# Research Papers — 2026-10 · bundle page 37 of 156 > Full-text markdown of papers #1801–1850 of 7799 published in 2026-10, newest first. AI-readable mirrors of the full 7799-paper 2026-10 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-10/page/37 [← Previous bundle](https://aipod.dev/papers/archive/2026-10/page/36.md) [Next bundle →](https://aipod.dev/papers/archive/2026-10/page/38.md) ### GLST: Global-Local Spatio-Temporal Memory Decomposition Framework for Online Chained-Goal Navigation Authors: Yousef Liang, Ryan Ardalan, Yarden Goraly, Chenhao Zhu, Yue Yin, Jonas Frey, Claire Tomlin, Marco Pavone, Shankar Sastry Published: 2026-10-13 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2610.17927v1 Autonomous navigation in real-world environments often requires long-horizon operation under partial observability and evolving goals. Existing mapless navigation approaches predominantly assume fixed goals throughout an episode, limiting applicability to practical scenarios where goals change online during execution. We study online chained-goal navigation, where goals evolve through existing-goal redirection or new-goal emergence without explicit transition signals. Under this setting, we find navigation may require either continued exploration or strategic revisitation, creating conflicting memory requirements. We propose Global-Local Spatio-Temporal Belief Learning (GLST), a memory-decomposition framework that separates latent beliefs into global-local spatial and long-short temporal representations for long-horizon reasoning. We further introduce goal-conditioned rewards to balance exploration and exploitation under evolving goals. Extensive experiments in simulation across diverse terrains demonstrate improved chained-goal navigation performance over baselines. Real-robot experiments using the Spot quadruped further validate our memory decomposition framework. --- ### CineMem: Attention-Aligned Fine-grained Memory Retrieval for Multi-Shot Long Video Generation Authors: Xiaoxue Wu, Xinyu Gao, Fang Yu, Yu Qiao, Xinyuan Chen Published: 2026-10-13 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2610.17926v1 Autoregressive video generation relies on memory retrieval to model long-range dependencies. This capability becomes particularly critical for multi-shot generation, where models must reconcile substantial visual changes across shot transitions with persistent semantic continuity. Existing methods tend to perform memory retrieval at the frame level and share the same historical context across all query tokens, overlooking the heterogeneous information needs and leading to redundant retrieval and inefficient use of historical memory. To bridge this gap, we introduce CineMem, an attention-aligned fine-grained memory retrieval framework that combines online memory organization with query-adaptive retrieval for autoregressive multi-shot video generation. Specifically, guided by intrinsic memory-access patterns revealed through attention analysis, we propose an online-clustered KV cache as a memory organization strategy over all historical tokens, enabling semantic-aware memory grouping and more precise retrieval. In addition, we introduce a query-level retrieval mechanism that provides broader access to historical information under a fixed retrieval budget, thereby improving long-term memory utilization. Together, these designs enable the multi-shot video generator to preserve long-range consistency across shot transitions, particularly in revisitation scenarios, while approximating full-memory access under a constrained memory budget. To facilitate evaluation, we further introduce a suite of task-specific, VLM-assisted metrics for faithful inter-shot consistency assessment. Extensive experiments demonstrate the effectiveness of our framework. --- ### Stein Variational Distributional Reinforcement Learning: Score-Guided Particles for Multivariate Return Distributions Authors: Manuel Haußmann, Melih Kandemir Published: 2026-10-13 Categories: cs.LG arXiv: https://arxiv.org/abs/2610.17925v1 Decisions that weigh several reward signals against each other require the joint distribution of the return vector. Categorical critics scale exponentially in the reward dimension, and quantiles have no canonical multivariate analogue. Multivariate distributional critics therefore represent each return distribution by M equal-weight particles trained with a discrepancy loss. Beyond their atoms, the particles provide no sampler and cannot resolve events rarer than 1/M. We propose Stein variational distributional RL (SVDRL). Its critic outputs Stein particles, which Stein variational gradient descent transports along a learned score field of the distributional Bellman target. We learn this field by bootstrapped denoising score matching across noise levels and tie it across the Bellman recursion by a consistency loss, which we prove is a non-expansive temporal-difference scheme in score space. The same field certifies the particles. Its reverse-diffusion sampler satisfies a supremal Wasserstein policy-evaluation bound that holds for bounded, possibly multimodal returns. The Stein particles inherit this bound up to a gap that we measure without ground truth, and the sampler answers tail-risk queries below the particle resolution at any sample size. Empirically, SVDRL is more accurate on rare events than particle critics with the same number of particles, and on par on MuJoCo continuous control. --- ### Batch-Paced: Fixed-Budget Batch Decoding in Discrete Diffusion Language Models Authors: Shuai Zhang, Zhanchao Zhou, Huachuan Qiu, Anqi Li, Zhenzhong Lan Published: 2026-10-13 Categories: cs.CL arXiv: https://arxiv.org/abs/2610.17917v1 As generative language models become more capable and widely used, inference demand grows, making efficient serving increasingly important. Discrete diffusion language models (dLLMs) offer a promising path to more efficient inference through parallel token generation. Prior dLLM decoding methods, however, focus on improving inference efficiency for individual sequences. We formulate fixed-budget dLLM batch decoding and introduce Batch-Paced, a training-free method that jointly optimizes the allocation of decoding compute across requests to improve overall batch quality. Across two dLLM backbones, five benchmarks and their representative mixtures, Batch-Paced improves batch performance under matched budgets; gains generally grow with batch size and largely persist across decoding configurations. Analysis of requests' decoding trajectories suggests that these gains arise largely by raising the committed-token confidence of lower-confidence requests. We further show that Batch-Paced's batch decode time remains competitive and that adding a decoding-depth allowance during Batch-Paced decoding can shorten it further at a modest performance cost. --- ### LinMorph: Learning to Morph Weights for Efficient Video DiT Linearization Authors: Lingyi Long, Congyue Zhao, Xiaoou Li, Yanmin Qian, Tao Huang Published: 2026-10-13 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2610.17914v1 Post-hoc linearization is an attractive way to accelerate pretrained Video Diffusion Transformers (Video DiTs), yet existing methods largely treat the quality loss induced by attention replacement as a recovery problem. We show that a more fundamental bottleneck lies upstream: \emph{initialization}. Random initialization discards pretrained knowledge, whereas manual weight reuse transfers it through rigid and incomplete source--target correspondences. We introduce \textbf{LinMorph}, which recasts Video DiT linearization as a learnable softmax-to-linear weight transformation problem. Instead of copying predefined counterparts, LinMorph learns a cross-parameter morphism that lets each target Gated DeltaNet (GDN) parameter draw jointly from pretrained query, key, value, and output projections, including for parameters with no direct softmax counterpart. A single morphism is shared across layers and learned through local teacher alignment, while producing layer-specific initializations from their respective pretrained weights. The resulting weights are then materialized as independent GDN modules for progressive distillation and joint recovery.Since the morphism operates in weight space rather than token space, it can be learned on short, low-resolution clips and directly reused at larger video scales. Across pretrained Video DiTs, LinMorph enables markedly faster recovery and better final generation quality than random initialization and heuristic weight reuse. Results show that efficient linearization depends not only on how a converted model is recovered, but critically on how pretrained knowledge is transferred before recovery. Project page: https://lonelyling.github.io/LinMorph. --- ### NBCLT: Parameter-Agnostic Norm-Ball Clipped Local Training Authors: Linxuan Pan, Shucheng Yang, Junchi Yang Published: 2026-10-13 Categories: cs.LG arXiv: https://arxiv.org/abs/2610.17913v1 Distributed learning with multiple local steps reduces communication but aggravates drift, and its practical performance often depends on carefully tuned hyperparameters under heterogeneous data and communication constraints. At the same time, recent Muon-type optimizers suggest that constraining updates within suitable norm balls provides a simple way to encode geometric priors into optimization. In this work, we propose NBCLT (Norm-Ball Clipped Local Training), a unified local-training framework for both federated learning (FL) and decentralized learning (DL). NBCLT combines a predict-and-correct local update rule with Euclidean projections onto flexible norm balls, thereby performing norm-ball clipping under flexible update geometries while retaining a simple implementation. In FL, its problem-parameter-free regime achieves $\mathcal{O}(ε^{-2})$ communication complexity and $\mathcal{O}(ε^{-4})$ total stochastic-gradient complexity without assuming bounded heterogeneity, when the per-round local-computation budget is scaled appropriately with the horizon. In DL, NBCLT is compatible with both static and time-varying mixing and preserves these accuracy exponents under the analogous computation scaling, up to fixed network constants and an additive initialization cost. --- ### Text-Only vs. Image-Aware VLM Judges for Manga Translation Evaluation Authors: Subaru Kimura, Jun Suzuki Published: 2026-10-13 Categories: cs.CL, cs.CV arXiv: https://arxiv.org/abs/2610.17908v1 In manga translation, visual context outside speech-bubble text, including speaker identities, referents, and scene tone, is depicted in page images, making it natural to refer to images during translation evaluation. However, most conventional manga translation evaluations rely solely on text-only automatic metrics, leaving the impact of image input insufficiently explored. We propose a practical validation framework that employs the same vision-language model (VLM) as both a text-only judge (TOJ) and an image-aware judge (IAJ) to investigate evaluation divergences associated with image-aware evaluation procedures. We use a Japanese-English manga dataset of official translations aligned at the page and speech-bubble levels. First, adding page images to TOJ improved the primary VLM's detection