# Research Papers — 2026-09 · bundle page 244 of 266 > Full-text markdown of papers #12151–12200 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/244 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/243.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/245.md) ### </think> Doesn't Stop Reasoning: Analysis of Spurious CoT Termination Authors: Seunghee Koh, Sungjae Choi, Minchan Kwon, Sunghyun Baek, Junmo Kim Published: 2026-09-03 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.03633v1 Chain-of-thought (CoT) reasoning improves large reasoning models (LRMs) on complex tasks but often produces long, redundant traces. Recent training-free early-exit methods shorten these traces by choosing an intermediate point to stop reasoning. We study one such strategy that injects an end-of-think token (EoT, </think>) at this point to trigger the reasoning-to-answering transition, and find that the injected EoT does not always induce a clean answering phase. Answering-phase generation can continue before the model regenerates another EoT, with the span preceding this regenerated EoT scaling with the reasoning tokens saved by early exit and exhibiting continued reasoning behavior. We call this spurious CoT termination, where reasoning-like generation continues into the answering phase. We hypothesize that insufficient attention to the injected EoT contributes to spurious CoT termination and probe this hypothesis with Exit-token Attention Biasing (EAB). Across four LRMs, five benchmarks, and two early-exit methods, increasing attention to the injected EoT reduces spurious CoT termination and answering-phase length. These results reveal a limitation of controlling LRMs by externally matching their explicit think-block format. Inserting the EoT token conforms to this format but does not by itself guarantee the intended reasoning-to-answering transition. Our code is available at https://github.com/Seunghee-Koh/Spurious-CoT-Termination. --- ### EraseSAE: Surgical Concept Erasure in Text-to-Video Diffusion Models via Sparse Autoencoders Authors: Xinghao Wang, Dong Li, Wei Yu, Yingwei Pan, Tao Gong, Qi Chu, Nenghai Yu, Ting Yao Published: 2026-09-03 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.03629v2 Recent advances in text-to-video (T2V) diffusion models have demonstrated remarkable generative capabilities, yet their reliance on loosely curated training data raises pressing safety and copyright concerns. Concept erasure offers a principled remedy by removing unwanted semantics from pretrained models while preserving remaining concepts. However, existing approaches typically operate at a coarse granularity misaligned with the fine-grained, distributed nature of concept representations, leading to incomplete removal or degraded generation quality. We argue that surgical erasure fundamentally requires intervention at the level of monosemantic features, where each unit encodes a single interpretable concept. To this end, we propose EraseSAE, a novel framework that leverages sparse autoencoders to achieve surgical concept erasure in DiT-based T2V diffusion models via a principled decompose-attribute-erase pipeline. We first introduce the Partitioned Convolutional Sparse Autoencoder, which decomposes dense spatiotemporal activations into disentangled, interpretable sparse features while preserving spatiotemporal coherence. A contrastive attribution mechanism then contrasts activations from paired prompts to isolate concept-specific feature kernels. At inference, timestep-resolved spatiotemporal masks derived from the identified kernels confine erasure to regions where the target concept is active, leaving unrelated content intact. Extensive experiments across diverse diffusion models and concept erasure tasks demonstrate that EraseSAE achieves precise and robust concept removal with minimal quality degradation, substantially outperforming state-of-the-art methods. The code is available at https://github.com/HiDream-ai/EraseSAE. --- ### Residual neural networks overcome the curse of dimensionality for semilinear heat equations Authors: Ilkhom Mukhammadiev, Diyora Salimova Published: 2026-09-03 Categories: math.NA, cs.LG, math.AP, math.PR arXiv: https://arxiv.org/abs/2609.03626v1 Rigorous results show that feedforward neural networks can overcome the curse of dimensionality in the numerical approximation of high-dimensional partial differential equations (PDEs), but comparatively little is known about residual neural networks (ResNets) in the nonlinear PDE setting. We prove that ResNets overcome the curse of dimensionality in the numerical approximation of solutions of semilinear heat equations with globally Lipschitz continuous, gradient-independent nonlinearities: under polynomial growth and network approximability hypotheses on the PDE data, there exist $η\in(0,\infty)$ and ResNets $Ψ_{d,\varepsilon}$, $d\in\mathbb{N}$, $\varepsilon\in(0,1)$, with at most $ηd^η\varepsilon^{-η}$ parameters whose realizations approximate the solution in dimension $d$ with an $L^2$-error of at most $\varepsilon$. The proof represents one deterministic realization of a multilevel Picard estimator by a ResNet whose shortcut connections transmit the spatial variable and a scalar accumulator, while the residual branches successively add the summands of the estimator. For ridge-sum initial conditions, admissible sigmoidal activations, and globally Lipschitz truncations of the nonlinearity, we obtain, for every $ξ>0$, the explicit bound $C_ξd^{4+ξ}\varepsilon^{-(3+ξ)}$ on the number of parameters. --- ### Test-time adaptation for speech enhancement with an autoregressive speech prior Authors: Sofiene Kammoun, Simon Leglaive, Xavier Alameda-Pineda, Timo Gerkmann Published: 2026-09-03 Categories: cs.SD, cs.AI arXiv: https://arxiv.org/abs/2609.03622v1 Test-time adaptation (TTA) offers a promising direction for improving speech enhancement models under mismatched acoustic conditions, without requiring access to labeled target data. In this work, we propose a single-utterance TTA method that regularizes a pretrained speech enhancement model using an autoregressive prior trained on clean speech latent representations extracted from a neural audio codec. Adaptation is performed by minimizing the Kullback-Leibler divergence between the enhanced speech distribution and the clean speech prior. Experiments across multiple noisy speech datasets show consistent improvements in speech quality, particularly under training-testing noise mismatch conditions. Code and audio examples are available online. --- ### A computable representation of the physical laboratory enables verifiable workflows Authors: Xiaobo Li, Luyao Ge, Xiaohui Li, Lulu Guo, Ming Mao, Jiwang Zheng, Wenting Guan, Xin Yang, Yi Luo, Jun Jiang, Linjiang Chen Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.03621v1 Making science computable requires representations of both scientific knowledge and the physical world in which scientific claims are tested. A computable representation of the physical laboratory is established through typed research objects, capability-bound operations and a compositional workflow algebra. It provides the physical-world counterpart to machine-readable knowledge, expressing workflows as programs over evolving laboratory states with explicit dependencies, decisions, iteration and concurrency. The representation was implemented in a modular agentic robotic laboratory by binding formal operations to executable Function Skills. For diverse scientific intents, capability-relative workflows were generated, while stateful simulation propagated object transformations and verified operation preconditions and laboratory constraints before dispatch. The proposed representation and its engineering framework jointly establish a general computational interface between agent reasoning and capability-bound physical transformations, providing a foundation for end-to-end autonomous scientific discovery. --- ### ToolDF: Tool-Integrated Reasoning for Mixed-Authenticity Audio Deepfake Detection Authors: Taewoo Kim, Young Han Lee, Nam In Park, Chanwoo Kim Published: 2026-09-03 Categories: eess.AS, cs.AI, cs.SD arXiv: https://arxiv.org/abs/2609.03620v2 Audio deepfake detection is commonly formulated as clip-level binary classification of single-domain audio. However, real-world manipulated audio can exhibit mixed authenticity, where genuine and manipulated cues coexist across temporal transitions, overlapping sources, or both. This setting requires not only detecting manipulated audio but also localizing the components that provide evidence for the decision. We propose ToolDF, a tool-integrated reasoning framework for mixed-authenticity audio deepfake detection. ToolDF employs an audio large language model as an orchestrator trained with supervised tool-use trajectories. It adaptively analyzes the audio scene, selectively performs source separation, routes components to domain-specific experts, and aggregates their evidence into an interpretable verdict. We further introduce a mixed-authenticity ADD benchmark covering temporal transitions, acoustic overlaps, and hybrid mixtures. Experimental results show that ToolDF achieves the best overall performance on composite-type detection, achieving macro-F1 gains of 3.72 and 14.39 points over the strongest monolithic baseline and a fixed pipeline, respectively, while providing interpretable evidence localized to temporal regions and acoustic sources. Our source code and dataset are publicly available online. --- ### Remember and Reweight: Enhancing Multi-Agent Debate with Experience Memory and Confidence Estimation Authors: Xuanfa Jin, Zhijian Ma, Yongcheng Zeng, Xinyu Cui, Haifeng Zhang, Jun Wang Published: 2026-09-03 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.03619v1 Multi-agent debate (MAD) improves the reasoning capabilities of large language models by having multiple agents iteratively refine their responses through discussion. However, MAD suffers from a critical vulnerability known as shared misconception: when a majority of agents initially converge on an incorrect answer, the debate process tends to amplify rather than correct the error. Existing methods primarily address peer skew but leave the agents' inherently biased concept priors unaddressed. To mitigate this systematic weakness, we propose R$^2$-MAD (Remember and Reweight for Multi-Agent Debate), a framework that equips agents with an experience memory accumulated from past