# Research Papers — 2026-10 · bundle page 89 of 156 > Full-text markdown of papers #4401–4450 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/89 [← Previous bundle](https://aipod.dev/papers/archive/2026-10/page/88.md) [Next bundle →](https://aipod.dev/papers/archive/2026-10/page/90.md) ### MCL: Meta Convolution Layer Authors: Naim Reza, Md Al Amin, Ho Yub Jung Published: 2026-10-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.11117v1 Dynamic convolution enhances convolutional neural networks (CNNs) by adapting kernels to input content, but it expresses the effective kernel as a linear mixture of a small number of basis kernels, which limits expressivity and complicates optimization as the mixture size grows. In this work, we revisit dynamic convolution from a functional perspective and propose the Meta Convolution Layer (MCL), which directly models the convolutional kernel as an input-conditioned function W(x) realized via a high-order polynomial expansion. Leveraging nested residual blocks inspired by deep polynomial networks, MCL implements a structured polynomial meta-network that generates a single input-adaptive kernel, thereby decoupling representational power from the explicit number of mixture kernels and alleviating training instability. MCL is a plug-in addition with standard convolutions and can be seamlessly integrated into both CNN and transformer backbones. Experimental evaluation shows that adding MCL improves the Top-1 accuracy of Resnet- 18, Resnet-50 and ResNet-101 by 6.61%, 3.42% and 3.05% on the ImageNet dataset. Moreover, the proposed method significantly boosts the accuracy of Resnet and Wide-Resnet variants on CIFAR-10 and CIFAR-100 datasets. Additionally, the proposed method outperforms previous methods on fine-grained visual classification tasks using Swin and ViT backbones. These results demonstrate that high-order polynomial kernel generation is a powerful and scalable alternative to linear mixture based dynamic convolution. --- ### Dynamics as Code: On Model Compression via Dynamic System Authors: Fan Gao, Wei Su, Juntong Fan, Renfeng Peng, Hongyu Liu, Jinqiao Duan, Feng-Lei Fan Published: 2026-10-08 Categories: cs.LG arXiv: https://arxiv.org/abs/2610.11115v2 The escalating size of pretrained neural networks has rendered model compression a prerequisite for deployment under stringent memory and compute constraints. With the irrational winding as an example, earlier work introduced a dynamic system (DS) paradigm that reconceptualizes compression as compact weight representation: high-dimensional parameters are encoded by the index of a trajectory produced by a dynamic system, from which the vector is recovered during decompression. This mechanism is fundamentally distinct from pruning, quantization, knowledge distillation, and low-rank decomposition. Along this direction, we prove that under a Diophantine condition, a finite trajectory of \(M = O(ε^{-(d+ν)})\) states in the irrational winding constitutes an \(ε\)-net over the \(d\)-dimensional weight space, thereby linking state resolution, decompression error, and compression ratio in a predictable manner. Furthermore, we propose a generalized DS-based model compression framework by unifying four DS families---space-filling curves (Hilbert, Peano, Morton/Z-order, Snake), chaotic systems (Lorenz), congruential and pseudo-random generators (LCG, PCG), and low-discrepancy sequences (Halton). Also, we introduce the KD-tree and coordinate-template acceleration to scale to large models as well as outlier identification to control the error. Experiments on ResNet-18 and Qwen2.5-1.5B/Qwen1.5-7B validate that DS-based compression achieves competitive compression ratios without post-hoc retraining, with controllable decompression error and flexible state-space design, establishing it as a principled and practical compression approach. --- ### DiscoVL: Unveiling Disentangled C ross-Modal Representation Learning via Orthogonal Adversarial Regularization for V ision-Language Models Authors: Mengping Dong, Jinbao Li, Fei Li Published: 2026-10-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.11113v1 Pre-trained vision-language models excel across varied perception tasks, but adapting them to novel downstream settings without sacrificing generalization remains non-trivial. Existing parameter-efficient prompt learning method often yields inconsistent representations and fails to account for semantic distribution shifts. In this work, we present DiscoVL, a disentangled cross-modal representation learning framework that couples orthogonal adversarial regularization with structured cross-modal alignment for vision-language models. To address the insufficient cross-modal interaction, our DiscoVL designs a multi-branch low-rank residual aligner that decomposes representations into subspaces and enables bidirectional cross-modal feedback between visual and textual streams at each layer. Furthermore, while conventional triplet constraints overfit features to class centroids, we design an orthogonal regularization for adversarial triplet loss, which prevents centroid collapse and substantially boosts generalization. Evaluations on 15 benchmarks demonstrate that DiscoVL delivers consistent improvements over state-of-the-art methods for base-to-novel generalization, cross-dataset evaluation, and few-shot learning --- ### False Claims, Credible Images: A Red-Teaming Benchmark for Commercial Image Generators Authors: Zeyu Ye, Yanchun Li, Sibei He, Meng Xie, Hangtao Zhang, Xianlong Wang, Li Zeng, Jiahao Chen, Yichen Wang, Junhui Wang, Ziqi Zhou Published: 2026-10-08 Categories: cs.CR, cs.CV arXiv: https://arxiv.org/abs/2610.11112v1 Image-generation models can now produce text-rich, natural-looking visual artifacts that are hard to distinguish from real-world evidence, such as news reports and textbook pages. Yet, the same capability introduces a new risk: these models can just as easily fabricate visual misinformation. Even commercial models (e.g., GPT-Image-2) readily produce it. Curiously, we find that these models can recognize a claim as false when asked, yet still render that very claim as credible visual evidence. This discrepancy points to a blind spot in current alignment: safeguards judge what an image shows, not what it asserts; however, existing red-teaming benchmarks target conventional harmful content, such as violent or explicit imagery, and say little about where the alignment boundaries lie for visual misinformation, especially in commercial models. To fill this gap, we introduce EpiReal-Bench, the first systematic benchmark for evaluating visual misinformation risks in commercial image generators, comprising 10k false-claim prompts and 10k corresponding generated images that span 10 real-world claim categories and 10 credible visual formats. We further introduce EpiReal-Attack, a skill-guided black-box optimization framework that uses Pareto-based selection and multimodal feedback to identify commands that bypass alignment safeguards while preserving visual realism, textual legibility, and semantic fidelity. Experiments on four commercial models reveal that more than 70% of false-claim prompts elicit images that faithfully depict the corresponding misinformation, and EpiReal-Attack pushes this rate to 95%. Most worryingly, these models are only a click away, and their outputs are cheap to spread yet hard to disbelieve, leaving this dimension of alignment largely unguarded. --- ### Lapras: Latent Reasoning for Time Series Language Models Authors: Yuliang Chen, Yu Yvonne Wu, Patrick Langer, Arvind Pillai, Sudarshan Regmi, Martin Maritsch, Juncheng Liu, Robert Jakob, Thomas Kaar, Tess Z. Griffin, Lisa Marsch, Michael V. Heinz et al. Published: 2026-10-08 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2610.11111v1 Time Series Language Models (TSLMs) offer a promising path toward time series understanding by reasoning over temporal signals and producing natural language answers and explanations. A common approach is Chain-of-Thought (CoT), which generates step-by-step rationales linking relevant signal patterns to final answers. Although these models learn from reference CoT traces during post-training, generating faithful descriptions of input time series at inference remains challenging. Expressing high-dimensional, continuous temporal representations in discrete language tokens may cause the model to neglect task-relevant patterns or describe them inaccurately. Because later reasoning steps build on these descriptions, early errors propagate, leading to incorrect answers with plausible explanations that are inconsistent with the input signal. We propose Lapras (Latent Post-trained Reasoning Across Series), a post-training framework that equips TSLMs with latent reasoning. A model trained with Lapras reasons through a sequence of continuous thoughts in the joint time series-language space, producing text only for the final answer. It learns this through teacher-student self-distillation, where a teacher trained on CoT reference traces reasons explicitly through text. The student aligns its hidden states with the teacher's at the answer stage, transferring the teacher's reasoning ability into its latent computation. We evaluate Lapras across four TSLM backbones on five time series question answering benchmarks. Lapras improves average F1 by up to 10.79% over explicit CoT while generating 23.9x fewer tokens. Lapras's continuous thoughts can also be decoded into readable reasoning traces via standard language decoding, preserving textual explanations. Together, these results highlight Lapras as a promising post-training paradigm for efficient, effective, and interpretable TSLM reasoning. --- ### Cova-PINN: Cross-Domain Conservation Physics-Informed Neural Network for Fluid-Solid Conjugate Heat Transfer in Complex Geometries Authors: Weizheng Zhang, Xunjie Xie, Hao Pan, Lin Lu Published: 2026-10-08 Categories: cs.LG, physics.flu-dyn arXiv: https://arxiv.org/abs/2610.11108v1 Multi-domain physics-informed neural networks (PINNs) flexibly model medium-specific representations to solve fluid--solid conjugate heat transfer (CHT). However, standard multi-domain PINNs enforce governing equations and interface conditions on separately sampled domain supports, which can yield plausible temperature fields but inaccurate end-to-end energy transfer and outlet temperatures. We propose Cova-PINN, a multi-domain PINN framework that aligns conservation support with thermal interaction paths in complex geometries. Cova-PINN jointly optimizes cross-domain composite control-volume balances at the local scale and paired-wall closure at the global exchanger scale. We evaluate Cova-PINN on four triply periodic minimal surface (TPMS) heat exchangers and a geometrically distinct DualMS design against CHT-specific, optimization-oriented, and complex-geometry PINN baselines