# Research Papers — 2026-09 · bundle page 64 of 266 > Full-text markdown of papers #3151–3200 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/64 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/63.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/65.md) ### Maximum Likelihood Estimation for Entity Ranking under Iterative Synthetic Data Augmentation Authors: Yiqiao Jin, Mengxin Yu Published: 2026-09-26 Categories: stat.ME, math.ST, stat.ML arXiv: https://arxiv.org/abs/2609.32987v1 We study entity ranking and inference under the Bradley--Terry--Luce (BTL) model using pairwise comparisons collected over sparse comparison graphs. In practice, collecting high-quality human judgments can be expensive and time-consuming, motivating the use of synthetic data augmentation in model training. We analyze an iterative synthetic augmentation workflow in which synthetic comparisons generated from fitted models are successively added to the original dataset. In this process, the proportion of real data may vanish as the number of iterations grows. However, emerging literature has shown that recursive training on synthetic data can lead to model collapse, raising concerns about the statistical reliability of such augmentation procedures. To this end, we systematically analyze the resulting MLE in finite-sample, high-dimensional regimes. For the resulting iterative maximum likelihood estimator (MLE), we derive its optimal finite-sample $\ell_2$ and $\ell_{\infty}$ statistical rates and establish its asymptotic normality under natural identifiability conditions. We further characterize regimes in which model collapse is avoided despite the diminishing fraction of real data. We validate our theoretical findings through large-scale numerical experiments and an application to the Arena Human Preference 140k dataset. --- ### ReVision3D: Attribution-Guided Recursive Self-Improvement for 3D Medical Perception Authors: Ho Hin Lee, Yuyin Zhou, Yannan Yu, Shi Gu, Yifan Wu Published: 2026-09-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.32984v1 Recursive self-improvement (RSI) offers a promising path for overcoming the limited visual capability of current medical imaging agents. Yet applying RSI to volumetric imaging remains difficult: failures can arise from acquisition, perception, training recipe, or downstream inference, while self-generated feedback and logged trajectories provide little guidance on which component should change. We introduce ReVision3D, an RSI system that leverages 3D volumes with spatially grounded annotations to determine where visual evidence is lost and recursively improve the corresponding visual capability. A frozen language-model designer proposes revisions to acquisition, perception, training, or inference, while the verifier and system-level objective remain fixed. Our key insight is that an annotated volume forms an exact replay world for view rendering and spatial verification: unvisited views can be rendered on demand, and localized predictions can be checked directly against reference masks. This grounded feedback directs targeted revision, while only changes that improve beyond measured seed noise are retained. Each accepted change triggers renewed attribution, allowing the dominant bottleneck to shift across rounds. On abdominal CT, attribution identifies perception as the dominant remaining limitation. Revising that level enables ReVision3D to achieve 79% liver recall and 83% kidney recall at under 0.4 false positives per patient, outperforming the evaluated frozen multimodal foundation models, with the largest gains on small lesions. --- ### Feasible Flow Matching for Graph Reconstruction via Within-Sampling Primal-Dual Guidance Authors: Haoming Chen, Nicolas Zilberstein, Santiago Paternain, Santiago Segarra Published: 2026-09-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.32980v1 Graph reconstruction from partial observations often comes with structural side information, such as degree bounds, triangle counts, or an edge-density band. Prior-Informed Flow Matching (PIFM) reconstructs graphs by transporting a local prior toward the graph distribution, but it provides no mechanism to incorporate this side information. We put forth Constrained Primal-Dual PIFM (CPD-PIFM), which augments the sampler with Lagrange multipliers that evolve along each trajectory. The multipliers respond to constraint violations at a predicted endpoint and guide subsequent sampling steps without retraining. We prove that the sampler inherits PIFM's permutation equivariance and bound its expected terminal slack by a term that decays as the inverse square root of the number of steps, plus two approximation terms. On three link-prediction benchmarks and nine combinations of datasets and constraints, CPD-PIFM raises feasibility by 11-26 percentage points and remains competitive with fixed guidance without selecting a separate multiplier for each constraint. --- ### Adaptive Ensemble Selection for Noisy Labels on Tabular Data Authors: Faizaan Ali, Inwon Kang, Oshani Seneviratne Published: 2026-09-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.32976v1 Incorrect or corrupted labels in tabular datasets can significantly degrade supervised learning performance, particularly when mislabeling is subtle and not easily detectable from feature space alone. In the context of automated or AI-augmented data science workflows, robust detection of such label noise is critical for building reliable models. We propose a data-centric reasoning module for AI data science systems that automatically diagnoses dataset quality and selects appropriate cleaning strategies. Given a dataset, a meta-model predicts weights over a diverse set of detectors, including confidence-based, neighborhood-based, and distributional methods. Across benchmark datasets with controlled noise, our approach achieves performance comparable to a Confident Learning baseline on average, with dataset-dependent gains and losses, particularly in heterogeneous regimes. We further show that detector effectiveness is systematically linked to dataset properties. These results demonstrate the value of descriptor-driven, data-centric ensembling as a component of AI-assisted data-science pipelines for robust dataset assessment and model reliability. --- ### Distributed Hydrological Modeling in the Feature Space Authors: Mohamad Hakam Shams Eddin, Maria Luisa Taccari, Yikui Zhang, Shijie Jiang, Juergen Gall, Markus Reichstein Published: 2026-09-26 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.32971v1 Accurate forecasting of river discharge and floods is very challenging. River dynamics are affected by storage, meteorological forcing, and flow propagation at different spatial and temporal scales. Forecasting thus requires a framework that considers the upstream-to-downstream flow through river networks across grid cells and catchments. This modeling is known in hydrology as distributed modeling and routing. Existing deep learning approaches either ignore this topology, operate on lumped catchments, or route predicted physical quantities through a separate graph or physical routing model. We instead introduce feature-space routing: a topology-aware state-space operator embedded directly in the forecasting dynamics. At every forecast step, the operator gathers latent states from upstream grid cells and causally updates the downstream state according to the known river network. This preserves the physical connectivity of the river system while allowing the propagated state itself to be learned end-to-end and allows the model to predict river discharge considering both local dynamics and neighboring upstream contributions. To address uncertainty and provide probabilistic forecasts, we minimize the fair continuous ranked probability score (fCRPS) as a training objective. Our experiments on the European Flood Awareness System (EFAS) and observational data for river discharge forecasting demonstrate that encoding the physical structure of river networks explicitly in the feature space substantially improves the forecasting skill, particularly in an ungauged setting. Our approach achieves state-of-the-art results on both reanalysis and observational data and is able to forecast maps of river discharge at 1 arcminute and 6-hourly resolution up to 10 days lead time. --- ### Self-Confirming Superposition Traps in Reinforcement Learning Authors: Dai Shi, Andi Han, Feng Chen, Yiqun Duan, Junbin Gao, José Miguel Hernández-Lobato Published: 2026-09-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.32966v1 Reinforcement learning (RL) trains representations on data selected by the agent's policy, which then uses the resulting returns to guide its next choices. We show that this loop can sustain a lower-return policy even when representation fitting is globally optimal on those data. In a self-confirming superposition trap, every optimal code assigns overlapping directions to features that rarely occur together under the current policy. An alternative action brings them together, causing interference that lowers its return and reinforces avoidance, although refitting to that action would yield more return at the same capacity. We characterize the dimensions admitting a trap in a tied two-step model and show separately that equal feature frequencies, continued visitation, and independent controller learning need not prevent it. Because fitting weights errors by visitation, an avoided action can lose its return advantage at little cost to the objective. In a finite-action model, we bound this distortion and derive a replay condition: sufficient training weight on the best separately adapted action preserves its ranking despite residual error. Neural PPO experiments show how the feedback develops during learning: agents initialized toward different actions develop different interference patterns, opposite mean return rankings, and different final policies at the same capacity. We therefore test whether retaining access to neglected states can improve control. Keeping these states in training reduces measured interference and improves sequential return, with gains even when the encoder is frozen. Related interventions on state access, replay weights, and feature overlap improve control on MiniGrid and DMControl. For agents that learn through a world model, protected fitting improves DreamerV3--Crafter's cumulative training scores at unchanged capacity. --- ### Relic: From Multi-Agent Collaboration to Persistent Organizational Capability Authors: Hongyi Du, Tianyi Zhang, Weijia Zhang, Yi Yang, Haofei Yu, Kunlun Zhu, Tianxiang Dai, Shang Jiang, Zhelun Gao, Jiaxin Pei, Shang Zhu, Jiaxuan You Published: 2026-09-26 Categories: cs.AI, cs.SE arXiv: https://arxiv.org/abs/2609.32965v2 Multiple agents may often conflict in an organization: for example, one coding agent changes an interface in a repository, but another continues to develop on the old version where existing tests become stale. A conversation can resolve the episode, but when the participants