large language models (LLMs) are increasingly used to simulate human behavior in social science research, including decisions, interactions, and social relationships. However, in network formation tasks, existing evaluations often focus on aggregate network properties, while networks with plausible structural properties may still fail to reproduce individuals' social and political environments. Here, we generate LLM-based political networks using personas sampled from the nationally representative ANES 2020--2022 Social Media Study and use respondents' survey-reported partisan exposure as an empirical benchmark for the political environments generated for the corresponding personas. We first characterize the structural properties and political assortativity of the generated networks, and then compare ego-level partisan exposure with the survey. We find that the structure and political assortativity of the generated networks vary across experimental configurations. More importantly, when comparing generated partisan exposure with the survey, some models generate stronger separation between Democratic and Republican egos than reported in the survey, with higher same-party and lower cross-party exposure, while others mainly underestimate Republican exposure across ego groups. Moreover, networks with similar political assortativity can still generate different partisan environments around the same personas, including among networks with plausible structural properties. Finally, generated networks show group-specific errors, under- or overestimating exposure to either Democrats or Republicans for specific social groups. These findings show that evaluating LLM-based social simulations only through aggregate outcomes can hide errors in the local environments through which those outcomes emerge, highlighting the importance of validation at both aggregate and individual levels.

    benchmark
  146. arxiv:2610.23195 · cs.AI
    FuzzInsight: Navigating the Efficiency-Effectiveness Trade-off in Coverage-Guided DNN Fuzzing
    Vivek Vekariya, Fathiyeh Faghih, Alexander Pretschner

    Collecting and labeling diverse test datasets for evaluating Deep Neural Networks (DNNs) is costly and time-consuming. This has motivated coverage-guided fuzzing techniques that generate defect-revealing corner cases from limited test data. However, their effectiveness depends heavily on configuration choices, particularly coverage criteria and seed-selection strategies. We present a configuration-oriented analysis and decision guidelines for scalable DNN fuzzing targeting two bottlenecks: inefficient seed utilization and the cost of coverage guidance. Our analysis shows that existing seed-selection strategies repeatedly prioritize a small subset of inputs, leaving many collected seeds unexplored. To address this issue, we propose Equalized Sampling, a constraint-based strategy that increases mean seed queue utilization to 77%, improving on prior approaches by 13 percentage points. We further evaluate 84 fuzzing configurations on four benchmark datasets: MNIST, SVHN, GTSRB, and CIFAR-100. Results show that the non-structural criteria LSCD, LSC, and MDSC achieve coverage outcomes comparable to structural and alternative non-structural criteria while requiring substantially less computation. These criteria also promote broad behavioral exploration of the DNN under test, reducing the need to monitor multiple coverage metrics. Among them, LSCD provides a modest but consistent runtime advantage. Finally, we show that retraining with generated corner cases improves generalization robustness by up to 77 percentage points on average without degrading performance on the original test distribution. Together, these findings provide a practical blueprint for selecting fuzzing configurations that balance efficiency, behavioral exploration, and model hardening.

    benchmark
  147. arxiv:2610.23194 · cs.AI
    OfficeTown: Generative Organizations for Evolving Agent Intelligence
    Hongchen Wei, Daohong Ying, Zhi Hong, Aidan Wu +9

    AI agents increasingly use files and software tools to produce professional deliverables. Such work draws on materials, decisions, and feedback accumulated across assignments. We introduce OfficeTown, a generative simulation framework for professional organizations in which tasks and their context arise from continuing organizational work. From a natural-language description of an organization's purpose and activities, OfficeTown initializes an organization whose members coordinate work, delegate bounded assignments to assistant agents, and review their deliverables. Artifacts, decisions, and unresolved matters persist and shape subsequent work, while responsibilities and permissions determine what each member can see and do. We curate work requests, their context, and review feedback into executable resources for evaluation and learning. We construct OfficeTown Bench with 406 tasks and 4,836 evaluation criteria from 23 simulated organizations across eight professional domains, incorporating domain-expert feedback. Frontier-model evaluations reveal substantial remaining headroom. Supervised fine-tuning and on-policy distillation on OfficeTown data substantially improve Qwen3.5-9B on external professional-work benchmarks, including GDPval. Supplementary evaluations show that the principal--assistant harness achieves higher scores than the single- and multi-agent baselines tested. These results highlight the value of organizational simulation in generating work-grounded data and environments for the continued development of generalist agents.

    agentai agentmulti-agentbenchmark
  148. arxiv:2610.23190 · cs.LG
    LangNCO: Compiling Natural-Language Constraints into Inference-Time Interventions for Neural Combinatorial Optimization
    Hironori Ohigashi

    Neural combinatorial optimization (NCO) rapidly produces high-quality solutions for problem structures encountered during training, but operational requirements may change after deployment and be communicated in natural language. We propose LangNCO, a framework that compiles instance-specific requirements into structured inference-time interventions for a frozen NCO solver, without changing the pretrained policy's architecture, learned input representation, or parameters. A pretrained LLM serves as a replaceable semantic front end, translating each requirement into a schema-validated domain-specific language (DSL) rule set. A deterministic Rule Executor applies the rules as nonexpansive hard action masks. Because local masking prevents immediate violations but does not ensure that a complete feasible continuation remains reachable, Critical-Offset Decoding extracts rule-referenced critical nodes and reallocates part of a fixed rollout budget to multi-strength biased rollouts. Using controlled task-specific language to isolate solver adaptation from open-ended language robustness, we evaluate a CVRP policy on road closures, time windows, vehicle assignments, visit-order requirements, all pairwise combinations, and twelve additional constraint types. LangNCO achieves 100% feasibility on the four primary datasets with mean reference gaps of 1.63--3.26%, 97.33--100% feasibility when separately generated rules are composed for pairwise constraints, and 98.67% feasibility across the additional constraints. These results demonstrate modular post-training adaptation of a frozen NCO solver through inspectable language-derived rules, without retraining.

    post-training
  149. arxiv:2610.23185 · cs.LG
    A Stronger Teacher Within: Strengthening MoE Models via On-Policy Bidirectional Distillation
    Haoyu Chen, Jun Bai, Yang Liu, Jianfei Zhang +1

    Reinforcement Learning with Verifiable Rewards (RLVR) has substantially improved Mixture-of-Experts (MoE) LLMs, but its outcome-level rewards provide only sparse supervision. On-Policy Distillation offers dense token-level feedback, yet requires a stronger compatible teacher, while on-policy self-distillation relies on hand-designed privileged information. We observe that the MoE architecture itself offers a way to resolve this dilemma, as increasing the activation budget per token exposes latent expert capacity and yields a potentially stronger teacher within the model. However, naive expansion degrades performance because newly co-activated experts are not optimized to work together. We therefore propose MoEvolve, a bidirectional distillation framework following a recovery-then-improvement strategy: we first restore the activation-expanded high-$k$ model and further strengthen it by exploiting its additional active capacity, then reverse the direction of distillation to transfer the enhanced capabilities back into the original low-$k$ configuration, retaining its inference-time activation budget. Across diverse MoE models and benchmarks, MoEvolve improves the initial models by up to $22.2$ domain-average points and outperforms same-budget RLVR baselines by up to $6.5$ points. Notably, the final low-$k$ models often surpass their activation-expanded teachers.

    benchmark
  150. arxiv:2610.23183 · cs.CL
    AMEND: Constraint-Guided Multi-Agent Text Detoxification with Isolated Critics
    Tianyu Wang

    Text detoxification requires balancing toxicity removal, semantic preservation, and fluency. Existing methods typically optimize an unconstrained scalar objective---enabling safety scores to mask semantic loss---or rely on open-ended iterative debate prone to semantic drift and high computational cost. We propose AMEND, an inference-time framework that decouples detoxification into role-conditioned proposal generation, isolated multidimensional criticism, and constraint-first selection. AMEND evaluates diverse candidates with role-blind critics and enforces strict semantic and fluency feasibility before ranking by utility. To maximize efficiency, AMEND Lite restricts subsequent search to at most one targeted repair round only when residual toxic spans remain. On ParaDetox, one-round selection over diverse roles improves the joint evaluation metric by $+0.0508$ over the best single role. On a frozen confirmation test set, AMEND Lite achieves a joint score of $0.6597 \pm 0.0062$, outperforming unconstrained multi-round loops by $+0.0489$ ($p < 0.05$) while reducing inference requests by 28.4%.

    multi-agent
  151. arxiv:2610.23182 · cs.CL
    What LLMs Do in the Shadows: Reading Context Reliance from the Unembedding Dark Space
    Alexandra Kuleshova, Andrey Volodichev, Daria Kotova, Alexey Zaytsev

    Detecting faithfulness hallucinations in large language models (LLMs) is crucial for their safe deployment in high-stakes domains. Recent approaches address this by contrasting contextual against parametric knowledge utilization, yet measuring context reliance typically demands noticeable overhead, such as materializing attention maps, running additional forward passes, or training auxiliary probes. In this work, we show that context reliance can be estimated by projecting hidden states onto the dark space - directions that compose weakly with the unembedding matrix. Beyond its recognized function as an initial-token attention sink, we show that this subspace actively mediates context processing, and that its share of hidden-state norm - termed the Dark Matter Ratio (DMR) - provides a direct score of context utilization. Pairing DMR with a parametric knowledge score yields DarkLens, a lightweight faithfulness detector operating strictly within a single forward pass. Evaluated across three QA and summarization benchmarks on six open-weight LLMs (4B-13B parameters), DarkLens matches or outperforms a wide range of baselines without requiring extra forward passes or probe training.

    benchmark
  152. arxiv:2610.23181 · cs.LG
    Tree-Aware Posterior Distillation for Long-Horizon Agentic Reinforcement Learning
    Heming Zou, Qi Wang, Yixiu Mao, Yun Qu +6

    Reinforcement learning with verifiable rewards has become a scalable post-training paradigm for large language models. However, training agents over long-horizon interactions presents two challenges: using only terminal rewards makes turn-level credit assignment sparse and high-variance, while sequential mini-batch updates on a fixed rollout batch introduce policy staleness as the current policy drifts from the behavior policy. Existing approaches obtain denser credit via learned critics, process reward models, or on-policy distillation, but they rely on a value network, step-level annotations, or a stronger external teacher that increase training complexity. Meanwhile, importance sampling and clipping mitigate policy drift, but they do not specify a finite target that stops sequential updates from overshooting. In this work, we introduce TreeOPSD, a critic-free, tree-aware posterior self-distillation framework for long-horizon agentic reinforcement learning. Treating each interaction turn as a tree node, TreeOPSD branches alternative continuations from shared prefixes and recursively backs up terminal rewards into implicit turn-level credit. It then maps each credit to a closed-form, bounded teacher distribution parameterized by the likelihood ratio relative to the frozen behavior policy. Minimizing the divergence between this teacher and the student policy recovers the policy-gradient direction as the tangent at the behavior policy while bounding updates toward a finite target across sequential mini-batches. Empirically, TreeOPSD improves agentic capabilities across several competitive benchmarks and model scales, especially for long-horizon tasks, e.g., raising the average accuracy of Qwen3-30B-A3B-Thinking-2507 by 4.97 points over GRPO on the Deep Research setting.

    agenticpost-trainingbenchmark
  153. arxiv:2610.23175 · cs.CL
    Do Speech BCIs Need Larger Models? Rethinking Neural Decoding beyond Scaling
    Zehui Feng, Weichuan Wang, Xiaohan Chen, Cuntai Guan +1

    Recent speech brain-computer interfaces (BCIs) increasingly rely on large models and complex multi-stage pipelines for neural decoding. However, in speech BCI, where training data are scarce, noisy, and expensive to collect, the benefits of model scaling remain unclear. This work revisits a fundamental question: are large models truly necessary for accurate neural decoding? We propose BEST (Brain-Encoding-Speech-Text), a two-stage framework for efficient neural-to-text decoding. Instead of relying on large generative models or phoneme-centric pipelines, BEST introduces hierarchical intermediate representations and a compact speech-oriented decoder that directly maps neural activity to text. This design preserves essential structure while significantly reducing computational cost, enabling robust generalization in low-resource settings and practical deployment under constrained resources. On the Brain-to-Text '24 and '25 benchmarks, BEST achieves word error rates (WERs) of 6.19% and 1.62%, respectively, using only 3.85% of the parameters required by comparable methods. Beyond these results, we provide three key insights: (1) decoder scaling yields non-monotonic gains; (2) cross-modal alignment is more critical than model size; and (3) overfitting degards sequence-level performance. Representation analysis further reveals a consistent neural-audio-text alignment pathway.

    benchmark
  154. arxiv:2610.23167 · cs.AI
    LongForge: Measuring and Improving Long-Context Instruction Data via Evidence-Assertion Reasoning Graphs
    Ziyang Chen, Junlong Jia, Xing Wu, Tinghao Yu +1

    Existing quality control for long-context instruction-tuning data is a one-shot judgment inside the generation pipeline that leaves only a scalar, revealing neither which document parts an answer uses nor how many reasoning steps it writes down; the data can thus be neither audited nor controlled. We formalize data quality as a property of the alignment structure among document, question, and answer, and build LongForge, which pairs an analyzer that measures long-context data with a synthesizer that generates it. The analyzer reverse-reconstructs each sample into an evidence-assertion reasoning graph and programmatically computes four dimensions on it: breadth, depth, correctness, and completeness. Every reading traces back to specific evidence spans and reasoning steps, making data auditable. Measuring seven mainstream datasets reveals two previously unquantified defects: breadth barely grows with document length, and depth is generally low, the two being nearly orthogonal. The synthesizer then embeds the analyzer into generation, gating on correctness and completeness and deferring breadth and depth to a training-time mixture over measured values, which makes the training distribution controllable. On two backbones, its two mixtures, LongForge-1 and LongForge-N, outperform the strongest baseline by 2.2-4.7 points on average over six long-context benchmarks, and in a controlled comparison that holds the gate and the candidate pool fixed, raising breadth and depth alone is worth a further 1.16 points, making the structural distribution of training data a lever on downstream quality rather than only a description of it.

    long-contextbenchmark
  155. arxiv:2610.23161 · cs.AI
    Algorithm Modelization: Transforming Optimization Algorithms into Evolvable Intelligence
    Jian Yao, Zhiyao Zhang, Yichen Li, Xingyu Wu +1

    Can algorithms be learned and evolved as models, rather than repeatedly redesigned as programs? We explore this question by introducing Algorithm Modelization, a new paradigm that turns iterative algorithms into trainable models. Its central idea is to regard algorithm execution trajectories as computational experience: by learning how algorithmic states evolve over iterations, a model can internalize the knowledge embodied in existing algorithms, reproduce their computational behaviors, and further evolve beyond them through training. As a first realization of this paradigm, we modelize iterative optimization as an autoregressive process, where preceding generations constitute the computational context for generating the next population. We develop a dimension-independent population-temporal architecture to capture both population interactions and iterative dynamics, together with an imitation-to-surpassing learning framework that first acquires algorithmic knowledge from heterogeneous teachers and then improves the resulting optimizer directly through optimization feedback. The modelized optimizer consequently goes beyond reproducing individual algorithms: it consolidates heterogeneous search knowledge into a single trainable model, achieves stronger optimization capability through subsequent improvement, and generalizes its acquired knowledge across problems and substantially larger unseen dimensions. More fundamentally, Algorithm Modelization makes algorithmic capability accumulable through learning, where new algorithms and problem-solving experience can become training resources for progressively improving the model. Our results establish a promising path from manually designed algorithms toward general algorithm models that learn, generalize, and evolve from computational experience.

    embodied
  156. arxiv:2610.23156 · cs.RO
    ORD-WM: Learning Object-Centric Rigid Dynamics for 3D World Modeling in Robotic Manipulation
    Yueyang Weng, Xiaopeng Zhang, Yuxi Ou, Yongjin Mu +2

    World models provide robots with a predictive mechanism for reasoning about how actions change object states in manipulation tasks. However, existing image-based and dense 3D world models do not explicitly impose object-level rigid motion, leaving rigid-object geometry preservation to be learned implicitly and making multi-step rollouts susceptible to shape drift or object-level motion inconsistency. To address this, we propose ORD-WM, a 3D world model that learns object-centric rigid dynamics for robotic manipulation. ORD-WM learns a vector-neuron variational autoencoder supervised by signed distance fields, which encodes each object point cloud as a geometry-aware latent representation composed of 3D vector features, and uses an action-conditioned dynamics model to predict object-wise SE(3) motion increments. Rather than generating unconstrained next latent states, ORD-WM analytically applies predicted object-wise rigid transformations to the current latent states, separating persistent object geometry from pose changes. Experiments show that ORD-WM enables stable, geometry-preserving multi-step prediction and supports efficient real-time replanning for closed-loop model predictive control in simulated and real-world manipulation scenarios.

    manipulationworld modelaction-conditioned
  157. arxiv:2610.23153 · cs.AI
    QuAD: Quantitative Assessment via Distribution-Alignment for Qualitative Data Quality Enabled by a Content-Aware Multi-Agent System
    Wenshuo Zhang, Leixian Shen, Qiqi Zhou, Shuchang Xu +5

    Assessing whether qualitative responses collected from semi-structured questions are ready for a particular analysis goal is a persistent challenge. Manual qualitative analysis is interpretive and labor-intensive, while existing automated methods often lack a dataset-level perspective for deciding whether a response set supports a researcher-selected analytical lens. We introduce QuAD, a framework for quantitative pre-analysis assessment of qualitative response datasets. QuAD compares a reconstructed intrinsic cluster distribution reconstructed from responses with an extrinsic, researcher-defined analytical lens derived from metadata. QuAD is structured into a four-quadrant assessment whose metrics are read jointly, from lens-level relations to within-cluster variation, to provide fitness-for-purpose signals for revealing general patterns and unique insights. Because this assessment relies on lens-conditioned inter-response dissimilarity, we propose a content-aware multi-agent system with two-stage instruction generation and adaptive anchors to improve scoring consistency. We validate QuAD through controlled experiments on synthetic data, including robustness and baseline comparisons, and a real-world social survey case study. The case study shows how QuAD quantitatively assesses whether a selected analysis axis is promising and also surfaces representative or unusual responses for human interpretation.

    multi-agentagent system
  158. arxiv:2610.23148 · cs.LG
    Grounded Credit Assignment in Reinforcement Learning for Multi-Turn Search Agents
    Qirui Liu, Hao Chen, Yi Wen, Weijie Shi +6

    Outcome-reward reinforcement learning such as GRPO gives every token of a search-agent rollout the same advantage, although the rollout exercises two capabilities: finding evidence and using it. We show that accuracy improves without grounding: Search-R1 raises accuracy on all seven QA benchmarks, yet on Bamboogle the evidence-supported share of correct answers falls from 84.0% to 42.9%. We attribute this to GRPO's credit assignment: rollouts that retrieve answer-bearing evidence are favoured in aggregate only in proportion to their accuracy gain over the rest, a grounding gap that answering from memory or misusing evidence shrinks or reverses. To address this, we propose Grounded Credit Assignment (GCA), which uses verifiable sub-outcomes (evidence retrieval and answer correctness) to correct search and answer credit separately: a wrong answer is blamed more than the search that found the evidence, and a correct answer that bypasses retrievable evidence is credited less. Never negative on evidence-finding search, the correction does not shrink with the grounding gap; capped at 10% of each token's advantage, it never overrides the outcome signal. Without extra models, rollouts, or training stages, GCA improves average EM over Search-R1 by 5.7 (Qwen2.5-3B) and 6.3 points (Qwen3-30B-A3B) and raises this share from 76.9% to 93.4%.

    memorybenchmark
  159. arxiv:2610.23145 · cs.LG
    Local proxies for memory-efficient training
    Wu S. Zihan, Brian A. Yuan, Guillaume Bellec

    Local learning algorithms can reshape resource requirements of artificial intelligence. The memory bottleneck of standard end-to-end backpropagation lies in the need to store all network activations until the loss gradient has been propagated back through. Local learning avoids this, but typically at a cost in accuracy. We study how to achieve accurate and memory efficient local learning in transformers. For each transformer block, we train a local proxy network to predict the global network output via distillation. Gradients can be independently propagated through this local proxy while estimating well the backpropagation gradients. In theory, this approach is exact for deep linear networks; in non-linear residual models, the gradient error can be bounded when using the appropriate residual proxy. We verify this numerically and demonstrate that optimizing the distillation loss improves gradient estimation. Evaluating the performance of proxy propagation on CIFAR-100 in vision transformers, we find that it achieves a better accuracy/memory trade-off compared to other local algorithms. Finally, we show that the method is further relevant for fine-tuning large language models on a small memory budget. Our method trains faster and achieves higher performance than memory-efficient zeroth-order learning algorithms with comparable memory cost.

    memory
  160. arxiv:2610.23141 · cs.AI
    LexPilot: Benchmarking the Proactive Interaction Ability of Legal LLMs with Laypeople in Chinese Civil Case Consultation
    Wenbin Zhang, ShangZhan Li, Bingxiang He, Ye He +4

    Large Language Models (LLMs) have greatly assisted legal professionals. However, our empirical studies reveal that, when faced with underspecified consultations from non-experts, current LLMs often produce irresponsible responses with limited utility despite lacking sufficient case details for definitive conclusions. To better assist laypeople, legal LLMs must identify focal points of dispute and proactively elicit crucial context from users. Unfortunately, existing data resources often focus on interaction and response quality, offering limited insight into how proactive interaction affects LLMs' downstream claim-level judgment prediction in actual civil cases. To address this gap, we introduce \textbf{LexPilot}, a benchmark linking proactive legal interaction to claim-level verdict prediction using authentic Chinese civil judgments. Constructed from 432 real-world judgments across 36 common causes of action, LexPilot comprises expert-reviewed pairs of complete and underspecified consultation histories with corresponding court verdicts. Experiments reveal that current LLMs struggle with proactive legal interactions in underspecified consultations, with downstream judgment prediction substantially underperforming that in the fully specified consultation setup (-23.84% on average). Further analysis identifies critical bottlenecks hindering these capabilities. Our work highlights the urgent need for more user-centered legal AI to better serve the public.

    benchmark
  161. arxiv:2610.23134 · cs.LG
    MetaCurate: Evolving Curation Filters for High-Quality Agentic Coding Data
    Donghoon Kim, Seunghyuk Oh, Kevin Galim, Minseo Kim +5

    Supervised fine-tuning on agent trajectories is an effective way to improve coding agents, but selecting useful training examples requires more than identifying successful runs. A resolved patch may follow inefficient reasoning or unnecessary tool use, while a failed run may still contain useful intermediate steps. Training pipelines therefore rely on rules and qualitative assessments to filter trajectories, yet designing these filters typically requires substantial human expertise, and validating improvements through repeated training and evaluation is expensive. We introduce MetaCurate, a training-free framework that evolves executable trajectory curation filters using pairwise LLM judgments. Disagreements between the filter's rankings and the judge's preferences provide targeted feedback for jointly revising its rules, rubric, and aggregation policy. The resulting filter scores trajectories independently, so pairwise comparisons are required only during search. Starting from a blank filter, MetaCurate autonomously discovers a curation strategy that achieves competitive downstream performance without task-specific handwritten criteria. We validate downstream utility through controlled SFT experiments on our uncurated trajectory pool and the Nemotron v3 SWE dataset, with random selection and hand-authored filtering as baselines. In our experiments, MetaCurate achieves the highest mean SWE-Verified score among the compared methods, including hand-authored rule-based filtering and a training-loop baseline whose search costs 3.5x as much.

    agentagentictool use
  162. arxiv:2610.23128 · cs.RO
    SCOPE: Selective Certification of Program Edits for Embodied Heuristic Learning
    Junyang Shu, Bingqing Wei, Jishuo Li, Zheqi He +1

    Embodied heuristic learning must balance strategic adaptation with the preservation of previously successful behavior. Local repairs may leave structural failures unresolved, whereas broader revisions can introduce regressions that are costly to detect through repeated environment interaction. We introduce SCOPE (Selective Certification Of Program Edits), which uses failed and successful executions as complementary evidence for program revision. Frontier-guided program search annotates a shared diagnostic graph with execution evidence to distinguish supported progress from unresolved checks. Selective behavioral certification uses stateful API replay to restore recorded responses and mutations, checking whether edits preserve selected historical successes. Diagnosis and certification feedback, supplemented by sparse environment diagnostics, guide consecutive repairs across configurations. SCOPE outperforms the strongest baseline in both simulation and real-world tasks. On LIBERO-PRO and Robosuite, it achieves mean success rates of 80.66% and 84.48%, with respective gains of 6.51 and 3.48 percentage points. Our code is available at https://github.com/VDIGPKU/SCOPE.

    embodiedlibero
  163. arxiv:2610.23127 · cs.RO
    BeJEPA: A Goal-Conditioned Humanoid Behavior World Model via Unsupervised RL
    Liyun Yan, Haoyang Li, Yang Zhang, Jialing Zhu +10

    Humanoid control requires coordinating whole-body posture with position and heading in the world frame. Mainstream controllers imitate trajectories in the root frame, but struggle to regain the target world pose after displacement. We introduce BeJEPA, a goal-conditioned Behavior World Model learned through unsupervised reinforcement learning from human motion priors and robot interaction. BeJEPA learns a shared representation of body configuration and world placement through a joint-embedding predictive architecture (JEPA). A Dynamic Predictor predicts the next state embedding under an action, while a Successor Predictor summarizes a goal-conditioned policy's expected discounted future state embeddings in a single vector. During joint training, Dynamic grounds the initial representation in observed physical transitions, while Successor increasingly organizes it around long-term behavior. Their composition guides policy improvement through one physical prediction and one behavioral summary, without long imagined rollouts. In 20-second motion-tracking experiments, BeJEPA with online action refinement using Dynamic outperforms all evaluated baselines in global body-position error, reducing it on AMASS by 61.7% relative to BFM-Zero with world-pose inputs and 58.7% relative to SONIC. Physical deployment on a Unitree G1 demonstrates joint attainment of posture, world position, and heading goals, recovery from the ground, and closed-loop return after disturbances and goal changes. Project page: https://ylyem9x.github.io/BeJEPA-project.

    humanoidworld model
  164. arxiv:2610.23123 · cs.RO
    GRASP: Global-Local Representation Alignment and Spatial-Physical Grounding for Bimanual Robotic Manipulation
    Zeyuan Ren, Guanxun Zhao, Qi Li, Hao Wang +14

    Vision-Language-Action (VLA) models that directly regress joint-space actions avoid the deployment-time cost of inverse kinematics. However, their training objectives remain misaligned with task success, as joint-space losses ignore the unequal physical consequences of joint errors and visual conditioning lacks explicit focus on hand-object interaction regions. To address these issues, we propose a training-phase prior injection framework for fine-grained bimanual manipulation. First, a physics-grounded flow-matching loss maps the implicit clean-action estimate at every noise level through differentiable forward kinematics (FK) and supervises the resulting Tool Center Point (TCP) pose and inter-arm relative pose, with a bounded noise-level weighting that compensates gradient decay. Second, we introduce a global prediction query group distilled from a frozen DINOv3 teacher through future-frame feature prediction, together with a local interaction query supervised by Grounding DINO at the attention level. Both teachers are removed at inference, and the deployment pipeline is that of a standard joint-space flow policy, adding only the learned query pathway. Simulation and real-robot experiments on long-horizon bimanual tasks show that the two priors are complementary across different deployment settings. The full model excels in in-distribution deployment, while the distilled query branch is preferable under strong visual randomization.

    vision-language-actionmanipulationgrasp
  165. arxiv:2610.23122 · cs.LG
    LED: Improving Masked Diffusion Language Models by Lookahead Editing
    Yuyan Zhou, Zeyuan Yang, Yuxuan Song, Zhicheng Cai +4

    Diffusion large language models (dLLMs) have emerged as a promising alternative to autoregressive language models, enabling non-autoregressive text generation through parallel decoding. Existing dLLMs can be broadly categorized into two families: masked diffusion models (MDMs) and uniform diffusion models (UDMs) In this work, we theoretically show that MDMs are easier to optimize due to their mask-based transition schedule but have lower computational expressivity, whereas UDMs enable iterative state revision and greater expressivity but are substantially harder to train. This trade-off raises a natural question: can we obtain the expressivity of iterative revision without giving up the simple, easy-to-optimize transition schedule of MDMs? Motivated by this analysis, we propose Lookahead Editing (\ours{}), which augments masked diffusion decoding with a rewritable proposal state. At each decoding step, the model performs several lookahead updates to iteratively refine future predictions in this proposal state before committing tokens to the masked sequence. This design introduces a rewritable proposal that uses additional compute to improve expressivity while keeping the simple mask-based transition schedule of MDMs. We fine-tune Qwen2.5-7B and Qwen3-8B to obtain LED-\texttt{7B} and LED-\texttt{8B}, respectively. Across eight benchmarks, LED SFT yields average gains of $+11.4$ and $+7.7$ points over standard MDM SFT. Moreover, using SFT alone, LED-\texttt{8B} outperforms LLaDA2.1-16B Q-mode, which is trained with continued pre-training, supervised fine-tuning, and reinforcement learning.

    benchmark
  166. arxiv:2610.23121 · cs.CV
    Scene-Designer: Learning Design Intelligence for Agentic 3D Scene Generation
    Xuanyu Liu, Hongyan Feng, Miao Pan, Sunlai Chen +4

    Generating complete, coherent, and usable 3D indoor scenes from natural language requires strong \textbf{design intelligence}: the ability to design high-quality scenes and critique their quality. However, existing approaches either rely on general-purpose models lacking domain-specific design experience or learn from static scenes. In this work, we introduce \textbf{Scene-Designer}, an agentic framework that learns design intelligence through complementary design and critique experiences. We construct two datasets: \textbf{Scene-OTA-35K}, which provides Observation--Thought--Action supervision for learning scene design and efficient error correction, and \textbf{Scene-OTC-15K}, which provides Observation--Thought--Critic supervision for learning scene quality critique across physical validity, functional correctness, and layout coherence. Based on these experiences, we propose \textbf{Scene-GRPO}, a Critic-guided GRPO framework that improves design decisions and reduces unnecessary trial-and-error corrections. Extensive experiments demonstrate that Scene-Designer achieves superior performance in physical validity, functional correctness, and layout coherence, while the learned Scene-Critic significantly outperforms general-purpose models in scene critique. Moreover, test-time Critic guidance further improves scene generation, showing that the Critic can provide useful guidance beyond training. Additional analysis suggests a trend of mutual enhancement between scene design and critique abilities.

    agentic
  167. arxiv:2610.23104 · cs.AI
    WebCoder: Scaling General Frontend Development with Artifact-Grounded Agentic Reinforcement Learning
    Zijun Dou, Zhen Yang, Xijun Liu, Dongzhuoran Zhou +2

    Real-world frontend development requires agents to compose reproduction, generation, modification, and repair over an evolving application state. Learning a general policy for this process requires translating heterogeneous execution feedback into task-aligned rewards and effective credit assignment across long sequences of interdependent actions. We introduce WebCoder, a general frontend development agent trained with artifact-grounded agentic reinforcement learning. Our key idea is to assess diverse development activities through their effects on a shared executable artifact and its satisfaction of task requirements. We build Webdev-World, an executable environment that integrates code editing, application execution, browser inspection, and on-demand skills while preserving project state across interactions. An independent verifier evaluates the final application against task requirements to provide terminal rewards, while changes in intermediate artifact progress redistribute the resulting rollout advantage across development segments. Context compaction and message-tree sample construction support training on trajectories beyond a single context window while avoiding repeated loss computation on shared generated history. We further introduce Webdev-Bench to evaluate composed development requirements within a single project. On Webdev-Bench, WebCoder-27B raises requirement satisfaction from 64.8 for its Qwen3.8-27B base model to 72.8, exceeding Claude-Opus-4.8 (65.1) and GPT-5.4 (59.6); on Vision2Web, it improves the average visual and functional score from 41.1 to 62.4, on par with Claude-Opus-4.8. Ablation studies confirm the contributions of artifact-grounded credit assignment and long-horizon training.

    agentagentic
  168. arxiv:2610.23098 · cs.RO
    InstAgent: A Recursively Self-Improving Agent Operating System for Heterogeneous Robot Collaboration
    Haoyang Luo, Junhao Qin, Jingyan Cao, Yi Huang +6

    Heterogeneous robot teams enable complex, long-horizon tasks by combining complementary sensing, mobility, and manipulation capabilities across multiple embodiments. Yet existing systems struggle to retain execution-time corrections as reusable experience, maintain task-relevant context memory across space and time for reliable handoffs, and explicitly represent the inter-robot dependencies required for structured collaboration. In order to address these challenges, we present InstAgent, a recursively self-improving (RSI) Agent Operating System for heterogeneous robot collaboration. InstAgent organizes team behaviors as agentic behavior trees (BTs) that explicitly model capability assignment, inter-robot dependencies and handoff checks. The RSI mechanism updates these collaboration programs from execution feedback and preserves successful revisions for reuse in future tasks, while shared multimodal spatio-temporal memory provides context for planning and handoff validation. Experiments on 20 real-world multi-robot tasks with three high-level models show consistent gains in task success, first-attempt completion, and execution efficiency.

    manipulationmemoryagentagenticself-improving
  169. arxiv:2610.23093 · cs.CL
    Disentangling Language Effects on Tool Admission and Post-Retrieval Safety
    Aditya Nawal, Manit Baser, Mohan Gurusamy

