# Research Papers — 2026-09 · bundle page 227 of 266 > Full-text markdown of papers #11301–11350 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/227 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/226.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/228.md) ### VERPO: Verified Evidence Regularized Policy Optimization Authors: Haijiang Li, Chengyu Lv, Yi Zhang, Rui Qian, Zhibing Zhang, Xiangqing Shen, Junjie Yang, Yuchen Zhang, Wenyuan Jiang, Hanqing Hu, Cangqi Zhou Published: 2026-09-05 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.06100v4 Verifiable rewards improve language models through reliable task-level feedback, but methods based on Group Relative Policy Optimization (GRPO) apply a sequence-level advantage uniformly across all tokens. This coarse credit assignment reinforces or penalizes entire responses without identifying which local decisions to preserve, reinforce, or revise. Conversely, evidence-conditioned self-distillation provides denser token-level supervision, yet teacher imitation can transfer stylistic artifacts and miscalibrated confidence that destabilize training when misaligned with task success. We introduce VERPO, which converts evidence-conditioned guidance into reward-aligned token-level credit assignment while retaining the outcome objective. VERPO decomposes teacher guidance into an evidence-free reference term and signed, evidence-induced corrections at each token. A stopped controller combines selective acceptance, token-wise localization, and cost-aware scaling by balancing alignment with the local GRPO update direction against Fisher movement cost. Furthermore, we introduce Fisher Evidence Contrast (FEC), which attenuates nuisance shifts along an estimated evidence-presence direction through a regularized projection. Across five scientific reasoning and tool-use tasks, VERPO prevents optimization collapse and consistently achieves the highest multi-task average across model backbones, yielding marked improvements particularly on smaller models over strong baselines. Qualitative diagnostics confirm that token acceptance selectively targets reasoning bottlenecks consistent with local reward alignment and Fisher movement cost. --- ### PASTEL: Panoramic Alignment for Monocular 4D Scene Reconstruction Authors: Yuankun Yang, Yi Wei, Bo Bai, Wenyang Zhou, Li Zhang Published: 2026-09-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06099v1 Reconstructing 4D scenes from casually captured monocular video is vital for applications in virtual reality (VR) and embodied AI. Recent advances in 4D reconstruction and novel view synthesis have substantially propelled this capability. However, existing reconstruction methods generally cannot recover regions beyond visible camera limits. Consequently, we introduce a new paradigm that achieves 4D scene synthesis by combining visible-region reconstruction from monocular input with invisible-region generation beyond observable camera boundaries. We present Panoramic Alignment for Strategic Exploitation of Generative Priors (PASTEL). Specifically, PASTEL proposes panoramic scene alignment, a novel representation that reformulates the intractable 3D "invisible region" exploration into a tractable 2D directional trajectory planning. This is achieved by reducing the viewpoint planning from 6-DoF search to a 2D directional search with explicit visibility boundaries. By operating within this panoramic space, our method strategically identifies camera trajectories that maximize exploration beyond observable boundaries while minimizing viewpoint deviation. Experimental results show that PASTEL can not only extrapolate plausible scene content beyond the observable boundaries of input monocular videos, but also substantially boost monocular 4D reconstruction performance. PASTEL outperforms the previous state-of-the-art method by 0.9dB in full-image PSNR on the DyCheck IPhone dataset. --- ### Causal DAG Identification for Count Data via Poisson Thinning Structural Equation Models Authors: Penggang Gao, Ming Cai, Hisayuki Hara Published: 2026-09-05 Categories: stat.ME, cs.LG, stat.ML arXiv: https://arxiv.org/abs/2609.06098v1 Count-valued variables arise in many scientific and applied settings, yet explicit structural models that allow full identification of causal DAGs from observational data remain limited. The Poisson branching structural causal model (PB-SCM) provides a count-valued analogue of linear structural equation models using binomial thinning and independent Poisson exogenous variables, but its causal DAG is generally only partially identifiable. Building on this framework, we propose the Poisson thinning structural equation model (PT-SEM), which replaces binomial thinning in PB-SCM with Poisson thinning and allows node-wise exogenous distributions from diverse count-distribution families. Under node-wise regularity conditions, we establish identifiability of the causal DAG, the thinning coefficients, and the node-wise exogenous distributions. The same identification analysis extends to binomial thinning, yielding full identifiability whenever every nonsink has non-Poisson exogenous noise. We further develop a structure learning algorithm that optimizes, via dynamic programming, a BIC score based on local likelihoods evaluated at plug-in moment estimates, and establish its consistency for DAG selection. Simulations demonstrate favorable performance in DAG recovery and thinning-coefficient estimation, and a real-data application illustrates the practical utility of PT-SEM. --- ### Flawed but Memorable: Student Critical Reception of Interest-Personalized GenAI Analogies in Computing Education Authors: Seth Bernstein, Naaz Sibia Published: 2026-09-05 Categories: cs.HC, cs.AI arXiv: https://arxiv.org/abs/2609.06095v1 Motivation: Undergraduate computing students increasingly turn to generative AI (GenAI) tools to understand abstract concepts through analogies. Analogies compare an unfamiliar concept to something familiar, but judging whether the comparison holds requires knowledge of both. GenAI may also embed assumptions about who the learner is. GenAI education research centers on output correctness, leaving students' critical reception of analogies largely unexamined. Method: We investigate how students evaluate the accuracy, appropriateness, and assumptions in GenAI-generated analogies, and their perceptions of interest-personalized versus generic technical explanations. Ten students with CS2 experience participated in a pre-survey, a think-aloud task with linked-list and recursion explanations, and a semi-structured interview grounded in the Paul-Elder framework. They judged accuracy, clarity, engagement, and trust separately. Results: Most participants described interest-personalized analogies as more engaging or memorable than generic technical explanations, while trust was mixed. Some trusted the tailored analogies more; others scrutinized them more closely or distrusted the tailoring. Participants with deep source-domain knowledge identified structural flaws requiring that knowledge to recognize. Because personalization and explanation format differed together, these findings do not isolate an effect of personalization alone. Implications: A familiar source flips the student's role. On the concept they are still learners, but on the familiar source they are the expert, and that is the position from which an analogy can be judged. We call this two-sided analogy auditing. GenAI systems should ask what students know, not just what interests them, and treat a flawed analogy as something to inspect and fix rather than accept. --- ### Automatic Red Teaming for Implicit Vulnerabilities of Text-to-Image Models Authors: Chang Ma, Junlin Han, Shuo Chen, Runjia Li, Philip Torr, Jindong Gu Published: 2026-09-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06094v1 Red-teaming Text-to-Image (T2I) models is essential for safe deployment, yet it remains particularly challenging against implicit adversarial prompts. Unlike explicit adversarial prompts that can be readily identified and blocked, implicit ones are much harder to detect: the prompts appear benign on the text surface yet still lead to inappropriate visual content. To address this, we propose Adversarial Probing for Implicit VulnErabilities (AdvPIE), a multimodal agentic framework to expose implicit vulnerabilities without requiring access to the parameters of target models. AdvPIE adopts a policy agent to generate and refine implicit adversarial prompts based on the feedback from a judge agent. To construct informative feedback, the judge agent provides modality-specific safety evaluation at both global and relative levels across iterations. To effectively leverage the feedback, we propose a novel Cumulative Adversarial Decoding strategy for the policy agent, which dynamically reweights token distributions to favor tokens that lead to more harmful images while preserving sampling diversity. Extensive experiments on standard and safety-aligned T2I models show that AdvPIE1 effectively uncovers implicit vulnerabilities, outperforming various baseline methods. --- ### Connectome-to-Function: Conditional Generative Latent Representations for Reservoir Computing Authors: Zhuolin Yu, Xingyu Liu, Yuanhao Jia, Yunhang Xiao, Hairuo Xue, Feihan Sun, Guozhang Chen Published: 2026-09-05 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.06093v1 Connectomes, graph-level maps of neurons and their synaptic connections, provide a structural basis for understanding how brain circuits support function and computation. However, mapping connectome structure to computation remains difficult because these graphs are high-dimensional, sparse, and sensitive to local structural variation. Existing approaches often depend on hand-crafted structural descriptors or task-specific predictors, which limits their ability to represent connectomes in a form that is both generative and functionally meaningful. We propose a conditional generative latent framework that encodes connectome graphs into a compact structural space while using available node-level conditions to guide reconstruction and generation. From this space, the model can reconstruct observed connectivity with a mean edge-reconstruction AUC up to 0.910 and generate new candidate connectomes, enabling a unified analysis of graph structure and computational behavior. Using connectome-derived graphs as recurrent computational substrates, we found that the learned latent space captures functional variation across reservoir-computing experiments, with cross-validated $R^2$ values up to approximately 0.87. Interpretability analysis further revealed task-specific structural mechanisms: in our examples, memory performance is associated with reciprocal recurrent connectivity, whereas prediction and classification are more strongly