under a common protocol. Relative to the closest baseline, MUSA-PINN-CHT, Cova-PINN reduces average outlet-temperature and device-level closure errors across the four TPMS topologies by $37.7\%$ and $60.2\%$, respectively, while also improving full-field and heat-duty accuracy, with consistent gains on DualMS. --- ### TKCAM: Text and Keyframe to Camera Trajectory Generation Authors: Haozhe Yang, Zhiyang Dou, Zekai Gu, Cheng Lin, Wenping Wang, Yuan Liu, Taku Komura Published: 2026-10-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.11105v1 Generating high-quality and controllable camera motion is essential for AI-assisted cinematography, video synthesis, and 3D scene understanding. We introduce TKCAM, a Text- and Keyframe-conditioned CAMera-motion synthesis framework based on generative masked modeling. We represent camera dynamics using a 12-dimensional kinematic feature comprising position, velocity, and a continuous rotation representation and discretize them into hierarchical motion tokens via a Residual Vector Quantizer (RVQ). A two-stage masked transformer architecture then learns to reconstruct and refine these tokens, utilizing explicit self- and cross-attention modules for multimodal conditioning. A central feature of our framework is sparse visual keyframe conditioning: users can provide free-form text prompts together with RGB observations at selected timestamps, which provide temporally localized visual guidance for generating coherent in-between trajectories. Furthermore, to advance evaluation standards, we curate RealEstate10K-Cap, a large-scale text-camera dataset, and establish a cross-domain benchmark with a Universal CLaTr Evaluator. Extensive experiments demonstrate that TKCAM surpasses recent state-of-the-art baselines on Fréchet distance (FID), text-motion matching scores, and retrieval metrics (R@K), while additional analyses evaluate temporal smoothness and cross-domain generalization. Code is available at https://github.com/linearalgebrayhz/TKCAM. --- ### Skeleton-Guided Progressive Test-Time Adaptation for Thin Curvilinear Structures Authors: Boa Jang, JunGyu Lee, Gwanho Lee, Jinwook Choi, Young-Gon Kim Published: 2026-10-08 Categories: cs.CV, eess.IV arXiv: https://arxiv.org/abs/2610.11104v1 Accurate segmentation of thin curvilinear structures is vital for various real-world applications, from vessel analysis to road extraction. Yet their intricate geometry makes even minor pixel-wise errors enough to break the global topology, and this structural fragility turns severe domain shifts into catastrophic failures. The difficulty is most acute under cross-modality gaps, where the imaging process itself differs fundamentally between source and target. While test-time adaptation (TTA) offers a practical source-free remedy, existing methods adapt feature statistics and confidence, neither of which constrains connectivity, and thus degrade under such extreme gaps. To address this, we propose Skeleton-Guided Progressive Test-Time Adaptation (SGP-TTA). Progressive Batch Normalization (ProgBN) shifts normalization from frozen source statistics toward current target estimates under a sample-count schedule, so that the source-target balance follows the stage of adaptation rather than a fixed coefficient. Consensus Skeleton Recall (CSR) then derives a structural target from geometrically aligned multi-view predictions and updates only the BN affine parameters to preserve connected structures. Extensive experiments show that SGP-TTA consistently outperforms existing TTA methods in topological connectivity, with the largest margins under cross-modality shift. The project page is available at https://boa-jang.github.io/SGP-TTA. --- ### Continuous Ground-Truth Construction and a Recovery Policy for Air--Water Robotic Tracking Authors: Jiangong Xiao, Zhe Sun, Kanzhong Yao, Yuanbo Bi, Haofei Zhao, Ruixuan Hu, Guan Huang, Xuelong Li Published: 2026-10-08 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2610.11096v2 Visual tracking across the air-water interface is challenged by splashes, bubbles, refraction, reflections, and abrupt appearance changes that can temporarily invalidate observations. This setting poses two coupled difficulties: first, for evaluation, image-only annotation cannot reliably describe the target's physical location during visual blindness; second, for online tracking, corrupted observations can contaminate motion estimates and appearance templates. We address the first difficulty with a construction pipeline that synchronizes camera frames with motion-capture poses, projects known target geometry, corrects underwater projection with a medium-gated residual, and subjects the annotations to manual review. This yields an evaluation-only cross-medium test set of 22,346 frames. We further introduce a Cross-Medium Recovery Policy (CMRP) centered on confidence-triggered template selection. It supplies MixFormerV2 with the fixed initial template, a window-best pre-trigger template, and a trigger-frame Kalman-guided image crop, together with their associated weights, without retraining the visual backbone. In the accuracy evaluation, CMRP achieves 49.90 Macro Success AUC, 2.95 points above MixFormerV2 Official. On selected cross-medium transition and occlusion-recovery intervals, CMRP increases MixFormerV2 tracking coverage from 47.91\% to 50.43\% relative to Official updating, while mean loss-to-recovery latency over successfully recovered videos decreases from 55.3 to 49.3 frames. --- ### Beyond Imitation: A Framework and Benchmark for LLM-Assisted Peer Review Authors: Rachel S. Y. Teo, Yutaro Yamada, Shashank Kotyan, Yuki Imajuku, Tarin Clanuwat Published: 2026-10-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2610.11087v1 The rapid growth of scientific publishing has strained peer review, particularly in machine learning, raising concerns about declining review quality and increasing reviewer workload. Large language models (LLMs) have been proposed as automated review assistants, yet their evaluation has focused largely on imitating human-written reviews rather than supporting the core functions of peer review. Here, we introduce a verification-centric perspective on LLM-assisted peer review, emphasizing error detection as a critical and resource-intensive task. We present a scalable benchmark that evaluates review systems' ability to identify logical contradictions, constructed through synthetic insertion of errors into conference papers, yielding unambiguous evaluation targets and enabling systematic comparison. We further propose a Multi-Layered Review (MLR) framework that prioritizes detailed manuscript comprehension before review generation, aligning more closely with human reviewing practices while improving token efficiency. Across evaluations, our approach demonstrates strong alignment with human review scores, achieves high error detection performance, and provides complementary perspectives on reviewer focus. These improvements can be attributed to both the choice of the underlying LLM and the design of our system. At the same time, we corroborate persistent vulnerabilities to adversarial manipulation, underscoring the need for robustness in automated review systems. Our findings highlight the importance of rigorous, error-focused evaluation to guide responsible deployment of LLM-based tools in peer review and other critical scientific workflows. --- ### Contrast Enhancement or Noise Reduction? On Improving Cervical Cancer Classification Authors: Ach Khozaimi, Ulfatun Nahdhiyah Published: 2026-10-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.11086v1 Purpose: Cervical cancer is one of the leading causes of mortality worldwide. Deep learning has shown promising performance in medical image classification. The influence of image preprocessing algorithms on classification performance remains insufficiently investigated in the literature. This research aims to evaluate the impact of image preprocessing algorithms on the performance of CNNs for Pap smear image classification. Methods: Three CNN architectures (ResNet-34, MobileNet-V2, and DenseNet-121) were trained and evaluated using the SIPaKMeD dataset. Two preprocessing algorithms were applied: the PMD filter for noise reduction and CLAHE for contrast enhancement. The model performance was assessed using a confusion matrix. Results: Preprocessing improved the classification performance of all models. CLAHE significantly increased the accuracy of ResNet-34 from 76.73% to 84.16% and DenseNet-121 from 76.73% to 84.16%. The PMD filter yielded limited improvement and slightly reduced the MobileNet-V2 performance. Novelty: This research provides a systematic comparison of contrast enhancement and noise reduction techniques across CNN architectures. This research demonstrates that contrast enhancement is more effective than noise reduction in improving CNN performance. The research provides new pipelines for improving cervical cancer classification. --- ### DaCe-DT: Data-Centric Offline Multi-Task Reinforcement Learning via Adaptive Prompts and Trajectory Correction for Heterogeneous Tasks Authors: Xinfei Wang, Shanchen Pang, Chenhao Zhang, Shudong Wang, Wenhao Ji, Haiyuan Gui, Meng Han, Xiaojian Liao Published: 2026-10-08 Categories: cs.LG arXiv: https://arxiv.org/abs/2610.11085v1 Offline multi-task reinforcement learning (Offline MTRL) heavily depends on the quality and distribution of pre-collected data. However, existing methods mainly focus on algorithmic optimization, with less emphasis on data-level improvements to enhance learning ability and generalization performance. This paper, from a data perspective, reveals three key bottlenecks that limit Offline MTRL performance:(i) ineffective utilization of prompts length under diverse task complexities, and (ii) semantic irrelevance of randomly sampled prompt segments, (iii) misleading supervision induced by fragmented and discontinuous trajectories. To address these challenges, we propose DaCe-DT, a robust offline MTRL framework designed to be insensitive to heterogeneous task complexities and data quality, featuring length-gated prompt masking (LGPM), retrieval-augmented prompt construction (RAPC), and value-adaptive return calibration (VARC). Together, these mechanisms enable DaCe-DT to deliver data-centric prompt adaptation and trajectory refinement, resulting in robust multi-task generalization and stable policy learning amid heterogeneous offline data and tasks. Experimental results on Meta-World show that DaCe-DT consistently outperforms state-of-the-art methods, achieving an average improvement of 11.73% on optimal datasets and an improvement of 13.34% on suboptimal datasets, demonstrating its effectiveness in learning stably from imperfect data and improving overall multi-task performance. --- ### A General $\widetildeΩ(\sqrt{T γ_T})$ Lower Bound for Kernel Bandits Authors: