# Research Papers — 2026-09 · bundle page 30 of 266 > Full-text markdown of papers #1451–1500 of 13298 published in 2026-09, newest first. AI-readable mirrors of the full 13298-paper 2026-09 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-09/page/30 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/29.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/31.md) ### The Safety Operator: Modulating the Expression of Safety Instructions via Spectral Optimization Authors: Benoit Dherin, Michael Munn, Xavier Gonzalvo, Adrian Goldwaser, Blaz Bratanic, Ananth Balashankar, Andrey Vlasov, Pinzhi Huang, Cecile Loge, Nicole Mitchell, Andre Fernandes, Trilok Acharya et al. Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36434v1 Context tokens in a transformer-based language model can be absorbed into the model's weights as a multiplicative operator. We study this operator in the setting of safety instructions and show that influencing its dominant eigenvalue modulates how strongly the instruction shapes generation. We derive a Contrastive Safety Loss with a suppression weight that controls the tradeoff between emphasizing the safety instruction on harmful queries while suppressing it on harmless queries. Varying the suppression weight maps a relationship between the attack success and the over-refusal rates, supporting the hypothesis that the operator's eigenvalue acts as a continuous dial for the instruction's influence. Moreover, this relationship holds relatively independently of how the Safety Loss is parameterized, yielding Pareto-improved safety instructions for appropriate values of suppression weight. --- ### RA-CFGCache: From Branch-Level Criteria to Guided-Risk Control under Classifier-Free Guidance Authors: Yiming Liu, Ben Wan, Tongxuan Liu, Ao Wang, Yuqi Xiong, Fan Zhang, Hui Chen, Guiguang Ding Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36433v1 Diffusion models enable high-quality visual generation, but iterative denoising remains computationally expensive, especially under classifier-free guidance (CFG), which requires both conditional and unconditional evaluations. Training-free caching reduces this cost by reuse of previously computed features or predictions. However, existing branch-local reuse criteria do not explicitly account for how cache errors combine under CFG or how local perturbations affect the final output. We identify two misalignments in cache control: a branch-guided mismatch, where guided error depends on both the magnitudes and alignment of branch errors, and a local-final mismatch, where the downstream impact of a local error varies across timesteps. We propose RA-CFGCache, a Risk-Aligned Caching framework under CFG that incorporates both factors while keeping the sampling schedule and guidance rule fixed. CFG-aware Guided-Risk Composition combines existing branch-wise proxies using CFG coefficients and offline-calibrated cross-branch alignment. Propagation-Aware Rescaling further weights the resulting guided-risk estimate with a timestep-dependent propagation prior calibrated from isolated reuse perturbations. An online threshold controller then determines when to jointly refresh or reuse both branches. Experiments on FLUX.1-dev, Wan2.1-T2V-1.3B, and CogVideoX-2B demonstrate improved efficiency--fidelity trade-offs over evaluated training-free caching baselines. Moreover, RA-CFGCache is compatible with diverse base proxy families, including TeaCache-, DiCache-, and MagCache-style estimators, and consistently improves fidelity at nearly unchanged latency. Code is available at https://github.com/yiming-l21/RA-CFGCache.git. --- ### Temporal-Aware Fusion for Robust Outdoor LiDAR Localization Authors: Minghang Zhu, Zhijing Wang, Yuxin Guo, Chen Liu, Yongshu Huang, Wen Li, Sheng Ao, Cheng Wang Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36432v1 LiDAR relocalization aims to estimate the global 6-DoF pose of a sensor in the environment. However, existing regression-based approaches often encounter limitations in dynamic or ambiguous scenarios, as they typically prioritize single-frame inference, leaving the potential of spatio-temporal consistency across scans not fully explored. In this paper, we propose a Temporal-aware Localization framework (TempLoc) designed to enhance the robustness of outdoor localization by effectively modeling sequential consistency. Specifically, a Global Coordinate Estimation module is first introduced to predict point-wise global coordinates and associated uncertainties for each LiDAR scan. A Prior Coordinate Generation module is then presented to estimate inter-frame point correspondences by the attention mechanism. Lastly, an Uncertainty-Guided Coordinate Fusion module is deployed to integrate both predictions of point correspondence in an end-to-end fashion, yielding a more temporally consistent and accurate global 6-DoF pose. Experimental results on the NCLT and Oxford RobotCar benchmarks show that our TempLoc outperforms state-of-the-art methods by a large margin, demonstrating the effectiveness of temporal-aware correspondence modeling in LiDAR relocalization. --- ### Towards Scalable Context-Aware Single-Cell Spatial Transcriptomics Prediction from Histology Images Authors: Zijun Gao, Chunbin Gu, Jinxi Xiang, Xiangde Luo, Pheng-Ann Heng Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36429v1 Predicting gene expression from H&E-stained histology images offers a scalable alternative to costly spatial transcriptomics, yet most existing methods operate at the spot level, where signals from multiple cells are aggregated and critical cellular heterogeneity is obscured. Extending this paradigm to single-cell resolution is non-trivial. Naively applying pathology foundation models faces a scale mismatch: their patch-level representations mix multiple cells, whereas per-cell cropping or resizing distorts morphology and removes local context. Conversely, segmentation-based models without strong pretrained visual encoders often lack the morphological representation capacity needed for accurate molecular prediction and inherit errors from imperfect cell boundary masks. Here, we present CELLO, an efficient end-to-end framework that performs a single pathology foundation model forward pass per image and uses grid sampling to extract location-specific features for all cells simultaneously. We further introduce a distance-decay cross-attention module that refines each cell representation using spatially biased local morphological context. Using 52 public Xenium-H&E pairs from HEST-1k that span 12 organs and approximately 10 million cells, CELLO improves the average predictive accuracy over the evaluated baselines while reducing the mean whole-slide inference time compared to DeepSpot2Cell, a 14.0x speed-up on average that excludes upstream cell segmentation. Our work establishes a scalable foundation for single-cell gene expression prediction from H&E images. --- ### Losing the name before the box: measuring and repairing what narrow fine-tuning costs a detector outside its deployment vocabulary Authors: Trung Minh Bui, Jongsul Moon, YoungOuk Kim, Jung-Hoon Hwang, Dongin Shin Published: 2026-09-29 Categories: cs.RO, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.36426v1 A detector pretrained on a broad corpus is fine-tuned on a narrow domain, its in-domain accuracy improves, and it ships. We ask what happens meanwhile to its coverage of objects the vocabulary never names, which in obstacle detection and inspection carry the risk. No in-domain test set holds an example of one. We give a longitudinal protocol: one pretrained checkpoint against its own fine-tuned descendants. It tracks held-out top-$K$ proposal coverage $C_τ$: of categories pretraining covered and the vocabulary omits, the share of boxes a detector's top $K$ regions still cover. The quantity is the open-world proposal literature's; the longitudinal reading is not. $C_τ$ falls while in-domain accuracy rises, on four architectures and three domains, by $5.12$ to $63.35$ points on boxes above $1024$ px$^2$. No in-domain number identifies the fall, and neither does detection average precision, which charges a missed and a misnamed box alike. On the one architecture scoring both, adaptation costs $87\%$ of the AP against a fifth of the coverage, and the naming goes first at all six depths of its freeze ladder, every run. What breaks is structured: three architectures sharing no pretraining run agree on which categories lose coverage, and those a model never learned do not lose any. A repair follows and needs no training: mixing a quarter of the pretrained state back, normalisation statistics included, raises coverage on every cell swept for at most $2.47$ points of in-domain accuracy. Seeing it costs one extra evaluation pass. --- ### Sample Complexity of Equivariant Reinforcement Learning Authors: Rayan Mazouz, Haibo Zhao, Chris Hillar, Christian Shewmake Published: 2026-09-29 Categories: cs.CC, cs.LG, math.GR arXiv: https://arxiv.org/abs/2609.36421v1 Reinforcement learning (RL) is a powerful framework for robotic control, yet its practical application is often hindered by high sample complexity. This is particularly restrictive in physical domains where interaction data is costly. While the world often exhibits geometric and physical symmetries, standard RL algorithms typically fail to exploit this structure. In this paper, we demonstrate that exploiting group symmetries significantly reduces the sample complexity of RL. Focusing on finite-horizon Markov decision processes, we find that leveraging homomorphisms induced by group symmetries significantly reduces the theoretical upper and lower bounds on the number of environment interactions required to reach an optimal return. We further extend these bounds to continuous state and action spaces, providing corresponding sample-complexity guarantees under appropriate regularity assumptions. Beyond theory, we validate our findings through controlled experiments and demonstrate the advantages of symmetry-aware policy learning on high-dimensional continuous robotic simulations. Our results show that integrating symmetry into the learning pipeline yields substantial gains in sample efficiency and performance, offering a principled path toward more data-efficient robotics. --- ### Support-Set Target Leakage in Relational Foundation Models during In-Context Learning: Model Dependence and Evaluation Reliability Authors: Roshan Reddy Upendra, Alexandre