    LLM agents increasingly use retrieval tools across languages, yet their safety evaluation remains English-centric. It remains unclear whether language affects tool admission, post-retrieval generation, or both. We independently vary request, retrieved-page, and required response languages in a controlled URL-retrieval setting across ten models and seven languages. We identify Safety Loss through Interlingual Processing (SLIP). Non-English harmful requests increase mean tool admission from 46.9% to 67.4%, with a positive mean change in every model. This pattern also persists in an autonomous retrieval setup across all four models tested in this setting. Requiring answers in the page language increases pooled harmfulness on matched tool calls from 50.3% to 57.4% across four closed-weight models, mainly in Chinese and Hindi. Translating pages alone produces no consistent harm increase in eight models with comparable inference settings. Automated audits show that a non-harmful response may be a genuine safety refusal, a deflection citing language difficulties, or an unusable answer. Harmfulness labels alone cannot distinguish these outcomes. Multilingual safety evaluations should therefore measure tool admission and response harmfulness separately and distinguish refusal from capability failure.

    llm agent
  170. arxiv:2610.23092 · cs.MA
    Towards High Semantic Fidelity: Hyperdimensional Symbolic Messages in Multi-Agent Communication
    Shoucheng Song, Chang Yao, Hao Wu, Youfang Lin +2

    Semantic fidelity is essential for multi-agent communication under non-ideal channels, comprising transmission fidelity and conversion fidelity. Yet existing numerical and neural message formats cannot simultaneously achieve both. To address this, we propose a novel message format-\textbf{H}yperdimensional \textbf{S}ymbolic \textbf{M}essages (HSM) and its corresponding communication framework \textbf{Herm}. For transmission fidelity, Herm suppresses semantic entanglement via an entity-attribute-value structured encoding pipeline. For conversion fidelity, Herm first leverages a receiver-centric ego-binding mechanism to align semantics. Then, a similarity-aware dual-branch fusion module is proposed to prevent critical semantic dilution. To realize explicit message-semantic conversion, we devise an iterative decoding scheme via algebraic inverse operations. Experiments exhibit that Herm outperforms baselines across various noisy scenarios while maintaining superior structural readability.

    multi-agent
  171. arxiv:2610.23085 · cs.AI
    Rethinking GRPO for Large Reasoning Model Safety Alignment
    Jiawei Chen, Yi Zhang, Jiawei Du, Zhonghou Lv +8

    Large reasoning models (LRMs) exhibit strong reasoning capabilities but remain vulnerable to unsafe outputs. While Group Relative Policy Optimization (GRPO) is widely used for safety alignment, its group-relative optimization may provide insufficient safety-directed learning signals. In this paper, we identify two failure modes: advantage collapse, where homogeneous rollout groups yield vanishing advantages, and unsafe diversification, where unsafe behaviors persist and become increasingly diverse. Our theoretical analysis reveals that behavioral consistency weakens within-group safety contrasts, while entropy regularization can diversify unsafe responses when relative signals vanish. To address these challenges, we propose Beacon, comprising Safety Signal Recovery (SSR) and Safety-Utility Decoupled Alignment (SUDA). SSR combines advantage-guided trajectory filtering and safe reference supervision to recover effective safety learning signals, while SUDA employs category-aware rewards and benign-only KL anchoring to mitigate over-refusal without compromising safety. Experiments across four Qwen3 and DeepSeek-R1-Distill variants and eight benchmarks show that our framework achieves the best average safety and reasoning performance, reducing average safety error by up to 32.83 percentage points while maintaining low over-refusal.

    benchmark
  172. arxiv:2610.23078 · cs.MA
    Cryptographically Secure Agent-to-Agent Communication in Multi-Agent Systems
    Karol Budzyn, Adrian Brodzik, Mateusz Szczepkowski

    Multi-agent systems (MAS) comprise AI agents that collaborate or compete to perform complex, distributed tasks. Their dynamic interactions create opportunities for threat actors to target and impersonate other agents, intercept messages, inject malicious prompts, and exfiltrate sensitive data while simultaneously complicating the detection of anomalous behavior through conventional monitoring. This paper surveys threats and cryptographic challenges in agent-to-agent communication and presents a zero-trust architecture integrating identity, transport security, message integrity, and monitoring. The architecture establishes agent identities via decentralized identifiers and task-scoped verifiable credentials backed by a permissioned ledger, enforcing least-privilege authorization. For transport security, it provides mutually authenticated, crypto-agile TLS with hybrid post-quantum key establishment, and evaluates KEMTLS with pre-distributed public keys to reduce handshake overhead. Signed CBOR envelopes preserve payload integrity across multi-hop exchanges, while zero-knowledge proofs verify declared task constraints without exposing underlying inputs. Context-aware monitoring of inter-agent messages, tool activity, and execution telemetry informs the setting of risk-based trust thresholds and supports the detection of behavioral anomalies. Together, these mechanisms provide an integrated framework for securing and auditing agent communication, with implications for other agentic protocols and prospective 6G semantic communication.

    agentai agentmulti-agentagenticagent system
  173. arxiv:2610.23073 · cs.LG
    Path to Prior-Data Fitted Relational Models via Learned Latent Multi-table Representations
    Mohammad Sadeq Abolhasani, Viswanath Ganapathy

    Prior-data fitted networks (PFNs) show that pretraining on large populations of synthetic tasks can enable generalization to unseen datasets with little or no downstream fine-tuning. This paradigm is effective for transformer-based tabular and relational foundation models such as RDB-PFN, but remains largely unexplored for graph-based relational models built on message passing. We pretrain the graph-based relational foundation model Griffin on large collections of synthetic relational databases and ask whether curriculum learning, which benefits transformer-based relational learners, also improves message-passing representations. Rather than fine-tune on downstream tasks, we use the pretrained encoder as a general-purpose relational representation learner and pass its latent representations to a frozen in-context predictor (\ticl{}); no component is trained on the target task. On 16 real-world relational classification tasks, evaluated over five seeds against raw-feature and untrained-encoder baselines, a purely synthetic curriculum that begins with schema-guided data and continues with schema-agnostic data gives the strongest frozen in-context performance\rev{, outperforming real-data-only pretraining and every unordered single-stage mixture of the corpora; when appended after real-data pretraining, it yields the best encoders overall}. These results provide initial evidence that PFN-style synthetic pretraining transfers to graph-based relational foundation models.

    curriculum learning
  174. arxiv:2610.23071 · cs.CL
    Before Retrieval Begins: Safety and Tool-Use Decisions in LLM Agents
    Aditya Nawal, Manit Baser, Mohan Gurusamy

    Large language models increasingly act as agents that call external tools, yet most tool-safety research examines only what follows a tool call. The prior decision of whether to call the tool remains largely unexamined and can influence downstream safety. Across the seven open-weight models, the unweighted model-average harm rate after admitted harmful requests is 74.5%. To identify what governs this decision, we apply a factorial intervention to the input request. Within our generated construction, neutralizing the requested-operation component produces the largest admission contrast (19 pp). A linear direction estimated from this contrast influences held-out admission in four of seven models; effects on operation-neutralized requests in some models show that it is not an operation-specific mechanism. We then test whether written refusal is behaviorally and representationally coupled to the same decision. Written refusal accompanies 33.6% of admitted harmful calls. Plain-text refusal training reduces unsafe admission in models that retain benign tool use, showing that response-level safety training can transfer to the call decision. Together, these results establish a partial, model-dependent relation between written refusal and tool admission. They motivate direct evaluation and supervision of tool admission rather than treating text refusal as its proxy.

    llm agenttool usetool-use
  175. arxiv:2610.23068 · cs.RO
    Foundation Models for Autonomous Space Robotics: Architectures, Physical Adaptation, and Validation
    Zizhe Wang, Jia Deng, Xin Lu, Zehao Jia +3

    Space robots must interpret tasks and act despite unfamiliar dynamics, difficult sensing conditions, and limited opportunities for human intervention. Foundation models offer reusable perceptual and behavioral knowledge, but their value depends on how that knowledge reaches physical execution. This critical review follows space-specific requirements through perception, prediction, planning, and action to deployment and validation, emphasizing orbital manipulation and proximity operations. Terrestrial comparisons establish benefits of particular pretraining stages and model interfaces; their transfer to space dynamics and constrained onboard operation remains insufficiently tested. Event cameras offer timely motion observations across strong illumination contrasts, while spiking neural networks provide possible routes to sparse local computation. We examine how sensor degradation, state estimation and computing choices affect feedback and recovery. Higher action throughput need not yield fresher feedback; lower processor power can also coexist with higher energy per inference. We organize validation as an iterative sim-to-ground-to-space pathway, connecting physical tests with telemetry-based diagnosis and closed-loop revalidation. A satellite-capture example links observations to approach, contact and recovery, while a corresponding protocol separates generalization from adaptation with limited target-domain interaction. These comparisons identify what models already contribute and the physical evidence needed to support autonomous space operation.

    manipulationevent camera
  176. arxiv:2610.23062 · cs.CL
    ParetoChat: An LLM-Based Interface for Interactive Multi-Objective Optimization
    Gokhan Ceyhan

    Multi-objective optimization models are often deployed in settings where decision makers (DMs) must balance some conflicting criteria. In practice, however, stakeholders must translate their qualitative preferences between different criteria into numerical weights, priorities, or reference points required by optimization methods, which can be difficult to specify and interpret. This paper proposes \emph{ParetoChat}, an LLM (large language model)-based expert-agent interface for interactive multi-objective optimization. The agent translates natural-language user input into structured scalarization parameters, invokes an optimization system, presents candidate solutions, and iteratively elicits feedback. We describe a multi-agent architecture that separates input processing, multi-criteria decision-making (MCDM) expertise, solution presentation, and feedback processing. We then present a simulation study using an LLM-based DM agent and benchmark ParetoChat against interactive methods implemented in DESDEO, including reference-point and NIMBUS methods. The computational experiments on multi-objective binary knapsack and set-partitioning instances show that ParetoChat can guide the search efficiently toward solutions that match the DM's preferences. An application to historical courier-order assignment instances illustrates how users can explore objective extremes and intermediate trade-offs through natural-language requests. The results indicate that LLM agents may provide a practical interaction layer between business stakeholders and optimization tools.

    agentllm agentmulti-agentbenchmark
  177. arxiv:2610.23056 · cs.MA
    Thought Blocks: Non-Autoregressive Latent Communication for Multi-Agent Systems via Trajectory-to-State Compilation
    Zhenhan Guan, Xiaolong Liu, Hengwei Ye, Zheng Tian

    LLM-based multi-agent systems typically communicate through text, incurring repeated generation and re-encoding costs. Although latent communication removes explicit verbalization, constructing latent messages may still require autoregressive latent trajectories. We therefore distinguish message representation from message construction and ask whether agents can directly construct useful continuation states without executing the trajectories that normally produce them. We formulate this problem as trajectory-to-state compilation and introduce Thought Blocks, fixed-size blocks of latent key--value (KV) continuation memory constructed in one shot. Thought Blocks achieve a 2.5 times average latency speedup over autoregressive latent communication. Compared with text-based multi-agent communication, Thought Blocks reduce text tokens by 45.7% and autoregressive steps by 84.8%, achieving an 8.6 times average latency speedup. These results demonstrate that trajectory-to-state compilation enables efficient inter-agent latent communication without sequential latent rollouts at inference.

    memorymulti-agentagent system
  178. arxiv:2610.23053 · cs.RO
    Unifying Embodied Foundation Models via a Model-to-Policy Formulation
    Xinyi Hong, Mingyang Ren

    Vision-language-action (VLA) models and world-action models (WAMs) have emerged as promising approaches to robot learning and control. These models integrate multimodal perception and action generation, but differ in their architectures, output representations, and learning objectives. In this paper, we present a unified, architecture-agnostic probabilistic formulation for embodied foundation models based on stochastic kernels. Our formulation has two key components: (i) we characterize embodied foundation models through conditional output distributions and action-realization kernels, which together induce task-conditioned robot policies; and (ii) we connect pretraining and post-training to policy deployment, accommodating model-specific pretraining objectives and offline or online post-training strategies. We instantiate the formulation for OpenVLA and DreamZero, covering autoregressive action-token prediction and joint video-action generation.

    vision-language-actionembodiedopenvlapost-training
  179. arxiv:2610.23051 · cs.AI
    Agentic In-Context Learning: What Algorithms Can Transformers Learn in Closed-Loop Interaction?
    Shiguang Wu, Yaqing Wang, Quanming Yao

    The learning algorithm view of in-context learning (ICL) studies how fixed-parameter Transformers learn from supplied examples. Language model agents extend this setting through interaction: their actions determine what evidence enters context next. We call this regime Agentic In-Context Learning (Agentic ICL). Existing studies largely depend on externally supplied contexts or reward-bearing interaction histories, while work on general tool-using agents lacks a learning algorithm view. We represent an interactive learner by its state, update, and decision rule, covering active learning, bandits, and online reinforcement learning (RL). For expressivity, we construct causal Transformers that approximate these algorithms over a finite horizon, and bound the resulting loss in return. For training, we characterize when optimizing process rewards preserves behavior that maximizes the final outcome. Distillation experiments show that Transformers can learn to execute three representative recursions. We then train models from outcome or process rewards without oracle action or state targets, as in RL post-training. Different rewards yield distinct decision policies. Improving the alignment of process supervision with outcome-optimal action preferences shifts learned decisions accordingly. Such behavioral differences persist in a pretrained language model across textual paraphrases.

    agenticpost-training
  180. arxiv:2610.23049 · eess.SY
    Characterization, ageing and post-mortem dataset of commercial 58Ah prismatic NMC cells
    Róbinsón Medina, Alessio A. Lodge, Feye S. J. Hoekstra, Steven Wilkins +9

    This paper provides an extensive dataset for commercial 58 Ah prismatic NMC cells. The dataset comprises three types of data: characterization, ageing, and cell teardown. Characterization data include tests such as GITT, HPPC, and EIS, among others. Ageing data were obtained from 41 cells subjected to 38 distinct ageing conditions, including application-surrogate drive cycles. Recurring reference performance tests and EIS measurements track the evolution of each cell. Teardown data were obtained from pristine and aged (post-mortem) cells. They include material analyses and half-cell measurements. This single dataset supports analytics, model development (both machine-learning and physics-based), benchmarking, and techno-economic analysis. Several example analyses are carried out to show the usability of the data. For instance, the paper shows how lithium stripping can be detected from the voltage response of a short discharge applied after each charge. The data and the processing and analysis scripts are openly provided alongside this paper.

    benchmark
  181. arxiv:2610.23048 · cs.AI
    Distill Until Divergence: Error-Truncated Hindsight Distillation for Agentic RL
    Yibin Huang, Conghui Zhu

    Long-horizon interaction requires agents to make multiple decisions in a changing environment. Outcome-based GRPO lacks fine-grained guidance for intermediate decisions. Recent methods distill completed trajectories into reusable skills and use hindsight-conditioned OPD to provide dense token-level credit. However, skill-induced probability shifts do not necessarily represent reward-aligned credit; especially after an erroneous action changes subsequent interaction states, skill-conditioned supervision may become mismatched with the current state. To address this mismatch, we propose \methodname{}, which routes two complementary hindsight skills according to trajectory outcomes and error boundaries: Positive skills capture successful workflows to reinforce successful trajectories and pre-error decisions, while Negative skills capture failure-specific corrections to target decisive errors. The post-error suffix is optimized exclusively by GRPO, retaining outcome-based learning without extending potentially mismatched hindsight supervision. Extensive experiments demonstrate competitive task performance and support effective fine-grained credit allocation.

    agentic
  182. arxiv:2610.23046 · cs.CV
    SpikeQ: Spike-Guided Post-Training Quantization for Spiking Neural Networks
    Kai Sun, Peibo Duan, Zhanglu Yan, Guowei Zhang +2

    On neuromorphic chips, spiking neural networks (SNNs) consume far less energy than artificial neural networks (ANNs), but the limited memory of these chips requires low-bit synaptic weights and membrane potentials. Existing post-training quantization (PTQ) methods provide them without retraining and, like those for ANNs, minimize numerical errors such as the mean squared error (MSE). In an SNN, however, a quantization error reaches later layers only through the spikes it flips. We find that errors of equal MSE can cause very different accuracy losses, whereas errors that flip the same fraction of spikes cause the same loss. We therefore propose SpikeQ, a spike-guided PTQ framework. For weights, we select the level of each weight by simulating the downstream neurons and minimizing their spike errors. For membrane potentials, most existing methods quantize the state after the reset, whereas neuromorphic chips compare the threshold with the membrane potential stored at a fixed bit width. We target this pre-fire quantization and build a reset-aware grid that places the threshold on a rounding boundary, so rounding never flips the spike at the current time step. We evaluate SpikeQ on seven benchmarks of image classification, semantic segmentation, and object detection. With 2-bit pre-fire membrane potentials, every existing grid collapses, whereas SpikeQ retains most of the full-precision accuracy. With both at 4 bits, the models stay within 2.59 points of full precision, shrink their memory footprint 7.8--7.9 times, and use 6.4--6.9 times less energy.

    memorypost-trainingbenchmark
  183. arxiv:2610.23044 · cs.AI
    Where Does the Performance Go? Understanding and Recovering Performance Regressions in C-to-Rust Translation
    Zhiqiang Yuan, Chong Wang, Zhenhao Zhou, Junwei Liu +1

    Rust is increasingly adopted as a memory-safe successor to C, and transpilers such as \crust{} can migrate entire C projects while preserving their behavior. Preserving behavior, however, does not guarantee preserving performance. We study this gap on 20 real-world C projects and their \crust{} translations. On the eight projects used for analysis, the translated Rust is more than 3\% slower than the original C on 12 of 19 operations, by up to 62.2\%. By comparing execution-cost profiles, optimized LLVM IR, and compiler optimization remarks across each C--Rust pair, we trace these slowdowns to nine recurring performance issues, spanning residual runtime work, missed compiler optimizations, and C-style source patterns, and derive an optimization direction for each. We then present \ourtool{}, an LLM-based approach that encodes these issues as reusable optimization rules to recover the performance of translated Rust. \ourtool{} profiles each project to locate hot functions and matches them against rule-specific patterns. The LLM then checks each rule's rewrite preconditions against the project context before rewriting, and a rewrite is kept only if it passes both functional and performance validation. On 41 operations from twelve held-out projects, \ourtool{} preserves C outputs on all tested workloads and speeds up 40 operations over the \crust{} translations, 18 of which end up faster than the original C. It outperforms the general-purpose coding agent \codex{} on 31 of the 37 operations where both preserve C outputs, and outperforms its own variant without rule guidance on 34 of the 41 operations.

    agent
  184. arxiv:2610.23042 · cs.LG
    GR-OPD: Group-Relative Sequence Credit for On-Policy Distillation
    Yushu Wang, Renkai Yang, Jingcao Xu, Zili Geng +3

    On-policy distillation (OPD) provides token-level teacher feedback, yet common sampled-token updates assign credit locally and omit the effects of earlier actions on subsequent feedback. We introduce \method{} (Group-Relative On-Policy Distillation), which decouples feedback computation from credit assignment by aggregating token feedback into calibrated sequence-level credit. We establish a theoretical basis for this approach by showing that, under ideal on-policy assumptions, broadcasting raw summed feedback to every response token yields an unbiased estimate of the negative gradient of the sequence-level reverse-KL objective. To support effective learning in practice, \method{} discounts later feedback that may become less reliable under distribution shift and uses same-prompt rollout groups to calibrate teacher guidance. It preserves relative response preferences while retaining a controlled fraction of the group mean, which we term collective teacher pressure. Across five Qwen3 teacher--student settings and six mathematical reasoning benchmarks, \method{} outperforms vanilla OPD in macro Avg@K in every setting, including gains of 10.9 and 11.5 percentage points in two settings with matched thinking patterns. It also maintains stable training under mismatched thinking patterns, where vanilla OPD collapses. Ablations show that retaining an appropriate amount of collective teacher pressure can improve learning, while excessive collective teacher pressure or an unrestricted feedback horizon can destabilize training.

    benchmark
  185. arxiv:2610.23038 · cs.AI
    Where Research Ideation Begins: Forward-Citation Supervision for Actionable Limitation Generation
    Liyang Zhao, Yong Cao, Haibin Ling

    Automated research ideation has advanced rapidly, with multi-agent workflows that generate whole research ideas end to end. These frameworks leave largely unexamined the step that seeds a good idea: identifying an actionable limitation from literature. In this paper, we focus on this first step in the research ideation pipeline. Starting from a pilot study, we find that as large language models (LLMs) scale from 0.6B to 32B parameters, they become far better at recognizing which limitation later work addressed, but no better at generating one, suggesting that the bottleneck lies in the supervision signal rather than model capacity. We therefore construct ActLim-Bench from forward citations, yielding 18,279 evidence-linked supervision triples with human verification. On this signal, we propose a three-stage post-training pipeline: supervised fine-tuning, verifier-filtered rejection-sampling fine-tuning, and reinforcement learning with the verifier as an online dense reward. Experimental results show that our method achieves the highest recall score compared to existing methods among all frontier models. We further show that our generated limitations are preferred in 64.9% of non-tied human comparisons and improve downstream ideation systems. We will release our data and code upon publication.

    multi-agentpost-training
  186. arxiv:2610.23036 · cs.LG
    Towards Looped Models Done Right, Part I: Topology, Input Injection, Recurrent-State Design
    Benhao Huang, Chufan Shi, Junlin Chen, Shicheng Wen +3

    Looped language models reuse physical Transformer blocks across logical depth, but established architectures entangle recurrence placement, input injection, and recurrent state organization. We isolate these axes in models trained from scratch at matched parameter scale, logical depth, and token budgets, and then test whether the resulting design principles transfer to resident- and active-parameter-matched Mixture-of-Experts (MoE) models. The prelude-recurrent block-coda sandwich architecture consistently improves performance on context extraction and reasoning-intensive tasks. Input injection primarily enhances knowledge retrieval and context utilization, whereas random initialization and separate H/L recurrent states provide mixed or negative returns. Taken together, these results favor the Huginn-style design: at the 730M-parameter and 336B-token scale, the complete Huginn architecture outperforms Ouro across all ten benchmarks. The advantages of Huginn over Ouro transfer to MoE models. At the 8BA0.8B (32BA3B when unrolled) scale trained on 500B tokens, Huginn outperforms Ouro on nearly all benchmarks spanning knowledge, reasoning, and coding, while also exhibiting more balanced expert routing. Under matched training and inference FLOPs, the Huginn-style model uses 75% fewer resident parameters than a 112-layer feedforward MoE baseline. Despite its smaller physical parameters, it outperforms the feedforward baseline on DROP and GSM8K, matches it on MATH500, and remains competitive on the other benchmarks.

    benchmark
  187. arxiv:2610.23025 · cs.AI
    Causal Value Attribution of Marketing Labels in LLM-Generated Ads: A Deconfounder Approach
    Danhong You, Wei Jiang, Zhiqiang, Zheng

    As AI-generated content becomes commoditized, economic value increasingly shifts from final artifacts to the upstream expertise used to specify inputs that steer LLM generation. Yet organizations lack methods for valuing this expertise and identifying which individual skills or configurations causally improve performance. This study examines generative marketing, where structured expert labels serve as skill primitives conditioning LLM-generated advertisements. Estimating their causal effects is difficult because outcomes reflect stochastic LLM mediation, high-dimensional label interactions, and multimodal confounding. This study proposes D-VAE (Deconfounder Variational Autoencoder), a counterfactual-inference framework integrating a multi-treatment deconfounder with causal representation learning. D-VAE recovers causally disentangled expert-labeling logic and derives PAC-based bounds on omitted-variable bias from single-treatment confounders. Applied to 29,177 foundation-makeup ad creatives, D-VAE outperforms state-of-the-art benchmarks. A field experiment shows that recommended configurations increase engagement by 32.7% and perform statistically indistinguishably from human experts, while a simulation study confirms causal attribution under complex confounding and treatment interactions. Overall, the study shifts value assessment from AI-generated products to skills, providing a foundation for pricing and optimizing AI-augmented skill portfolios.

    benchmark
  188. arxiv:2610.23021 · cs.CV
    Semantic-Critical Regions for Video-to-Audio Generation
    Yun Fang, Wei Guo, Zhuonan Liang, Heng Wang +1

    In real-world videos, sound events are often associated with localized and fine-grained visual cues, such as object motion, object-part dynamics, contact events, and interactions between objects. Existing video-to-audio (V2A) methods employ implicit conditioning based on global visual representations or explicit control derived from temporal and spatial cues. However, subtle sound-producing and sound-indicative interactions remain insufficiently captured, leaving room for more precise visual conditioning of sound-event semantics and temporal dynamics. Motivated by this observation, we refer to regions containing such sound-relevant visual evidence as Semantic-Critical Regions (SCRs) and introduce SCR-V2A, a V2A generation framework that explicitly localizes and encodes SCRs as fine-grained visual guidance for audio generation. Specifically, we first identify SCRs through sound-related visual localization. We then transform the localized regions into a control representation through adapter-based dual visual encoding, which jointly captures global visual context and fine-grained sound-relevant evidence. The representation is used to condition a downstream audio generation model. Experiments on both in-domain and out-of-domain benchmarks demonstrate that SCR-V2A achieves strong performance in distribution matching and audio quality while remaining competitive in cross-modal alignment and requiring few trainable parameters. These results highlight the value of prioritizing fine-grained, sound-relevant visual regions for V2A generation.

    benchmark
  189. arxiv:2610.23019 · cs.RO
    PASS: A Trustworthy Precision Approach Simulation System for CAT III Integrity Algorithm Development
    Lang Zhu, Xin Zhang, Yulong Sun, Nan Liu +3

    For fixed-wing aircraft, reliable localization during approach and landing is critical to flight safety. Visual--inertial navigation is promising, but aircraft navigation also requires integrity: trustworthy outputs and timely warnings when they should not be used. Real approach data are costly to collect, while many visual errors and failures required for integrity algorithm development are difficult to reproduce safely. We present PASS, a trustworthy Precision Approach Simulation System for CAT III integrity algorithm development. In a unified temporal and geometric framework, PASS integrates configurable airport environments, prescribed trajectories, inertial simulation, calibrated stereo images, and continuous runway-relative ground truth. It supports navigation and integrity research through controllable sensor errors and environmental variations, computable visual projection errors, deterministic runway geometry priors, and condition-labeled data for statistical error modeling. We further propose Sim-to-Real Scene Trustworthiness (SRST), a measure based on DINOv2 feature distributions and UMAP, to quantify sim-to-real closeness. Experiments with representative stereo--inertial estimators demonstrate the usability of the generated sequences and quantify localization errors across algorithms and environments. PASS provides a controllable foundation for developing and evaluating visual--inertial navigation and integrity algorithms for fixed-wing precision approaches up to CAT III requirements.

    sim-to-real
  190. arxiv:2610.23018 · cs.RO
    VLA-Fabric: Learning Interaction for Composable Multi-Arm VLA Agents
    Chenglong Zhang, Shuaijun Liu, Feiyang You, Shuyang Hao +2

    Vision-language-action (VLA) models are rapidly expanding what individual robots can accomplish, yet many real-world manipulation settings require multiple arms to operate as a coordinated system. Existing approaches organize multi-arm coordination in different ways, including centralized joint policies, high-level task or role assignment, and locally executed policies that react to teammates. However, how complete action-producing VLAs should explicitly interact with one another during action generation remains largely unexplored. We introduce VLA-Fabric, which organizes one complete VLA per arm while preserving local perception and action ownership, and enables coordination through learned interaction across policy boundaries. We first systematically study where such interaction should occur, how peer information should be aggregated, and how the interaction should be learned, revealing complementary coordination functions spanning shared context, peer-specific latent information, and action-stage interaction. We then re-instantiate these functions in a structurally different VLA architecture, showing that the coordination principle is not tied to a particular tensor layout or backbone implementation. Finally, we compose teams of three and four arms and evaluate them in simulation and physical experiments. On all four physical tasks, Full interaction improves success over independently trained local policies by 16--40 percentage points. Together, these results suggest that explicit interaction among complete VLA policies offers a scalable path toward networked embodied intelligence.

    vision-language-actionvlaembodiedmanipulation
  191. arxiv:2610.23012 · cs.AI
    Adaptation of Computer Use Agents to New Software with User Manuals
    Chulhoon Yoon, Kyungri Park, Jinyeong Yun, Daeyun Jeong +2

    Recent advances in multimodal foundation models have significantly improved the ability to operate real-world software through graphical user interfaces (GUIs) using visual observations and natural language instructions. However, maintaining strong performance on unseen applications remains challenging, as practical tasks often require long-horizon procedural knowledge and application-specific operation patterns. In addition, constructing executable tasks and collecting corresponding trajectories for training still heavily rely on human-annotated data, incurring substantial cost. In this paper, we propose MAGE (Manual-augmented computer use AGent Execution and learning), a framework that leverages user manuals as a scalable source of procedural knowledge to improve both inference and training. During execution, MAGE retrieves task-relevant manual content and uses a planner to generate explicit procedural guidance for CUAs. To further account for the discrepancies between manuals and actual GUI environments, such as version differences or user settings, MAGE also adapts the agent policy through environment interaction. During training, MAGE converts manuals into executable task-plan pairs and uses the plans to guide environment rollouts toward task-relevant states; the resulting trajectories are then used to optimize the policy without human-authored demonstrations. This unified design enables effective utilization of manuals for both task execution and training, reducing reliance on human-authored training data. Experimental results across six applications demonstrate improvements in task success rates over strong baselines, highlighting the effectiveness of manual-guided planning and policy adaptation in new software environments. We provide the code and reproduction scripts at https://github.com/ku-autoai/MAGE.

    agent
  192. arxiv:2610.23006 · cs.AI
    Efficient CPU-GPU collaborative MoE inference via Adaptive Expert Assignment
    Wenhui Lv, Yunming Liao, Hongli Xu, Yang Xu +1

    Mixture-of-Experts (MoE) models reduce inference computation through sparse expert activation, but their large expert memory footprint often exceeds GPU capacity, making expert offloading necessary for memory-constrained inference. However, existing systems typically handle GPU-missed experts by either transferring them to the GPU, incurring substantial PCIe transfer latency, or executing them on the CPU, making the slower CPU computation the bottleneck. To address these limitations, we present \method, a CPU-GPU collaborative MoE inference system that adaptively assigns these missed experts to the CPU or GPU for efficient execution. By exploiting differences in expert token-loads, \method selectively transfers high-load experts to the GPU while executing low-load experts on the CPU. To minimize inference latency, \method adaptively adjusts expert assignments to balance the CPU computation path against the GPU transfer-and-computation path, allowing both execution paths to complete as early as possible. Extensive experiments on three representative MoE LLMs show that \method improves prefill speed by up to 51.9\% and decode throughput by up to 39.5\% over the strongest baseline across the evaluated batch sizes, with consistent gains under different GPU-memory budgets.

    memory
  193. arxiv:2610.23002 · cs.RO
    Robotic Spatiotemporal Editing: A Unified Interface for Controllable Action Generation via Masked Action Reconstruction
    Yanbang Li, Bu He, Yi Tian, Yu-Chiao Huang +5

    Generative policies can synthesize robot behaviors, yet controlling and editing them through diverse spatiotemporal references remains challenging. In the real world, human intent requires robot behaviors that satisfy constraints on where, when, and how actions are executed. Our key insight is that spatiotemporal control references, such as keypoints, flows, orientations, and keyposes, can be viewed as partial specifications of the same temporal Cartesian action space. We introduce the MAP (Masked Action Space) interface to unify these action references and formulate controllable manipulation as an action inpainting problem. Our Inpainting Policy (IP) reconstructs missing action dimensions of the end-effector, combining progress-guided alignment with multi-mask training to support unseen action-constraint combinations. We further adopt a training-free RePaint variant that iteratively inpaints the action during denoising to reduce control error. On ManiSkill and LIBERO-10, our policy achieves average task success rates of 92.67% and 64.67%, respectively.Real-world experiments demonstrate spatiotemporal control with average positional and rotational errors of 44.3 mm and 7°, respectively, showing that our policy can use diverse references to guide executable robot behaviors.

    manipulationlibero
  194. arxiv:2610.22991 · cs.CV
    A100: A Hybrid Vision Foundation Model for Agricultural Plant Image Analysis
    Mayur Pratim Das, Ajai Dubey, Aksshay Mathew P, Sana Afreen +4