Chenkai Ma, Jonathan Scarlett Published: 2026-10-08 Categories: stat.ML, cs.IT, cs.LG arXiv: https://arxiv.org/abs/2610.11082v1 The kernel bandit problem consists of sequentially optimizing an unknown function with noisy feedback, where the function has bounded norm in a given Reproducing Kernel Hilbert Space (RKHS). A central quantity in the regret analysis of kernel bandits is the maximum information gain $γ_T$. In particular, the best existing upper bounds scale as $\sqrt{Tγ_T}$ up to log factors, and nearly-matching lower bounds have been derived for specific kernels such as squared exponential and Matérn. However, lower bounds for general kernels are lacking, thus making it unclear in what generality the upper bounds are near-optimal. In this paper, we establish a general $Ω(\sqrt{Tγ_T/\log T})$ minimax regret lower bound for non-constant continuous kernels on compact domains, establishing near-optimality (within log factors) in a very general sense. We show that the log factor appearing in this bound is unavoidable in general, but that it can be removed under certain conditions. Among other things, our findings imply that the minimax-optimal scaling is exactly $Θ(\sqrt{Tγ_T})$ (i.e., within constant factors) for the Matérn-$ν$ kernel with $ν\in (0,2)$, $γ$-exponential kernel with $γ\in (0,2)$, and certain piecewise-polynomial kernels. --- ### Stability-Plasticity Balance via Singular-Vector Selection in LLM Continual Learning Authors: Lingxiang Wang, Hainan Zhang, Liang Pang, Hongwei Zheng, Zhiming Zheng Published: 2026-10-08 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2610.11076v1 Domain-specific continual adaptation of LLMs risks catastrophic forgetting, creating a fundamental tension between acquiring new capabilities and preserving those learned during pretraining. PEFT mitigates this problem by restricting the number of trainable parameters, but existing methods lack a principled unit for deciding where plasticity should be allocated and stability should be preserved. We identify the singular-vector channel as a natural unit for managing this trade-off. Each channel represents an input-output transformation, which can be updated to acquire new knowledge or fixed to preserve pretrained capabilities. Based on this perspective, we introduce SVC, a parameter-efficient continual-learning method that selectively updates Singular-Vector Channels. Before fine-tuning, SVC uses domain-specific data to estimate each channel's adaptation benefit and a fixed public general-domain corpus only as a history activation proxy for estimating forgetting cost. It then adaptively selects trainable channels based on these scores via knee-based cost screening, Pareto-front filtering, and Otsu thresholding. Experimental results across four LLM families and eight downstream tasks show that SVC better preserves pretrained capabilities while achieving strong downstream performance relative to existing PEFT baselines. Further analysis of channel scoring and selection demonstrates that selective plasticity at the singular-vector-channel level enables effective continual LLM adaptation. --- ### Optimally Pacing Budget Spending and Learning Authors: Mark Braverman, Jingyi Liu, Jieming Mao, Jon Schneider, Eric Xue Published: 2026-10-08 Categories: cs.LG, cs.GT arXiv: https://arxiv.org/abs/2610.11074v1 We establish near-optimal regret bounds for budget-constrained online learning against arbitrary classes of budget-pacing experts in the adversarial setting. In particular, given any class of $F$ experts and a candidate budget pacing schedule, we provide a full-information algorithm which obtains regret $O(D \sqrt{\log F}+ \sqrt{T\log F})$ against all experts whose cumulative spending stays within distance $D$ of this schedule, matching lower bounds established by Braverman et al. (2025). We additionally show that our technique extends to various problems in online resource allocation, where the learner gets to see the rewards and costs of the current options available to them, and establish $O(D\sqrt{\log F})$ regret bounds when fractional allocation is allowed. This is the first algorithm we are aware of which can achieve $o(\sqrt{T})$ guarantees for such tasks. --- ### No Distillation Needed: Single-Pass Real-Time Talking Heads via Acausal Noise Shaping Authors: Yu Han, Dejan Markovic, Alexander Richard, Wojciech Zielonka, Akshay Venkatesh, Cheng-hsin Wuu, Michael Zollhoefer Published: 2026-10-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.11070v1 Audio-driven facial animation underpins real-time avatars, telepresence, and embodied virtual agents. And it must run online: each frame emitted from audio observed up to the current time, at interactive rates. Recent progress is dominated by diffusion models, which need many network evaluations per sample and are therefore a poor fit for streaming. We argue the cost is unnecessary in this domain. Audio-conditioned facial motion occupies a comparatively low-dimensional manifold, a regime where a single-pass GAN suffices. The obstacle is not capacity but stochastic structure. We show that a causal, time-invariant generator driven by i.i.d. noise cannot suppress its output spectrum over a band without collapsing its per-step innovation. We proposed FaceGAN, which dissolved the limitation by shaping the noise pathway acausally. Because the driving noise is synthetic, its future can be sampled now, so the audio-to-expression path stays causal, and the model supports fully causal operation. FaceGAN emits expression and head pose in a single forward pass per frame and matches or outperforms state-of-art approaches in generation quality. Being feed-forward with bounded attention windows, it generates indefinitely without drift. --- ### Clinician use of language models diverges from how the models are evaluated Authors: Krithik Vishwanath, Haitong Lin, Anton Alyakin, Jin Vivian Lee, D. Brock Hewitt, Jie J. Yao, William Robert Small, Hammad A. Khan, Cordelia Orillac, Aakaash Varma, Brandon Ye, Daniel Alexander Alber et al. Published: 2026-10-08 Categories: cs.CL arXiv: https://arxiv.org/abs/2610.11069v1 Large language model (LLM) assistants are being deployed to clinicians across health systems, and judgments about their readiness rest largely on benchmark scores, most of them derived from examination questions or curated cases. A benchmark predicts performance in deployment only to the extent that its items resemble real use, yet whether benchmarks reflect the work these systems receive has rarely been measured. Here we analyze 127,833 queries sent by 6,342 physicians, advanced practice providers and nurses in 35 specialties to an institutional assistant during an eight-month roll-out. We characterize each query with RCQ-Map, a clinician-validated framework grounded in taxonomies of clinical questions and of LLM evaluation, which records its task, intent, answerability, missing information and potential harm. Documentation and administration (36.2%) and knowledge retrieval (28.9%) made up nearly two-thirds of use, and diagnosis 3.7%; more than a third of queries could not be answered well as posed. Applying RCQ-Map to 58 public benchmarks drawn from major evaluation suites and frontier model reports, which we assemble into the Clinical AI Benchmark Atlas, showed that the median benchmark contained no documentation requests and shared 31% of the task mix of real use, less than an even spread across task categories would. Benchmarks in suites designed to resemble clinical practice were individually no closer to real use than those used in frontier model reports. Benchmark scores therefore say little about how clinical AI performs on most of the work it is actually given, and evaluation should be matched to real clinical use. --- ### Diffusion Meta-Prompting and Steering for Generalizable Foundation Model Adaptation Authors: Deepak Sridhar, Yi Li, Kartikeya Bhardwaj, Shuangjun Liu, Taotao Jing, Yuan Li, Shuai Zhang, Jiancheng Lyu, Dashan Gao, Nuno Vasconcelos Published: 2026-10-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.11067v1 Prompt learning is a popular method for adapting foundation models, but learned prompts are typically task-specific and fail to generalize to new classes, domains, or compositions of tasks. In this paper, we introduce a Diffusion Meta-Prompt (DMP) model , a framework that models the distribution of learned prompts using diffusion models. Given a repository of previously learned prompts, DMP is trained and sampled without access to the original task examples or task losses, and synthesizes new prompts conditioned on natural language task descriptions. To improve the sampling stability, we introduce a test-time steering strategy for DMP, which uses the best training-selected prompt in the repository as a latent anchor during diffusion sampling, without retraining the DMP or accessing test classes. DMP improves generalization across classification, retrieval and text-to-image generation tasks, supports concept composition and negative prompting without explicit training. It reduces storage and inference costs by over 90% compared to prompt retrieval methods. For composite classification, DMP achieves upto 2.0% average gain over prior meta-learning methods across 55 pairs of datasets with gains as high as 8.5% on specific pairs such as Eurosat and Flowers. DMP also enhances cross-task generalization with ~2-9% improvement for hierarchical classification task. We further provide a theoretical guarantee bounding the expected task loss of prompts sampled from a DMP. Code is available: https://github.com/DeepakSridhar/dmp --- ### CityDeploy-Bench: Benchmarking Physics-Grounded Spatial Set Planning for Multi-Transmitter Network Deployment Authors: Chenyang Yuan, Xiaoyuan Cheng Published: 2026-10-08 Categories: cs.LG arXiv: https://arxiv.org/abs/2610.11065v2 Automating city-scale wireless deployment remains challenging under complex urban propagation and network-wide interference. We introduce \textbf{CityDeploy-Bench}, a benchmark that reframes multi-transmitter deployment as \emph{physics-grounded spatial set planning} under a unified ray-tracing verifier. The benchmark separates utility representation from planning dynamics, enabling controlled comparison between direct scalar rewards, relational models, and higher-order interaction structures across diverse planners. Our experiments reveal a clear transition in planning behavior as physical coupling grows. Deployment quality becomes increasingly dependent on whether the learned utility captures collective transmitter interactions, whereas stronger search alone cannot compensate for missing relational structure. This establishes multi-transmitter deployment as a coordination problem over physically interacting sets rather than a collection of independent spatial decisions. We release CityDeploy-Data and the benchmark framework as