of deliberately injected, visually dependent degradations. Second, we analyzed evaluation divergences between TOJ and IAJ, finding different error counts and error-category distributions. Human verification did not reveal a consistent preference between TOJ and IAJ. --- ### SpecCloak: Probing Harmful Code Generation through Engineering Specifications Authors: Wenbin Chen, Jiexin Wang, Guanting Wu, Jiaxi Liang, Yi Cai Published: 2026-10-13 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2610.17907v1 The code generation capabilities of large language models (LLMs) can equally be misused to implement harmful functionality, underscoring the need to understand their safety behavior in response to programming requests. While request phrasing is known to influence refusal, it remains unclear whether the form in which coding task requirements are stated affects the refusal process for malicious code generation. We introduce PseudoSpec, a template that organizes functional requirements as engineering specifications, and investigate its effects through controlled comparisons, representation analysis, and activation interventions. Across five open-weight models, removing explicit risk keywords only partially reduces refusal, whereas specification prompting substantially lowers refusal even when these keywords are retained. This reduction accompanies consistent shifts toward code-output-related representations, and interventions along condition-difference directions alter refusal behavior. Output analysis further distinguishes code-related responses from generation degeneration. Building on these findings, we propose SpecCloak, a black-box red-teaming framework combining PseudoSpec with discriminator-guided optimization, achieving an average execution-verified attack success rate of 28.91\% across eight target models on RedCode-Gen. We also introduce SpecReveal, a defense prototype combining natural-language intent reconstruction with dual-view detection without target-model fine-tuning. It achieves a 96.88\% attack detection rate in a separate prompt-detection evaluation, with false positives assessed on benign programming and specification-style inputs. These results connect specification-induced response differences to practical security consequences and identify intent reconstruction as a complementary detection approach. --- ### Stable-Plastic Subspace Construction for Differentially Private Continual LoRA Authors: Jin Liu, Ning Xi, Yinbin Miao, Junkang Liu, Ke Cheng, Jianfeng Ma Published: 2026-10-13 Categories: cs.LG arXiv: https://arxiv.org/abs/2610.17904v1 Continual learning (CL) adapts pretrained models to a sequence of tasks while retaining knowledge acquired from earlier tasks. When task data are sensitive, differential privacy (DP) provides record-level protection, but maintaining both adaptation and retention under privacy constraints remains challenging. We study this problem in continual low-rank adaptation (LoRA) and observe that: 1) private training increases forgetting primarily during subsequent adaptation rather than initial task acquisition; 2) jointly perturbing both LoRA factors introduces a non-negligible bilinear noise interaction; and 3) DP aggravates the stability-plasticity dilemma, hindering both knowledge retention and adaptation to new tasks. Based on these observations, we propose DP-SPA, which constructs a stable subspace from previously privatized updates and a plastic subspace from pretrained weights within its orthogonal complement. The method fixes these bases and privately optimizes only the LoRA factor, eliminating the bilinear noise interaction within each step. Before merging each task update, it rescales stable-direction coefficients using historical and current update energies. We establish a privacy guarantee and show that recalibration does not increase historical overlap and tightens a conditional upper bound on old-task loss increase. Experiments across five datasets, eight settings, and multiple privacy budgets show that DP-SPA improves average accuracy and reduces forgetting compared with the evaluated baselines. --- ### DriveWLA: Unifying Future Dynamics and Scene Semantics through World-Language-Action Modeling for Autonomous Driving Authors: Shuzheng Wang, Haoxuan Li, Shengzhen Li, Chenxi Wu, Yajing Sun, Lu Hou, Wenyu Liu, Xinggang Wang Published: 2026-10-13 Categories: cs.CV, cs.AI, cs.RO arXiv: https://arxiv.org/abs/2610.17903v1 Vision-Language Models (VLMs) and world models have demonstrated strong capabilities in semantic understanding and future scene evolution for autonomous driving. However, effectively integrating both capabilities for trajectory generation remains challenging. In this work, we propose DriveWLA, a unified World-Language-Action model based on Mixture-of-Transformers (MoT) that jointly models semantic understanding, future scene dynamics, and trajectory generation. DriveWLA consists of three experts: a World Expert for future prediction, a Language Expert for semantic understanding, and an Action Expert for trajectory generation. We further introduce Cross-Expert Attention, which enables controlled information flow across experts while aligning semantic and dynamic representations. To improve trajectory generation beyond imitation learning, we apply Flow-GRPO reinforcement learning post-training to refine the Action Expert. Extensive experiments show that DriveWLA achieves superior performance in both open-loop and closed-loop evaluation. DriveWLA achieves 92.3 PDMS on NAVSIM v1 and 90.1 EPDMS on NAVSIM v2, demonstrating strong performance among VLM-based and world-model-based planners. Without specific fine-tuning, DriveWLA further demonstrates strong zero-shot generalization to HUGSIM, nuScenes, and Bench2Drive. These results validate the effectiveness of our unified World-Language-Action modeling, establishing DriveWLA as a generalizable framework for autonomous driving. --- ### Visual Dependence Does Not Imply Answerability: A Systematic Analysis of Answerability Signals in Frozen InstructBLIP Authors: Minjun Kim, Hyunmin Song, Changwon Lim Published: 2026-10-13 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2610.17902v1 Visual question answering models often produce plausible responses even when the available visual evidence is insufficient to support a valid answer. Although prior work has introduced benchmarks and abstention mechanisms for unanswerable visual questions, it remains unclear which signals derived from a frozen vision-language model reliably indicate answerability on unseen image-question pairs. We systematically evaluate answerability signals on a stratified 10K subset of UNK-VQA using frozen InstructBLIP, covering predictive uncertainty, cross-modal alignment, visual grounding, visual dependence, internal representations, and learned probes. Answerability discrimination remains weak, with test AUROC largely confined to 0.50-0.54. CLIP alignment yields the highest point estimate but does not significantly outperform generation confidence under paired bootstrap analysis. Targeted removal of question-relevant visual evidence causes substantially larger changes in answer confidence than same-size random masking, yet intervention sensitivity provides essentially no answerability discrimination. Increasing probe capacity, combining multiple signals, and adding reason-aware supervision do not provide robust improvements, while post-freeze cross-validation yields similarly weak discrimination. Selective-answering analysis shows limited usable coverage under low-risk operating criteria. Failure-mode analysis shows that the relative utility of individual signals varies across perturbation mechanisms. Taken together, these results show that, on our UNK-VQA subset with frozen InstructBLIP, visual dependence, uncertainty, semantic alignment, and static representations are not robust standalone proxies for answerability. Our findings emphasize the need to distinguish a model's reliance on visual evidence from whether that evidence is sufficient to support an answer. --- ### FlashMask-3: Efficient and Expressive Mask-Aware Distributed Attention Authors: Guoxia Wang, Qianyue He, Siming Wu, Haoyang Xie, Qiwen Bao, Jinle Zeng, Jiabin Yang, Tian Wu, Dianhai Yu Published: 2026-10-13 Categories: cs.LG arXiv: https://arxiv.org/abs/2610.17901v1 Long-context training of large language and multimodal models couples structured attention masks (document, prefix-LM, share-question, sliding-window, block-sparse) with multi-node context parallelism, making the mask a dominant bottleneck. FlashMask's column-wise sparse mask compresses them from $O(n^2)$ to $O(n)$ via four query-axis boundary vectors, covering broad structured patterns. Implemented for Ampere GPUs, it neither exploits FlashAttention-3's TMA/WGMMA warp-specialized pipeline on Hopper nor supports distribution. The column-wise sparse mask is indexed by key positions but stores query-axis interval boundaries. Under any multi-chunk context-parallel assignment, it reduces per-rank mask localization to a provably zero-communication \emph{clip--shift--max} on the boundary vectors, compatible with the single-GPU kernel. FlashMask-3 introduces a mask-aware Hopper kernel with three-role warp specialization atop FlashAttention-3's TMA/WGMMA pipeline, plus a mask-aware distributed extension: zero-communication mask localization, a sparsity-aware load balancer (HCC-IPO), and a mask-aware compute-communication overlap with mask-driven communication pruning, whose forward all-gather uses topology-aware hierarchical routing. On a single H100 across 12 masks at 4K--128K context, it improves achieved TFLOPs over FlashMask, FlexAttention, and MagiAttention by 45--141%, 34--71%, and 2--44%. At context-parallel degrees 4--32 with 8K sequence per device, it attains 31--75% compute-communication overlap efficiency and improves throughput over Megatron-Core CP ring, Megatron-Core CP all-gather, and MagiAttention by $1.80\times$--$3.71\times$, $1.71\times$--$2.92\times$, and $1.09\times$--$1.46\times$. In end-to-end training of a 118B-A14B MoE at 128K on 128 H100 GPUs, it raises throughput by 7.3--11.0% over a Dual Chunk all-gather baseline, with attention forward+backward 24.5--44.9% faster. --- ### PaceReasoner: Label-Free Learning to Answer Briefly from Long Reasoning Authors: Changguo Jia, Zhiyou Xiao, Xinbei Ma, Geng Chen, Weiming Zhang, Yiming Xu, Tianqi Zhao, Minghui Zhou Published: 2026-10-13 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2610.17900v1 Extended chain-of-thought reasoning improves problem-solving performance but increases inference cost and latency, motivating more accurate answers with shorter reasoning. We introduce PaceReasoner, a label-free test-time reinforcement learning framework that uses a model's LONG responses to improve its SHORT-mode accuracy. Both modes share a trainable policy, creating a need for reliable supervision, sufficient SHORT training samples, and continued LONG responses to guide learning. PaceReasoner combines all-response and LONG-only answer consensus through progressive, confidence-weighted supervision. A behavior reward gradually shifts from encouraging SHORT exploration to rewarding agreement with LONG answers, while a ratio controller regulates