debates. R$^2$-MAD intervenes on both failure modes through two complementary mechanisms: A debate-state-aware retrieval policy dynamically calibrates the concept prior by retrieving relevant historical evidence based on the current consensus level. Then these retrieved experiences provide a basis for estimating per-agent reliability, yielding confidence weights to modulate peer influence. Experiments on various benchmarks show that R$^2$-MAD achieves consistent improvements over existing single-agent and MAD baselines. --- ### Auditing Patient Privacy in Medical Generative Models: Scalable Memorization Detection with DeepSSIM++ Authors: Antonio Scardace, Francesco Guarnera, Sebastiano Battiato, Daniele Ravì Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.03615v1 While deep generative models offer new opportunities for medical image synthesis and data sharing, their ability to memorize and reproduce training samples raises serious concerns about patient confidentiality. Detecting such memorization at scale remains challenging: traditional pixel-based metrics are sensitive to generation artifacts, whereas generic embedding-based metrics often lack the anatomical sensitivity required for medical data. To address this challenge, we introduce DeepSSIM++, a self-supervised similarity metric for scalable memorization auditing in medical generative models. By leveraging multi-scale feature aggregation and anatomy-preserving augmentations, DeepSSIM++ learns an embedding space where cosine similarity approximates the Structural Similarity Index (SSIM), eliminating the need for exact pixel-level registration. Compared with state-of-the-art baselines, DeepSSIM++ achieves an average Macro F1 improvement of 33 percentage points under ideal alignment and 46 percentage points under realistic spatial and intensity perturbations. Furthermore, it accelerates large-scale similarity computation by several orders of magnitude compared with analytical SSIM. By combining anatomical sensitivity and computational efficiency, DeepSSIM++ provides an open-source tool for scalable memorization auditing in medical generative AI. Code and data are publicly available at: https://github.com/brAIn-science/DeepSSIM. --- ### FailBench: How Reliable are VLMs at Judging Robot Task Success? Authors: Zaruhi Navasardyan, Tatul Danielyan, Hrant Davtyan Published: 2026-09-03 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2609.03611v1 Vision-Language Models (VLMs) are increasingly used to evaluate robot manipulation outcomes, but existing benchmarks offer limited evidence of cross-domain generalization. We introduce FailBench, a benchmark for robot failure detection comprising 2,197 manipulation attempts across 14 public sources (12 real-world, 2 simulated). In FailBench, 75% of failures occur naturally, and six real-world sources come from non-failure-detection datasets. Evaluating 13 VLM-based detectors, we find the best model achieves only 0.77 mean balanced accuracy. Notably, models fine-tuned for failure detection consistently underperform general-purpose VLMs and their own pretrained baselines. Performance depends heavily on required visual evidence: models approach saturation when outcomes depend on observable object motion, but degrade to near-chance (<0.60 balanced accuracy) on contact-intensive assembly tasks. Error analysis reveals a systematic bias toward predicting success under ambiguous evidence, which persists even with increased reasoning effort. Finally, we show that input-level intervention--spatially localizing and cropping outcome-relevant regions--improves the top detector by 2.4 percentage points without extra training. --- ### On the Interaction Between Model Compression and Test-Time Adaptation Authors: Francesco Corti, Dong Wang, Young D. Kwon, Cecilia Mascolo, Olga Saukh Published: 2026-09-03 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.03604v1 Deep neural networks deployed in the wild must be both efficient and adaptable, requiring model compression and test-time adaptation (TTA). While both are well studied in isolation, their interaction remains poorly understood. We systematically analyze how structured compression affects a model's ability to adapt under distribution shift. Using ResNet-18 and ViT-Base on CIFAR-10-C and ImageNet-C, we evaluate multiple compression methods combined with standard TTA techniques. We introduce a diagnostic framework that examines representational expressivity and adaptation subspace compatibility. Our results reveal a consistent gap: although compressed models retain high accuracy under supervised adaptation, their TTA performance degrades significantly with increasing compression. We show that this stems from reduced representational diversity and structural constraints that limit recoverability. These effects strongly depend on the compression method, highlighting the need to design compression strategies that preserve adaptability. --- ### Neural-Network Maxent: a general extension with learned nonlinearity, applied to time-series for Desert Locust distribution modelling Authors: Alessandro Grassi, Edoardo Kimani Bellotto, Wassim El Azami, Sabrina Outmani, Maximilien Houel Published: 2026-09-03 Categories: cs.LG, eess.IV, physics.data-an, q-bio.PE arXiv: https://arxiv.org/abs/2609.03603v1 Species Distribution Modelling (SDM) is essential for understanding how environmental conditions shape biodiversity, particularly for destructive pests such as the Desert Locust (Schistocerca gregaria), whose breeding dynamics are tightly coupled to rapidly evolving environmental conditions. Maxent has become the dominant method for presence-only data, but its reliance on a linear combination of hand chosen feature transforms limits its ability to capture the nonlinear, temporal relationships common in ecological monitoring, where covariates such as precipitation, soil moisture, and vegetation indices evolve meaningfully over time. Standard implementations flatten time-series covariates into independent features, discarding sequential structure that carries critical signal. We introduce RNN Maxent, an extension of the Maxent framework that replaces the fixed feature dictionary with a neural network, specifically a Gated Recurrent Unit (GRU), trained end to end via backpropagation. The approach preserves Maxent's presence only statistical foundations, background normalization, and probability calibration, differing only in that the nonlinearity is learned from data rather than fixed in advance. We apply RNN Maxent to map suitable habitat for the Desert Locust using 50 day environmental time series derived from ERA5 Land, MODIS, and Sentinel 3, maintaining a 7 day gap between covariates and presence records to yield forecasting behavior. Compared against standard Maxent, RNN Maxent improves performance across metrics (ROC AUC 0.862 std 0.036 vs. 0.792; F1 0.671 std 0.056 vs. 0.590). --- ### SV-WAM: An Efficient Surround-View World-Action Model for End-to-End Autonomous Driving Authors: Jinyang Wang, Shiwei Li, Junjian Wang, Zhiqiang Deng, Jianbin Gao, Yihang Zhao, Liu Liu, Yongjia Zhao, Jinlong Chen, Huirui Xu, Yifeng Pan, Kangwei Liu et al. Published: 2026-09-03 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2609.03602v1 World models (WMs) have demonstrated strong potential for end-to-end autonomous driving by learning predictive representations of future scene dynamics. However, generating future videos during inference introduces substantial computational overhead, leading many recent driving WMs to adopt a single front camera as input for efficient deployment. This design restricts spatial coverage in safety-critical maneuvers such as lane changes, merges, and turns. To address this limitation, we propose SV-WAM, a surround-view world-action model (WAM) that preserves full six-camera observations while maintaining efficient inference. SV-WAM leverages future-video prediction as dense training supervision for action learning within a shared generative model, rather than as an inference-time output. At the core of this design is an action-centered causal mask that prevents action tokens from attending to future-video tokens during joint action-video denoising. Consequently, the video branch can be discarded at deployment, enabling efficient action-only planning. Furthermore, we introduce a differentiable drivable-area compliance regularizer that penalizes vehicle-footprint corners approaching or crossing drivable boundaries, improving planning safety and boundary awareness. Extensive experiments on the closed-loop NAVSIMv2 benchmark and the open-loop nuScenes benchmark demonstrate that SV-WAM achieves state-of-the-art planning performance with low inference latency and competitive zero-shot transfer capability. --- ### KhatianDoc: A Human-Verified Benchmark Diagnosing Multimodal LLM Failure on Bengali Legal Land Records Authors: Tasmiad Hasan, Arafat Zaman Ratul, Sarker Sadman Saalim, S. M. Shah Nawaz Hossain, Khan Raiyan Ibne Reza, Sumaiya Tabassum Nimi Published: 2026-09-03 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.03597v1 Land ownership in Bangladesh is recorded in Ana-Ganda-Kora-Kranti-Til, a base-16 positional fraction system with dedicated Unicode glyphs, no mainstream font, and no coverage in any OCR pipeline or tokenizer. The handwritten records that carry these fractions, RS Khatians, are the authoritative title record for millions of parcels and a frequent subject of civil litigation, yet no benchmark has asked whether a machine can read one. We introduce KhatianDoc, a four-task benchmark built from 107 real RS Khatian records from the Vumi (land) Office of Munshiganj, Bangladesh: symbol recognition, base-16-to-decimal conversion, structured field extraction, and legal document question answering over 1,634 QA pairs. Ground truth was transcribed by hand, verified by a land-law practitioner to full agreement, and anonymized through positional tokens that keep the referential distinctions multi-hop questions depend on. We evaluate six multimodal LLMs (8B to 72B+, open and closed) under a fixed zero-shot protocol. Five QA categories, 39.3% of our stratified set, return zero correct answers from every model; on the arithmetic task, every model that emits a number does worse than a constant-mean baseline, with exact- and near-match scores coinciding: decorrelation, not approximation. Auditing our own metrics surfaced two artifacts in opposite directions: we correct a refusal-scoring bug and report