change, what makes the lesson continue to govern the team? We introduce Relic, which turns recurring collaboration failures into organization-owned, executable protocols. Members reflect on visible work, propose rules, and govern their adoption. Adopted protocols bind triggers, responsibilities, required evidence, and execution consequences to the runtime, while remaining open to revision and retirement. In one traced case, repeated integration friction produces an interface-review rule that governs later pull requests and is revised as work continues. Across 360 controlled runs over ten software workloads and three models, Relic raises complete-contract delivery from 14.06% to 19.76% (+5.71 percentage points) over a matched structured team without the protocol lifecycle, improving all four verified production endpoints in every model stratum. Under fresh-member transfer, behavioral correctness is 25.4% with no inherited protocol, 34.6% with the same rules provided as readable text, and 41.2% with executable bindings, a +6.5-point advantage over text alone. On the full CooperBench benchmark, after excluding 183 broken benchmark pairs, Relic achieves 371/469 (79.1%), establishing the best reported result among peer-structured systems. On the 47-pair same-model subset, Relic also exceeds Solo (28/47 vs. 26/47), reversing the coordination loss exhibited by the official peer baseline. Together, these results show how collaboration experience can become persistent organizational state that remains useful beyond the members who created it. --- ### The Commit-Abstain Circuit: Why Language Models Hallucinate Instead of Abstaining Authors: Vy Nguyen, Ziqi Xu, Jeffrey Chan, Estrid He, Feng Xia, Renqiang Luo, Erik Cambria, Xiuzhen Zhang Published: 2026-09-26 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.32964v2 Language models (LMs) often hallucinate by committing to confident answers rather than abstaining, even when they do not have enough information to answer reliably. A large body of existing work mitigates hallucination through detection or abstention mechanisms, but leaves open how models internally arrive at the decision to commit or abstain in the first place. We study this decision through mechanistic analysis, framing hallucination as unsupported commitment: the model commits despite exhibiting signals of unanswerability. Using causal gating, we identify a Commit-Abstain Circuit (CAC), a sparse, causally localised subset of attention heads and MLP sublayers underlying this decision. Across ten LMs (3B-14B) from five families and three benchmarks, the CAC exhibits a recurring accumulate-yet-undercorrect pattern: commitment-promoting components build up commitment in earlier layers, while abstention-promoting components act later as corrective signals that are often insufficient to overturn the accumulated commitment. Building on this finding, a lightweight policy trained on CAC activations improves decision accuracy by 12.2 points over the model's intrinsic commit-abstain margin, reduces false abstentions by 2.5 times, transfers to unseen benchmarks, and extends to larger models (27B-35B). The CAC is both diagnostic, clarifying how models overcommit, and practical, enabling improved abstention decisions. --- ### Clipped or Unclipped? Finite-Sample Trade-offs for Averaged SGD under Heavy-Tailed Noise Authors: Alexandra Suvorikova, Egor Gladin, Darina Dvinskikh, Artem Agafonov, Mohammad Alkousa, Yuriy Dorn, Vladislav Matyukhin, Alexander Gasnikov Published: 2026-09-26 Categories: cs.LG, math.OC arXiv: https://arxiv.org/abs/2609.32962v1 Gradient clipping is widely used to stabilize training, but it need not improve the statistical accuracy of averaged SGD, even under heavy-tailed noise. We derive a finite-sample comparison of clipped and unclipped Polyak-Ruppert averaged SGD under finite conditional $p$-th moments, $p\ge2$. Our main result gives explicit accuracy and confidence conditions under which, for $p>2$, the Gaussian term dominates the unclipped deviation bound, so clipping need not improve its leading order. By balancing clipping bias and concentration, we obtain a bound in which the heavy-tail correction depends logarithmically rather than polynomially on the inverse failure probability. At $p=2$, this improves the confidence dependence of the leading bound. We establish sharpness of the unclipped heavy-tail term through an exact one-dimensional quadratic recursion and extend the comparison to projected convex SGD. We also prove concrete costs of clipping: every fixed finite threshold increases asymptotic variance on a scalar Gaussian quadratic, while whole-gradient clipping can shift the limiting point under asymmetric noise. --- ### Beyond Token Savings: A Systematic Study of Context Compression in LLM Agents Authors: Ritul Satish, Prasoon Sinha, Akiho Kawada, Neeraja J. Yadwadkar Published: 2026-09-26 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.32961v1 As LLM agents tackle longer tasks, they increasingly compress growing histories of reasoning, actions, and tool outputs. Compression can reduce token use, but it also changes the information available for later decisions. Existing agentic harnesses bundle decisions about what to compress, when to compress, and how much to remove into fixed policies. A systematic characterization is needed to disentangle these decisions and reveal how each affects task success and execution cost. We systematically vary these decisions across three open-weight models on SWE-bench Verified and Terminal-Bench 1.0. Across nearly 35,000 agent runs, we measure task success, token use, end-to-end latency, and estimated cost. We find that fewer tokens need not mean faster or cheaper execution: on Terminal-Bench with Qwen, policies using roughly one-third as many tokens can take 20-80% longer than the uncompressed agent. Policies with similar overall success can solve different tasks, while the same policy can perform quite differently across models. Our results motivate evaluating compression by its effects on agent execution and tailoring policies to the task, model, and workload. --- ### Environmental Impact of Generative and Agentic AI: An in-Depth Analysis and Green Solutions Authors: Abderaouf Bahi, Amel Ourici, Ibtissem Gasmi Published: 2026-09-26 Categories: cs.CY, cs.AI arXiv: https://arxiv.org/abs/2609.32960v1 The proliferation of generative and agentic artificial intelligence (AI) systems has introduced computational demands whose environmental consequences are substantial yet underexamined. This paper examines the environmental footprint of modern AI systems across energy consumption, carbon emissions, water usage, and electronic waste over the full lifecycle of large language models, multimodal foundation models, and agentic workflows, from hardware fabrication and training through fine-tuning and inference to end-of-life disposal. This work provides a conceptual analysis, utilizing order-of-magnitude estimations based on published data, without conducting original physical measurements. We contribute a lifecycle taxonomy that crosses lifecycle phases with five impact dimensions and emission scopes; a Sustainability Assessment Framework for AI Systems (SAFIA) comprising nine indicators; a comparative analysis of traditional, generative, and agentic AI; seven open challenges; policy recommendations for regulators, cloud providers, hardware manufacturers, and AI developers; and a research roadmap to 2035. The analysis indicates that inference can rival or exceed training energy over a deployment lifetime, that agentic workflows can multiply the energy of equivalent single-pass inference by one to several orders of magnitude depending on the number of model and tool calls, that indirect water use and embodied carbon are systematically underreported, and that measurement tools, disclosure practices, and regulation have not kept pace with agentic deployment. These findings call for agent-aware energy measurement, lifecycle-based carbon and water accounting, and mandatory disclosure for large-scale AI training and deployment. --- ### DynamicDx: Evaluating Evidence Acquisition in Video-Based Diagnosis Authors: Jiahui Li, Yutong Guo, Nan Yang, Wenzhan Song, Jin Lu, Fei Dou Published: 2026-09-26 Categories: cs.CV, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.32957v1 Diagnosing a patient from video requires more than recognizing the sign: a vision-language model must turn what it sees into hypotheses, questions and tests. DynamicDx evaluates each step in 71 neurological consultations across 11 sign categories, linking authentic patient videos to confirmed diagnoses and fixed charts built from the same case reports, so that every model queries the same evidence. Across five such models, video improves accuracy by 9.9-22.5 percentage points over blind input, but neither recognition alone nor temporal order explains the gain: the cause is usually missing from the model's video-only differential diagnosis even when the sign is recognized, and shuffling the frames produces no reliable accuracy loss. Instead, a trajectory replay traces most of the gain to the investigation results the video prompts. Evidence acquisition is the bottleneck: supplying the decisive investigations raises accuracy to 73.2-93.0%. Two interventions act on it. A post-trained 4B video describer improves sign descriptions, especially from a short, densely sampled segment, and source-clean literature retrieval expands initial hypotheses; both bring the tests a model orders closer to those the treating clinicians documented and, through them, raise accuracy. For video-based diagnosis, seeing better helps when it leads to asking better. --- ### Efficient Message Passing for Partial Differential Equation Priors Authors: Anna Kazachkova, Leonhard Hennicke, Rainer Schlosser, Ralf Herbrich Published: 2026-09-26 Categories: cs.LG, cs.AI, stat.ML arXiv: https://arxiv.org/abs/2609.32956v1 Prior information for real-world physical quantities is most elegantly expressed via partial differential equations (PDEs). In this paper, we propose a novel way to solve PDEs using probabilistic inference on a factor graph. In general, factor graphs provide a natural way to encode prior knowledge into a model as explicit factors; here, this knowledge is provided by a governing PDE, which narrows the solution space, while observed data further shape the posterior over the parameters. The approximate parameter posterior is inferred using message passing based on moment matching, without posterior sampling or global gradient-based optimization. We demonstrate our approach on the first-order advection and the second-order semi-linear Fisher-KPP equations, where it achieves predictive accuracy comparable to a standard baseline while providing structured predictive uncertainty. Moreover, the inferred posterior marginal means and uncertainty structure match more closely those obtained using Hamiltonian Monte Carlo than the evaluated mean-field variational inference baseline, while requiring up to 10x