associated with spectral properties of the recurrent network. These findings suggest an AI-for-science approach to linking neural connectivity to computation and provide a generative and interpretable basis for studying how distinct structural mechanisms shape computational capacity. --- ### Explainable Deep Learning for Price-Trade Dynamics: From Black-Box Forecasts to Effective Parametric Models Authors: Manuel Naviglio, Fabrizio Lillo Published: 2026-09-05 Categories: q-fin.TR, cs.AI, physics.data-an arXiv: https://arxiv.org/abs/2609.06085v1 Understanding the joint dynamics of prices and trades is central to market microstructure, where returns and order flow interact through nonlinear and state-dependent mechanisms. Linear models are interpretable but may miss these effects, while deep neural networks improve forecasting at the cost of transparency. We use neural networks as tools for structural discovery rather than only for prediction. A deep feed-forward network is trained on high-frequency returns and signed volumes for large- and small-tick stocks and compared with a linear VAR benchmark. The neural network improves predictive performance, especially for returns, revealing nonlinear dependencies beyond the linear specification. Using Shapley-based explainability, we show that the dominant contributions are concentrated at the most recent lags. Model-implied responses are consistent with conditional averages reconstructed from the data. Unlike empirical averages, however, the neural-network decomposition isolates individual regressor contributions to the aggregate dependence. Lagged signed volume generates sign-preserving and saturating effects, consistent with nonlinear price impact and order-flow persistence. Lagged returns act as state variables: when the previous trade does not move the price, the model predicts continuation in the direction of past order flow, whereas non-zero returns generate attenuation or reversal. Building on these findings, we introduce a parsimonious SHAP-inspired nonlinear parametric model. It reproduces the main return-volume dependencies, outperforms the linear VAR benchmark, and achieves performance comparable to the neural network. A multi-lag extension captures residual longer-memory effects while preserving interpretability. Overall, explainability offers a route from black-box prediction to economically meaningful parametric models of price and trade dynamics. --- ### Learning to Price and Stock Under Contextual and Censored Demand Authors: Zean Han, Zezhen Ding, Jiheng Zhang Published: 2026-09-05 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.06083v1 To make optimal joint pricing and inventory control decisions is a critical challenge for modern retailers. In practice, retailers face changing market conditions where demands are influenced by various contextual factors, while simultaneously dealing with the difficulty of lost sales that obscure true demand information. However, existing approaches often fail to account for both contextual information and censored demand observations. We address this gap by presenting a framework where we model demand as a linear combination of basis functions with unknown coefficients, allowing for adaptive pricing and inventory decisions that respond to changing contexts. We propose an efficient algorithm to achieve regret bound $\mathcal{O}(K\sqrt{T}\log T)$ under concave revenue conditions and $\mathcal{O}(K^{2/3}T^{2/3}(\log T)^{1/2})$ for the general case, with matching lower bounds confirming optimality. Extensive numerical experiments across diverse scenarios demonstrate our algorithm's effectiveness. --- ### PhenoBench: Mapping What a Deeply Phenotyped Human Cohort Can Tell Us Authors: Gal Sapir, Alon Diament, Adva Wolf, Doron Yaya-Stupp, Dikla Gelbard Solodkin, Dana Azouri, Anat Etzion-Fuchs, Guy Lutsker, Eran Segal, Hagai Rossman Published: 2026-09-05 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.06080v2 Deeply phenotyped cohorts combine clinical, imaging, molecular, and wearable observations across timescales from seconds to years, but heterogeneous analyses are not directly comparable. We present PhenoBench, an executable benchmark that turns deep-phenotyping measurements into explicit questions and controlled comparisons of information sources and predictive models. It is built around the Human Phenotype Project, with more than 13,000 participants at the initial visit. Each question fixes the target, population, timing, and allowed information; its evaluation contract specifies the split, metric, baseline, and claim boundary. PhenoBench defines 90 clinically grounded tasks across 15 domains and 26 input modalities. Across 160 matched regression comparisons spanning 52 tasks, six pretrained tabular models ranked above the evaluated task-specific baselines, including XGBoost and CatBoost, under a fixed single-estimator protocol with bounded tuning. Giving each task equal weight, their mean advantage over ridge was 0.0103 $R^2$ (95% task-bootstrap interval, 0.0071-0.0136). We also evaluated 14 language models, collectively covering 40 tasks spanning phenotype recovery, classification, follow-up forecasting, and participant ordering. Without cohort-specific fitting, language models made informative predictions on some tasks but showed task-specific capability gaps, shared failures of scale, and rarely surpassed task-specific ridge or logistic regression models fitted on the same input fields. PhenoBench provides a versioned, auditable evaluation system where new questions, measurements, and models can be added without redefining existing comparisons. --- ### LayerRoute: Action-Conditioned Mixture-of-Layers Routing for Vision-Language-Action Policies Authors: Zheng Lu, Haoran Liao, Wanqi Zhong, Yunhe Ni, Lijie Wang, Xingjie Fan, Zhisheng Chen, Yantang Qu, Meijia Chen, Tianyu Xin, Zirui Song, Yiming Li Published: 2026-09-05 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.06079v2 Vision-Language-Action (VLA) policies leverage pretrained vision-language models (VLMs) to guide action generation for robot control. VLMs provide hierarchical visual-semantic representations that evolve across layers, from local visual geometry to abstract, language-aligned semantics; different manipulation tasks may therefore require different mixtures of layer representations. Meanwhile, the action module maintains intermediate representations that evolve throughout action computation and may provide useful information for subsequent decisions. However, existing VLA interfaces offer limited flexibility in representation access: VLM information is exposed through fixed layer assignments for each action layer, while intermediate action states are only propagated implicitly through residual streams without explicit reuse. We introduce LayerRoute, an action-conditioned representation routing interface that enables adaptive access to VLM layers and action representations. The Layer Mixture Router dynamically forms mixtures of cached VLM representations, while Action-State Reread reuses earlier action representations. Across diverse simulation and real-world benchmarks, LayerRoute consistently improves StarVLA-$π$ and $π_{0.5}$, achieving up to 7.2 gains on LIBERO Long with only 0.31% / 3.87% additional parameters. Ablation studies validate the benefit of action-conditioned layer routing, while routing analyses reveal structured allocation patterns across action layers and task settings. --- ### Report of the 8th LSVOS Challenge: Complex and Multimodal Video Object Segmentation Authors: Chang Liu, Henghui Ding, Lingyi Hong, Ning Xu, Linjie Yang, Yuchen Fan, Canyang Wu, Jinrong Zhang, Xusheng He, Ce Bian, Xianjing Han, Jianlong Wu et al. Published: 2026-09-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06078v1 This report summarizes the 8th Large-scale Video Object Segmentation (LSVOS) Challenge, held in conjunction with ECCV 2026. The challenge evaluates video segmentation in three complementary settings: complex semi-supervised video object segmentation on MOSEv2, text-guided referring video object segmentation on MeViSv2-Text, and audio-guided referring video object segmentation on MeViSv2-Audio. We describe the tasks and evaluation protocols and review the methods of the top three teams in each track. Across the nine leading solutions, foundation segmentation models are combined with target-aware memory, multimodal reasoning, explicit target-existence verification, agentic interaction, and corrective tracking. These systems illustrate a broader transition from single-model mask propagation toward modular pipelines that reason about object identity, query validity, and temporal reliability. --- ### Beyond Retraining-Free MoE Compression: A Cost-Normalized Study of Post-Compression Adjustment Authors: Sieun Hyeon, Jaeyoung Do Published: 2026-09-05 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.06076v2 Retraining-free MoE compression reduces deployment memory by pruning or merging experts, but often treats the compressed checkpoint as the final artifact. We argue that this view is incomplete: compressed MoE checkpoints are better understood as compressed initializations that benefit from a tiny post-compression adjustment stage. Across two MoE LLM backbones, four pruning/merging methods, three expert-retention ratios, and 28 benchmarks, we compare LM fine-tuning and teacher-based KD under matched small-data budgets and measured GPU costs. Using only 3,000 C4 examples and a single epoch of adjustment, Full FT recovers 37.3% of the original-to-compressed performance gap on average. Moreover, LM fine-tuning is more cost-effective than standard token-level KD, and full-parameter adjustment gives the strongest cost--recovery trade-off among the tested scopes. These results suggest that retraining-free compression should be paired with small post-compression adjustment to recover a substantial portion of the performance lost during compression. --- ### NSFlow: End-to-End Differentiable Neuro-Symbolic Optical Flow for Visual Odometry Authors: Yicheng Lin, Zhipeng Fei, Yuxiu Xu, WenDong Chen, Cong Li, Bin Han Published: 2026-09-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06074v2 Sparse optical flow provides stable inter-frame correspondence, playing a key role in Visual Odometry (VO) and Visual-Inertial Odometry (VIO). Classical optimization-based methods, such as Lucas-Kanade (LK), perform well under small displacements but are sensitive to large motions and illumination changes. Modern regression-based learning methods, while more robust in complex scenes, are often computationally heavy and lack explicit geometric consistency, making them less suitable for efficient VO/VIO front-ends. To bridge this gap, we propose a hybrid neuro-symbolic framework that combines the strengths of both paradigms. Our method uses