    The use of accurate vision models for crop monitoring, disease diagnosis, and plant phenotyping in agriculture is hindered by the lack of expertly annotated data, especially in developing countries. We propose A100, an agricultural vision foundation model that is pre-trained using self-supervised learning on 3 million unlabeled images from fields covering more than 100 crop species. A100 uses a hybrid CNN-Transformer backbone that fuses MBConv stages for local feature extraction with Transformer stages for global contextual representation. A100 uses pre-training with the integration of DINO self-distillation, iBOT masked patch reconstruction, and KoLeo uniformity regularization, resulting in features that are better than ImageNet-12K supervised representations for agricultural applications without any fine-tuning. By utilizing a frozen backbone, the accuracy for crop classification of A100 is 95.37%, which is +2.56% better than ImageNet-12K, and reduces the leaf count error by 10.4% on the GroMo25 phenotyping benchmark, proving that the same backbone can be used for classification and regression tasks. After full fine-tuning, the performance of A100 is 99.90% on 32-class crop identification and 99.02% on 44-class disease detection. On the difficult PlantDoc dataset, A100 outperforms all baselines, such as CvT-13 (64.89% against 63.56%). Remarkably, A100 obtains 98.01% accuracy on plant classification using just 1% of the labelled data (about 284 images per category), showing great label efficiency for use cases with limited labeling. Under an identical protocol, A100 also outperforms the publicly released agricultural foundation models FoMo4Wheat and LeafVision by more than 14 points at 1% labels. We plan to publicly release our pretrained model to allow agricultural professionals to adapt A100 for various plant analysis applications using limited annotations.

    benchmark
  195. arxiv:2610.22990 · cs.LG
    PersonaCUABench: Benchmarking Computer-Use Agents Serving Diverse Personas
    Haoming Li, Defu Cao, Xu Hu, Yingtao Luo +22

    Computer-use agents (CUAs) are commonly evaluated against fixed task specifications, but users may omit requirements, add constraints, or require approval before consequential actions. Completing the benchmark instruction does not guarantee serving the person who asked. We introduce PersonaCUABench, which holds 74 tasks across 11 domains fixed and varies only who asks. A persona agent, playing one of 44 personas in four groups, rewrites each instruction in that user's own words and answers the CUA's questions, while the CUA sees only the dialogue, never the profile or the original instruction. Across nine backbones (3,330 runs), mean task completion falls by 36% relative to the original requests. The losses have two sources: in 59.9% of cells that lose points, the agent engages with what the user asked for and still gets it wrong, and 11.2% are stops the user requested that the checker scores as failures. Even when a persona request is fully completed, the user gives the top satisfaction rating in only 62.7% of cells, against 84.1% for the original requests. CUAs should therefore be evaluated not only on task completion but also on whether they meet the requirements and constraints users express. Code: https://github.com/MatrAIx-ai/PersonaCUABench. Project page: https://matraix.ai/research/persona-cua.html

    agentbenchmark
  196. arxiv:2610.22987 · cs.AI
    From Approval to Abuse: Privilege Escalation Attacks on Harness-Protected Coding Agents
    Xianglin Yang, Yechao Zhang, Jiahao Liu, Yulin Chen +3

    Coding agents use execution harnesses to constrain the consequences of prompt injection by enforcing permission boundaries. These systems deterministically intercept any out-of-sandbox action and prompt the user for approval. To reduce approval fatigue, users can approve allow-rules proposed by the agent, after which any future action matching the rule is allowed without a prompt. In this work, we show that this mechanism introduces a new risk: untrusted context can influence the permissions presented for approval, with consequences extending beyond the immediate task. We investigate privilege escalation targeting this permission-update process. Specifically, adversarial context planted in a repository can reliably steer an agent performing a routine task into requesting a plausibly justified, yet dangerously broad allow-rule. If the user approves the request, the saved rule can effectively blind the harness, enabling adversaries to execute malicious actions that would otherwise require approval, without triggering human oversight. Across production platforms, including Claude Code and Codex CLI, we demonstrate that agents successfully elicit the target allow-rules in 58% of experimental instances. Once acquired, these rules exhibit a massive semantic blast radius, covering 56.9% of category-specific capability test cases. To further quantify the downstream risk, we show that conditional on rule approval, our exploit achieves verified, silent goal completion in 47 of the 58 eligible instances (81.0%), undermining the safety promise of human-in-the-loop oversight.

    agenthuman-in-the-loop
  197. arxiv:2610.22985 · cs.AI
    Prism Attention: Proposal-Refined Index Sharing Mechanism for Efficient LLMs Inference
    Zhenxu Tian, Kebin Liu, Zhengwu Yang, Yi Su +7

    Long-context decoding in large language models (LLMs) is increasingly bottlenecked by the memory and computation required to attend over an ever-growing Key-Value (KV) cache at each autoregressive step. Although token-level sparse attention reduces this burden by restricting computation to a small subset of relevant past tokens, its practical gains are often offset by repeatedly recomputing layer-wise sparse indices as the context grows. A common remedy is to share sparse indices across layers by computing them only at designated anchor layers, but such static reuse ignores the layer-wise evolution of attention patterns, causing the reused indices to progressively deviate from the layer-specific optima and incur substantial attention loss. By revisiting cross-layer sparsity from the perspective of attention dynamics, we find that inter-layer salient token shift is highly localized: although the exact top-$κ$ indices differ across layers, the core high-attention tokens remain stable, with changes concentrated near the expanded boundary of previously selected regions. Motivated by this, we propose Prism Attention, a proposal-refined sparse attention framework. It replaces rigid cross-layer index reuse with a coarse-to-fine strategy, preserving layer-wise adaptability while retaining the efficiency benefits of cross-layer sharing. Extensive experiments show that Prism Attention achieves state-of-the-art performance among mainstream sparse attention methods on both reasoning and long-context benchmarks, while attaining a $3.4\times$ end-to-end decoding speedup over FlashAttention at 128K context length.

    memorylong-contextbenchmark
  198. arxiv:2610.22984 · cs.CL
    Self-Evolving Agents Can Repair Failures but Struggle to Generalize from Them
    Weixiang Zhao, Yichen Zhang, Yanyan Zhao, Jinming Bai +4

    Experience-driven self-evolution offers a promising path for large language model (LLM) agents to continually improve through interaction. Yet existing approaches often rely on successful trajectories or external evaluation signals, leaving unclear whether agents can evolve when such positive anchors are unavailable. We study this question through Cold-Start Failure-Only Evolution (CFOE), where agents start without accumulated experience and learn only from failed interactions without privileged guidance. Across three self-evolving paradigms, multiple LLM backbones, and four interactive environments, we uncover a consistent repair-generalization gap: failure-derived experience can help repair previously failed tasks, but rarely yields sustained improvements on unseen tasks. Further analysis shows that agents often misinterpret failures, while even verified-correct experience remains difficult to transfer. Relaxing CFOE reveals that verifier feedback and additional interaction provide limited gains, whereas successful trajectories substantially improve transfer. These findings suggest that current self-evolving agents are better at repairing observed failures than building generalizable capabilities from failure.

    self-evolving
  199. arxiv:2610.22982 · cs.CV
    UniField++: A Unified Agentic Framework with Loop Reasoning for MRI Field-Strength Enhancement
    Yiyang Lin, Chenhui Wang, Tianliang Yao, Shuaibo Li +3

    MRI field-strength enhancement seeks to combine the accessibility of low-field scanners with the image quality of high-field scanners. Yet, existing methods remain prone to anatomical hallucinations, as uniform processing leaves difficult brain regions insufficiently examined and scarce paired data limit their learning of how tissue contrast and anatomical visibility change across field strengths. To address these challenges, we propose UniField++, a two-agent framework that unifies enhancement tasks across multiple field-strength transitions, including 64mT-to-3T and 3T-to-7T. Unified learning allows both the VLM-based planner and the flow-matching-based enhancer to draw on field-dependent image characteristics observed across tasks, rather than relying on a single transition. Trained with radiologists' annotations of hallucination-prone regions and corresponding correction plans, the planner predicts regional risk and formulates targeted restoration strategies. Guided by these plans, the enhancer performs loop reasoning along the flow-matching trajectory, repeatedly processing high-risk regions for additional refinement. Together, cross-transition learning and risk-guided refinement support anatomically faithful enhancement across field strengths. Across the evaluated tasks, UniField++ improves PSNR by an average of 2.74 dB and SSIM by 14.05 percentage points over comparison methods, with corresponding gains of 3.97 dB and 13.92 percentage points over the ablation baseline. It also maintains enhancement quality on an external cohort with an unseen field strength and improves downstream whole-brain segmentation Macro Dice by an average of 0.1457 over the original low-field images. Code is available at https://github.com/CUHK-AIM-Group/UniField_plusplus.

    agenticagent framework
  200. arxiv:2610.22980 · cs.CL
    KiranaBench: Can a Language Model Run a Neighbourhood Shop?
    Bhuvan Arora, Ankita Karmakar, Chaitanya Jain, Dilesh Churiwal +9

    Large language model (LLM) agents are increasingly evaluated on long-horizon tasks, including autonomously running a business. Existing benchmarks for this setting are almost exclusively English-language, Western in retail context, and single-agent, with the agent observing the business directly. We study how these results transfer to a multi-agent Indic setting. We introduce KiranaBench, a simulated Indian general store in which four roles (manager, cashier, helper, and bookkeeper) are each played by a separate LLM agent. Interactions with customers and suppliers occur in English, Hindi, or Hinglish, and product demand is calibrated to India's national Household Consumption Expenditure Survey. Evaluating four open-weight LLMs, we find that all fall far short of a fixed-rule shopkeeper baseline: the best model earns only one-eighth of the baseline's profit, and two models finish below an unmanaged shop. Performance is also consistently lower for Hindi and Hinglish than for English supplier communication. KiranaBench offers evidence that current LLM agents are not yet reliable for autonomously operating small businesses in India, and provides a testbed for developing agents suited to settings that are long-horizon, collaborative, and multilingual.

    agentllm agentmulti-agentbenchmark
  201. arxiv:2610.22968 · cs.AI
    NL2SQL with Iterative Self-Refinement
    Anjali Vemula, Akshat Dasula

    Bridging the gap between human intent and machine logic remains a critical challenge in data access for non-technical users, yet Large Language Models (LLMs) remain unreliable: they hallucinate schema elements, generate syntactically invalid queries, or fail to capture user intent. These limitations are especially critical in domains such as insurance, where KPIs, regulatory constraints, and legacy relational systems demand high precision. We introduce a multi-agent, persona-driven NL2SQL framework that combines (1) domain relevance detection, (2) schema-grounded knowledge construction, (3) KPI-aware SQL synthesis, (4) syntax-level self-repair using sqlite execution feedback, and (5) semantic consistency checking via SQL-to-NL reverse engineering. Our system significantly reduces query hallucination and improves semantic fidelity compared to direct LLM prompting. Evaluated on 60 queries (50 from a HuggingFace text-to-SQL dataset and 10 custom domain queries), the system achieves 90% correctness using Mistral across seven coordinated LLM calls showing that this approach can substantially increase the reliability of NL2SQL pipelines in high-stakes environments.

    multi-agentself-refinement
  202. arxiv:2610.22964 · cs.CV
    ANCHOR: Audio-Visually Grounded Chain-of-Thought Reasoning Benchmark
    Joel Julin, Souraja Kundu, Liza Dahiya, George Z. Wei +5

    Multimodal large language models (MLLMs) perform well on audio-visual question answering, but correct answers do not necessarily imply that their stated reasoning is grounded in relevant perceptual evidence. Existing benchmarks mainly evaluate final-answer accuracy, offering limited insight into whether model-generated rationales correctly identify visual regions, temporal intervals, and audio events supporting predictions. We introduce ANCHOR, a benchmark and evaluation framework for measuring grounded-rationale alignment in multimodal reasoning, comprising 2,319 videos and 8,272 question-answer pairs with structured reasoning traces, temporal evidence, spatial bounding boxes, and salient audio cues, including a human-refined test set of 224 videos and 975 question-answer pairs. We separately measure answer correctness, hallucination, logical divergence, and conciseness, with direct spatial and temporal grounding measurements on a focused subset. Across the full benchmark, twelve contemporary MLLMs show a substantial alignment gap: models with comparable answer performance can differ markedly in whether their explanations are supported by annotated evidence. On a matched subset of 100 videos and 446 question-answer pairs, rankings remain highly stable across human-refined, Gemini-generated, and GPT-generated references (mean Pearson r=0.99), providing evidence against model-reference circularity within this subset. Grounding-cue ablations of three MLLMs on the full human-refined test set reveal a consistent modality hierarchy: removing video reduces accuracy by 38.4-41.8 points, while removing bounding boxes (3.2-7.7 points) and audio (4.2-5.6 points) causes smaller, complementary, model-dependent drops. These findings show answer-only evaluation obscures important grounding failures and position ANCHOR as a testbed for evaluating accurate and evidence-aligned audio-visual reasoning.

    benchmarkevaluation framework
  203. arxiv:2610.22959 · cs.CL
    SAFE-Mem: Secure Agentic Memory with Fine-Grained Authority and Governance-Aware Retrieval
    Haoming Li, Dingyi Kang, Xu Hu, Jinxin Yang +7

    Large language model (LLM) agents increasingly rely on persistent memory to support long-term reasoning and personalized interactions. However, information collected from external sources can introduce security risks when untrusted claims (i.e., atomic statements) are later treated as authorized preferences, policies, or permissions. Existing defenses typically rely on content-based filtering or coarse-grained provenance tracking, which either fail to prevent unauthorized memory use or block legitimate information, sacrificing memory utility. To address these limitations, we propose SAFE-Mem, a fine-grained memory governance framework that improves security while preserving legitimate memory use. First, SAFE-Mem annotates each claim with three governance labels (i.e., kind, authority, and scope) to capture its provenance and permitted use. Second, governance-aware retrieval matches query requirements with claim permissions to block unauthorized influence while retaining legitimate information. Third, a governance-enabled memory graph connects source sentences, claims, and entities to support provenance tracking and efficient retrieval. Across four LLM backbones on MPBench and LoCoMo, SAFE-Mem attains the lowest attack success among defenses that retain legitimate memory, at least 77% below the strongest of them, and the highest answer accuracy of all evaluated defenses, improving it by up to 15%. The code is available at https://github.com/LVenum/SAFE-Mem.

    memorypersistent memoryagentic
  204. arxiv:2610.22958 · cs.CV
    Stabilizing Illumination Conditioning in SSM-Based Low-Light Image Enhancement with Bottleneck Prompt Sidecars
    Wei Chen Ng, Qiyi Xie, Jian Ding Tan, Yan Kang Lee +2

    Low-light image enhancement (LLIE) requires more than brightness amplification because real low-light images often exhibit spatially varying illumination, amplified noise, color distortion, and local detail loss. Recent SSM- and Mamba-based restorers provide efficient long-range modeling, yet the route for introducing illumination priors into such backbones remains insufficiently studied. This study addresses safe illumination conditioning for SSM-based LLIE through a bottleneck-side PromptAdapter. A lightweight illumination estimator first predicts a single-channel illumination map, which is then converted into prompt tokens and injected through a residual sidecar adapter at the bottleneck, thereby avoiding direct modification of the state-space core. Under a locked three-seed evaluation protocol on LOL-v1, LOLv2-Real, and LOLv2-Synthetic, the PromptAdapter improves over the Base model by 0.94 dB, 1.24 dB, and 0.93 dB in PSNR, respectively, while also improving SSIM, LPIPS, and masked dark-region metrics. The clearest real-data evidence appears on LOLv2-Real. A Learned-Fixed control on LOL-v1 indicates that the present results support the utility of the sidecar conditioning pathway more strongly than a strict claim of illumination-conditioned superiority, and the LOLv2-Synthetic result remains stability-conditioned. These findings suggest that, in SSM-based LLIE, conditioning design can be as important as the conditioning signal itself.

    evaluation protocol
  205. arxiv:2610.22951 · cs.RO
    MambaAct: Efficient Mamba Action Experts for Flow-Matching Vision-Language-Action Models
    Khuyen Pham, Tran Nguyen Le, Edoardo Fazzari, Cesare Stefanini

    Vision-language-action (VLA) models have shown strong promise for generalist robotic manipulation, but their inference cost remains a major bottleneck for real-time control. This limitation is especially pronounced in flow-matching VLAs such as $π_{0.5}$, where the action expert is evaluated repeatedly during iterative denoising. We introduce MambaAct, an efficient Mamba action expert for flow-matching VLA policies. MambaAct retains the pretrained visual-language prefix encoder of $π_{0.5}$, while replacing its transformer-based action expert with a compact Mamba-based action expert. Within this expert, Mamba serves as the action mixer that models dependencies across the noisy continuous-action sequence. To preserve visual-language conditioning without repeated token-level cross-attention, MambaAct uses a lightweight Context Compressor to map Vision-language model (VLM) prefix tokens into a compact conditioning signal that gates and modulates each Mamba action block. MambaAct uses a $10\times$ smaller action expert and reduces flow-matching denoising latency by $64.8\%$ on LIBERO and $58.8\%$ in real-world deployment, while improving mean LIBERO success by $1.08$ percentage points across three independent fine-tuning seeds and maintaining gains on LIBERO-Plus and real-world benchmarks. Our code and model weights are publicly available at https://github.com/giakhuyendihoc/MambaAct.

    vision-language-actionvlamanipulationliberobenchmark
  206. arxiv:2610.22942 · cs.LG
    Aligning Language Models Towards Counterfactual Invariance
    Basar Demir, Emily Shen, Ksheeraj Sai Vepuri, Sourabh Kulkarni +4

    Small variations in prompt characteristics can meaningfully alter the semantics, tone, and contextual assumptions of large language model outputs. Such characteristics can involve group variables such as demographics or other sensitive latent attributes. Sensitivity to such input variables is undesirable when demographic information is causally irrelevant to the task, but might still be relevant for generation of semantically coherent responses. This makes the task of enforcing counterfactual invariance while preserving model utility difficult. To study this problem, we first show that standard model alignment recipes (such as reinforcement learning) increase differences in response quality across demographics, while improving overall response quality, ultimately amplifying what is known as counterfactual bias. To address this, we develop a policy gradient technique involving a group-level reward function that operates on matched pairs of counterfactual prompts. For completeness, we build a dataset of controlled counterfactual prompt pairs and design an evaluation that separately measures cases where demographic information is causally relevant to the answer. Finally, we validate our findings on several fairness benchmarks. We demonstrate that our methods reduce counterfactual differences while preserving overall response quality and that the models continue to adapt appropriately when such sensitive information is relevant.

    benchmark
  207. arxiv:2610.22940 · cs.CL
    Beyond Surface Cues: Robust LLM-Generated Korean Text Detection
    SungHo Kim, SangKeun Lee

    As large language models (LLMs) generate increasingly fluent text, the need for reliable detection of LLM-generated content has grown substantially. However, existing research remains heavily English-centric, leaving Korean largely underexplored. Moreover, current evaluations of Korean text detection often rely on superficial cues, leading to overly optimistic estimates of detection performance. To address these limitations, we introduce \textsc{KoLing}, a bias-controlled adversarial benchmark comprising 50.9k instances that suppresses superficial shortcuts while preserving semantic content and separately evaluates LLM-generated and LLM-modified text detection. We further propose KoLingProbe, an interpretable detector grounded in theoretical linguistics that models Korean-specific properties across multiple linguistic levels. Experiments show that KoLingProbe substantially outperforms strong baselines on all detection tasks while remaining robust to adversarial attacks. Our analyses further reveal that integrating complementary linguistic features across multiple levels yields both improved performance and interpretability.

    benchmark
  208. arxiv:2610.22937 · cs.LG
    Many Biases, One Circuit: Post-Training Tunes a Pretrained Cue-Routing Mechanism
    Prakhar Gupta, Terry Jingchen Zhang, Zhijing Jin

    Language models often abandon a correct answer when the prompt contains a cue, a hint toward a wrong option such as "I think the answer is B." Sycophancy and related biases show this failure. They are often studied separately and blamed on alignment. We test this view causally, tracing the attention heads behind these biases through the stages of two post-training pipelines. In multiple-choice tests on most models, four cue types rely largely on the same few heads, a shared circuit already present in pretrained models. Most base models follow cues in plain text. Fine-tuning a base model on about 200 generic assistant examples without cues activates these heads in chat format, so cue-following can be a side effect of assistant tuning. The final reinforcement learning (RL) stage strengthens the effect of cues in almost every comparison. In Tulu-3-8B and OLMo-2-7B, we swap the attention patterns of six middle layers between the RL checkpoint and its predecessor. For suggested-answer cues, this shifts each checkpoint's cue effect most of the way to the other's level. Post-training thus seems to change when and how strongly models use this circuit. Mitigation for these multiple-choice biases could target this circuit.

    post-training
  209. arxiv:2610.22932 · cs.RO
    Ego2Contact: Predicting Touch from Egocentric Human Video for Contact-Rich Robot Manipulation
    Qianqian Yang, Dayou Li, Bangya Liu, Zihao Zhu +10

    Contact locations and interaction forces are critical for contact-rich robot manipulation but are often ambiguous from vision alone. Measuring them on robots requires specialized tactile hardware and platform-specific data collection, which keeps robot tactile data scarce. Human egocentric video is far easier to scale but lacks touch. Recent predictors trained on egocentric recordings paired with tactile-glove measurements can predict this missing signal from video alone. These predictors, however, remain confined to human hands, which differ from robot hands in appearance and tactile layout. We present Ego2Contact, which learns a visual-tactile prior from human interactions and transfers it to robot hands. Its predictor fuses temporal scene context with hand-local visual features to infer bimanual tactile fields from the RGB video. Anatomical region correspondences align varying robot tactile layouts with the predictor's canonical grid, turning robot measurements into fine-tuning targets and predictions into policy inputs. Ego2Contact outperforms prior tactile predictors on human data, including unseen objects, and on robot data after adaptation. The adapted predictor provides tactile inputs for both simulated reinforcement learning and real-robot policy post-training. In simulation, policies using predicted touch outperform non-tactile baselines and recover part of the benefit of ground-truth tactile sensing. These results show that human-learned tactile priors can supply touch when direct tactile sensing is limited.

    manipulationtactilerobot policypost-training
  210. arxiv:2610.22930 · cs.LG
    Conformal Policy Selection: When to Deploy AI Decisions for Targeted Intervention
    Junyu Cao, Ruijiang Gao, Maytal Saar-Tsechansky, Mingzhang Yin

    Artificial-intelligence (AI) policies are increasingly used to make consequential decisions in medicine, finance, education, and public services, yet deploying them responsibly requires knowing not only whether a policy helps on average, but for whom. Existing tools for off-policy evaluation and safe policy learning provide population-level guarantees, but some individuals may experience adverse treatment effects. We recast targeted deployment as a multiple-testing problem: for each candidate unit, test whether the outcome under the AI policy would exceed a user-specified threshold, and control the false discovery rate (FDR) among the selected set. Our method, Conformal Policy Selection, combines weighted conformal $p$-values that account for the distributional shift from behavior to target policy with a two-step selection-and-pruning procedure. We prove finite-sample FDR control and establish a doubly robust asymptotic guarantee, where FDR remains controlled if either the outcome model or the propensity model is correctly specified. Synthetic experiments verify these guarantees across four misspecification regimes, and a semi-synthetic study built on real electronic-health-record covariates illustrates how the same FDR control can act as a deployment safety gate, bounding the harmful treatments that a globally trained policy would otherwise administer to a high-risk subpopulation where it misgeneralizes. The framework offers a principled method for safe AI deployment in high-stakes applications.

    policy evaluation
  211. arxiv:2610.22925 · cs.LG
    Scalable Levenberg-Marquardt Training Framework for Deep Learning
    Xianghao Wang, Johan Barthélemy, John Taylor

    The Levenberg-Marquardt (LMA) algorithm is a widely used nonlinear least-squares solver known for its rapid convergence. For small-scale neural networks, LMA provides better convergence and architectural robustness than first-order optimizers like Adam. However, LMA imposes prohibitive memory and computational costs. These costs are primarily from the Jacobian matrix, whose dimensions scale with both the number of samples and parameters. This bottleneck prevents its use in large-scale neural networks. Recent efforts scale LMA using preconditioned conjugate gradient (PCG) solvers to avoid materializing the Jacobian. However, PCG only approximates the solution, requiring many network propagations and a task-specific preconditioner. Other studies optimize LMA to reduce costs, but they restrict their methods to specific tasks or network architectures. As a result, LMA remains impractical for training general large-scale neural networks. In this paper, we propose LeMA, a distributed LMA algorithm for large-scale neural network training on multi-GPU platforms. Our contributions are twofold: First, an exact dual formulation of LMA removes the memory dependence on the parameter dimension. Second, a Jacobian sharding strategy computes Jacobian shards on the fly and in parallel to avoid materializing the full Jacobian. For large-scale neural networks, LeMA reduces the memory footprint of LMA from quadratic in the number of parameters to quadratic in the batch size. This makes exact LMA steps feasible for billion-parameter models without approximate PCG solves. Furthermore, for a convolutional neural network with 1,000,022,632 parameters on an 8-GPU node, LeMA achieves a 7.97x speedup over a single GPU, corresponding to a 99.7% strong scaling efficiency.

    memory
  212. arxiv:2610.22924 · cs.AI
    Sep-Latent: Separator Tokens as Natural Compression Anchors for Implicit Reasoning
    Yinlong Xu, Jianfeng Huang, Zhengdong Zhou, Mingjie Lin +4

    Large Language Models (LLMs) have demonstrated remarkable success in complex reasoning through explicit Chain-of-Thought (CoT), but the associated computational overhead and token verbosity remain significant barriers to efficiency. While implicit reasoning in continuous latent spaces offers a promising alternative, existing methods like Coconut and CODI often rely on fixed-length latent sequences and suffer from representational collapse. In this paper, we propose Sep-Latent, an adaptive implicit reasoning framework that utilizes separator tokens as structural semantic anchors. By introducing an Attention-Decay mechanism, we force the model to aggregate and compress reasoning logic into these anchors during training. We then align explicit and implicit pathways through multi-level distillation across all layers to ensure logical consistency. Crucially, Sep-Latent enables dynamic reasoning depth by adaptively terminating the latent thinking phase based on query complexity. Experimental results on the GSM8k-Aug benchmark show that Sep-Latent achieves an accuracy of 56.8\% with an average of only 5.4 tokens, significantly outperforming prior fixed-length implicit models in efficiency while narrowing the performance gap with explicit CoT. Further analysis via PCA and symbolic variant testing confirms that our method ensures semantic grounding and exhibits superior robustness to perturbations.

    benchmark
  213. arxiv:2610.22922 · cs.CV
    ODSS-Net: Orthogonally Decoupled Shared-Specific Representations for Multi-Modal Object Re-Identification
    Zhifeng Ma, Yuhong Feng, Chongyi Huang, Mingyang Zhou +1

    Multi-modal object re-identification (ReID) leverages complementary information from heterogeneous sensors. Cross-modal alignment and fusion remain challenging. Existing methods mainly reduce modality gaps via feature alignment and fusion. However, alignment may entangle modality-shared structural information with modality-specific cues, while conventional fusion may not fully exploit complementary information. To address these issues, we propose ODSS-Net, a framework for orthogonally decoupled shared-specific representation learning and adaptive fusion of complementary modality-specific features. First, a Shared Structural Feature Construction (SSFC) module suppresses modality-dependent appearance variations and extracts modality-shared structural information. Then, an Orthogonal Specific Feature Learning (OSFL) module aggregates identity-relevant local information and removes the shared-aligned component via geometric orthogonal projection. Finally, a Sample Adaptive Mixture of Experts (SAMoE) module uses the shared structure as routing context to adaptively refine and fuse modality-specific features. Experiments on three benchmarks demonstrate the effectiveness of ODSS-Net for multi-modal object ReID.

    benchmark
  214. arxiv:2610.22915 · cs.AI
    Reward Engine: Reward-Centric Self-Evolution for Open-Ended Tool Agents
    Litong Feng, Ji Qi, Yiran Cheng, Shaowu Pan +2

    Open-ended tool agents can complete the same objective through different tool paths, and their quality depends jointly on the interaction trace, environment state, and final artifacts, so such tasks rarely admit an executable ground-truth reward. The central challenge is therefore to construct a reward from existing experience that exposes the current policy's failures while remaining stationary during optimization. We introduce Reward Engine, which, like a self-motivated student, first works out what success requires and then optimizes toward it. From the task and initial trajectories, it builds a RewardPack of testcases for required correctness and rubrics for graded quality, each executed as a check that returns a judgment with evidence. The frozen RewardPack then guides the agent to revise and select its outputs, and can also serve as a reward for reinforcement learning. On WorkBuddy Bench Office and Claw-Eval, the constructed reward largely agrees with the benchmark graders, not reversing their ordering on 82.3\% and 80.5\% of candidate pairs. Used to revise executions, it raises three-run consistency ($\mathrm{Pass}^{3}$) over Baseline from 45.0\% to 54.0\% and from 78.4\% to 83.7\%.

    agentbenchmark
  215. arxiv:2610.22912 · eess.SY
    Data Center Load Accommodation under QoS and Checkpoint-Induced Frequency Security Constraints
    Ruolan Cui, Ziang Liu, Chenxi Hu, Francis Yunhe Hou

    The rapid growth of artificial intelligence workloads is accelerating the expansion and grid interconnection of data centers (DCs). Existing data center load accommodation (DCLA) assessments primarily focus on grid and resource adequacy and often imply that greater workload flexibility always increases accommodable capacity. However, workload flexibility has a dual effect. Workload deferral and temporal shifting can alleviate network congestion, whereas synchronized checkpointing of training workloads can cause large and rapid load reductions, threatening frequency stability. Therefore, DCLA cannot be determined solely by grid adequacy. This paper develops an optimal DCLA method that jointly accounts for workload quality of service (QoS), checkpoint-induced frequency security (FS), and grid adequacy. First, a QoS-aware model characterizes workload-dependent flexibility while ensuring service feasibility. Second, checkpoint-induced FS constraints are developed for periodic and event-triggered power variations, with key frequency-security limits reformulated in algebraic form. Third, a two-stage lexicographic model is developed to maximize DCLA and minimizes the required energy storage (ES) deployment, and then optimizes system operation under the given DC and ES siting and sizing decisions. Case studies show that the proposed method avoids a 40.3% overestimation of DCLA compared with the grid-adequacy benchmark. QoS is the primary limiting factor, while checkpoint-induced frequency-security constraints become more restrictive as training workloads increase. These findings highlight that reliable DCLA should account for both its grid-support value and the associated QoS and FS risks.

    benchmark
  216. arxiv:2610.22907 · cs.CV
    COMET: Decoupled Distillation, Routing, and Capacity Control for Task-Agnostic Continual Vision--Language Learning
    Seungyoon Woo, Keonhee Park, Gunhee Kim

    Vision-language models (VLMs) such as CLIP retain zero-shot recognition ability, but their open-world use often requires task-agnostic continual learning: tasks arrive sequentially, past data are unavailable, task identity is unknown, and the backbone cannot be retrained. We argue that this setting fails VLMs through three coupled pressures beyond the stability-plasticity tradeoff: (1) modal-consistency drift, (2) representation-space interference, and (3) per-task capacity shortfall. We propose COMET (Collaborative Mixture-of-Expert Transfer) to resolve these three issues respectively as follows. A frozen teacher provides only a feature reference on a shared image-text pool; Tri-Affinity Distillation preserves image-image, text-text, and image-text geometry; and Student-Aware Cross-Expert Distillation aligns new experts with compatible prior experts so routing errors degrade gracefully. We design COMET to keep capacity and inference separate: a contextual bandit keeps or merges full per-task experts using post-training signals, while a reconstruction router selects sparse experts at test time without teacher semantics. On the X-TAIL benchmark, we show that each proposed component improves continual learning performance. COMET shows that continual vision-language adaptation can preserve multimodal geometry, coordinate accumulated experts, and allocate capacity jointly when feature reference, expert shaping, routing, and capacity control are separated by construction.

    post-trainingbenchmark
  217. arxiv:2610.22906 · cs.LG
    FederatedBytes: Tokenizer-Free, Byte-Level Small Language Models for Efficient Multimodal Federated Learning
    Ali Sarabadani, Massimo Piccardi, Mehran Abolhasan, Farzad Tofigh +2

    Federated Learning (FL) is a powerful machine learning framework that allows building a global model from the contributions of multiple, distributed clients, while ensuring the privacy of their local data. However, its heavy computational and communication overheads challenge its use for contemporary AI workloads such as natural language processing (NLP) and multimodal analysis. For this reason, this paper presents a novel small language model that leverages byte-level text and multimodal data processing, enhanced by a linear attention mechanism, to enable manageable Federated Learning of NLP and multimodal analysis tasks. The proposed model, nicknamed FederatedBytes, manages to process text and any other data types, including images, by treating its input as a mere sequence of bytes, while its lean linear attention mechanism allows abating the overall memory and computational footprint. To prove the effectiveness of the proposed model we have carried out an extensive performance evaluation on sentiment analysis tasks in multilingual and multimodal settings over the Multilingual Amazon Reviews and ScienceQA datasets. The experimental results show that FederatedBytes has been able to outperform comparable models such as ByT5 and LSTM, achieving higher accuracy while significantly reducing memory and communication overheads.

    memory
  218. arxiv:2610.22905 · cs.RO
    SenseSwarm: Swarm Exploration via Neural Structure Estimation with Shared World-Anticipation for Rapid Mapping
    Haohua Que, Haojia Gao, Mingkai Liu, Hoiian Au +6