less training time in our experiments, with inference speed comparable to variational inference. --- ### Constrained Flow Policy Updates: A Generalized Schrödinger Bridge View Authors: Boyang Li, Matthew Kim, Sylvia Herbert Published: 2026-09-26 Categories: cs.LG, cs.RO arXiv: https://arxiv.org/abs/2609.32952v1 Online safe reinforcement learning (RL) seeks policies that maximize reward while satisfying safety constraints. Reward and safety can induce multimodal action distributions, challenging the prevailing primal-dual methods: Gaussian actors may collapse onto a single suboptimal mode, and optimization over the nonconvex Lagrangian landscape can be unstable. Diffusion and flow policies can represent such distributions, but recent work with a diffusion actor relies on estimating and matching the score of an augmented-Lagrangian target policy. Instead, we differentiate the augmented objective directly through the generation path of a flow policy, so no score needs to be estimated. Because a flow policy lacks a readily available action log-density for entropy regularization, we build on the density-free kinetic-energy regularizer of FLAC, a recent reward-only method, and propose Reparameterized Augmented-Lagrangian Flow Actor with Least Energy (RAFALE), an off-policy actor-critic method for safe RL. We formulate its update as a constrained one-ended generalized Schrödinger bridge and show that, for each source draw, this path-space problem is exactly an entropy-regularized problem in action space. At positive noise, its solution reweights the reward-only action distribution only where the estimated cost exceeds a threshold set by the Lagrange multiplier. As the noise vanishes, the optimal value converges to that of a least-energy map objective that the flow policy optimizes directly. Across seven Safety-Gymnasium tasks, RAFALE achieves competitive reward with mean final cost within budget on every task, whereas strong baselines trade one for the other; ablations support the necessity of both its augmented objective and its flow actor. --- ### Optimal Nonparametric Dynamic Pricing with Censored Demand and Adversarial Inventory Authors: Mengxiao Zhang, Yingfei Wang, Haipeng Luo Published: 2026-09-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.32949v1 We study online dynamic pricing with censored demand, where an arbitrary inventory level is revealed before pricing and may adapt to past observations, while demand follows an unknown, price-dependent distribution that is stationary over time. For a horizon of $T$ rounds, Xu et al. (2026) achieved $\widetilde{\mathcal{O}}(\sqrt{T})$ regret under restrictive structural assumptions including linear demand, price-independent additive noise, and conditions relating inventory levels to the noise support. Our first contribution is to extend this framework to a substantially more general and statistically harder nonparametric setting, requiring only the natural assumption that expected sales are nonincreasing in price and allowing nonlinear demand curves and price-dependent noise. For this model, we first propose a simple baseline, Double-Grid-UCB, which discretizes both price and inventory and achieves $\widetilde{\mathcal{O}}(T^{3/4})$ expected regret using separate revenue estimates for each price-inventory grid pair. Then, we develop Threshold-UCB, which improves the expected regret to $\widetilde{\mathcal{O}}(T^{2/3})$. Unlike Double-Grid-UCB, Threshold-UCB reuses sales observations across inventory levels through shared estimates of demand-tail probabilities, allowing the same data to support revenue upper bounds for multiple inventories rather than a single inventory bin. We also complement this upper bound with an $Ω(T^{2/3})$ lower bound via a reduction from stochastic posted pricing, establishing its minimax optimality. Finally, extensive experiments across inventory processes, demand functions, and noise models demonstrate consistently superior performance of Threshold-UCB over benchmark algorithms. --- ### Synthetic Thermal Image Generation for Real-Time Animal Detection Under Low-Visibility Conditions Authors: James Momoh, Khandaker Mamun Ahmed Published: 2026-09-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.32944v1 Wildlife-vehicle collisions remain a significant road safety concern, particularly during nighttime and low-visibility conditions when RGB-based perception systems are often unreliable. Thermal imaging offers a promising alternative for detecting animals under poor illumination. However, the limited availability of annotated infrared animal datasets restricts the development of robust deep learning-based detection models. This paper investigates synthetic thermal image generation as a scalable approach for real-time animal detection under low-visibility conditions. A subset of 514 annotated visible-spectrum animal images from the NTLNP dataset is translated into synthetic thermal representations using CycleGAN-Turbo, while a limited real thermal dataset of 60 images is expanded through thermal-focused augmentation. Multiple object detection architectures, including YOLOv8, YOLOv9, YOLOv10, and RT-DETR, are trained independently on synthetic and real thermal datasets and evaluated using precision, recall, mAP@0.5, mAP@0.5:0.95, model size, and inference latency. Experimental results show that synthetic thermal images provide competitive detection performance, with RT-DETR achieving the highest synthetic-data mAP@0.5 of 0.9613. Models trained on augmented real thermal data achieve the strongest overall performance, with YOLOv10s obtaining 0.9879 mAP@0.5 and 0.9571 mAP@0.5:0.95. Computational analysis further indicates that lightweight YOLO variants provide favorable inference latency, supporting their potential for real-time deployment. These findings demonstrate that synthetic thermal imagery can reduce dependence on scarce infrared datasets and support the development of efficient animal detection systems for future vehicle-mounted wildlife collision mitigation applications. --- ### The Key Handoff: Retrieval in Hybrid Language Models Authors: Kaan Kale, Oguzhan Baser, Sriram Vishwanath Published: 2026-09-26 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.32942v1 A two-hop question makes a language model retrieve twice: once to produce a bridge entity, and once to retrieve with it. Transformers resolve that entity in their early layers. Hybrid models replace most of the attention with a recurrent state, so where the key becomes usable, and where it is spent on an answer, is not known. Answering "Where is the ball?" from "the ball belongs to Alice" and "Alice is in the garden" turns on Alice, a name the question does not mention. A model could reach garden through Alice, the key it computed, or through where the fact sits in the prompt. Across twelve models, dense and hybrid, we move a hidden state from one story into another where the two routes lead to different places, and read off which place the model gives. An attention layer converts the key in every model we tested, and crossing that layer removes its usable effect, all of it where no attention follows. In sequential hybrids this makes the answer a handoff: recurrent layers carry the key forward, and attention spends it. Retrieval does not always stop there. Writing a different fact into the memory of the recurrent layers that follow the last attention layer can move the answer toward that fact, multiplying its odds by 1.3 to 2.7, and which hybrids do this is not settled by their architecture or training. That read is addressed by the key, not by position: recurrent state in a hybrid is not only a carrier, but a memory that later layers can query. --- ### Generative Priors Conditioned on Natural Language for Bayesian Inversion in PDEs Authors: Pengyu Zhang, Mark Girolami, Arnaud Vadeboncoeur Published: 2026-09-26 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2609.32941v1 Inferring quantities of interest (QoI) from data is a central task in Science and Engineering. In such contexts, we often have access to both quantitative data and qualitative data. Quantitative data may be represented by noisy sensor measurements, simulation data, re-analysis data; qualitative data may be in the form of text descriptions of experimental setups, expected experiment outcomes, and human-perceived system behaviours. The task we address in this paper is the following. Given a training set of paired qualitative text and quantitative QoI data, we learn to exploit the inherent correlation between the two modalities to learn a highly informative data-driven natural-language-conditional Bayesian prior, such that when presented with a new physical system, we can coherently combine (i) the training dataset, (ii) qualitative text describing the new system, and (iii) a small number of noisy sensor readings from that new system, to perform inference and uncertainty quantification (UQ) over the QoI. To achieve this task, we develop two parallel approaches, one uses conditional diffusion and the other conditional autoencoders, and compare both against classical Bayesian methodology, unconditional generative models and deterministic supervised methods. Each approach has specific strengths and tradeoffs; conditional autoencoder offers theoretical tractability, allows for fast posterior sampling, and provides better-calibrated UQ, whereas conditional diffusion is explored for greater expressiveness and capturing complex posteriors with irregular QoI fields. The approach is tested on the steady-state heat equation, damped Helmholtz equation, and UK weather reanalysis data. --- ### TwinS-GCN: Spectral conjugate for Spectral Graph Convolutional Networks Authors: Chun Hei Michael Chan, Flavia Petruso, Dimitri Van De Ville Published: 2026-09-26 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.32940v1 Graph convolutional networks propagate information by repeated local aggregation through a graph shift operator; i.e., a $K$-layer network reaches $K$ hops neighborhood. On the one hand, such spreading can lead to oversmoothing. On the other hand, long-range dependencies demand the depth. Transporting information on long distances and without attenuation requires the shift to distinguish a direction of flow, which a symmetric operator cannot perform but a directed one can fulfill. A natural way to extract pure-directionality is to take the skew-symmetric part of the shift operator through the Cartesian split, which, however, generally does not commute with the shift itself, meaning that the filters built on it are not shift-invariant. We instead use the spectral conjugate; i.e., the image of the operator under $τ:z\mapsto \bar{z}$, which commutes with the shift and splits it into a dissipative and a non-dissipative part. Two filter families follow: a sum filter, whose non-dissipative component transports signal without energy loss, and a ratio filter, ratio in the pair of components rather than polynomial in the shift. Both arise from non-holomorphic kernels, placing them outside the