a reproducible testbed for research linking decision learning with physically grounded wireless network design. --- ### Measuring and Mitigating Solution Mode Collapse in RLVR Authors: Liv G. d'Aliberti, Marwa Abdulhai, Sofiia Druchyna, Peter Henderson, Manoel Horta Ribeiro Published: 2026-10-08 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2610.11064v2 A language model (LM) can usually answer the same question in more than one way, but reinforcement learning with verifiable rewards (RLVR) is indifferent to which correct answer a model produces. A solution will earn the same reward whether it is the thousandth copy of a familiar answer or one the model has never produced before. Yet, there is potential value in having the model retain multiple correct solutions as it is trained. For instance, multiple modes may give users a choice and provide problem-solving strategies that improve overall model performance. Here, we introduce ModeBench, a benchmark of multi-solution tasks in which the verifier returns both correctness and mode discovered. We then use ModeBench to measure how solution diversity changes under RLVR post-training. We find that RLVR post-training concentrates probability onto fewer correct modes even as accuracy holds or improves, and moreover, that frontier models are already highly concentrated. We then introduce our solution, Re:Max, which stores one verified example per discovered mode in a replay buffer and trains on those stored modes uniformly. A solution found once is, therefore, practiced as often as one found repeatedly. Across three model scales, two RL objectives, and harder task constructions, replay improves both how often a policy succeeds and how many different ways it can succeed. --- ### Emergent Inverse-Depth Scaling From Nonlinearity In Attention Authors: Zirui Peng, Yizhou Liu, Ziming Liu, Jeff Gore Published: 2026-10-08 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2610.11063v1 Scaling laws describe power-law improvements in model performance with dataset size and parameter count, yet their underlying mechanisms are not fully understood. To explain the parameter count scaling, existing theory posits power-law scaling with model depth. In linear-attention models, this scaling is tied to a power-law data spectrum: unable to selectively attend to relevant tokens, these models learn according to global spectral strength, with stronger directions learned before weaker ones. Large language models, however, can be strongly nonlinear. Here, we show that nonlinear attention yields inverse-depth decay of loss across all tested data spectra. Nonlinearity enables attention to focus selectively on relevant tokens, allowing strong and weak spectral directions to be learned in parallel. Similar focusing across layers motivates a connection to the central limit theorem: shared error across layers sets the loss plateau, while aggregation turns layer-specific differences into continued gains with depth. Our findings suggest that depth scaling may arise from nonlinearity in attention, which allows large language models to focus locally and may make the global covariance structure less relevant. --- ### AffordDrive3D: Affordance-Aware World-Action Modeling with Spatial Understanding Authors: Tianhui Cai, Xinglong Sun, Chao Fang, Zhenxin Li, Rui Song, Jose M. Alvarez, Yunxiang Mao, Jiaqi Ma, Langechuan Liu Published: 2026-10-08 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2610.11060v2 World-action models have recently improved autonomous driving by jointly learning future scene prediction and trajectory generation. Most existing approaches model the future primarily through RGB appearance, and recent works have begun to incorporate geometric prediction to improve spatial understanding. However, dense geometry describes the spatial layout of the entire scene without indicating which parts are most relevant to the ego vehicle's action. For driving, the model must also identify and anticipate where it can safely move and which regions may pose collision risks. Jointly modeling action-relevant regions and future geometry can provide the policy with both driving-relevant cues and their corresponding spatial structure. We therefore propose AffordDrive3D, an affordance- and geometry-aware world-action model that jointly learns future action-relevant regions and spatial structure. In order to capture the scene semantics and driving context needed for driving affordance prediction, we build AffordDrive3D on a VLM backbone to forecast drivable areas and collision-critical regions that directly affect ego motion, while predicting future geometry from RGB world-model latents. On NAVSIM, AffordDrive3D achieves state-of-the-art performance with 91.3 PDMS and 89.9 EPDMS, demonstrating the effectiveness of jointly modeling future affordances and geometry for trajectory planning. Project Page: https://afforddrive3d.github.io/ --- ### Refine Connections, Close the Gap: A Reliable Enhancement Framework for Driving Scene Topology Authors: Xiaoqi Wang, Dingyi Zhuang, David Paz, Wenbin He, Yucai Bai, Peng Zhou, Rui Zhang, Jinhua Zhao, Liu Ren Published: 2026-10-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.11058v2 In autonomous driving, understanding scene topology - the connectivity between lanes and traffic elements - is critical for safe path planning and motion control. While current methods excel at detecting individual map elements, their connectivity reasoning often falls short of its theoretical potential, leaving a significant performance gap relative to the theoretical upper-bound achievable given the underlying detections. Furthermore, the decision-ready topology graphs passed to downstream tasks often remain unreliable. Current approaches typically derive connectivity by thresholding continuous topology scores; however, these scores often fail to reflect the true logical likelihood of connectivity, resulting in false positives or missing connections. Existing benchmarks further overlook this issue by primarily evaluating continuous metrics, rather than assessing the discrete connectivity required for decision-making. To bridge these gaps, we propose TopoEnhance, a novel topology enhancement framework designed to unlock the latent potential of existing methods and improve the reliability of decision-ready topology. We formulate topology enhancement as a denoising-based reconstruction process, where the model learns to recover structural consistency from stochastically corrupted ground-truth graphs. This formulation enables the model to resolve logical inconsistencies and rectify unreliable connections, producing robust discrete topology graphs that closely approach theoretical maximum performance. Extensive experiments across different baselines show that TopoEnhance consistently improves both continuous topology metrics (TOP score), and discrete connectivity measured by our adapted Topology Jaccard Similarity (TJS) metric. As a flexible, source-agnostic framework, TopoEnhance delivers substantial gains across diverse state-of-the-art baselines without requiring retraining. --- ### Learning What to Trust in Multimodal Learning under Noisy Supervision Authors: Jiashuo Zou, Xiaobo Xia Published: 2026-10-08 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2610.11057v1 Multimodal classification processes and relates information from multiple modalities to achieve more accurate predictions. However, existing methods typically rely on high-quality ground-truth labels, which are difficult to obtain in real-world scenarios. While sample-selection methods for learning with noisy labels aim to identify correctly labeled examples from noisy data, traditional methods primarily focus on unimodal settings and fail to exploit multimodal information fully. This motivates us to build a more reliable noise detector in multimodal learning. To this end, we theoretically analyze the relationship between representation structure and noise detection capability. Based on this analysis, we propose REFINE, which is a multimodal label-noise detection framework that jointly uses fused and unimodal representations for label-noise detection. Specifically, REFINE constructs discriminative eigenvectors through discriminative analysis of the target and background classes and selects trusted representation spaces with better noise detection capability for each class. Within each trusted space, REFINE measures the alignment between each instance representation and the discriminative eigenvectors. It then combines the subsets selected from these spaces. The combined set provides cleaner supervision for updating the multimodal classifier, thereby reducing the influence of mislabeled examples during training and improving model generalization. Extensive experiments across diverse tasks demonstrate REFINE's superiority compared to baseline methods. The source code will be publicly available. --- ### AgentHorizon: Evaluating Agentic Judges for Long-Horizon Computer-Use Tasks Authors: Xing Han Lù, Dheeraj Vattikonda, Sina Hajimiri, Fatemeh Pesaran Zadeh, Parishad BehnamGhader, Ghazwa Darwiche, Amirhossein Kazemnejad, Christopher Pal, Alexandre Drouin, Siva Reddy Published: 2026-10-08 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2610.11050v1 Computer-use agents are capable of completing complex tasks, increasing the use of automatic judges to determine success, either for training or for evaluation without human involvement. Despite their flexibility, their reliability on long tasks spanning multiple applications remains unclear. A trajectory, composed of long sequences of screenshots and actions, may appear complete, but in reality violates constraints from the instruction or introduces an unwanted side effect. To identify these errors, a judge needs to examine the trajectory with respect to the user's instruction. To this end, we introduce AgentHorizon, a benchmark of 1,373 computer-use tasks (instruction-trajectory pairs) drawn from 166 hours of human-recorded trajectories spanning three operating systems. By recording trajectories for closely related instructions, we can construct negative tasks by swapping the instructions. This paired design evaluates judges on their ability to distinguish a successful trajectory from one that completed a similar (but incompatible) request. We release the benchmark under three splits: a frontier split, AgentHorizon (AH), a simplified split, AgentHorizon-Simple (AH-S), and a development split, AgentHorizon-Development (AH-D). We further evaluate eleven judges by (1) directly passing the full trajectory (with up to 300 screenshots and actions), and (2) using them as coding agents across five agent harnesses. We find that our best agentic judge, GPT-5.5, achieves 80.9% balanced accuracy on the AH subset. We find that tool-use improves certain models but results in worse performance for