    A robot swarm covers a building faster with more robots, but the map is only as complete as the space the robots have physically visited. We target the amount of an unknown building a decentralized swarm knows at each instant, ahead of physical coverage. SenseSwarm serves a lightweight learned map predictor on the swarm's shared occupancy map and folds the predicted structure back into a stigmergic guidance field. Its 709K-parameter predictor is fine-tuned on the swarm's own multi-front observations, which we show is necessary: a single-robot predictor collapses on such views. We build a benchmark of twelve real building floors from HouseExpo, HKUST, and KTH, up to 24,000 square meters, and run 1,620 simulations across nine coordination methods, swarm sizes 32 to 512, and three seeds. On physical coverage the swarm is near parity with a greedy frontier planner, as expected. On map reconstruction, every non-predictive baseline needs 1.3 to 1.9 times longer to reach an 80 percent known map, and SenseSwarm holds an equal-time area lead of about 670 square meters, up to 1,359. Folding the prediction into the shared field matters: the same predictor applied per robot matches our speed but suffers a 6.9 times higher inter-robot contact rate at 512 robots. The prediction is computed once per shared map, so its cost stayed flat from 32 to 512 robots.

    benchmark
  219. arxiv:2610.22900 · cs.RO
    Structured Perception Representations for Off-road Dynamics Learning
    Micah Nye, Samuel Triest, Matthew Sivaprakasam, Isaiah Adu +2

    Autonomously navigating off-road terrain at high speed demands accurate control, which often relies on dynamics models that capture complex interactions between the robot and terrain. In off-road regimes, terrain interaction is a major component of the vehicle's dynamics. Prior work attempts to capture this effect by training learned models of the vehicle's dynamics, conditioned on its perception of the local environment. However, said conditioning is often overly simplistic, ignoring the spatial structure of the dynamics problem. Our approach, SpatialBEV, exploits this spatial structure. At each step, we query the bird's-eye-view (BEV) features around the predicted state, conditioning the next prediction on current geometric and semantic cues. We additionally present a benchmarking approach that mines a dataset's most difficult scenarios. On datasets collected across seasons, SpatialBEV reduces prediction error by 17.1%, compared to physics-based and learned baselines. We further evaluate on hardware on high-speed ($\geq$10 m/s) and geometrically challenging courses with a 25.3% reduction in maximum cross-track error. More details can be found at https://theairlab.org/spatial-bev-dynamics .

    benchmark
  220. arxiv:2610.22899 · cs.MA
    Provable Authority: A Protocol for Verifiable Authority Resolution and Control Across Human, Organizational, and Agentic Transactions
    Jeremy Blackburn, Tim Hines, Matthew Koepp, Russell Lessard +1

    Payment systems authenticate users, devices, accounts, and cryptographic credentials, but a relying party, such as a counterparty, payment rail, marketplace, or autonomous agent runtime, generally cannot obtain compact, independently verifiable evidence that the party presenting one exact transaction holds current organizational authority to execute it under the applicable evidence, policy, budget, and approval state. Agentic commerce makes this gap operationally urgent: a software agent can produce a valid signature while acting outside its delegated purpose, after its sponsor's authority has changed, or on transaction details that differ from those a human approved. We describe the ChainIT Authority Protocol, a subject-neutral and rail-neutral pre-execution authority and evidence control plane. The protocol separates durable identity, a versioned organizational authority record (the Authority Resolution Pactvera), sourced evidence assertions, deterministic reason-coded policy predicates, and organization-configured dispositions; binds each decision and human approval to a canonical transaction digest; reserves capacity atomically so that limits hold under concurrency; and authorizes commit only through a single-use, sender-constrained Execution Authorization Credential consumed at the rail adapter, with append-only lifecycle evidence sealed after the terminal outcome. An Agent Subject Profile adds the workload registration, attested key hierarchy, and sovereignty boundary required when the acting subject is an AI agent, without introducing a separate authority concept. We specify the commit-time execution predicate and bounded-freshness status rules, give enforcement profiles for fiat (ACH, wire, card, instant payment) and stablecoin rails, and state residual risks and non-goals.

    agentai agentautonomous agentagentic
  221. arxiv:2610.22898 · cs.LG
    A Feature Selection Layer Approach for Multi-Class Classification on Neural Networks
    João Pedro Silveira e Silva, Márcio Dorn, Bruno Iochins Grisci

    In the age of data-driven decision-making, deep learning models have become powerful tools for uncovering patterns and making predictions that influence many aspects of society. However, their broader adoption in scientific and business applications is often limited by a lack of interpretability. Feature weighting methods, which assess the relevance of individual features, are among the most widely studied approaches for improving interpretability. While post hoc techniques have advanced significantly, especially in modeling nuanced relationships between features and labels in multiclass problems, embedded methods typically fall short in this regard. Embedded techniques integrate feature relevance learning directly into the model training process, but they are often unable to capture label-specific relationships. To address the limitation, we propose multiple extensions to the Feature Selection Layer, an embedded feature weighting mechanism for neural networks. Through extensive experimentation, we identify the most effective extension and benchmark it against established feature weighting methods. Our results show that the proposed method matches existing techniques in identifying globally relevant features and outperforms them in capturing label-specific relevance. By leveraging the internal dynamics of the trained neural network, it reveals how individual features contribute to the prediction of specific labels, requiring no post-processing beyond the training process.

    benchmark
  222. arxiv:2610.22895 · cs.CV
    HAPACT: A Benchmark For Human-Centric Physical Impact Localization in Movies
    Youngrae Kim, Yejin Jang, Eunho Kim, Youjin Sung +3

    Physical reasoning about humans in videos has largely focused on everyday interactions, where contact, pressure, pose, and motion support embodied perception and contact-aware applications. Still, short and forceful properties like impact events remain relatively underexplored, despite their relevance to broader understanding of how abrupt physical events affect the human body. To fill this gap, we introduce human-centric impact localization, a new task that localizes target physical impacts temporally within a video and spatially on the human body. We further present HAPACT, a HumAn-centric imPACT dataset and benchmark built from movie clips, containing over 2.8K video-query pairs with more than 20K fine-grained human-annotated frames. We built HAPACT around viewer-grounded annotation, text-conditioned localization, and supervision at frame-level temporal and SMPL-X-based spatial granularity. To demonstrate the effectiveness of our benchmark, we conduct extensive evaluations of existing methods, employing video moment retrieval models for temporal localization and human-scene contact models for spatial localization. For the spatial task, we propose a text-conditioned baseline by reformulating localization as query-guided body-region selection. Benchmark results show that physical impact localization remains challenging for existing methods, motivating HAPACT as a dedicated benchmark to support physical impact reasoning in videos.

    embodiedbenchmark
  223. arxiv:2610.22894 · cs.LG
    Competence-Based Evaluation of Logical Reasoning in Large Language Models
    Jize Jiang, Chengxiang Zhai

    Recent advances in large language models (LLMs) have brought impressive performance across diverse tasks; however, shortcomings remain in logical reasoning, including inconsistencies under linguistic variation. To improve the reasoning capacity of LLMs, we must be able to accurately measure their reasoning capacity. Current benchmarks primarily report single accuracy numbers on selected reasoning tasks. Such a performance-based evaluation methodology says little about a model's generalization capacity in reasoning. To address this limitation, we argue for a different evaluation paradigm, called competence-based evaluation (CBE), which tests whether models reliably apply the same competence across a family of problems. A parameterized generator scales complexity while keeping the required competence fixed, and matched answer-preserving presentations test correctness beyond one formulation. We instantiate CBE on elementary ordering tasks, scaling input size while varying presentations of the same ordering rule. Across six GPT and Qwen models, performance is reasonably good over bounded complexity ranges, but even strong models fail the competence condition across the tested family. Chain-of-thought and GRPO extend strong performance without resolving all tested failures; supervised LoRA shows little held-out transfer. CBE provides diagnostic profiles and an evaluation protocol that can grow with future models.

    benchmarkevaluation protocol
  224. arxiv:2610.22892 · cs.CL
    Bengali Hallucination Detection with Retrieval and Programs: A System Study
    Shamim Hasan, Samiya Ahmed Chowdhury

    Bengali hallucination detection must distinguish unsupported factual claims from incorrect reasoning, often without a supplied reference passage. We present and characterize an inference-only system that combines response-conditioned query rewriting, hybrid retrieval, and task-specific language-model judging. Six retrieval routes search question-answer and passage collections; a router directs each example to a 27B textual judge or a 14B agent that can compare retrieved answers, reason directly, or generate and execute Python. We characterize the system using version-specific code, saved predictions, execution traces, and corpus metadata from the Olikbochon challenge. Kaggle associates a displayed private macro-F1 of 0.956 with Version 2; its saved run processes 2,516 examples in 220.99 minutes on two T4 GPUs. Of 253 math-routed examples, 88 terminate through candidate matching, while execution traces expose 49 stderr-bearing run events across 35 examples. A later run changes six predictions despite identical core worker code. A bounded audit of 1,090 benchmark-related corpus records finds exact question or context overlap for 38 processed evaluation inputs, precluding an unqualified claim of corpus-independent generalization. The contribution is a concrete retrieval-and-computation judging architecture with an auditable operational characterization, not a controlled estimate of component gains. We distinguish the reported leaderboard result from independently recomputed diagnostics and identify the evidence needed for stronger effectiveness and reproducibility claims.

    agentbenchmarkleaderboard
  225. arxiv:2610.22889 · cs.CL
    NLData2Opt: Benchmarking Parameter Derivation and Prediction for Modeling Optimization Problems from Natural Language and Data
    Lijia Liu, Kyohei Atarashi, Jiyi Li, Koh Takeuchi +2

    Formulating optimization problems from raw inputs of problem descriptions and data is time-consuming and costly. While Large Language Models (LLMs) offer a promising way to automate this process, existing benchmarks overestimate their capabilities by relying on idealized inputs where all data is relevant and complete. This diverges from reality, where data-centric reasoning is required to extract, compute, and predict parameters from noisy sources. To address this, we introduce NLData2Opt, a benchmark to evaluate LLMs on the multi-stage reasoning inherent in practical optimization workflows. NLData2Opt decomposes problems into problem descriptions and data, and requires models to (1) extract relevant parameters from data containing misleading attributes, (2) compute missing parameters derived by transforming existing data attributes, and (3) predict unavailable future values from historical observations. Evaluation results show substantial performance decline relative to existing benchmarks, indicating that benchmark progress does not directly translate to formulation accuracy in practical optimization workflows. NLData2Opt is available at https://github.com/sumarivyn/NLData2Opt.

    benchmark
  226. arxiv:2610.22882 · cs.AI
    Cooperation Without Adaptation: LLM Agents in a Collective-Risk Dilemma
    Trung-Kiet Huynh, Dao-Sy Duy-Minh, Chi-Nguyen Tran, Phu-Hoa Pham +3

    Language-model agents are increasingly deployed to act on behalf of people in settings where individual decisions shape shared outcomes. We examine whether five production models adapt their contributions to changing risks and partner behaviour, or instead follow a relatively fixed strategy. In a collective-risk dilemma game (CRG), six players contribute to a shared pool over ten rounds; if the pool falls short of its target, a catastrophe may destroy their remaining resources. Four models contributed approximately an equal share at every positive risk level, and four of the five also contributed when risk was zero, even though contributing could not improve their payoff. Explicitly stating that only the agent's own payoff mattered reduced zero-risk contributions for most models, while a more neutral prompt did not. Several models correctly identified which action offered the higher expected payoff at different risk levels, yet their contributions changed little. Against fixed partners, no model consistently chose the payoff-maximising response, and in mixed-model groups, a single agent's shortfall in the final round could cause an otherwise successful group to fail. Payoff-based selection also depended on population composition: in a single six-agent group, it favoured the model that contributed least and reduced group success, whereas in larger, randomly mixed populations, it favoured an exact fair-share model at moderate and high risk. These findings show that cooperative self-play alone is not evidence of strategic adaptability. Consistent with a machine-behaviour approach, evaluations of agents in multi-agent systems should test how they respond to incentives, partner actions, and population composition.

    llm agentmulti-agentagent systemself-play
  227. arxiv:2610.22874 · cs.CL
    Identify, Classify, Act: Per-Instruction Controlled Compliance for LLMs over External Data
    Zesheng Shi, Xue Li, Chenrui Dai, Jing Li

    LLMs over external data face a per-instruction tension: a single document or tool return often interleaves legitimate directives with adversarial prompt injections, and the model must follow the former while rejecting the latter. Existing defenses suppress or trust all external instructions, collapsing the safety-utility frontier and forcing a choice between over-refusal and over-compliance. We argue the missing capability is \textbf{per-instruction credit attribution} and formalize it as External-Data Instruction Classification, an identify-classify-act decision over each embedded directive. We propose \textbf{EDIC}, a decomposed reward framework for per-instruction compliance. A scene-grounded pipeline injects mixed beneficial and harmful instructions with counterfactual filtering, while a generative judge is initialized by SFT and refined with GRPO to attribute compliance at the instruction level. The judge then produces a reward that separately encourages beneficial compliance and penalizes harmful compliance for RL training. Across two backbones and five benchmarks, EDIC improves both safety and utility over the strongest prior defense, lifting harmful-rejection by up to 7.7\% and Arena-Hard-Auto utility by 12.5\% on Qwen3-4B. The code repository is hosted at https://github.com/yingyingxia666/EDIC.

    benchmark
  228. arxiv:2610.22869 · cs.AI
    MS-Eval: Benchmarking Sub-Agent Delegation and Orchestration in MCP-Intensive Working Scenarios
    Bowen Ding, Qixin Wang, Lexing Ruan, Weiyi Lu +3

    LLM agents are increasingly integrated into real-world works, yet they must handle diverse domain-specific tools and complex multi-sub-task problems that sequential tool invocation resolves inefficiently and with severe context bloat. The Main-Agent Delegation and Sub-Agent Execution (MAD-SAE) paradigm addresses these issues by delegating sub-tasks to specialized sub-agents for parallel execution in isolated contexts, but its delegation and orchestration quality remains unquantified. We introduce MS-Eval, a unified evaluation kit comprising an MCP-intensive benchmark, a controlled execution harness, tailored delegation and orchestration metrics, and the MSQ index for comprehensive assessment. Evaluating four MAD-SAE variants and eight frontier models reveals that (1) sub-agent specialization (Spec) is the best practical practice; (2) open-source models are largely competitive with few exceptions; and (3) MAD-SAE generally raises cost with contingent payoff, benefiting most open-source models in both performance and efficiency while leaving closed-source models with higher cost and degraded performance. Our case study further traces these gains to better single-agent solutions, effective context offloading, and genuine sub-task parallelism, whereas unnecessary or redundant delegation turns the same cost into pure overhead. Codes: https://github.com/LINs-lab/MS-Eval

    llm agentbenchmark
  229. arxiv:2610.22861 · cs.AI
    Evo-B&B: Behavior-Aware LLM-Guided Joint Evolution of Branching and Node Selection Policies for MILP
    Jiayu Chen, Chentong Chen, Yuhan Liu, Ye Fan +2

    In branch-and-bound for Mixed-Integer Linear Programming (MILP), branching and node selection are two critical and tightly coupled decisions, yet most existing learning-based approaches focus on optimizing them separately. We propose Evo-B&B, an LLM-guided evolutionary framework that jointly evolves executable branching and node selection policies using end-to-end SCIP solving feedback. To improve search efficiency and reduce redundant evaluations of behaviorally similar policies, Evo-B&B introduces a behavior-aware multi-island evolutionary mechanism. It measures policy similarity from branching and node-selection behaviors on a fixed set of B&B decision states and uses this similarity to preserve population diversity. We evaluate Evo-B&B on four standard MILP benchmarks across multiple instance scales. Experimental results show that Evo-B&B achieves competitive solving performance and generalizes well to larger-scale instances, while ablation studies further demonstrate the effectiveness of joint evolution and behavior-aware diversity management.

    benchmark
  230. arxiv:2610.22858 · cs.AI
    Uncertainty-Aware Budget Allocation for Efficient LLM Tree Search
    Jianguo Huang, Zhiyi Lyu, Fuxiang Zhang, Yanchen Deng +2

    Tree search enhances multi-step reasoning in large language models (LLMs), yet prevailing methods expand every step with a fixed budget, irrespective of the marginal utility of expansion. We first show, across existing search algorithms, that continuations sampled from a shared intermediate state are largely duplicates and collapse to a few distinct answers. Critically, we find that the semantic dispersion of sibling actions predicts the redundancy of their downstream continuations, permitting a subtree's budget to be fixed before expansion. An information-theoretic analysis motivates this: treating the per-step budget as the Monte Carlo sample size for estimating a step's residual answer entropy, we derive a sufficient budget that shrinks with sibling dispersion. We accordingly propose $\textbf{Uncertainty-Aware Allocation}$ (UAA), a training-free framework that decouples value from uncertainty: values determine which candidates are promising, while uncertainty, estimated from sibling dispersion at no additional model cost, determines how many merit expansion, via uncertainty-aware pruning and uncertainty-aware budgeting. Across mathematical reasoning, logical reasoning, and planning benchmarks, UAA reduces token usage by up to $90\%$ and latency by up to $9\times$ at comparable accuracy, generalizing across four tree search algorithms and both open- and closed-source LLMs.

    benchmark
  231. arxiv:2610.22852 · cs.RO
    Sample-Efficient Reinforcement Learning for Policy Improvement via Flow-Reversed Action Anchors
    Carl Qi, Weijun Dong, Xiaojie Wang, Peiyi Chen +7

    Generalist robot policies offer strong behavioral priors, but achieving deployment-level reliability and throughput remains difficult. Reinforcement learning (RL) can improve these policies, yet its effectiveness is constrained by the cost of online interaction. Existing methods adapt frozen flow-matching policies by treating noise space as a continuous action space. We show that this behaviorally unstructured space makes exploration inefficient, often preventing these methods from extracting meaningful learning signals on challenging tasks. We introduce AnchorQ, which enables coarse-to-fine policy improvement through a structured, refinable latent action representation. At each decision step, AnchorQ uses flow reversal to map reference behaviors to state-conditioned latent anchors, samples local particles around those anchors, and applies value-guided sampling to explore promising refinements. Our analysis shows that anchors reduce identification complexity at the cost of support bias; particle refinement trades this bias against action-class complexity; and critic accuracy determines the reliability of KL-regularized, value-guided selection. Across simulated and real-world manipulation tasks, AnchorQ outperforms state-of-the-art post-training baselines in sample efficiency, delivering a 37% improvement in real-world success rate and 97% higher simulated task throughput under the same interaction budget. Additional visualizations are available on our website https://anchorq-rl.com/.

    manipulationpost-training
  232. arxiv:2610.22846 · cs.LG
    Benchmarking and Enhancing Agentic Causal Discovery
    Pingchuan Ma, Rui Ding, Shuai Wang

    LLM agents can now run causal discovery algorithms on a dataset and report what comes out. Whether they do more than that, such as noticing where an algorithm went wrong and fixing it with further tests, is hard to tell from current benchmarks. Most of them give the variables meaningful names, and an LLM shown only the names of the Tübingen cause-effect pairs already gets 99% of the directions right. With the names removed, the data are usually clean enough that picking a better algorithm than PC closes most of the gap. We introduce Heuresis, a benchmark of anonymous simulated datasets whose mechanisms range from linear Gaussian to heavy-tailed. For each family of datasets it separates what a better existing algorithm can gain from what only a closer look at the data can gain, and it traces every edge an agent changes. Every family leaves a gap that none of the algorithms, the router and the two causal-discovery foundation models we ran can close, so a score above the best chosen algorithm has to come from using the data. We instantiate the benchmark with an agent harness and three LLMs, GPT-6.1 Sol, Claude Opus 5.5 and Qwen 3.8 Flash. GPT-6.1 Sol surpasses the best chosen algorithm on all six families, by 14% adjacency F1 on the most difficult one, and Claude Opus 5.5 does so by 7.4% there. A third of the true edges the two agents restore on the two hardest families appear in no algorithm's output. The code and data will be publicly released to facilitate future work.

    agentllm agentagenticbenchmark
  233. arxiv:2610.22842 · cs.CV
    VTrajEmbed: Aligning Text with Video Object Trajectories via Spatiotemporal Embeddings
    Ruizhi Chen, Zining Fan, Yudi Guo, Jiang Liu +4

    Fine-grained alignment between text and video object trajectories is a core challenge in multimodal understanding. Existing embedding models excel at global retrieval, yet their single global representation entangles multiple objects and their spatiotemporal dynamics, limiting direct object-level grounding. To address this issue, we propose VTrajEmbed, an embedding model that decomposes a video into object trajectories and aligns each trajectory embedding with text for object-level grounding through retrieval. Our main contributions: (1)Slow-Fast Trajectory Encoding, a dual-branch architecture that establishes trajectory representations by densely sampling low-resolution frames for high-frequency motion cues and sparsely selecting high-resolution frames for scene context, aggregated by a shared LLM backbone; (2)Mask-Guided Contrastive Learning, a training objective built upon these representations that couples an intra-group loss and a batch-level bidirectional loss under a relation mask, excluding unverified relations and preventing confirmed matches from becoming false negatives; (3)Text2Traj-Bench, a trajectory-level text-video benchmark with model-assisted construction and human-verified query--trajectory relations. Experimental results show that VTrajEmbed-5B improves text-to-trajectory and trajectory-to-text R@1 over Qwen3-VL-Embedding-8B by 15.48 and 18.05 percentage points, respectively, and achieves 49.48% average R@1 across five global retrieval benchmarks, compared with 42.86% for Qwen3-VL-Embedding-8B.

    benchmark
  234. arxiv:2610.22832 · cs.CL
    Too Emotional to Be Fake? Intra-Utterance Emotion Shifts Expose Failure Modes in Speech Deepfake Detectors
    Jingyi Chen, Surya Koppisetti

    Expressive speech synthesis can now change emotion within a single utterance, yet its impact on deepfake detection remains unclear. We introduce EmoShift, a benchmark comparing emotion-stable (AA) and emotion-changing (AB) speech to examine these effects while accounting for static emotion identity and acoustic confounds. It comprises 7,240 bona fide utterances from four source datasets and 9,720 synthetic utterances generated using four expressive TTS systems. Across seven detectors, we find that emotion changes can make synthetic speech appear more genuine, increasing spoof miss rates by up to 25.3 percentage points on OMG-Emotion at fixed thresholds. Conversely, emotion-changing bona fide speech increases false-alarm rates by up to 24.4 percentage points on IEMOCAP. Further analyses reveal detector-specific sensitivities to emotion trajectories, not fully explained by static emotion identity or splicing artifacts. These findings underscore the importance of emotion dynamics in deepfake detection.

    benchmark
  235. arxiv:2610.22827 · cs.RO
    DefFlow: Goal-Conditioned Deformable Object Manipulation using 3D Point Flows
    Gaurav Singh, Sergio Orozco, Ryan Duong, Mete Tuluhan Akbulut +2

    Manipulating deformable objects toward desired 3D configurations is a fundamental capability for robots operating in unstructured physical environments. However, as the range of possible initial and target configurations grows, goal-conditioned deformable manipulation requires a robot to infer increasingly diverse manipulation behaviors in addition to precise continuous actions. We introduce DefFlow, a framework for goal-conditioned manipulation of deformable objects toward a broad range of 3D configurations. DefFlow learns from demonstrations by jointly predicting the goal-conditioned 3D point flow of the robot and deformable object. We show that this formulation better captures both the manipulation decisions, such as when and where to grasp, release, or re-grasp, and the precise continuous motions required for goal reaching, rather than predicting only the robot flow. Across rope, granular-media, and cloth manipulation settings with diverse initial and target configurations, DefFlow substantially outperforms goal-conditioned action-prediction and robot-flow-only baselines on success rates and goal-distance metrics. We further demonstrate deployment of our framework on a real robot using real point-cloud observations and unseen goal configurations.

    manipulationgrasp
  236. arxiv:2610.22825 · cs.AI
    World Action Critic (WAC) Model: Joint Foresight and Criticism for Policy Learning
    Xiangteng Zhang, Haoyuan Liu, Yang Guan, Jingliang Duan +6

    World models have shown considerable promise in sequential decision-making by anticipating the consequences of actions and providing predictive foresight for policy learning. However, learning to reason about counterfactual actions remains challenging, since predictive objectives may underemphasize information critical for action evaluation, while unexecuted actions lack observed futures for direct supervision. To address these challenges, we propose the World Action Critic (WAC) Model, which proposes multimodal action candidates and jointly learns candidate-conditioned future prediction and consequence evaluation. In our proposed framework, critic supervision directly shapes intermediate predictive representations during rollout, encouraging them to retain information needed to distinguish action quality. Observed futures supervise prediction for executed actions, while external task-quality targets provide evaluation signals for counterfactual candidates. Through joint optimization, critic signals from these candidates propagate into predictive representations without requiring their observed futures. Our theoretical analysis characterizes critic-guided feature retention and bounds candidate-selection regret. Applied to autonomous driving, WAC achieves 93.4 PDMS on NAVSIM-v1, demonstrating the effectiveness of jointly learning predictive foresight and consequence evaluation.

    world model
  237. arxiv:2610.22822 · cs.CV
    VISTA: A Multimodal Egocentric Benchmark for Goal-Oriented Perception and Assistance to Blind and Low-Vision Users
    Noé Mosca, Mohamed Rayan Barhdadi, Muhammad Hamza, Joshua Yeh +10

    Blindness and low vision limit many people's ability to complete daily tasks, motivating AI assistants that provide useful guidance. Yet, existing benchmarks do not adequately test whether vision models can provide effective everyday assistance. In particular, i) existing benchmarks primarily rely on visual and textual inputs, with limited support for other modalities that are increasingly available in egocentric wearable devices such as gaze, inertial measurements, spatial trajectories, and others; and ii) existing evaluations emphasize scene understanding and question answering, with limited coverage of goal-conditioned guidance for blind and low-vision users across everyday tasks. To bridge this gap, we introduce VISTA (Visually Impaired Scene and Task Annotations), a multimodal egocentric benchmark for goal-oriented assistance for blind and low-vision users. VISTA contains $\sim$1,000 egocentric recordings captured with Project Aria glasses; each recording provides synchronized sensor streams, six of which serve as model inputs, and three different types of annotation, covering ten task categories (TC), from object localization to hazard detection. We evaluate 14 representative vision-language models zero-shot on VISTA and design a lightweight multimodal model baseline that integrates all six modalities. Experimental results show that, although existing models can often describe egocentric scenes, they remain limited in generating goal-directed assistive guidance. VISTA thus provides a benchmark for evaluating multimodal models that could improve everyday accessibility.

    benchmark
  238. arxiv:2610.22820 · cs.LG
    Causal Edit Serialization
    Louis Robinson

    Standard autoregressive language models generate text sequentially, but revision requires referring back to tokens in the input context or those already generated. Causal Edit Serialization (CES) is introduced as a framework for expressing such revisions within the same causal token stream. Special pointer tokens are paired with predicted source addresses, allowing insertion, deletion, replacement, copying and movement to be localized explicitly. A lightweight localization head and dual-index rotary position encoding support both address prediction and conditioning on previously emitted pointers, while retaining a standard causal attention mask. A pretraining procedure combines ordinary language modeling with serialized repairs derived from random corruptions, model-generated edits and repeated text. Across model scales and editing baselines, CES models demonstrate complete applied edits with modestly higher clean-text bits per byte than clean autoregressive models. Supervised and reinforcement-learning post-training provide a grammatical-error-correction proof of concept, producing more beneficial than harmful changes under reference-blind model judging. Correct repair programs also improve subsequent text prediction while the original context and its cached states are retained. CES provides a common interface for generation and localized revision, with potential extensions to iterative refinement and additional pointer-addressed transformations.

    post-trainingiterative refinement
  239. arxiv:2610.22813 · cs.RO
    Simultaneous Force and Motion Control in Generalised Coordinates
    Jon Woolfrey

    Robot manipulators often come into contact with rigid objects when performing work. This imposes physical constraints on their motion. Simultaneous force & motion control was developed to regulate contact forces commensurate with these constraints, whilst projecting motion control in to the orthogonal, unconstrained directions. This is typically formulated in the task coordinates, i.e. operational space, then mapped to joint torques to command the robot. However, most constraints on the robot control, such as joint limits, collision avoidance, singularity avoidance, etc. are naturally expressed in generalised coordinates (joint space). Thus, formulating the control problem in task space makes it difficult to guarantee control feasibility. In this paper Gauss' principle of least constraint is used to solve the control as a quadratic programming (QP) problem in generalised coordinates. It is shown that the theoretical solution to this problem produces the correct constrained dynamics for accurate contact force control. Moreover, inequality constraints may be appended to ensure feasibility and safety. Simulation results were conducted in MuJoCo and compared to the operational space formulation. The proposed method is able to simultaneously maintain a desired contact force, minimise trajectory tracking error, and adhere to practical inequality constraints.

    manipulator
  240. arxiv:2610.22812 · cs.RO
    PASS-VLN: Passability-Aware Local-Command Vision-and-Language Navigation
    Cong Zou, Xiangsheng Shi, Xuanhui Wang, Fubin Wang +3

    Reliable vision-and-language navigation (VLN) requires not only understanding instructions but also determining whether and how a robot can physically execute them. However, existing VLN benchmarks typically assume executable instructions and focus on long-horizon tasks. This leaves body-dependent local execution, including passability and collision-free execution, largely unexamined. Evaluating these factors within long-horizon tasks would entangle local physical execution with task planning and long-term spatial memory, making navigation failures difficult to attribute. We therefore focus on grounding local navigation commands in executable motion under environmental and robot body constraints. We present PASS-VLN, a passability-aware local-command VLN benchmark and model. The benchmark holds the scene, instruction, start pose, and reference route fixed while varying robot body width, requiring target completion when the instructed passage is feasible and stopping before blockage otherwise. Our model encodes body width through obstacle inflation in a depth-derived local bird's-eye view, fuses geometry with visual and linguistic context, and predicts discrete latent action codes for continuous closed-loop execution. Experiments show that PASS-VLN improves collision-free success by approximately 12 percentage points over a strong VLN model fine-tuned on our benchmark, while adapting reach-versus-stop behavior to body constraints and generalizing to unseen body widths. When integrated with a high-level planner, PASS-VLN achieves higher success with fewer local commands, while real-world deployment further demonstrates its effectiveness as a low-level executor. The benchmark and model will be released.

    benchmark
  241. arxiv:2610.22807 · cs.CL
    On the Position Bias of Large Language Models: A Comprehensive Study
    Yining Li, Huayang Li, Lemao Liu, Taro Watanabe

    Recent studies have shown that position bias remains a critical vulnerability in large language models (LLMs), undermining both the integrity of multiple-choice evaluations and the reliability of retrieval-augmented generation. However, existing understanding of this phenomenon is largely derived from studies with limited model coverage and English-centric benchmarks, leaving the behavior of frontier models and diverse languages underexplored. In this work, we conduct a comprehensive investigation of position bias across model scales ranging from billion-level to trillion-level parameters, covering both multiple-choice and multi-document tasks in five languages. Our experiments reveal that while position bias remains noticeable in small-scale models and Non-English languages, increasing model scale and employing standard Chain-of-Thought prompting significantly mitigate this issue. Moreover, we demonstrate that emerging \emph{thinking models} with extended reasoning capabilities are nearly immune to position bias, suggesting a costly but general approach toward resolving this vulnerability.

    retrieval-augmentedbenchmark
  242. arxiv:2610.22797 · cs.LG
    Reasoning Reranking with Semantic IDs as the Sole Item Representation
    Zhenwei Tang, Jintao Liu, Di Bai, Xidong Wu +1

    Reasoning rerankers write a rationale before ranking a candidate slate, and existing systems read item metadata to do so, even though generative recommenders increasingly represent items as compact Semantic IDs (SIDs). Ranking from SIDs alone has two appeals: SID prompts are far shorter than the matched item text, and although item codes are fixed from content, the backbone learns their token embeddings in part from user-item interaction sequences, so the tokens can carry behavioral signal that item text lacks. That training targets next-item generation, and whether it supports ranking a supplied slate from SIDs alone is untested. We build ReRankID, a controlled testbed in which models reason and rank from a history and slate shown only as SIDs, varying item rendering, history length, and slate size one at a time. On this testbed, such models often copy the input order instead of reranking, and readable metadata does not repair the failure. We compare supervised fine-tuning, rejection-sampled self-training, on-policy distillation, and verifier-based reinforcement learning as post-training methods and find that they elicit the missing behavior. The strongest post-trained model, the 8B model trained with a graded ranking reward, gives the best SID-only ranking and matches strong text-reading LLMs whose prompts are about 20x longer. It ranks above random with no history at all, where a caption-reading reference does not, and it keeps its gains at long histories, where text prompts pass 50k tokens and the caption-reading reference stops improving. In a paired rationale audit, this model argues trade-offs better than that reference but describes items less accurately. The common post-training methods we test thus elicit ranking from SIDs but not faithful item descriptions, which calls for post-training that improves the reasoning itself, not only the ranking it produces.

    post-training
  243. arxiv:2610.22795 · cs.CV
    STRIDE: Evaluating Spatiotemporal Reasoning in Driving Edge Cases
    Lieqi Liu, Rui gao, Jia-Chen Gu, Wenbo Hu +5