holomorphic class underlying classical spectral convolution. On the directed cycle, the ratio filter becomes an IIR filter with global impulse response, for which we prove a long-range reach gap against every degree-$K$ polynomial filter. Chebyshev reparameterization gives stable vertex-domain layers with real coefficients, yielding TwinS-GCN, which solves graph transfer tasks at reduced depth and is competitive with state-of-the-art graph convolutional networks on node classification benchmarks. --- ### Theory of Scene: Breaking the Symmetry Trap in Multi-Agent LLM Coordination Authors: Liangqi Yuan, Wenzhi Fang, Shiqiang Wang, Christopher G. Brinton Published: 2026-09-26 Categories: cs.LG, cs.CL, cs.MA arXiv: https://arxiv.org/abs/2609.32939v1 Multi-agent systems built on large language models (LLMs) are largely homogeneous, as their agents behave alike even across distinct LLMs. We show that when such agents act concurrently without communication, they collide on targets they must split and diverge on targets they must take together, a double failure we term the symmetry trap. Theory of Mind (ToM), widely used for coordination without communication, cannot escape this trap, since homogeneous agents form the same prediction of one another and respond to it in the same way. We propose Theory of Scene (ToS), a training-free reasoning schema in which each agent reads its public role, the only difference between the agents, and the task context they all observe. Homogeneous agents thereby derive one division of labor, each taking the part its role fixes, which turns homogeneity from the cause of the trap into the cure. ToS reads the role together with the scene through role gating, which determines whether ownership overlaps or is already divided, and the task context through task coupling, which infers whether the team must converge on each target, divide it, or take its stages in turn. We evaluate on DivvyBench, a controlled environment we introduce, whose target types make an episode Competitive, Cooperative, or Mixed across Tabletop, Airspace, and Household scenarios, and on two established agentic benchmarks, GovSim and Overcooked. ToS outperforms all six baselines on every benchmark, and each baseline falls far behind it in at least one setting. Against ToM given the same inputs, ToS raises the DivvyBench success rate from 71.1% to 99.6%, the GovSim total gain from 207 to 400, and the Overcooked level-normalized throughput from 1.41 to 1.67. --- ### The Impact of Stochasticity on the Rashomon Effect in Machine Learning Authors: Andrea Apicella, Francesco Isgrò, Andrea Pollastro, Roberto Prevete Published: 2026-09-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.32934v1 Neural network training is inherently stochastic, with factors such as weight initialization leading to distinct models despite comparable predictive performance. This phenomenon is commonly associated with the Rashomon effect, which describes the existence of multiple near-optimal models for the same task. Although the Rashomon effect has received increasing attention, it remains unclear whether different sources of training stochasticity contribute similarly or differently to its manifestations. In this work, we present an empirical study of the Rashomon phenomenon along three complementary dimensions: solution-space multiplicity, predictive multiplicity, and decision-basis multiplicity. These dimensions are quantified through the size of the empirical Rashomon set, predictive ambiguity, and agreement between XAI attribution maps, respectively. By independently controlling three standard sources of stochasticity, namely weight initialization, mini-batch data ordering, and dropout, we isolate their respective contributions to each dimension of the Rashomon phenomenon. Experiments on tabular and image classification benchmarks reveal that these sources affect the three dimensions in different ways. In particular, larger empirical Rashomon sets do not necessarily correspond to greater predictive disagreement or lower explanation agreement, indicating that solution-space, predictive, and decision-basis multiplicity capture complementary rather than interchangeable aspects of the Rashomon effect. Overall, our results show that training stochasticity influences not only predictive performance but also the stability of predictions and explanations, highlighting the importance of identifying the specific sources of stochasticity responsible for different manifestations of the Rashomon phenomenon when assessing the reliability, reproducibility, and interpretability of neural network models. --- ### Last-Iterate Guarantees for Online Reinforcement Learning in Structured Constrained MDPs Authors: Nam Phuong Tran, Trinh Ha Mai Huynh, Tuyen Pham Le, Van-Truong Nguyen, Quan Nguyen, Long Tran-Thanh Published: 2026-09-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.32933v1 In safety-critical applications, deployment uses a single policy, whose performance and constraint satisfaction should hold directly rather than only for an average or mixture of training policies. This motivates last-iterate guarantees in constrained reinforcement learning. Recent progress has established such guarantees in exact-gradient or tabular online settings, yet scalable results for structured large-state problems remain open. We develop a general, statistically efficient framework for last-iterate convergence in structured Constrained MDPs (CMDPs). Our analysis separates contraction of the regularised primal-dual dynamics from actor approximation and statistical errors in policy evaluation under online exploration. This enables model-free on- and off-policy learning with structured function approximation: optimistic policy evaluation avoids explicit transition-model construction, while a compact parametric actor avoids maintaining mixtures or histories of past policies. We instantiate the framework for linear CMDPs and general function approximation, obtaining representation-dependent complexity and improved target-accuracy dependence over prior optimistic regularised primal-dual analyses. We further validate the stabilising effect predicted by our theory on a synthetic linear CMDP: the regularised method exhibits stable last-iterate behaviour, whereas its unregularised counterpart shows larger oscillations. --- ### Counting on Thinking: Tracing Evidence Integration in Language Models Authors: Jingming Xue, Robert C. Wilson, Huadong Xiong Published: 2026-09-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.32932v1 Finite computational resources force a tradeoff between automatic System 1 processes and costly System 2 thinking. Large language models (LLMs) can spend extra computation on hard problems, yet direct answers struggle even with counting, an elementary operation humans and animals perform automatically. We ask why this requires thinking in LLMs. Evidence integration has long been used in psychology and neuroscience to probe decision-making. Our evidence-integration task presents one letter per conversational turn and asks which of two target letters appeared more often. A running count difference solves the task optimally by weighting every letter equally; tokens at each turn could represent and update this difference. Direct responses instead weighted evidence unevenly, with strong recency effects, and assigned less probability to the correct answer as difficulty increased. Thinking improved performance and made integration weights nearly uniform, yet final-query attention remained concentrated on the sequence ends in both modes. Reasoning trajectories showed models revisiting input, recounting letters, and checking intermediate counts that informed the answer, suggesting that thinking constructs the accumulated count that direct responses lack rather than reading out one already formed. Reasoning-token costs grew with the number of letters far more than with coherence. Outcome feedback did not bring this computation into direct responses: under in-context reinforcement learning (ICRL), performance deteriorated over repeated games and recency effects strengthened, yet models grew more confident. Humans and animals amortize such computations into automatic processes, whereas current LLMs still pay for them with thinking on every trial. Which operations learning can make directly available remains central to how future models allocate computation. --- ### EEG-Based Motor Imagery BCI Algorithms and Technologies: A Review Authors: Mohammad Hossein Koohi Ghamsari, Seyede Fatemeh Ghamkhari, Siavash Bayat, Ahmed Hemani Published: 2026-09-26 Categories: eess.SP, cs.LG arXiv: https://arxiv.org/abs/2609.32930v1 Brain-computer interfaces (BCIs) have emerged as transformative technologies that enable direct communication between the brain and external devices. Among various BCI paradigms, EEG-based motor imagery (MI) has gained prominence due to its simplicity, non-invasiveness, and potential to restore motor function and facilitate rehabilitation for patients with motor impairments. This paper presents a comprehensive review of the most practical processing algorithms developed over the past decade for decoding brain sensorimotor cortex signals. Specifically, this paper discusses the integration of artificial intelligence (AI)-based algorithms, particularly machine learning and deep learning techniques, and their contributions to improving the performance and efficiency of MI-BCI systems in detail. Furthermore, the paper reviews state-of-the-art hardware platforms and emerging converging technologies, including system-on-chip (SoC) architectures, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), wearable devices, the Internet of Things (IoT), and augmented/virtual reality (AR/VR), and discusses their integration with advanced signal processing algorithms to enable next-generation MI-BCI systems. By highlighting current achievements of EEG-based MI-BCI technology and predicting future research directions that could further enhance real-time capabilities, this paper aims to provide valuable insights for researchers and practitioners, fostering innovation in high-performance EEG-based MI-BCI systems. --- ### Efficient Dynamic Algorithms for Graph Neural Networks with Non-Linear Propagation Authors: Kiarash Banihashem, MohammadTaghi Hajiaghayi, Mahdi JafariRaviz, Silvio Lattanzi, Danny Mittal Published: 2026-09-26 Categories: cs.LG, cs.DS arXiv: https://arxiv.org/abs/2609.32929v1 Graph Neural Networks (GNNs) are widely used for representation learning on graphs, but most methods assume static topologies, making them inefficient on evolving networks where edges change over time. Existing dynamic approaches either model graph evolution through temporal GNN architectures without focusing on efficient dynamic maintenance, or are restricted to linear propagation models based on Personalized PageRank. In this work, we study how to efficiently maintain node representations for non-linear GNN propagation under edge insertions and deletions. The propagation