Dorais, Joe Meyer, Andrew Pouret, Anastasios Lambrianos Stappas, Dinesh Katupputhur Ramprasath, Viswanath Ganapathy, Tom Palczewski, Minghua Li Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36417v1 Relational in-context learning (ICL) uses labeled support examples and their linked relational context to predict labels for new queries. This creates a failure mode when target-derived features are present in the support context but unavailable for the query. We study this setting as support-set target leakage. We construct 20 controlled target-derived features that vary in signal fidelity, representation, semantic transparency, coverage, and zero-, one-, and two-hop relational placement, and evaluate them across 13 RelBench tasks and five relational ICL configurations that vary the ICL head, message-passing depth, pretraining cohort, or relational encoder architecture. We evaluate matched 0-hop, 1-hop, and 2-hop leakage settings, together with a Full leakage condition containing all 20 leaker columns. Within the tested configurations, target-table (0-hop) and Full leakage produce the largest aggregate deviations from clean evaluation, while higher-hop effects are often weaker, consistent with differences in effective exposure associated with temporal reachability, sampling, and aggregation fidelity. Leakage effects are strongly task- and model-dependent and can reverse relative conclusions between model variants even when aggregate changes are small. For leaker detection, we compare an Integrated Gradients (IG)-based screening method with mutual information (MI) and leave-one-column-out (LOCO) on a common Baseline subset. Ranking quality is strongest in the high-impact 0-hop and Full leakage conditions, but detector-based removal does not consistently restore the clean evaluation. A four-task rel-salt case study further shows the same evaluation concern with native-schema leakage candidates from the original relational schema. These results identify the support/query information boundary as an important component of reliable relational ICL evaluation. --- ### FineART: Fine-grained Annotated Robotic Trajectory Dataset and Vision-Language-Action Model for Bimanual Manipulation Authors: Jade Choghari, Pepijn Kooijmans, Mansi Agarwal, Yusuf Umut Ciftci, Aseem Doriwala, Catherine Weaver, Mouli Sivapurapu, Kai Yang, Jackson Lee, Thomas Wolf, Pragna Mannam Published: 2026-09-29 Categories: cs.RO, cs.AI, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.36416v1 Robots operating in real-world environments must execute complex, multi-step bimanual tasks over long horizons rather than single, isolated actions. Current manipulation datasets struggle to support this capability: although single-arm datasets reach hundreds of thousands of trajectories, they typically provide only one high-level instruction per episode while the rare bimanual effort that does label subtasks annotates only a fraction of its hours. We present FineART, a densely annotated bimanual manipulation dataset of 40,543 episodes, 1,718 hours, and 533,913 subtasks across 151 tasks. We also introduce FineART-VLA, a vision-language-action policy that predicts its own next subtask, and show that mid-training it this way yields substantial gains. Specifically, success on a spatial disambiguation task increases from 32.0% to 100.0%, and step-by-step human subtask guidance lifts success on an unseen long-horizon task from 16.0% to 76.0%. Furthermore, after minimal fine-tuning on a new robot, the policy requires one-tenth the data of baselines without mid-training and generalizes zero-shot to completely unseen tasks on the new hardware. We open-source the full dataset, model weights, and training code. --- ### Eternal Sunshine of the Spotless Mind: Systematically Erasing LLM's Memories Authors: Olga Ohrimenko Published: 2026-09-29 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.36414v1 We consider persistent LLMs that accumulate memories of their interactions with a user over time. Such LLMs maintain memories using external storage, which they can query to overcome the limitations of a fixed context window. Such systems have numerous practical applications, as they can draw on all past interactions when responding to user queries. In this paper, we ask whether LLMs can forget information shared with them upon a user's request. We find that current LLMs fail to delete such information---even when they claim to have forgotten it and even when operating with a limited context. To this end, we consider a new direction of study: Deletion of LLM Memories. We show that naively removing messages that match a user's deletion request is insufficient, since conversations naturally introduce message dependencies that cause information to persist. To correctly handle deletion requests, we propose the DeLLM framework. It dynamically constructs relevant context for each LLM query and maintains a provenance graph of messages to determine which ones must be removed during deletion. Our experiments show that DeLLM achieves a high deletion rate while maintaining utility. --- ### Longer Records, Broader Invariance: The Hidden Scaling Problem in Longitudinal Contrastive Learning Authors: Rameen Mahmood, Xuhai "Orson" Xu, Zachary Beattie, Jeffrey Kaye, Danny Yuxing Huang Published: 2026-09-29 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.36409v1 Longitudinal data are valuable because people change. Yet the objectives used to learn from these data can inadvertently erase that change. In person-level contrastive learning, observations from the same person are treated as positives; as records grow, those positives can span increasingly distant---and increasingly different---behavioral states. More history can therefore produce not only more data, but broader invariance. We show that this distinction is fundamental. We separate \emph{record span}, how much history the learner sees, from \emph{supervision span}, how far across that history positive-pair supervision reaches. Across in-home sensing records spanning up to 2.7 years, broader supervision systematically suppresses recoverable changing-state information, even when the available history is held fixed. At the broadest span, less than 10\% of the information recoverable from an untrained encoder remains. Yet keeping positives local is not sufficient: as records grow, even distant states that are never paired become increasingly similar. Explicitly contrasting other observations from the same person reverses this loss without shortening the record, revealing a second route by which longitudinal scale can broaden invariance. Finally, we prospectively reproduce the supervision-span effect in 199 GLOBEM participants. Longitudinal scale therefore presents a choice: more history need not mean more invariance. By controlling what is held invariant as records grow, we can preserve the change that made the longitudinal data valuable in the first place. --- ### What Makes High-Magnification Knowledge Transferable? A Study of Cross-Resolution Distillation in Whole-Slide Imaging Authors: Zhiyuan Yang, Jiahao Cheng, Mahdi S. Hosseini Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36407v1 Cross-resolution knowledge distillation aims to improve low-magnification whole- slide analysis by transferring high-magnification representations, yet the conditions for useful transfer remain unclear. We develop a decomposition-based analysis of teacher access, representation loss, and model excess, motivating three questions: whether (a) teacher targets help the task, (b) low-magnification students can predict them, and (c) slide models benefit from those predictions. We investigate them through controlled experiments across ten pathology cohorts spanning classifi- cation, grading, and survival prediction. In the main comparison, providing teacher regional means alongside native low-magnification features improves downstream performance in all ten cohorts. Direct prediction achieves lower reconstruction error than residual prediction, yet the predicted features underrepresent variation in the teacher targets. Moreover, better reconstruction does not consistently improve downstream scores, and retaining native features changes performance even when the predicted teacher features are held fixed. Together, these findings expose a gap between reconstructing teacher representations and realizing their downstream value. They challenge the sufficiency of reconstruction error as a measure of cross-resolution transfer and provide a diagnostic framework for examining where that transfer breaks down. Future distillation designs must account for both what students can predict and how slide models use those predictions. --- ### From Retrieval to Reasoning: Agentic Mechanism Prediction from Cell Painting Profiles Authors: Jiayuan Chen, Botao Yu, Tianyu Liu, Thai-Hoang Pham, Meng Wu, Ping Zhang Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36406v1 Cell Painting is a high-content morphological profiling assay widely used for phenotype-based biological inference, with mechanism of action (MOA) prediction as a central application. Existing approaches largely formulate Cell Painting-based inference as representation matching, assigning predictions from nearby reference perturbations in morphological feature space. However, retrieved neighbors are often noisy and partially misleading evidence due to batch effects, non-specific cytotoxicity, phenotypic convergence, and source-dependent variability. We reformulate Cell Painting-based MOA prediction as a calibrated evidence reasoning problem, where retrieved neighbors are treated as uncertain observations that must be evaluated, compared, and sometimes rejected before supporting a mechanistic conclusion. We propose PhenoAIR, a reliability-aware multi-agent framework that maintains a candidate-centric evidence memory and performs controller-guided refinement over phenotype- and mechanism-side evidence. PhenoAIR uses offline reference-set calibration to weight evidence by source reliability, phenotype stability, and mechanism-level confusion. We evaluate PhenoAIR on a benchmark constructed from JUMP Cell Painting profiles and annotations, covering controlled, realistic, and discovery-oriented open-world MOA prediction settings. PhenoAIR outperforms representation-matching and LLM-based baselines across all settings. --- ### Layer-Informed Fine-Tuning via Three-Stage Functional Segmentation of LLMs Authors: Junning