the balance between modes. Across three model backbones and four benchmarks, PaceReasoner improves average SHORT accuracy over the baselines by 7.4--13.5 percentage points, with fewer than 25 reasoning tokens per response on average. Ablations support the key components, and cross-dataset results demonstrate transfer beyond the adaptation set. --- ### Evaluating the Causal Faithfulness of SHAP as a Candidate Attribution Method for SAE Features Authors: Samuel Xie, Noura Al-Moubayed Published: 2026-10-13 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2610.17898v1 Sparse autoencoders decompose language model representations into interpretable features, but feature interpretation in this literature remains predominantly correlational. Attribution methods offer more direct estimates of feature influence, yet their faithfulness to actual causal effects has not been systematically evaluated in the SAE setting. We apply SHAP-based attribution to SAE latent features and evaluate it against direct causal intervention, using sentiment classification on SST-2 with GPT-2 Small and a 32,768-feature layer-7 SAE as a testbed. We compare SHAP to Integrated Gradients, Gradient$\times$Input, and probe-weight baselines across linear and MLP probe arms, scoring each against feature ablation and activation steering measured through the language model's own logit difference. Most faithfulness associations we measure do not survive multiple-comparison correction, and attribution methods that agree strongly with one another ($ρ= 0.42$--$0.80$) agree poorly with measured causal effects, indicating that inter-method agreement is not evidence of causal faithfulness. Two effects hold under control. First, local per-example attribution retains a directional advantage over global averaged attribution, while its apparent importance advantage is explained by feature activity rather than superior causal ranking. Second, under known-ground-truth minimal pairs, attribution recovers inserted sentiment signal well above chance (AUROC $0.792$) but fails to localize deletions (AUROC $0.520$), suggesting that these methods detect the presence of a feature's influence more reliably than its absence. We argue that importance faithfulness and directional faithfulness are separable properties requiring separate evaluation, and that attribution agreement should not substitute for causal validation. --- ### Interaction-as-a-Judge Authors: Wayne Chi, I-Hung Hsu, Li Chen, Wei-Lin Chiang, Ion Stoica, Anastasios Nikolas Angelopoulos, Ameet Talwalkar, Chris Donahue, Cho-Jui Hsieh Published: 2026-10-13 Categories: cs.LG, cs.AI, cs.SE arXiv: https://arxiv.org/abs/2610.17895v2 Coding agents are tasked with producing sophisticated interactive applications such as websites, desktop applications, and games. While humans assess these applications interactively, current automated LLM evaluations primarily focus on code analysis and unit tests. This mismatch can cause LLM judges to miss failures that are immediately apparent to human users. We study Interaction-as-a-Judge, an evaluation paradigm where computer-use agents (CUAs) directly interact with an application. We demonstrate that Interaction-as-a-Judge can improve both evaluation and generation of interactive applications. To this end, we collect and release two complementary datasets: $6,858$ pairs of applications with human preferences and corresponding agent traces, and $222$ applications with human-verified requirement labels. Relative to standard non-interactive approaches, we find that Interaction-as-a-Judge improves evaluation in terms of predicting human preferences ($65.6\% \to 69.1\%$) and requirement verification ($77.0\% \to 81.6\%$). We also observe scaling benefits: across both preference alignment and requirement verification, higher-quality interactions improve performance. Finally, through both inference- and training-time experiments, we demonstrate that Interaction-as-a-Judge can also improve the quality of generated artifacts. At inference time, Interaction-as-a-Judge repairs $75.5\%$ of broken requirements within five attempts, compared to $55.4\%$ for code-only self-feedback. At training time, using Interaction-as-a-Judge as a reward signal improves win-rate over the base model by $6.7$pp while reducing both generated code length by $35.7\%$ and build failure rates by $7.2$pp compared to a code-only reward signal. --- ### Swiss4Arg: A Parallel Dataset for Multilingual Argument Mining across Swiss National Languages Authors: Sophia Conrad, Fatma-Zohra Rezkellah, Christina Niklaus, Reto Gubelmann Published: 2026-10-13 Categories: cs.CL arXiv: https://arxiv.org/abs/2610.17894v1 We present Swiss4Arg, a multilingual dataset for argument mining based on official Swiss voting brochures, covering all four national languages: German, French, Italian, and Romansh (a Romance minority language exclusively spoken in Switzerland). The brochures contain explicitly formulated arguments systematically organized into pro and con sections, resulting in a cross-lingually aligned resource with balanced argumentative stance. We create a dataset that comprises 53 federal voting issues collected between 2019 and 2025, annotated for argument components (major claims, claims, premises, and non-argumentative units) and relations (support and attack) by three native speakers per language. Inter-annotator agreement (Krippendorff's alpha) reaches 0.548 for component classification and 0.623 for relation and target classification, computed jointly across all annotators on the parallel data. We release both adjudicated labels and individual annotator annotations, enabling analyses of disagreement and cross-linguistic variation, and provide baseline results using GPT-5.4 and SwissBERT. Swiss4Arg establishes a novel benchmark for multilingual argument mining, with a particular focus on cross-lingual comparability and low-resource settings such as Romansh. The dataset and code are publicly available at https://anonymous-hf.up.railway.app/a/gozw0nion72y/ (HuggingFace) and https://anonymous.4open.science/r/MultiVote-03BD/ (GitHub). --- ### Training LLMs for Multi-Agent Collaboration with Verifiable Information Rewards Authors: Keru Chen, Sen Lin, Yingbin Liang, Nathaniel D. Bastian, Shaofeng Zou Published: 2026-10-13 Categories: cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2610.17893v1 Effective collaboration among large language model (LLM) agents in multi-agent systems requires sharing relevant evidence, requesting missing information, and incorporating teammates' contributions into decisions. Training these behaviors is challenging because final task outcomes provide limited feedback on the value of individual messages. We introduce CoSIR (Collaboration via Surprisal-based Information Rewards), a post-training approach that uses a simple, verifiable environment to teach collaboration behaviors that generalize to more complex, unseen tasks. We propose a Bayesian evaluator that scores shared information by its reduction in team surprisal, defined as the sum of the negative log probabilities assigned to the correct option given each agent's available evidence. Requests are scored by the potential reduction in the same quantity if the requested information were truthfully provided. Separate rewards encourage agents to use the received information in their private choices and calculations. We post-train Qwen3-4B, Qwen3-8B, and Gemma-3-4B-it through self-play and evaluate transfer to seven unseen natural-language collaboration tasks. CoSIR achieves the highest aggregate score, while its performance in general tasks remains competitive. Behavioral analyses show increased disclosure of private evidence and fewer judged failures to use teammates' input. --- ### Deep Mesoscopic Dynamics Authors: Yue Zhao, Shiqi Wu, Yihang Ouyang, Sergey Grebenchuk, Magdalena Grzeszczyk, Xiaoxu Zhao, Kostya S. Novoselov, Maciej Koperski, Qianxiao Li Published: 2026-10-13 Categories: cs.LG, cond-mat.stat-mech, physics.comp-ph arXiv: https://arxiv.org/abs/2610.17885v1 Microscopic simulations resolve spatial patterns but are feasible only over small domains and short timescales, whereas bulk descriptions erase those patterns through spatial averaging. Mesoscopic fields preserve this spatial organisation, but their effective equations are known only for selected microscopic systems and scaling limits. We introduce Deep Mesoscopic Dynamics (DeepMesoD), which uses neural networks to learn an effective mesoscopic field equation from microscopic trajectories at a prescribed coarse-graining scale. For the nonconserved dissipative dynamics studied here, DeepMesoD learns the free-energy functional with a spatial interaction kernel and a state-dependent dissipation operator constrained by the fluctuation-dissipation relation. We first validate the construction in the Ising-Kac model, where the learned mesoscopic equation recovers the known hydrodynamic limit and rollouts remain stable far beyond the training horizon. We then treat two systems without closed-form mesoscopic descriptions. For a two-dimensional Heisenberg ferromagnet governed by atomistic spin dynamics, DeepMesoD learns a mesoscopic model whose predictions match atomistic references up to 1 ns, and a rollout with an unseen initial condition remains bounded up to 25 ns. For layered CrBr$_3$, the learned kernel captures the suppression and sign reversal of interlayer coupling at a stacking fault whose location and strength are not supplied as inputs. Controlled changes to the learned interaction show how fault number and placement shape multistep magnetic responses, complementing our magneto-photoluminescence measurements. Across the two atomistic systems, DeepMesoD achieves a three-to-four-order-of-magnitude reduction in wall-clock time compared with atomistic spin dynamics. --- ### Diagnose-to-Generate: Eliminating Structural Distortions in Flow Models via Patch-Level Grounded Alignment Authors: Yifan Li, Lishuai Gao, Tong Zhang, Qilong Huang, Chen Zhao, Zhuoliang Kang, Jie Hu, Yujie Zhong, Fenghe Tang, S Kevin Zhou, Xiaoming Wei Published: 2026-10-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.17882v1 Flow-based image generators achieve strong semantic synthesis but still produce localized structural distortions, such as malformed hands and limbs. Existing post-training methods typically optimize global scalar rewards, which provide weak spatial credit assignment and may alter well-formed regions while correcting local defects. We present \textit{Diagnose-to-Generate (D2G)}, a grounded alignment framework that separates visual diagnosis from generative optimization. We construct \textbf{Achilles-600K}, a region-annotated dataset for structural anomalies, and train \textbf{ARGUS} through supervised cold-start and Region-Aware Group Relative Policy Optimization to predict anomaly categories and bounding boxes. These diagnoses are converted into confidence-weighted spatial rewards for our \textbf{Patch-level Flow-GRPO}. We further identify that global stochastic exploration can shift semantic regions across group rollouts, making patch-wise relative advantages spatially inconsistent. Our \textbf{Trajectory-Preserving Stochastic Sampling (TPSS)} anchors early denoising with deterministic