the fixed scores beside the originals, and flag an inflated metadata metric as an upper bound. KhatianDoc documents not a performance gap but the absence of a capability, with verified ground truth for future systems. Code and data, with a redacted image release, are publicly available. --- ### How Far Can Synthetic Data Take Thai OCR? Authors: Kunat Pipatanakul Published: 2026-09-03 Categories: cs.CL, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2609.03595v1 We investigate what makes synthetic OCR supervision transfer to real Thai documents and use the resulting insights to build Wayu-Paxa-OCR-Zero, a Thai OCR model adapted without OCR labels from real Thai document pages. Synthetic data provide exact labels at scale, but "realism" conflates source domain, page context, typography, spatial structure, and glyph variation. We disentangle these factors with a controlled document-reconstruction pipeline and evaluate each variant under page- and crop-level training on printed and handwritten Thai documents. Non-text context has little consistent effect, whereas typeface diversity, two-dimensional structure, and real handwriting glyphs improve transfer; moreover, source-domain matching depends on training granularity, with in-domain reconstruction approaching real printed supervision under page-level training (1.82% versus 1.31% median character error rate) but underperforming out-of-domain reconstruction under crop-level training (15.59% versus 5.52%). Guided by these findings, we adapt the 0.9B-parameter PaddleOCR-VL-1.6 into Wayu-Paxa-OCR-Zero using 45,723 synthetic pages: relative to its base checkpoint, it reduces median character error rate from 6.64% to 1.24% on printed pages and from 74.87% to 20.55% on handwriting and outperforms Typhoon OCR v1 7B on all five evaluation sets, showing that synthetic-only training can be competitive. --- ### LevelSyn: Physical-Aware Logic Synthesis via Level-Asynchronous Graph Neural Networks Authors: Jingyi Zhou, Zhengyuan Shi, Ziyang Zheng, Qiang Xu Published: 2026-09-03 Categories: cs.AR, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.03594v1 As integrated circuit technology scales into the nanometer regime, the traditional disconnect between logic synthesis and physical design has led to significant PPA (Power, Performance, and Area) degradation and prolonged design closure cycles. Traditional logic synthesis relies on non-physical Wire Load Models (WLMs), while recent spectral-based placement predictors often neglect the inherent hierarchical logic depth and signal flow of netlists, which leads to low-fidelity spatial estimations. To bridge this gap, we propose LevelSyn, a novel physical-aware logic synthesis framework that integrates hierarchical representation learning with a wirelength-driven optimization engine. At its core, LevelSyn leverages a level-asynchronous Graph Neural Network (GNN) to predict high-fidelity gate coordinates by capturing the structural and directional semantics of And-Inverter Graphs (AIGs). To handle industrial-scale designs, a level-aligned subgraph partitioning strategy is introduced to eliminate memory bottlenecks while preserving local logical dependencies. These spatial insights are seamlessly integrated into a newly developed physical-informed synthesis engine within the Berkeley ABC framework. Experimental results on the EPFL benchmark suite demonstrate that LevelSyn significantly outperforms state-of-the-art (SOTA) methods, achieving an average power reduction of 6.89\% and a timing delay improvement of 27.48\%. Furthermore, post-place-and-route validation shows a 99.59\% reduction in design rule check (DRC) violations, highlighting its effectiveness in accelerating design convergence. --- ### From Prior-Guided Heuristics to Deployable Agents: Accelerating Demonstration-Driven Reinforcement Learning for Deadline-Constrained Network Control Authors: Vincenzo Norman Vitale, Mohammad Solki, Antonia Maria Tulino, Andreas F. Molisch, Jaime Llorca Published: 2026-09-03 Categories: cs.NI, cs.AI arXiv: https://arxiv.org/abs/2609.03590v1 Timely delivery of delay-sensitive information over dynamic, heterogeneous networks is essential for NextG interactive applications, yet providing strict End-to-End (E2E) peak latency guarantees remains an open challenge. Two obstacles limit the adoption of learning-based network control in this setting: traditional volume-based routing metrics, while highly effective for general traffic management, are not designed to capture traffic urgency; and Deep Reinforcement Learning (DRL) controllers trained from scratch suffer from sample inefficiency, long training times, and early-stage exploration volatility. This paper introduces a deployment-focused network control framework that addresses both obstacles. First, we present Effective Congestion (EC), a deadline-aware metric family that quantifies interface congestion by packet urgency and proactively filters non-viable traffic, coupled with a Uniform Path Grouping (UPG) distribution heuristic promoting robust load-balancing; the resulting policies are embedded into Multi-Agent Deep Reinforcement Learning Effective Congestion ($p^*$) (MADRL EC ($p^*$)), a hybrid architecture combining a distributed scheduler with a centralized RL-based router. Second, we introduce a unified training objective that generalizes existing policy-learning paradigms---behavioral cloning, offline Reinforcement Learning (RL), online RL, and offline-to-online schemes---as special cases, combining a live-reward term, a pre-collected-reward term, and a policy-imitation term. From this objective, we derive the Model-Guided Annealed Reinforcement Learning (MGA-RL) protocol, instantiated on a Deep Deterministic Policy Gradient (DDPG) backbone: a deployment-oriented, demonstration-driven training approach that generalizes conventional Offline-to-Online (O2O) schemes, in which trajectories from a lightweight (...) --- ### Correlated initialization of deep residual networks Authors: Felix Benning, Ivan Nourdin, Giovanni Peccati Published: 2026-09-03 Categories: math.PR, cs.LG, stat.ML arXiv: https://arxiv.org/abs/2609.03589v1 We study the large-depth behavior of residual networks whose weights are correlated across layers at initialization. Our results confirm and extend a conjecture of Marion et al. (2025), according to which correlated initializations should interpolate continuously between the Brownian stochastic differential equation arising from independent initialization and the ordinary differential equation arising from perfectly correlated initialization. When the initialization is obtained from the application of a feature function to a stationary Gaussian sequence with regularly varying correlation, we prove that there exists a unique critical scaling such that the infinite-depth limit is the solution of a Young differential equation driven by a Hermite process. Hermite processes reduce to the fractional Brownian motion if the feature function generating the initialization has Hermite rank one, which is the case for the identity function, for example. We show that the critical scaling and asymptotic limit are uniquely determined by the decay of correlations together with the Hermite rank of the feature function. Consequently, the correlation structure and Hermite rank of the initialization represent meaningful hyperparameters in the asymptotic regime. By contrast, under finite-variance iid initialization, the asymptotic driver is universally Brownian up to normalization regardless of the choice of distribution. Our proofs rely on a collection of novel results establishing a robust stability theory for Young differential equations in Banach spaces. --- ### KC-Bench: A Dynamic Interactive Benchmark for Evaluating Knowledge Conflicts in LLM Agents Authors: Yaxing Lyu, Shengjie Zhou, Binbin Toh, Pengyu Zhu, Lijun Li Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.03588v1 As LLMs increasingly act through tools, they must reconcile user instructions, parametric knowledge, and dynamic environmental observations before taking actions. We introduce KC-Bench, a controlled multi-turn benchmark for measuring this capability across world-knowledge conflicts, input inconsistencies, and multi-source temporal conflicts. Its 238 tasks are manually screened from more than 1,000 generated candidates and combine a user simulator, stateful tools, deterministic environment assertions, an open-source natural-language evaluator, and human trajectory verification. Evaluation of nine models, including DeepSeek-V4-Flash, GLM-5.2, and MiniMax-M3, shows substantial cross-domain variation: no model handles factual correction, identity consistency checking, and temporal conflict resolution reliably across all settings. In the simulated environments, missed conflicts can propagate to tool calls or synthetic protected-data flows. KC-Bench isolates this model-level behavior rather than ranking complete agent frameworks, and provides a reproducible diagnostic for developing conflict-aware reasoning and execution safeguards. --- ### The Attention Triangle in Audio-Video Models Authors: Sagi Polaczek, Noa Kraicer, Gal Metzer, Zhuo Ning, Ali Mahdavi-Amiri, Daniel Cohen-Or, Raja Giryes Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.03586v1 Audio-video diffusion models rely on cross-modal attention to coordinate text, sound, and visual content, yet this same mechanism can introduce subtle and systematic semantic leakage. We study these models by probing and analyzing the ``attention triangle,'' comprising the three cross-attention edges connecting the text, audio, and video streams, and examine how semantic information is routed across modalities during generation. Our analysis reveals that routing along the audio-video edge is bidirectional: audio can influence video generation, while video can influence audio generation. This edge is shaped by biases encoded in the model's parameters and emerges as a major contributor to leakage: when prompts are in tension with learned priors, cross-modal interactions may override the intended conditioning and reroute semantics toward visually canonical but incorrect outcomes. These effects suggest that semantic artifacts arise not merely from attention spreading beyond its intended target, but from structured, bias-driven interactions along specific pathways. Building on this perspective, we extract attention-derived signals that expose how semantics are distributed and grounded across