Shao, Siwei Wang, Zhixuan Fang Published: 2026-09-28 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.38027v1 In recent years, the performance of large language models (LLMs) on reasoning tasks has been remarkable, even surpassing human capabilities on various benchmarks. However, there remains a lack of clear understanding in the academic community regarding how the structure and internal parameters of LLMs progressively solve complex reasoning problems. In this study, we investigate the inference process of LLMs on cross-linguistic materials and propose the hypothesis that LLM layers exhibit a structured division of labor across conceptualization, reasoning, and textualization. Based on this hypothesis, we introduce a bottleneck identification mechanism using sensitivity analysis to pinpoint the most critical functional stage for a specific task. Leveraging this insight, we propose a novel approach, Layer-Informed Fine-Tuning (LIFT), which achieves efficient and effective fine-tuning by selectively updating only these functionally critical layers. We then conduct extensive experiments to show that the LIFT method not only accelerates the training process but also significantly improves model performance. --- ### Evaluating Bounded Autonomy in Regulated Agentic AI: A Diagnostic Harness with Constitutional Rewards, Escalation Labels, and Runtime Governance Authors: Dipankar Sarkar Published: 2026-09-28 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.37501v1 We propose RegLLM, a diagnostic harness for bounded autonomy in regulated agentic workflows. It instruments six trustworthiness signals: citation validity, source grounding, schema compliance, escalation correctness, constitutional alignment, and unsafe-action rate. Signals are distinguished by their source of supervision: programmatic verifiers, task-level escalation labels, or AI-judge scores. A deterministic runtime supervisor blocks ungrounded answers and forces escalation, logging interventions. The same domain constitution informs evaluation, training rewards, and serving guardrails. Task-level should-escalate labels make the act-versus-defer decision a measurable training signal. We demonstrate the harness at smoke scale. An offline reference run (n=12) lifts escalation recall from 0 to 0.67 and reduces unsafe-action rate from 0.33 to 0.08 when governance is enabled. Two single-GPU Qwen2.5-3B LoRA/DPO pilots (n=8, same seed and evaluation split) expose substantial variation: nominally identical RL-base configurations yield task success of 0.25 versus 0.12 and escalation recall of 1.0 versus 0.5. An answer-quality adapter changes recall from 1.0 to 0.5 in Run A, but from 0.5 to 1.0 in Run B. An escalation-aware variant produces no measurable change in Run B. These small pilots do not establish reliable adapter effects or production readiness. Their contribution is diagnostic: configuration variance can overwhelm apparent tuning effects on bounded-autonomy metrics, motivating larger evaluation sets and repeated runs. --- ### REVO: Rollout-Efficient Off-Policy Distillation via Variance-Guided Reuse Authors: Yuxiao Yang, Shangzhe Li, Tianrun Yu, Kaixiang Zhao, Taylor W. Killian, Weitong Zhang Published: 2026-09-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.37500v1 On-policy distillation (OPD) trains language models using dense token-level teacher supervision on student-generated trajectories. However, its reliance on frequently refreshed student rollouts often incurs substantial generation cost. We introduce REVO, an off-policy distillation framework that improves rollout efficiency by reusing each student rollout for multi-step learner updates. REVO addresses prefix-level and current-token policy mismatch through stabilized prefix weighting and one-step resampling from the current student, which enables repeated updates without regenerating full trajectories. To prioritize informative token positions within reused rollouts, REVO uses the variance of the student-teacher log-probability ratio to quantify the remaining token-level learning signal and guide repeated optimization. Across multiple student-teacher scales, REVO with only 50 rollout iterations matches or exceeds OPD baselines trained for 200 iterations on both in-domain and cross-domain reasoning benchmarks. --- ### Your Benchmark Is Not Saturated: Reviving Multiple-Choice Evaluation with Answer Pooling Authors: Mohamed Eltahir, Abobaker Ahmed, Nawaf Barebood, Hussain Bu Subayt, Tanveer Hussain, Naeemullah Khan Published: 2026-09-28 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.37494v1 Multiple-choice benchmarks are cheap to grade and are running out of room, and the standard remedy, writing harder items, is slow and repeated for every benchmark. A saturated benchmark still holds a harder task. Each question's wrong options are written for that question alone, so a model can score by eliminating a few options. We propose AnswerPool: take $N$ questions that share a context, pool all their options into one list, and ask the model to assign every question its answer. No item is written and no label changes. The chance of guessing a group right falls from $10^{-3}$ to $5\times10^{-7}$ for five four-option questions, and a model that recognizes its answers keeps its multiple-choice score, so the accuracy lost to pooling measures the credit the format gave for elimination. Deleting answers from the pool makes questions unanswerable with exact ground truth, so abstention is scored in the same pass. Across eight text, image, and video benchmarks and eighteen models, pooling is harder for every model, the elimination credit is largest for the weakest models, and seven of eight open-weight models answer 87 to 100% of unanswerable questions. --- ### Attention-Scoped Guidance: Training-Free Spatial Control for Image Editing Authors: Zeyan Li, Wei Zhou, Hadi Amirpour, Minghao Zou, Panqi Yang, Jianfeng Xu Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.37492v1 Instruction-guided image editing should change what the instruction names and leave the rest of the image untouched. In dual classifier-free guidance (CFG), an editor combines two directions at every denoising step, one that pushes toward the instructed edit and one that pulls back toward the source image, using global weights. We introduce Attention-Scoped Guidance (ASG), a sampler wrapper that makes these weights spatial. It reads a soft support map from the instruction attention that the editor already computes, then weakens text guidance where support is low and strengthens image anchoring where support is high. The wrapper requires no training, no external mask, and no additional network evaluation. On the full MagicBrush and PIE-Bench++ splits, ASG improves preservation-oriented metrics, leading three of four MagicBrush metrics and PIE-Bench++ background PSNR. A dose-matched control that removes the spatial placement loses up to 0.73 CLIP on PIE-Bench++, confirming that the spatial allocation itself carries the gain. --- ### KT-EGO: A Knowledge Transfer Assisted Efficient Global Optimization Algorithm for Solving High-Dimensional Expensive Black-Box Problems Authors: Qineng Wang, Liming Song, Yun Chen, Guangjian Ma, Zhendong Guo, Jun Li Published: 2026-09-28 Categories: cs.CE, cs.LG arXiv: https://arxiv.org/abs/2609.37473v1 Many engineering problems involve optimizing a high-dimensional expensive black-box (HEB) design space. To solve such problems efficiently, we propose a knowledge transfer assisted efficient global optimization (EGO) algorithm, labeled as KT-EGO, which extends the EGO algorithm for solving problems over higher dimensions (i.e., $d>20$). Specifically, the original design space is divided into several low-dimensional subset design spaces. More importantly, in order to extract information from the subset design spaces to accelerate the progress of full optimization, we propose a surrogate-based data fusion strategy in KT-EGO. And further, a searching strategy with an adaptive variable range is devised to enhance the exploitation of promising areas. To show the effectiveness of our proposed algorithm, it is compared against the state-of-the-art algorithms over 12 benchmark functions and a 28-dimensional engineering optimization for the design of compressor blade, which fully validates the effectiveness of the KT-EGO for solving HEB problems. --- ### Commitment Hierarchies under Intent Revision: A Belief-Revision Account of Salvage in Tool-Use Agents Authors: Spandan Ghose Chowdhury Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.37453v1 When a user changes their mind partway through a task, an agent that has already split the task into sub-goals and paid for tool calls must decide, per cached sub-result, whether to keep, patch, or discard it (salvage), restarting wastes valid work and continuing unchanged answers the old question. Our main finding is that salvage quality is a matter of role design rather than model capability: a language model asked the keep/patch/discard question one node at a time is unreliable, but asked to classify the revision once, with a deterministic layer propagating the decision, it reaches the cost-optimal oracle on all three models tested, from two vendors. Modeling the plan as a commitment hierarchy and the intent change as a belief-revision operator with AGM style postulates, we prove that no policy observing only a node's local view can be both safe and cost optimal, while the single classification design is both. Across three environments the policy recovers the full achievable savings, 43% cheaper than restart, at 100% correctness. --- ### CAMEO: A Class-Activation-Mapped Equitable Overlay Framework for Fair and Robust Deep Learning-based Skin Condition Diagnosis Authors: Youssef Attia, Debasmita Mukherjee Published: 2026-09-28 Categories: eess.IV, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.36400v1 Deep learning classifiers for dermoscopic skin lesions often reach high in-distribution accuracy while quietly relying on spurious background cues such as skin tone, device vignetting, and embedded rulers, rather than on lesion morphology. This undermines robustness and fairness across skin tones. This work asks whether Explainable AI (XAI), typically used only to audit a finished model, can instead be repurposed as an active training signal that corrects this shortcut without sacrificing diagnostic accuracy. We introduce CAMEO (Class Activation Mapped Equitable