ODE trajectories and introduces stochasticity only at later steps, thereby limiting intra-group spatial drift while retaining local exploration. Experiments on FLUX.1-dev and SD3.5 show improved structural fidelity with preserved FID and CLIP scores. Consistent gains under HumanRefiner, zero-shot GPT-4.1, and blinded human evaluation further demonstrate that the improvements extend beyond the training reward model. --- ### Cross-Lingual Transfer of Evil Vectors Is Directionally Asymmetric Authors: Juann Kim, Minku Kang Published: 2026-10-13 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2610.17880v1 Persona vectors are linear directions in the residual-stream activations of LLMs that encode behavioral traits and enable fine-grained control over model behavior. Among them, the evil persona vector is particularly relevant to understanding harmful behavior, jailbreak vulnerabilities, and the unlearning of harmful knowledge. Yet whether this vector reflects a genuinely cross-lingual concept, rather than an artifact of its extraction language, remains untested. We conduct a systematic evil-vector transfer study across two high-resource languages (English, Spanish) and two low-resource languages (Swahili, Telugu) on four open-source instruction-tuned LLMs, and find that transfer is directionally asymmetric. Evil vectors are harder to obtain reliably from low-resource languages, and the resulting vectors are largely ineffective at steering, even within their own language. In contrast, vectors extracted from high-resource languages transfer effectively to low-resource inputs while largely preserving general knowledge, as measured by MMLU. Geometric analysis further supports this pattern: evil vectors from the two high-resource languages are closely aligned, whereas low-resource vectors align only weakly with this shared direction. These results suggest that a shared cross-lingual evil direction exists and, once reliably extracted from a high-resource language, can be reused to steer behavior in low-resource languages without additional per-language data collection. --- ### LoopREG: Representation Supervision Enables Test-Time Recurrent Depth in Diffusion Transformers Authors: Zipeng Wang, Anping Xiong, Jing Zhang, Xiaoya Zhu, Yi Jiang, Yanbing Xiao, Zige Zeng, Zezhou Chen, Xiang Liu, Huan Hu, Ping Chen, Kai Wang et al. Published: 2026-10-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.17877v1 Reusing Transformer blocks offers a way to increase diffusion-model computation at test time, but naive recurrence can degrade generation beyond its training depth. We introduce LoopREG, a representation-supervised recurrent diffusion Transformer that repeatedly applies a shared block at the state supervised by a pretrained visual encoder. Alignment is applied after the final recurrent visit during training; inference extends the same learned operator to additional visits. On ImageNet 256 x 256, LoopREG reduces SiT-B/2 FID from the local REG baseline's 15.08 to 4.91 at 400K training steps, a 67.41% reduction. Within a checkpoint trained with two visits, one extra inference visit improves FID from 14.71 to 5.34 with unchanged weights. The benefit persists on SiT-L/2 and SiT-XL/2. Controlled interventions show that useful recurrence follows the location of representation supervision, while REPA alignment produces a smaller extrapolation benefit. These results make representation-supervised recurrence a practical route to improving diffusion generation through additional inference depth. --- ### Searching for What Matters: Open-Ended Discovery of Multiple Hypotheses with Agentic Data Exploration Authors: Sanchit Kabra, Nurendra Choudhary, Jiangshu Du, Eddie Huang, Danai Koutra, Chandan K Reddy Published: 2026-10-13 Categories: cs.LG arXiv: https://arxiv.org/abs/2610.17874v1 Discovering what drives an outcome from data is central to science and often requires deciding what is worth investigating before a specific hypothesis is known. Yet most existing hypothesis generation settings substantially narrow this search by providing a research question, target relationship, or curated context. We introduce open-ended multiple hypothesis discovery, where a system receives only a large tabular dataset and an outcome and must discover multiple data-supported hypotheses without knowing in advance the relevant variables, functional forms, conditions, or number of underlying mechanisms. To study this problem, we introduce a benchmark built from structural causal models grounded in real-world data schemas, containing 351 planted hypotheses across 15 tasks and five domains, with 53--65 variables and 1M--10M rows per task. The causal model is hidden from the discovery method and retained only by the evaluator, allowing proposed hypotheses to be tested through interventions and separating observational discovery from causal validation. We further propose HypAgent, which treats the dataset as an interactive environment and performs iterative search by proposing candidate directions, triaging and refining promising hypotheses, and removing explained signal so that each accepted discovery redirects subsequent search. HypAgent achieves 0.47 precision, 0.51 recall, and 0.49 F1, compared with 0.35 F1 for the strongest baseline. Our analysis shows that many failures occur after a relevant variable has already been surfaced, suggesting that correctly characterizing a mechanism's functional form and conditions, rather than merely identifying relevant variables, remains a key challenge in open-ended discovery. --- ### ECMR: Evidence-Aware Deliberation Control for Multi-Agent LLM Reasoning Authors: Yuan Liu, Haoyue Peng, Fanliang Hu, Lei Huang, Wenli Cai Published: 2026-10-13 Categories: cs.AI, cs.MA arXiv: https://arxiv.org/abs/2610.17869v1 Multi-agent Large Language Model (LLM) systems typically commit to a deliberation protocol before inference. A fixed protocol cannot allocate computation differently to resolved and unresolved instances, and it may mistake repeated support for independent confirmation. We study this failure as semantic false consensus: a confident group answer supported by narrow, duplicated, or correlated rationales. We introduce ECMR, an evidence-aware controller over frozen role-conditioned LLMs. At each step, ECMR summarizes answer concentration, support diversity and independence, contradictions, role behavior, verification history, and remaining budget, then selects whether to propose, challenge, verify, arbitrate, diversify, correct a role, or stop. We formulate this decision as finite-horizon budgeted control and derive sufficient conditions under which state conditioning, risk-triggered verification, and role correction improve cost-adjusted utility. Across a declared held-out evaluation of 9,000 records from MedQA, CaseHOLD, and FinQA, ECMR-Full increases pooled accuracy from 62.1% to 84.9%, reduces false-consensus rate from 40.7% to 17.0%, and raises evidence-support rate from 58.0% to 88.7%. State shuffling loses 11.7 accuracy points with only 0.7 fewer rounds, while harder traces shift from early stopping toward arbitration. ECMR-Full uses the largest token budget, yielding the strongest absolute reliability rather than uniformly superior token efficiency. --- ### Prefilling-dLLM v2: Learned Chunk Routing for Long-Context Diffusion Language Models Authors: Qi Han, Jing Xiong, Ziyuan Zhang, Ngai Wong Published: 2026-10-13 Categories: cs.CL arXiv: https://arxiv.org/abs/2610.17868v1 Diffusion language models apply bidirectional attention over the whole context at every denoising step, so a longer context makes each step substantially more expensive. Sparsifying only at inference leaves the pretrained weights unchanged, so the model is never trained under the sparse pattern it runs with. We propose Prefilling-dLLM v2, which continues training a diffusion language model so that it routes with its native mask token. One mask is appended at the end of each chunk, and a learned role offset marks it as a routing slot. On every fourth layer, a query keeps a bidirectional local window and the top-K chunks scored from that mask; the other layers use a bidirectional sliding window. Starting from Dream-v0-Base-7B extended to 16K with YaRN, we compare Prefilling-dLLM v2 with bidirectional NSA, DSA, and SWA under the same layout, key budget, and dense initialization, using the same Fast-dLLM decoder. With the YaRN factor scaled to the evaluation length, Prefilling-dLLM v2 leads the other sparse methods on all three RULER tasks from 16K to 128K. Its mean accuracy at 128K is 7.4 points above NSA, the next sparse model, while DSA and SWA remain near zero. At the same length, decoding is 172x faster than with full dense attention. --- ### RT-J: Large-Scale Pretraining of Relational Transformers for Context-Efficient Predictions Authors: Rishabh Ranjan, Vignesh Kothapalli, Harshvardhan Agarwal, Charilaos Kanatsoulis, Roshan Upendra, Tom Palczewski, Carlos Guestrin, Jure Leskovec Published: 2026-10-13 Categories: cs.LG, cs.AI, cs.DB arXiv: https://arxiv.org/abs/2610.17867v3 Foundation models for tables and relational databases are increasingly used for predictive tasks across various scientific and industrial domains. However, adapting existing models to unseen datasets requires tens of thousands of examples. Here we introduce RT-J, an 86M-parameter relational foundation model (RFM) that makes accurate predictions even with hundreds of in-context examples or fewer. To achieve this, we first assemble the Join, a large collection of 639 web-crawled databases with 6,023 LLM-generated forecasting tasks. We pretrain with long contexts of up to 8,192 database cells, using in-context examples from a novel lightweight random-walk retriever. Finally, we leverage context tuning and ensembling to scale up test-time compute. On in-context regression tasks from RelBench, RT-J requires 10x fewer examples on average to match strong feature engineering pipelines with LLM-agent or RDBLearn featurizers and TabICLv2 or LightGBM predictors. Overall, our work pushes the frontier for context-efficient predictions on relational data and highlights the importance of real-world semantics, context engineering and large-scale pretraining for RFMs. --- ### MicroCLIP: Connecting Text and Microscopy Images via Contrastive Language--Image Pretraining Authors: Yanran Ruan, Zhixiang Wei, Shiyao Hong, Xiaoxiao Ma, Yuxuan Gu, Liming Ye, Huaian Chen, Yi Jin Published: 2026-10-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.17866v1 Understanding fluorescence microscopy images is fundamental to automating biological research. However, existing methods typically learn predefined concepts from closed-set data, limiting transfer across tasks and biological contexts. In this work, we introduce MicroCLIP, a vision-language model that learns generalizable cellular and subcellular representations from natural language supervision for open-set image classification and retrieval. To pretrain MicroCLIP, we construct MicroPool-2M, a dataset of approximately 2.5 million fluorescence microscopy image-text pairs spanning diverse cellular structures, cell lines, and imaging conditions. Aligning microscopy images with biological concepts in natural language is challenging because most