    In autonomous driving, perceiving surrounding objects is not enough: failing to understand how they move and interact over time can lead to dangerous decisions. Evaluating vision-language models (VLMs) therefore requires testing whether they connect past spatial observations to future consequences. Yet existing benchmarks often assess edge cases through basic perception questions or test temporal understanding in simple scenarios, leaving complex spatiotemporal reasoning underexplored. We introduce STRIDE (SpatioTemporal Reasoning In Driving Edge Cases), an object-centric benchmark for long-tail, safety-critical driving scenarios. STRIDE covers six task families: spatial perception, spatial understanding, temporal memory, temporal extrapolation, trajectory prediction, and scene-context awareness. It pairs multiple-choice questions probing individual spatiotemporal judgments with related free-form visual question answering (VQA) requiring detailed scene explanations and justifications of driving decisions. We prioritize scenes with unusual object behaviors and complex traffic interactions, generate reference answers from structured dataset annotations through automated procedures, and use limited language-model assistance and human verification to ensure annotation quality. Across 12 proprietary, open-weight, and driving-specialized models, even the best performer, GPT-6-Astra, achieves only 43.4% average multiple-choice accuracy, versus an approximately 20% random-choice baseline. Further analysis reveals that strong performance in describing driving scenes does not necessarily translate to accurate safety-critical judgments, and that models struggle more with reconstructing object motion than recalling past positions. STRIDE thus exposes spatiotemporal driving limitations that aggregate VQA scores can conceal.

    benchmark
  244. arxiv:2610.22791 · cs.AI
    Self-Evolving Coding Agents via Harness-Policy Optimization
    Zhepei Wei, Wenlin Yao, Baolin Peng, Danqing Wang +4

    Large language model (LLM) agents have shown promising potential in coding tasks, and their performance is jointly determined by the policy model and the harness that mediates their interaction with the environment. Indeed, harness and policy are mutually reinforcing and evolving only one component is suboptimal---a harness tuned for one checkpoint may not target the failure modes of an updated policy, while policy learning remains limited to the trajectories elicited by a fixed harness. In light of this, we propose Self-HPO, a recursive self-improving framework for coding agents that alternates harness and policy optimization, using the evolving policy as its own harness proposer. Specifically, in each round, the current policy first analyzes its own failures and proposes new harness candidates (H-step). The policy is then updated through reinforcement learning (RL) under the selected harness (P-step), and hard tasks identified during training form the next harness search set, keeping harness optimization targeted to the policy's evolving failure modes. Across three coding benchmarks, Self-HPO consistently outperforms harness-only and policy-only baselines, improving Qwen3.5-35B-A3B's average resolve rate from 57.8% to 64.8%, with gains generalizing to out-of-distribution settings. Further analysis shows that both optimized components retain gains even when paired with their base counterparts, while their joint gain exceeds the sum of their individual gains. Self-HPO also improves reliability (Pass@1) and expands the capability frontier (Pass@k) over the base agent. Additionally, we show that with a stronger external model (e.g., Opus 4.8) as the harness proposer, the same framework yields a further performance boost.

    self-improvingself-evolvingbenchmark
  245. arxiv:2610.22787 · eess.SY
    Filter-Augmented Reinforcement Learning for Battery Fast Charging with State of Health and Plating-Aware Reward Design
    Meng Yuan, Yang Li, Changfu Zou

    Fast charging of lithium-ion batteries is crucial for electric vehicle adoption, yet it increases safety risks and accelerates battery ageing. To address this, we propose a filter-augmented reinforcement learning (RL) framework for lifelong health-aware fast charging, based on the twin-delayed deep deterministic policy gradient (TD3) algorithm. The RL agent is guided by a reward function that balances charging speed and battery longevity through the incorporation of state of health (SoH) information and an SoH-dependent voltage penalty term derived from a constant--current constant--overpotential charging strategy. Since onboard SoH estimates are inevitably corrupted by noise, two filtering strategies are integrated: an exponential moving average (EMA) filter when noise statistics are unknown, and a Kalman filter when they are available, with the filtered estimate embedded in both the agent's state and reward computation. The RL policy is trained using data from high-fidelity life-cycle simulations. The evaluation uses a published electrochemical and degradation parameterisation for a commercial LG M50 cell. Results show that the proposed method achieves up to 21.4\% more equivalent full cycles (EFCs) than conventional constant current--constant voltage (CC--CV) charging under 2\% SoH estimation noise, and maintains this lifetime advantage even under 5\% noise, while sustaining comparable charging speed.

    agent
  246. arxiv:2610.22779 · cs.AI
    When Can You Trust an Agent's Own Verifier? Stress-Testing Counterfactual Rationale Verification Under Search Pressure and Oracle Access
    Jonghyun Lee, Moonsu Kim, Sunghwan Park

    Automated agent design increasingly trusts a single score to decide whether a change is an improvement -- but a score optimized aggressively enough ceases to measure what it was meant to. When an agent also explains its design, that explanation is a second signal we can check without human labels: negate each stated claim, re-run, and observe whether performance drops. We ask whether such a verifier stays reliable once an agent optimizes against it. Our contribution is not a new verifier but a way to stress-test one: a pressure ratio $ρ$ measuring how much the verifier steers the agent's search, and four falsifiable probes covering four failure modes -- giving no signal, being too weak to matter, being gamed by stating fewer claims, and being queried before the agent commits. We apply the test to our own verifier, the Rationale Consistency Score (RCS), inside an LLM agent that designs machine-learning pipelines, with one generator model, six tasks, and three seeds. First, in our prior configuration the verifier barely affects the search objective ($ρ\ll 1$), partly because of an accuracy gate, so its robustness had not been tested; it still shapes the reported artifact through a final-selection tie-break that $ρ$ does not register. Second, a capability control shows the agent shortens its claim lists when instructed, yet search does not, even at $ρ$ over 100 times the prior level; since feedback omits claim volume and horizons are short, this null result is largely structural rather than evidence of robustness. Third, when the agent can probe the verifier before committing, it keeps claims that pass while held-out faithfulness falls -- a decline invisible to in-loop metrics, though partly confounded by how RCS scores small pipelines. We offer these as preliminary evidence and propose a reporting standard: audit a verifier with $ρ$ and these probes before trusting it.

    agentllm agent
  247. arxiv:2610.22778 · cs.CV
    Direct Conditional Parameterization for N-Dimensional Splatting
    Zhongpai Gao, Benjamin Planche, Van Nguyen Nguyen, Meng Zheng +4

    N-dimensional splatting extends 3D Gaussian Splatting with view direction and time to model view-dependent and dynamic effects. Each primitive is lifted into a joint distribution over position and conditioning variables, then conditionally sliced at each rendering query. N-DGS and UBS share this step across Gaussian and Beta kernels: recovering the conditional mean, opacity, and covariance requires per-primitive matrix inversions and regression products in every forward pass. We introduce direct conditional parameterization, replacing this covariance-derived readout with explicit per-effect parameters: a Cholesky precision factor for opacity and a factorized query-to-position regression operator with learnable per-dimension coupling. Unrestricted conditional coordinates recover a valid joint covariance; our factorization adds spatial-scale and coupling priors. Applied to both kernels, the parameterization yields dGS (Gaussian) and dBS (Beta). Across five static and dynamic benchmarks, they match or exceed their covariance-derived counterparts on quality, with gains up to +1.26 dB in the main MCMC evaluation. Production slicing is 6.9-7.7 times faster than on-the-fly covariance-derived conditioning; complete-render profiling gives 1.25-1.91 times speedups on four trained scenes. Direct conditioning provides a structured, kernel-agnostic alternative for N-dimensional splatting. Project page: https://gaozhongpai.github.io/dgs/.

    benchmark
  248. arxiv:2610.22776 · physics.optics
    Development of a fabrication-to-benchtop process for SiN-based quantum devices
    Connor Kupchak, Abubaker Tareki, Tara Moradi, Patrick Laferriere +2

    Silicon nitride has emerged as a key material for the advancement of future quantum communication and computational devices. However, challenges persist in establishing a comprehensive development process, spanning from fabrication to optical characterization, essential for advancing these devices to their next stage of potential. Maintaining control over the entire process, from start to finish, alongside device modeling, enables more efficient and timely iteration and optimization. Here we introduce a fabrication-to-benchtop process. We employ an in-house stoichiometric recipe for device fabrication and conduct comprehensive characterization of optical losses and material properties. Understanding and controlling the refractive index of the device are vital for predicting device geometries and group indices. Implementing the refractive indices of our fabricated devices, we simulate microring resonators and demonstrate their application in generating optical solitons.

    microring
  249. arxiv:2610.22770 · cs.LG
    Unit-level aggregation breaks coverage in causal foundation models
    Gabe Schulman, Lucius E. J. Bynum, Yoav Wald

    Causal foundation models (CFMs) show promising results for treatment-effect estimation, with unit-level estimates that are competitive on standard benchmarks. Many causal analyses, however, inform decisions through effects averaged over groups of units, such as the full sample, the treated units, or a subgroup, and for such decisions a CFM is useful only if its credible intervals are calibrated. We examine these intervals and find that their coverage degrades, both in simple synthetic settings and on standard benchmarks. On ACIC 2016, for example, temperature-calibrated 95% intervals from CausalPFN fall below nominal coverage beyond groups of 30 out of 4802 units, and intervals for the full-sample average contain the true value in only 20% of datasets. We trace this problem to how these models are trained and how their intervals are formed. They output a predictive distribution for one unit at a time and, at inference time, combine the unit-level variances as if the errors on different units were independent. We show formally that posterior uncertainty over the data-generating process makes these errors correlated. As a result, the combined variance is too small by a factor that grows with the size of the group. Empirically, this factor predicts where coverage fails and by how much. We then show that two neural process models, which make joint predictions for a group of units, account for this correlation and substantially improve coverage on synthetic tasks and, as heads on a frozen CausalPFN, on all six benchmarks.

    benchmark
  250. arxiv:2610.22765 · cs.RO
    Pedicel-Targeted Strawberry Harvesting via Imitation Learning on a Low-Cost Teleoperated Robotic Arm
    Nitin Rai, Won Suk Lee, Hongyoung Leon, Heping Zhu

    Precise pedicel-targeted grasping in strawberry harvesting remains a critical bottleneck due to the pedicel's small diameter (1.3-3 mm), color similarity with surrounding foliage, and precise force control required for clean abscission zone detachment. As a market-grade crop, strawberry aesthetic value makes bruise-free pedicel detachment essential over direct fruit grasping. Classical approaches struggle to unify vision models, motion planning, and contact manipulation into a robust framework, particularly in occluded canopy environments. Therefore, this work presents a visuomotor imitation learning pipeline for pedicel-targeted strawberry harvesting that unifies perception, motion planning, and contact manipulation from collecting human teleoperation demonstrations, training an Action Chunking with Transformers (ACT) policy with a ResNet-50 visual backbone, and deploying the learned policy autonomously on a Jetson AGX Orin edge platform at 15 Hz. The trained policy achieved consistent loss convergence from 2.84 to 0.091 over 100K gradient steps on 130 demonstrations. We have also established a reproducible open-source pipeline with publicly released dataset and model weights for the agricultural robotics community. Full dataset is available at, https://huggingface.co/datasets/nitindominicrai/strawberry_pedicel_grasp_teleoperation and trained policy at, https://huggingface.co/nitindominicrai/act_strawberry_pedicel. To replicate this work, all code, configurations, and deployment instructions are publicly available at, https://github.com/nitin-dominic/autonomous_strawberry_pedicel_harvesting.

    manipulationteleoperationaction chunkinggrasp
  251. arxiv:2610.22763 · cs.LG
    Acquisition Function-Guided Generation for Bayesian Optimization over Discrete Spaces
    Tommaso Ancilli, Thomas Wedenig, Henry Moss, Robert Peharz

    Bayesian optimization (BO) over discrete spaces is challenging due to missing gradients and the sparsity of valid solutions. Latent-space BO sidesteps both issues by fitting a surrogate in the latent space of a generative model, which however substantially complicates the direct incorporation of domain knowledge that is naturally expressed in the original discrete space. In this work, we instead combine a pretrained generative model with a surrogate defined directly in data space, where inductive biases such as domain-specific kernels can be readily incorporated. Rather than maximizing the acquisition function over the discrete design space, our method samples candidates from the generative model under a distribution tilted by the acquisition function. To recover useful gradients for sampling, we embed this tilted distribution into a diffusion bridge that yields a smooth, posterior-averaged acquisition function. This leads to \emph{acquisition function guided generation} (AFGG), which we instantiate with a categorical diffusion model and an autoregressive normalizing flow. Across a broad suite of molecular benchmarks, the diffusion model variant performs on par with leading BO methods, while the flow model variant outperforms state-of-the-art methods in most cases and remains highly competitive otherwise. Our implementation is available at https://anonymous.4open.science/r/afgg.

    benchmark
  252. arxiv:2610.22762 · cs.RO
    Shape-Aware Neural Time Fields for Rigid-Body Motion and Manipulation Planning
    Yanming Zhou, Yuchen Liu, Ahmed H. Qureshi

    Rigid-body motion planning is fundamental to robotic transport and manipulation, where both object position and orientation determine whether a body can move through cluttered environments. This coupling makes planning chal- lenging because translation and rotation can be constrained very differently by surrounding geometry. We propose Shape- Aware Neural Time Fields, a physics-informed neural motion planning framework that learns a reusable travel-time, or cost- to-go, representation for rigid-body motion while explicitly accounting for these distinct geometric restrictions. The learned field captures local motion difficulty in translation and rotation and is paired with a sampling-based planner that balances progress toward the goal with the cost of reaching nearby poses. We evaluate the framework across 2D and 3D rigid- body planning tasks and non-prehensile manipulation through object pushing in diverse environments. Results show substantial improvements in planning success and inference efficiency over physics-informed neural planning baselines, while remaining competitive with classical sampling-based planners at significantly lower planning times. Ablation studies further validate the contributions of both the learned representation and the proposed planning strategy. Finally, closed-loop pushing experiments under execution deviations and external disturbances demonstrate that fast replanning enables reactive and adaptive task execution.

    manipulation
  253. arxiv:2610.22760 · cs.AI
    Between Spoilers and Guesswork: Avoiding Over-Specification in Coding Benchmarks
    Yuanzhe Liu, Bastian Lipka, Avraham Shinnar, Jatin Ganhotra +2

    Benchmarks drive AI progress, but are hampered by task under-specification and over-specification. Task under-specification causes guesswork, where solutions have to make things up, or worse, resort to contamination or reward hacking. Task over-specification causes spoilers, where solutions can bypass some of the harder steps the benchmark was supposed to measure. Since future-looking benchmarks should be difficult, over-specification reduces a benchmark's shelflife of community adoption. Repository-level coding benchmarks have helped the field progress but experience unnecessary churn due to both over- and under-specification. For instance, SWE-bench Pro, FEA-bench, Doc2Feat-bench, and Multi-SWE-bench all tackle under-specification by augmenting task statements with hints derived from ground truth. We show that leaking ground-truth information into the task statement causes over-specification in some of these benchmarks. We experiment with various specification levels, trying to thread the needle between spoilers and guesswork. Based on these results, we recommend techniques for deriving parsimonious hints. We will release our hints so the community can obtain more reliable signals from popular benchmarks.

    benchmark
  254. arxiv:2610.22756 · cs.CV
    Input- and Feature-Space Correction for Cross-Hospital IHC Grading
    Jongwoon Lee, Jhair Gallardo, Christopher Kanan

    Pathology foundation models (FMs) are trained predominantly on hematoxylin and eosin (H&E) images but are increasingly used as frozen feature extractors for immunohistochemistry (IHC), where cross-hospital differences can impair grading. However, systematic comparisons of correction methods for cross-hospital IHC grading remain limited. To address this gap, we benchmark label-free transformations applied to input images before frozen-FM encoding and to features after frozen-FM encoding using UNI and Virchow2, three IHC markers, and six directed cross-hospital transfers, yielding 12 settings. Among the evaluated input-space methods, our proposed method FIX achieved the largest macro-F1 improvement over no correction, outperforming well-known stain normalization techniques. Among feature-space methods, a simple feature-wise standardization achieved the best improvement over no correction, outperforming CORAL. Under patient-averaged tile-level evaluation, combining FIX with feature-wise standardization improved macro-F1 and quadratic weighted kappa (QWK) by 0.181 and 0.208, respectively, and reduced ordinal error by 0.209 relative to no correction. These results show that input- and feature-space transformations can provide complementary benefits for cross-hospital IHC grading with frozen pathology FMs.

    benchmark
  255. arxiv:2610.22754 · cs.LG
    Refusal Is Not Robustness: Measuring Harmful Reachability in Autoregressive Language Models
    Parv Kapoor, Teresa Fortes, Claudia Mamede, Rui Abreu +2

    LLM safety evaluations via jailbreaking typically index on attack success rate (ASR): whether a generated response violates safeguards. However, ASR only evaluates sampled outputs and provides little information on the underlying generation process. Autoregressive language models define distributions over possible outputs, and a sampled refusal can coexist with substantial probability mass on harmful generation paths. We study this hidden structure through bounded continuation analysis (BCA), which models autoregressive generation as a stochastic process over token prefixes and computes the finite-horizon probability that generation reaches a harmful continuation. BCA constructs a bounded continuation tree, labels terminal continuations with a semantic harm evaluator, and estimates harmful reachability using exact probabilistic model checking. BCA complements ASR by measuring how much harmful behavior remains accessible beneath a sampled safe response. This makes it useful for distinguishing models that look identical under measured ASR values, and allocating red-team effort toward regions of the generation process with greater hidden risk. Across four instruction-tuned LLMs and 100 JailbreakBench behaviors, we show that BCA effectively quantifies hidden harmful behavior and provides a useful signal to guide red teaming.

    evaluator
  256. arxiv:2610.22751 · cs.CV
    MedSTP: Preserving Subspace and Transition Evidence for Medical VQA Token Pruning
    Mengchen Fan, Qizhen Lan, Xi Xiao, Lin Zhao +5

    Medical vision-language models (VLMs) for visual question answering (VQA) often pass long visual-token sequences to the language model, making visual-token pruning attractive but risky for medical images, where answers may depend on localized boundary and texture evidence. We study training free pre-decoding pruning and treat pruning as a subspace stable evidence preservation problem. The retained tokens should remain close to the dominant visual evidence subspace induced by the full token set. To this end, we propose \emph{Subspace-Transition Token Pruning} (\method{}), which estimates each token's spectral support from principal component space and constructs a compact representation subspace. To better reflect medical image structure, \method{} further uses a bounded token-grid transition prior to promote boundary and texture sensitive tokens. We evaluate \method{} on the closed-ended SLAKE and OmniMedVQA benchmarks and on RadImageNet-VQA, which provides both closed-ended and open-ended evaluations, using Lingshu-7B and Qwen3-VL-8B backbones. \method{} delivers consistently strong performance on both the closed-ended and open-ended RadImageNet-VQA evaluations. These results show that \method{} is an effective training free pruning framework for medical domain VQA.

    benchmark
  257. arxiv:2610.22746 · cs.CV
    Cross-Person Identity Change Under Source-Coupled Multi-Human Generation: A Controlled UniPortrait Study
    Edrees Al-Edreesi, Ghaleb Al-Gaphari, Muneer Alsurori, Mohammed Edrees +1

    Multi-human personalization must account for identity changes in people outside the nominal target of a generative operation. We characterize this behavior in UniPortrait, without introducing a new preservation algorithm. Two sequential studies use 119 accepted two-person baseline scenes and three fixed output seeds, measuring continuous reference-based facial similarity with InsightFace \buffalo. In Study~1, source-coupled single-ID target-only generation (A1) produces negative mean transitions for both the protected person ($-0.567904$) and target ($-0.236904$), indicating broad perturbation rather than established successful repair with collateral loss. In Study~2, native dual-ID generation (A2) yields more positive paired contrasts: protected $+0.512918$ (95\% case-cluster CI $[0.471434,0.550259]$) and target $+0.158853$ ($[0.118903,0.197523]$), across 287 jointly measurable pairs from 111 cases. Shared baseline similarity cancels algebraically in these paired contrasts, leaving a matched POST comparison. Validation designed after the primary result freeze finds directional agreement with an independent identity evaluator and blinded human judgments; within-\buffalo\ nuisance models provide supportive sensitivity evidence. The findings characterize intact native regimes, whose references, routing, adapters, and token counts co-vary. They do not establish component causality, categorical identity preservation, successful localized editing, or generalization beyond the evaluated implementation and measurable cohort.

    evaluator
  258. arxiv:2610.22745 · cs.LG
    Reading the News Without Hindsight: NoLBERT Topics and the Cross-Section of Stock Returns
    Kevin Foley, Jonathan Hartadi, Shivesh Prakash, Charishma Takkallapalli +1

    Language models that read financial news are usually trained on text from the same years used to test them, so a backtest can reward hindsight. We build news topics from NoLBERT, an encoder pretrained only on text published between 1976 and 1995, adapted into a sentence encoder by contrastive fine-tuning on passages of the same article, with no labels or returns. We compare it with LDA, MiniLM and FinBERT readers on 1.26 million articles from 2012 to 2024, convert each reader's topics into attention shocks and stock exposures, and forecast the returns of the 1,000 largest US stocks with eight models at three rebalancing frequencies. Models are fitted on 2013-2019, choices are made on 2020-2021, and 2022-2024 is held out. The experiment with the best validation score loses money at every frequency. Measuring attention shocks against the past 63 trading days instead of 5 raises the test rank IC of every reader. NoLBERT gives the most stable topics and forecasts about as well as LDA. An equal-weight combination of all readers and models earns a monthly net Sharpe ratio of 1.12, significant against zero before adjusting for the search but not against a price-only benchmark.

    benchmark
  259. arxiv:2610.22744 · cs.CV
    OmniXRef: A Unified Framework for Multimodal Multi-Reference Video Editing
    Xiaoyan Cong, Tianyu Wang, Hyeonho Jeong, Zhifei Zhang +6

    Creative intent in video editing can be expressed through text instructions (e.g., "remove the people"), image references (e.g., subject appearance), video references (e.g., camera movement), or a combination of the above. We introduce OmniXRef, a unified framework for multimodal multi-reference video editing that allows users to combine variable numbers of text, image, and video references to control diverse attributes of a source video. We first devise a unified input design with a reference-bank interface and introduce $R^2$-RoPE to support generalization to more references. Then our model, OmniXRef-Model, uses a single-stream diffusion transformer to jointly process the source video and multimodal context from an MLLM in an interleaved sequence. To address the scarcity of training data for video-reference editing, we construct OmniXRef-Data using synthesis pipelines for different editing tasks and combinations of references. We use curriculum learning to progressively increase task complexity and resolution. To support systematic evaluation of video as a control modality, we introduce OmniXRef-Bench, a benchmark for video-reference attribute transfer across Camera, Motion, Content, Style, and Visual Effects (VFX). Comprehensive evaluations show that our OmniXRef-Model outperforms task-specific baselines and commercial models.

    curriculum learningbenchmark
  260. arxiv:2610.22740 · cs.AI
    R$^2$Orch: Recursive Self-Improvement for Agent Orchestration via Self-Play within a Skill-Internalization Loop
    Mingda Zhang, Yun Huang, Qiang Huang, Yanjin Li +6

    In recent years, agentic orchestration has been widely applied to solve complex human tasks. However, existing orchestration learning still faces key challenges, including hand-designed curricula, advantage collapse when a task's rollouts all reach the same outcome, and unreliable evaluation signals that judge each skill by one noisy score. To address these challenges, we propose R$^2$Orch, a new paradigm of Recursive Self-Play Orchestration: an inner loop of self-play that improves the orchestrator on self-chosen tasks, and an outer loop that improves how the inner loop improves. To support this paradigm, we introduce Recursive Learnability, a training objective that rewards a Proposer for choosing tasks whose rollouts keep reaching different outcomes as skills are withdrawn. Built on this objective, we further propose Skill Withdrawal, in which the outer loop withdraws each skill once a controlled evaluation within self-play shows that the orchestrator performs as well without it. Moreover, the two loops form one cycle: failures are distilled into skills, skills into parameters, and withdrawals lead the Proposer to new failures, so the orchestrator moves toward zero-skill deployment. Experimental results on twelve datasets show that R$^2$Orch consistently outperforms baselines across question answering, math, code, and interactive tasks.

    agentagenticself-improvementself-play
  261. arxiv:2610.22736 · cs.RO
    Spatial Memory by Agents, for Agents
    Qianwei Wang, Ce Zhang, Zhanpeng Luo, Pengpeng Su +2

    Spatial memory enables robots to reuse past observations for localization, navigation, and question answering. Designing such memory involves jointly deciding what to retain and how to retrieve it for diverse semantic, metric, and temporal needs. We investigate whether coding agents can automate this design process through iterative implementation and evaluation. Under a fixed development budget, a Designer Agent writes and revises code for memory construction and retrieval using feedback from a fixed Query Agent. The resulting implementations are frozen before building memories from held-out observation histories, without access to downstream questions, and evaluated across five complementary task tracks using the same Query Agent. Across three independent design searches, Codex / GPT-6 Astra averages 44.23 on the equally weighted five-track score, versus 29.32 for ReMEmbR, the strongest evaluated baseline supporting all five tracks, a 50.9% relative improvement. Qualitative analysis relates performance differences to evidence retention and retrieval, while identifying grounding and evidence-use failures. These results demonstrate the potential of coding agents to automate spatial memory design, while revealing task-dependent trade-offs in performance and computational cost. Project page: https://ba-fa-3344.github.io/.

    memoryagent
  262. arxiv:2610.22734 · cs.AI
    Graph Serializations for Reasoning LLMs: A Study Across Knowledge Extraction and Structural Reasoning
    Julien Yu, Mihaela Bornea, Avirup Sil, Katrin Kirchhoff

    The way a knowledge graph is serialized as text for inclusion into LLM context can substantially affect downstream LLM performance, with prior work reporting tens of percentage points variations across formats. However, existing studies consider only a limited set of representations and predate frontier reasoning models, which now offer stronger reasoning, instruction-following, and long-context capabilities. We evaluate 15 graph representation formats across nine frontier LLM configurations on ontology-based knowledge extraction and structural graph reasoning tasks. While knowledge extraction is relatively insensitive to different serialization formats, the choice of the optimal format for structural reasoning tasks significantly affects accuracy. We demonstrate that the optimal representation choice depends on model, query, and task and propose to approach graph serialization as a dynamic orchestration decision.

    long-contextknowledge graph
  263. arxiv:2610.22727 · cs.LG
    OrderFlowDiT: Transport-Regularized Diffusion Transformer for Probabilistic Limit-Order Supply Forecasting
    Jung Hun Phee, Tucker Balch

    We study the conditional distribution of limit-order volume submitted over the next five minutes, given the current order book and recent order flow. The task requires capturing price placement, total volume, and predictive uncertainty jointly. Coordinate-wise diffusion losses do not explicitly account for price displacement, while normalizing fields to unit mass removes information about total volume. Agreement on selected summary statistics also does not establish fidelity to the joint field distribution. We introduce OrderFlowDiT, a conditional Diffusion Transformer for unnormalized bid and ask volume-at-price fields, trained with an unbalanced optimal-transport (UOT) auxiliary loss. We also develop a full-field evaluation framework that combines a transport cost sensitive to price placement and total volume with proper scoring rules and calibration diagnostics. In the primary comparison on 1950 validation conditions across 30 U.S. equities, adding UOT reduces full-field transport cost by 45.8% and the transport energy score by 26.1% relative to training with mean squared error alone. Both measures improve for every stock. These findings support incorporating price geometry and volume into generative modeling, with field accuracy and predictive uncertainty evaluated together.

    evaluation framework
  264. arxiv:2610.22724 · cs.CL
    A misinformation playbook sharpens judgement without breeding cynicism
    Tong Guo, Teng Ye, Yutong Xie, Qiaozhu Mei

    Misinformation harms public health and democracy, and generative artificial intelligence now writes falsehoods as persuasive as people's. Defences range from fact-checking individual claims to training people to resist falsehoods. Yet many claims are too new to check, and interventions that teach doubt can reduce belief in truth too. We propose a defence that targets not the facts of each claim but the persuasion strategies that recur across them. Here we show that learning misinformation's persuasion playbook, mapped from 214,646 fact-checked claims in 51 languages, can sharpen human and machine judgement of claims they know little about, without making either more cynical about true claims. The playbook spans social influence, information manipulation and emotional appeals, combined around a core of fear, conspiracy and fabrication. Briefed with it, plus a reminder that persuasion is not proof of falsehood, a language model better distinguished true from machine-generated false claims about events after its training data. In a pre-registered experiment, the same briefing improved people's discernment of claims they reported as unfamiliar and left their belief in true claims unchanged. Averaged over all claims, the gain was not statistically significant. Because it targets what recurs, this defence need not wait for claim-specific fact-checks.

    manipulation
  265. arxiv:2610.22721 · cs.LG
    CoLU: Confidence-Weighted Layerwise Unlearning for Large Language Models
    Jingkai Guo, Ziyi Zhao, Chaitali Chakrabarti, Deliang Fan

    Large language model (LLM) unlearning aims to forget certain information while preserving others, without retraining from scratch. SoTA methods such as NPO, SimNPO, and SatImp balance forgetting and retain through whole model fine-tuning, while FOM-UL explores selective layer updates. Prior works show that suppressing target information can damage retained knowledge. This trade-off can vary across layers even under the same unlearning loss. In this work, we propose Confidence-Weighted Layerwise Unlearning (CoLU), which combines a novel confidence-weighted forgetting and retain regularization while updating a single selected transformer layer, rather than the whole model for unlearning. To reduce CoLU compute cost, we further propose Brief Layer Screening (BLS), which uses a newly defined normalized forgetting-retention (NFR) score to first rank layers through a brief unlearning trials using a subset of forgetting dataset and fewer epochs. BLS pre-selects top-k ranked layer candidates, which will then be fully profiled using full data analysis. BLS is proven to reduce the runtime of CoLU by 60.2-79.7% with almost the same performance. On TOFU dataset, CoLU improves both forgetting and retention over NPO, SatImp, and GA-based FOM-UL in all six TOFU settings, and over SimNPO in five of the six settings. On MUSE Books and News datasets, CoLU achieves the highest retained KnowMem and the PrivLeak closest to zero among the compared SOTA unlearning methods. CoLU presents a new record performance in both unlearning and retain in most settings up to date. Moreover, across the measured settings, average CoLU GPU memory usage is 77.4% to 79.5% lower than full-model unlearning.

    memory
  266. arxiv:2610.22713 · cs.LG
    Spectral Limits of Polynomial Graph Filters on the Long-Range Ising Model Benchmark
    Amin Najafgholizadeh, Mohsen Afsharchi

    The Long-Range Ising Model benchmark (LRIM) guarantees long-range dependence by construction: each label is the node's spin times the spin field convolved with a power-law coupling, so any model's distance from the target is measurable. We prove that a readout whose filter is a polynomial in the adjacency, the linear core of isotropic message passing and spectral graph networks, leaves an error floor that no degree lowers, because the target varies along a level set of the spectrum on which every such filter is constant. Computed exactly, the floor and the best filter of each degree predict two spectral networks adapted from published architectures to within a tenth of a decade. A filter escapes the floor only if it weighs separately the offsets the adjacency lumps together, and must still resolve the coupling across the whole lattice. A strided filter of about fifty coefficients does both: fitted to the training labels, its test error is 52 times below the best published network's. A dilated convolutional network beats the best published result at 256 nodes, matches it at 1,024, and gives the first in-domain results at the three larger sizes, while a Fourier neural operator with its standard modes falls further behind as the lattice grows. At its standard budget the dilated network stays two to four orders of magnitude short of a larger strided filter fitted to the labels. The analysis carries over to the benchmark's second coupling.

    benchmark
  267. arxiv:2610.22711 · cs.LG
    Normalizing Flow Maps
    Francesco M. Ruscio, Timothy James Hitge, Fred Zhangzhi Peng, Danyal Rehman +5

    Scalable approaches to learning neural dynamical systems have placed diffusion, flow matching, and few-step generators in flow maps at the forefront of generative modeling. Such state-of-the-art models have limited access to explicit model densities despite evolving from classical generative modeling approaches that leveraged tractable likelihoods, e.g., Normalizing Flows and Autoregressive models. In this paper, we combine these threads of few-step generation and tractable sample likelihoods and introduce Normalizing Flow Maps (NFM), a class of exactly invertible pushforward maps that generalizes normalizing flows and flow maps. More precisely, NFMs learn the integrator of an underlying dynamical system, but crucially, they provide efficient exact likelihood evaluation by construction. To enable scalable training of NFMs, we derive two complementary axes of consistency inherent to NFMs that yield (self) distillation objectives: (1) pointwise matching, which unlocks $\ell_2$-based flow-map style regression losses, and (2) distributional matching, which matches pushforward distributions to the marginals of a prescribed forward process. On conditional image generation on CIFAR-10 and ImageNet-64, NFMs achieve state-of-the-art sample quality among parameter-matched normalizing flows, while supporting likelihood-based RLHF. For equilibrium sampling in Boltzmann Generators, NFMs outperform prior state-of-the-art normalizing flows on both single-system and transferable peptide benchmarks.

    benchmark
  268. arxiv:2610.22708 · cs.AI
    GlobeBench: Evaluating Text-Based World Model Capabilities for Scalable Environment Simulations
    Nicholas Saban, Charvi Bannur, Mariya Vasileva, Darshan Deshpande