Overlay), a framework that improves skin-lesion classification by selecting stable model explanations and using them to separate lesions from their backgrounds. It then replaces the background with realistic synthetic skin while keeping the lesion unchanged. On HAM10000 and dark-skin ISIC images, CAMEO maintained accuracy while reducing background-driven errors by nearly four times. It also made the model's attention more consistent when backgrounds changed. Results across multiple tests show that reducing reliance on background information improves robustness, with Fitzpatrick-based backgrounds providing a realistic and interpretable approach. Results show that XAI-guided augmentation can make dermoscopic classifiers measurably more robust and fair at no cost to accuracy. They also clarify that it is the mechanism and not the specific tone palette that matters, and that the lasting contribution of XAI here lies in stability-screened, annotation-free lesion localisation rather than in the robustness number itself. --- ### Calibrated to Whom? Persona and Language Effects on Cultural Values in JEV Authors: Bushra Asseri, Abdulaziz Asseri Published: 2026-09-28 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.36399v1 Decision-only language models return a probability for every answer option instead of generating text, which makes them attractive as survey respondents and as judges. We audit the cultural values of one such model, TypeSafe's JEV, with the Values Survey Module 2013. We asked it the 24 items as 12 matched Saudi and 12 matched American personas and without a persona, in English and Arabic, under eight ways of formulating the request (288,000 answers). JEV's answers were highly repeatable (ICC 0.997), and without a persona they resembled those of its own American personas. When the persona was Saudi rather than American, the answers moved in the direction of the human Saudi-US difference, reproducing 87% of its size in English but 62% in Arabic, with long-term orientation reversed. A language cross shows that the smaller difference in Arabic comes from the language of the items, not from the language of the persona description. Age shifted the profiles about as much as nationality, gender shifted them more for Saudi than for American personas, and JEV was less confident in Arabic and for Saudi personas. These patterns held in every request design, although the model never generates text. --- ### Where Should Physics Enter a Molecular Crystal Generator? Authors: Haocheng Tang, Junmei Wang, Wengong Jin Published: 2026-09-28 Categories: q-bio.BM, cs.AI, cs.CE arXiv: https://arxiv.org/abs/2609.36398v1 Generative models make molecular crystal structure prediction fast, but their samples still exhibit geometric and packing violations. Physics can be introduced during training, post-training, or inference, yet these choices are rarely compared with the generator and physical signal held fixed. We introduce CrystAF, an all-atom crystal flow-map generation model, and use it with the UMA interatomic potential to systematically study where physics should enter. Post-training learns physical preferences directly into CrystAF, improving molecular validity and crystal packing while leaving sampling unchanged: physics is paid for once during training rather than repeatedly at deployment. In contrast, UMA relaxation is effective at repairing local clashes but makes generation 6--26$\times$ slower, while learning from relaxed targets provides little benefit. These routes are complementary rather than competing. Physics-informed post-training first shifts the generated distribution toward more physically reasonable structures, after which inexpensive inference-time corrections further remove clashes and restore stereochemistry that the generator cannot represent. Importantly, the same post-training strategy also improves the multi-step all-atom Clari-M and rigid-body MolCrystalFlow generators, demonstrating transfer across architectures and representations. Together, our results suggest a simple principle: learn reusable physical alignment into the generator, and reserve inference-time physics for residual constraints that are better corrected than learned. --- ### LOCO-AdaMP: Built-in LOCO Inference for Adaptive Minipatch Ensembles with Enhanced Prediction Authors: Yinan Cheng, Lili Zheng Published: 2026-09-28 Categories: stat.ML, cs.LG, stat.ME arXiv: https://arxiv.org/abs/2609.36396v1 As black-box machine learning models become increasingly common, extracting interpretations with uncertainty quantification has become a critical challenge. One popular type of interpretation is leave-one-covariate-out (LOCO) feature importance, while prior LOCO inference methods often require data-splitting or model-refitting. A recent ensemble framework, LOCO-MP, addresses these challenges using minipatches that subsample both observations and features, but massive feature subsampling can hurt prediction in high-dimensional sparse settings. Motivated by this limitation, we consider minipatch ensembles with adaptive feature sampling guided by LOCO importance, and propose LOCO-AdaMP, which enables free LOCO inference for the resulting adaptive minipatch ensemble. We show that LOCO-AdaMP yields substantially improved predictive models while retaining asymptotically valid feature importance inference without data-splitting, despite the complex dependence between the adaptive sampling distribution and the LOCO importance statistics. Our analysis relies on a careful leave-two-out perturbation bound for the iteratively updated sampling probabilities together with the stability of LOCO scores induced by observation subsampling. Empirical results on synthetic and real datasets demonstrate advantages of LOCO-AdaMP over existing methods in predictive performance, inferential power, and stability. Overall, LOCO-AdaMP provides a flexible ensemble framework (agnostic to base models) that delivers both strong predictive performance and asymptotically valid, powerful feature importance inference for regression. --- ### Reward-rate Policy Gradient for Efficient Machine Learning Engineering Agents Authors: Muhang Tian, Sherry Yang Published: 2026-09-28 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.36393v1 Traditional reinforcement learning (RL) techniques focus on maximizing expected cumulative reward, where each action assumes to take a constant unit of time. However, this assumption does not hold for agentic RL tasks such as machine learning engineering (MLE) agents, where actions involve data loading, feature engineering, and model training that take variable durations. Efficiency matters in modern agentic RL where actions are costly. To address this limitation, we adapt from continuous-time RL and Semi-Markov Decision Process (SMDP) formulation and propose Reward-rate Policy Gradient (RPG), where we focus on optimizing the reward rate -- the long-term reward per unit of time. RPG estimates the reward rate from off-policy samples, then charges each action for the time it consumes at that rate. We first conduct theoretical analysis in the bandit setting to establish that RPG approximates the optimal reward rate and empirically demonstrate it outperforms baselines while avoiding enumeration over the policy space, a known issue for an existing method. We then further apply RPG on a small language model (Qwen3.5-4B) with self-improvement loops and empirically show it obtains higher rewards within a fixed time budget than vanilla RL on MLE-Bench and NanoGPT, with a 19.2% and 85.7% margin, respectively. Our method provides a practical solution for optimizing performance under wait time considerations in modern agentic RL tasks, where actions interact with external environments and cost time. --- ### ARCagent: An Adaptive Retrieval Calibration Agent for Clinical Question Answering Authors: Yuyan Chen Published: 2026-09-28 Categories: cs.AI, cs.IR arXiv: https://arxiv.org/abs/2609.36392v1 In diseases where clinical guidelines are incomplete, contested, or mutually contradictory, knowledge completeness and dynamic conflict-aware synthesis are two safety-critical properties that standard Retrieval-Augmented Generation systems do not provide. Therefore, we present \sysname, an adaptive retrieval calibration clinical question-answering agent for ME/CFS, a disease where diagnostic frameworks coexist and major guidelines actively contradict each other on treatment. ARCagent contributes three components. First, a 1,706-chunk, 10-source knowledge base with a structured inter-guideline conflict registry spanning all active ME/CFS diagnostic frameworks. Second, a conflict-aware retrieval calibration pipeline that re-ranks retrieved evidence using query-specific focus and conflict signals. Third, a benchmark scored by LLM-as-Judge, avoiding systematic underestimation averaging 10.1 percentage points caused by keyword matching. ARCagent achieves 95.3%, outperforming all base LLMs. Code is available at https://github.com/Yukyin/ARCagent. --- ### Receptive-field-constrained stimulus optimization for human early and intermediate visual cortex Authors: Junru Zhao, Hanfei Guo, Andrew Luo, Margaret M. Henderson Published: 2026-09-28 Categories: q-bio.NC, cs.LG arXiv: https://arxiv.org/abs/2609.36391v1 An ongoing challenge in sensory neuroscience is to characterize the feature dimensions encoded by cortical populations. Recent approaches probe feature selectivity in a data-driven way, by synthesizing a most-exciting-input (MEI) for a target neural population. While this approach has been successfully applied to human higher visual cortex using fMRI data, generating MEIs for early- and mid-level retinotopic visual areas requires additional modeling constraints due to small receptive field sizes. To address this challenge, we introduce two novel MEI generation frameworks, Receptive Field Diffusion for Visual Exploration (RF-DiVE) and Receptive Field Gradient Optimization (RF-GO). Both methods use a population receptive field (pRF)-constrained voxelwise encoding model; RF-DiVE combines this with a pretrained latent diffusion model, while RF-GO uses regularized gradient ascent. When applied to single voxels in retinotopically defined areas V1-hV4, using data from the Natural Scenes Dataset, we obtain MEIs that exhibit consistent structure within the pRF, suggesting selectivity for local features like contour, color, and texture. We systematically compare MEIs generated by RF-DiVE and