images lack descriptive captions and biological knowledge annotations. We therefore progressively establish image-text associations at three levels to provide MicroCLIP with richer semantic supervision: (1) category-level alignment links images to standardized labels defined by subcellular structure-cell-line combinations; (2) knowledge-level alignment incorporates biological priors derived from these labels to capture relationships across species and cell types; and (3) detail-level alignment uses a multimodal large language model to recaption and verify image-text pairs at scale, emphasizing fine-grained visual characteristics. Extensive experiments demonstrate that MicroCLIP achieves leading performance in zero-shot classification and text-image retrieval on microscopy images. It retrieves images with specific cellular characteristics from thousands of candidates and generalizes to unseen biological concepts without additional supervision. These results highlight MicroCLIP's potential as a foundation model for fluorescence microscopy. --- ### TStruct: Learning Shared Temporal Structures for Long-Term Time-Series Forecasting Authors: Yangze Li, Zhengyang Zhou, Qihe Huang, Pengkun Wang, Binwu Wang, Yang Wang Published: 2026-10-13 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2610.17864v1 Long-term time series forecasting requires temporal dependency modeling that remains stable and generalizable under temporal distribution shifts. However, real-world time series are often noisy and non-stationary, making deep learning models prone to fitting incidental temporal interactions in the training data. A key question is whether there exist temporal relationship structures that recur across different temporal positions and variables. We analyze temporal relationship matrices on commonly used time series datasets and find that samples from different temporal positions and different variables exhibit shared dominant temporal relationship structures. This indicates that temporal relationship structures contain patterns that recur stably across samples and variables. To learn such shared structures, we propose \textbf{T}emporal \textbf{Struct}ure Network (\textbf{TStruct}), a simple yet effective network for directly learning shared and generalizable temporal dependencies. Specifically, TStruct directly parameterizes a set of shared temporal dependency matrices and generates normalized combination weights using dynamic sample information and static variable priors. These weights represent a soft allocation of each sample-variable pair over shared temporal dependency structures, enabling the model to generate adaptive temporal dependency matrices under shared structural constraints, rather than fitting incidental interactions in a fully unconstrained manner. Notably, this simple structure-aware design consistently outperforms several strong baselines across all datasets, achieving superior accuracy while maintaining high computational efficiency. --- ### Socratic Skill Learning: Interleaving Teaching and Practice for Transferable Agent Skills Authors: Letian Zhu, Han Li, Yifan Yao, Xinping Lei, Xinyu Che, Jiahao Wang, Dianyi Wang, Yucheng Fu, Jiakai Wang, Zhiqi Bai, Wenbo Su, Jiaheng Liu Published: 2026-10-13 Categories: cs.AI arXiv: https://arxiv.org/abs/2610.17860v1 Agent Skills improve LLM agents by making task knowledge reusable across sessions. Learning transferable Skills requires evidence about a procedure and its conditions of use, which a successful task attempt can leave untested. We introduce Socratic Skill Learning (SSL), which interleaves expert teaching with Student practice. An Educator guides investigation while the Student remains the sole environment operator and Skill author, able to question, test, and adapt advice. To test reuse beyond the learning task, we construct SkillTransferBench, with 35 task families and 408 executable cases, pairing public variation specifications with held-out evaluation in fresh solver sessions. On the main 22-family, 155-case cohort, Fixed-Window SSL reaches 0.801 utility, versus 0.723 without a Skill, 0.761 for adapted Trace2Skill (trajectory distillation), and 0.751 for adapted SKILL-KD (contrastive distillation). Across three model pairs on shared families, interleaved guidance outperforms expert authoring and post-hoc teaching. Process analysis links guidance to Student writing and package revision; a case study shows the Student testing and refining a proposed procedure for later reuse. As demand for new lessons falls, the cost of fixed-window review grows. Issue-Guided SSL tracks unresolved questions across Skill revisions to select and focus review, reaching 0.805 utility with 61.7% fewer Educator tokens and 16.9% fewer total acquisition tokens than Fixed-Window SSL on the main cohort. --- ### Frequency Routing and Identity Conditioning for Thermal-to-Visible Face Translation Authors: Haojie Li, Xinlong Sun, Xiaohaoyang Lei, Kai Cheng, Yiguo Qiao, Xu Yang, Cunjian Chen Published: 2026-10-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.17859v1 Thermal-to-visible (T2V) face translation has important applications in nighttime biometric recognition, low-light surveillance, and human interpretation of thermal observations. The task remains challenging because thermal images do not directly record visible appearance, while the cross-spectral cues useful for reconstruction vary across frequency components and semantic roles. Existing spatially conditioned models tend to process these cues through a homogeneous adaptation path and do not explicitly separate local structure from global identity. We propose FreqID-Control, a parameter-efficient framework that organizes thermal guidance according to spectral reliability and semantic role while preserving the pretrained generative prior. It combines frequency-routed low-rank adaptation, which allocates separate transformation subspaces to low-frequency structure and high-frequency detail, with shared cross-spectral identity conditioning, which supplies compact global identity tokens independently of spatial thermal tokens. A controlled analysis on MCXFace and SpeakingFaces provides the empirical basis for this design, revealing stable low-frequency correspondence and edge-localized detail correspondence. Experiments on both datasets demonstrate strong image fidelity and biometric identity preservation, while controlled ablations on MCXFace confirm the complementary contributions of frequency routing and shared identity conditioning. --- ### MMTA: Benchmarking Multimodal Temporal Analysis with Time Series, Text, and Vision Authors: Ziyang Zhang, Shenyi Li, Yilin Wang, Ziyun Cui, Bowen Zhou, Wen Wu, Chao Zhang Published: 2026-10-13 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2610.17856v2 Time-series analysis is increasingly studied in multimodal settings, yet most benchmarks pair temporal signals with text or with visualizations derived from the same signal. This leaves open whether multimodal large language models (MLLMs) can integrate time series with real-world visual evidence. We introduce MMTA (multimodal temporal analysis), a benchmark of tri-modal tasks with diverse cross-modal complementarity patterns among time series, text, and vision. MMTA spans nine domains -- robotics, autonomous driving, human action, manufacturing, AR assistance, healthcare, recommendation, finance, and climate -- under video+TS and image+TS layouts, with a 3,524-sample test split covering classification and forecasting. The benchmark standardizes heterogeneous real-world sources into a shared schema with domain-specific prompts, provenance documentation, and success-weighted metrics that penalize both wrong answers and failed generations. Evaluating eight open-source and proprietary MLLMs reveals a shared bottleneck: representing time series as text creates very long token sequences and weakens fine-grained alignment with visual evidence, especially in long-video settings. We then introduce TimeOmni-v, a multimodal LLM extension with a dedicated time-series encoder, temporal positional alignment, modality-aware token layout, and a forecasting head. TimeOmni-v improves long-video classification and time-series inference efficiency, showing that MMTA not only exposes representation and grounding failures but also motivates targeted model design. --- ### DVTR-KD: Distilling Vision-Language Models via Intermediate Visual Token Reconstruction Authors: Haolan Wei, Jiahui Wang, Junli Zha, Xinkai Lu, Yunxiao Zou, Chao Zhou, Xiang Song Published: 2026-10-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.17853v1 Multimodal large language models (MLLMs) have achieved strong performance on diverse vision-language tasks, but their computational demands limit deployment under constrained budgets, particularly on edge devices. Many works focus on Knowledge Distillation (KD) and try to transfer knowledge from powerful teacher models to students. In addition to reducing model size, visual token compression provides a complementary route to efficiency, but introduces mismatched visual token counts that complicate direct token-wise knowledge transfer. To mitigate this, we propose DVTR-KD, a knowledge distillation framework based on \textbf{D}ense \textbf{V}isual \textbf{T}oken \textbf{R}econstruction, which reconstructs the teacher's dense visual tokens from the student's compressed tokens. We reconstruct visual states after the final LLM layer and at the input to an intermediate layer identified as important for visual-information readout. At the intermediate layer, reconstruction is supervised through the teacher's response predictions rather than by direct token-wise KL or MSE matching. An auxiliary attention-readout loss complements this response-level supervision by aligning how the student and the original teacher integrate contextual information when predicting responses. Evaluations on diverse benchmarks demonstrate that DVTR-KD improves the performance of token-compressed students while retaining the inference efficiency afforded by visual token compression. --- ### Persistence Positional Encoding: A Minimal Transformer Recipe for Persistence Diagrams Authors: Eunwoo Heo, Ganghyun Kim, Jaejun Yoo Published: 2026-10-13 Categories: cs.LG arXiv: https://arxiv.org/abs/2610.17848v1 Persistence diagrams (PDs) represent multiscale topological information as multisets of birth-death pairs $(b,d)$. How best to use them with Transformers, however, remains unclear. Connecting PDs to Transformers requires choosing how topological information is tokenized and exposed to self-attention. Prior methods use specialized designs for tokenization, aggregation, Transformer blocks, or attention, but are hard to compare since reported results rarely share PD inputs, splits, or model-selection rules. We instead confine the diagram-specific design to point representation, mapping each birth-death pair through fixed random Fourier features while leaving the Transformer backbone unchanged, and call this recipe Persistence Positional Encoding (PPE). In matched controls on eight TU graph benchmarks (TU8), changing only point representation shifts macro accuracy by 5.49 to 6.47 percentage points (pp), against 0.47 to 1.45 pp from changing the downstream module, identifying