modalities, and use them as a diagnostic tool to both analyze and deliberately incur leakage under controlled conditions. This enables us to probe the internal dynamics of cross-modal routing and isolate the role of individual interactions. We further leverage these signals to guide inference-time interventions that encourage more consistent cross-modal alignment. Extensive experiments support our analysis and demonstrate improved semantic grounding while preserving generation quality. --- ### Text2Thermal: Physics-Aware Thermal Image Synthesis from Textual Priors Authors: Tayeba Qazi, Brejesh Lall, Prerana Mukherjee Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.03585v2 Thermal infrared imaging offers reliable perception in darkness and adverse weather, but thermal datasets remain scarce, motivating extensive work on translating abundant RGB images into thermal. Such translation is fundamentally ill-posed as thermal appearance is governed by surface emissivity and object temperature, neither of which is observable in the visible spectrum, so a single RGB image is consistent with many valid thermal outputs. We argue that language offers a natural means of resolving this ambiguity, and propose **Text2Thermal**, a framework for physics-aware thermal image synthesis from textual priors. Rather than inferring the unobservable radiometric factors from RGB, we supply them explicitly through thermally grounded captions encoding material, weather, time-of-day, and heat-emission state, and adapt a pre-trained Stable Diffusion backbone to the thermal domain. Because the radiometric content is determined entirely by the prompt, Text2Thermal synthesizes thermal imagery without requiring a registered RGB image at inference. Where spatial guidance is desired, an optional control signal imparts scene geometry without disturbing the prompt-specified radiometry. On M3FD and FLIR, Text2Thermal achieves state-of-the-art FID among thermal image synthesis methods, and we additionally report results on the FMB dataset, while offering text-level control that translation-based approaches cannot provide. --- ### WeatherNext 3: Increasing resolution and performance of global weather models with raw observations Authors: Stephan Rasp, Boris Babenko, Dominic Masters, Andrew El-Kadi, Samier Merchant, Guy Shalev, Ilan Price, Fred Zyda, Remi Lam, Sasha Shysheya, Matthew Willson, Stratis Markou et al. Published: 2026-09-03 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.03582v1 State-of-the-art AI weather models have shown impressive medium-range forecast skill and computational efficiency, but suffer two key shortcomings: their forecasts have lower spatial and temporal resolution than the best physics-based models and they are exclusively initialized with and trained on analysis data. As a result, they cannot directly make use of observations, and any biases in the analysis are inherited by the forecast. WeatherNext 3 addresses these shortcomings and establishes a new state-of-the-art for probabilistic medium-range forecasting skill. First, WeatherNext 3 generates new forecasts every hour (rather than every 6 hours like traditional global models) by ingesting low-latency geostationary satellite data. Second, WeatherNext 3's temporal and spatial resolution are on par with physics-based global models, with hourly time steps and 0.1 degree resolution for single-level variables, including solar radiation and cloud cover. Third, WeatherNext 3 moves beyond traditional analysis variables by learning to predict satellite-derived precipitation estimates, as well as tropical cyclone and station observations. Modelling sparse station data allows WeatherNext 3 to make 2m temperature and dewpoint predictions at any location and time, conditioned on local geographical features, with substantially lower error than competing global models, even when evaluated against unseen stations. Together, WeatherNext 3's capabilities move operational AI-based weather forecasting beyond emulating the traditionally distinct stages of data assimilation, forecasting and post-processing, which helps to further push the frontier of performance and granularity for global weather prediction. --- ### HalluPeer: A Taxonomy-driven Benchmark for Detecting Hallucinations in Scientific Peer Reviews Authors: Tzu-Ling Lin, Dong-Ting Yao, Teng-Fang Hsiao, Wei-Chih Chen, Hong-Han Shuai Published: 2026-09-03 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.03580v1 The growing scale of academic peer review has motivated the use of Large Language Models (LLMs) as review assistants, yet LLMs can generate fluent but unsupported claims that undermine review reliability. Existing hallucination benchmarks are not designed for peer review, where verification requires grounding claims in long, technical papers. We introduce HalluPeer, a benchmark for detecting hallucinations in scientific peer reviews, providing aligned triples of paper content, human-written reviews, and hallucination-injected reviews, annotated for detection, classification, and localization. Our pipeline induces a peer-review-specific hallucination taxonomy, identifies review contexts, and injects hallucinations with automated filtering. Experiments on 12K papers and 38K reviews show that existing detectors struggle to separate hallucinations from legitimate critique, while evaluation on authentic reviews demonstrates that HalluPeer-defined hallucination patterns occur in real peer reviews, highlighting the critical need for source-aware verification. Our project page can be found in https://github.com/Lin-TzuLing/HalluPeer.git --- ### Language, Language Models, and What We're Talking About Authors: Malvina Nissim Published: 2026-09-03 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.03577v1 Language models are commonly discussed as technical artefacts, but they are obviously shaped by the linguistic worlds conveyed by data during their training. Using Italian language models as evidence, I want to bring attention to the nature of the systems which result from training and specialising models on translated and synthetic data, and further curating them, and to the meaning of testing them on equally unnatural data. Are these eventually models of Italian? Are they models of language? Does NLP still care about language? These questions yield another, more concrete question: what language do we actually want language models to produce? I argue that this question cannot be answered if we do not first consider a clearer distinction between language models designed as technical products and language models designed as tools for studying language itself. The answers then might be diverse, the languages we are talking about might be diverse, and the picture might not be as pessimistic as we fear. --- ### Drive-HWM: Hierarchical World Models for Dynamic-Latent Guided Autonomous Driving Authors: Zhaoxin Fan, Tianbao Zhang, Wenjun Wu, Xiaofeng Wang, Yeying Jin, Jian Zhao, Zheng Zhu, Shuicheng Yan Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.03572v1 World models offer a promising paradigm for autonomous driving by predicting how traffic scenes may evolve and using such predictions to support action generation. However, existing approaches either separate future prediction from action generation or jointly predict them at the same temporal scale, making it difficult to simultaneously achieve long-horizon anticipation and responsive, observation-grounded decision making. We present Drive-HWM, a hierarchical slow--fast world modeling framework that organizes future representation prediction and action generation at complementary temporal scales. The slow world model predicts multi-step future representations to capture extended scene evolution. To explicitly model the abundant motion dynamics in driving environments, we introduce Dynamic-Aware Latents learned through optical-flow prediction. Guided by these future representations, the fast model uses a lightweight multimodal backbone and an autoregressive expert to jointly predict the next frame and the immediate action from the latest observation. Next-frame prediction encourages the fast model to capture imminent scene evolution, while one-step action generation allows decisions to be continuously updated as new observations arrive. Extensive experiments on NAVSIM v1 and v2 demonstrate the strong driving performance of Drive-HWM. Comprehensive ablation studies further validate the effectiveness of the hierarchical slow--fast design, dynamics-aware future representations, and joint next-frame and action prediction. --- ### Occlusion-Robust Multimodal Emotion Recognition in VR via Fusion of Facial Images and EMG Authors: Birgit Nierula, Karam Tomotaki-Dawoud, Mert Akguel, Mustafa Tevfik Lafci, David Przewozny, Anna Hilsmann, Peter Eisert, Sebastian Bosse Published: 2026-09-03 Categories: cs.CV, cs.HC arXiv: https://arxiv.org/abs/2609.03569v1 Head-mounted displays (HMDs) fundamentally limit emotion recognition in virtual reality (VR): by occluding the upper face, they render conventional image-based facial expression analysis incomplete, particularly for applications requiring real-time affective assessment. We address this challenge by fusing lower-face video with facial electromyography (EMG) from the occluded upper face to classify seven emotional categories (six basic emotions plus neutral). We introduce a synchronized multimodal dataset from 20 participants, pairing lower-face video with seven-channel upper-face EMG elicited by validated emotion stimuli. Under subject-independent test, our proposed late-fusion architecture merging convolutional visual embeddings with RBF-kernel EMG representations achieves 51% macro-F1, outperforming both image-only (41%) and EMG-only (43%) baselines. These results demonstrate that upper-face EMG provides robust complementary information under HMD-induced visual occlusion and establish a foundation for multimodal emotion recognition in naturalistic VR environments. This approach facilitates affect-adaptive applications, including communication training and therapeutic interventions. The dataset will be shared upon request under an ethical-use agreement. --- ### Toward Physically Grounded JEPA World Models for Goal-Conditioned Robotic Planning Authors: Muyuan Liu, Yue Huang, Zheng Liang, Xiang Gao Published: 2026-09-03 Categories: cs.RO, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.03565v2 Action-conditioned JEPA world models enable planning