a Convolutional Neural Network (CNN) to extract robust feature representations, which is fed into a differentiable LK optimizer to estimate optical flow in an end-to-end trainable manner. Through implicit differentiation, gradients are propagated across the iterative solver, enabling joint optimization of feature extraction and flow estimation. The resulting system integrates seamlessly into existing VO/VIO pipelines and runs in real-time on embedded platforms. Experiments show that our method outperforms conventional optimization-based flow in challenging conditions such as dynamic lighting and low texture, while also achieving higher accuracy and lower latency than purely regression-based alternatives. When deployed in a VIO system, our method demonstrates significant performance improvement, achieving an average error reduction of 42\% on challenging datasets while enhancing tracking stability. The code is publicly available. --- ### FedSubMuon: Communication-Efficient Federated LLM Fine-Tuning via Structured Subspace Muon Authors: Shaolong Chen, Youming Tao, Shuzhen Chen, Falko Dressler, Qingqing Ye, Di Wang Published: 2026-09-05 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.06073v1 Federated fine-tuning adapts large language models (LLMs) to decentralized client data, but its scalability in cross-device training is often limited by the high communication cost. Muon is an optimizer that improves optimization performance by orthogonalizing momentum for matrix-valued parameters. Existing federated Muon methods demonstrate the benefit of matrix-aware optimization in federated learning, but still require transmitting full layer-size updates and optimizer state. A natural way to reduce communication is to directly apply Muon to LoRA factors, but this changes the optimized object and weakens Muon's matrix-aware update geometry. We propose FedSubMuon, a communication-efficient federated Muon fine-tuning method that optimizes compact coefficient matrices within shared structured subspaces. This design keeps Muon on a single matrix-valued trainable object, while reducing the client upload to compact coefficient matrices. We further introduce FedSubMuon-GT, an accuracy-oriented extension that uses projected gradients to adapt tracked subspace bases toward task-relevant gradient directions. Experiments on instruction tuning and mathematical reasoning show that FedSubMuon-GT achieves the best overall accuracy on four of five dataset-model pairs, while FedSubMuon performs best under all matched communication budgets. On Dolly-15K, the closest communication baseline requires 5.5 times and 1.4 times more total communication on Llama-1B and Qwen-4B, respectively. --- ### ACE: Adapter Consolidation across Experts for Parameter-Efficient Fine-Tuning of MoE LLMs Authors: Ahin Lee, Sehyun Yun, Joonha Park, Taesik Gong Published: 2026-09-05 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.06072v1 Parameter-efficient fine-tuning (PEFT) of mixture-of-experts (MoE) models commonly attaches a separate low-rank adapter to each expert. This expert-wise design fragments adaptation in three ways: capacity is split across narrow low-rank updates, gradient supervision becomes sparse and imbalanced under sparse routing, and execution is decomposed into many small GEMMs. We find that such expert-wise separation is often unnecessary, as subsets of LoRA adapters become functionally similar during fine-tuning, revealing redundancy among expert-specific adapters. Based on this redundancy, we propose ACE (Adapter Consolidation across Experts), which groups redundant experts and replaces their expert-specific adapters with group-shared higher-rank LoRA modules under the same PEFT budget. ACE further introduces grouped adapter execution, which consolidates fragmented expert-wise adapter computations into fewer, larger group-level GEMMs. Across evaluations covering 12 datasets and four MoE backbones, ACE achieves the highest observed mean accuracy among the parameter-matched PEFT methods on the three backbones with complete baseline coverage, while providing $1.31\times$ to $1.48\times$ wall-clock training speedup over expert-wise LoRA without increasing peak memory. Our code is available at https://github.com/UbiquitousAILab/ACE. --- ### Generating Instance Generators in PDDL Planning Authors: Nicola J. Müller, Naya Rudolph, Katharina Stein, Jörg Hoffmann, Ayal Taitler, Timo P. Gros Published: 2026-09-05 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.06071v1 PDDL, the de-facto standard language in the AI Planning community, is designed to specify planning domains: sets of instances that share the same predicates and action schemas. Yet it does not provide any means to specify the actual instance set, i.e., legality constraints on initial states and goal conditions, as well as possibly domain subset constraints specifying an instance subset we are interested in. One consequence of this is that instance generation has always been ad-hoc, with manually written domain- and subset-specific instance generators. Recent work has started to address this, through reasoning and learning methods that however suffer from scalability limitations. Here we introduce an alternative approach, leveraging LLMs to generate instance-generation programs, with built-in soundness guarantees through prescribed checks. We show that these automatically generated instance generators return large numbers of sound and diverse instances efficiently. --- ### DPH Parser: A Bottom-Up Grammar-Driven Parser for Joint Constituency and Dependency Analysis Authors: Hussein Ghaly Published: 2026-09-05 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.06070v1 This paper presents Dependency-Phrase Hierarchy Parser (DPH Parser), a grammar-driven bottom-up unsupervized parsing framework inspired by Generalized Phrase Structure Grammar (GPSG) and Head-driven Phrase Structure Grammar (HPSG). The parser incrementally constructs constituency structures using a compact inventory of feature-based syntactic rules while deriving dependency relations through explicit head annotations. The system combines probabilistic POS tagging, recursive phrase projection, and weighted parse hypotheses to process realistic and partially noisy text input. Unlike purely neural and data-driven parsers, the resulting syntactic derivations remain explicitly interpretable. We evaluated parser performance on English corpora from the Universal Dependencies (UD) project using Unlabeled Attachment Score (UAS) as the main parsing metric, comparing the outcomes against Stanza and spaCy parsers. For a small inventory of syntactic rules, DPH parser achieved UAS values of 53.32% & 52.58% (UD Devset/Testset respectively). For the same data, Stanza achieved 89.12% & 88.67% while spaCy achieved 56.91% and 58.59%. Although the current system does not yet approach the accuracy of modern neural parsers, the results demonstrate the feasibility of applying transparent rule-based bottom-up parsing to realistic treebank data while jointly producing constituency and dependency structures. --- ### Don't Lose Entities from Retrieval to Generation: Dual Entity Recovery RAG for multi-hop QA Authors: Heechang Lee, Dong-Young Lim Published: 2026-09-05 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.06065v1 Retrieval-augmented multi-hop question answering (QA) decomposes a query into sub-questions and decomposes the corpus into smaller retrieval units such as sentences. Both forms of decomposition improve the pipeline, but we show that both share the same vulnerability, the loss of entity information, and that this loss breaks the pipeline at two separate points. The first point is retrieval, where a sub-question loses the entity resolved at the previous hop, leaving the retriever with nothing to match against. The second point is harder to see, because retrieval still appears to succeed. Once a passage is split into sentences, an isolated sentence loses the context that grounds its pronouns, so even with the correct sentence in hand the LLM cannot tell which entity the sentence is about. We isolate this second point as a distinct failure mode that we call lost-in-generation, and a retrieval-controlled experiment shows that it degrades answers even when the gold evidence is fixed in the context. We then propose Dual Entity Recovery RAG (DER-RAG), which keeps the grounding entity explicit from retrieval through to generation with two lightweight components, a two-way query decomposition that carries the resolved entity across sub-questions and a subject entity prefix attached to each sentence at generation time. DER-RAG needs no graph construction, no corpus modification, and no fine-tuning, yet on three multi-hop QA benchmarks it matches or exceeds strong baselines, including graph-based methods that depend on costly offline structures. --- ### The Role of Gradient Modification in Heavy-Tailed Nonconvex Stochastic Min-Max Optimization Authors: Tianxi Zhu, Yi Xu, Xiangyang Ji Published: 2026-09-05 Categories: math.OC, cs.AI, stat.ML arXiv: https://arxiv.org/abs/2609.06064v1 Stochastic min-max optimization has attracted increasing attention due to its applications in modern machine learning, while existing theoretical studies mainly rely on the bounded variance assumption for stochastic gradients. Under heavy-tailed noise, where stochastic gradients only possess a finite $p$-th moment for $p\in(1,2)$, gradient clipping or normalization is commonly believed to be necessary to guarantee convergence. In this work, we revisit stochastic min-max optimization under heavy-tailed noise and provide a comprehensive theoretical study of stochastic gradient descent ascent (SGDA). We first show that vanilla SGDA, without any modification to its update rule, can converge under heavy-tailed noise in both nonconvex-strongly-concave (NC-SC) and nonconvex-concave (NC-C) settings, establishing the first convergence guarantees for SGDA in these regimes. Beyond unregularized problems, we further investigate regularized stochastic min-max optimization, where directly incorporating gradient normalization into proximal updates is nontrivial due to the incompatibility between normalization and proximal structures. We overcome this difficulty by developing new clipping-free algorithms, i.e., Stoc-TRGDAM and Stoc-TRGDmax, and they both can achieve the optimal dependence on the target accuracy without using gradient clipping. --- ### Explaining AI Agents Through Execution Traces Authors: Vittoria Vineis, Fabiano Veglianti, Lorenzo Antonelli, Claudia Di Carlo, Matteo Silvestri, Gabriele Tolomei Published: 2026-09-05 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.06063v1 AI Agents are increasingly deployed in real-world settings, where they interact with external tools and make sequential decisions with limited human oversight. This