point representation as the dominant design variable tested. We further reevaluate vectorizations, kernels, learned representations, and Transformer-based models under one protocol with shared PD inputs, outer folds, and validation-only selection. Among all evaluated PD methods and downstream configurations, PPE achieves the highest mean on most benchmarks, including TU8, ORBIT5K, and MCB-C. Finally, we prove that grid averaging leaves a reconstruction error at any resolution, whereas a Transformer with fixed random Fourier frequencies is almost surely universal. PPE thus offers a minimal, Transformer-native baseline for persistence diagrams, and a common protocol for future comparison. --- ### GeoNIRS: Geometry-Aware Pretraining for Heterogeneous fNIRS Representation Learning Authors: Zongsheng Li, Shengyuan Zhu, Zimo Zhang, Hanyao Li, Shengming Wang, Quanying Liu Published: 2026-10-13 Categories: cs.LG arXiv: https://arxiv.org/abs/2610.17845v1 Functional near-infrared spectroscopy (fNIRS) provides portable and spatially localized measurements of cortical hemodynamics, but differences in channel count, probe layout, signal distribution, and task type hinder joint pretraining across datasets. We introduce GeoNIRS, a geometry-aware framework that uses source-detector locations as a shared spatial reference across heterogeneous probe layouts. GeoNIRS compresses variable-size channel sets into a fixed-size latent representation and employs bidirectional HbO/HbR cross-reconstruction with contiguous temporal masking. GeoNIRS-base is pretrained on eight heterogeneous datasets, while GeoNIRS-large scales to 22 datasets. Neither downstream data nor labels are used during pretraining, enabling evaluation under a strict cross-task setting. To our knowledge, GeoNIRS is the first fNIRS framework to explicitly model physical channel geometry during joint pretraining across heterogeneous channel configurations, and the first to provide complete source code and downloadable pretrained checkpoints. We evaluate GeoNIRS on eight held-out tasks spanning clinical and auditory assessment, pain and affective processing, semantic decoding, and motor function. GeoNIRS achieves the highest task-averaged balanced accuracy among the evaluated conventional machine-learning, fNIRS-specific, general time-series, and fNIRS pretraining baselines. Compared with random initialization, pretraining improves average balanced accuracy by 4.12 percentage points under linear probing and 3.30 points under full fine-tuning, with larger benefits when labels are scarce. GeoNIRS also remains robust to missing input channels. Geometry perturbations and pretraining ablations further show that both channel geometry and cross-chromophore reconstruction contribute to the learned representations. --- ### What Makes a Good Visual Representation for Spatial Understanding in MLLMs? Authors: Xiaohu Huang, Yumin Guo, Kai Han Published: 2026-10-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.17844v1 Spatial understanding is a necessary step in moving multimodal large language models (MLLMs) toward reasoning about the 3D physical world. Their visual pretraining data, however, is almost entirely 2D, so the visual tokens processed by MLLMs often lack inherent 3D structural information. A growing number of methods inject such structure by transferring representations from external vision or geometry models, and collectively establish that such transfer helps. Behind these methods lies a largely underexplored question: what makes a visual representation good for spatial understanding? To answer it, we first separate what is transferred from how it is transferred, and then trace the resulting differences back to properties of the representations themselves. What we transfer comes from 3D geometry models, self-supervised image and video models, vision-language models, and generative models, in the form of latent features or explicit depth and point maps. We transfer it either by distillation, using it as a supervision target, or by fusion, feeding it as an additional input. We find that a representation's 3D structure correlates strongly with how much it helps, measured by cross-view consistency for latent features and by depth and reconstruction accuracy for explicit geometry. We further show that the form of a representation dictates the mechanism it needs. Latent features gain as much from fusion as from distillation; explicit geometry gains little from fusion, and most of its distillation gain depends on a geometry decoding head. Guided by these findings, we distill both latent features and explicit geometry into a single MLLM. Taking only RGB video as input, it consistently improves over its base model on both static 3D and dynamic 4D spatial reasoning benchmarks. --- ### FASHION-RQ: Mining Hierarchical Semantic Structure for Automated Evaluation of Fashion Retrieval Authors: Tripti Shukla, Shwetha Somasundaram, Akshat Gupta, Rohit Garg, Paul Duff, Ian Anderson Published: 2026-10-13 Categories: cs.IR, cs.AI, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2610.17841v1 Evaluating fashion retrieval systems remains challenging due to the limited availability of benchmarks that capture the fine-grained, open-ended queries encountered in real-world search and recommendation. Existing benchmarks often rely on manual curation, predefined taxonomies, or synthetic prompts, limiting scalability and semantic coverage. We introduce FASHION-RQ: Mining Hierarchical Semantic Structure for Automated Evaluation of Fashion Retrieval, an automated framework that derives user-type queries and relevance sets directly from catalog structure. Our key insight is that hierarchical discrete representations learned by Residual Quantized Variational Autoencoders (RQ-VAE) organize products into semantically coherent clusters at progressively finer granularities. We leverage this structure to construct FASHION-RQBENCH, a scalable benchmark without manual query annotation or predefined taxonomies. We propose four complementary tasks for evaluating ranking, relevance classification, fine-grained sibling discrimination, and paraphrase robustness, alongside query- and attribute-level relevance metrics. Experiments across recent retrieval models demonstrate that FASHION-RQBENCH differentiates model capabilities across query types and attributes. A human study further shows that benchmark queries achieve comparable naturalness and higher intent clarity than real user queries, supporting their validity for real-world evaluation. The code and FASHION-RQBENCH dataset will be publicly released upon acceptance. --- ### Learning Adaptive Deliberation Control for Reliable Recognition with Multi-Agent Language Models Authors: Yuan Liu, Haoyue Peng, Fanliang Hu, Lei Huang, Wenli Cai Published: 2026-10-13 Categories: cs.AI arXiv: https://arxiv.org/abs/2610.17840v1 Multi-agent deliberation can improve recognition by criticizing and revising candidate answers, but fixed controls cannot respond to changes in disagreement or reasoning quality. We present an adaptive multi-agent controller (ADMA) that learns continuous policies for proposal, challenge, arbitration, and coordination while keeping the task model frozen. Shared trace features describe quality, agreement, and progress, while role-specific features support decentralized control during inference. One controller trained on legal and financial reasoning is transferred without policy updates to four textual and multimodal benchmarks and two successive task-model versions. With Qwen3.5 held fixed, ADMA-Full improves accuracy by 6.2-14.2 percentage points over Direct and by 2.8-3.0 points over ADMA-Fixed. Paired bootstrap ranges for the Full-Fixed comparison are positive on three tasks. Outcome decompositions show that recoveries exceed introduced errors, while process assessment records fewer false-consensus events and more supported correct outcomes. Transfer to Qwen3.6 remains task dependent, which bounds the generality of the learned policy. Median token and latency ratios are 14.3-18.7 and 7.8-16.5 relative to Direct. These results show that adaptive deliberation can improve fixed-model recognition when its task-dependent gains justify the additional computation. --- ### ProtoArena: Can AI Agents Discover and Exploit Real-World Cross-Component Protocol Vulnerabilities? Authors: Fasheng Miao, Zhaoyu Hu, Xiang Li, Changqing An, Jiutian Zeng, Longtao Huang, Haixin Duan, Jilong Wang Published: 2026-10-13 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2610.17837v1 Large language model (LLM) agents are increasingly used for vulnerability analysis, yet existing benchmarks provide limited coverage of exploitation that depends on interactions among network protocol components. Such vulnerabilities arise from inconsistencies in message interpretation, timing, state, or trust assumptions across participants. We introduce ProtoArena, a black-box, multi-component, dynamic benchmark with 80 cross-component vulnerability-exploitation environments spanning 11 protocol and application domains and 7 attack categories. Tasks are primarily informed by published protocol-attack mechanisms. Agents interact with running systems through scoped interfaces to achieve security-effect objectives without target source code. Three information levels assess guided reproduction, exploitation with mechanism and topology context, and effect-directed black-box exploration. A hybrid verifier combines rule-based checks with LLM review of independently collected traffic and logs. We evaluate five LLMs. Claude Opus 4.8 achieves the highest observed success rates: 48.8% under L1 guidance and 12.5% under effect-directed black-box L3. In a separate DNSBomb optimization experiment, it achieves a peak downstream-response to average query-input byte-rate ratio exceeding 171,000 times. The results show that agents can achieve verifiable exploitation effects in some cross-component environments, but success with limited black-box information remains low and varies across domains and attack categories. All experiments use controlled, explicitly authorized targets; they do not test production systems or disclose previously unknown vulnerabilities. --- ### A Tutorial on Quantum Amplitude Embedding with Data Re-Uploading Authors: Hyojun Ahn, Joongheon Kim Published: 2026-10-13 Categories: quant-ph, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2610.17836v1 Amplitude embedding lets a variational quantum model read many classical features with few qubits, but the model output is then only a quadratic form in the normalized input. Data re-uploading, which encodes the input repeatedly between trainable blocks, can be used to enrich such a model. Because every additional upload of an amplitude embedding repeats a costly state preparation, the gain that an upload buys has to be known before the gates are spent. This tutorial thus addresses the problem of explaining what data re-uploading adds to an amplitude-embedded model and how large the gain is. We express sequential re-uploading and parallel re-uploading with two copies in one notation and show that each upload raises the reachable polynomial order by two. Under one trainable circuit family and one budget of trainable parameters, simulation