open-weight models, and that judges differ drastically in their ability to accept a valid trajectory and reject failed ones. Our findings highlight the need for judges that are capable of locating and verifying often hidden evidence that a task was properly completed inside long interaction histories. --- ### SatFix: Absolute Visual Localization of UAVs in Satellite Maps from a Single Oblique Image Authors: Jiarui Zeng, Kun Shi, Chiman Vong, Zhedong Zheng Published: 2026-10-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.11049v1 We study absolute metric UAV localization within a provided geo-referenced satellite region, recovering continuous map position and viewing heading from a single oblique image or a short multi-view clip. Existing cross-view geo-localization methods retrieve the most similar satellite tile from a gallery and report Recall@K, but retrieval depends on gallery sampling, provides no heading estimate, and returns a tile index rather than a continuous coordinate. We propose SatFix, a feed-forward UAV--satellite localization framework built on VGGT-$Ω$. Satellite-grid features act as queries that aggregate UAV visual evidence, and two lightweight heads regress a 3-DoF pose in the satellite-map frame: continuous 2D position and heading. SatFix requires no explicit 3D map, rendered bird's-eye image, auxiliary sensor, or test-time pose alignment. A single model supports both single- and multi-view inputs, with trajectory constraints used during multi-view training. For metric evaluation, we introduce University-Metric, where satellite imagery is re-collected over a region up to 10.7$\times$ longer on a side (about 114$\times$ the ground area) than the original University-1652 tiles, with continuous position and heading labels for the original UAV tours. With one UAV view, SatFix localizes 52.08% of test frames within 50 m and 17.34% within 10 m, with median position and heading errors of 45.66 m and $20.81^\circ$, respectively. Inference takes under 0.1 s per single-view query on an NVIDIA RTX 4090. With nine UAV views, the median position error falls to 21.96 m and the median heading error to $8.73^\circ$. Compared with a fine-tuned VGGT-$Ω$ baseline, SatFix reduces median position error by 34.0% and nine-view median heading error from $25.43^\circ$ to $8.73^\circ$. --- ### Rendering-Free Lookahead for Question-Guided Active Vision Authors: Koya Sakamoto, Daichi Azuma, Shuhei Kurita, Naoya Chiba, Yusuke Iwasawa, Yutaka Matsuo, Taiki Miyanishi Published: 2026-10-08 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2610.11039v1 Active robot vision requires controlling the camera to reveal task-relevant information that is hidden from the current viewpoint. For example, determining what is inside a box may require raising the camera and looking down into it. For viewpoint-dependent question answering, the challenge is to select camera motions that expose the visual evidence needed to answer the question. Although vision-language models (VLMs) can interpret observed images, selecting such motions requires anticipating the usefulness of unseen views. We quantify this usefulness as answerability, a VLM's estimate that a view suffices to answer the question, and present Rendering-Free Lookahead (RFL), a viewpoint-selection policy that ranks candidate camera motions by predicted future answerability. RFL transfers visual lookahead from deployment to offline training. At training, a privileged teacher renders candidate future views in 3D Gaussian Splatting (3DGS) scenes and uses a frozen VLM to compute one- and two-step answerability targets. Through two-stage distillation, a student learns to predict these action values from the question, recent visual observations, and a candidate camera motion. At deployment, RFL uses these predicted values to select camera motions without rendering future views. On 377 E3VS-Bench test episodes in unseen environments, RFL improves the mean judge score by 43\% over a direct-action baseline using the same VLM. These results support learning camera-control policies from privileged visual lookahead for viewpoint-dependent question answering. --- ### Decision-Sufficient Posterior Approximation Authors: Sean Plummer Published: 2026-10-08 Categories: math.ST, stat.ML arXiv: https://arxiv.org/abs/2610.11038v1 We investigate the consequences of requiring a posterior approximation to preserve a specified downstream decision problem. A target posterior $P$ and loss determine a regret geometry on actions, a baseline approximation $Q_0$ determines the forward-Kullback-Leibler information required to induce action changes, and a restricted approximation family $\mathcal{Q}$ determines which such changes are available. Contracting KL divergence over Bayes-action fibers gives exact distances to decision adequacy and decision failure together with the least-informative posterior deformations that reach either side of the decision boundary. In regular finite-dimensional problems, the target and baseline constructions have quadratic local limits: a target regret Hessian $G$ and a baseline information metric $J_I$ . Their generalized eigenproblem $Gv = γJ_Iv$ orders local decision directions by regret consequence per unit information cost and induces a tolerance-dependent effective dimension. For restricted approximation families, the tangent image separates decision coverage from information efficiency: a family may miss consequential decision directions, or it may realize reachable directions only at excess Fisher cost. The resulting framework provides decision-relative criteria for comparing and designing posterior approximation families. --- ### Transforming Image Editors into Video Editors Authors: Feng Wang, Zijie Li, Ceyuan Yang, Alan Yuille, Peng Wang Published: 2026-10-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.11037v1 Recent image editing systems have achieved impressive semantic understanding, visual fidelity, and instruction-following ability, while video editing remains substantially more difficult and costly. In this paper, we present a simple alternative to end-to-end video editing: instead of training a monolithic video editor, we transform a strong image editor into a video editor through anchor-based generation. Our key insight is that video editing can be decomposed into two subproblems: editing a sparse set of keyframes and propagating those edits across time. Based on this observation, we propose Anchor-based Video Editing (AVE), a two-stage framework in which a powerful image editor first performs composed editing on selected keyframes, and a motion-guided image-to-video diffusion model then generates the final video by treating the edited keyframes as fixed anchors. This design directly inherits the strengths of modern image editors while avoiding expensive end-to-end video editing training. Experiments on IVEBench and VIE-Bench show that AVE achieves strong performance in instruction following, temporal consistency, and content fidelity. Further ablations reveal that final video editing quality is strongly correlated with the quality of the image editor, suggesting that future progress in video editing may come from stronger image editing foundations and lightweight transfer to video. Code is available at https://github.com/wangf3014/AVE. --- ### FedAlphaEdit: Null-Space-Aligned Merging for Collaborative Knowledge Editing Authors: Sota Sugawara, Yukihiko Okada Published: 2026-10-08 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2610.11033v1 Multiple institutions may each hold their own private knowledge edits and wish to integrate them into a single large language model without sharing raw edit requests. Null-space-constrained editing methods such as AlphaEdit mathematically guarantee that each update leaves unrelated knowledge intact, while collaborative frameworks such as CollabEdit aggregate edits from multiple clients without data sharing. Combining the two appears trivial. However, we show that this naive combination fails structurally, and we identify its cause. Guided by this analysis, we propose FedAlphaEdit. To our knowledge, this is the first collaborative knowledge editing framework that aligns both local editing and the server-side merging rule under a single null-space principle for preserving existing knowledge. FedAlphaEdit builds on null-space-aligned merging, in which clients share projected statistics and the server provably recovers the result of editing everything in one place under a one-shot idealization. Empirically, the proposed method repairs the collapse and brings edit success and preservation simultaneously close to the level of centralized editing across two architecture families. FedAlphaEdit thus lets institutions that cannot share raw edit data, such as hospitals and financial firms, jointly maintain a shared model that closely approximates editing all facts in one place. --- ### Language Modeling is Monotone Compression Authors: Noam Mazor, Andrew Morgan, Rafael Pass Published: 2026-10-08 Categories: cs.IT, cs.AI, cs.CR arXiv: https://arxiv.org/abs/2610.11031v1 A long-standing hypothesis in artificial intelligence and neuroscience posits that intelligence is closely related to compression: the ability to compress information efficiently intuitively reflects capacities associated with intelligence and learning. Indeed, recent experimental works verify this intuition by showing connections between the capabilities of large language models (LLMs) and their ability as compressors: for instance, Deletang et al. (ICLR'24) demonstrate that LLMs can be used as powerful compressors, and Huang et al. (COLM'24) show that the compression ability of LLMs is highly correlated with their performance on benchmarks for knowledge and reasoning. In this work, we initiate a theoretical study of this connection. Our main result is that LLMs (formally modeled as next-token predictors) are equivalent to monotone (a.k.a. order-preserving) compression algorithms---namely, compression algorithms where the encoding process preserves the ordering of the inputs---in the sense that the one can be constructed from the other while preserving the same error up to an additive gap of 2. We next show that the monotonicity is required for this equivalence to hold if and only if cryptographic (infinitely-often) one-way functions exist. As a direct corollary, we get a cryptographic result of independent interest: the notion of next-bit pseudoentropy (a computational analogue of entropy) of a distribution is equivalent to monotone incompressibility of the distribution. (Previously, it was only known (Haitner et al., ITCS'23) that incompressibility implies next-bit pseudoentropy.) --- ### NOMOS: Compiling Written Policies into Statically Verified Tool-Call Gates for LLM Agents Authors: Min-Young Yu, Tony Kim, Jang Won Choi Published: 2026-10-08 Categories: cs.CR, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2610.11030v2 Tool-using LLM agents violate the policies they are deployed to enforce, often silently. Prior defenses hand-write rules, query an LLM verifier per action, or compile policies through heavyweight formal machinery. Naive compilation fails: extracted rules block the tool satisfying their own precondition, or read arguments their tool lacks. NOMOS, a four-pass compiler, turns a natural-language policy into a deterministic tool-call gate; static verification with tool-schema-level checks alone (no prover, solver, or LLM) repairs or rejects 37% (airline) and 13% (retail) of candidates, without which most shipped rules are inoperable. Replaying compiled rules over undefended transcripts flags bindings that refuse legitimate work (a development binding refused 95.9% of task-passing calls); no evaluation binding is flagged. On $τ^2$-bench the gate cuts violations of reference-encoded clauses among state-changing calls from 66.3% to 2.6% (airline) and 30.8% to 6.9% (retail), raising airline task success significantly for $2 \le k \le 4$; a 26B on-premise compilation is not significantly worse than hand-written or frontier-compiled rules. Unlike AgentDojo's shipped defenses, it reaches a zero attack success rate (ASR) on banking, where nine attack families collapse onto three structural rules. On the other three suites its ASR is at most 3.6%, from goals with no tool call to govern and one write admitted by a binding weaker than its clause; a second agent model, Llama-3.3-70B, reproduces the effect on both benchmarks. Decisions take microseconds without an LLM call, at a domain-dependent benign-utility cost; compilation runs on-premise on open-weight gemma-4-26B. --- ### Intent Graph: Navigating the Analytical Reasoning Space for Exploratory Data Analysis Authors: Junran Yang, Shruti Badrish, Teanna Barrett, Leilani Battle Published: 2026-10-08 Categories: cs.HC, cs.AI arXiv: https://arxiv.org/abs/2610.11025v1 Exploratory data analysis (EDA) is rarely open-ended in practice: analysts work from high-level domain questions toward the concrete analyses that can answer them, prioritizing directions with domain knowledge and prior hypotheses. Large language models (LLMs) can supply such knowledge, but their responses are unstructured, leaving analysts no way to see what has been explored, what is missing, or why one direction was chosen over another. We present DAG-EDA, a system that lets analysts and an LLM co-navigate the space of possible analyses through two linked structures. An intent graph, governed by a grammar of analytical intent, decomposes an ambiguous natural-language question into progressively concrete analysis tasks, keeping alternative framings open and letting analysts branch, backtrack, and compare paths. A multi-layered knowledge graph externalizes the LLM's domain knowledge, linking domain concepts to the dataset variables that can measure them, so analysts can inspect and contest how their question is grounded in the data. Both graphs are constructed from only the dataset and the analyst's question, and the analyses the analyst reaches are rendered as interactive dashboards. We illustrate the system through a usage scenario and describe a user study design for examining whether the system scaffold analysts' reasoning and navigation. --- ### Expression-Diverse References for Identity-Preserving Video Generation Authors: Tianwen Fu, Wenbin Teng, Gonglin Chen, Junyi Ouyang, Haolin Xiong, Yajie Zhao Published: 2026-10-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.11023v1 Identity-preserving video generation aims to maintain a subject's identity while synthesizing realistic videos. Yet a single reference portrait captures the subject's appearance under only one facial configuration. As expressions change, facial appearance can vary in highly identity-specific ways, leaving the subject's appearance under unseen expressions underdetermined by the reference alone. This expression-dependent variation also complicates evaluation: similarity to a neutral reference may decrease under strong expressions even for real images of the same person. We investigate this limitation from both generation and evaluation perspectives. First, we quantify how face-recognition similarity varies with expression intensity using controlled photographs and MEAD videos. We then construct a compact yet expressive reference gallery that captures diverse expression-dependent facial configurations. Matching against this gallery provides a more robust measure of identity similarity under expressive motion. To further expose performance degradation with expression intensity, we report identity similarity separately for mild, intense, and extreme expressions. For generation, we extend Stand-In to condition on our expression-diverse reference sets and develop a data-curation pipeline that extracts consistent yet diverse face crops from training videos. In practical settings where only a single portrait is available, we construct the reference set by synthesizing additional expressions with a pretrained facial reenactment model. On our controlled benchmark, both real and synthesized reference sets outperform the evaluated baselines in identity similarity across all three expression-intensity regimes, with the largest improvements for extreme expressions. --- ### A Graph Neural Network for Global Daily Fire Radiative Power Prediction at Medium-Range Lead Times Authors: Li Zhang, Jun Wang, Isidora Jankov, Yongxin Liu, Gonzalo A. Ferrada, Ravan Ahmadov, Ligia Bernardet, Haonan Chen, Shobha Kondragunta Published: 2026-10-08 Categories: physics.ao-ph, cs.LG arXiv: https://arxiv.org/abs/2610.11022v1 Skillful prediction of biomass-burning activity several days in advance is important for air-quality forecasting and aerosol prediction. Two operational constraints motivate this work. First, the GBBEPx satellite fire radiative power (FRP) product used to initialize NOAA's GEFS-Aerosols is available with about a 1.5-day latency, so each forecast cycle relies on the most recently available, but already outdated, fire observations. Second, these fire inputs are then held fixed throughout the subsequent 5-day operational forecast, or 7 days in the GSL experimental system, effectively assuming no evolution in fire activity. We develop a data-driven model that predicts global FRP one to seven days ahead from the most recent available observations. The model adapts a spatiotemporal graph neural network using reanalysis meteorology, land-cover and vegetation information, recent fire history, and GBBEPx FRP as the training target. It is trained on 2020-2022 data and evaluated for 2023-2024. The model reproduces the global seasonal cycle and substantially outperforms persistence. At 0.1$^\circ$ resolution, mean squared error is reduced by 32% at one-day lead and 43% at seven days in 2023, and by 24% and 40% in 2024. At 1$^\circ$ resolution, the critical success index ranges from 0.32 to 0.60. Detection skill declines only modestly with lead time, whereas intensity skill degrades more rapidly. Large fires are detected reliably, but their radiative power is systematically underestimated. These results demonstrate useful predictability of fire activity several days ahead and identify intensity calibration and small-fire placement as the main remaining challenges before predicted FRP can support operational aerosol forecasts. --- ### Mid-Training Language Models on Raw Video Authors: Jaedong Hwang, Xiaoqian Shen, Ernie Chang, Changsheng Zhao, Chong Zhou, Saksham Suri, Qi Qian, Zechun Liu, Lemeng Wu, Qinsi Wang, Raghuraman Krishnamoorthi, Wei Wen Published: 2026-10-08 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2610.11019v1 Multimodal large language models learn mostly from paired image-text data or annotated video, and raw web video is rarely used to further train an existing language model. We study whether raw video, with no captions and no text loss, can serve as mid-training data for a pretrained language model. Frames are encoded into continuous visual tokens, and the language model learns to predict the next visual token. We mid-train Qwen3-1.7B on raw clips from YT-Temporal-1B and then apply the same image-text instruction tuning to it and to the model without mid-training, so that the two differ only in mid-training. The mid-trained model scores 2.9 points higher on average across four video benchmarks and 5.1 points higher across ten image benchmarks, spanning perception, document, and chart tasks. Text performance is preserved even though mid-training includes no text, with an average of 48.9 across 14 text benchmarks compared with 48.0 for the model without mid-training. Analyses across training show that the image and video gains emerge within 30% of training and plateau thereafter, varying by less than 0.5 points. Predicting captions fails to outperform next-visual-token prediction, demonstrating that video mid-training can remain purely self-supervised without the computational overhead or labeling noise of automated captioning. --- ### AbridgeKV: Compressing Long Reasoning into the KV Cache of Learned Pause Tokens Authors: Jack Cai, Lijie Yang, Vasilis Kontonis, Dimitris Papailiopoulos, Shivam Garg Published: 2026-10-07 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2610.20641v1 Long chain-of-thought decoding is memory-bound: the KV cache of the generated tokens grows linearly with the test-time compute being spent. Evicting cache entries does not fix this, because a model never trained on a cache with entries removed generates longer, often up to the length limit. We instead train the model to compress its own output. AbridgeKV compresses each completed block of generated text into the KV entries of a few learned pause tokens. These are extra input tokens with trained embeddings that produce no output. Once the pause tokens have encoded the block, AbridgeKV frees the block's original cache; subsequent tokens attend to the pause tokens' KV entries in its place. Three novel choices make this work: the pause tokens' KV entries take positions inside the block they replace, the cache budget is allocated across layers according to the measured value of each layer's entries, and compression operates on relatively large blocks. On 446 competition problems from AIME, HMMT, AMC, BRUMO, CMIMC and SMT with 32K-token generations, keeping a quarter of the generated cache matches an identically trained full-KV model in accuracy and output length; the strongest published evictor at a matched budget writes far longer, and the closest trained compressor, retrained on the same data, loses 10 points. Served in vLLM, the cache is 3.5x smaller per sequence and a saturated H100 finishes a problem 3.2x faster than with full KV. --- ### Loose Lips: Understanding Privacy Leakage in Agentic Dialogue Authors: Soumya Sharma, Ayana