RF-GO using two encoding backbones, performing in-silico validation of predicted responses to MEIs using independent encoding models. Across all methods and all visual areas, MEIs elicit higher model-predicted responses than the most activating natural images. We further find that the choice of generation framework and encoding backbone differentially affects MEI properties, including their visual appearance, structural interpretability, and cross-model generalizability. These results offer a new approach for performing data-driven characterization of spatial and feature selectivity across human visual cortex. --- ### Finite-Sample Theory for Fitted Q-Iteration When Actions Are Functions Authors: Gefei Lin, Rui Miao, Xiaoke Zhang Published: 2026-09-28 Categories: stat.ME, cs.LG, math.ST, stat.ML arXiv: https://arxiv.org/abs/2609.36390v1 Offline reinforcement learning seeks optimal decision rules from previously collected data. In some applications, a decision can be an entire function, such as a fluence map in radiation therapy or a smooth movement trajectory in robotics. In this paper, we study the finite-sample theory for fitted Q-iteration (FQI) with functional actions in a discounted infinite-horizon setting. Three major difficulties arise in this setting: first, the absence of a Lebesgue probability density for functional actions complicates coverage descriptions; second, conventional coverage requirements can be restrictive; and third, the large functional action space makes greedy optimization in FQI challenging. To address these difficulties, we study smoothness-regularized policy search under a critic-relative coverage condition. This condition measures how well logged data distinguish relevant action-value differences without requiring an action density. Our main theorem gives finite-sample guarantees for learned-policy regret relative to the best value within a fixed smooth class of functional-action policies. The results allow trajectory lengths to be either bounded or growing and Q-functions to be fitted by either functional-input kernel ridge regression or adaptive functional neural networks. For a few examples, we can obtain polynomially decaying regret bounds in the number of logged transitions, up to logarithmic factors, with logarithmically many FQI iterations. Numerical experiments show gains of learned functional-action policies over constant-action policies and support our adoption of a critic-relative coverage condition. --- ### Persona Dosing: Calibrated Activation Steering for Graded Trait Control Authors: Zehao Jin, Junran Wang, Ruixuan Deng, Jiahao Chen, Jingyuan Zhang, Yuxuan Zhang, Xinjie Shen Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36388v1 An activation-steering coefficient sets intervention strength, but requesting a particular degree of persona expression requires a behavioral scale. We study persona dosing: controlling a language model through a trait description and a requested mean intensity. PersonaDose specializes a shared, description-conditioned FLAS controller on persona responses, then calibrates its flow time against measured trait expression. Training responses are not paired with requested target intensities. Across Llama-3.1-8B, Qwen3-8B, and Gemma-3-4B, PersonaDose raises core-trait expression at the Persona Vectors coherence floor of 75 by 33.2, 18.3, and 17.8 points over contrastive activation addition. Calibration-selected settings retain an expression advantage on held-out questions, although the coherence floor does not hold for every trait there. Across seven trained traits, calibrated requests yield mean targeting errors of 4.7-6.2 points over 14-22 calibration-reachable targets out of 28 per model. These results separate the behavioral range learned by a controller from the accuracy of requests within that range. --- ### Stealth Is a Relation, Not a Property: How Event Representations Create Blind Spots for Timing Attacks in Event-Based Perception Authors: Shoaib Ahmed Dipu, Md. Shaown Miah, Kamrul Hasan, Sayeed Shafayet Chowdhury Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36386v1 An event camera produces an asynchronous stream, but what is visible in that stream depends on how a downstream consumer, such as a model or detector, processes time. The same timestamp change may leave a coarse temporal representation unchanged while changing the response of a model that preserves finer timing. We characterize this dependence as observer-relative stealth. For recorded event streams, retiming an event within its protected accumulation window leaves the accumulated integer tensor exactly unchanged. We use this exact blind space to construct Null, a gradient-guided timestamp-retiming attack, and define SC-ASR_A(tau) to measure attack success while bounding the change visible to observer A. On DVS Gesture at a 10% event budget, Null reaches 81.56 +/- 5.81% ASR on ConvSNN and 98.67 +/- 0.45% on a GRU while preserving the protected tensor exactly. On DailyDVS-200, a protocol-scale Multi-View Fusion Network variant reaches 99.28 +/- 0.11% exact-null ASR, compared with 9.70 +/- 1.06% for its matched control. In a five-attack comparison, Null is the only method with nonzero attack success at exact observer equality, reaching 81.4% on DVS Gesture and 87.35% on DailyDVS-200. We also search the same exact blind space with an independently implemented constrained projected-gradient optimizer, C-PGD. At matched victim-gradient evaluations, C-PGD reaches 84.50 +/- 2.89% ASR on DVS Gesture and 89.55 +/- 4.39% on DailyDVS-200, again with exact protected equality. Perturbations that are exactly hidden from the protected observer become visible under shifted, finer, overlapping, and randomized temporal views. Adding observer constraints reduces the real-valued blind-space fraction from 87.5% to 75.0% to 62.5%, while DVS ConvSNN ASR falls from 74.9% to 61.9% to 37.2%. These results show that stealth is not a property of the perturbation alone. --- ### Support-Set Target Leakage in Relational Foundation Models during In-Context Learning: Impact, Detection, and Mitigation Authors: Roshan Reddy Upendra, Alexandre Dorais, Joe Meyer, Andrew Pouret, Anastasios Lambrianos Stappas, Dinesh Katupputhur Ramprasath, Tom Palczewski, Minghua Li Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36384v1 Relational in-context learning (ICL) conditions predictions on the labeled support examples and their linked tables, creating a failure mode when the support set contains target-derived features that are unavailable for the query. We formulate this problem as support-set target leakage, distinct from leakage during dataset construction, temporal splitting, or representation learning. Here, the target-derived (leaker) columns are present only in the labeled support set during relational in-context inference, while queries remain clean. We construct 14 synthetic leaker types, corresponding to 20 columns, spanning proxies with different noise levels, coverage, modalities, semantic transparency, and relational distances. We evaluate a frozen relational encoder with an ICL head on held-out RelBench databases and use Integrated Gradients (IG) to rank and remove suspicious columns. Our results show that the effect of support-set leakage varies across tasks and relational distances. Target-table leakers cause the clearest degradation, while one- and two-hop leakers are not consistently used by the model. IG ranks target-table leakers highly across datasets and partially recovers performance in settings where leakage has the largest effect. --- ### LEGO-Anything: Coding Agents for 3D Scene Reconstruction Authors: Xirui Li, Peng Shi, Mingwen Dong, Sheng Zhang, Zhuoyan Xu, Dongkyu Lee, Shuaichen Chang, Yi Xiang, Lin Pan, Jiarong Jiang Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36380v1 A 3D scene reconstructed from a single image is most useful when represented not as a rendering or a fixed 3D output, but as an explicit scene program whose execution yields a scene that can be inspected, edited, and queried. We present LEGO-Anything, an Image-to-Code framework in which a coding agent iteratively writes and executes Blender code, inspects scenes and renderings, and revises the program. To evaluate end-to-end scene recovery, we introduce LEGO-Bench, a simulator-grounded benchmark with 208 images from 104 diverse indoor and outdoor scenes. LEGO-Bench separately scores artifact validity, visible-surface geometry, and rendered appearance. Its simulator-grounded design enables extensibility and precise automatic evaluation. Among evaluated agents, GPT-6-astra achieves the strongest overall results, with 53.4% indoor and 39.6% outdoor scores, yet substantial gaps remain between delivering valid scene artifacts and faithfully recovering scene geometry and appearance. Analysis of agent construction trajectories reveals three recurring issues: weak scene initialization, regressive edits during iteration, and unreliable self-evaluation. These findings motivate LEGO-Plugin, a training-free harness plugin for more controlled iterative scene construction, which improves all six evaluated models, with relative gains of up to 62.7% in overall score. Finally, we test whether reconstructed scenes can represent natural images and support vision tasks. In LEGO-World, we derive object detections, instance masks, and relative depth as deterministic queries on scenes reconstructed by GPT-6-astra. These readouts show non-trivial performance across all three tasks but fall well short of specialized vision models, suggesting that program-constructed scenes from current coding agents are a promising but not yet sufficiently precise representation of natural images. --- ### Quantization Enables Private Dense Retrieval against Malicious Service Providers Authors: Louis Tremblay Thibault, Sofiane Azogagh, Marc-Olivier Killijian, Ulrich Aïvodji Published: 2026-09-28 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2609.36376v1 Dense retrieval, the key component of Retrieval Augmented Generation (RAG), retrieves the most relevant documents by comparing dense vector representations of queries and passages from a large corpus. In privacy-sensitive applications, the server observes the query and controls which evidence is returned, creating both confidentiality and integrity risks. We formulate private dense retrieval as providing query privacy and retrieval integrity against a malicious server, and develop a two-round