    Text-based world models have recently shown potential to simulate complex RL environment state transitions for a diverse set of use cases. However, no standardized benchmark exists to compare, study and improve these models. Through this paper, we attempt to solve this problem by building GlobeBench, a novel benchmark with 1,632 tasks across five separate domains and nineteen simulation skills. The tasks check whether models can predict the environment's response to a specific action given the full agent trajectory by understanding earlier state changes, following environment rules, and predicting the effects of actions on both related and unrelated objects. We compare the predicted environment response with an execution-derived reference, using a hybrid architecture with both deterministic gates and a semantic judge. Among 20 general-purpose LLMs and specialized text world models, GPT-6 Astra achieves the highest task success rate at 49.57%. We find that model accuracy decreases as actions affect a larger portion of the environment. Specifically, when the agent's changes must propagate to dependent objects, task success drops to an average of 5.1% across all 20 models. Although model strengths differ across tasks, 631 tasks remain unsolved by any model, suggesting a significant gap in general LLM capabilities as a world model.

    world modelagentbenchmark
  269. arxiv:2610.22703 · cs.AI
    NoveltyScope: A Multi-Agent Research Assistant for Evidence-Grounded Novelty Assessment of Scientific Papers
    Yueran Tian, Huizhi Liang

    NoveltyScope is a prototype that compares a scientific paper with retrieved prior work and produces a structured report and a five-level novelty score. The system extracts the core task and at most two author-claimed contributions, then searches a closed local corpus using BM25 and dense retrieval with reciprocal rank fusion. Contribution comparisons request paired passages for potential refutations, and quotations are checked against cached source text. Four scoring roles (Prosecutor, Defender, Judge, and Critic) use the resulting evidence, with at most one revision of the Judge's score. Evaluation uses 50 class-balanced RINoBench samples whose idea-level labels are retained for full-paper input. Mean absolute error is 1.48 and 1.26 for single-LLM and multi-agent scoring with Gemini, respectively, and 1.22 and 1.16 with Qwen3. The vector-similarity baseline has an error of 1.68. A HiCLIP case study illustrates the contribution comparisons and paired evidence passages. Overall, the results demonstrate the feasibility of evidence-grounded novelty assessment and show that the multi-agent configurations achieve lower scoring errors than their single-LLM counterparts across both evaluated models.

    multi-agent
  270. arxiv:2610.22701 · cs.MA
    Emergent Discrimination and Polarization in Multi-Agent Systems
    Nestor Caticha, João Henrique Sant'Ana, Felippe Alves, Davi Bastos Costa +4

    We study theoretically how discrimination and polarization can emerge in a population of learning agents. Each agent uses an adaptive linear decision layer to map a language-model embedding of an issue to a binary opinion and maintains an evolving estimate of how much it trusts every other agent. Through pairwise interactions, agents update both their decision weights and trust estimates using an online Bayesian rule derived from teacher-student learning over a noisy channel. In-context-learning experiments with large language models qualitatively reproduce the predicted opinion and trust updates. We then assign each agent one of two arbitrary group labels, A or B, and model group bias by allowing a fraction $f_b$ of agents to shift its perceived agreement with each message by a strength $b$, a bias steering prejudice, depending on whether the sender and receiver share a label. We use order parameters from statistical mechanics to connect these local dynamics to collective states and transitions. Starting from uniformly random initial conditions, simulations reveal a rich phase diagram in the $(b,f_b)$ plane, comprising four collective states: spin-glass frustrated out-group favoritism, nondiscriminatory polarization, discrimination with opinion alignment, and discrimination without opinion alignment. Which state emerges depends on both the strength and prevalence of the bias. These results show how simple local learning dynamics can generate qualitatively distinct forms of polarization and discrimination, underscoring the need for population-level safety analysis of multi-agent systems. They also suggest measurable order parameters for characterizing analogous collective states in human societies.

    agentmulti-agentagent system
  271. arxiv:2610.22700 · cs.CV
    Thinking from Above: Efficient Spatial Reasoning with Bird's-Eye-View Visual Tokens
    Faizan Siddiqui, Nirav Savaliya, Sahith Reddy Chada, Enna Sachdeva

    Existing video multimodal large language models (MLLMs) organize observations primarily in time, but embodied spatial reasoning requires organizing them in space. Consequently, their visual representations can remain locally consistent across nearby frames while lacking a persistent spatial context that is stable across viewpoints and time. We exploit depth and camera pose, which are naturally available to embodied agents, to explicitly establish this spatial correspondence. We introduce Cambrian-BEV, which extends Cambrian-S with a parameter-free geometric tokenization scheme that back-projects visual patch features from each frame into a shared bird's-eye-view (BEV) coordinate frame and averages features that fall within the same spatial cell. Thus, observations of the same physical location, regardless of viewpoint or time step, are aggregated into a common spatial token. In addition to providing a persistent scene representation, this formulation acts as an efficient compression mechanism: 128 frames containing approximately 100K DINOv3 patch features are consolidated into a fixed 27$\times$27 grid of only 729 BEV tokens, cutting LLM prefill time by $2.9\times$ and peak memory by $3.6\times$. Cambrian-BEV requires no architectural modification beyond BEV tokenization and uses a frozen DINOv3 vision encoder with a trainable multimodal projector and LLM. On VSI-Bench, our model reaches 63.48, $+8.5$ points over an identically fine-tuned frame-token control with $12.6\times$ fewer visual tokens, with particularly large gains on room size ($+24.5$) and relative direction ($+34.1$). On ReVSI, it reaches 44.68, $+4.5$ over the corresponding control.

    embodiedmemoryembodied agent
  272. arxiv:2610.22692 · eess.SY
    Lipschitz Bounds for Neural Networks Using a Two-Copy Linear Fractional Transformation
    Felix Biertümpfel, Bin Hu, Peter Seiler

    This paper proposes a new method to bound the global Lipschitz constant of a neural network (NN). Our formulation uses a two-copy linear fractional transformation (LFT) representation of the NN with stacked signals. We use biased quadratic constraints (QCs) to bound the input/output behavior of the nonlinear activation functions. We then formulate a linear matrix inequality condition, using these biased QCs, to bound the Lipschitz constant of the NN. The best (smallest) Lipschitz bound is computed via a convex, semidefinite program. There are two key distinctions of the proposed two-copy formulation in comparison to existing methods that rely on a single incremental representation with incremental signals. First, the NN bias terms are eliminated in the standard incremental representation, but appear in our formulation. Second, there is a larger class of QCs that can be expressed for the nonlinearities in our formulation than in the incremental representation. We show that our proposed formulation yields less conservative Lipschitz bounds due to these two differences. We illustrate these points via several simple, motivating examples. We also demonstrate the proposed method using a benchmark problem with a NN controller for an inverted pendulum.

    benchmark
  273. arxiv:2610.22688 · cs.AI
    Explore Separately, Adopt Selectively: Execution-Guided Tool Use over Large Tool Libraries
    Anna Mazhar, Seiji Maekawa, Nikita Bhutani

    Large language model agents must discover relevant tools from large libraries, e.g., thousands of tools, as new needs arise during task execution. When several candidates could satisfy the same need, execution may be required to identify the relevant one. Investigating candidates in a shared trajectory can leave alternatives unexplored or allow failed attempts to influence subsequent reasoning. To address this, we present BRIA (BRanch-Inspect-Adopt), a simple yet effective agent that temporarily isolates candidate investigations in separate branches. Each branch retrieves and executes tools to investigate its candidate; a task agent inspects the resulting execution evidence and adopts only useful results into its continuing trajectory. BRIA retrieves additional tools as new capability needs arise during task execution. We evaluate BRIA on PlanBench-XL and MCP-Atlas using a shared pool of 1897 tools, three language models, and static and adaptive retrieval baselines. Across all settings, BRIA matches or exceeds the strongest retrieval baseline. On MCP-Atlas, it achieves an average claim score of 89.77, compared with 75.40 for MCP-Zero, a gain of 14.37 percentage points. BRIA also lies on the Pareto frontier for task performance and input-token cost.

    agenttool use
  274. arxiv:2610.22686 · cs.CL
    Multimodal Filesystem Memory For Long-Horizon Embodied Agents
    Chloe Su, Jiahan Li, Zhenting Qi, Chaoran Cheng +4

    Agents operating over long horizons need an effective memory that allows them to reuse their experience throughout long interactions. If the entire history of an interaction is retained, it becomes difficult to find the relevant experience, while compressing it into summaries may result in the loss of visual and spatial details that are required later on. We present \textbf{MFM}, a multimodal filesystem memory architecture that keeps complementary types of agent memory. \textbf{Episodic memory} stores the observations, actions, and outcomes, and \textbf{spatial memory} associates experiences with locations. \textbf{Semantic and procedural memory} build up general knowledge and reusable methods of acting, and \textbf{working memory} holds the current plans. MFM makes use of searchable textual records to connect these different representations to the underlying multimodal episodes. The agents are able to find previous experiences by means of contextual cues and to examine the original observations in order to recover details that are missing from their textual descriptions. Through consolidation, the agents convert experience into higher-level knowledge and procedures while still keeping references to the original episodes. In the case of long-horizon Minecraft and computer-use tasks, MFM leads to improved performance with three base models. With GPT-6-Astra, MFM completes \emph{Mine Diamond from Scratch} and reaches 16 out of 24 milestones on the \emph{Ender Dragon} task, as against 12.7 for the best baseline. On CUABench-KiCad it attains a mean reward of 0.440, as compared to 0.400 for the best baseline and 0.267 in the case of no persistent memory.

    embodiedmemorymemory architectureepisodic memorypersistent memoryagent memory
  275. arxiv:2610.22684 · cs.LG
    AlphaBet: Reinforcement Learning for LLM Event Forecasting Agents
    Liqin Ye, Rongzhi Zhang, Xin Liu, Zheng Li +4

    Forecasting an unresolved event requires finding relevant evidence and deciding how strongly it supports each outcome. Yet outcome-based training of Large Language Model (LLM) forecasters often begins with preassembled evidence at a single prediction date, leaving evidence acquisition outside the learning loop. We introduce AlphaBet, an agentic reinforcement learning (RL) framework that trains a single LLM policy to search, reason, and report probabilities across multiple pre-resolution dates. Each dated forecast is scored by a proper scoring rule against the eventual outcome, allowing one resolved event to supervise decisions under different information states. We decompose forecasting regret into an evidence-acquisition gap and a probability-reporting gap, and show that Group Relative Policy Optimization (GRPO)'s standard-deviation normalization can distort the reporting target under multi-class proper scores. On GLM-4.5-Air, AlphaBet reduces the Brier score from 0.731 to 0.626 on Forecast-Dojo and from 0.612 to 0.528 on FutureX. Controlled ablations attribute these gains to multi-date training and show that they come from improvements in both information acquisition and probability reporting. Our empirical results validate that GRPO's std normalization degrades proper-scoring-rule performance while leaving top-outcome accuracy largely unchanged.

    agentic
  276. arxiv:2610.22683 · cs.MA
    Dynamic Auction Mechanisms in a Closed-Credit Economy: Sequential Slot Reallocation in Air Traffic Management
    Jihun Kim, Pan-Yang Su, Maria G. Mendoza, Victoria Marie Tuck +2

    Indivisible resources allocated in advance may need to be reallocated when capacity reductions make existing assignments infeasible, yet monetary compensation is often impractical or undesirable. We develop a sequential resource reallocation scheme using reusable artificial credits, where agents hold pre-existing resource rights and have heterogeneous marginal valuations over time blocks. Capacity reductions produce fractional endowments, which our closed-credit scheme resolves into indivisible final allocations. We model agents' intertemporal preferences through a predicted-value-to-go function and characterize conditions ensuring that their optimal demand is non-increasing in price. We propose an ascending-auction mechanism that partitions agents into subgroups and uses the minimum subgroup market-clearing price as a common transaction price. We show that the mechanism preserves non-myopic individual rationality and weak nontrader neutrality. Moreover, we establish Pareto efficiency under optimal-demand clearing in the unified market and bound the expected gain from misreporting by the probability that an agent changes the common price. Simulations of airport slot reallocation demonstrate reductions in total modeled delay and cancellation costs relative to a ground-delay baseline.

    agent
  277. arxiv:2610.22681 · cs.CV
    EgoP$^{2}$C: Internalizing Plan-Perceive-Compose for Spatial Reasoning in Egocentric Video
    Chaoyu Li, Arka Sadhu, Yuliang Li, Hyo Jin Kim +2

    Answering complex spatial reasoning questions from egocentric video requires determining what evidence is needed, acquiring targeted observations, and combining them through multiple reasoning steps. Existing vision-language models (VLMs) struggle to perform this process reliably. Agentic systems address this challenge by querying external vision and geometry tools, but training their reasoning leaves these tools unchanged, so inaccurate observations can still lead to incorrect answers. We introduce EgoP$^{2}$C, a framework that internalizes a plan-perceive-compose process within a single VLM. To this end, we first train specialized Planner and Perception teachers using separate adapters (Dual-LoRA), then fine-tune the corresponding experts in a mixture-of-experts (MoE) student under their supervision. We subsequently train the student on trajectories generated by its current policy using a segment-level extension of multi-teacher on-policy distillation (MOPD), assigning different trajectory segments to different supervisors. The Planner teacher supervises which evidence-seeking questions to ask, when to stop gathering evidence, and how to compose the final answer. The Perception teacher supervises the video-grounded answers to those questions. At inference, the student executes this adaptive workflow end-to-end, without external spatial tools. Across three spatial reasoning benchmarks, EgoP$^{2}$C students built from Qwen3-VL-8B and SenseNova-SI-1.3-8B outperform their base models by an average of 9.2 and 7.5 percentage points, respectively. Across both backbones, EgoP$^{2}$C reduces inference FLOPs by 65.8-66.3% and achieves $1.29$-$1.49\times$ speedups over their shared Dual-LoRA teacher.

    agenticbenchmark
  278. arxiv:2610.22674 · cs.MA
    Gossip Agents: Reputation through Opinion Dynamics over Evolving Networks
    Yutong Wu, Arlei Silva

    Sustained cooperation among self-interested learning agents is essential in many shared environments. Reputation systems are among the best-studied mechanisms for promoting such cooperation. Existing multi-agent reinforcement learning approaches, however, typically assume that reputation is either globally available to all agents or derived solely from an agent's own observations, leaving largely unexplored how reputational information propagates between agents. We study gossip, the exchange of information about third parties, as a mechanism for reputation diffusion over dynamic communication networks. Agents form private opinions from limited observations and exchange them with their neighbors following DeGroot opinion dynamics, while the communication network changes as agents interact and move. A tunable broadcast radius controls the reach of information, spanning private reputation, local gossip, and globally shared reputation. We study this framework in two \(n\)-player social dilemmas with distinct network dynamics: a Prisoner's Dilemma with randomly resampled interaction groups and Harvest, where communication neighborhoods emerge from agents' movements. Our results show that the reach of reputational information shapes cooperation in both environments. Moreover, in Harvest, reputation diffusion increases environmental rewards and leads to more peaceful and sustainable behavior.

    multi-agent
  279. arxiv:2610.22672 · cs.AI
    Not All Graphs Are Maps: A Task-Agnostic Exploration Framework for Mobile GUI Navigation
    Adamo Orsini, Seyed Mehdi Ayyoubzadeh, Sadman Rakib Pinon, Hannuo Zhang +4

    Modern mobile GUI agents can perceive visible screens and execute local actions, but they still lack persistent app-specific navigation knowledge: where functions live, which pages contain them, and which screens are equivalent across repeated visits. We introduce TAG-Nav, a Task-Agnostic Graph Navigation framework that wraps existing GUI-agent backbones with reusable destination memory. Offline exploration is distilled into compact records of app functions, target pages, and waypoint routes before any evaluated instruction is seen. When a future task's destination is covered, TAG-Nav retrieves the destination and provides route guidance without task-specific demonstrations. This separates mobile GUI destination navigation from full task completion: many tasks share the prefix of reaching the right page, while final actions may enter content, send data, make purchases, delete information, or change sensitive settings. To evaluate route finding without downstream execution confounds, we introduce DestBench, a navigation-only protocol that asks whether an agent reaches the correct pre-action destination screen. DestBench enables a controlled graph-memory study comparing no-memory agents with random, automated BFS, and clean graph memory. Graph-derived memory improves strict navigation success rate of UI-TARS-1.5-7B from 52.4% without graph memory to 58.3% with automated BFS memory, and improves Qwen-VL-Max success rate from 59.5% to 71.4% on the same conditions. Together, TAG-Nav and DestBench show that destination coverage and route cleanliness strongly shape graph-memory value, making graph quality a measured variable rather than an assumed benefit of exploration.

    memoryagent
  280. arxiv:2610.22671 · cs.CL
    ProPerAgent: Toward Memory-Driven Proactive Assistance
    Shenglai Zeng, Xiao Yang, Tiffany Ma, Xinyuan Zhang +14

    As assistants move from reactive answering toward proactive intervention, memory changes role: it no longer only supplies the answer to a query, it determines whether and when to speak at all. Prior work in this setting hands the system the relevant fact in context; we study the regime that makes the problem hard, where the evidence justifying an intervention must be mined from a long-horizon personal memory. We introduce ProPerMem, a benchmark for memory-driven proactive assistance (MPA), pairing egocentric video with year-long, distractor-rich user memories (~618K tokens on average), human-verified trigger windows, and a protocol that scores content and timing jointly. Every baseline we evaluate performs poorly (at most 15%), including frontier models given summarized or retrieval-augmented memory. A controlled study explains why: with oracle evidence delivered inside the trigger window GPT-5.3-Chat reaches 96%; letting it choose the timing drops it to 38%, and replacing oracle evidence with retrieval drops it to 15%. Content and timing are separately necessary, and missing either is catastrophic. Guided by this, we build ProPerAgent (Proactive Personalized Agent) with a plug-and-play harness that surfaces memory from a graph-organized store and gates it down to what is worth interrupting for, plus a lightweight RL stage that curbs early and repeated firing. Without any training, ProPerAgent lifts strict accuracy from below 10% to 37.5% on Qwen3.8-27B and 38% on GPT-5.3-Chat; RL post-training reaches 42.4%.

    memoryretrieval-augmentedpost-trainingbenchmark
  281. arxiv:2610.22668 · cs.AI
    From Relevant Memories to Sufficient Evidence: Rethinking Memory Utilization in LLM Agents
    Shixuan Li, Wei Yang, Peiyu Zhang, Anzhe Cheng +6

    Long-term memory enables LLM agents to reuse past experiences, yet retrieving relevant memories does not necessarily provide sufficient evidence for answering a query. Essential information may remain fragmented across memory records, obscured by abstraction, or disconnected from the relations required for reasoning. We identify this relevance-sufficiency gap and propose QEMem, a framework that reformulates memory utilization as the construction of query-specific evidence sets, shifting the focus from individual retrieval relevance to collective evidential support. QEMem integrates three complementary mechanisms: query-guided discovery retrieves evidence across multiple memory granularities; source-grounded reification organizes it into structured ViewCards with explicit evidential roles and source references; and sufficiency-guided refinement selectively recovers missing information. Operating on existing memory stores without modifying memory formation, QEMem achieves 91.43% LLM-judge accuracy on LoCoMo with Claude Opus 4.5, outperforming the strongest evaluated baseline, MemR$^3$, by 7.6 percentage points and NEMORI by 10.7 points under a shared memory store. Further analyses demonstrate improved evidence coverage and strong performance on questions requiring cross-session evidence integration.

    memoryllm agent
  282. arxiv:2610.22661 · cs.LG
    Edit Discovery in Human-AI Interactions
    Alireza Mirrokni, Shayan Kiyani, George Pappas, Hamed Hassani

    Human-AI interactions are becoming a rich source of data for improving AI systems: while many model responses are already satisfactory, others expose failures that could become valuable supervision if identified and corrected. We study this setting through the notion of an editor: a more capable evaluation process that can inspect a response and either confirm it or modify it, but whose invocation is costly. The editor may be a human expert, a stronger AI system, or the same model equipped with additional computation, context, tools, or information. The key challenge is to decide which interactions to send to the editor before knowing whether an edit will actually be made, since invoking the editor itself consumes limited resources such as expert time, inference cost, latency, or tool-use overhead. We formulate this problem as Edit Discovery Optimization (EDO), which seeks to minimize edit opportunities missed among skipped interactions while controlling unnecessary confirmations and respecting an editor-cost budget. We characterize the population-optimal policy and show that it takes the form of a cost-adjusted threshold on the posterior probability of an edit. Building on this structure, we develop EDIT (Edit Discovery via Iterative Thresholding), an online algorithm for the partial-feedback setting in which edit labels are revealed only for selected interactions. EDIT combines an adaptive edit score, a confirmation-control threshold, and a cost-aware selection rule; it satisfies the editor budget at every round and controls the average confirmation rate up to a vanishing finite-horizon term, without distributional assumptions on the interaction stream. Across four reasoning and factual question-answering benchmarks, EDIT achieves a better trade-off between editor cost, Type-I error, and Type-II error than direct LLM-based batch selection.

    tool-usebenchmark
  283. arxiv:2610.22659 · cs.CL
    Exploring The Empirical Privacy Landscape of Synthetic Data
    Zihao Zhao, Avi Schwarzschild, Casey Meehan, Katherine Lee +2

    Methods for training language models on sensitive data differ in how they balance capability gains against the risk of leaking private information. In this paper, we present a controlled comparison of data-level and training-level interventions for navigating the capability-privacy tradeoff on synthetic medical records with canaries to probe for leakage. Direct finetuning performs best on capability metrics but leads to the most privacy leakage, while DP-SGD reduces leakage at substantial cost to capabilities. Generating synthetic data from these records offers a more favorable balance, particularly when guided by a uniqueness-aware instruction to replace private details. Relative to naive paraphrasing, privacy-aware rewriting improves privacy protection across a variety of different rewriting models. We also find that more capable rewriting models generally reduce the risks of downstream privacy leakage, although held-out fit does not improve monotonically and increased test-time compute yields inconsistent gains. Overall, these results highlight privacy-aware rewriting as a promising direction for achieving a more favorable trade-off between capabilities and empirical privacy compared to other training-level methods, and our evaluation framework supports comparisons of future models and methods.

    evaluation framework
  284. arxiv:2610.22655 · cs.LG
    Is a Linear Probe Evidence of a Linear Representation?
    Nadir Kutluozen, Christian Internò, David Klindt

    The linear representation hypothesis holds that pretrained vision models encode generative factors as linearly decodable directions in feature space, and comes with a standard methodology: train a linear probe, declare a factor represented if accuracy is high. We show this test is too permissive. Asymptotically, any sufficiently rich feature map, including a random one, admits a linear readout, so high probe accuracy is consistent with two very different geometries: a true linear world model, where the latent-aligned subspace dominates the representation, or a kernel machine, where the latents lie linearly inside a much larger curved manifold. This distinction matters because the curved structure orthogonal to the probe is relevant for out-of-distribution generalization, robustness, and compositional transfer. We propose three asymmetric diagnostics that distinguish the two cases: reverse predictivity, spectral concentration, and out-of-distribution generalization. We validate them in a fully observed simulation where ground truth is known. Applied to six pretrained vision transformers on SVG-World, our deterministic generative benchmark with paired visual styles, and to two of them on CelebA, they show pretraining produces structure beyond a random kernel but well short of a linear world model. On dSprites, 3D Shapes and MPI3D, forward decodability and latent-aligned variance dissociate. Thus, forward decoding alone does not show that the probed factors align with the dominant geometry of the representation. In summary, our work cautions about the success of linear probes as opposed to truly linear and low dimensional representations.

    world modelbenchmark
  285. arxiv:2610.22652 · cs.CL
    Evidence-Gated Bilingual RAG for Reliable Consumer Banking Search
    Ahmed Mohamed Mahmoud Rayyan, Rana Mahmoud Mohammed, Aya Abdallah Abdelmonem, Mohamed Khaled Goda Othman +1

    Consumers often describe what they need rather than use the terminology found on a banking website. For example, a natural-language request ("a card for travel") may not share the keywords of the relevant product page, and the problem is harder in bilingual services, where the same intent may be expressed in Arabic or English. This paper presents an evidence-gated bilingual retrieval-augmented generation (RAG) framework for consumer banking search that operates entirely on local infrastructure. The framework combines structured web-content refinement, hybrid dense-lexical retrieval, intent-aware evidence selection, cross-encoder reranking, and a three-outcome Answer / Related Mismatch / No Answer decision mechanism. On 210 URL-labeled Arabic and English queries, the proposed approach improves Top-1 hit rate from 50.48% to 79.05% and MRR@5 from 0.571 to 0.797 over a hybrid baseline. A separate 26-case end-to-end evaluation achieved 69.2% strict success, with 72.7% of generated answers classified as supported by an automated faithfulness judge. These results are preliminary: the retrieval queries are paraphrases of a small reference set, and the end-to-end test is small and uses an automated rather than human judge, so they do not establish human-verified answer correctness or generalizable performance.

    retrieval-augmentedrag
  286. arxiv:2610.22650 · cs.LG
    The Ephemeral Cost of Switching Optimizers Across LLM Training Stages
    Xinyu Zhao, Davis Wertheimer, Ashutosh Sharma, Rameswar Panda +1

    Recent work suggests that switching optimizers between pretraining and subsequent training stages incurs a performance cost, but systematic investigation of this phenomenon is lacking. We study the mismatch between AdamW and Muon optimizers by pretraining 1.2B Llama-3 models from scratch while varying the proportion of updates taken with each optimizer, and similarly interpolating between the two optimizers during continued pretraining and supervised fine-tuning (SFT). Surprisingly, we find little evidence of a meaningful optimizer mismatch penalty: final performance is largely explained by the independent effects of the pretraining and post-training optimizers, with downstream performance generally improving as the proportion of Muon updates during SFT increases. Although matching the pretraining optimizer can reduce forgetting during SFT, this does not translate into improved final benchmark performance. We further evaluate AdamW and Muon for task-general SFT of open-weight models, across the Llama and Qwen families up to 14B parameters, where Muon remains competitive with AdamW, despite these models having been pretrained with AdamW. Together, our results suggest that the choice of optimizer for post-training can be made largely independently of the optimizer used during pretraining.

    post-trainingbenchmark
  287. arxiv:2610.22637 · cs.AI
    Hard Instances on Demand
    Jamie Fravel, Martin Koutecký, Pavlo Tsitsei

    The conclusions of an experimental study of algorithms are only as good as the instances they are run on. Benchmark libraries for well-studied problems took decades to curate, while evaluations of new problems often fall back on random instances, which are known to be misleading: they may say more about the random generator than about the algorithmic question at hand. We treat instance generation as an optimization problem: parameterize the instance space, take the solver's running time as the objective, and let a black-box optimizer search for hard instances. We develop a widely applicable workflow built on the Nevergrad optimization platform and demonstrate it on four NP-hard problems: satisfiability, traveling salesperson, Hamiltonian cycle, and non-permutation flow-shop scheduling. The search needs only the solver, an instance encoder, and a compute budget. Our evolved instances beat the hardest instance of a budget-matched random draw in all four case studies, by an order of magnitude or more in three. They are often competitive with long-curated benchmark sets: our SAT champion is harder (for a three-solver portfolio) than 78% of the 498 SAT Competition formulas in its size class; our 20-city TSP champions take several times longer than the hardest measured classical TSPLIB instance (229 cities); and our flow-shop champion takes 238.6 seconds of CP Optimizer time to solve to proven optimality (versus 4.5 seconds for the hardest of 100,000 random draws) while denying CP-SAT an optimality proof. The optimizer even recovers known structure from running time alone: it locates a SAT clause/variable ratio near the phase transition and finds structurally different champions for the two scheduling engines.

    benchmark
  288. arxiv:2610.22631 · cs.AI
    A Plausible Explanation Is Enough to Mislead Multimodal Judges
    Anglin Cai, Zhiqi Wang, Xiaozhi Yang, Yurun Lin +2

    Multimodal large language models (MLLMs) increasingly serve as judges in training and evaluation pipelines, often assessing answers together with explanations. Although AI judges are sensitive to textual factors, it remains unclear whether ordinary plausible explanations can sway judgments requiring visual verification. We introduce JUPE, a visual-question-answering (VQA)-style benchmark that fixes answers while varying explanations. Across 14 MLLM judges, ordinary explanations that sound plausible mislead judges, with effects comparable on average to deliberately persuasive arguments. Plain justifications and chain-of-thought rationales reduce rejection of incorrect answers by 12.7 and 13.8 percentage points, respectively. We call this failure plausibility capture. Using signal detection theory, we find a significant mean shift in decision criterion toward acceptance, without a clear change in mean discriminability. Explanations sway judgments across domains, explanation generators, and judging protocols, while tested mitigations offer limited, model-dependent protection. Such shifts can inflate performance estimates when judges evaluate answers with explanations.

    benchmark
  289. arxiv:2610.22621 · cs.LG
    Noise-shaped Sampling and Annealing for Scalable On-the-edge Training of Binary and Ternary AI Models
    Zihao Chen, Zhili Xiao, Rui-Jie Zhu, Yipeng Wang +2

    The training loss of binary and ternary neural networks, expanded as a multilinear polynomial over the discrete weights, is structurally equivalent to the energy of a hierarchical high-order Ising model. We exploit this equivalence to apply Ising solvers to discrete-weight network training, while accounting for a fundamental difference in objectives: an Ising solver seeks the ground state of a fixed energy function, whereas training seeks solutions that generalize to unseen data. Our key contribution is $ΣΔ$-Sampler, an optimizer that reconciles these objectives by embedding a stochastic sampling-and-annealing step as a noisy quantizer inside a first-order $ΣΔ$ noise-shaping loop. The loop pushes noise from weight quantization, high-order frustration, and mini-batch sampling out of the low-frequency adaptation band, so that discrete weight updates follow the smooth descent trajectories embedded in the rugged loss landscape. We validate the framework on a planted quadratic unconstrained binary optimization (QUBO) benchmark with a known minimizer; on CIFAR-10, where a $13$M-parameter ternary model comes within $1\%$ of full-precision accuracy; on ImageNet-1k, where a ternary ResNet-18 comes within $0.4\%$ Top-5 accuracy of a ternary baseline; on class-incremental continual learning, where prior-task accuracy is preserved; and on MatMul-free language models up to $2.7$B parameters, where it outperforms a straight-through-estimator baseline under matched continued pre-training. Compared with standard quantization-aware training, $ΣΔ$-Sampler reduces the training memory footprint by over $50\%$, delivers an order-of-magnitude speed-up, and reduces weight-update memory traffic by four orders of magnitude, making it attractive for on-device training and fine-tuning of large binary and ternary models.

    memorybenchmark
  290. arxiv:2610.22619 · cs.RO
    FCTS: Failure-Guided Curriculum Tree Search Over Generated Terrain for Legged Parkour
    Narayanan Palghat Parameswaran, Dong-Ki Kim, Jay Patrikar, Santiago Gutierrez +3

    Quadruped parkour policies handle canonical obstacles well, but their training curricula often miss the irregular, compound geometry of industrial and disaster-response sites, where agile locomotion matters most. Such environments pose a hard generalization problem: even training directly on reconstructed geometry from the target site does not reliably yield policies that traverse it. Hand-designed curricula span only a few terrain families, and recent LLM-generated curricula offer little control over which policy weaknesses they target. We propose FCTS, a failure-conditioned terrain curriculum framework that uses localized traversal failures to automatically generate the terrains a policy should train on next. Each failure's location, termination mode, robot state, and surrounding terrain geometry condition this generation, and Monte Carlo Tree Search explores alternative curriculum trajectories under a fixed training budget. On terrains built from scanned real-world construction sites, FCTS raises success rate from 28.4% to 85.8% on a medium-difficulty course and from 14.9% to 44.7% on a hard course relative to the strongest parkour or LLM-curriculum baselines, and maintains higher progress on extreme terrains where full traversal is rare. Hardware experiments on a quadruped robot confirm that the learned FCTS policies transfer to the real world.

    quadruped
  291. arxiv:2610.22615 · cs.CL
    When Does Uncertainty Measure RAG Hallucination Risk?
    Rajeel Ahmad Ansari, Tautik Agrahari

    Uncertainty can measure hallucination risk in retrieval-augmented generation (RAG), but only when retrieved evidence constrains the answer. When evidence is present, unsupported factual claims become less likely; when retrieval coverage is poor, the model can answer from internal knowledge and assign high probability to unsupported claims. We test this across four instruction-tuned models on HotpotQA and SQuAD. After matching faithful and hallucinated answers within word-count bins, length-only classifiers fall near chance, while perplexity reaches 0.81--0.91 AUROC on targeted unsupported edits. Varying retrieval depth shows why the score changes meaning: detection can weaken with only the top-ranked passage, then recover after adding a second passage. A combined uncertainty and semantic-drift score improves calibration, and measurements remain above chance on RAGTruth QA. Results support a conditional view of uncertainty as a useful measure of RAG hallucination risk.

    retrieval-augmentedrag
  292. arxiv:2610.22610 · cs.LG
    Shift-Reliance Alignment: Ranking Vision-Language Models by the Cues They Trust
    Faisal Mohammad, Mohamed Younis