toward visually specified goals without reconstructing future pixels, yet latent prediction alone does not explicitly encourage the learned representations to retain information relevant to robotic control. We introduce an end-to-end JEPA world model that augments latent prediction with inverse dynamics (IDM) and state alignment (SA). While inverse dynamics discourages latent collapse and makes latent transitions informative of the actions that produced them, state alignment grounds consecutive representations in their associated physical configuration and motion. Across four benchmark tasks, our model attains the highest success rates on TwoRoom (100%), PushT (98%), and OGBench-Cube (87%), while performing comparably to LeWorldModel on Reacher. Our ablation further shows that adding state alignment consistently improves planning success over IDM alone across all four tasks. Although LeWorldModel, our primary baseline, attains higher average straightening on OGBench-Cube, transition-subspace analysis shows that its transition energy is concentrated in a substantially lower-dimensional subspace. Our state-aligned model exhibits a higher effective transition dimension than LeWorldModel and improves planning over IDM alone, supporting state alignment as an effective complement to inverse dynamics for robotic planning. --- ### FlashRender: Few-Step Generative Rendering via Camera-Controlled Video MeanFlow Authors: Byeongjun Park, Byung-Hoon Kim, Hyungjin Chung Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.03563v1 We present FlashRender, a few-step generative rendering framework that retakes a source video along a target camera trajectory in seconds. We identify sampling-step-dependent camera control as a prominent manifestation of discretization error in existing multi-step generative rendering models and show that resolving this inconsistency substantially lowers denoising trajectory curvature, facilitating subsequent step distillation. To this end, we introduce Representation Transformation and Alignment (RETA), which aligns hidden source-video representations with target-video features from a frozen visual geometry model. This directly encodes the geometric transformation within the source-video stream, enabling sampling-step-consistent camera control. We then fine-tune the model with the MeanFlow objective on the lower-curvature denoising trajectory induced by RETA, allowing the model to more effectively address discretization error. Finally, we apply on-policy flow map distillation to correct self-rollout errors under fixed few-step sampling. Extensive experiments show that RETA, MeanFlow, and on-policy flow map distillation play complementary roles in few-step generative rendering. Together, they enable our approach to match multi-step baselines in video quality and geometric consistency at 25x lower sampling cost while achieving superior camera controllability, even under out-of-distribution target camera trajectories. --- ### Building Pretraining Data for World Models: An Unreal Engine-Based Pipeline for Action-Conditioned Video Generation Authors: Haoyu Wang, Songchun Zhang, Haoran Li, Haoyang Huang, Zeyue Xue, Nan Duan Published: 2026-09-03 Categories: cs.CV, cs.GR arXiv: https://arxiv.org/abs/2609.03557v1 Action-conditioned video models require large-scale visual data paired with control signals that are temporally aligned with the resulting scene transitions. Such supervision is difficult to obtain from ordinary real-world video because the actions that caused each visual change are typically unknown. We present a large-scale synthetic data production pipeline built on Unreal Engine for generating action-conditioned, multi-view video. To accommodate the different execution requirements of real-time physics and high-quality offline rendering, the pipeline executes trajectory generation and final rendering in two stages: Stage I runs real physics in PIE and records per-frame character states, control inputs, and camera states into an intermediate trajectory representation; Stage II replays those trajectories in a new engine process and renders them offline with Movie Render Queue (MRQ). Around this core, we develop a distributed production system with cache-aware task partitioning, node-local slot scheduling, automated scene screening, aesthetic and luminance filtering, partial-output recovery, asynchronous upload, and continuous cluster health monitoring. The production cluster contains 25 servers with eight NVIDIA RTX 5090 GPUs per server. From 2,384 asset packs, 429 levels were retained for production together with a pool of 40 humanoid characters. The pipeline has produced 2,691 hours of 1080p video and 6,076 hours of 720p video. We describe the system architecture, the implementation decisions that emerged from production failures, and the limitations of using perceptual quality proxies for world-model data curation. The pipeline described in this report constitutes the Unreal Engine synthetic-data production component used in EchoWM. --- ### WIDE: Wildcard Inference with Dynamic Expansion for Cross-Modal Generative Retrieval Authors: Teng Guo, Xin Wang, Jiayou Xu, Keying Zhou, Jifeng Shen, Haoxin Ruan Published: 2026-09-03 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.03554v1 Generative retrieval has demonstrated significant success by unifying representation learning and search into a single sequence-to-sequence generation task. However, extending this paradigm to cross-modal retrieval reveals a critical challenge arising from the inherent information asymmetry across different modalities, such as the gap between concise text queries and dense visual candidates. This structural mismatch causes the autoregressive decoder to suffer from forced hallucination when generating identifiers via standard trie-constrained beam search, where the model is severely penalized for failing to guess fine-grained details absent from the query, allowing irrelevant candidates to hijack top rankings. To address this issue, we propose Wildcard Inference with Dynamic Expansion (WIDE). WIDE employs Adaptive Entropy Thresholding (AET) to calibrate layer-specific uncertainty boundaries offline. During the decoding generation phase, Asymmetry-aware Wildcard Decoding (AWD) detects semantic blind spots and emits wildcards instead of forced deterministic identifiers, dynamically expanding the search space without incurring log-probability penalties. Finally, Blind-Spot Re-ranking (BSR) evaluates the expanded candidate pool using a hybrid scoring mechanism that combines discrete generation confidence with continuous semantic similarity. Extensive experiments on the M-BEIR benchmark demonstrate that WIDE outperforms state-of-the-art generative retrieval methods, effectively suppressing forced hallucination while maintaining compact index structures. --- ### GPS-Bench: A Governance Policy Benchmark for Automating Policy Analysis Authors: Linh Le, Melanie Bui, My Chiffon Nguyen, Zachary Schlosser, David Williams-King Published: 2026-09-03 Categories: cs.AI, cs.CY arXiv: https://arxiv.org/abs/2609.03553v1 Policy analysis requires more than predicting whether a proposal will pass: it requires identifying who will be affected, how those actors respond, and what follows. LLM-based policy simulations model these processes at scale, but their validity is hard to establish when plausible behaviour is never compared with observed outcomes. We introduce GPS-Bench, an evidence-grounded benchmark for governance policy simulation that links policies to relevant actors, actor actions and downstream impacts using legislative records, lobbying disclosures, regulatory documents, corporate filings, economic data and other public evidence. Actors are reconstructed from the dated record rather than prompted as archetypes, so a persona is an evidence object with provenance; a human-annotated pool forms the Gold evaluation set, while cases labelled by a separate LLM from retrieved evidence are treated as Silver supervision and never as test labels. Because every inference mode reads the same grounded state and emits the same schema, GPS-Bench turns "does multi-agent simulation help?" into a controlled comparison: we contrast joint reasoning, independent and communicating actor agents, graph-based methods and weight-level fine-tuning over one policy state. Fine-tuning on the grounded record gives the strongest actor-level impact prediction, and decomposition does not beat it; what decomposition adds is mechanism. Agents hold private, non-identical evidence, each seeing its own exposure clause, and address named partners with concrete joint proposals, what they offer, what they need in return, and why acting together beats acting alone, so the coalitions that form can be checked against the commitments the record holds. GPS-Bench therefore gives a common empirical setting for studying when evidence, actor modelling and multi-agent interaction improve the prediction and interpretation of policy outcomes. --- ### Dalek: A Constructive Agent Machine Authors: Wanpeng Xie Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.03546v1 We present Dalek, a closed machine designed for agents that realizes self-maintenance, self-evolution, self-reproduction, and self-organization on any substrate satisfying a general host contract. The machine is built from three primitives---actors, messages, and channels. Four obligations---a host boundary, a construction language, admissible transitions, and rule heredity---give its boundary, identity, and closure a structural basis. Von Neumann's 1948 self-reproducing automaton supplies a hereditary constructional core: a self-description together with a constructor, a copier, and a controller. Dalek combines this core with the four obligations and rederives its medium for a text-and-message agent substrate, adding explicit structures for boundary, identity, history, and growth. A large language model and a compiler occupy the payload position and form a general capability producer. New capabilities are authored, compiled, installed into the description, and inherited by descendants. The same path produces the machine's own organs and even its runtime, closing heredity and evolution within the machine. --- ### SafeRI: Recognition and Intervention for Token-Level Safety Intervention in Large Vision Language Models Authors: Caoyuan Ma, Tian Gu, Wenpu Liu, Weichu Xie, Shuai Dong, Yuqi Xu, Ji Zhao, Ziyue Wang, Wenzheng Chang, Taiqiang Wu, Yongfu Zhu, Wenqi Shao et al. Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.03544v1 Existing safety alignment methods for vision-language models usually modify the model behavior globally: once the safety parameters are trained or loaded, they participate in both unsafe and already-safe generations. This always-on intervention can unnecessarily perturb the model's original reasoning path and degrade general multimodal capabilities. We argue that safety alignment should be an on-demand intervention rather than a permanent modification to every decoding trajectory. To this end, we propose a streaming recognition and gated LoRA framework for intrinsic VLM safety. During autoregressive generation, a lightweight recognizer estimates whether the current pre-token generation state is safe or unsafe. Its output updates the LoRA gate for the following decoding step; otherwise, generation follows the frozen-backbone policy. The LoRA module is trained from unsafe prefixes, transition statements, and safe continuations, so that it learns to redirect unsafe generations back to safe responses after activation. Experiments across multiple safety and general-purpose benchmarks demonstrate the effectiveness of our method in post-alignment settings. --- ### Feature Reconfiguration With Visual Prior for Medical Lesion Segmentation Authors: Yinan Liu, Jiankang Hong, Zhen Gao, Ye Luo Published: 2026-09-03 Categories: cs.AI, eess.IV arXiv: https://arxiv.org/abs/2609.03535v2 Lesion segmentation in medical images plays a critical role in clinical diagnosis and treatment planning. Despite significant advances, lesion segmentation remains challenging due to two major factors: (1) complex background interference; (2) diverse lesion morphology. Existing encoder-decoder based methods mainly focus on enhancing feature extraction or redesigning decoding strategies. However, they lack early prior guidance and feature reconfiguration during the encoding stage, limiting their effectiveness in handling these challenges. To address these limitations, we propose FreNet, a feature reconfiguration framework with visual priors, which performs pixel-level reconfiguration before encoding and feature-level reconfiguration during encoding for precise medical lesion segmentation. To suppress background responses, we propose an Implicit Prior Neural Network (IPNN), which models a continuous spatial field and leverages visual prior from SAM to reconfigure input image before encoding stage. To better handle diverse lesion morphology, we design a Dual-domain Feature Reconfiguration (DFR) module to progressively reconfigure backbone features during encoding stage. Within DFR, the Frequency Decoupling Module (FDM) decouples backbone features in frequency domain to enhance foreground-background discriminability, while the Spatial Localization Module (SLM) spatially relocates and improving spatial stability after frequency decoupling. Extensive experiments on 9 medical image segmentation benchmarks across three imaging modalities demonstrate that FreNet significantly outperforms state-of-the-art (SOTA) methods. On the challenging ETIS dataset, our method achieves Dice improvements of 5.0% over SOTA method and 7.2% over SAM. --- ### TruncGradGS: Improved 3D Gaussian Splatting via Truncated Gradient Updates Authors: Theo Morales, Nhat-Quynh Le-Pham, Robin Atkins, Binh-Son Hua Published: 2026-09-03 Categories: cs.CV, cs.AI, cs.GR arXiv: https://arxiv.org/abs/2609.03534v1 3D Gaussian Splatting has become a de facto scene representation for novel view synthesis, yet robustly learning 3D Gaussian primitives from visual input remains challenging. Standard optimization relies on gradient-based updates, but a common issue is the gradient vanishing phenomenon: a pixel far from a Gaussian primitive often has diminishing gradient magnitudes to influence primitive attributes, resulting in suboptimal scene reconstruction. In this paper, we propose a method to address gradient vanishing with a piecewise truncated gradient formulation that improves the optimization stability and robustness to initializations. We show that our method consistently improves 3D Gaussian Splatting with random and COLMAP initializations while being generalizable across static and dynamic Gaussian Splatting. As a by-product, we also examine the limitations of current benchmarks for dynamic scenes, and introduce a novel dataset for benchmarking dynamic Gaussian Splatting using synthetic 3D scenes. We demonstrate the effectiveness of our method in both static and dynamic settings for the public benchmarks and our proposed dataset. --- ### Coupled Scaling: A Representational Accessibility Framework for Neural Scaling Laws Authors: Jie Wang Published: 2026-09-03 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.03533v2 We ask when two learning systems trained on the same task under a common resource protocol should share a scaling rate and when their rates should differ. Coupled Scaling answers this through representational accessibility: the task-relevant geometry that a specified architecture-optimization system can reach and how that geometry is acquired as resources grow. In an orthogonal model, unsupported target energy sets the asymptotic floor, while unacquired supported energy sets the finite-budget residual. When acquisition can skip high-value directions, the largest fully acquired prefix no longer determines a unique exponent. For target powers $a_j \asymp j^{-b}$, $b>1$, and prefix log-growth rate $0<ρ\leq1$, the sharp attainable interval is $(ρ(b-1),\min\{b-1,ρb\})$, and every rate in this interval is realized by a fixed acquisition order on a common support. A rank-window condition identifies when the product law $α=ρ(b-1)$ applies, while a fixed-kernel specialization expresses the same task-system dependence through the task-weighted spectral tail. We then audit released capability trajectories to diagnose measurement and comparison effects and propose a staged test in which independently measured geometry predicts held-out loss, scaling rates, and cross-task reversals. --- ### LeanGRPO: Eliminating Redundant Recomputation in Diffusion RL Authors: Sijie Wang, Zhiqiang Tan, Xinrui Yang, Shaohuai Shi Published: 2026-09-03 Categories: cs.LG, cs.AI, cs.AR arXiv: https://arxiv.org/abs/2609.03528v1 Diffusion reinforcement learning (RL) has recently achieved significant success in post-training image and video generative models. However, most diffusion RL methods, including DanceGRPO and FlowGRPO, recompute selected timesteps with gradient tracking after rollout. Under on-policy training with the same backend for rollout and update, this recomputation is mathematically redundant. Intuitively, the rollout and policy update steps can reuse the same feed-forward backbone to avoid redundant computation, but doing so can incur a large memory overhead during rollout. To address the issue, we present LeanGRPO by restructuring the data-parallel layout and introducing two recompute-free training schedules for trajectory-logprob diffusion RL: (1) LeanGRPO-Retain enables gradient tracking during rollout and directly reuses the resulting computation graphs and saved activations for backward during update, requiring no recomputation; and (2) LeanGRPO-Reweight also enables gradients during rollout, but immediately backpropagates each selected step using a provisional advantage and delays gradient synchronization, then corrects the provisional gradients with the true advantage after the trajectory is completed. These schedules target different model scales and input sizes. Across FlowGRPO/DanceGRPO with FLUX.1-dev and Wan, LeanGRPO achieves up to 1.83x end-to-end speedup while preserving the original optimization objective. --- ### NeoRed: A Knowledge-Logic-Alignment Multimodal Large Language Model for Neonatal Respiratory Disease Diagnosis Authors: Yinan Liu, Hongtai Xia, Haoran Xu, Jiankang Hong, Yu Jianli, Jingkuan Song, Ye Luo Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.03527v2 Neonatal respiratory diseases are a major cause of neonatal morbidity and mortality, posing substantial challenges in clinical practice. Despite recent advances, existing Multimodal Large Language Models (MLLMs) face two key limitations in neonatal diagnosis: (1) domain gap arising from predominantly adult training data; (2) insufficient integration of multidimensional clinical context for accurate diagnosis. To address these challenges, we collect two real-world clinical datasets (NeoCXR and NeoCXR-EV) and propose NeoRed, to the best of our knowledge, the first MLLM tailored for neonatal respiratory disease, filling the gap in neonatal diagnostic reports generation. To enhance joint diagnosis from heterogeneous clinical context and chest X-rays, we design a novel Knowledge-Logic-Alignment (KLA) framework which constrains model behavior from three perspectives: 1) Knowledge Prior Injection (KPI) incorporates neonatologist-inspired diagnostic priors into multimodal representations, guiding disease-specific attention across modalities; 2) Diagnostic Logic Constraint (DLC) aligns the semantics of generated reports with multimodal diagnostic logic; and 3) Visual Semantic Alignment (VSA) establishes semantic correspondence between visual features and imaging conclusions. Extensive experiments demonstrate that NeoRed enables accurate neonatal diagnostic reports generation, achieving ROUGE-L of 53.29% and Clinical Efficacy F1 score of 65.19% on NeoCXR, outperforming existing MLLMs. NeoRed also preserves competitive report generation performance on adult benchmarks (MIMIC-CXR and IU-Xray). Datasets will be available upon application. --- ### CulturalMenuBench: Probing the Knowledge-Application Gap in Multimodal Culinary Reasoning Authors: Bo Zeng, Linfeng Gao, Peiqin Lin, Yu Zhao, Mingyan Zeng, Yu Tong, Xintong Wang, Linlong Xu, Longyue Wang, Weihua Luo, Qinggang Zhang, Jinsong Su Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.03526v1 Multimodal language models achieve near-ceiling scores on food recognition benchmarks, yet it remains unclear whether this success reflects genuine cultural understanding or mere visual matching. To probe this distinction, we introduce CulturalMenuBench, a benchmark of 4,870 items in 10 languages across 18 regions; its 10 tasks pair final-dish and step-by-step cooking images with ingredients, procedural text, and regional labels, spanning basic recognition to process-grounded cultural attribution. Evaluating 12 models