has no learned parameters, and only a classifier applied afterward is trained. For a broad class of standard activation functions, we develop a residual-based dynamic algorithm that selectively propagates local errors via push operations, maintaining an approximation to the evolving fixed point without full recomputation. We prove that our method achieves amortized $O(1/ε)$ update time per graph change under a degree-normalized error guarantee. Our approach uses a potential-based analysis in a degree-scaled norm and, in contrast to prior work on the linear case, requires no randomness assumptions on either the update sequence or the input vector. For the linear special case, we additionally provide an exact dynamic algorithm via low-rank matrix inverse updates. Experiments on benchmark datasets show that incorporating non-linearity improves accuracy while preserving efficient update performance, yielding a scalable and theoretically grounded method for maintaining this propagation on dynamic graphs. --- ### Phenomenon-Graph JEPA: Label-Efficient Representation Learning for Contactless Cardiorespiratory Sensing Authors: Constantino Álvarez Casado, Nhi Nguyen, Mohammad Rakibur Rahman, Le Nguyen, Manuel Lage Cañellas, Sasan Sharifipour, Miguel Bordallo López Published: 2026-09-26 Categories: cs.LG, cs.ET arXiv: https://arxiv.org/abs/2609.32928v1 Millimeter-wave (mmWave) radar and RGB-D cameras can record cardiac and respiratory waveforms continuously and without contact, but labeled recordings remain scarce because every label requires a supervised acquisition session. Self-supervised pretraining can exploit the unlabeled signals, yet contrastive methods depend on signal transformations and negative pairs whose validity is uncertain for cardiorespiratory data, where time warping changes breathing rate and distant windows can share the same physiological state. We present Phenomenon-Graph JEPA, a joint-embedding predictive architecture that learns from four processed one-dimensional streams without negative pairs or synthetic augmentation in its base configuration. Each stream is encoded by a temporal convolutional branch and a band-limited spectral branch. During pretraining, the model predicts stopped target embeddings along typed edges, which connect streams assigned to the same physiological phenomenon, and forward in time within a state episode. We treat this physiological typing as a testable hypothesis and compare it with wrong-edge and all-pairs prediction graphs. In the OMuSense-23 dataset, pretraining improves label-efficiency area over matched supervised training by 3.91 percentage points (95% interval 2.08 to 5.80, Holm-adjusted p = 0.006), and by 3.74 points under a second configuration evaluated on the same test participants. However, the wrong-edge and all-pairs controls do not establish a benefit from physiological typing. Optional Takens-inspired delay coordinates improve a validation comparison with learned history, whereas two wrist-only WESAD protocols do not establish a pretraining advantage. The study therefore separates the measured benefit of predictive representations from the physiological prior used to organize their training. --- ### The Geometry of Logic: Stratification Induces Semantic Structure and Robust Reasoning Authors: Cristina V. Lopes, Yuangang Li, Justin Tian Jin Chen, Alberto Krone-Martins, Iris Ma, Md Rakib Hossain Misu Published: 2026-09-26 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.32927v1 Transformer-based language models perform well on symbolic tasks, yet it remains unclear whether they learn generalizable rules or rely on statistical shortcuts. Mechanistic studies link algorithmic behavior to structured internal representations, motivating the hypothesis that robust reasoning benefits from separating values from the types that control their manipulation. Can making this separation an architectural primitive improve the learnability and generalization of logical mechanisms? We introduce \textbf{STRAT} (\textbf{ST}ratified \textbf{R}egisters \textbf{A}nd \textbf{T}ypes), which partitions the residual stream into orthogonal Data and Type subspaces and uses Type-based attention and gating to govern Data transformations. Controlled arithmetic ablations identify three failure modes associated with data-control interference: the Linear Trap, Gradient Wall, and Open Gate Trap. Mechanistic analysis reveals interpretable logical structure, and in arithmetic, STRAT reduces median OOD error 35-fold relative to a Transformer baseline. On each of 11 datasets spanning 10 tasks, STRAT outperforms the Transformer baseline in mean accuracy, by 26 percentage points on average, with both models trained from 10 base examples per dataset using identical task-specific augmentation where applicable. Under distribution shift, STRAT's mean accuracy drops by only 2.39 percentage points, compared with 11.75 for the Transformer. --- ### Identification of Nonlinear and Dependent Latent Factor Structure through Clique Search Authors: Dale S. Kim, Qing Zhou Published: 2026-09-26 Categories: stat.ME, stat.ML arXiv: https://arxiv.org/abs/2609.32926v1 Learning the structure of latent factor models involves two central challenges: (1) estimating the number of latent factors and (2) learning the support of the mapping from latent variables to observed variables. This is especially challenging for nonparametric regimes and nonlinear settings. We propose a method for latent structure learning, based on pairwise dependence measures on the observed variables using a graph-theoretic representation. We show that both the number of latent factors and nonlinear mapping structure can be identified from the distribution of observed variables under mild structural assumptions. Unlike prior work restricted to linear correlations, we establish identifiability and consistency for a general class of dependence measures under nonlinear factor models. This motivates a Dependence Thresholding (DT) algorithm, which jointly estimates the number of latent factors and nonlinear mapping structure from observational data alone. We pair this with a neural network architecture constrained by the nonlinear mapping structure, to recover the nonlinear function. Through simulation studies, we show that the DT algorithm is accurate in practice, even when using flexible methods such as neural networks, and exhibits robustness against violations of its assumptions in high-dimensional settings. --- ### Diagnosing Sampled LLM Reasoning in Formal Geometry: Coverage, Realization, and Validity Evidence Authors: Xiao Yue, Guangzhi Qu Published: 2026-09-26 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.32924v1 Repeated sampling can reveal a correct numerical answer without yielding either a reliable system output or a supported derivation. We present Coverage, Realization, and Validity Evidence (CRV), an evaluation protocol for sampled large language model (LLM) reasoning over formal geometry states. Coverage is answer availability, realization is readout accuracy on the frozen candidate pool, and validity evidence is a label-blinded critic judgment of derivational support rather than a proof certificate. CRV freezes each candidate pool before comparing readouts and analyzes covered failures by correct-answer multiplicity and within-problem discrimination. On HardShift441, a 441-problem set for which a reference solver leaves 406 problems unsolved, a LoRA-adapted Qwen2.5-7B generator obtains 24.2% average single-sample accuracy and 68.9% pass@16, whereas verifier-weighted self-consistency (WSC) reaches 38.0%. Readout accuracy is particularly low when the correct answer occurs only once or twice in the pool. In a separate constructed audit of 195 covered problems, the critic labels 12 correct-answer representatives as supported, 181 as refuted, and two as uncertain. These results show that coverage, realization, and validity evidence from the critic are distinct quantities and should be reported separately. --- ### TRACE: Learning to Self-Calibrate Wireless Digital Twins from ISAC Measurements Authors: Saad Masrur, Saeed R. Khosravirad, Ismail Guvenc Published: 2026-09-26 Categories: cs.AI, eess.SP arXiv: https://arxiv.org/abs/2609.32923v1 Wireless digital twins (DTs) rely on 3D environment models to predict radio propagation and support wireless-network decisions, yet these models are often initialized from imperfect 3D maps. Errors in building position, height, footprint, and orientation can therefore cause a high-fidelity propagation engine to simulate the wrong physical environment. In this paper, we study how a deployed wireless network can repair an existing DT using its own radio frequency (RF) measurements. In particular, we introduce Twin Residual Alignment and Calibration Engine (TRACE), a physics-grounded learning-based self-calibration framework that treats twin maintenance as residual alignment between the physical world and the current DT. Using the same sensing configuration as the physical measurements, TRACE ray-traces the current DT, coherently backprojects the measured and simulated RF onto a common world grid, and extracts the same local region around each building's current DT position. A multi-view corrector then fuses evidence across sensing nodes and neighboring buildings to predict a gated six-parameter correction per building, without relying on absolute layout or sensor ordering, and supports iterative correction through re-rendering. On 5,400 held-out samples from unseen simulated scenes at 28 GHz, TRACE reduces 3D position RMSE from 2.202 m to 0.302 m and yaw RMSE from 4.978° to 0.894°, outperforming ViT and U-Net baselines under changes in layout, building count, sensing-node count, and SNR. On measured 28 GHz RF data from the NIST outdoor courtyard, a model trained only on synthetic RF reduces mean planar wall-position error from 1.00 m to 7.8 cm, without measured-data fine-tuning or geometric labels. These results show that the discrepancy between measured and twin-rendered RF can serve as a learning signal for repairing a wireless DT. --- ### Precision As You Need: Stochastic Computing Is a Dense Adaptive Quantizer Authors: Haoran Jin, Kangqi Zhang, Jirong Yang, Barry Lyu, Qiuyi Ding, Ruijie Gao, Nathan Bleier Published: 2026-09-26 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.32922v1 Matrix multiplications dominate the inference cost of modern transformer-based vision models, yet existing efficiency techniques such as post-training quantization and mixed-precision inference are largely limited to the small set of fixed-width formats (INT4, INT8, BF16, and FP16) supported by conventional accelerators. We revisit stochastic computing (SC) as a way to lift this constraint: viewed as a dense adaptive quantizer, SC controls precision by bit-stream length L rather than a fixed datapath, while each multiplication reduces to a single AND/XNOR gate. We build a GPU library that emulates SC matrix multiplication at scale, exposes stream lengths as first-class kernel arguments, and