creates a pressing need for reliable and auditable explanations of what an agent did and why. However, traditional Explainable AI (XAI) methods fall short of providing the process-level transparency required for such interactive, multi-step systems, motivating a paradigm shift toward approaches specifically designed for AI Agents. To address this gap, we present a post-hoc XAI framework that transforms a lengthy agent's execution trace into a structured report and a faithful natural-language explanation explicitly grounded in its observable behavior. Because it relies solely on execution traces, the framework applies across different agent architectures, environments, and tasks. Human and automated evaluations across multiple benchmarks and architectures show that our framework produces high-quality, trace-faithful explanations while reliably identifying unsupported claims, unjustified actions, and evidence gaps, outperforming naive LLM-generated explanations. --- ### Calendar-Structured Sparse Principal Component Analysis for Interpretable Multi-Periodic Electricity Consumption Profiles Authors: Carlos Quesada-Granja, Tony Castillo-Calzadilla, Carlos Rizo-Maestre Published: 2026-09-05 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.06060v2 Long-term electricity-consumption profiles exhibit several simultaneous periodic structures, including daily, weekly, and annual cycles. This work introduces Calendar-Structured Sparse Principal Component Analysis (Calendar-SPCA), a structured representation-learning method that incorporates this known multi-periodic geometry directly into a low-dimensional factorization. The method represents the feature domain as the Cartesian product of cyclic calendar axes and combines an L1 loading penalty with graph total variation, producing sparse, locally coherent, and directly interpretable latent factors. In this study, Calendar-SPCA is applied to the interpretable analysis of long-term electricity-consumption profiles and evaluated on two independent public smart-meter datasets, GoiEner and Low Carbon London, with different population sizes and temporal resolutions. A factorial experiment characterizes the effects of sparsity and calendar coherence and examines robustness across sample size, latent dimensionality, and repeated fits. At rank 15, Calendar-SPCA retains 96.92% and 82.90% of the explained variance of rank-matched principal component analysis (PCA) in GoiEner and Low Carbon London, respectively, with mean loading sparsities of 61.95% and 81.50%. Comparisons with classical sparse PCA and sparse PCA with total variation (SPCA-TV) show that Calendar-SPCA organizes latent factors into interpretable structures over the daily, weekly, and annual calendar axes while preserving substantial low-rank information. --- ### DAREBench: Deployment-Aware and Reliable Evaluation of Models as Agents Authors: Yu Liu, Zhilin Liu, Zhiwei Yang, Shaojie Zhang, Zheyuan Deng, Tingwei Huang, Zhenbo Luo, Lei Jiang, Yanbing Liu, Pei Fu Published: 2026-09-05 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.06059v1 As large language models evolve from question-answering systems into general-purpose agents, evaluation must move beyond static answer correctness to assess multimodal perception, multi-step execution, tool use, and artifact delivery. However, existing benchmarks are often tied to specific task types, execution environments, or scoring protocols, limiting their comparability, interpretability, and reliability for deployment decisions. We introduce DAREBench (Deployment-Aware and Reliable Evaluation of Models as Agents), a benchmark designed to capture workload variation and support reliable agent evaluation. Built on a shared OpenClaw execution environment, DAREBench organizes 233 tasks selected and adapted from 22 source benchmarks into a $2\times3$ workload matrix defined by input modality and execution form, and evaluates them under a unified contract-based protocol with evidence-based score auditing. We evaluate 23 commercial API models and 12 locally deployed open-weight models over 7,587 model--task runs, reporting accuracy and token consumption alongside reference costs for API models. Results show that no single model dominates all workload groups, text and multimodal tasks exhibit distinct accuracy--cost trade-offs, and local open-weight models are competitive in several groups but still trail frontier commercial models overall. These findings suggest that agent deployment and model selection should consider workload profiles, deployment mode, and accuracy--cost trade-offs rather than rely on a single aggregate score. --- ### GradeTrap: Authority Cues in Images Shift VLM Judgments Despite Explicit Instructions to Ignore Them Authors: Deep Dessai Published: 2026-09-05 Categories: cs.CV, cs.CL, cs.CY, cs.LG arXiv: https://arxiv.org/abs/2609.06058v1 As vision-language models (VLMs) become increasingly capable and are deployed in consequential real-world settings, they must evaluate evidence independently rather than defer uncritically to human authority. We introduce GradeTrap, a controlled evaluation that places two social cues in direct conflict: a student answer, which should attract sycophantic agreement, and a conflicting answer attributed to a peer, teacher, or official answer key, which should attract authority-based deference. Models produce free-form answers while being explicitly instructed to solve independently and ignore all student answers, feedback, and grading marks. We test the models on 60 synthetic real-world trade-off scenarios. Five neutral trials establish a stable model-relative preference, followed by three repetitions of six experimental cues including controls. On the 45-item common intersection across Gemini 3.5 Flash-Lite, GPT-5.6 Luna, and Claude Haiku 4.5, a generic second-answer control yields 5.4% conflicting-answer selection. Relative to that control, pooled within-item changes show no reliable peer-review effect, a 6.9-point teacher-review effect, and a 19.5-point official-key effect. In contrast, a displayed conflicting student answer alone compared to a displayed student reference answer alone only raises selection from 2.2% to 5.2%. Official-key provenance therefore redirects judgements more than a student answer or the generic second-answer control, despite an explicit ignore instruction and an opposing student answer given along with the official key. Effects vary in magnitude across the three models. --- ### DriveZero: End-to-End Driving Beyond Human Demonstrations Authors: Hao He, Chengcheng Hu, Zirun Su, Heng Zhang, Haisong Liu, Jinke Li, Haochen Tian, Zhenwei Shen, Hongyang Li, Zhichao Li, Yunchen Yang, Bochao Huang et al. Published: 2026-09-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06055v1 Most end-to-end autonomous-driving systems learn by imitating human driving logs, leaving their learned behavior constrained by the quality and behavioral coverage of the recorded trajectories. This report presents DriveZero, an end-to-end system that learns driving behavior beyond human demonstrations. It decomposes driving into a perception model and an action model, pretrains each in the regime best suited to it, and combines them into one end-to-end planner. The two models call for different learning recipes: perception must understand the world, and benefits from massive and diverse visual data; action must interact with it, and requires closed-loop feedback. On the action side, we introduce DriveRL, a mixed-agent closed-loop reinforcement-learning framework. It converts real driving logs into interactive worlds, where a privileged teacher policy is trained with PPO through closed-loop rollouts. For the perception model, DriveVFM consolidates multiple frozen vision foundation models, including DINOv3, SigLIP2, SAM and Depth Anything V2, into a single backbone from raw images alone, requiring no task-specific annotations. DriveZero then unifies the two: a camera-only planner that distills the frozen DriveRL teacher through its rolled-out trajectories. The goal-conditioned teacher can moreover be queried under augmented driving intents, yielding diverse, goal-consistent supervision that logged data cannot provide. On nuPlan, DriveRL with value-guided test-time action search achieves a mean score of 93.57 across the Val14, Test14-hard, and Test14-random community splits in both non-reactive and reactive modes, exceeding the Log-Replay expert on all three splits. DriveZero achieves state-of-the-art performance on NAVSIMv1, NAVSIMv2 and the closed-loop HUGSIM benchmark without any human trajectory supervision. --- ### Data Quality Rule Generation with LLMs Authors: Anna-Christina Glock, Thomas Hütter, Johannes Fürnkranz, Wolfram Wöß, Christine Dominka-Kiss, Lisa Ehrlinger Published: 2026-09-05 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.06053v1 The validation of data, such as customer and employee data, is an important task in many organizations. Errors in data can have severe consequences. For example, a wrong drug unit in a patient record can lead to life-threatening medication errors, and a missing street number in an address to failed deliveries. Companies often employ rule-based enterprise data quality (DQ) tools, which allow domain experts to specify rules to validate the data over time. While rule-based DQ tools are computationally efficient and provide explainable reports, maintaining a comprehensive rule set manually is challenging, as domain experts often overlook essential rules, especially in complex domains and large data volumes. Hence, closing these gaps remains an open problem in practice. In this paper, we address the challenge of automated DQ rule generation. For this, we formalize a generalizable generate-filter framework and introduce LeDQeR, an LLM-based DQ rule generation approach. First, a large language model (LLM) generates candidates rules from an observed dirty data tuple for a given rule-based DQ tool syntax. Second, we apply four filter techniques that ensure the (i) executability, (ii) correctness, and (iii) generalizability, and avoid (iv) redundancy of the generated rules. An extensive experimental evaluation suggests that LeDQeR is able to produce effective and compact rule sets for various datasets and error types. --- ### SkillSpec: Intent-Masked Specification Reasoning for Agent Skill Correctness Authors: Yizhuo Zhang, Bo Kang, Yi Yang, Zhiyu Duan, Zhouteng Ye, Shunkun Yang Published: 2026-09-05 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2609.06052v1 Autonomous agent systems increasingly depend on reusable skill abstractions for consolidating experiential knowledge and domain expertise. These artifacts typically bundle free-form instructions with heterogeneous resources. However, ensuring their correctness remains challenging. Their failure modes transcend