cryptographic protocol that provides both guarantees. Our protocol reduces private and verifiable retrieval to multiplication of a committed matrix by an encrypted vector and uses low-bit quantization to make this computation practical. We evaluate the resulting trade-off between cryptographic cost, retrieval quality, and downstream RAG accuracy across six embedding models, four language models, and corpora of up to 2.68 million passages. Our results show that, with a clipped quantizer, three-bit quantization largely preserves retrieval quality and downstream accuracy, while a private query over a corpus the size of a clinical reference requires one to three minutes of server time. These results suggest that private dense retrieval is already practical for moderately sized, privacy-sensitive corpora when minute-scale latency is acceptable. --- ### Neural Succession: A Mesoscopic Theory of Invasion, Coexistence, and Stabilization in Continual Learning Authors: Shoaib Ahmed Dipu, Md Salman Shamil, Sayeed Shafayet Chowdhury Published: 2026-09-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.36375v1 Continual learning is usually studied through mechanisms that preserve old knowledge. We develop Successional Learning Theory (SLT), a mesoscopic account in which the current representation is a resident community, the incoming task is an invader, forgetting is resident displacement, joint retention is coexistence, replay is resident reinforcement, and training moves from establishment toward stabilization. Its empirical coordinate is directional pre-invasion compatibility, measured on the resident model before the incoming task is learned. Across eight experiments, compatibility orders later forgetting on the 20 directed Split-CIFAR-10 transitions (three-repeat r=-0.789, incoming-task cluster 95% CI (-0.90,-0.72), every repeat alone r<=-0.67), forecasts held-out forgetting with 24% lower error than a no-information baseline, and reproduces under controlled MNIST permutations and CIFAR-10 rotations (r=-0.804, -0.718). On an 84-transition suite, compatibility separates coexistence from exclusion at every retention threshold (AUC 0.93-0.97). Replay repairs every transition with at most 325 stored examples and is most efficient where displacement is largest. Compatibility reaches |r|=0.720, while activation, representation, Jacobian, and fixed-coefficient Lotka-Volterra specializations do not. Plasticity and feature turnover fall reliably from early to late training (15/15 and 14/15 runs). We formalize a minimum habitat-modification bound, a displacement floor, a sufficient coexistence condition, an identifiability law with a range-restriction corollary, successional stabilization, and local reinforcement. The identifiability law also predicts where the coordinate loses leverage, and the prediction matches three CIFAR-100 partitions and five optimizer regimes. SLT is a pre-adaptation diagnostic that complements replay, regularization, and projection methods. --- ### OTT3R: Multi-View 3D Reconstruction and Fast Dataset Generation at 1% Compute Authors: Brandon Leblanc, Charalambos Poullis Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36374v1 Feed-forward 3D reconstruction models have achieved impressive performance by scaling model and dataset size, but their cost excludes most research groups and precludes edge deployment. Additionally, generating 3D supervision without sensors still relies on slow, unreliable Structure-from-Motion, as the community lacks a COLMAP-like system for neural 3D pseudo-label generation. We present OTT3R (RGB-Only Tiny Transformer for 3D Reconstruction), a knowledge distillation framework that addresses both problems on a single workstation equipped with 2 GPUs. Distilling $π^3$ (959M parameters) into a 102M-parameter student yields 9.4$\times$ compression and up to 7$\times$ faster inference, trained at 1.6% of VGGT's training compute. An integrated pseudo-label pipeline offers a reliable, high-throughput alternative to COLMAP, generating dense per-pixel point maps and SE(3) camera poses for a 667K-image corpus in 3.5 hours on two commodity GPUs and succeeding on every sequence we tested, including those where COLMAP fails. The general student tracks the teacher on in-distribution monocular depth and, zero-shot, outperforms COLMAP on 7-Scenes and on DTU completion, but it does not replace the teacher on out-of-distribution multi-view geometry. The deployable artifact is the domain-specialized student: after specialization at 0.2% compute, it is 4$\times$ more accurate than COLMAP on 7-Scenes at 980$\times$ throughput, with near-teacher completion. Code is available at https://github.com/TheFourthKaramazov/OTT3R --- ### Audience-Bound Persistent Memory: Authorization Across the Memory Lifecycle Authors: Sibo Liu Published: 2026-09-28 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2609.36373v1 A personal language agent that acts for its owner across private and shared conversations can learn a fact from one audience and later place it in the context it assembles for another. We study authorization before context across the whole memory lifecycle. Each memory item carries the audience present when it was recorded; derived items are partitioned by audience, receive the intersection of their sources' audiences, or are suppressed; an audience widens only by an explicit, object-specific grant; and an item enters a model attempt only when every current viewer belongs to one of its authorized audiences, with unresolved viewers failing closed to public-only. Under explicit identity, provenance and complete-mediation assumptions, this admission is sound and policy-complete on the exact assembled context, enforced by exclusion rather than by model behavior. We realize it in two independently persisted reference architectures, a flat store and a relationship graph, and, descriptively, in a native agent-memory runtime. In a prospectively frozen confirmation over 10,000 multi-party histories, no forbidden item entered any architecture's context, whereas unscoped retrieval exposed forbidden items in 82% of its contexts. Entitled recall matched policy-equivalent baselines exactly and exceeded unscoped retrieval by 0.30 Recall@5, with a Holm-confirmed advantage that grows with distractors. No architecture produced a wrong-principal substitution, but unscoped substitutions were too rare to establish the prespecified joint decision. --- ### TTMark: Pairwise Distortion-Free Watermarking Beyond Single-Token Entropy Authors: Ruibo Chen, Zhengmian Hu, Donghang Lu, Xuehao Cui, Georgios Milis, Yihan Wu, Jian Du, Heng Huang Published: 2026-09-28 Categories: cs.CR, cs.CL arXiv: https://arxiv.org/abs/2609.36372v1 Distortion-free watermarking enables reliable attribution of machine-generated text while preserving output distribution. However, existing methods operate independently on each generated token, making their detection capability fundamentally constrained by the entropy of the next-token distribution. We present Tandem Token WaterMark (TTMARK), a general pairwise watermarking framework that extends distortion-free watermarking from individual tokens to adjacent token pairs. By watermarking the joint distribution of consecutive tokens, TTMARK enlarges the effective watermarking alphabet from V to $V^2$, allowing the detector to exploit both token entropy and conditional entropy while preserving distortion-freeness over the joint distribution. We further introduce a branch-isolating concatenated tandem generation algorithm that efficiently constructs the joint distribution in a single forward pass. Theoretically, we show that pairwise watermarking achieves better expected detection strength in low-entropy regimes. Extensive experiments across multiple language models, datasets, and three representative distortion-free watermarking schemes demonstrate that TTMARK consistently improves detectability without degrading generation quality, while also improving robustness to edits and substantially enhancing localized watermark detection. --- ### LatentSift: Policy-State Filtering for Token-Efficient Verification of Software Engineering Agents Authors: Yuning Han, Yangchenchen Jin, Tyler Jandreau, Jingwei Sun Published: 2026-09-28 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2609.36371v1 Test-time scaling improves software engineering agents by generating multiple candidate trajectories and selecting the best one. Verifying and selecting among these long interactions can consume as many tokens as generation itself. Existing hybrid workflows first apply an LLM-based execution-free (EF) verifier to filter candidates before running tests, which adds another model pass over every trajectory. We introduce LatentSift, a token-free and execution-free filter that replaces this first stage with hidden states the policy already produces while generating the candidates. It represents each candidate through its reasoning, observation, and function-call states, compares them with positive and negative banks of such states collected from successful and unsuccessful trajectories during policy training, and fuses the resulting distance scores with a learned linear score to retain promising candidates for the execution-based stages. On SWE-bench Verified, across three agents and two policy sizes, LatentSift cuts EF-verifier tokens by 66.6--81.0% and total verification tokens, which include test generation, by 49.1--62.1% at K=16, while hybrid Best@16 matches or improves on each agent's reference workflow, rising from 59.26% to 60.06% on DeepSWE-Preview. --- ### AdaKerNet: Neural Kernel Decoding for Task-Adaptive Prediction with Multimodal Large Models Authors: Konstantinos D. Polyzos, Eleni Oikonomou, Tara Javidi Published: 2026-09-28 Categories: cs.LG, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2609.36368v1 Large foundation models have been introduced with the promise of efficient adaptation to downstream tasks. Yet, under limited supervision, MLLMs, an important class of large foundation models, remain challenging to adapt to various downstream tasks. Adaptation typically relies either on MLLM parameter fine-tuning or on training neural-based decoders. Both approaches struggle under limited supervision, while fine-tuning additionally requires access to model parameters, which is often unavailable for closed-source models. We introduce AdaKerNet, a novel learnable task-adaptive neural kernel decoder. AdaKerNet is fully