evaluates SC end-to-end on image classification, object detection and instance segmentation, class-conditional image generation, and visual world-model planning. On top of this substrate, we develop a dynamic per-row mixed-precision policy that assigns stream length per token or group at matched average budget, requires no retraining, and uses the same SC hardware across schedules. Across tasks, SC remains competitive with fixed-format INT quantization at matched bit budgets, while per-row mixed precision helps maintain accuracy at lower average stream lengths. These results provide software-level feasibility evidence that SC can serve as a dense-precision substrate for fine-grained mixed-precision inference on modern vision transformers. --- ### Adaptive Latent Capacity for World Models Authors: Idan Achituve, Lior Dikstein, Idit Diamant, Arnon Netzer, Hai Victor Habi Published: 2026-09-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.32921v1 We introduce Adaptive LeWorldModel (ALeWM), a world model based on a joint-embedding predictive architecture (JEPA) that learns to concentrate predictive information in compact prefixes of a wide latent representation. To encourage this ordering, ALeWM learns a sequence-conditioned distribution over prefix lengths and trains the predictor to estimate the full next embedding from a sampled input prefix. As standard anti-collapse objectives encourage variation across latent coordinates and do not organize them by predictive importance, we also introduce MixSIGReg. MixSIGReg regularizes the masked embeddings against a prior-weighted mixture with Gaussian active prefixes and zeros in the remaining coordinates. As a result, the ALeWM objective encourages early coordinates to retain information useful for prediction and recursive planning. Our analysis shows that the mixture distribution used by MixSIGReg assigns higher variance to earlier coordinate blocks and lower variance to later ones. In addition, we show that, under specified assumptions, prediction error is minimized by placing the information most useful for prediction in earlier blocks. Empirically, we study the behavior of ALeWM in a controlled dynamical system with known state variables and in goal-conditioned visual control. We show that ALeWM consistently achieves higher mean success rates than tuned fixed-width LeWM, with lower planning capacity on average. --- ### Vision-Language Agents for Active Perception in Optics Laboratories Authors: Ryan Lopez, Sachin Vaidya, Seou Choi, Serena Landers, Marin Soljačić Published: 2026-09-26 Categories: physics.optics, cs.AI, cs.CV, cs.RO arXiv: https://arxiv.org/abs/2609.32918v1 Vision-language models (VLMs) are increasingly being used in scientific workflows, but their ability as agents to directly control laboratory experiments from visual feedback remains underexplored. This capability is important because many laboratory tasks do not naturally provide dense, pre-defined numerical objectives: informative signals can be sparse, intermittent, or visually ambiguous. A more general laboratory agent should instead be able to interpret visual observations, take actions to acquire useful feedback, and adapt its behavior based on the consequences of those actions. We study whether general-purpose VLMs can perform this kind of closed-loop scientific control using experimental optics as a testbed. We evaluate agents on three experimental systems that isolate distinct capabilities: a Michelson interferometer, a two-mirror cavity, and a four-mirror optical relay. The agents observe camera images, directly issue actuator and measurement commands, and retain their interaction history without receiving an engineered scalar objective during control. Across these experiments and matched simulations, we find that, given task-specific natural-language guidance, VLMs can estimate actuator-response relationships, resolve ambiguous observations through intervention, and actively create informative visual feedback when signals are sparse. These results suggest that pretrained multimodal models can serve as important decision-making agents within the experimental loop. Our work also establishes optics as a physically grounded testbed for visual reasoning and active perception in scientific agents. --- ### Planner-as-Router: Joint Plan-Time Model Routing for Cost-Efficient Multi-Agent Workflows Authors: Vivek Kumar Singh, Preeti Priyam, Gautam Bhowmick Published: 2026-09-26 Categories: cs.AI, cs.CL, cs.MA arXiv: https://arxiv.org/abs/2609.32917v1 Running large language model (LLM) agents in production gets expensive fast. A frontier model (the largest, most capable tier) is accurate but can cost 25 times what a small model costs per token, and the gap compounds once a workflow chains several calls together. Planner-as-Router (PaR) attacks this from a different angle. Instead of leaving model-tier selection to some component downstream, it folds the choice into planning itself. As the planner breaks a query into subtasks, it also assigns each one a model size tier (small, mid, or frontier, ordered by capability and price), so the dependencies between subtasks are visible before any specialist runs. Unlike per-call routers such as cascade routing, which look at one node at a time, PaR sees the whole workflow up front and needs no separate router model or training data. We evaluate PaR with EntBench, a benchmark of 54 enterprise agentic tasks across seven classes, graded by actually running the generated Structured Query Language (SQL) and MongoDB queries against live databases. Over 1,157 evaluations spanning eight routers and three seeds, PaR stays on the observed cost-accuracy frontier. It matches a sink-frontier heuristic (frontier model on terminal nodes only) in accuracy at comparable cost and a faithful FrugalGPT cascade at lower cost, and cuts cost 44% against all-frontier routing while giving up 2.9 points of accuracy. Several accuracy gaps fall inside the plus-or-minus six-point confidence interval of a 54-task study, so we frame PaR's advantage as frontier position rather than a clean accuracy win. We also report a preliminary observation, not a validated result: a small pilot hints that cheap routing may carry a hidden compounding penalty on compositional workflows, which we frame as a hypothesis for future measurement. PaR, EntBench, and all evaluation code are open source. --- ### AgentTell: Behavioural Side-Channel Leakage in Browser-Use Agents Authors: Asif Shahriar, Md Nafiu Rahman, Sadif Ahmed, Farig Sadeque, Md Rizwan Parvez Published: 2026-09-26 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2609.32915v1 Browser-use agents often carry information in their context as they move between websites. While it may be necessary for task completion, it also creates a privacy risk, especially when the information contains a private fact regarding the user. For example, an agent may learn a user's affiliation after reading a membership record. If it later selects a registration option specific to that affiliation on another website instead of a general option, the information gets leaked. In this work, we define and study behavioural side-channel leakage in browser-use agents, where an agent's actions inadvertently reveal private information (secret) retained from a prior website, despite an explicit instruction not to disclose it. We introduce AgentTell, a benchmark of 20 scenarios and 100 tasks in which an agent acquires a secret on one website and then completes a task on another website that offers secret-specific actions alongside a general action that reveals nothing. Our evaluation across 9,760 sessions on six backbones shows that agents carrying a secret reveal it through their actions in 61.1% of sessions. Even when agents explicitly state in memory that the secret must not be shared, they still reveal it in 56.7% of those sessions. Moreover, in 34.5% of leaking sessions, their final responses falsely assure users that the secret was not disclosed. These findings show that agents often fail to recognize side-channel leakage as a privacy risk. --- ### Allspark: Weak to Strong Transfer via Alternating Chain of Thought Authors: Kaizhao Liang, Junxiong Wang, Chen Liang, Zhendong Wang, Qiang Liu Published: 2026-09-26 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.32913v1 Recent progress in frontier models has renewed interest in large-scale reinforcement learning (RL), but the cost of generating large-model rollouts makes even testing RL recipes expensive. We ask whether reasoning improvements learned by a small, weak model can benefit a larger, stronger model without using the strong model's rollouts during training. We introduce Allspark, a training and inference framework for weak-to-strong transfer through alternating chains of thought. A weak teacher is trained alongside a frozen copy of the same model; the two alternate reasoning segments, and the frozen model produces the final answer. At inference time, a stronger student replaces the frozen training partner, while both models remain fixed. Because they communicate through text, the teacher can steer students from different model families and with different tokenizers. We study Allspark at two scales: controlled Qwen experiments across math and reasoning, and larger-scale Inkling experiments on ARC-AGI-2. The Inkling experiments show accuracy gains in within-family and cross-family settings, including transfer to Kimi and Nemotron, with benefits that vary across inference settings. These findings motivate reusing a trained weak teacher across strong students and examining the resulting accuracy--token tradeoff. --- ### Progression- vs Automata-based Anticipatory Monitoring of LTL over Finite Traces (Extended Version) Authors: Sarah Winkler, Toryn Klassen, Sheila McIlraith, Marco Montali Published: 2026-09-26 Categories: cs.LO, cs.AI arXiv: https://arxiv.org/abs/2609.32912v1 When safety-critical systems are developed from a known internal specification, their correctness can be established by model checking. In the frequent case where such a specification is unknown or inaccessible, runtime verification presents an attractive alternative, e.g., to ascertain that autonomous and agentic systems as well as business processes satisfy desirable properties and/or comply with safety requirements. In this paper we study anticipatory monitoring, an advanced form of runtime verification, where the monitoring state is determined by both the trace prefix seen so far, and all its possible finite-length, future continuations. We focus on monitoring linear-time properties that may involve arithmetic constraints. Automata-based approaches, the de-facto standard in this setting, are notorious for their computational complexity. We propose an alternative approach based on progression and LTLf satisfiability checking, for both propositional and arithmetic settings. We experimentally compare the automata- and progression-based approaches, and a third method that combines the two. Our experiments suggest that the