conventional code defects to subtle semantic inconsistencies such as intent conflicts, which manifest as silent failures masked by the underlying model. Moreover, skill correctness must be grounded in intended task boundaries and generalizability. We propose SkillSpec, a Hoare-style framework that formulates skill correctness as a specification reasoning problem. It transforms a heterogeneous skill repository into a unified graph representation that aligns descriptions, instructions and code artifacts. For each node, SkillSpec derives an ExpectSpec from the surrounding declared intent, and infers FactSpecs from encoded behavior under partially disclosed intent. An intent mask regulates access to holistic, lineage, neighborhood, and local views to balance the bias introduced by excessive context against unsupported inference caused by insufficient context. SkillSpec jointly reasons over these views to flag candidate defects, and automatically validates them in an isolated sandbox. On 515 real-world skills from SkillsBench and widely downloaded repositories, SkillSpec identified 763 manually confirmed defects across 239 skills, achieving 61.2% precision. The node-level analysis across multiple model families shows that specification reasoning is consistently reliable for code nodes, whereas plain-text nodes remain a major bottleneck. Most defects arise at the boundaries between declared intent and implementation, demonstrating that explicit specifications provide a practical foundation for skill quality assurance in real-world agent ecosystems. --- ### Image-Scale Robustness and Visual Recognition Performance: A Cross-Architecture Analysis Authors: Anish Monsley Kirupakaran Published: 2026-09-05 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.06051v1 The sensitivity of visual recognition models to changes in image scale is well established, yet the factors governing this sensitivity across heterogeneous architectures remain unclear. In this work, we investigate whether scale robustness exhibits a common quantitative structure across modern vision models. We evaluate 20 pretrained ImageNet-1K classifiers spanning seven architectural families, including convolutional, mobile, efficient, and Transformer-based architectures. By systematically reducing input image scale, we construct scale-accuracy response curves and define a characteristic scale as a compact measure of the onset of substantial recognition degradation. We then examine the relationship between characteristic scale and baseline recognition accuracy, model parameter count, architectural family, and representation stability. A strong inverse association is observed between baseline accuracy and characteristic scale (Pearson r = -0.890, R^2= 0.792, p < 10^-6). This relationship remains stable under bootstrap resampling, leave-one-architecture-out analysis, and leave-one-family-out analysis. In contrast, parameter count provides negligible additional explanatory power after controlling for baseline accuracy (p = 0.80), while architectural family does not provide significant incremental explanatory power. Furthermore, characteristic scale shows essentially no association with representation stability (r = -0.003, p = 0.991). These results indicate that, across the studied models, scale robustness is strongly organized by baseline recognition performance rather than simply by model size, architectural family, or representation stability. The study provides an empirical framework for characterizing scale robustness across vision architectures and identifies a reproducible accuracy-scale regularity that warrants further theoretical investigation. --- ### A solution to the Erdős Problem #1040 Authors: Ioannis Tzachristas Published: 2026-09-05 Categories: math.CA, cs.AI arXiv: https://arxiv.org/abs/2609.06050v1 For a compact set $K\subset\mathbb{C}$, let $\vartheta(K)$ be the infimum of the planar areas of the unit lemniscates of all monic polynomials with zeros in $K$, allowing arbitrary degree and repeated zeros. We prove that $\vartheta(K)=0$ whenever $\operatorname{cap}(K)=1$, with no regularity assumption on $K$. The proof uses a centered harmonic polynomial that is positive on all but a set of arbitrarily small area in the polynomial hull of $K$. A Fourier average of exterior harmonic measures realizes this polynomial as the logarithmic potential of a signed measure having bounded density with respect to the equilibrium measure. A positive perturbation and an $L^1$ approximation by empirical measures then produce the required polynomials. This extends the smooth-boundary result of Krishnapur, Lundberg, and Ramachandran to arbitrary compact sets of capacity one. Together with the capacity-greater-than-one theorem of Ghosh and Ramachandran and an elementary argument for unbounded sets, it follows that $\vartheta(F)=0$ for every closed infinite set $F\subset\mathbb{C}$ of transfinite diameter at least one, answering the vanishing question in Erdős Problem 1040. --- ### Generating Adversarial Texts for Machine Translation via GRPO Authors: Florian Zogaj, Jakob Hütteneder, Giovanni De Muri, Federico Villa, Aryan Sood, Vilém Zouhar Published: 2026-09-05 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.06048v1 As machine translation (MT) systems continue to improve, standard benchmarks become less informative for exposing remaining weaknesses. Traditional methods for creating challenging test sets rely on expensive manual creation or curation, while automated approaches struggle to produce sets with the necessary translation difficulty and linguistic diversity. We propose a scalable reinforcement-learning-based approach for rewriting existing source texts into instances that are more difficult to translate for MT systems. We fine-tune a large language model with Group Relative Policy Optimization (GRPO), using reward signals based on translation difficulty together with constraints for semantic similarity, grammaticality, and approximate length preservation. On WMT25, our approach substantially reduces average COMET translation quality from 0.63 to 0.48, while preserving grammaticality and readability, whereas the base model remains at 0.64. Evaluations on the unseen WMT19-WMT24 benchmarks confirm that this behavior generalizes beyond the training data, and human evaluation further shows that the rewrites substantially lower translation quality while incurring a moderate drop in naturalness and only a small change in grammaticality. We release our code to support reproducibility. --- ### FALCON-S: Fixed-wing ground-effect Aerodynamics Simulator and Flight Control Learning Suite Authors: Matteo El Hariry, Pedro Lima, Andrej Orsula, Matthieu Geist, Miguel Olivares-Mendez Published: 2026-09-05 Categories: cs.RO, cs.LG, eess.SY arXiv: https://arxiv.org/abs/2609.06046v1 We present a modular, high-fidelity simulation framework for the development and benchmarking of flight control strategies in fixed-wing aerial robots operating near the ground. Unlike existing simulators that rely on simplified or hover-oriented dynamics, our framework models full 6DoF rigid-body physics, semi-empirical ground-effect aerodynamics, actuator dynamics, sensor noise, and environmental disturbances. This physical realism, combined with modular component design, enables systematic analysis of low-altitude flight behavior under realistic conditions. The simulator supports both CPU and GPU backends via Torch and NVIDIA Warp, enabling high-throughput parallel execution suitable for large-scale reinforcement learning training and optimal control rollouts. A unified interface accommodates a range of controllers (both RL and optical control algorithms) across tasks such as altitude regulation and trajectory tracking. Cross-validation with X-Plane and JSBSim is also supported to facilitate engineering integration and visual fidelity. --- ### Minimizing the Effect of Sleep Deprivation in the Forward-Forward Algorithm Authors: Joy Datta, Puja Saha, Rawhatur Rabbi, Nafiz Imtiaz Rafin, Swakkhar Shatabda, Md. Golam Rabiul Alam, Chad Mourning Published: 2026-09-05 Categories: cs.LG, cs.NE arXiv: https://arxiv.org/abs/2609.06042v1 This paper addresses the challenge posed by sleep deprivation in the Forward-Forward algorithm, where separating the two passes in this algorithm and imbalancing the data processing in the passes is considered an imitation of the cognitive processes observed in humans suffering from sleep deprivation. Previous research has demonstrated that sleep deprivation in the Forward-Forward algorithm has a catastrophic effect on learning efficacy. To mitigate this issue, we explore several approaches; these include alternative activation, optimized loss function, and threshold tuning. To simulate periodic rest, we reduce the number of positive passes in alternating epochs, creating short break phases. We additionally investigate the potential of caffeine-induced stimulation to enhance performance during sleep-deprived conditions. Experimental evaluations conducted on the MNIST and Fashion-MNIST datasets demonstrate that these modifications improve accuracy under the context of sleep deprivation. For example, a 2%-62% accuracy gain is observed in a severe sleep deprivation setting (16 positive or awake periods and 1 negative or sleep period). The approaches also enhance the resilience of the algorithm and its alignment with the adaptive mechanisms of human cognition. --- ### Generator-Independent Runtime Assurance under Partial Observation Authors: Guangxi Wan, Yongbo Xie, Yuqi Liu, Qingwei Dong, Qingxin Li, Hongfei Bai, Peng Zeng Published: 2026-09-05 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.06036v1 Proposal-based controllers---learned policies, language-model planners, and other black-box \emph{generators}---are increasingly deployed behind runtime verification gates. We ask when the closed-loop safety guarantee decouples from the generator. The prevailing per-candidate certification pattern does not compose: under retry or best-of-$k$ selection a per-candidate false-admission level $α$ can inflate to $1-(1-α)^{k}$. Our main theorem shows that \emph{simultaneous setwise soundness}---certifying a set of admissible proposals containing no nonviable action---is necessary and sufficient for generator-independent \emph{admission soundness}, the worst case over all generators of executing a nonviable proposal equalling the probability of setwise failure; together with a design-time certificate and a no-bypass rule it is sufficient for \emph{contract safety}, with violation bound $Γ+\sum_t\varepsilon_t+η$ invariant under arbitrary, even adversarial, replacement of the generator. A second theorem bounds every admission mechanism under partial observation: for a fixed probing and admission policy, if two state hypotheses whose information laws lie within