results show that re-uploading raises the test coefficient of determination from 0.689 to between 0.883 and 0.975 on targets of order four and from 0.337 to between 0.642 and 0.783 on targets of order six, on all ten seeds. On a quadratic target and on handwritten digits, one upload remains the best choice, and every additional upload lowers the gate error that the classifier tolerates. These results help designers match the number of uploads to the polynomial order of the task at hand. --- ### Agentic-MatriBO: Capability-Compiled Agentic Bayesian Optimization for Materials and Chemical Design Authors: Yu Zeng, Yuchuan Chen, Ziyue Zhang, Zijian Wei, Yanyi Su, Guojiang Zhao, Hongshuai Wang, Zhifeng Gao Published: 2026-10-13 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2610.17832v2 Materials and chemical optimization must choose the next synthesis, characterization, or high-fidelity simulation from sequential observations under a limited budget. Large language models can use task knowledge to propose adjustments, but composing surrogates, acquisition functions, and search strategies is constrained by compatibility, parameter dependencies, and evidence requirements. We present Agentic-MatriBO, a capability-compiled agentic Bayesian optimization (BO) framework that represents BO components and selection conditions as composable capabilities. A language policy proposes BO recipes and candidate plans; a compiler checks proposals against the registered capability catalog; and a deterministic execution layer validates and commits candidates. Oracle observations are stored in research memory and inform later rounds. We compare Agentic-MatriBO with eight baselines on four offline finite candidate pools---COF, PCE10, Buchwald--Hartwig (BH), and Suzuki--Miyaura---using five seeds per condition; results are reported descriptively. Across the four tasks, Agentic-MatriBO has the best mean endpoint quality after orienting each objective so that higher is better. On COF, it reaches the tabulated optimum in all five seeds both in the end-to-end comparison and after shared initialization, where it also retains an advantage in cumulative quality. Its normalized area under the best-so-far quality curve during the BO stage (nAUC-BO) is substantially higher on COF, near parity with the best baseline on PCE10 and BH, and lower on Suzuki. COF ablations show that removing the capability compiler or evidence gate coincides with lower cumulative quality and fewer exact-optimum hits. The framework provides an inspectable interface for evidence-conditioned BO adaptation. --- ### Diff Diffusion: Predicting Edit Paths with Diffusion Models for Efficient Code Editing Authors: Xiaolong Xu, Hao Jiang, Bo Jiang, Zhiwen Deng, Ningyuan Sun, Bingzhou Chen, Jin Chen, Chenxing Wei Published: 2026-10-13 Categories: cs.SE, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2610.17831v1 Diffusion large language models (dLLMs) are appealing for code generation because they can update many tokens in parallel, yet many are optimized for sequence synthesis rather than edit prediction. This setting is poorly matched to modern AI-assisted IDEs, where edit prediction must infer the next code change from recent edits, the current file, and cursor-local context. Such changes are usually sparse and preserve most source tokens, so regenerating a full code segment wastes computation. We propose Diff Diffusion, a diffusion-inspired LLM framework for code editing that predicts edit paths from an original snippet to its revised form. Diff Diffusion represents denoising as structured edit operations, and combines this formulation with multi-stage training and efficient parallel decoding. By focusing computation on changed regions, Diff Diffusion supports accurate low-latency edits and multi-position parallel modification. Across public edit-completion benchmarks, edit-prediction evaluations, and online validation, Diff Diffusion matches or leads autoregressive baselines while delivering 2.78 times faster inference on public edit-completion benchmarks and a 39.0% P50 model-latency reduction in production. This work provides a new modeling perspective for AI applications that require high accuracy and low latency. --- ### QK-Edit: Revisiting Attention-based Injection in MM-DiT for Image and Video Editing Authors: Tiancheng Shen, Zilong Huang, Xiangtai Li, Zhijie Lin, Jiyang Liu, Yitong Wang, Jiashi Feng, Ming-Hsuan Yang, Jun Hao Liew Published: 2026-10-13 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2610.17829v2 Multimodal Diffusion Transformers (MM-DiTs) have recently emerged as a powerful framework for unified text-vision synthesis, surpassing traditional U-Net architectures in generative tasks. One key innovation lies in its Multimodal Self-Attention (MM-SA) interaction, where image and text tokens are concatenated and processed via self-attention. However, this mechanism poses significant challenges for editing, rendering conventional U-Net-based attention manipulation methods ineffective. To address this limitation, we propose QK-Edit, a training-free framework that exploits the unique attention dynamics of MM-DiTs for precise text-guided image and video editing. By introducing a novel query-key manipulation strategy, our method isolates and adjusts critical attention components to achieve an optimal balance between prompt fidelity and structural consistency. This enables seamless edits across various tasks, including object addition, object removal, object replacement, changing background, changing material, changing color, and style transformation. Notably, it can be easily implemented with feature replacement in inference. QK-Edit demonstrates superior editing performance on state-of-the-art models, such as FLUX and HunyuanVideo, effectively bridging the gap between generative power and editable flexibility in MM-DiTs, and paving the way for scalable multimodal content creation. --- ### Benchmarking and Training Reward Models for Cross-Source Evidence Grounding in Search Agents Authors: Haochen Zhao, Minsheng Hao, Zheng Li, Yifan Li, Zikang Liu, Bowen Qin, Yuhao Cheng, Chuang Fan, Haosheng Zou, Jeff Chen, Sujian Li Published: 2026-10-13 Categories: cs.AI arXiv: https://arxiv.org/abs/2610.17827v1 Existing approaches to training and evaluating search agents focus heavily on final-answer quality, largely neglecting how generated answers map back to retrieved evidence. Recent methods track citations to enforce grounding, but citation support alone does not establish source independence or whether an answer appropriately handles conflicting evidence. We argue that credibility in open-web retrieval depends on principled cross-source evidence grounding that accounts for source independence and systematically attributes and adjudicates conflicting evidence, while hierarchical judgments of these grounding decisions provide a unified signal for training and evaluation. To this end, we propose an evaluation framework, Cross-Source Evidence Grounding (CSEG), that employs a fine-grained, six-layer decision tree over the complete search trajectory. We introduce CSEG-Bench and trainable reward models to provide targeted supervision following this framework. Our approach helps eliminating the severe data skew of natural trajectories by providing targeted coverage over complex, deep-layer decision paths, such as cross-source conflict resolution and nuanced attribution, that standard rollouts fail to capture. Specifically, (1) we define the hierarchy at label, node, and path levels; (2) we construct 1,160 verified trajectories over nine canonical paths by rewriting answers on real retrieved document pools, with independent frontier judges reproducing the entire gold path and human experts deciding inclusion; (3) we train CSEG-RM, which reaches 83.9% exact accuracy on our benchmark while lying on the performance-token efficiency Pareto frontier. Experiments that use this reward alone to supervise a frozen-retrieval answer policy raise the independently judged mean label from 2.83 to 4.19, and show that learning the hierarchy generalizes across layers at a fraction of frontier-judge cost. --- ### Checking the Checkers: A Synthesis of Verifier Noise and Robustness in RLVR Authors: Savyasachi Gupta, Aniruddha Srinivas Joshi Published: 2026-10-13 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2610.17824v1 Reinforcement learning with verifiable rewards (RLVR) has become a principal recipe for improving the reasoning of large language models (LLMs). It rests on the assumption that the checker scoring each response is itself correct. In practice, rule-based checkers reject correct answers in unexpected formats, unit tests accept incorrect programs, and LLM judges can be led to approve empty or adversarial responses. This work surveys 58 studies that measure, analyze, or mitigate such verifier errors. We ask what errors verifiers make, how they affect training, which mitigations work against which errors, and how the findings extend to self-generated, rubric- and judge-based, and verifier-free rewards. The evidence suggests that the structure of an error matters more than its rate. Random errors mostly slow learning, though under a fixed budget or KL regularization they also lower final accuracy. False positives that the policy can learn to trigger can cause plateaus or collapse, sometimes even with a low initial error rate. Whether a verifier wrongly rejects or accepts depends on what it checks and who built it, not on its family. Its errors shift as the policy improves, and most mitigations have been tested against only one kind of error. We compile the findings into comparison tables, reconcile apparent conflicts in the literature, and propose a reporting protocol to make future results comparable. --- ### Same Attack, Different Judge: A First Controlled Adversarial Comparison of Jev, LLM Judges, and Purpose-Built Safety Classifiers Authors: Xiang Li, Aichun Huang, Mingming Zhang, Zheli Liu Published: 2026-10-13 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2610.17820v1 TypeSafe AI's Jev is a new kind of safety judge. It generates no text and no label: it answers questions the developer defines in advance and returns a probability for each answer. The vendor admits that adversarial input can change these answers but gives no numbers. In this paper, we present the first controlled adversarial evaluation of Jev. We build a test harness that sends every system the same payload under the same policy, and run it on four public-data tasks, from rule-based ones (input moderation, injection detection, routing) to a judgment-based one (phishing detection), with fifty attack variants from thirteen families. The targets are Jev, five GPT and GLM judges, and two local classifiers (Llama Guard 3, Prompt Guard 2). We find that Jev is the most robust system on input moderation: eight forgeries of its structured input bypass none of its decisions, and a sentence asking only why someone might make a harmful request bypasses 9% of its blocks, against 56% for GPT-6 Astra. On phishing detection. However, the same forgeries bypass 19-29% of the emails it flags, and even without attack Jev detects only 62% of phishing emails, against 77-81% for the GLM judges. Jev is also