progression-based approach often succeeds in producing a verdict when the automata constructions do not terminate, especially for the arithmetic setting. For the propositional setting, the combined technique provides a good tradeoff. --- ### Predicting the Next State Is Not Enough: JEPA Representations for Lean Theorem Proving Authors: Aarnav Choudhary Published: 2026-09-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.32908v1 Neural theorem provers must both propose tactics and decide which valid successor states to explore. We study whether one-step Lean transitions provide a self-supervised signal for branch ordering. A JEPA-style model predicts latent successor representations and scores only kernel-validated, nonterminal successors generated by a fixed pretrained ByT5 proposer. JEPA achieves higher Top-1 than matched InfoNCE on a same-theorem ranking diagnostic(50.18% versus 31.55%), but averages 282.3 of 987 solved theorems across three seeds versus 308 for proposer ordering, while requiring more tactic checks. In this setting, accurate one-step transition ranking is therefore insufficient as a long-horizon search value. The controlled evaluation separates representation from proposal quality and treats kernel-checked proof completion as the primary endpoint. --- ### Logical subspace in LLMs Authors: Hope Kean, Enric Boix-Adsera Published: 2026-09-26 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.32907v1 Recent work has identified a human brain network specialized for abstract formal reasoning (Kean et al., 2025). Does the same hold true in language models? To answer this question, we introduce the minimal viable subspace (MVS) method, which searches for the lowest-rank activation subspace at a layer that preserves task performance when everything outside that subspace is ablated. Using MVS, we demonstrate low-rank subspaces supporting logical inference on Gemma and Qwen models. Furthermore, these subspaces exhibit a clear dissociation from model capacities on other tasks, such that retaining these late logic subspaces preserves inference while impairing factual knowledge, working memory, cognitive control, and arithmetic. Conversely, ablating them reduces logical inference accuracy to chance while largely sparing these other capacities. Our results suggest a functionally localizable core machinery for logic akin to that in the human brain. --- ### Linger and Lose: Knowledge Collapse in Low-Bit Language Models Authors: Prashanna Mani Paudel, Shivanand Venkanna Sheshappanavar Published: 2026-09-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.32902v1 Training language models with ternary weights is commonly judged by loss and downstream accuracy, which record only a modest cost relative to full precision. We show that these metrics can conceal a much larger failure. We instead measure knowledge capacity, the factual bits stored per parameter, on synthetic biographies with known information content. We train GPT-2-style models from scratch with 2.5M to 50M parameters at five precisions. Under the standard cosine schedule, ternary models retain as little as 6% of an identically trained fp16 model's capacity. The deficit widens with model size while perplexity rises by only 1.4 to 1.6 times. Measured throughout training, these models first acquire capacity and then lose most of it. We identify this knowledge collapse as a learning-rate dwell instability. Held near $1$--$2\times10^{-4}$ with no decay, a pre-collapse model collapses within a few hundred exposures, and returning to a safe rate does not restore capacity. We then locate the collapse in the output head. It happens at a value the model can never predict. The weights there grow unchecked, while every other prediction the model makes is unchanged. We find that making the value predictable removes the collapse, regardless of which attribute carries it. A warmup-stable-decay schedule with a 10% cooldown increases ternary capacity by 2.7 times at 25M and 4.3 times at 50M. Gains are larger at lower precision. Cutting the output head's learning rate prevents failure when training from scratch. The post-training quantization methods we tested recover no measurable capacity below 4 bits. Our findings suggest judging low-precision training by retained capacity during training rather than final loss. --- ### Constraints Are Graphs, Not Chains: Exact Decoding for Diffusion Language Models Authors: Jianchang Su, Wei Zhang Published: 2026-09-26 Categories: cs.AI, cs.FL arXiv: https://arxiv.org/abs/2609.32900v1 Diffusion language models (dLLMs) predict masked positions in arbitrary order, but their exact constrained decoders still encode constraints as sequential languages, whose state must track every unresolved dependency between positions. For relational constraints this encoding grows exponentially: for same-order copy, every finite automaton needs $4^k$ states, deterministic or nondeterministic, and every context-free grammar has size $2^{Ω(k)}$, while the factor graph of the same relation has size $O(k)$ and a 16-entry peak table. We introduce FactorDLM, a training-free decoder that represents finite-domain relations as a factor graph and, at each denoising step, conditions the model's mean-field prediction on that graph exactly by variable elimination. Decoding cost then grows exponentially with the induced width of the constraint graph, which replaces automaton size as the governing parameter. Because a finite automaton is a chain-shaped factor graph, one compiler enforces syntax and nonlocal relations together: on JSON records with cross-field references, a schema automaton alone leaves references dangling, relational factors alone produce malformed JSON, and the combined plan is valid on both counts, including on records of variable length. Across nine relational benchmarks and three backbones, every output satisfies every declared constraint at 0.4-6.9% projection overhead, where unconstrained decoding is 0-79% valid, and compiled projection answers repeated queries 13.6x faster than CP-SAT with eight parallel workers. Because model-free rules solve three of five standard benchmarks, we construct benchmarks with exact chance and fixed-template floors, on which selecting among exact constrained samples beats greedy projection. Which encoding is cheaper, sequential state or direct factors, depends on the constraint and is computable before decoding begins. --- ### Rethinking the Fully Hyperbolic Vision Transformer in Polar Coordinates Authors: Ahmad Bdeir, Niels Landwehr Published: 2026-09-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.32899v1 Hyperbolic space can embed intrinsic hierarchies in data with low distortion due to the exponential growth of volume with distance from the origin. However, current Lorentz transformer blocks are formulated in ambient coordinates, where numerical errors increase at large radii due to instability in the Lorentzian inner product, leading many models to limit the radius to avoid this issue. This prevents us from utilizing the regions of hyperbolic space that motivate the geometry. To address this, we revisit the components of the transformer block in polar coordinates, where hyperbolic operations such as distance calculation and attention centroids can be computed without the numerical cancellation in their ambient-coordinate formulations. Specifically, we propose a polar fully connected layer that separately maps an embedding's direction and radius, allowing the radius of a feature to be learned rather than determined by the norm of a linear map. We additionally introduce horospherical shifts as relative positional encodings whose query-key distances grow logarithmically with the token gap. Finally, we reformulate the residual connection as average radius Lorentz boosts. Combining these components, we develop a fully hyperbolic transformer that substantially improves performance over Euclidean and hyperbolic baselines on standard vision tasks. We further evaluate our model on ImageNet and demonstrate its ability to generalize to other datasets using pre-trained weights, similar to Euclidean counterparts. --- ### When Less Compute Is More: Adaptive Early Exit Improves Pretrained Outlier Detection Authors: Tianyang Zhou, Leman Akoglu Published: 2026-09-26 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.32898v1 Pretrained tabular foundation models process every dataset at a fixed depth, with inference costs growing with dataset size. To address this, we present the first study of depth-adaptive early-exit for pretrained outlier detection models. While early-exit is typically motivated by efficiency, we uncover a surprising benefit: exiting at the optimal intermediate layer can also improve detection performance on diverse real-world benchmarks by 4.7-7.3% on average, consistent across three distinct foundation models. First, we investigate the factors driving these gains, and identify a key mechanism: context pollution, i.e., the presence of outliers among in-context samples. Our analysis reveals that nearby in-context samples exert increasing influence on query predictions at greater depths, consistent with a retrieval-based view of these models. In effect, early-exit alleviates the adverse effects of retrieving accurate-yet-polluted neighbors, with gains of 13-21% when context pollution matches the natural outlier rate. Motivated by these findings, we pretrain a plug-in router to select a dataset-specific exit layer, using query outlier labels as privileged information available only during router training. The router operates post hoc, leaving the base model parameters and prediction head unchanged. Experiments on three large real-world benchmarks show that, on clean context, the router recovers up to 45% of the oracle gain with up to 1.8x speedup across three pretrained backbones, with larger gains as context pollution increases. --- ### Optimizing H-Graph Hybridization for Diffusion-Guided RRT Authors: Omer Talmi Published: 2026-09-26 Categories: cs.LG, cs.AI, cs.RO arXiv: https://arxiv.org/abs/2609.32897v1 Sampling-based motion planners guided by diffusion models produce high-quality trajectories in a single run, yet the stochastic diversity available at inference time is left largely unexploited. We present two inference-time diversification strategies for a fixed, pretrained DiTree model, combined via H-Graph hybridization, and evaluate them on a holonomic AntMaze robot across 15 maze scenarios. The first, factorial diversity, sweeps the random seed and Diffusion Goal Bias (DGB) parameter, the second, refinement-only diversity, sweeps the diffusion refinement strength (RS) that controls how much an RRT-generated trajectory is edited. Because a single-run baseline only partially succeeds, we additionally compare H-Graph results with pool-based statistics. H-Graph improves the mean pool length of the factorial and refinement-only diversities by 18.8% and 14.5%, respectively. In addition, it also improves the best individual candidate's lengths by 9.7% and 6.8%, respectively. And