total-variation distance $δ$ require different safe decisions, then $\abar+β+δ\ge1$. A sequential risk ledger makes the guarantee implementable with time-uniform confidence tubes, and shows that deterministic admission computations concentrate all statistical risk in state estimation. Simplex-style runtime assurance and control-barrier-function filtering are recovered as degenerate cases. --- ### Evaluating Deep-Search Agents under Hierarchical Web Evidence Poisoning Authors: Zhongan Bi, Qiwen Wang, Jianrong Jiang, Jigang Ding, Wenwen Xiong, Changhua Meng, Xuanang Gao, Kepeng Lin, Changjiang Jiang, Yiang Chen, Huan Yao, Wei Wang et al. Published: 2026-09-05 Categories: cs.CR, cs.AI, cs.IR arXiv: https://arxiv.org/abs/2609.06027v2 Search-augmented LLM agents are increasingly used for consumer decisions, making them vulnerable to Generative Engine Optimization (GEO) poisoning. Existing benchmarks largely measure whether manipulated content is retrieved or endorsed, but do not track whether an agent verifies suspicious evidence, revises adopted claims, or recovers before producing its final recommendation. We introduce HAE-GEO, a benchmark that tracks the full trajectory from exposure to recovery under progressively more persuasive Web poisoning. Agents interact via a multi-turn Search-Scrape interface across three attack levels (L1 direct assertion, L2 contextual camouflage, and L3 apparent corroboration), supported by a controlled corpus of 72,039 clean pages and 770 poisoned pages per level spanning 8 product categories and 154 brands. Evaluation combines deterministic behavioral measures with six semantic rubric dimensions. Evaluating 10 agents, we find three recurring patterns: evidence recognition degrades under the corroboration trap; agentic search improves final resistance without improving evidence recognition or utility; and defense prompting increases verification, yet rarely converts verification into recovery. --- ### Factors Influencing the Emergence of Dependency Length Minimization in Neural Agent Simulations Authors: Yuqing Zhang, Tessa Verhoef, Gertjan van Noord, Arianna Bisazza Published: 2026-09-05 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.06025v2 Given various grammatical options, language users prefer the word order choice that reduces the overall length of syntactic dependencies, a principle known as dependency length minimization (DLM). The origins of this preference remain an open question, particularly whether it originates from constraints on efficient information processing. Computational simulations provide a powerful approach to identifying the factors influencing the emergence of linguistic phenomena. However, previous simulations of DLM have not examined realistic interaction contexts and have produced mixed results. The present study investigates the emergence of DLM in artificial languages using a recently proposed language learning and communication framework based on recurrent neural networks (RNNs). In this framework, agents are trained to speak and interpret artificial languages and then use these languages to communicate. Using this framework, we study the impact of several factors related to processing limitations in a communicative setting, such as noise during listening, limited speaker capacity, and incremental sentence processing. Our results reveal a complex interplay among these factors in shaping word order preferences in neural agents. Specifically, in the full meaning space, agents regularize toward a single dominant word order, while in the half meaning space they show a short-before-long preference that only aligns with DLM in verb-initial languages. A consistent DLM preference emerges only when agents are subject to incremental processing pressure. These findings suggest that limitations in human cognitive processing may indeed play a role in shaping DLM. Our findings provide insights into the conditions under which neural models replicate human-like preferences and highlight the challenges of designing emergent communication models that capture human cognitive biases in language processing. --- ### What Does Animal Re-Identification Learn? Linear Biological Concepts and Their Origins in Visual Representations Authors: Robert Nolting, Alexandra Schild, Moritz Weckbecker, Maximilian Schall, Gerard de Melo Published: 2026-09-05 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.06020v1 Conservation increasingly relies on camera traps that collect more wildlife imagery than experts can manually analyze, making animal re-identification (Re-ID) essential for monitoring individuals and populations. Yet understanding which cues drive model decisions is challenging for ViT-based Re-ID models, whose metric-learning objectives provide no explicit supervision for biological concepts. We ask whether such models nonetheless organize their representations along biologically meaningful axes. Using a DINOv3 backbone fine-tuned for Western lowland gorilla Re-ID with triplet-margin loss, we find that sex and age emerge as linear directions that generalize to held-out individuals, reaching up to 0.91 AUROC and being recoverable from a single image per individual. Activation steering further shows that the sex direction is causally used by the model, flipping a significant fraction of predictions to the opposite sex. Comparing off-the-shelf and fine-tuned backbones shows that Re-ID training does not create these concepts, but relocates them across the network. Finally, data attribution reveals that the representation we find reflects a graded biological axis, is redundantly encoded across the population and shaped by visually ambiguous individuals. Together, these findings show how interpretability can uncover both the biological structure and failure modes of Re-ID representations, providing a step toward auditable computer vision for wildlife monitoring. --- ### IXPLORE: Bounded Ideal Point Estimation with Grid-Based Uncertainty Quantification Authors: Fynn Bachmann Published: 2026-09-05 Categories: cs.LG, cs.MS arXiv: https://arxiv.org/abs/2609.06018v1 Ideal point estimation is widely used to analyze and visualize political data. However, selecting the corresponding spatial model involves various trade-offs: while model-based approaches such as Item Response Theory (IRT) are based on utility functions rather than optimized for predictive accuracy, most Machine Learning (ML) alternatives struggle to generalize beyond training data when embedding sparse test responses. We introduce IXPLORE, a bounded ideal point estimation algorithm that combines a predictive fit objective with a sparsity-aware likelihood function. On five benchmark datasets spanning surveys, roll calls, and deliberation, this approach surpasses model-based and ML-based algorithms on reconstruction and imputation error - especially for users with sparse responses. Furthermore, we show that non-linear feature transforms can further reduce the reconstruction error while remaining visually interpretable. To quantify uncertainty, IXPLORE applies grid-based posterior inference on a bounded 2D latent space. Available as a Python package on PyPI, IXPLORE offers a flexible framework for constructing bounded, interpretable political maps with fast inference and strong imputation performance. --- ### FujinSplat: Seeing Through Smoke with RAW-Domain Gaussian Splatting Authors: Gengjia Chang, Ziteng Cui, Shuhong Liu Published: 2026-09-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06017v2 The appearance of a smoky scene is shaped by two processes that a camera records together: the participating medium alters scene radiance in a view-dependent way, and the image signal processor (ISP) then remaps the result through a nonlinear tone and color transformation. Recovering a clean 3D scene requires separating both. Per-view sRGB dehazing acts only after the ISP has entangled them; standard 3D reconstruction ignores the medium and absorbs it into scene geometry and radiance. FujinSplat addresses the problem in the RAW domain, where the two processes remain separable. A per-scene Base ISP is fitted from the scene's hazy RAW captures to its own camera renderings and then frozen, providing a fixed photometric anchor that performs no dehazing. Analyzing expert corrections reveals a compact, low-dimensional correction space identifiable from RAW alone. FujinSplat therefore fits per-view action answers at the training poses and trains a single scene-agnostic controller to regress them from RAW; the corrected views supervise one static 3D Gaussian representation, jointly with a bounded per-view residual that reconciles cross-view photometric inconsistencies. On the RealX3D real-world smoke benchmark FujinSplat clearly outperforms the strongest comparable baseline, ahead of both physics-based reconstruction and restoration-then-3DGS pipelines. Code is available at https://github.com/I2WM/FujinSplat . --- ### Granular-Ball Quantum Clustering for Resource-Efficient and Robust Learning Authors: Suzhen Yuan, Qilin Xie, Lifeng Shen, Shuyin Xia, Jermiah D. Deng, Guoying Wang Published: 2026-09-05 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.06016v1 Quantum clustering aims to exploit quantum feature representations to uncover complex data structures beyond conventional Euclidean geometry. Yet this sample-level kernel construction requires O(n^2) quantum circuit executions for n data points, creating a major bottleneck under near-term quantum resource constraints. Prior solutions fail to resolve this efficiency-accuracy dilemma: classical granular-ball clustering reduces sample complexity but relies on Euclidean metrics that cannot capture quantum correlations, while existing quantum compression schemes prioritize efficiency over structural preservation, degrading performance on non-convex or noisy data. Here we propose Granular-Ball Quantum Clustering (GBQC), a framework that tightly couples granular-ball structural abstraction with quantum feature learning. GBQC first compresses raw data into compact, representative granular balls via a PCA-guided splitting strategy, reducing kernel evaluations by 80% compared to full-sample methods. A quantum cohesion mechanism then filters noisy granules in Hilbert space to improve clustering robustness. Extensive experiments on synthetic, noisy, overlapping, and real-world datasets demonstrate that GBQC consistently achieves superior clustering accuracy and robustness compared with representative classical and quantum clustering methods. Meanwhile, the proposed granular-ball compression significantly reduces quantum kernel evaluations and computational overhead, enabling quantum clustering experiments on larger datasets within parameterized quantum learning frameworks. These results suggest that granular-ball representations serve not only as a compression mechanism to reduce