agnostic to the parameters of the underlying MLLM and operates solely on its (frozen) rich representations obtained from the diverse available modalities. AdaKerNet relies on (i) a set of learnable, Lipschitz-controlled multimodal features derived from these MLLM representations; (ii) a reference kernel that provides a soft structural prior on those features; and (iii) a lightweight nonlinear neural predictor that adaptively deforms that structure. Learning the kernel representation and the neural predictor jointly within a unified optimization framework allows AdaKerNet to capture features and geometric relationships relevant to the downstream task. Numerical tests across four MLLMs: BLIP-2, LLaVA-1.5, Qwen2.5-VL, and Gemini Embedding 2, and multimodal inputs spanning text, audio, images, and tabular measurements demonstrate significant and consistent improvements over direct MLP, attention-, autoencoder- and kernel-based decoders, across a range of scarce-label budgets, with average error reduction of up to 41% across baselines. These results establish AdaKerNet as an effective approach for prediction from frozen multimodal representations in the scarce label regime. Additional structural ablations highlight the complementary contributions of AdaKerNet's components. --- ### Cross-attention encoding models reveal dynamic spatiotemporal routing across human higher visual cortex Authors: Iishaan Inabathini, Margaret M. Henderson Published: 2026-09-28 Categories: q-bio.NC, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.36366v1 Understanding how the brain parses actions and events from time-varying natural inputs is a central challenge in neuroscience. Recent work has used deep neural network (DNN) models to build stimulus-computable fMRI encoding models that predict single-voxel responses to complex natural videos. However, the majority of video-computable encoding models predict responses using simple linear mappings from model tokens, overlooking the spatiotemporal structure shared by video representations and neural responses. Recent cross-attention encoding models address this limitation for static images, enabling flexible stimulus-dependent weighting of image content across space. Here, we extend this framework to naturalistic video, using per-parcel cross-attention to dynamically route features from a self-supervised video model (V-JEPA-2) across both space and time, fitting this model to fMRI responses to short video clips. We compare joint spatiotemporal attention with factorized and selectively constrained alternatives, and find that joint routing improves predictions of brain responses to held-out videos across higher visual regions, most consistently in lateral and dorsal visual areas associated with dynamic motion perception. Moreover, our method provides interpretable, stimulus-specific attention maps that dynamically follow moving objects, revealing which locations and temporal moments contribute to each neural response. We further show that attention maps from parcels in different category-selective networks (face-, body-, scene-selective) differentially weight content in accordance with expected semantic selectivity. Together, this work provides a new computational framework for understanding how visual information is adaptively weighted by cortical populations during dynamic visual perception. --- ### Engineering Simplicity: Simple Mechanism Interfaces Steer LLM Agents Authors: Kehang Zhu, Anand Shah, David Parkes Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36365v1 Can interaction formats and textual scaffolds help large language model (LLM) agents make better decisions, and do better decisions come with better explanations? We study these questions in auctions and matching, multi-agent environments with explicit rules and known optimal strategies. These settings let us vary how a decision problem is presented while retaining a benchmark for evaluating behavior. Drawing on human-motivated theories of simplicity, we compare interfaces that elicit a complete bid or ranking with sequential interfaces that make safe choices easier to identify. We then hold the interaction format fixed and vary reasoning scaffolds and rule descriptions. Across four model families, the ascending auction interface substantially reduces bid deviations. The matching comparison also shows why sequential responses require different error accounting from complete rankings. Laying out payoff contingencies and explaining why truth-telling is safe also improve choices, whereas prompts to plan through matching rounds or form beliefs about opponents worsen play overall. In auctions, these behavioral gains are not accompanied by corresponding improvements in measured verbal indicators of strategic understanding in the agents' short stated plans. Other prompts change those indicators without improving bids. Our findings suggest that human-motivated theories of simplicity can inform the design of decision environments for artificial agents. They also show why scaffolds should be evaluated through realized choices as well as explanations: improvements in one need not appear in the other. --- ### Compress to Remember: Learning Compact Memory via On-Policy Distillation for Long Video Generation Authors: Xiaoyu Wu, Weihang Guo, Yifei Wang, Xinze Feng, Lydia E. Kavraki, Zhiwei Steven Wu Published: 2026-09-28 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.36364v1 Standard video generators do not natively compact historical context into reusable memory tokens. As generation continues, the growing history makes it increasingly difficult to retain information from earlier frames due to long-context degradation. Key-frame-based approaches address this challenge by retaining selected past frames, but can discard information needed for future generation. Rather than relying on frame selection alone, we study whether a frozen video generator can supply the supervision needed to learn a compact representation of the history. We propose Prediction-Aligned Context Compaction (PACC), which uses a learned compressor to aggregate information across past frames into compact memory tokens. We train the compressor through on-policy distillation, using the same frozen generator both as a student when conditioned on compressed memory and as a teacher when conditioned on the full history. The student generates continuations, while the teacher provides targets for the same noisy inputs at each denoising step. Only the compressor is updated to align the student's predictions with these targets. We evaluate PACC on MBench, which jointly measures memory-event coverage and consistency. PACC outperforms the strongest baseline by 6.63 points on Causal-rCM and 3.19 points on Causal Forcing. Evaluation on VBench-Long using MovieGen prompts further shows that PACC produces minute-long videos with generation quality competitive with baselines. Together, these results show that learning to compact historical context can improve long-video memory without modifying the underlying generator. --- ### Better Nearest Neighbor Graph Indices via (Efficient) LLM-Guided Pruning Authors: Fangzhou Wu, Haike Xu, Sandeep Silwal Published: 2026-09-28 Categories: cs.AI, cs.IR arXiv: https://arxiv.org/abs/2609.36359v1 Graph-based approximate nearest neighbor search (ANNS) is widely used for large-scale semantic search. Its indices are constructed primarily based on geometric relationships among embeddings of an input dataset (e.g., documents or images), rather than explicitly optimizing for semantic relevance. However, when using these indices for downstream query retrieval, performance is evaluated based on the semantic relevance of the retrieved results to the query. This creates a fundamental "geometry-semantic" mismatch between how the indices are constructed and how their retrieval results are evaluated. While existing LLM-based reranking methods can partially mitigate this mismatch at query time, they leave this underlying structural problem in the graph unresolved. We therefore propose LLM-Guided Graph Pruning (LGP), a general framework that addresses this mismatch directly by leveraging LLM reasoning to refine an existing ANN graph index itself. LGP identifies structurally "low-value" neighbors of nodes and replaces them with LLM-selected alternatives that provide useful semantic information while retaining desired geometric structures of the original graph, including sparsity and efficient navigability. Experiments on representative semantic retrieval benchmarks show that LGP consistently improves end-to-end retrieval performance over both vanilla greedy graph search and LLM-based reranking across widely used graph-based ANN indices such as DiskANN and HNSW. --- ### HyperZip: Efficient Data Compression through Personalized Diffusion LLMs with Hypernetworks Authors: Thai Nguyen, Khang Tran, NhatHai Phan Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36357v1 Large language models (LLMs) have shown strong potential for lossless data compression, but existing approaches are constrained by the high computational cost and low throughput of autoregressive decoding. We propose HyperZip, an efficient and scalable LLM-based compression framework that leverages diffusion-based LLMs (dLLMs) with Multi-Token Prediction (MTP) to accelerate LLM-based data compression processes. We identify a trade-off in diffusion-based compression, where increasing decoding throughput degrades the compression rate. To mitigate this trade-off, HyperZip employs a hypernetwork to generate data-specific updates from a context representation, adapting the dLLM to the target data without costly fine-tuning, resulting in a low compression rate and high throughput. Extensive experiments show that HyperZip achieves a superior trade-off between compression rate and speed compared with state-of-the-art baselines. --- ### Explainability from Training with Applications to TCR-Epitope Prediction Authors: Jiarui Li, Zixiang Yin, Samuel Landry, Zhengming Ding, Ramgopal Mettu Published: 2026-09-28 Categories: cs.LG, cs.AI, cs.CE, q-bio.BM arXiv: https://arxiv.org/abs/2609.36354v1 Deep learning models have achieved strong performance in artificial intelligence for science, yet their black-box nature limits our understanding of how they learn scientific tasks. Existing methods for interpretability provide limited insight into how models organize evidence and evolve during learning. We introduce explainability