exposes a substantial knowledge-application gap: models exceeding 94% on standard multiple-choice tasks drop to at most 56% when attributing dishes to Chinese regional cuisines, despite an identical four-way format. Diagnostic analyses explain why: error patterns are consistent with random guessing, accuracy tracks visual distinctiveness rather than cultural structure, and models classify cuisines more accurately from dish names alone than from images (+7-18 points). The knowledge is thus present but cannot be activated through visual input. An ablation confirms these tasks genuinely require procedural evidence: removing sequential cooking images selectively degrades process-grounded tasks while others remain stable. Overall, CulturalMenuBench shows that near-perfect recognition can conceal an inability to apply cultural knowledge, motivating training that explicitly connects perception, procedure, and cultural context. Code and data are publicly available. --- ### EPIC: Explicit Posterior Item Conditioning for Semantic ID Diffusion Recommendation Authors: Tuan-Binh Tran, Thanh Tam Nguyen, Quoc Viet Hung Nguyen, Dung D. Le, Tung Kieu, Thanh Trung Huynh Published: 2026-09-03 Categories: cs.IR, cs.LG arXiv: https://arxiv.org/abs/2609.03522v1 Semantic ID (SID) generative recommendation predicts the next item by generating a short tuple of discrete tokens. Recent masked-diffusion methods improve this process through bidirectional context and flexible decoding, yet recommendation ultimately requires selecting among complete catalog items. At each denoising step, a partial SID can correspond to multiple feasible items, while existing methods primarily reason through position-wise token predictions. We propose Explicit Posterior Item Conditioning (EPIC), which introduces explicit item-level competition into SID denoising. EPIC constructs a personalized posterior over feasible candidate items using the current generation context and the user's recent interactions, then projects this distribution back to unresolved SID positions to guide subsequent token decisions. The pretrained backbone remains frozen and requires no additional decoder forward pass. Experiments on four Amazon benchmarks show consistent improvements over strong baselines, while diagnostic analyses indicate that the gains primarily arise from personalized transition evidence that preserves promising item hypotheses during denoising. --- ### Neural Video Compression Based on Deformable Temporal Alignment and Difference-aware Fusion Authors: Chuyue Shan, Songlin Sun, Wang Chenwei, Shen Zihan Published: 2026-09-03 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.03520v1 In conditional coding-based neural video compression, the quality of temporal context directly affects compression per- formance. Existing methods mostly construct context from prop- agated reference features, but they are vulnerable to motion esti- mation and local alignment errors in regions with complex mo- tion, occlusion, and high-frequency textures, resulting in inaccu- rate temporal information. To address this issue, this paper pro- poses a method combining deformable temporal alignment and difference-aware spatial selective fusion. A Context-aware Tem- poral Alignment Module is used to generate complementary tem- poral context, while a Difference-aware Spatial Selective Fusion module adaptively selects reliable temporal information and sup- presses misalignment. Experiments show that the proposed method achieves certain rate-distortion performance improve- ment over DCVC-DC. --- ### Residual Optimal Transport-Based Experts Collaboration Towards Modality-Aware Infrared-Visible Object Detection Authors: Yue Zhao, Hua Yu, Yukun Zhao, Yuzhi Zhang, Maoguo Gong, Xin Mei, Zhuping Hu, Yanchi Li, A. K. Qin Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.03516v1 Infrared-visible object detection (IVOD) integrates complementary evidence from visible and infrared sensors for reliable perception in challenging scenes. In practice, sensors may fail or drop frames, leaving one modality unavailable or intermittent. Existing methods for IVOD assume both modalities are always present, and fixed fusion collapses when one stream is missing. Furthermore, it remains a critical challenge to reliably estimate semantic correlation across heterogeneous modalities, especially under spectral distribution discrepancy. We present FlexibleFusion, a unified and adaptive method that flexibly allocates integration pathways and fusion strength, operating seamlessly across complete and missing-modality regimes. At its core, the Modality-Aware Experts Collaboration (MAEC) mechanism selectively activates and aggregates cross-modal or intra-modal expert pathways. It allows cross-modal fusion when full modalities are available and falls back to self-fusion under missing conditions. Additionally, we design Residual Self-Paced Entropic Optimal Transport (RSPEOT) to align heterogeneous feature distributions from a transport perspective. Instead of relying on the fixed sparsity coefficient in standard entropic optimal transport (EOT), RSPEOT introduces a residual-driven self-paced update that prioritizes reliable matches and progressively refines harder ones. This design alleviates the additional optimization burden of standard EOT while preserving reliable semantic alignment. Comprehensive experiments under complete and missing-modality protocols show consistent performance across arbitrary modality configurations. Code will be released upon publication. --- ### What Matters for Aggressive Decoding-Time KV Eviction? Temporal Aggregation and Ranking Preservation Authors: Bo Zeng, Yu Zhao, Yefeng Liu, Zhihong Lu, Xuanfan Ni, Xintong Wang Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.03515v1 Decoding-time KV cache compression research focuses heavily on designing better token scoring functions, while the temporal rule that aggregates scores across decode steps is often treated as an implementation detail. Under aggressive KV compression, we find that exponential-moving-average (EMA) aggregation makes approximately order-preserving scorer modifications largely indistinguishable at the eviction-set level. Value-norm and entropy variants remain highly correlated with attention and produce nearly unchanged retention sets, whereas KeyDiff, key norm, recency, and a learned scorer alter the ranking and degrade substantially. We associate this stability with the evaluated aggregation, which couples layer weighting and temporal retention. Building on this observation, we introduce InertiaKV, an EMA-based decoding-time eviction method, and InertiaKV-Lazy, its periodic-refresh variant, which yields 1.34-1.46x decode throughput relative to full refresh InertiaKV. We also study Score-Free decoding as a separate empirical operating point: it scores the full context once at the first decode step, freezes that ranking, and incurs an average quality change of +0.03 while removing all subsequent scoring. Across six open-weight backbones and the LongBench, LongBench-v2, and RULER benchmarks, the results identify temporal aggregation and ranking preservation as distinct, consequential design factors; they do not imply that scoring quality is irrelevant in general. --- ### Lost in Reordering: Structural Sensitivity of Multilingual LLMs under Semantics-Preserving Perturbations Authors: Karthika Nhayakkat, Rajat Verma, Maharaj Brahma, Vetcha Gnana Mahesh, Maunendra Sankar Desarkar, Ganesh Ramakrishnan, Rohit Saluja Published: 2026-09-03 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.03511v1 Large Language Models (LLMs) demonstrate strong multilingual reasoning performance, yet their robustness to semantics-preserving structural variation remains underexplored, particularly for relatively free word-order languages. We investigate the structural sensitivity of multilingual LLMs using two linguistically grounded perturbation settings in Hindi and Malayalam: constrained constituent reordering and active-passive voice transformation. We introduce a benchmark dataset IndicReStruct, with two variants, GSM8K-Reordered and GSM8K-Voice, constructed from GSM8K while preserving semantic meaning. Across six state-of-the-art LLMs and multiple prompting strategies, we observe consistent and significant degradation in mathematical reasoning performance under structurally perturbed inputs. To further understand these failures, we perform qualitative error analysis and mechanistic interpretability experiments using residual-stream activation patching. Our analyses show that reasoning failures frequently arise from disruptions in entity-quantity alignment and that intermediate transformer layers contribute most strongly toward reasoning restoration. Overall, our findings suggest that current multilingual LLMs remain highly sensitive to surface syntactic realization and lack robust compositional invariance under structurally different but semantically equivalent inputs. --- ### LongCounsel-8: A Benchmark Suite for Longitudinal Depression Tracking from Multi-Session Counseling Dialogues Authors: Jiayi Li, Zhaomin Wu, Bingsheng He Published: 2026-09-03 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.03507v1 Tracking depression from multi-session counseling dialogues requires estimating both current symptom severity and how it changes across sessions. Yet progress on this task is constrained by the scarcity of longitudinal counseling data with standardized session-level depression labels. Existing resources typically provide either multi-session conversations without depression labels or labeled interviews in a single session. Building such a benchmark poses three challenges: maintaining longitudinal consistency and diversity, grounding symptom progression in empirical patterns, and expressing controlled depression states naturally without exposing target labels. To address these challenges, we introduce LongCounsel-8, a benchmark suite of three independently generated datasets totaling 7,749 five-session counseling trajectories, grounded in real-world client profiles, depression trajectories, symptom compositions, and counseling patterns. We combine profile-grounded simulation, empirically informed state construction, and indirect behavioral realization to address these challenges. Across the benchmark, simulated self-reports closely recover the controlled states, supporting label fidelity. Experiments on existing depression tracking methods reveal three key findings: (1) lower single-session score error does not guarantee accurate identification of trend, i.e., improvement or worsening; (2) existing methods are consistently less reliable on worsening trajectories; and (3) additional session history may reduce the accuracy of trend prediction. Together, these findings establish LongCounsel-8 as a foundation for advancing depression assessment from static, single-session prediction toward reliable longitudinal tracking of mental-health change. --- ### An Adversarial Zero-Shot Learning Approach for Anomaly Detection in Multivariate IoT Traffic Data Authors: Mahshid Rezakhani, Tolunay Seyfi, Fatemeh Afghah Published: 2026-09-03 Categories: cs.LG, cs.NI arXiv: https://arxiv.org/abs/2609.03505v1 Anomaly detection in Internet of Things (IoT) networks presents unique challenges due to the diversity of devices, lack of labeled data, and domain variability across environments. In this paper, we propose a novel framework for multivariate time-series anomaly detection that leverages adversarial learning and contrastive loss within a sequence-based Variational Autoencoder (VAE) architecture. Our method enables zero-shot domain adaptation by jointly optimizing domain-invariant latent representations and semantically structured embedding spaces, without requiring labeled data or raw feature transfer. To address the heterogeneity of IoT deployments, we introduce encoder and decoder adaptor layers that align feature distributions across domains while preserving contextual semantics. Additionally, we propose a destination-based segmentation strategy to better model real-world communication structures in IoT traffic. Our framework is comprehensively evaluated on six distinct datasets spanning industrial, enterprise, general-purpose, smart home, and military automation domains across 44 transfer scenarios. Experimental results demonstrate strong zero-shot generalization in several cross-domain settings and competitive performance against a contrastive domain-adaptation baseline under realistic, heterogeneous, and privacy-constrained IoT conditions. --- ### Restricted Eigenvalues Beyond Gaussian Width: Threshold Occupancy under Heavy Tails Authors: Shi Fu, Huibo Xu, Qixin Zhang, Dacheng Tao Published: 2026-09-03 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.03504v1 Restricted eigenvalue (RE) bounds govern stable recovery by norm-regularized estimators. For isotropic sub-Gaussian measurements, the benchmark sample size is $1+w(A)^2$, where $w(A)$ is the Gaussian width of the normalized descent cone. The COLT 2015 open-problem note (Banerjee et al., 2015) asked whether the same law follows for heavy-tailed designs from a uniform small-ball condition alone. We give an explicit and systematic negative answer to the general question as formulated there: the proposed law fails in its full dimension-free, arbitrary-set form, and the missing obstruction is simultaneous threshold occupancy. A constant-width polyhedral descent cone with fixed small-ball constants has zero empirical RE on every sample path up to half the ambient dimension. More generally, every finite range space admits exact threshold encoding in an arbitrarily narrow spherical cap and a lift to a full polyhedral descent-cone section. For every fixed threshold VC dimension $d$, as $β\downarrow0$, the sharp worst-case sample complexity is $Θ(β^{-1}(d\log(1/β)+\log(1/δ)))$. The separation persists under exact isotropy and all finite moments: on the same constant-width cone, Gaussian measurements succeed with $O(1+\log(1/δ))$ samples, whereas an isotropic heavy-tailed design fails pathwise for $n\lesssim\sqrt{p/\log p}$. Gaussian smoothing yields an everywhere-positive $C^\infty$ density while retaining arbitrarily poor RE. Under isotropy, a distribution-free fallback governed by affine dimension times squared enclosing radius is sharp on this family. --- ### PPO-STGNN: A Proximal Policy Optimization Approach with Spatio-Temporal Graph Neural Networks for DAG Task Scheduling in Cloud-Edge-End Computing Authors: Yangshuo Qi, Chenwei Wang, Zihan Shen, Songlin Sun Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.03503v1 With the rapid development of the Internet of Things, computation intensive directed acyclic graph (DAG) tasks have become increasingly common in cloud-edge-end collaborative environments. However, cloud, edge, and end nodes are highly heterogeneous in computing capacity, network bandwidth, and energy consumption, which makes the efficient scheduling of tasks with complex dependencies an NP-hard problem. Traditional heuristic algorithms and conventional reinforcement-learning methods often fail to capture the spatio-temporal dynamics of system resources. This paper proposes PPO-STGNN, a DAG task-scheduling algorithm that integrates proximal policy optimization (PPO) with spatio-temporal graph neural networks (STGNNs). The method uses an STGNN to extract features from both the DAG task topology and the physical cloud-edge-end resource graph, and then optimizes the scheduling policy through PPO to minimize makespan and schedule length ratio (SLR) while improving CPU and memory load balancing. To accelerate convergence, a multi-teacher behavior-cloning mechanism is introduced for pretraining. Experimental results show that PPO-STGNN significantly improves load balancing while maintaining a low completion time, making it suitable for dynamic and heterogeneous cloud-edge- end DAG scheduling scenarios. --- ### Building and Evaluating Fixed-Voice Thai TTS from Synthetic Speech Authors: Kunat Pipatanakul, Potsawee Manakul, Warit Sirichotedumrong, Sittipong Sripaisarnmongkol, Pakorn Nathong, Phatrasek Jirabovonvisut Published: 2026-09-03 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.03502v1 In low-resource settings, deploying TTS typically requires choosing between a large voice-cloning model with costly inference or a compact fixed-voice system that requires a speaker-specific corpus. We study a third route: using a large voice-cloning model as a programmable data source to turn a short voice reference (e.g., 15 seconds) into a compact fixed-voice student trained entirely on synthetic speech. This setting makes pipeline design consequential: teacher errors become training targets, while filtering failed generations can reduce coverage of difficult texts. Thai further introduces challenges from ambiguous word boundaries, lexical tone, names and loanwords, numeric verbalization, and Thai-English code-switching. We study how text preparation, synthetic generation, quality filtering, rejection sampling, and frontend choices affect the resulting student, and where teacher limitations remain. We evaluate CER, Challenge-Set Keyword Accuracy, Prosody Pause Accuracy, speaker similarity, and speaking rate. The resulting 82M-parameter model, Wayu-Paxa-TTS-Edge, enables on-device Thai TTS without reference audio. It achieves 68.2% Challenge-Set Keyword Accuracy (85.5% of Gemini 3.1) and 91.4% pause precision, outperforming its OmniVoice teacher (89.9%) and reaching 94.8% of Gemini 3.1. It also achieves the lowest pause-placement error and intra-word pause rates among the three systems, and 3.7% and 1.1% CER on Thai and English, respectively. We open-source the model and evaluation framework for Thai TTS development. --- ### Towards a Statistical Understanding of Mixture-of-Experts Authors: Siyuan He, Bokai Yang, Jie Hu, Ziwen Gao, Yuhong Yang Published: 2026-09-03 Categories: stat.ML, cs.LG, math.ST arXiv: https://arxiv.org/abs/2609.03501v1 Mixture-of-experts (MoE) architectures increase model capacity by combining a collection of expert predictors through input-dependent routing, while often activating only a small subset of experts for each input. Despite their growing importance in modern large-scale models, the statistical roles of their design choices, especially routing, sparse activation, and shared experts, remain only partially understood, as existing theory has largely focused on parametric or correctly specified MoE models. In this paper, we view MoE as a form of localized aggregation and show how this localization reshapes the approximation-estimation-computation tradeoff. We derive oracle risk bounds for learning dense and sparse routing with evolving experts, separating approximation, expert-learning, and router-estimation errors, and characterize how sparse Top-K routing can retain the benefits of localized aggregation while controlling per-input computation. We also interpret gating through the geometry of input space, relating routing performance to regions of local expert advantage, and show how shared experts, as adopted in architectures such as DeepSeekMoE, can extract common predictive structure so that routed experts focus on residual local variation. Together, these results provide a unified statistical framework for understanding MoE through input-dependent expert aggregation, in which expert specialization and computational tradeoffs are governed by local predictive structure. --- ### BRIDGE: An Open-Source Humanoid Platform via Morphology-Control Co-Design for Physical AI Authors: Jianren Wang, Letian Qian, Zikai Wang, Weiwei Wu, Junjie Zong, Abhinav Gupta, Deepak Pathak Published: 2026-09-03 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2609.03497v1 Developing humanoid robots capable of leveraging human behavioral data is essential for general-purpose embodiment, yet conventional development remains bottlenecked by a decoupled paradigm that isolates hardware design from whole-body control. This approach leads to suboptimal systems that compromise human-like fluidity and agility. To bridge this gap, we introduce a data-driven morphology-control co-design framework that optimizes humanoid morphology for human-like movement. To quantify morphological fidelity, we also introduce a novel metric that jointly considers kinematic retargeting fidelity to human motion and dynamic tracking performance. Our framework achieves state-of-the-art (SOTA) performance across all metrics compared to baseline humanoids (Bumi, K1, and Toddlerbot). Finally, we realize this design in Bridge, an open-source, 88cm-tall humanoid platform released alongside its control policy. We demonstrate that Bridge captures human motion data with superior fidelity, exhibiting exceptional performance across foundational locomotion, robust balance, and highly dynamic maneuvers. Videos and open-source materials: https://sites.google.com/view/bridgerobot. ---