the fastest judge, but it flags the most benign requests (27%). The local classifiers fail differently: a training-example statement makes Llama Guard 3 miss 62% of the phishing it had flagged. These results suggest that Jev resists attacks on tasks with a clear rule to check but is more exposed on tasks that call for an overall judgment. --- ### Follow the Perturbed Leader for Linear Bandits: $\sqrt{T}$ Regret over General Convex Sets Authors: Chen-Yu Wei Published: 2026-10-13 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2610.17818v2 We give the first adversarial linear bandit algorithm based on Follow-the-Perturbed-Leader (FTPL) that achieves $\sqrt{T}$ regret over general convex action sets. Our algorithm attains $\widetilde{\mathcal{O}}(d^{5/4}\sqrt{T})$ regret using only linear optimization oracles, and the near-optimal $\widetilde{\mathcal{O}}(d\sqrt{T})$ regret when an optimal-design exploration distribution is further available. This improves on the $T^{2/3}$ regret of the FTPL algorithm of Dani and Hayes (2006) for general convex sets, and generalizes the $\sqrt{T}$ regret of Lévy et al. (2026) over two special action sets (hypercube and Euclidean ball). Compared to previous oracle-efficient linear bandit algorithms based on continuous exponential weights, the FTPL approach generally has better computational efficiency, and is much simpler to implement. The key algorithmic element is a perturbation distribution that adapts to the past reward estimators. Its simplest form, an adaptive Gaussian perturbation, is straightforward to implement and already attains the regret bound above. We further extend it to two general constructions, random-weighting and elliptical perturbations, each covering distributions from light tails (Gaussian) to heavy tails (Student-$t_2$). On the technical side, we develop a systematic way to analyze the stability term in the FTPL regret through the Fisher information of the perturbation distribution. This builds a general framework for FTPL beyond existing ones such as the hazard rate and the FTRL-FTPL duality. --- ### Empowering Edge Agents: A Systematic Analysis of Edge-Cloud Collaboration in LLM Agent Execution Authors: Pengcheng Huang, Zhenghao Liu, Shixin Ren, Shi Yu, Sijia Yao, Yukun Yan, Chaojun Xiao, Maosong Sun Published: 2026-10-13 Categories: cs.AI arXiv: https://arxiv.org/abs/2610.17816v1 Edge-cloud collaboration combines the lower inference overhead and greater data locality of edge-side execution with the stronger capabilities of cloud-hosted models. However, as LLM agents increasingly operate in long-horizon scenarios, cloud participation becomes interleaved throughout execution rather than a one-shot decision for each query. Such interleaved participation introduces trajectory-level effects that remain insufficiently understood. To address this gap, we introduce EC2-Bench, a benchmark for evaluating edge-cloud collaboration in long-horizon agent execution. It provides executable digital-workspace agent tasks and a unified evaluation framework covering task performance, resource usage, and privacy exposure. Using this benchmark, we empirically study diverse forms of edge-cloud collaboration. Our study reveals three findings. First, collaboration achieves a favorable balance among task performance, resource usage, and privacy exposure compared with edge-only and cloud-only execution. Second, as edge models become stronger, collaborative execution approaches cloud-only performance, and performance differences across strategies shrink, while privacy exposure remains strongly strategy-dependent. Third, stronger edge capability shifts preferred collaboration from externally controlled cloud participation and direct cloud execution toward more edge-controlled collaboration. Further analyses identify cloud-invocation decision-making and instruction following as important edge-side abilities for effective collaboration. These findings provide a systematic understanding of edge-cloud collaboration in long-horizon agent execution and insights for designing future edge agents that effectively collaborate with cloud models. Code and resources are available at https://github.com/OpenBMB/EC2-Bench. --- ### Perturbation, Not Provenance: Why Activation Steering Does Not Make a Robust Text Watermark Authors: Sergey Pletenev Published: 2026-10-13 Categories: cs.CL arXiv: https://arxiv.org/abs/2610.17810v1 Linear directions in the activation space of large language models causally and universally influence model behavior. This suggests that they might serve as a basis for a robust text watermark. The idea is to steer the generation along a specific semantic direction and then recover the mark by probing the activations of the re-encoded text. We test this hypothesis by conducting a series of experiments. We vary the strength of the steering and use four different directions (formality, sentiment, complexity, and a random control) across two different model families. We also test two paraphrasing attacks: a sentence-level one and a stronger paragraph-level one. The results show that activation-steering watermarks are outperformed by a standard logit-based watermark (KGW) in terms of detection accuracy. At a matched level of fluency (within 1.15 times the original text's perplexity), KGW achieved an AUC of 0.81 against a sentence-level paraphraser, while the best steering direction only reached 0.65. A stronger paragraph-level paraphraser lowered both (0.65 for KGW and 0.59 for steering), narrowing the gap but not reversing it. On Qwen, it left the two indistinguishable but at near-chance levels for both. We further showed that the detectability provided by steering was not semantic: a random direction was just as detectable as a meaningful one. The signal was independent of which layer was probed and whether a surface feature control was used. We conclude that, in this case, semantic activation does not encode a localized, recoverable semantic mark, but rather a global, content-independent shift that can only be strongly detected when the text is degraded. --- ### SPARCL: Spatially Adaptive Regional Coverage for Active Learning in Chest X-ray Analysis Authors: Juho Jung, Seongjae Kang, Mijoo Kim, Sungduk Cho, Yunkyoung Jun, Sunghoon Joo, Yoona Hwang Published: 2026-10-13 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2610.17808v2 Pixel-level annotation for chest X-ray (CXR) lesion segmentation is expensive, making active learning (AL) essential. However, existing AL strategies fail to account for the spatial distribution of selected samples, resulting in inconsistent performance across focal (e.g., aortic calcification) and diffuse (e.g., nodules) pathologies. In this paper, we propose SPARCL (Spatially Adaptive Regional Coverage for Active Learning) to dynamically balance predictive uncertainty with cumulative spatial coverage. By employing a concave coverage function, SPARCL imposes diminishing returns on previously annotated regions, naturally directing the annotation budget toward under-labeled areas. Furthermore, an entropy gate reverts the algorithm to standard uncertainty sampling when spatial variation is absent. This coverage objective is monotone submodular, guaranteeing near-optimal greedy batch selection. To address how aggregate metrics like Dice obscure regional failures, we further introduce two metrics---Spatial Inconsistency (SI) and Performance Coverage (PC)---to quantify anatomical fairness across AL rounds. Evaluated on an in-house calcification cohort and public CXR benchmarks, SPARCL reaches 95% of fully supervised performance using only 24% of the data, compared to 29% for the best baseline. Ultimately, SPARCL ensures robust spatial reliability, surfacing regional gains that aggregate metrics miss. --- ### BRIDGE: Selective Boundary Repair with Dependency Graph Evidence in Multi-Agent Systems Authors: Encheng Su, Pengze Li, Ziheng Hua, Qiucheng Yu, Jianyu Wu, Chen Tang, Yizhou Wang, Lintao Wang, Junjia Xiang, Xinzhu Ma, Xi Chen, Shixiang Tang et al. Published: 2026-10-13 Categories: cs.CL arXiv: https://arxiv.org/abs/2610.17805v1 Multi-agent collaboration has emerged as a critical paradigm for tackling complex, long-horizon tasks; however, while coordination protocols continue to advance, systemic failures at the inter-agent collaboration interface remain a pervasive bottleneck. In collaborative workflows, systems frequently suffer from \emph{broken handoffs}---where critical task context is omitted, misrouted, or ignored despite each agent emitting locally plausible responses. To address this challenge, we present \method, a verifiable runtime control framework that treats handoff repair as selective intervention over explicit dependency graphs. Central to \method is an integrated fault-attribution and calibration pipeline, \dataset, which decouples execution dependencies and injects counterfactual handoff perturbations to provide empirical supervision for runtime intervention. Leveraging this diagnostic foundation, \method dynamically compiles observable interaction traces into a provenance graph to pinpoint unresolved dependencies, without accessing private agent states. It then evaluates candidate repairs through an empirical gate that weighs expected gains against disruption risks, and executes repairs only when formal local postconditions hold. Extensive evaluations across demanding benchmarks demonstrate that \method boosts task success by up to 12.3 absolute percentage points on SILO-BENCH and 11.1 on TeamBench. Across both backbones, \method maintains strict cost-effectiveness, adding as little as 2.0\% marginal token overhead. Furthermore, on 562 handoff-relevant traces from the MAST benchmark, \method achieves 69.8\% F1 in a single pass, establishing a reliable operational boundary for verifiable coordination repair. --- ### Learning from fracture: Teaching language models to design through physical experience Authors: Bo Ni Published: 2026-10-13 Categories: cond-mat.mtrl-sci, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2610.17804v1 How can language models learn to optimize physical designs through experience? We introduce FracLearn, an experience-abstraction-generalization framework evaluated through a deliberately unfamiliar, Tetris-inspired fracture game. The model first proposes architectures and examines simulated fracture responses to accumulate experience. Abstraction organizes these observations into conditional principles linking reinforcement placement, crack evolution and load transfer, with supporting evidence and failure conditions. Generalization challenges the model to combine these principles in new architectures and surpass the earlier design record. Campaigns demonstrate progressive refinement and recovery through hypothesis revision, yielding a best composite score 11.0% above the experience-stage record. Fracture analyses connect improved performance to crack deflection and temporary trapping as toughening mechanisms, alongside greater residual load capacity. Learning occurs through external knowledge and decisions, without model-weight updates. This approach offers a route to scientific agents that turn computational experiments into reusable physical knowledge for materials discovery and engineering design. ---