from training (EFT), a model-agnostic paradigm that traces model interpretation during training to explain why models rely on specific features and how they organize these features as predictive evidence. We apply EFT to four state-of-the-art T cell receptor (TCR)-epitope prediction models, TCR-SRIM, TULIP, MixTCRpred, and NetTCR-2.2, spanning post-hoc and interpret-by-design approaches as well as transformers and CNNs. To investigate how structural information affects model explanations, we introduce a benchmark, TCR-XAI2, containing 388 unique experimentally resolved TCR-epitope structures, complemented by structures predicted using AlphaFold3, Boltz-2, TCRModel2, tFold-TCR, and OpenFold3. Using EFT with TCR-XAI2, we demonstrate that (1) CNN and transformer models exhibit distinct learning trajectories; (2) TCR $α$ and $β$ evidence can conflict during learning, limiting the benefits of jointly modeling both chains, while MHC information mitigates this; and (3) real versus predicted structural data for TCR-epitope prediction exhibits distinct TCR and peptide feature preferences as well as differing trajectories of model certainty. --- ### StructRL: Online Structured Reinforcement Learning for Long-Horizon Vision-Language-Action Tasks Authors: Ziyi Yin, Sangmin Woo, Kang Zhou, Sungyeon Kim, Aosong Feng, Haibo Ding, Jun Huan Published: 2026-09-28 Categories: cs.CV, cs.LG, cs.RO arXiv: https://arxiv.org/abs/2609.36352v1 Vision-language-action (VLA) models perform well on shorter-horizon manipulation tasks but still struggle with long-horizon tasks that require multiple dependent manipulations from a single command. Online reinforcement learning (RL) can improve these policies through environment interaction, yet many existing methods provide reward only after the complete task succeeds. However, such terminal supervision is sparse and does not distinguish early failures from rollouts that make substantial partial progress. We propose StructRL, an online RL framework that constructs structured intermediate supervision from verifiable subtask completions. StructRL decomposes each task into verifiable subtasks, grants intermediate rewards only after the prerequisite subtasks have been completed, and scales each reward according to completion pace. Across RoboCasa365 and LIBERO-Long with GR00T-N1.5 and pi 0.5, StructRL consistently outperforms evaluated online RL baselines. These results show that verifiable, structured intermediate rewards improve long-horizon VLA post-training. Code is available at https://github.com/amazon-science/StructRL. --- ### Representation by Design in Generation: Cross-View Class-Token Alignment in Diffusion Transformers Authors: Xiaoyu Wu, Yifei Wang, Chen Wei Published: 2026-09-28 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.36348v1 Generative and representation learning remain asymmetrically connected: semantic representations are used to improve diffusion generation, whereas the models' own representations are often treated as a by-product of synthesis. We ask whether diffusion models can instead be trained to learn substantially stronger semantic representations without sacrificing generation quality. SelfFlow takes a step in this direction by introducing self-supervised patch alignment into flow matching, but its main gains remain in faster convergence and improved generation. Inspired by DINO and iBOT, we extend this framework with cross-view class-token alignment to further strengthen semantic representations. Specifically, we form two independently noised, dual-timestep observations of each image and align each student class-token representation with the stop-gradient EMA-teacher target from the other observation. This objective is optimized jointly with the inherited flow-matching and local patch objectives. Notably, although the additional objective acts only on the class token, it strengthens both class-token and patch representations. Compared with a matched two-view baseline, ImageNet linear-probing accuracy improves by 9.4\% using the class token and 10.1\% using mean-pooled patch tokens, while frozen-backbone VOC2012 segmentation improves by 3.6 mIoU. These representation gains are achieved while maintaining comparable ImageNet generation FID. In text-to-image training, the same objective also improves generation FID, reducing it from 2.52 to 2.37 at matched checkpoints. Our results show that representation need not remain a by-product of generation or merely a tool for improving it: it can be directly optimized as a first-class capability of diffusion pretraining alongside generation. --- ### Massively Parallel Reinforcement Learning with a Chaotic Reconfigurable Clockless Chip Authors: Eric Oliveira-Gomes, Damien Rontani Published: 2026-09-28 Categories: cs.NE, nlin.CD, physics.app-ph arXiv: https://arxiv.org/abs/2609.36347v1 Hardware accelerators based on physical dynamical systems offer an attractive route toward energy-efficient reinforcement learning applications. However, their scalability is challenging because it requires many statistically independent entropy sources. Here, we introduce a quasi-analog decision-making architecture based on asynchronous Boolean networks (or lattices) implemented on a clockless reconfigurable chip. Each node in the network consists of a single logic element that acts as an autonomous entropy source. This architecture gives rise to distributed Boolean chaos, in which a spatially coupled network generates parallel streams of chaotic Boolean transitions with very low statistical dependence between nodes. We experimentally demonstrate parallel decision-making on a 1024-armed bandit problem, which is beyond the scale of previous hardware implementations, while significantly improving power-law scaling performance. Separately, we scale the proposed entropy source to 5120 parallel channels, yielding an aggregate sample generation rate of 2.14 TS/s. Our solution is implemented on a commercial reconfigurable CMOS chip and offers high integration density and ease of programmability. Our results pave the way for using distributed Boolean chaos as a valuable hardware substrate for large-scale reinforcement learning and for the development of fully integrated, high-throughput decision-making accelerators. --- ### DeepRewind: Predicting and Repairing Premature Commitments in Deep Research Agents Authors: Amirhossein Abaskohi, Amirhossein Dabiriaghdam, Lele Wang, Peter West, Giuseppe Carenini Published: 2026-09-28 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.36344v1 Deep-research agents conduct long-horizon investigations through iterative search, evidence evaluation, belief revision, and synthesis. However, they may commit to claims before sufficient evidence is available, causing later reasoning to reinforce an incorrect interpretation. We introduce DeepRewind, an additive control layer for reversible deep research that represents the agent's evolving epistemic state as a typed graph of sources, evidence, claims, hypotheses, assumptions, commitments, plans, and drafts. Before accepting an intermediate conclusion, a prompt-based world model predicts its impact and estimates reversibility based on hypothesis narrowing, information loss, recovery cost, and contradiction-trigger coverage. A binary controller blocks risky commitments, while a consistency monitor performs dependency-aware rollback when later evidence invalidates them. Across DRBench and LiveDRBench, DeepRewind improves insight recall by 3.6 percentage points and reduces premature commitments by 59.1% relative to Open Deep Research. --- ### ThuRunel: Dynamic Decoupling for Structured Advisory Dialogue Authors: Yuyan Chen Published: 2026-09-28 Categories: cs.AI, cs.HC, cs.IR arXiv: https://arxiv.org/abs/2609.36340v1 High-stakes advisory domains such as medical aesthetics, legal consultation, and educational planning exhibit a two-phase structure. The early phase requires empathetic elicitation and emotional support, and the late phase requires authoritative specialist judgment. Neither fully automated agents nor human junior consultants adequately address this structure at scale. We formalize the core design challenge as dynamic decoupling, asking how an AI advisory agent should decide what to ask, when to stop, what to resolve autonomously, and what to forward to the specialist. We present ThuRunel, an advisory agent combining a finite-state belief management framework, a chain-of-thought teacher synthesis protocol, and learned generation adapters. Against eleven baselines, ThuRunel achieves consistent improvements in elicitation completeness and specialist brief quality. ThuRunel is publicly deployed as a bilingual web application in which the same decoupling decisions operate from the client's side, grounded in a curated knowledge base that cites its sources in every answer. --- ### Adapting Linear-Time Architectures for Tabular In-Context Learning Authors: David Schnurr, Felix Sarnthein, Thomas Hofmann, Imanol Schlag Published: 2026-09-28 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2609.36337v1 Tabular foundation models achieve strong performance by conditioning on labelled examples in context, but softmax attention limits their use on large datasets. Existing linear-time alternatives, however, are mostly causal, and their potential for tabular in-context learning (ICL) remains underexplored. To address this, we (1) revisit causal training setups, (2) compare linear sequence mixers, and (3) investigate their ICL generalisation beyond the pretraining context length. First, we show that the best training setup for causal models resembles next-token prediction. Then, perhaps surprisingly, the most promising linear sequence mixer is causal: DeltaNet outperforms even non-causal linear attention. However, it degrades beyond $2$-$4\times$ the pretraining context length, and existing mitigation strategies such as bidirectionality defer the problem at best. A hidden-state oracle shows that this is not a capacity problem. Instead, our analysis points to an instability in the recurrent state, which drifts in deeper layers of causal models. Since DeltaNet's learned write rates overfit to the pretraining regime, we modulate them with a time-dependent decay schedule intervention to stabilise length generalisation. Finally, re-introducing non-causality by reading out from the final state allows us to closely match a controlled softmax attention baseline on OpenML-CC18 and TabArena. ---