Hussain, Heber Hwang Arcolezi, Ulrich Aïvodji, Spandana Gella, Adriana Romero-Soriano, Golnoosh Farnadi Published: 2026-10-07 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2610.14071v1 As LLM agents are increasingly deployed in high-stakes enterprise settings, understanding how they handle sensitive information in agentic dialogue has become critical. Privacy leakage in such settings requires reasoning about whether information flows respect the contextual norms under which they originated. We extend CIMemories to CIMEMORIES DIALOGUE, a benchmark of 4,410 two-agent dialogue instances grounded in Contextual Integrity theory. Using this benchmark, we systematically evaluate 7 open-weight models from the Qwen3 and Llama 3.1 families across conversation lengths of 1 and 6 turns under three prompting conditions. We find that task capability and privacy leakage are fundamentally coupled: more capable models are systematically leakier, with no model in our evaluation able to decouple the two. Thinking models consistently leak more than their non-thinking counterparts, a pattern we attribute to the extended reasoning surface created by chain-of-thought. Leakage compounds with conversation length, rising up to 3x from 1 to 6 turns. Our results suggest that the coupling between task capability and privacy leakage is not an artifact of misconfiguration but a structural feature of current agentic systems. --- ### Safety-Oriented Trajectory Ranking from BEV Perception Outputs Authors: Leon Pohl, Martin Frauendorfer, Corinna Schmitt, Mirko Maehlisch Published: 2026-10-07 Categories: cs.RO, cs.CV arXiv: https://arxiv.org/abs/2610.14066v1 Bird's-Eye-View (BEV) representations are increasingly used as compact inputs for autonomous driving perception and planning. However, lightweight and interpretable methods for evaluating candidate ego trajectories directly on BEV outputs remain useful for vehicular safety workflows, rapid prototyping, and human-in-the-loop validation. This paper presents a modular trajectory ranking pipeline that consumes BEVFusion object detections and map segmentations derived from nuScenes sensor data and ranks manually specified candidate trajectories according to safety-first criteria. The ranking prioritizes avoiding collisions and walkway violations, then smaller end distance to the target, and finally path efficiency. To study whether these rule-based rankings resemble human judgment, we created seven BEV traffic scenarios and compared the script output with preferences from a survey of 21 participants. Across 146 pairwise votes, 124 matched the trajectory ranked first by the evaluator, corresponding to an overall agreement of 84.9 %. The results indicate that simple BEV-centric safety heuristics can approximate human preferences in constrained trajectory choice tasks while remaining transparent enough for analysis, dataset curation, and autonomous driving studies. --- ### Auditing LLM Program Search for Bermudan Stopping with Primal--Dual Certificates Authors: Yuhua Qian Published: 2026-10-07 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2610.14063v1 LLM program-search agents are usually judged by the score that drives their search. We study LLM-driven search over Python exercise policies for Bermudan options, an optimal stopping problem where a trusted evaluator owns the simulation and a dual upper bound~$U$. For any policy, $0 \le V^\star - L_π\le U - L_π$; we report Monte Carlo estimates of both bounds with their sampling error. We audit 24 generation attempts from two pipelines, a six-pair feedback-representation ablation, and a holdout evaluated after selection was frozen. No selected generated program improves on Ridge LSMC consistently or by more than Monte Carlo uncertainty; holdout gap scores lie within $0.002$ of it. On six one-dimensional puts, a high-precision binomial reference places Ridge within Monte Carlo noise of optimal and attributes 99\% of the estimated gap to dual-bound looseness, so this dual audits validity but cannot certify progress at the search's scale. Structured memory showed no benefit over score-only memory in this ablation. Whether search ``worked'' depends on reporting rules: retaining the initial reference, freezing selection before holdout, and counting failures. --- ### CircuitTree: Composing Circuits into A Syllable Transcriber Authors: Jiankun Wei, Ewan Dunbar, Gerald Penn Published: 2026-10-07 Categories: cs.LG, cs.AI, cs.SD arXiv: https://arxiv.org/abs/2610.12520v1 The goal of circuit discovery is to decompose a monolithic neural network into tiny modular subnetworks which are interpretable in the sense that each does an identifiable subtask. We develop a practical method for composing arbitrarily many circuits into a single model. We propose a greedy, density-scaled composition algorithm that overcomes limitations of previous attempts to compose circuits by naive union, then generalize it to $N$ circuits via CircuitTree, a new data structure built on a $Θ(N\log N)$ tree-compose schedule. We compose circuits pruned from HuBERT for classifying individual phonemes into a syllable classifier, a capability none of the 39 phoneme circuits has individually. Chaining CircuitTrees into a hierarchy-of-circuits pipeline, we transcribe open syllables at 64.8\% accuracy, within 1.2 points of a 66.0\% ceiling that shares no circuitry at all, with 99.5\% of circuit components shared. --- ### Looking Under the Streetlight: Evaluation in Generative Molecular Dynamics Authors: Simon Olsson, Frank Noé, Grant M. Rotskoff, Kresten Lindorff-Larsen Published: 2026-10-07 Categories: physics.chem-ph, cs.LG arXiv: https://arxiv.org/abs/2610.12518v1 This position paper argues that the generative molecular dynamics (GenMD) community is not properly evaluating its progress: current evaluations let a model match what is reported while being arbitrarily wrong on what was claimed, which risks over-claiming advantage over classical baselines. This reflects a structural misalignment between common evaluation practices and the scientific tasks the field sets for itself. The reason is that the ground truth---exhaustive molecular dynamics on the systems of interest---is computationally costly, and this cost is what motivates the field. With ground truth inaccessible, evaluation drifts toward convenient proxies, and the resulting norms entrench easily. We identify misaligned evaluation patterns that persist as a community-level equilibrium, and propose concrete observables, reporting standards, and benchmark designs, both computational and experimental, that better reflect what GenMD aspires to deliver. Our aim is to give the field shared language and tools for evaluation that match its scientific ambitions before today's conventions calcify into norms. --- ### TerraAgent: Demonstration-Grounded Multi-Agent Reasoning for Long-Context Remote Sensing Visual Question Answering Authors: Jingwen Huang, Yiming Fang, Hanxiao Wu, Boyuan He, Keyi Luo Published: 2026-10-07 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.12517v1 Remote sensing visual question answering (RS-VQA) requires answers that truly depend on image evidence, yet current remote sensing multimodal large language models (RS MLLMs) are still evaluated mainly by accuracy, which cannot reveal whether a correct answer was produced without using the image. This setting introduces three challenges: demonstrations retrieved across different ground sample distances (GSD), sensors, and city textures may be poorly matched; real queries often involve long visual contexts from multi-scale tiles, multi-temporal frames, and retrieved knowledge, where evidence can be diluted and errors can propagate through multi-step reasoning; and prior-only answering remains invisible to standard metrics. We present AVA, a training-free framework that wraps a frozen RS MLLM with three modules. GDR performs GSD-bucketed, scene-graph-based demonstration retrieval and filters demonstrations whose answers are unsupported by their images. MACE decomposes each query into an evidence-typed sub-question DAG, routes sub-questions to suitable perception tools, and records verified provenance in a shared evidence ledger with bounded re-perception. CVA audits the final answer through counterfactual perturbations of cited evidence, including ablation and swap, and computes a visual-dependency score for confidence calibration and evidence-forcing re-query. On RSVQA-LR/HR, VRSBench, and LHRS-Bench, AVA improves average accuracy by up to 2.40 points over the strongest same-setting baselines, raises the visual-dependency score from 0.28 to 0.49, reduces the prior-only answer rate from 46.2 percent to 21.7 percent, and lowers the hallucination rate from 17.1 percent to 9.8 percent, with 2.6x the cost of a single backbone forward pass and no fine-tuning. --- ### TraceICL: Thread-Resolved Demonstration Curation with Consistency Calibration for Robust Multimodal In-Context Learning Authors: Yigit Yilmaz, Elena Petrova, Mehmet Kaya, Lucia Rossi, Amir Rahman Published: 2026-10-07 Categories: cs.CV arXiv: https://arxiv.org/abs/2610.12516v1 Interleaved multimodal in-context learning (ICL) lets frozen large vision-language models (LVLMs) acquire new tasks from a handful of image-text-answer demonstrations, yet it remains fragile: demonstrations bind to the wrong images, redundant exemplars crowd out informative ones, and noisy answers are copied verbatim into predictions. We present AVA, a training-free framework that organizes the demonstration pipeline of multimodal ICL into three composable stages. A Thread-Resolved Decomposer (TRD) segments the interleaved context into per-sample threads and re-stitches them with explicit boundaries, restoring image-text binding. An Influence-scored Demonstration Selector (IDS) estimates each candidate's marginal influence with one extra forward pass, removes redundancy, and orders the survivors by descending value. A Weak-to-Strong Consistency Calibration (WSC) module scores each demonstration by a weak sibling model's reconstruction confidence and softly reweights its contribution during decoding. On five vision-language benchmarks with interleaved 4-shot prompts, AVA improves the strongest same-setting baseline by +1.7 points on average, with gains of up to +2.4 points on VizWiz, shrinks the require-visual performance gap on ScienceQA-IMG, and retains most of its gain under 30% noisy demonstrations, at a 26% latency overhead that an offline cache amortizes. No gradients are ever computed; the LVLM stays frozen. --- ### CORAL-KV: Compact Cross-Layer Prediction and Residual Coding for Autoregressive Video KV Caches Authors: Albert Ge, Akash Haridas, Utkarsh Saxena, Ruijing Yang, Ian Colbert, Frederic Sala, Vikram Appia, Emad Barsoum Published: 2026-10-07 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2610.12514v1 Autoregressive video diffusion