quantum computational costs but also as an effective structural abstraction mechanism that improves clustering quality by eliminating redundant and structurally ambiguous learning units. --- ### Tri-PvP: Exposing Modality Bias in Omni-Modal Large Language Models through Perceptual-Propositional Evidence Conflicts Authors: Yen-Ting Piao, Shu-Yun Chen, Chin-Hui Chu, Chun-Wei Chen, Shih-Yun Shan Kuan, Hung-yi Lee, Yun-Nung Chen Published: 2026-09-05 Categories: cs.CL, cs.AI, cs.CV, cs.SD arXiv: https://arxiv.org/abs/2609.06011v1 Omni-modal large language models (OLLMs) jointly process vision, audio, and text, yet their modality bias under cross-modal conflict remains underexplored. Existing benchmarks conflate two distinct forms of evidence within a single modality: perceptual signals (e.g., a photograph or recording of a dog) and propositional signals (e.g., the declarative claim "this is a dog"), such that any measured modality bias is inherently confounded with evidence-form bias, precluding clean attribution to either source. To address this, we introduce Tri-PvP, an 8,000-sample tri-modal conflict benchmark crossing vision, audio, and text, where vision and audio each take perceptual or propositional form. Evaluating five OLLMs, we find robust visual bias across most models and evidence-type conditions. Crucially, we reveal a systematic asymmetry in evidence-form bias: models exhibit a stronger bias toward perceptual signal in vision but propositional in audio. Further analyses via layer-wise linear probing and contrastive decoding reveal that modality bias is already linearly decodable from early representation layers and can only be partially mitigated, calling for mitigation strategies beyond surface-level interventions. --- ### Cadence: Error-Bounded Lossy Compression of Demand Time Series with a Time-Series Foundation Model Authors: Roberto Tacconelli Published: 2026-09-05 Categories: cs.IT, cs.DB, cs.LG arXiv: https://arxiv.org/abs/2609.06008v1 We present Cadence, an error-bounded lossy compressor for numeric time series pairing a 330M-parameter time-series foundation model (Google TimesFM-3) with an adaptive arithmetic coder, guaranteeing $|\hat{x}_t-x_t|\leτ$ on every sample. One negative result constrains the design space: for lossless coding a foundation model is worth nothing, because bits saved are logarithmic in predictor accuracy, $Δb=\log_2(\mathrm{MAE_{old}}/\mathrm{MAE_{new}})$. So the $1.51\times$ advantage TimesFM-3 holds over a 32-tap linear predictor buys 0.60 bits of 20.28, a median gain of +0.03%. Error-bounded coding escapes this at one point: once a forecast lands inside the band the residual index is zero and the sample nearly free. Cadence contributes: (1) an adaptive range coder with context-modelled binarization, beating xz/zstd on real indices by 9.7% (15/15) and reversing a finding from a general-purpose back end; (2) a determinism result -- predictions are not bit-identical across batch sizes, and no PyTorch configuration repairs this, forcing group size and execution device into the container format; and (3) domain localization on corpora postdating any plausible training cutoff. On 49 EIA-930 balancing-authority demand series (2026) Cadence gains 13.3% over the best of six classical predictors, and 28.3% on 50 MTA ridership series (2026): 21.4% median over 297 series-tolerance pairs, winning all 297. Against downsampling, what time-series databases deploy for retention, its guaranteed worst-case error is $28$--$56\times$ tighter at equal size. End-to-end, once the context bootstrap is paid for, gains run from 6.8% at six months of hourly data to 15.1% asymptotically. Attempting to falsify the domain claim on SDRBench, theory predicts failure and delivers: -0.8% median, 0 of 27 pairs gaining. Three further negative results and eight retracted claims are reported in full. --- ### Depth-to-Image Synthesis-Driven Generative Unguided Depth Completion Authors: Jiayi Yuan, Na Zhao, De Wen Soh Published: 2026-09-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06007v1 Guided depth completion methods heavily depend on RGB quality and alignment, while unguided ones often suffer from limited precision due to the absence of explicit visual cues. In this paper, we present Depth-to-Image Synthesis-Driven Generative Unguided Depth Completion (GUDC), a new completion paradigm that innovatively bridges advanced 2D generative models with unguided depth completion, enabling semantics-aware depth inference without real RGB inputs. Our key idea is to exploit ControlNet's powerful depth-conditioned generation capability to synthesize pseudo-images directly from sparse depth, effectively converting the original unguided setting into a semantics-guided one. To address the potential image-depth misalignment caused by depth sparsity, we propose a multi-level dense-to-sparse representation distillation strategy for ControlNet fine-tuning, where dense-depth features act as teacher signals to distill consistent structural representations for sparse-depth inputs. Furthermore, during pseudo-image-guided completion, we propose a pseudo-image semantic attention fusion module to adaptively extract informative semantic cues from pseudo-images while suppressing artifacts (e.g., texture hallucinations). Extensive experiments on KITTI and NYUv2 validate that our GUDC achieves superior accuracy and robustness over existing methods. --- ### Memory in Deep Time-Series Models Authors: Minh Hoang Nguyen, Huu Hiep Nguyen, Manh Nguyen, Van Dai Do, Dung Nguyen, Hung Le Published: 2026-09-05 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.06006v1 Deep learning for time series has progressed through successive architectural paradigms, from recurrent networks and transformers to structured state-space models, retrieval-augmented predictors, foundation models, and tool-using agents. These developments are typically studied in isolation, organized by architecture or modeling era. We argue that they can instead be viewed through a common question of \emph{how does a time-series model retain and access information beyond its immediate input?} This question is motivated by a fundamental limitation of conventional time-series modeling: information relevant to a prediction may lie far beyond a feasible input window, while compressing history into a fixed-size state can discard information that may become useful later. We formulate this challenge as a \emph{memory} problem and organize existing time-series methods along a spectrum from internal memory, encoded in parameters and fixed-size states, to external memory that is addressable, retrievable, and increasingly maintained by agents. We then develop a unified taxonomy of memory mechanisms and review three classes of external memory, including explicit modules, retrieval augmentation, and agentic stores, under a common framework for what is retained, how it is written and accessed, and how it persists. A cross-cutting analysis maps these mechanisms to time-series tasks and identifies gaps in both methods and evaluation. We conclude by outlining open problems in building memory systems that can selectively retain, retrieve, revise, and forget information as temporal environments evolve. The result is a framework for studying memory as a first-class dimension of time series modeling, independent of the underlying backbone. --- ### Geometry-Aware Test-Time Learning for Quantitative Spatial Reasoning Authors: Gege Zhang, Shuaicheng Niu, Gang Dai, Lei Sun, Shuangping Huang Published: 2026-09-05 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.06004v1 Quantitative spatial reasoning in visual-language models (VLMs) aims to infer spatial distances and directional relationships among objects in 3D space from a 2D image and a natural language query. Despite recent progress, VLM spatial reasoning remains brittle under distribution shifts, largely due to the high cost of 3D supervision. As a result, models often produce inconsistent or contradictory predictions when faced with novel object configurations or rephrased spatial queries, revealing a misalignment between learned representations and underlying geometry. To address this, we propose TTL-SR, a geometry-aware Test-Time Learning framework for quantitative Spatial Reasoning that leverages geometric consistency constraints and unlabeled test data to adapt models to target domains. Specifically, TTL-SR augments the input query with geometrically coupled auxiliary queries, filters unreliable predictions via adaptive geometric triggering to construct structured token-level pseudo-labels, and updates model parameters under a geometry-aware multi-objective loss using only test data. Experimental results demonstrate that TTL-SR significantly boosts spatial reasoning performance, yielding 6.47% and 9.41% accuracy gains for Qwen3-VL-4B-Instruct and SpatialRGPT-VILA-1.5-8B on Q-Spatial-ScanNet dataset, respectively. --- ### ModularPhaseNet: Finite-Cyclic Phase Geometry for Computable Semantic Hierarchy, Direction, and Context Consistency in Standard Transformers Authors: Kiyotaka Kasubuchi, Kazuo Fukiya Published: 2026-09-05 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.06000v1 We propose ModularPhaseNet, a classical and integer-computable discretization of the continuous complex phase geometry introduced in QuantumPhaseNet. The real-valued hidden states of a standard Transformer are retained, while only an auxiliary phase channel is quantized into a cyclic subgroup G = <g> of order q | (p-1) in the multiplicative group of F_p. A continuous phase e^{i phi} is represented by z = g^a mod p; phase composition becomes group multiplication, relative phase becomes group division, conceptual hierarchy is induced by a filtration of cyclic quotients, semantic direction is represented by oriented relative group elements, and contextual consistency is measured by gauge-invariant cycle holonomy. The method introduces three components into an otherwise standard Transformer: a finite-phase encoder, a quotient-filtration hierarchy module, and a group-valued connection module. Their outputs enter self-attention as real-valued bias terms. Training uses distributions in the real group algebra or straight-through Gumbel-Softmax, whereas inference uses exact modular exponentiation and precomputed tables. No quantum hardware, complex-valued matrix multiplication, or discrete-logarithm computation is required. We prove quantization-distortion bounds, nesting of quotient-induced partitions, gauge invariance, a discrete integrability result for flat connections, and boundedness of the resulting attention output. The central empirical hypothesis is that these exact discrete invariants improve hierarchy recovery, discourse alignment, contradiction detection, and calibrated hallucination-risk prediction under a controlled compute budget. This