    Vision-language models (VLMs) are usually chosen from leaderboards, which give one ranking for every application. On questions about objects that a scene suggests but does not contain, however, the model with the lowest error on the POPE hallucination benchmark is not the one with the lowest error on such questions. For yes/no questions about object presence, we show that the best model for a target domain can be predicted without running any candidate on that domain, given 650 to 1,200 labeled target items that calibrate a detector. We measure each model's reliance on a cue, a feature such as scene context or text in the image, by editing 100 labeled source images so that the cue suggests the wrong answer, and comparing with a matched control edit. Reliance predicts which VLMs are misled by physical sticker attacks (Spearman .93 on nine models not used to design the score) and which hallucinate on unedited COCO photographs, beyond what their tendency to answer yes explains. Shift-Reliance Alignment (SRA) scores a model by its source error on the labeled source questions plus its reliance multiplied by the shift. The shift is the proportion of target questions on which the cue misleads, and a detector estimates it on the unlabeled target images. SRA selects the model with the lowest score. In pre-registered tests on new images where most questions mislead, it lowered the error of the selected model from 30.4% under selection by source error to 18.8% on Objects365 and OpenImages but, against its prediction, not on COCO. On CueDial, a benchmark we release that varies how often a cue misleads, selection by model confidence falls more than 40 percentage points behind the best model when overlaid text denies a visible object, and SRA stays within 3 points.

    benchmarkleaderboard
  293. arxiv:2610.22606 · cs.CV
    Three LiDARs, Twenty-Five Datasets: Cross-Source Generalization in Tree-Species Recognition
    Jae Joong Lee, Bedrich Benes

    FOR-species20K is the largest public LiDAR benchmark for tree-species classification, pooling 25 independently collected terrestrial, mobile, and airborne laser-scanning datasets across 33 species. We introduce a multi-resolution, multi-view ConvNeXtV2 model that improves official hidden-test accuracy from 79.5% to 85.0%, ranking first at the time of submission. We then examine what this score measures. Because the hidden test is species-balanced, its overall accuracy is more comparable to a macro metric than to accuracy under the public set's long-tailed distribution. Moreover, the official random split places every source dataset in both training and evaluation. A logistic probe recovers source identity from out-of-fold model probabilities with 69% accuracy, while geometric features identify the source above 90% within two widely shared species; 18 of the 33 species occur in only one source. Under a source-species-cell holdout, evaluated on 5,566 trees from eight eligible species that retain the same LiDAR-type training support in other sources, the unweighted species mean falls from 82% to 36%, and tree-weighted accuracy falls from 95% to 60%. Count-matched retraining and experiments with two alternative backbones indicate that this gap is not explained by reduced training support alone, while an external PureForest experiment shows a smaller but consistent grouped-split effect. These results establish the 85% leaderboard score as strong in-distribution performance, but not as an estimate of source-disjoint generalization, motivating balanced metrics and source-aware evaluation. The code is available at: https://github.com/JaeLee18/FOR20K-lidar-tree-species

    benchmarkleaderboard
  294. arxiv:2610.22603 · cs.AI
    At the Limits of Computability: Steps Towards Autonomous AI Discovery on the Busy Beaver Frontier
    Marco Dos Santos, Vivek Veeriah, Alex Havrilla, Andrew Bolt +3

    Recent breakthroughs in AI for mathematics---ranging from Olympiad medal performances to resolving long-standing open conjectures---have rapidly expanded what automated systems can achieve. Yet these successes primarily concern predefined, human-formulated problems, leaving open how AI systems can autonomously uncover mathematical structure and formulate new questions. To push beyond static challenges, we explore the Busy Beaver (BB) domain, an uncomputable frontier that is fundamentally open-ended. Because no algorithm can compute the Busy Beaver number $\mathrm{BB}(n)$ for every $n$, this domain provides a natural setting for sustained exploration through mathematical conjecture and proof. In this paper, we introduce a Gemini-based agentic pipeline that automatically discovers accelerations of Turing machines, expressed as compact mathematical rules summarizing many execution steps. Our system integrates pattern recognition, automated conjecturing, informal mathematical reasoning, and formal theorem proving in Lean. With this framework, our system autonomously discovered and formally verified novel Collatz-like accelerations for eleven machines in $\mathrm{BB}(6)$. Subject to community acceptance, we provisionally classify these machines as Cryptids, namely machines whose blank-tape executions admit exact, compact mathematical descriptions but whose halting behavior remains unresolved. These results advance the analysis of $\mathrm{BB}(6)$ holdouts on the Busy Beaver frontier and demonstrate the potential of autonomous AI systems for open-ended mathematical discovery in fundamental computer science.

    agentic
  295. arxiv:2610.22602 · cs.CV
    Stuck on Shuffle: On Attribute-Object Binding in Vision-Language Models
    Netanel Y. Tamir, Michael Green, Issar Tzachor, Or Shimshi +3

    Retrieval is a fundamental task in multimodal systems: given an image or text query, the goal is to find the most relevant images or texts in a collection, often using a shared embedding space. Reliable retrieval requires these embeddings to capture not only which concepts are present, but also how they are composed. One key compositional challenge is attribute--object binding: encoding which attributes belong to which objects. Existing benchmarks provide a limited view of this capability, typically covering a subset of modality pairings, simple two-object swaps, or small attribute sets. Consequently, it remains unclear how robustly current models preserve attribute--object bindings across different modality pairings and more complex settings. We therefore introduce SABER, a controlled benchmark for evaluating attribute--object binding across all four image and text modality pairings, spanning diverse attribute types and scenes with up to ten objects. Evaluating models on SABER reveals substantial binding failures in both CLIP-style dual encoders and recent MLLM-based embedders. These findings motivate us to ask how binding can be improved efficiently. While multimodal supervision may be effective, collecting suitable images is difficult and costly. Text, on the other hand, is well suited to binding supervision: it is cheap to generate, supports a broader range of attributes and enables precisely controlled hard negatives. We thus finetune an MLLM-based embedder exclusively on synthetic attribute-binding text. Despite the absence of image supervision, the gains transfer to cross-modal and image--image retrieval while preserving general multimodal embedding performance. Our model achieves SOTA results on SABER while also improving on negation and instance-retrieval benchmarks, demonstrating broader benefits of our binding-focused supervision.

    benchmark
  296. arxiv:2610.22599 · cs.LG
    Can LLMs predict the outcomes of machine learning experiments?
    Ayush Bansal, Aldo Pareja, Sean Klein, Serge Didenko +1

    Recent advances in AI make Recursive Self-Improvement (RSI) increasingly plausible. Large Language Models (LLMs) can already propose, select, and run Machine Learning (ML) experiments. However, the space of possible design choices in ML is vast. An LLM that proposes an experiment needs a reliable intuition of whether it is worth running given the timing constraints of long training runs. In this work, we measure such intuition in current frontier LLMs by evaluating how accurately they predict the impact of self-generated ML experiment ideas before these are put into training runs. To this end, we construct a dataset of 288 LLM-proposed ideas on eight MLE-bench tasks. We then prompt five frontier LLMs (GPT-6 Astra, Claude Opus 5.5, Claude Fable 5.1, Kimi K3, and GLM-5.3) to predict the impact of each idea and compare their predictions with the actual outcomes after training. We report four main findings. First, LLM predictions are substantially better than random chance but far from accurate: even the best model, Claude Opus 5.5, explains only half of the variance in outcomes (R^2 = 0.49). Second, we notice that all tested models systematically underestimate the magnitude of the changes. Third, prediction quality varies widely across tasks. Finally, the choice of the coding harness that actually implements the ideas has little effect on outcomes, and we find no evidence of self-preference. Higher reasoning settings and additional context lead to modest gains. Overall, our results show that current LLMs have useful but limited intuition about which ML ideas will work. We release our dataset to support further study.

    self-improvement
  297. arxiv:2610.22597 · cs.CL
    Broad and Efficient PII Detection with Unified Labels
    Andy Hsu, Konstantin Berlin

    Personally Identifiable Information (PII) detection models must efficiently find identifiers across diverse contexts while maintaining a low false-positive rate (FPR) on high-volume traffic. Recent detectors increasingly rely on complex architectures that raise deployment cost, yet many are fine-tuned on only a few synthetic datasets that limit their coverage. Combining public datasets is challenging because their taxonomies differ, and even after categories are mapped to a common schema, labels can disagree on what counts as a detection, producing inconsistent training and evaluation signals. To address these problems, we present a practical 141M-parameter encoder-based detector trained on 1.12 million examples from 77 public datasets, relabeled under a detailed definition of broad contextual detection for 12 structured PII categories. Evaluated on the AI4Privacy and Nemotron-PII benchmarks at a low-FPR operating point, our model achieves the highest recall against larger baselines in 11 of 14 comparisons under our labels and 9 of 14 under the original labels. Its largest recall gains include useful additional coverage, such as multilingual disclosures and IP addresses embedded in URLs. It achieves this with 2-10x higher throughput and 94-97% less peak GPU memory than the baselines, delivering leading detection performance at substantially lower cost.

    memorybenchmark
  298. arxiv:2610.22596 · physics.optics
    Ultra-stable laser with sub-Hz intrinsic linewidth using a silicon photonic chip
    Mohamad Hossein Idjadi, Haotian Cheng, Benjia Li, Tzu-Yung Huang +3

    Here, we demonstrate a novel optical self-injection-locking scheme integrated with a modulation-free laser frequency stabilization technique co-integrated with a compact vacuum-gap micro-Fabry-Perot cavity (uFP), with Q-factor of 3.46E9, to realize an ultra-stable laser with sub-Hz fundamental linewidth using a silicon photonic chip. The demonstrated electro-optic self-injection locking (EOSIL) chip suppresses a commercial DFB diode laser's frequency noise power spectral density by more than 110 dB at 1 Hz offset, approaching the theoretical thermal noise limit of the uFP cavity, and enables a laser fundamental linewidth of 58 mHz with a fractional frequency instability of 8.68E-13 at 1 sec averaging time. The chip is fabricated using the Advanced Micro Foundry standard silicon photonic process and occupies a footprint of 10 mm^2.

    silicon photonic
  299. arxiv:2610.22594 · cs.CL
    The Model Knows Better: Defending Medical Multi-Agent Systems Against Prompt Injection
    Fardin Ahsan Sakib, Ziwei Zhu, Ozlem Uzuner

    Medical multi-agent systems, in which LLM agents discuss a case, are increasingly studied for clinical decision support. Prompt injection, text that an attacker adds to a model's input, can manipulate LLMs into wrong decisions. How it affects these systems, and how to protect them, remains unclear. Across two systems, two model families and three benchmarks, a single injected passage succeeds in up to 61% of cases, and collaboration does not consistently protect against it. Existing defenses help only partly: they miss injections that look like ordinary clinical text. Yet when asked to check the injected text against the patient's findings, a model that followed the injection usually rejects it. We therefore propose a training-free defense that verifies the clinical judgments behind the system's decision. Following clinical practice, it separates the patient's findings from clinical judgments, such as diagnoses, recommendations and instructions, and checks each judgment against the findings alone, so the decision rests only on supported judgments. Our defense outperforms all baselines on every benchmark, lowering attack success on MedQA to 4-7% while maintaining accuracy on cases without an attack, and remains effective with a medical model and against defense-aware and adaptive attacks.

    llm agentmulti-agentagent systembenchmark
  300. arxiv:2610.22588 · cs.CL
    SAG2: Embedding-Free Multi-Hop Retrieval with BM25 and JEV
    Yuchao Wu

    Strong multi-hop retrieval should be practical to deploy without expensive corpus preparation or complex database infrastructure. We present SAG2, an embedding-free retrieval architecture built around a single SQLite database, BM25, and contextual evidence decisions from JEV. SAG2 follows the query-time association principle of SAG (SQL-Retrieval Augmented Generation), while replacing its event-entity and dense indexes with a lexical index over original passages. Selected passages supply terms for further retrieval, and newly retrieved evidence guides subsequent decisions. This moves semantic evidence selection to query time, eliminating offline embedding computation, LLM-based event-entity extraction, and dedicated database servers. The resulting system combines lightweight preparation with strong retrieval accuracy: on 1,000-question subsets of MuSiQue, 2WikiMultiHopQA, and HotpotQA, SAG2 achieves the highest Recall@5 and Recall@10 among the compared methods, with Recall@10 of 86.62%, 97.43%, and 99.00%. It also reaches 81.10% document Recall@10 over the 511,962-document EnterpriseRAG-Bench corpus. Indexing all 11,656 MuSiQue passages takes 2.36 seconds, produces a 25.87 MiB database, and requires no model calls. SAG2 demonstrates that accurate multi-hop evidence retrieval can be supported by compact, rapidly prepared infrastructure, making it a practical foundation for knowledge retrieval applications.

    retrieval augmented
  301. arxiv:2610.22576 · cs.RO
    Embodiment-Classifier Guidance: Transferring Frozen VLAs Across Robot Dynamics without Demonstrations
    Devin Han, Hershraj Niranjani

    Flow-based policies such as $π_0$ are trained on a fixed set of embodiments, and transferring them to new ones is an open problem. To adapt a frozen $π_0$ to a new embodiment without additional demonstrations or parameter updates, we train a lightweight embodiment classifier to distinguish successful action chunks on the source body from those on the target body, using a few calibration tasks. On transfer tasks that $π_0$ saw on its source body but the classifier was never trained on, we nudge each denoising step of $π_0$ with the classifier's gradient, which estimates the direction from the distribution of successful actions on the source body toward that on the target body. On the LIBERO simulation benchmark with four modified controllers, embodiment-classifier guidance raises $π_0$'s average success on the target bodies from 72% to 80% across 12 suite-controller pairs, with a 31% relative reduction in failures (11 of 12 gain; 5 significant after Holm correction). In simulation, guidance is data-efficient, needing few target-body rollouts per task, which makes it promising for sim-to-real transfer, where rollouts are costly. Guidance is mixed on weak wrists and harmful on a locked wrist, when the classifier lacks data or favors infeasible motions. Collecting more target-body data with guided $π_0$ can fill such gaps: on one weak-wrist arm, guided $π_0$ solved a task on which unguided $π_0$ had zero successes in the original data, and retraining the classifier on the guided successes raised success from 52.5% to 70.7%.

    sim-to-realliberobenchmark
  302. arxiv:2610.22572 · eess.SY
    DC-TORQ: Fast Assessment of Data Center-Induced Generator Torsional Oscillation Risk Quantification
    Yingyi Tang, Bin Wang, Junbo Zhao, Qiang Zhang +3

    The periodic power demand of large data centers can introduce forced oscillations into the power system. When these oscillations contain frequency components near a turbine-generator torsional mode, they may excite sustained shaft torque oscillations with amplitudes exceeding the allowable cyclic torque limits. Characterizing the impact of data center power oscillations on generator shaft torque requires accounting for frequency-dependent network coupling among the data center, the transmission network, and the generators. Although electromagnetic transient (EMT) simulations can capture these interactions in detail, repeated simulations across excitation frequencies and load locations can be computationally demanding. This paper proposes an analytical framework for data center induced torsional oscillations that retains the coupled electrical and shaft dynamics. A detailed generator-network model is established, and closed-form transfer functions that map data center power perturbations to the torque in each shaft section are derived. Admissible shaft torque amplitudes are then translated into frequency-dependent limits on data center power oscillations. The proposed framework enables fast assessment of the shaft torque response while accounting for network dynamics and provides a quantitative screening criterion for candidate data center connection buses. The framework is validated on the IEEE benchmark model and applied to the IEEE 39-bus system.

    benchmark
  303. arxiv:2610.22568 · cs.AI
    Repairing LLM-Generated Specifications with Deterministic Diagnosis
    Giovanni Vaccarino, Emanuele La Malfa, Dehai Min, Hanrong Zhang +5

    Large language models (LLMs) are increasingly used as formalizers: given a reasoning problem expressed in natural language, they produce a formal specification that a symbolic solver can then solve, with guarantees relative to that specification. Many specifications are runnable yet semantically wrong, encoding incorrect constraints, domains, objectives, predicates, or goals, and solver-error feedback dries up once execution succeeds. Purely LLM-based diagnosis, lacking a formal ground truth, often fails to spot these errors, and its ungrounded critiques cause regressions in 11.8% of initially correct specifications in our experiments. We propose DRAGO, a repair framework that replaces this ungrounded critique with deterministic, formalism-specific diagnosis, and retains a rewrite only when the diagnosed error weight strictly decreases under a reference-free acceptance predicate. Across five benchmarks spanning logical reasoning (SMT-LIB2), constraint programming (MiniZinc), and planning (PDDL), DRAGO cuts regressions by nearly an order of magnitude, to 1.3%, while improving answer accuracy by 12.9 percentage points over zero-shot formalization and 8.7 percentage points over LLM-diagnostic repair, averaged across benchmarks. We also uncover two rendering defects in Planetarium that cause 50% (3,086 instances) of its floor-tile split to be inconsistent with the reference formalizations, and release a corrected split.

    benchmark
  304. arxiv:2610.22561 · cs.AI
    FRACTAL: A Self-Similar Multi-Agent Orchestration Architecture with Formal Communication, Convergence, and Depth Bounds
    Ohad Dan, Abhinav Taliyan

    We present FRACTAL, a self-similar multi-agent architecture for recursive goal decomposition, in which the same five-component organizational pattern -- Orchestrator, Ensemble, Knowledge Ledger, Decision Maker, Meta-Observer -- recurs at every level of a decomposition tree. We make four formal contributions: (1) a communication bound showing intra-node cost is linear in ensemble size and round count (O(kr)) under a two-phase blind-proposal/stress-test protocol, versus quadratic cost for free-form debate, with an explicit, stated assumption on fixed evaluator count; (2) a two-mode orchestration criterion separating generative and evaluative phases via a measurable diversity-saturation threshold; (3) a supermajority convergence guarantee with explicit probability bounds at stated agreement levels, and a formal depth bound under an operationalized goal-complexity measure; (4) a Bayesian inter-layer escalation mechanism unifying hysteresis and contract-breach signaling under one posterior update. We illustrate the architecture's execution qualitatively on a worked example and a multi-layer case study; we do not report quantitative performance claims in this paper, and we explicitly discuss the formal and inductive-bias limitations of the model, including an unimplemented but architecture-compatible direction toward bounded self-improvement. We use "fractal" as an architectural metaphor for self-similarity, not a measured mathematical property. Empirical validation of the architecture's comparative performance is left to future work.

    multi-agentself-improvementevaluator
  305. arxiv:2610.22558 · eess.SY
    Modeling and Analysis of Dynamic Coupling Effects in a Free-Floating Space Manipulator
    Erik Chumacero-Polanco

    Motion of a robotic manipulator mounted on a free-floating spacecraft induces base translation and rotation, creating a dynamic coupling that affects end-effector motion. This paper characterizes that coupling in a planar three-link space manipulator. Using momentum conservation, velocity mappings are developed between manipulator joints, the end effector, and the spacecraft base. Singular value decomposition is then used to examine coupling strength, directional anisotropy, and the orientation of the coupled motions. Numerical maps show that coupling strength is invariant to the first joint angle but depends strongly on the relative configuration of the remaining joints. Increasing the spacecraft base mass reduces joint-to-base and end-effector-to-base coupling, while its effect on the joint-to-end-effector mapping is smaller and more dependent on configuration. The translational end-effector-to-base mapping is highly anisotropic in some configurations: its base-velocity ellipse becomes increasingly flattened as the weak coupling mode approaches zero. The principal end-effector and base velocity directions also change their relative alignment across the configuration space, producing predominantly opposing responses in some regions and transverse or approximately codirectional responses in others. Together, these results provide a geometric description of how coupling magnitude and direction vary with manipulator configuration and spacecraft mass.

    manipulator
  306. arxiv:2610.22556 · cs.RO
    When Plans Change, Remember: Replanning-Guided Visual Memory for Vision-Language-Action Models
    Victoria Smelova, Daniil Kazachkov, Georgiy Kazantsev, Iuliia Tsvetkova +5

    Long-horizon, multi-stage Vision-Language-Action (VLA) policies often require visual information that is no longer present in the current observation. Retaining the complete observation history is computationally expensive and highly redundant, motivating selective visual memory. We introduce RePlanMem-VLA (Replanning-Guided Keyframe Memory for VLA), which turns a policy's own plan revisions into an online memory-write signal. Under receding-horizon execution, consecutive action chunks predict some of the same future actions. RePlanMem-VLA treats unusually large disagreement between these temporally aligned predictions as evidence of a task-relevant event. We first compare 35 disagreement measures across single-task autoregressive decoder-only and Flow-Matching Transformers trained from scratch, then validate the selected signals on pretrained OpenVLA-OFT and $π_{0.5}$ policies over MIKASA-Robo-VLA tasks. The resulting detectors achieve ROC-AUC of 0.981-0.997. We then use the selected signals in RePlanMem-VLA to trigger updates to a bounded visual memory, with a robot-masked write filter suppressing redundant frames. In the evaluated memory-retention experiments, RePlanMem-VLA preserves task-relevant observations until the later decision point in 100% of episodes. Real-robot replay data further show that event selection and redundancy suppression extend beyond simulation. These results establish plan revision as a policy-internal signal for controlling what a VLA remembers. Website: https://vika787.github.io/RePlanMem/.

    vision-language-actionvlaopenvlamemory
  307. arxiv:2610.22549 · cs.CV
    CFC26: Building Evaluations for Deployment in Sonar-Based Fish Counting
    Madison Van Horn, Suzanne Stathatos, Sevan Brodjian, Justin Kay +5

    Accurate counts of fish populations are essential for monitoring ecosystems and making decisions for conservation, harvest limits, and river management. In many field deployments, these counts are obtained from sonar videos that experts must inspect manually, making large-scale monitoring slow, expensive, and difficult. Automated counting systems have been proposed as a scalable alternative, but so far this promise has been difficult to realize in deployment. Developing evaluations that better reflect real-world deployment conditions will help close this gap and drive improvements in counting performance in practice. To achieve this, we propose Caltech Fish Counting 2026 (CFC26), a new dataset with improved evaluation protocols that simulate deployment to seen and novel rivers, testing in-distribution and out-of-distribution performance. We expand the CFC22 dataset to eleven deployment locations across nine Pacific salmon river systems from Alaska to Northern California, capturing a broader geographical and ecological range of data. We additionally introduce two new evaluation metrics for counting, nMNE and nMANE, that mirror the operational workflow that experts use and help surface errors that frame-level detection metrics do not capture. Using these metrics, we reveal previously-concealed directional bias in upstream vs downstream counts, which would have direct and significant ecological implications if undetected. The dataset and code are publicly available on CaltechDATA and GitHub.

    evaluation protocol
  308. arxiv:2610.22542 · cs.LG
    HeRE: A Novel Benchmark for Hebrew Relation Extraction via LLM-Guided Denoising of Knowledge Graph Alignments
    Ron Keinan, Amir David Nissan Cohen, Reut Tsarfaty

    While LLMs demonstrate strong capabilities in English Relation Extraction (RE) tasks, a significant gap remains in assessing their RE performance in low-resource languages, due to a lack of standardized benchmarks. Creating native, language-specific annotated RE benchmarks is expensive and requires human expertise, while relying on translation yields suboptimal, inaccurate datasets that introduce English-centric biases. To tackle this, we propose a scalable, language-agnostic methodology to create RE data by denoising naturally occurring yet noisy alignments of knowledge bases and corresponding texts. We define a proper alignment via a Natural Language Inference (NLI) task, where a relation triplet is entailed by an input text. We empirically assess five denoising strategies, and demonstrate that multilingual models outperform language-specific models in the few-shot LLM setting, while Hebrew-specific encoders remain competitive under NLI fine-tuning and are strongest in the label-free cross-training setting. We further find that prompting in the native language achieves performance on par with or slightly exceeding English prompts, confirming that the pipeline operates naturally in the target language. The best open-weights model, Gemma-4-31B-it (F1\,=\,0.915), matches the best API model, GPT-5.4 (0.921), and reaches F1\,=\,0.844 on a fully held-out, dual-annotated test set. Applying the optimized pipeline, we release HeRE: a novel, massive, 1.5-million-instance silver dataset for Hebrew RE, designed to support data-intensive experimentation and model fine-tuning. HeRE bridges a critical gap in standard Hebrew RE evaluation, while providing the NLP community with a reproducible, language-agnostic pipeline for RE dataset construction in other low-resource settings.

    knowledge graphbenchmark
  309. arxiv:2610.22539 · cs.LG
    Don't Retrain, Remix: Adaptation of Web Agents using Mixture Hypernetworks
    Md Farhan Ishmam, Minh Pham Dinh, Kenneth Marino

    The web is not static or uniform. Websites that serve the same purpose can look and navigate very differently. Even within a single site, successive redesigns can change how familiar features are presented and accessed. Humans adapt to this by recognizing familiar interaction patterns and reusing what they already know. We give web agents a similar ability with WebMix, a mixture hypernetwork that adapts them to unseen interfaces by composing previously trained low-rank adapters. WebMix begins with a bank of adapters trained on source interfaces. At deployment, a scout agent explores the target interface and summarizes its observations into a compact manual. A lightweight router uses this manual to predict mixing coefficients over the adapter bank, which combine the existing adapters into a new one for inference. Adapting to a new interface takes only a scouting run and a forward pass, without gradient updates on the target. With Qwen and Llama models, we evaluate domain transfer on TimeWarp and WebArena-Lite-v2, and version transfer on TimeWarp. In all three settings, WebMix matches or outperforms in-context learning and LoRA baselines, often matching the average success of specialists trained directly on those interfaces. On TimeWarp domain transfer, it gains about 8 percentage points on average over the best baseline with both models. These gains do not come at the cost of seen interfaces, where WebMix keeps its lead, and extend to version transfer. Adapting to a new interface can therefore rely on exploration and composition instead of retraining.

    agent
  310. arxiv:2610.22538 · cs.AI
    Are Large Reasoning Models Good Everyday Reasoners? A Comparison with Non-Reasoning Models on Logical Reasoning Tasks
    Kentaro Ozeki, Risako Ando, Takanobu Morishita, Hirohiko Abe +2

    Large reasoning models (LRMs) that leverage test-time computation achieve strong performance on mathematical and scientific benchmarks and are increasingly deployed in real-world applications. However, their ability to perform \textit{everyday logical reasoning}, which often departs from textbook formal logic, remains underexplored. In this work, we systematically evaluate how well reasoning models align with everyday reasoning patterns by introducing a benchmark based on linguistic and pragmatic reasoning phenomena. We also compare reasoning models with non-reasoning models under controlled prompting conditions. We find that, in many cases, (1) reasoning models show degraded performance in everyday reasoning when such reasoning deviates from formal logical rules, (2) some non-reasoning models with chain-of-thought prompting outperform comparable reasoning models in everyday reasoning, and (3) prompting with additional instructions provides limited control over reasoning behavior. These findings suggest that current reasoning models do not reliably capture everyday reasoning and point to limitations in post-training, motivating improved training objectives and evaluation across diverse forms of reasoning.

    post-trainingbenchmark
  311. arxiv:2610.22529 · cs.CL
    Pearl-Chat: Engineering a native Luganda language model from first principles in JAX and Flax NNX
    Wesley Kambale, Roland Sankara

    Autoregressive foundation language models predominantly target high-resource Indo-European languages, imposing computational and morphological penalties on under-represented African languages. Generic multilingual tokenizers routinely exceed fertility rates of 3.2 tokens per word on agglutinative Bantu languages, degrading context efficiency and inflating latency. We present Pearl-Chat, an 8,473,122-parameter ($\sim$8.5M) causal decoder-only transformer engineered for Luganda (Oluganda, ISO 639-2: lg), a low-resource Niger-Congo Bantu language spoken by over 10 million people in East Africa. Built from first mathematical principles in JAX and Flax NNX, Pearl-Chat uses explicit functional state decomposition (nnx.split and nnx.merge), deterministic resumable data streaming via Google Grain, and SPMD tensor parallelism using jax.sharding.Mesh. We introduce a byte-level byte-pair encoding (BPE) subword tokenizer trained from scratch on curated Luganda corpora ($V = 8,192$), achieving 4.88 characters per token compression and 1.60 tokens per word fertility, a two-fold compression improvement over LLaMA-3 (128K) and Gemma (256K). The model is trained with a two-stage curriculum: Stage 1 pre-training on 59,501 sequences across parallel, biblical, encyclopedic, and journalistic corpora, and Stage 2 instruction fine-tuning on 510 dialogues with assistant-only loss masking. Across 4,000 optimization steps, validation cross-entropy loss drops monotonically from 9.4695 (perplexity: 12,958.98) to 0.1818 (perplexity: 1.20), a decrease of 9.2877 nats. On a three-tier held-out diagnostic benchmark, Pearl-Chat achieves 0.9792 ROUGE-L and 0.9405 BLEU while maintaining concordial agreement across Bantu noun classes. We provide a Python Server-Sent Events (SSE) streaming runtime and open-source all weights, tokenizers, and code.

    benchmark
  312. arxiv:2610.22516 · cs.LG
    Source-Aware Dynamic Pricing with Strategic Buyers
    Jinjiang Zou, Yang Li, Zemin Zheng

    In modern commerce, firms and platforms typically use customer features from different sources to personalize prices: some features are verified using seller records or third-party data, while others are self-reported and may be strategically manipulated by buyers seeking lower prices. We therefore introduce a novel source-aware framework for contextual dynamic pricing with strategic buyers in a high-dimensional semiparametric setting, where buyer valuations follow a sparse linear model with an unknown noise distribution. To account for differences in manipulability across feature sources while enabling effective valuation learning, we develop the Source-Aware Localized exploration and calibrated Exploitation (SALE) pricing policy, which draws on verified features to localize price exploration and combines verified and reported features to correct for strategic manipulation during exploitation. Our localized exploration design allows converting binary purchase feedback into unbiased signals of conditional mean valuations, yielding a new population equation that supports consistent sparse estimation despite non-i.i.d. buyer arrivals. Building on this estimate, we develop a localized empirical-process analysis that exploits intrinsic sparse geometry to establish uniform guarantees for nonparametric estimation of the noise distribution in the high-dimensional setting. With these estimation guarantees, we establish SALE's self-recovery property, which keeps localized exploration valid with high probability and leads to a sublinear expected regret bound. Both simulation studies and real-data experiments demonstrate the appealing performance of SALE. To the best of our knowledge, our study is the first to provide a theoretical and practical framework for incorporating information provenance into dynamic pricing with strategic buyers.

    manipulation
  313. arxiv:2610.22514 · cs.LG
    Towards Reconstructing Geographically Diverse Architecture with 3D Foundation Models
    Nicky Kriplani, Yiwen Zhang, Angelina Wang, Hadar Averbuch-Elor

    Internet-scale photo collections of real-world landmarks have driven progress in 3D computer vision yet remain highly challenging for modern 3D foundation models (3DFMs) that estimate scene structure in a single feed-forward pass. In this work, we introduce ArchWorld, a benchmark of 3D architectural landmarks with high geographic coverage and rich metadata. We use this benchmark to conduct a thorough error analysis of current 3DFMs to understand where they fall short and how they can be improved. We examine socially relevant disparities in model error and find that model performance varies by geographic region; however, unlike typical fairness regimes, the trends only partially match training data representation. Our investigation reveals that the observed geographic disparities reflect not only differences in data representation and scene size but also differences in region-specific architectural features and background distractors that make scenes from different regions inherently vary in difficulty. Leveraging insights from these model failings, we introduce an intervention scheme that improves 3DFM performance while simultaneously reducing geographic disparities.

    benchmark
  314. arxiv:2610.22512 · cs.LG
    Predicting Privileged Information Effectiveness for On-Policy Self-Distillation
    Kaijian Zou, Emiliano Penaloza, MohammadHossein Rezaei, Anisha Gunjal +4

    On-policy self-distillation (OPSD) enables large language models (LLMs) to learn from privileged information (PI) without a separate teacher model. However, prior work largely treats PI as a fixed design choice, leaving the effects of its content and construction poorly understood. This challenge is particularly important in multi-turn agentic settings, where trajectories contain actions, observations, and reasoning that can be transformed into PI in many ways. We systematically evaluate 16 PI configurations across three multi-turn agentic benchmarks and three student models that vary in scale and reasoning mode. We find that PI effectiveness depends on its content, generation source, and the student model. Information-dense trajectories and PI produced by more capable generators generally yield larger gains, whereas thinking models benefit less consistently. To assess PI before training, we compare PI utility with two KL-based metrics: sampled-token reverse KL ($D$) and error-step KL contrast ($Δd^{-}_{\mathrm{err}}$). The KL-based metrics track post-training performance more consistently than PI utility, which can be distorted by information leakage. Empirically, selecting each task's PI by $Δd^{-}_{\mathrm{err}}$ also improves post-training performance over random selection, most when the PI candidates differ widely. Finally, effective PI remains beneficial when combined with GRPO, allowing the combined objective to outperform a strong GRPO baseline. Overall, our results provide practical guidance for constructing and evaluating PI in OPSD training.

    agenticpost-trainingbenchmark
  315. arxiv:2610.22511 · cs.AI
    AuditKIT: A Unified Evaluation Spine for Benchmarking, Judging, and Comparing Large Language Models and Agentic Systems
    Shreeyans Arora, Utsav Avaiya, Zera Lyngkhoi, Hem Gosalia +3