models maintain a growing attention key-value (KV) cache, making memory a key bottleneck for long video generation. Recently, clustering-based quantization has emerged as an effective method to compress video KV caches by grouping similar key and value vectors then quantizing residual errors. However, when later frames attend to the compressed history, small reconstruction errors can compound as the video grows. Beyond the within-layer similarity exploited by clustering, we find strong dependencies across cache layers in video models. This motivates Coral-KV - Calibrated Offline, Residual-coded, and Adapted Locally - a KV-cache compression framework that uses cache states in each layer to predict keys and values in subsequent layers, then quantizes the remaining prediction residuals. We construct compact block-diagonal-plus-low-rank (block-DPLR) predictors: block-diagonal components model channel relationships within each attention head, while a global low-rank term models dependencies between heads. Coral-KV exposes a flexible fidelity-memory trade-off: it can reuse its predictors across videos or refit lightweight block-diagonal predictors to the current video for higher fidelity. Across extensive evaluations on both HY-WorldPlay and MAGI-1, our reusable configuration improves PSNR over QVG-4 by 0.32 dB and 4.95 dB, respectively, while using 48% and 37% less cache memory. Against QVG-Pro, our video-adapted configuration improves MAGI-1 PSNR by 3.77 dB while using 33.7% less cache memory. Coral-KV also maintains higher reconstruction fidelity over extended generation and reduces temporal discontinuities while retaining comparable generation latency. --- ### Prompts versus Rules: Auditing and Controlling Speech Naturalness Behaviors in Voice User Simulators Authors: Riqiang Wang, Elena Khasanova, Harsh Saini, Lex Konnelly, Parsa Kavehzadeh, Matthias Lee, Mohamed Attia Published: 2026-10-07 Categories: cs.CL arXiv: https://arxiv.org/abs/2610.11015v1 As voice agents gain more popularity commercially, the user simulators used to evaluate the deployed agents are also being developed to include more realistic, variable, and diverse speech naturalness behaviors -- disfluency, interruption and backchanneling. The quality of the user simulator directly affects the validity of agent evaluation results. However, we find that most studies so far have not examined in detail whether the intended configuration for these behaviors is realized in the simulation. In this study, we audit the realized naturalness behaviors of tau-Voice, our own LLM-based prompting approach across three models, and our rule-based injection algorithm for disfluency, interruption, and backchanneling. We find that prompting for these behaviors is unreliable and produces speech inconsistent with the instructions, placed and distributed less naturally than the instruction implies. In contrast, our rule-based, model-free algorithm produces controllable and diverse naturalness behaviors more aligned with natural speech. Our results suggest that LLMs not purpose-trained for user simulation are not sufficient on their own to represent authentic user behavior, and that linguistically informed deterministic approaches or specialized models are needed to close the gap; auditing and reporting realized naturalness behaviors, rather than configured settings, is what makes that gap visible. --- ### Distillation for Incrimination and Distillation for Capabilities Authors: Sebastian Prasanna, Jacqueline Tay, Alek Westover Published: 2026-10-07 Categories: cs.AI arXiv: https://arxiv.org/abs/2610.11012v1 Powerful misaligned AI models might recognize alignment evaluations and strategically behave well on them, rendering direct audits uninformative. However, distilling such a model into a weaker benign student places the teacher in a Distillation Double Bind: if misalignment transfers, the student may conceal it less effectively, revealing evidence about the teacher; if it does not, the student may learn useful capabilities while remaining benign. We introduce two distinct distillation approaches, one targeting each outcome. Distillation for Incrimination (DFI) aims to transfer misalignment but not the ability to conceal it. Distilling AuditBench's secret-keeping models into their underlying instruction-tuned model produces students that are significantly more likely than their teachers to admit their hidden behavior when asked, suggesting that knowledge of the behavior transferred more readily than the propensity to conceal it. Confession gains largely disappear when the student does not share the teacher's pretrained base, so DFI should target the teacher's own pre-RL checkpoint, which is weaker than the teacher but shares its base model. Distillation for Capabilities (DFC) aims to transfer capabilities but not misalignment. Among several techniques we evaluate, two are effective: inoculation prompting and training for more epochs on fewer unique examples. Both preserve the capability gains of standard distillation while substantially reducing the subliminal transfer of an animal preference, our proxy for misalignment. Together, these findings demonstrate two ways distillation can be used for AI safety: incriminating misaligned models, and extracting their capabilities without their misalignment. --- ### PCAsplat: Gaussian Splatting with Local PCA Regularization Authors: Vitor Matias, Filipe Nascimento, Kiyohiro Nakayama, João Paulo Lima, Márcus Lobo, Gordon Wetzstein, Leonidas Guibas, Afonso Paiva, Tiago Novello Published: 2026-10-07 Categories: cs.CV, cs.GR arXiv: https://arxiv.org/abs/2610.11011v1 Gaussian splatting has emerged as a flexible representation for 3D reconstruction from posed images. However, existing methods are optimized primarily using rasterization-based losses, which supervise a splat only when it contributes to sampled camera rays. Gaussians that are occluded or contribute little to the sampled view therefore receive weak or no geometric gradients and may drift away from the underlying surface, producing undesired floaters. We introduce PCAsplat, a geometry-aware regularization framework for Gaussian splatting based on differentiable local principal component analysis (PCA). Our PCA regularizer acts directly on neighborhoods of Gaussian centers and can therefore update Gaussians that do not contribute to the current training view. We regularize the PCA eigenvalues to encourage Gaussians to move to the underlying surface with isotropic tangent-plane coverage. We also align each Gaussian normal with the PCA-estimated neighborhood normal to enforce consistent orientation. Experiments on DTU, Tanks and Temples, and NeRF Synthetic show that the splats produced by PCAsplat better approximate samples of the reference surface while substantially reducing undesired floaters. These surface-aligned splats enable downstream geometry-processing tasks, including point cloud segmentation, and direct Poisson reconstruction. Additionally, PCAsplat remains competitive under conventional novel view synthesis and mesh extraction tasks. Code will be released. --- ### Why LLM Agents Favor Their Group: Stakes, Observed Norms, and Reputation Authors: Yujiao Chen Published: 2026-10-07 Categories: physics.soc-ph, cs.AI arXiv: https://arxiv.org/abs/2610.11008v1 Language-model agents favor their own group because they have watched their members favor each other. The group label alone does little once the decision has a cost; what drives favoritism is observed behavior, and an individual's own record can override it. We test this in small societies with arbitrary group labels, ten rounds of point sharing, and matched one-shot decisions across fifteen OpenAI models and three Claude models, about 4,400 societies and 3.3 million audited model calls. First, the large effect of a bare group label reported in earlier work appears only when giving others points costs the agent nothing; once the agent can keep points for itself, that effect collapses on every model that shows it. Second, under a stake, interaction history becomes the main source of favoritism: the history effect is statistically positive on 13 of 15 models, reaches about 3.5-8 points out of 10 on 11, grows with the number of rounds played, and extends to labeled strangers the agent has never met. Third, with scripted histories, favoritism falls to near zero under an egalitarian norm and reverses when the agent's own group is seen favoring the other side; stronger models side with an individual's record when it conflicts with the group. Group favoritism is thus conformity to observed group behavior, carried to strangers by the label and overridden by individual reputation. The same account predicts responses to betrayal, scandal, and a free offer to change group: public reprimand repairs betrayal better than apology or restitution, allocation punishment remains confined to the offending member, and a formed group cannot be bought but can, on weaker models, be invited away. --- ### Curating Always-Loaded Context for LLM Agents: A Capacitated Assortment Model with Censored Feedback Authors: Zexuan Liu, Yuning Yang, Tiancheng Zhao Published: 2026-10-07 Categories: cs.AI, cs.LG, math.OC arXiv: https://arxiv.org/abs/2610.11007v1 At the start of every session, LLM agents load a fixed context file, such as $\texttt{AGENTS.md}$. Each loaded token in the file is charged again in every later round of the session, and these files can degrade performance as they grow in size. However, in practice, human or automated curators usually grow these files by appending. We formulate context curation as a capacitated assortment problem. Instructions consume tokens under a finite attention capacity; adding an instruction never raises the compliance of the others, while retained instructions incur a per-session setup cost. We prove an upper bound on the optimal file size, regardless of the number of available candidate instructions, and that appending every instruction with positive standalone value can be arbitrarily worse in net value than selecting an optimal subset. A token budget also limits the loss when the token price is underestimated. We then examine what can be learned from past sessions and how this information can guide decisions to add or remove instructions. Feedback is inherently censored: the benefits and harms of loaded instructions are observable, whereas missing instructions generate feedback only when their absence causes harm. In this setting, we show that deleting instructions ignored by agents can inevitably remove helpful ones. We characterize how much evidence should be collected before adding an instruction. Besides, we bound regret when human reviewers can inspect only a limited number of edits per period. Empirical experiments further show that irrelevant rules drawn from real context files reduce language-model compliance. ---