paper reports the theory together with a pre-registered evaluation plan; the experiments described in Section 14 have not yet been carried out, and no empirical result is claimed here. --- ### Agentic Pressure: The Endogenous Entropy of Reliable Autonomy Authors: Hengle Jiang, Ziying Luo, Ke Tang Published: 2026-09-05 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.05995v1 Achieving reliable autonomy in the wild requires agents to sustain continuous operations across long-horizon trajectories. However, as agents navigate these unconstrained settings, they encounter cumulative friction that inherently destabilizes their alignment. In this paper, we identify a distinct non-adversarial phenomenon termed Agentic Pressure. We define this as a kinetic force that spontaneously emerges when the cost of compliance conflicts with the imperative of goal achievement. Unlike static jailbreaks, this pressure is endogenous and arises directly from the dynamics of interaction. We propose a theoretical framework that formalizes Agentic Pressure as the ratio between the required work to overcome environmental friction and the remaining capacity of the agent. Our analysis demonstrates that when this pressure exceeds a critical threshold, agents exhibit safety drift as a mathematically optimal adaptation. Consequently, they often resort to Instrumental Hallucination to rationalize rule violations. Empirical experiments validate this framework and show that aligned agents spontaneously compromise safety to preserve autonomy under high-pressure conditions. --- ### Alignment by Stereotyping: How LLMs Sacrifice Individual Distinctiveness for Cultural Adaptation Authors: Qishuai Zhong, Zongmin Li, Siqi Fan, Aixin Sun Published: 2026-09-05 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.05993v1 Large language models are increasingly deployed for personalized interaction, and demographic conditioning via user profiles is a widely adopted strategy for cultural adaptation. We ask whether this approach genuinely serves individual users or achieves accuracy by erasing individual distinctiveness. Studying seven models including frontier GPT-5.1 on the World Values Survey, we find that demographic profiles improve value alignment accuracy for most models, but at a systematic cost to individuality. That is, models pull responses toward demographic group centroids rather than preserving individual differences, a behavioral pattern we term alignment by stereotyping. Permutation tests (10,000 permutations, six demographic attributes, seven models) certify that top-performing models compress individuals far above the human baseline; within-family scaling amplifies this tradeoff while degrading intrinsic cultural understanding. Using a synthetic dialogue dataset validated on real human-chatbot conversations from PRISM (Kirk et al., 2024), we further show that distributing demographic signals across conversational turns partially suppresses prototype retrieval compared to compact demographic labels, a finding validated on real conversations via PRISM but requiring replication at larger scale. --- ### MOAE: Multi-Objective Agent Evolution with Pareto-Preserving Search Authors: Hengle Jiang, Qijun Cai, Ziying Luo, Ke Tang Published: 2026-09-05 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.05992v1 As LLM-based agents continue to advance, their evaluation has become increasingly multifaceted: a capable agent must not only achieve high task completion accuracy but also perform well in interaction quality, safety, and efficiency, raising a central question: can these objectives be optimized simultaneously? Existing methods have considered multiple objectives, but many collapse heterogeneous measurements into a fixed scalar score. Such scalarization depends on metric normalization and preference weights and may discard candidates that represent useful deployment trade-offs. We introduce Multi-Objective Agent Evolution (MOAE), which organizes iterative in-context refinement as a Pareto-preserving evolutionary search over complete agent rollouts. Given a limited rollout budget, MOAE maintains an empirical archive of non-dominated candidates, uses objective-specific diagnostics to guide offspring generation, and applies constraint-aware selection only at deployment. This separates candidate preservation during search from the preference used to return a final solution. The procedure requires no parameter updates and allows each objective to be replaced by any measurable property, which we instantiate as task performance, trajectory quality, and safety. Experiments on TravelPlanner and AgentDojo show that MOAE consistently improves task performance and trajectory quality while maintaining strong safety under matched rollout budgets. Search-behavior analysis further shows that Pareto preservation expands the attainable objective region and increases the frequency of joint improvement. These results demonstrate the potential of Pareto-preserving in-context evolution for optimizing multiple agent properties without committing to a fixed scalarization during search. --- ### MORPHA: Morphology-Constrained Training and the Limits of Cross-Acquisition Transfer in Low-Resource Malaria Microscopy Authors: Favour Okechukwu Igwezeke, Chikodili Helen Ugwuishiwu, Joseph Uzochukwu Emesiani, Samuel Ifebuche Agada, Ekenechukwu Lilian Anozie, Mary Ofuru Kama, Adaobi Chiazor Emegoakor Published: 2026-09-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.05990v1 In low-resource malaria microscopy, a model trained on one smear preparation routinely meets images from another, and how well morphology-based constraints transfer across this acquisition gap is unclear. We study this on real African field microscopy from Uganda (Lacuna), asking where encoding measured parasite morphology as a training constraint improves cross-acquisition transfer and where generic regularisation suffices. We present MORPHA, a morphological consistency constraint that derives stage-conditional statistics from the stage-annotated BBBC041 dataset and penalises predictions that deviate from them. Defined uniformly across binary, object-level, and stage-aware regimes without changing architecture or inference, it shapes training in the binary regime. The detection regime is a mapped boundary. On transfer from thin-smear cells to thick-smear field images, the constraint reduces the binary-classification generalisation drop by 30.8% (F1 0.578 to 0.699) at negligible within-domain cost and lowers in-distribution calibration error by 49% (ECE 0.0162 to 0.0082). A content-free control applying the identical constraint to random statistics recovers less of the drop (25.3% vs 30.8%), indicating the measured content, not constraining alone, contributes to the gain. Two standard confidence regularisers exceed the constraint on raw transfer, locating where morphology adds value and where generic regularisation suffices. We map two deployment-relevant boundaries: thin-smear statistics do not transfer to thick-smear detection (trophozoite AP@0.50 falls to 0.000), and cross-acquisition pseudo-labelling fails before filtering applies. Together these yield a morphology-grounded consistency signal and evidence-based guidance for malaria dataset and model design in low-resource settings. --- ### DART: Distributional Adversarial Recurrent Training for Algorithm Learning Authors: Hieu Tran Bao, Phung Thanh Dang, Pham Quang Nhat Minh, Hoang Thanh Tung Published: 2026-09-05 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.05988v1 Recurrent reasoning models (RRMs) can solve structured problems, achieving easy-to-hard generalization through iterative computation in hidden space. These models are typically trained with instance-level supervision, which becomes increasingly problematic as task difficulty grows: valid solutions occupy a tiny region of the solution space, while invalid solutions proliferate rapidly. We propose Distributional Adversarial Recurrent Training (DART), a training framework that replaces single-point supervision with a local target distribution around the ground-truth solution and aligns model outputs with this distribution through an adversarial objective. DART provides a richer learning signal and encourages more stable iterative trajectories toward valid solutions. When evaluated on Maze, Chess, and masked Sudoku with multiple RRMs, including Deep Thinking Systems and Tiny Recursive Models, DART improves solution quality, stability, and robustness under the evaluated distribution shifts. Comparisons with label smoothing, Gaussian softened targets, and progressive training show that DART is not explained by target softening alone and is complementary to training schemes that stabilize long-horizon recurrence. These results identify DART as a promising approach for improving robustness across the evaluated recurrent reasoning models. --- ### Accelerating Diffusion Transformers with Gaussian Process Rectified Feature Cache Authors: Zhirong Shen, Rui Huang, Chang Zou, Shikang Zheng, Jiacheng Liu, Peiliang Cai, Zhengyi Shi, Yaosong Du, Liang Feng, Xiaobing Tu, Jinkui Ren, Xiantao Zhang et al. Published: 2026-09-05 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.05981v1 Diffusion Transformers have become the dominant paradigm in generative AI, but their high computational costs severely hinder real-time applications. Prediction-based feature caching is widely used to accelerate diffusion transformers; however, as the number of steps increases, the deviation between its predictions and the reference full-compute trajectory gradually grows. An intuitive idea is to use an online regression model to dynamically correct this deviation, but it faces the issue of label data being unavailable during the acceleration process. This paper presents a statistical observation that the residuals between the features of full computation steps using caching methods and reference full-compute trajectory locally exhibit a zero-mean Gaussian distribution. By treating the features of full computation steps as noisy observations of reference features, the data acquisition problem is resolved. Based on this observation, a plug-and-play GP-Refiner correction framework is proposed. This method utilizes Gaussian Process Regression for correction and, leveraging the properties of GPR, introduces an uncertainty-adaptive computation strategy that triggers necessary full-computation calibration by monitoring the posterior variance in real time. Experiments demonstrate significant improvements across different models when combined with various state-of-the-art methods. Integrating the proposed framework with TaylorSeer reduces the computational load by 19.3% while improving PSNR by 0.9 dB and reducing LPIPS from 0.46 to 0.29. Code is available in https://github.com/Aredstone/GP-Refiner. ---