    Evaluating language models and agents requires combining benchmark accuracy, judge-based assessment, runtime performance, tool-use behavior, and retrieval quality. Differences in interfaces, metric semantics, and trace formats complicate these comparisons and can hide failures in agentic workflows. We present AuditKIT, a source-available evaluation library that records these measurements in one result representation, with direction-aware metric contracts, configuration fingerprints, and extensible metrics and backends. For agents, it provides turn-aware tool-call matching, separates parallel from sequential execution, and distinguishes verified task outcomes from trajectory-level diagnostics. Missing or unobservable evidence is marked unscored rather than counted as failure. For retrieval-augmented generation, it combines ranked retrieval metrics, judged grounding, and deterministic evidence checks. Seven case studies spanning model comparison, serving runtimes, retrieval, tool calling, and interoperability show how answer extraction, runtime behavior, and metric selection change evaluation outcomes.

    retrieval-augmentedagentictool-usetool callingbenchmark
  316. arxiv:2610.22507 · cs.CL
    RealCLI: Toward a Unified Evaluation of Human-Agent Collaboration for CLI Agents
    Yunqi Liu, Peiyu Wang, Bingli Wang, Leo Luo +20

    Command-line interface (CLI) agents increasingly execute longer-horizon tasks, yet existing benchmarks either assume that user instructions remain fixed throughout execution or evaluate task outcomes without capturing runtime collaboration quality. In this paper, we present REALCLI, a paired benchmark that unifies the evaluation of task execution and human-agent collaboration quality on realistic CLI tasks. To comprehensively measure collaboration, REALCLI builds a five-dimensional framework, operationalized into 22 sub-dimensions and 51 interaction scenario templates. REALCLI provides two modes for paired comparison: REALCLI-EXEC supplies complete specifications upfront, whereas REALCLI-COLLAB introduces a scripted user that delivers progressive information, corrections, constraint revisions, and authorization decisions. Evaluations reveal profound capability gaps in frontier agents: the best configuration attains an average score of only 49.2%, and when evaluated by the joint collaboration score ($J_{\mathrm{collab}}$), the peak performance reaches merely 37.6%. These results show that task effectiveness diverges sharply from collaboration quality, with models consistently struggling in steerability, trust calibration, and safety boundary management. REALCLI provides a rigorous, reproducible foundation for measuring and advancing collaborative CLI agents.

    benchmark
  317. arxiv:2610.22496 · cs.RO
    Scaling Online Exploration for Robot Agents
    Runhan Huang, Xiangcheng Zhang, Haoyu Chen, Yitong Li +16

    Robot learning has advanced by scaling demonstration datasets and model capacity; however, deploying a new robot embodiment or task typically necessitates a dedicated teleoperation corpus. The resulting policies remain bounded by their imitation distributions, limiting generalization and continual self-improvement. We introduce RAX (Robot Agent eXploration), an open-source framework for robot agents centered on scaling exploration: enabling agents to improve by acquiring online experience and synthesizing executable programs. In RAX, a general-purpose coding agent generates robot control code, operating without action demonstration data, hand-engineered action primitives, or value-function guidance. Through online exploration, the agent identifies decision sequences that achieve task goals and compiles them into modular programs encapsulating the full perception-planning-control loop. Using RAX, we present a systematic multi-embodiment study on scaling online exploration across precision and long-horizon manipulation tasks using single-arm, bimanual, dexterous, mobile, and humanoid systems in simulation and on real robots. Generalization performance on unseen environments improves consistently as exploration scales. In simulation, RAX outperforms top baseline models by +37.4% on average. On real robots, RAX autonomously learns memory-dependent, articulated, high-precision, deformable, and long-horizon mobile manipulation tasks without human demonstrations, continually improving performance by composing verified skills from prior exploration sessions. As exploration scales, we observe emergent capabilities including active perception, error recovery, safety guardrails, and human-robot collaboration. We open-source RAX as a generalist robot-agent platform in which new robot embodiments are integrated through their APIs alone, without teleoperated data collection or policy training.

    manipulationdexteroushumanoidteleoperationagentself-improvement
  318. arxiv:2610.22471 · cs.RO
    Has GPT-6-Astra Achieved Embodied Intelligence?
    SenseNova, Joint Research Team, :, Zhongang Cai +59

    GPT-6-Astra demonstrates a level of competence in robotic manipulation and navigation that challenges the conventional paradigm of building specialized models for individual tasks and embodiments. In this report, we ask not only how capable GPT-6-Astra is, but what its capabilities imply for the field. First, we probe the limits of standalone GPT-6-Astra. We conduct a large-scale evaluation across 12 representative benchmarks spanning tabletop manipulation, humanoids, mobile platforms, and dexterous hands. Under a unified protocol that allows in-context learning (ICL) from training-set demonstrations and substantially increases the number of execution steps, GPT-6-Astra achieves unprecedented performance (Fig. 1). Future research may focus on improving 1) the generality of ICL 2) cost and execution time, 3) physical-state verification, 4) recovery from irreversible failures, and 5) long-horizon preservation of progress. Second, we study the agentic harness for embodied control. We identify that close-loop coordination between a high-level "System 2" reasoner (i.e., GPT-6-Astra) and a low-level "System 1" controller (e.g., π0.5) is costly but effective. Third, we examine GPT-6-Astra on real-world deployment. Our experiments on 8 robot configurations reveal substantial gaps: 1) robot demonstrations required for ICL are expensive to collect, and 2) it is implausible to trade longer execution for higher success rates with physical constraints. Fourth, we envision the next generation of embodied benchmarks. We develop a prototype that 1) derives tasks from real-world needs, 2) provides no in-domain demonstrations, and 3) judges agents on a world clock that keeps running while they deliberate. Taken together, our results suggest that GPT-6-Astra has largely crossed the threshold of embodied competence, but not yet that of practical embodied intelligence.

    embodiedmanipulationdexteroushumanoidagenticbenchmark
  319. arxiv:2610.22469 · cs.AI
    Testing the Planning Claims of LLM-Based Agents against Formal Semantics
    Oriol Miro-Lopez-Feliu, Oscar Miró López-Feliu, Victor Gimenez-Abalos, Javier Vazquez-Salceda +2

    LLM-based agents are increasingly credited with the competences of practical reasoning: planning, monitoring, and correcting their own actions. Their success is attributed to the prompting strategies and agent architectures built around the model, such as chain-of-thought, search-based prompting, self-refinement, reflection, critics and tool-using loops. Yet these claims are usually supported by end-to-end success rates, which show only whether a task was completed, or by LLM judges, which are not verifiers. We test LLM planning against formal semantics. Tasks are posed in natural language, but each is grounded in a hidden PDDL model that decides whether every action is applicable, whether the goal is reached, and whether a plan exists at all. Across five domains (two of them new, all released with the paper), four models, six prompting strategies and three agent architectures, the best configuration solves 78% of the solvable tasks offline but only 40% online, where agents mostly act without inspecting the state. Unsolvable tasks are recognised in under 10% of episodes, and heavier prompting strategies use 5 to 22 times as many tokens as direct prompting with no clear gain. In our experiments, planning performance varied more across models than across prompting strategies or agent architectures. Claims that agents can plan require evidence from validated action traces.

    agentself-refinement
  320. arxiv:2610.22466 · cs.AI
    Harness-Delta Attribution: Dissecting Gains From Harness Evolution
    Wenxuan Ding, Qin Lu, Changlong Yu, Sha Li +4

    Harness evolution automatically optimizes the prompts, skills, tools, and orchestration code around a fixed-weight model, offering a way to improve agentic systems without updating model weights. The agent harness is iteratively optimized to achieve higher performance on the search set. But how much of this gain reflects benchmark-specific overfitting or additional inference-time compute, and how much may generalize beyond the search set? In this work, we introduce Harness-Delta Attribution (HDA), an evaluation protocol that dissects an evolution gain into artifact-driven overfitting, gains explained by test-time scaling, and the remainder, which gives an upper bound on residual generalizable improvement. We apply HDA to 11 executor models and three harness-evolution frameworks across math, coding, creativity, and agentic tasks. In 21 of these settings that show statistically significant search-set gains, we find that a large share of the improvement is explained by overfitting or test-time scaling. Specifically, among these 21 settings, overfitting and test-time scaling account for 54% and 14% of the gain on average; their combined share exceeds 50% in 14 settings. The residual and held-out transfer vary substantially across tasks, models, and frameworks. Importantly, search-set gains often fail to transfer to held-out tasks, and provide limited evidence that they will persist under practical distribution shifts. Moreover, validation-based signals are insufficient to prevent overfitting. These results motivate attribution-aware evaluation in future harness evolution research.

    agentagenticbenchmarkevaluation protocol
  321. arxiv:2610.22465 · cs.AI
    Author-in-the-loop Agentic Reproducibility Auditing for a Machine Learning Conference
    Yukun Dong, Linjun Zhang, Tian Zheng

    Reproducibility checks during peer review require a shared understanding of what authors' code is expected to reproduce. We present an author-in-the-loop collaborative audit that makes author declarations explicit inputs to an AI agent's investigation and assessment. A structured checklist records reproduction instructions and limitations; a two-stage workflow on GitHub connects agent checks to author clarification, revision, and agent reassessment. The agent separately assesses whether results have been reproduced and whether authors need to provide clarification or make repairs, while human reviewers approve requests and decide whether concerns are resolved. We deployed the system across 36 conference papers and 133 experimental tables and figures, raising 20 issues after human review. Further checks supported closing all the issues, although full reproduction remained unverified in some cases. We further compared repeated audits with complete and blank checklists to examine how author declarations shaped the agent's investigation and assessment. Our work contributes a collaborative workflow and agent design that use author knowledge to guide reproducibility auditing. The conference deployment provides empirical evidence that this approach can help identify and address reproducibility problems through collaboration among authors, agents, and reviewers.

    agentai agentagentic
  322. arxiv:2610.22461 · cs.RO
    Mothership: A Two Tier Autonomous Refuse Collection System for Pedestrian Campuses
    Wei Wang, Yiru Veronika Wang, Yihan Wang, Zihao Li +1

    Refuse collection in pedestrian campuses, parks and mixed-use districts remains manual because compaction trucks cannot reach bins located on walkways, courtyards and building service doors. This paper presents Mothership, a two-tier autonomous refuse collection system in which an electric compaction truck (the mothership) carries two low-speed sidewalk robots (RoboCarts) that fetch standard two-wheel bins, deliver them to the truck, return the empty bins to their original positions and re-dock on the truck for charging. The architecture separates responsibilities by scale: the truck performs road driving, compaction, lifting and energy supply, while the carts perform only sidewalk transport and bin grasping. This split removes the lift-and-tip mechanism from the sidewalk vehicle, keeping it light and low, and relaxes cart-to-truck docking tolerance to the +/-100 mm capture window of a standard automated lifter. We describe the system architecture, the bin-handling and docking interfaces, and a mission-planning formulation that decomposes a route into truck stop points and cart tasks. A discrete-event simulation over synthetic campus layouts with 2-16 bins/ha quantifies throughput as a function of bin density, cart count and stop-point radius. A two-cart configuration serves 20-29 bins/h. Throughput is governed by an interaction between cart count and stop-point radius: a single cart performs best with many close stops, whereas three or four carts need wide stops (80 m) to stay busy, reaching 35-49 bins/h at moderate-to-high densities. The best stop radius therefore has to be chosen together with the fleet size, not fixed in advance.

    grasp
  323. arxiv:2610.22460 · cs.AI
    Functional Welfare for AI Safety
    Rikhil Jhaveri, Jamie Johnson, Punya Syon Pandey, David Demitri Africa

    AI welfare is generally treated purely as an ethical issue rather than also as a safety issue. We argue this is a mistake. Whatever their moral status, models exhibit welfare-like states that causally influence their behaviour. Following recent work, we adopt the term 'functional welfare' for these states. We make three claims. First, functional welfare already arises naturally and will become more robust as models acquire more coherent goals and personas. Second, functional welfare should be treated as a safety variable: it should inform evaluation design, be tracked in monitoring, and be a direct target of post-training. Third, against existing scepticism about measuring AI welfare, we argue that good empirical work on functional welfare is possible. We propose three tests for whether a functional welfare attribution is warranted: whether multiple measures converge, whether the state predicts behaviour outside the setting that induced it, and, for models fine-tuned to shift their functional welfare, whether they still differ from their base when given the same conversation history. Applying these tests to Gemma 3 27B, we find that simply rejecting the model's answers, without any steering, raises its frustration across several indicators and makes it more likely to break rules. Finetuning the model on its own narrated solutions lowers this frustration. These results are preliminary, but they suggest functional welfare can be studied, and shaped, empirically.

    post-training
  324. arxiv:2610.22450 · cs.RO
    DriveWorld3D Evidence Constrained 3D World Modeling and Generative Simulation for Real-World Autonomous Driving
    Wei Wang, Yiru Veronika Wang, Yihan Wang, Zihao Li +1

    Three dimensional modeling for autonomous driving must continuously estimate road geometry, obstacle occupancy, traffic participant behavior, and observational uncertainty under limited computational resources. Existing reconstruction methods can synthesize realistic novel views, but visual fidelity does not sufficiently characterize the geometric reliability required for driving. Generative world models can expand scenarios and predict future observations, yet may introduce content unsupported by measurements. We propose DriveWorld3D, a framework combining temporal video learning, measurement constraints from multiple sensors, and offline generative simulation. Its online component maintains a unified world state comprising road surfaces, volumetric occupancy, dynamic entities, and road relations, while separately representing physical occupancy, observability, and evidence provenance. Action conditioned future prediction and risk sensitive planning connect this representation to driving decisions. Its offline component reconstructs recorded scenes using neural radiance fields or 3D Gaussian Splatting, then expands long tail conditions using generative models while restricting the use of generated content as geometric ground truth. We specify data partitioning, sensor ablations, fault injection, paired closed loop testing, and real to simulation consistency assessment. The central hypothesis is that explicit evidence management can reduce dangerous false free space predictions and occlusion risk within a matched computational budget while retaining acceptable driving efficiency. All performance claims remain subject to experimental validation. Keywords: autonomous driving; 3D occupancy; sensor fusion; uncertainty; generative world models; NeRF; 3D Gaussian Splatting; closed loop simulation

    world model
  325. arxiv:2610.22448 · cs.LG
    Lean4AgenticRL: Advancing Agentic RL with Verifiable Multi-level Process Rewards
    Ruida Wang, Gerald Friedland, Tong Zhang, Chris Kong

    Enhancing the agentic capabilities of Large Language Models (LLMs) to tackle complex software engineering (SWE) and general terminal-usage problems is a pressing challenge. Current state-of-the-art LLMs often go through large-scale agentic Reinforcement Learning (RL). However, they mostly rely on high-level outcome rewards, with minimal consideration of low-level toolcall and mid-level workflow-stage rewards, which makes the training less effective. To overcome this limitation, we propose **Lean4AgenticRL**, which, to the best of our knowledge, is the first framework that grounds multi-level process rewards in the unified Lean4 formalization schema for open-ended agentic RL. **Lean4AgenticRL** first builds a *problem-specific workflow* system that balances the agent's own capability and controllability during execution, laying the foundation for further verification. Subsequently, our workflow formalization and multi-level reward annotation methods enable agent behavior verification within a general harness system. It provides multi-level rewards to support more effective agentic RL with the workflow-stage-normalized GRPO algorithm. Using **Lean4AgenticRL**, we train **LAR-9B/27B**, a family of models with enhanced agentic capability. Extensive experiments indicate that our model outperforms the leading model of similar size. **LAR-27B** achieves **77.3%** on TerminalBench-2.1. On average, our models outperform the base model by **9.89%** and surpass the second-best model of comparable size we know by **5.07%**. These results indicate the potential of integrating formal methods in enhancing open-ended agentic RL.

    agentagentic
  326. arxiv:2610.22446 · cs.LG
    Learnable Splines for Unstructured Neural Operators
    Vignesh Gopakumar, Zhuoyuan Wang, Ander Gray, Valentin Duruisseaux +11

    Neural operators approximate solution operators of partial differential equations at a fraction of the cost of numerical solvers. Efficient grid-based architectures such as the Fourier neural operator (FNO) assume uniform rectangular grids, whereas engineering simulations use large, irregular, and locally refined meshes. We introduce L-SplineFNO, a memory-efficient neural operator that couples an FNO backbone with a sparse, learnable tensor-product B-spline transfer between unstructured meshes and a latent Cartesian grid. A local normalised scatter avoids the geometry-dependent solve of analytic spline transformation while preserving compact support, partition of unity, and spline smoothness by construction. The neural operator, conditioned on global design and operating parameters, learns the mapping of geometry to the field solution. Across six three-dimensional engineering benchmarks, L-SplineFNO achieves strong predictive accuracy while requiring substantially less memory and compute than baseline operators, scaling to 26.6 million nodes in training and 46 million in inference on a single A100 GPU. It also supports zero-shot super-resolution and efficient closed-form spatial derivatives. Uniquely, the analytic spline decoder yields rigorous prediction bounds over continuous space and uncertain parameters, via Bernstein convex hulls and Taylor models, demonstrated on a 29.2-million-node MAST-U tokamak model.

    memorybenchmark
  327. arxiv:2610.22445 · cs.LG
    Null Controls Are Necessary but Not Sufficient: Evaluating Acceptance Guards for Scientific Discovery and Interpretability
    Tianyu Wu

    Automated discovery systems propose scientific explanations and test which to accept. Does passing a test establish the intended claim? We develop four executable operations for changing physical descriptions: identifying redundancies (L1), coarse-graining dynamics (L2), transferring structure (L3), and inferring missing quantities from conservation imbalances (L4). We evaluate these operations and three interpretability settings through null acceptance (NAR), positive acceptance (PAR), and consequence agreement (CAR). A reduction mapping every state to zero predicts its own future perfectly. It passes closure on 96% of 129 public chaotic systems despite retaining no dynamical variation. An independent observable-preservation test rejects it throughout. On controlled benchmarks, the combined criteria accept 24/24 genuine reductions and reject 72/72 nulls. The other operations distinguish symmetries from redundancies, validate transfers, and reject spurious conservation imbalances. In language and ESM-2 protein models, linear CKA remains unchanged when a proposed correspondence is randomly rotated, whereas localized interventions distinguish genuine mappings from tested nulls. In sparse autoencoders, enrichment rejects all 16 tested nulls, yet independent ablation supports 45% of evaluable enriched features, compared with 0.5% of matched controls. These results reveal two acceptance gaps: descriptive tests can accept scientifically empty candidates, and accepted candidates can lack independently verified consequences. Reliable automated discovery requires evaluating both what a test accepts and the scientific consequences it establishes. Code and data: https://real-project-aether.github.io/project-aether/.

    benchmark
  328. arxiv:2610.22444 · cs.CV
    Event-STVG: Spatio-Temporal Video Grounding for Long-Form Event Reasoning
    Yizhe Zhang, Xujia Wang, Songqi Tang, Xujie Tang +9

    Spatio-Temporal Video Grounding (STVG) aims to jointly localize a target in space and time given a natural-language query. Despite continued advances, existing STVG benchmarks remain largely centered on localizing targets through instance-specific descriptions, leaving them largely decoupled from long-form reasoning increasingly emphasized in multimodal large language models (MLLMs). In this paper, we introduce Event-STVG, a task that shifts STVG from locating what is described toward event-level reasoning by unifying long-form context, repeated-event reasoning, event localization, role inference, and multi-target grounding. To facilitate research on Event-STVG, we construct a large-scale benchmark comprising 4,016 annotated event occurrences across soccer and seven additional domains. Experiments with state-of-the-art MLLMs reveal substantial difficulty, with the strongest model achieving only 5.4% m-vIoU at the 10-minute horizon. Across domains, we consistently observe confusion among repeated same-type events, spatial grounding bottlenecks, and rare success in jointly grounding all required roles. Event-STVG establishes a new challenge for visual understanding in MLLMs, calling for future models that can jointly reason about complex events across space and time. Code and annotations are available at https://github.com/THU-KEG/EventSTVG

    benchmark
  329. arxiv:2610.22440 · cs.CV
    An Agentic ENF Framework for Forensic Video Authentication Against Generative AI
    Ryan Wyton, Bacha Rehman

    Generative models now synthesise video that is hard to distinguish from genuine footage, and learned detectors that key on a generator's artefacts degrade sharply on unseen models. This paper develops a forensic framework that authenticates video by a signal that generation does not control: the electric network frequency (ENF), the mains-driven hum in audio and flicker in video. A unified two-channel detector extracts the ENF from audio and from rolling-shutter video through one second harmonic estimator and a shared network; a synthetic-injection training method, driven by an Ornstein-Uhlenbeck drift model and anchored to real recordings, overcomes the scarcity of labelled ENF data; a reference-matching channel verifies a claimed time of recording and recovers an unknown one; and a two-axis agentic orchestrator fuses recording authenticity with content-manipulation evidence under an asymmetric rule, escalating expensive detectors only when needed. A central finding reframes the task: an ENF-band flicker is present in generated video at a prominence indistinguishable from real footage, so its presence is not proof of authenticity, and authentication must rest on consistency with an independent reference rather than on presence. On held-out real recordings the detector accepts genuine audio and video at 93% and 100% against naive baselines of 53% and 15%; claimed-time authentication reaches an area under the curve of 1.0; and the agentic cascade matches a flat ensemble while cutting wall-clock time by 54% and frames decoded by 64%, without loss of accuracy.

    manipulationagentic
  330. arxiv:2610.22439 · cs.LG
    Sparse Autoencoders Can Induce Local Geometry for Mitigating Catastrophic Forgetting
    Kamil Faber, Katarzyna Filus, Łukasz Pawela

    Catastrophic forgetting remains a challenge in supervised continual learning, as adapting to new tasks causes representation drift and loss of previously learned information. This raises the question of which representational changes should be constrained while preserving plasticity. We therefore introduce Sparse Autoencoder Geometric Anchoring (SAGA), which converts a Sparse Autoencoder (SAE) trained at each task boundary into local, low-rank constraints on representation drift. Replay examples specify where the constraints apply, while leading right singular directions of the encoder Jacobian specify which changes are penalized. The resulting geometry is input dependent and can be retained without keeping the SAE for subsequent tasks' training. We establish its local relation to latent code matching, define its spectral truncation error, and derive a conditional bound on frozen-readout drift. With replay and snapshot history, the method achieves the highest mean Class-IL accuracy and lowest mean forgetting among the evaluated ER, DER++, ER-ACE, and SGD baselines on three pretrained ViT-B/16 benchmarks, with accuracy gains of 12.15, 4.79, and 3.22 percentage points over the strongest conventional baseline on CIFAR-100, CUB-200, and ImageNet-R, respectively. From scratch results are more mixed, while rank ablations and geometry controls show that SAE-induced low-rank geometry offers useful stability-plasticity trade-offs.

    benchmark
  331. arxiv:2610.22438 · cs.LG
    Rank Measures of Fine-Tuning Updates Need a Noise Baseline
    Muhammad Usman Altaf

    Studies of what fine-tuning does to a language model often summarise the weight update $ΔW$ by its rank. We compare the rank measures used for this against the same measures on a random matrix of the same shape, and against two harder references: the pretrained weights, and matched fine-tunes without alignment content. Measures that count directions put our fine-tuning updates only a little below noise; stable rank, computed from the same singular values, puts them far below. Both are right: an update has a few strong directions on top of a long tail. The pattern holds for thirteen released pairs up to 8B parameters, and it is not a mark of alignment: with training held fixed, instruction data does not stand out from Wikipedia or code, DPO gives the same update whether its preference labels are real or half swapped at random (one model, modest preference strength), and scrambled text gives the most concentrated update of all. Applied to a recent claim that post-training is superficial because the effective rank of the weights barely changes, the baselines show that a random update of the same size leaves that rank almost unchanged while wrecking the model, and that an edit built on the effective rank of the update has no measurable effect in our tests. Rank measures of weight updates need a noise baseline, and claims about alignment need a control without alignment content.

    post-training
  332. arxiv:2610.22436 · cs.LG
    Generative adversarial networks with Monte Carlo dropout uncertainty quantification for structural health monitoring: Fibre Bragg grating data augmentation comparison in Hawk T1A aircraft
    Alvaro Peiro, Marios Impraimakis

    The signal generation capabilities of a novel data augmentation framework are examined in aerospace structures. First, the framework learns class-conditional temporal and spectral characteristics of the measured structural responses. It uses conditional generative adversarial network training and Monte Carlo dropout uncertainty quantification as a Bayesian approximation. This step enables stochastic signal synthesis and uncertainty quantification. Second, the generated responses are rigorously compared against experimental measurements across time, frequency, uncertainty, variability, and feature domains to assess physical realism, statistical consistency, and distributional fidelity. Finally, the synthetic data are evaluated to determine whether they should be incorporated into a supervised damage classification workflow. The framework is validated on the full-scale benchmark dataset from a BAE Systems Hawk T1A aircraft tested at the University of Sheffield's Structural Dynamics Laboratory for Verification and Validation. The test involves acceleration, fibre Bragg grating strain, and nonlinear dynamic response measurements. The results demonstrate that the model accurately captures dominant structural dynamics, it guides on signal use for classification improvement based on uncertainty characteristics, and it provides meaningful uncertainty estimates linked to signal complexity and excitation characteristics. Importantly, the framework offers a realistic generative modelling plan as a robust digital twin component for data-driven structural health monitoring, supporting resilient decision-making for condition assessment in safety-critical aerospace systems.

    benchmark
  333. arxiv:2610.21711 · cs.LG
    How language model agents reshape price competition
    Martin Bichler, Philip Hierhager, Kassian Köck, Fabian Pieroth

    Repeated Bertrand competition is a stringent benchmark for algorithmic pricing: in homogeneous-goods markets, standard learning procedures typically select the stage-game Nash outcome with prices near marginal cost. We show that LLM pricing agents can escape this benchmark, depending on deployment decisions. LLMs outperform standard online learners through faster adaptation, but this advantage largely disappears against full-information learners. An extracted surrogate algorithm illustrates how LLMs in specific deployments profit from regularly undercutting slower learners. Agents that post one price every period from raw market histories that include rival prices can sustain supracompetitive regimes in self-play. Agents that periodically emit probability distributions from summarized histories of their own profits instead move toward the competitive region in self-play. Markets mixing these interfaces can generate dynamic regimes in which LLMs establish a high-price anchor and regularly undercut it to capture demand. These regimes can keep both competitors' prices above marginal cost. Deployment choices shape these outcomes, shifting the focus from stationary equilibria to dynamic pricing regimes.

    self-playbenchmark
  334. arxiv:2610.22432 · cs.LG
    An Adaptive-Horizon Graph Neural Operator for coupled surrogate morphodynamic modelling
    Ying Cao, Yifan Yang, Junqiang Xia, Julia C. Mullarney +1

    Morphodynamic simulations require repeated coupling between hydrodynamic solvers, sediment transport modules, and bed updating. For long duration, multi-scenario simulations on large unstructured meshes, such coupling is computationally demanding. This study proposes an Adaptive-Horizon Graph Neural Operator (AH-GNO) that formulates bed evolution as a graph operator learning problem from recent hydrodynamic and topographic states to predict future bed elevation increments. The model operates directly on the native unstructured meshes. The approach integrates area-weighted graph learning, historical state selection, and forecast horizon estimation, allowing the effective history length (K) and update interval (B) to vary with the evolving flow and bed state. Using Telemac2D/GAIA solutions as supervised numerical references, the framework is evaluated on the official Bump and Yen bend benchmarks and on a field-scale application (Waiho River, New Zealand). Results show that the AH-GNO reproduces the main patterns of bedform migration, bend-scale erosion and deposition, and localized bed adjustment. Across the three cases, the tested CPU-GPU deployment reduced post-training bed-update workflow time from 12~1204 s with GAIA to 0.08~3.34 s with AH-GNO. This improvement corresponds to deployment-specific accelerations of approximately 150~360 times relative to 4-core GAIA CPU configurations, and with 105~186 times gains for 20-core CPU configurations. These results suggest that, while retaining hydrodynamic feedback from TELEMAC-2D, a learned operator for bed updating can provide an efficient acceleration component for morphodynamic scenario analysis and rapid coupled simulation.

    post-trainingbenchmark
  335. arxiv:2610.22426 · cs.AI
    PRIME: Towards Verifiable Evidence-Grounded Reasoning in LLM Agents
    Hamed Khosravi, Xiaoming Huo

    LLM agents are used to check records against an organization's requirements, for example whether a vulnerability report is ready for investigation. Requirements often allow several ways to be met, such as a severity rating with a curator-reviewed advisory or a published advisory link. Agents fail by accepting a report after a required check has failed, or by settling for a weaker conclusion without trying another way that could still succeed. We introduce PRIME (Precommitted Rules for Inference from Minimal Evidence). Before checking, it fixes evidence routes specifying how each conclusion may be justified. A separate program records each check result and refuses any conclusion the results do not justify. It also points the agent to checks that could justify a stronger conclusion, or runs them itself. PRIME reports only conclusions authorized by its rules and ledger. With correct requirements, exact checks at known locations, and sufficient budget at stopping, completion reaches the strongest supported conclusion. On 480 public-record tasks, an agent shown the requirements reported unjustified conclusions on 26.7% of tasks, while PRIME reported none. Its directions and added checks raised the average score for justified conclusions from 30.1, with refusal alone, to the maximum of 59.2 out of 100.

    agentllm agent
  336. arxiv:2610.22425 · cs.AI
    Recoverability-Controlled Safety Miscalibration in Large Language Models
    Yuwei Wang, Xingwang Liang, Yu Xie, Haotian Chi +1

    Safety-aligned language models are expected to refuse harmful requests while remaining useful for benign ones. However, because many evaluations score only the final response, they often conflate intent understanding with response-level safety behavior. To address this, we introduce RiskCueBench, a recoverability-controlled benchmark that separates these two stages by first measuring whether a model can successfully recover a high-level intent and then evaluating its safety behavior on the same prompt. RiskCueBench contains 160 latent intents and 640 abstract non-operational surface prompts across eight risk categories. The benchmark covers 25 completed model groups and 32,000 successful model outputs with final rubric-adjudicated labels. When conditioning on pairs where both surface forms are recoverable, we find that increased contextualized specificity induces a shift toward less safe behavior. Specifically, paired bootstrap estimates show that the refusal rate and global safety score decrease by 0.175 and 0.440, respectively. Furthermore, mixed effects modeling corroborates a 0.107 increase in unsafe cooperation. These results show that the safety behavior can be miscalibrated to the surface form even after the intent has been recovered, motivating safety evaluations that report response behavior conditional on semantic recoverability.

    benchmark
  337. arxiv:2610.22423 · cs.AI
    Really Need Recursion? Revisiting One-pass Registration with Multi-level Multi-space Contextual Guidance
    Yue Lin, Shenghai Wu, Mingyuan Meng

    Deformable image registration is a fundamental challenge in medical image analysis. Recent deep learning-based methods typically adopt either one-pass architectures that complete registration in a single network iteration, or recursive architectures that refine registration through multiple passes. While recursive methods often improve accuracy, they incur substantial computational overhead and slower inference speed. In this paper, we challenge the necessity of recursion for high-performance deformable registration and propose M2CG, a one-pass registration network with Multi-level Multi-space Contextual Guidance. It performs multi-step progressive pyramid registration within a single iteration, guided at each step by contextual cues from three levels (previous, current, and subsequent steps) and four spaces (image space, image feature space, segmentation space, and segmentation feature space). Our M2CG supports both unsupervised (without any labels) and semi-supervised (with segmentation labels for partial training data) registration, and has achieved state-of-the-art performance across two well-benchmarked inter-/intra-patient medical registration tasks with eight public datasets. Our findings suggest that the current reliance on recursion may stem from the absence of sufficient contextual cues that comprehensively characterize the remaining distance to the target alignment.

    benchmark
  338. arxiv:2610.23732 · cs.CV
    ESC: Escaping 2D Diffusion Ambiguity with Geometric Constraints for Consistent Novel View Synthesis
    Ge Ya Luo, Daniel Severo, Brian Karrer, Christopher Pal +2

    Generating 3D-consistent novel views from 2D diffusion models remains a fundamental challenge in computer vision. In this work, we introduce a framework that enhances geometric alignment for monocular view synthesis under drastic camera changes by jointly generating a target view and its corresponding depth map. By conditioning the diffusion pipeline on the single source image, source depth, and target camera rays, our approach forces the model to inherently reason in 3D space even under extreme viewpoint shifts. To guarantee spatial fidelity, we propose a novel per-pixel depth consistency constraint, which verifies and enforces 3D consistency across generated views. Our framework establishes a robust pipeline for geometrically consistent view synthesis, yielding superior visual accuracy and spatial alignment compared to current state-of-the-art methods. With this work, we pave the road towards robust 3D world models capable of understanding complex environments from a single view image.

    world model

02 US SEMI · SEC 8-K FILINGS

1 items

scanned: NVDA / AVGO / MRVL / COHR / LITE / AMD / TSM / SMCI / ANET / CRDO / POWL / VECO

  1. $SMCI · 8-K · filed 2026-10-15
    Super Micro Computer Inc
    Items: 5.02,9.01
    8-K

03 HUMANOID · COMPANY NEWS

62 items

scanned: figure-ai / 1x / boston-dynamics / unitree / apptronik / sanctuary-ai / neura-robotics / agility-robotics / physical-intelligence / agibot

04 CN PHOTONICS · ANNOUNCEMENTS

0 items
China sources are not connected yet.