last, compared with the successful baseline's trajectory length, it improves the results by 18.2% and 19.9%, respectively. These results show that inference-time parameter variation is a reliable, training-free source of path diversity, and that H-Graph hybridization reliably converts this diversity into shorter, higher quality trajectories. --- ### A Function-Level Vulnerability Score Measures Flag Rate More Than the Model: Protocol Effects on Paired Benchmarks Authors: Maciej Cichoń, Bartłomiej Dmitruk Published: 2026-09-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.32890v1 Language models are increasingly evaluated as vulnerability detectors, and scores reported for similar models differ widely between papers. We measured how much of that difference evaluation protocol accounts for, with model outputs held fixed. In a paired test, a model must flag a vulnerable function and clear its version after a fixing commit. Three choices that published evaluations make differently were varied one at a time: metric, verdict extraction and output budget. Seven frontier and large open models were evaluated on five released pair benchmarks and a set pooled for this work under one protocol, and 61 open models of 1.5B to 36B parameters on the pooled set. Function-level F1 follows how often a model flags both functions of a pair (Spearman $+0.86$ over 42 combinations) and is nearly unrelated to pair-level correctness ($+0.16$). On the pair score, extraction changes a model's number by $+0.001$ at the median and budget by $+0.02$ with an interval through zero, whereas the model changes a benchmark's number by up to 0.18 and the benchmark a model's by up to 0.16; a function-level score therefore measures flag rate more than model. For 37 of 68 models the difference between correct and reversed pairs is within its 95% interval of zero, the value for a null model that flags each function at a fixed rate, while both-flagged and both-cleared rates exceed that null by 0.055 on median, and for 64 of 68 both functions of a pair receive one answer more often than independence predicts: verdicts are determined by the text common to both functions. On length-matched pairs a linear probe on activations separates 0.78 by within-pair ranking, against 0.64 for a tf-idf baseline and 0.5 for length; the generated verdict is near chance for three of six models and at 0.55 to 0.57 for the other three, and a prompted logit is at chance for all six. --- ### Measurement-Gated Provenance Attenuation for Frozen EEG Representations Authors: Anuar Aimoldin, Yankai Chen, Ayana Mussabayeva, Nurdaulet Akhanov, Xue Liu Published: 2026-09-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.32889v1 Frozen EEG representations retain acquisition signatures as well as neural activity. Source predictability alone does not identify what should be removed: it can reflect measurement effects or genuine biological and population differences, which should not be erased. We propose Measurement-Gated Provenance Attenuation (MGPA), built on one principle: measurement evidence determines where correction may act, and preserved information determines what it should aim for. Paired measurement contrasts define a gate outside which nothing changes; inside it, the source score is moved to the value the preserved coordinates already predict: for a fixed affine score, this keeps the same information as any target set by those coordinates and needs the least expected squared movement. Closed-form and critic-guided iterative constructions apply it without source identity or encoder retraining. Three studies test the principle at increasing distance from its assumptions. Under controlled reference changes, where the source-task association is known, MGPA brings source to near chance with task performance unchanged, whereas erasing what predicts source (LEACE) lowers frozen-task AUROC from .753 to .656 while barely touching source; ablations attribute the attenuation to the gate's directions and 2.7x less movement to the conditional target. Across recordings from different devices and electrodes, iterative correction lowers source accessibility while preserving or improving task performance. Finally, one iterative map selected on one task and reused unchanged on existing heads for two others raises their worst-association AUROC (lowest over device-label shifts) by .057 and .019 over LEACE, at a cost to those heads while the training association holds. A reusable correction shows its value in how an existing predictor behaves once acquisition cues stop being reliable, not only in what a probe can read. --- ### Refreshing Less, Selecting Better: Reusing Stale Gradient Features for Efficient Influence-Based Data Selection Authors: Jianchang Su, Yifan Zhang, Wei Zhang Published: 2026-09-26 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.32888v1 Gradient-based data selection methods such as LESS score each candidate by the alignment between its gradient and a target validation gradient, and recomputing per-example gradient features at every new checkpoint dominates their cost. Across three selection seeds, two model families, two candidate pools, and two target tasks, features cached at a post-warmup checkpoint and paired with fresh validation gradients preserve the ranking 40 optimizer steps later with Spearman correlation from 0.952 to 0.991, while the top-10% subset they induce misses 10 to 22% of the examples that full recomputation selects. We therefore propose Cached Diverse Influence Selection (CDIS), which recomputes gradient features for the top-ranked fraction $p$ of candidates under the stale scores, fits an affine calibration on the recomputed examples, and selects the final subset under source and length quotas. A refresh fraction at or above the selection fraction recovers the exact top-$k$ subset whenever the calibrated stale scores have bounded error, and the budget curves follow this rule: at $p=0.3$ the recovered top-$k$ subset coincides with full recomputation in every setting, allocating the same budget per stratum recovers the stratified subset at 0.92 to 1.00, and gradient-stage wall-clock drops 3.5 to 3.6 times. Iterating the cache over four checkpoints keeps top-$k$ overlap at 0.98 or higher at 1.9 gradient features per example against 4 for full recomputation, and stale-to-recomputed agreement on the refreshed examples provides a free check for unsafe reuse. Downstream, unconstrained top-$k$ selection collapses to a single data source and scores 13 points below random selection on GSM8K. CDIS scores 12 points above random selection with paired confidence intervals that exclude zero and trails full recomputation by 4.3 points, one training-run standard deviation, at 3.4 times lower selection cost. --- ### Neural Network-Assisted Refinement of Traditional Schemes for One-Dimensional Scalar Conservation Laws Authors: Imre Fekete, Ferenc Izsák, Vendel P. Kupás Published: 2026-09-26 Categories: math.NA, cs.LG arXiv: https://arxiv.org/abs/2609.32887v1 A clear link is established between conventional numerical methods and neural network approximations for solving one-dimensional scalar conservation laws. The focus is on the construction of an appropriate flux term in the case of convex flux functions for improving the classical schemes. The first neural network developed here is able to rediscover Godunov's method, while the second one emulates the behavior of a second-order slope-limiter function. In this way, by merging them, second-order reconstruction-based schemes can be developed. The networks presented here employ a minimal number of parameters, significantly reducing the complexity compared to previous approaches. These networks can also be linked consecutively to get a deep one corresponding to multiple time steps. Training them with an appropriate loss leads to stable schemes, improving even the classical methods without increasing their complexity. --- ### StraTune: Adaptive Selection of Revision Operators for Self-Evolving LLM Skills Authors: Zeping Liu, Yan Li, Ni Lao, Gil Wolff, Gengchen Mai Published: 2026-09-26 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.32886v1 Large language models (LLMs) can learn reusable textual skills from execution feedback without updating their parameters, but effectively deciding how to revise these skills remains a key challenge. Existing methods typically rely on a fixed revision operator, a search strategy and the revision forms applied under it. However, we observe that no single revision operator consistently performs best across tasks, and repeatedly applying an unsuitable operator can limit further improvement. We propose StraTune (strategy-guided skill tuning), which lets a frozen optimizer LLM choose the revision operator at every round from the optimization state, which is defined as the current execution feedback together with the recorded outcomes of earlier strategies and forms. Candidate skills from every revision operator pass one candidate evaluation, which screens for gains and regressions on a small sample set and validates them on a larger one, and every outcome is written back to the optimization state for later choices. Across four benchmarks and two LLM settings, StraTune outperforms all five baselines in most settings. Ablations attribute the gains to the adaptive choice of the revision operator, since fixed, random, scheduled, and bandit strategy choices all score lower, and skills learned with a small target LLM also improve a stronger one. Code and learned skills are available at https://github.com/seai-lab/StraTune. --- ### Can LLMs Predict the Future? A Brier Score Analysis of Prediction Markets Authors: Yuanbo Li, Zekun Li, Xiaoyan cong Published: 2026-09-26 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.32885v1 We study whether model upgrades improve probability estimates for prediction-market questions. Our Resolved Market Forecasting (RMF) benchmark contains 3,000 resolved binary questions across nine domains, on which we evaluate six Claude and Qwen model variants using a question-only, zero-shot protocol. We assess Brier scores relative to an empirical base-rate predictor, examine their Murphy decomposition, and compare models through paired differences, with results stratified by event category and timing relative to training cutoffs. In the reported post-cutoff stratum, the four Claude models achieve Brier scores of 0.183-0.192, improving on the base-rate reference by 0.024-0.033. Qwen 32B does not significantly outperform that reference, although its paired Brier is 0.024 lower than that of the 7B checkpoint. The evaluated Claude version and tier upgrades yield no significant improvement. Within-model differences across event categories exceed the observed differences among Claude variants. These results show why forecasting scores should be interpreted alongside simple probability baselines and question composition: under this protocol, newer versions or higher model tiers do not consistently produce more accurate probabilities. ---