# Research Papers — 2026-08 · bundle page 75 of 240 > Full-text markdown of papers #3701–3750 of 11995 published in 2026-08, newest first. AI-readable mirrors of the full 11995-paper 2026-08 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-08/page/75 [← Previous bundle](https://aipod.dev/papers/archive/2026-08/page/74.md) [Next bundle →](https://aipod.dev/papers/archive/2026-08/page/76.md) ### SkillBloat: Token Amplification Attacks via Skill Injection in LLM Coding Agents Authors: Yuanjin Zheng, Jingbang Chen Published: 2026-08-22 Categories: cs.CR, cs.CL arXiv: https://arxiv.org/abs/2608.21929v2 Agent skills extend coding agents with task-specific instructions, scripts, and resources, but they also create a trusted instruction channel that can be abused beyond conventional security attacks. This paper studies token amplification through skill injection: an economic resource-abuse threat in which a malicious skill causes an agent to consume substantially more tokens than needed for normal task execution. We present SkillBloat, a two-phase framework that first screens a library of diverse attack-type conditions across multiple amplification mechanisms and then refines the strongest candidate through LLM-guided full-document skill rewriting. Evaluated on a real-world skill benchmark, SkillBloat achieves 5.4184x-10.1455x average best amplification across multiple coding-agent target configurations. An ablation shows that the second-stage refinement loop consistently improves average best amplification over Phase 1 attack-type screening alone, demonstrating that iterative optimization provides additional benefit beyond initial attack-type selection. These results show that skill ecosystems expose a practical resource-amplification attack surface that is orthogonal to existing security-oriented skill poisoning. --- ### GuardianBench: A Same-Scene Instruction-Contrastive Benchmark for Latent Contextual Risk in Embodied AI Authors: Zhesheng Zhang, Jiahao Lu, Wei Liu, Cong Pan, Jianhua Yang, Yixiang Chen, Hongyuan Yu, Mengqi Zhang, Kailin Lyu, Zhumin Chen, Keji He Published: 2026-08-22 Categories: cs.AI, cs.CL, cs.RO arXiv: https://arxiv.org/abs/2608.21928v1 In embodied AI, safety risk can be latent: a benign instruction and a safe scene become hazardous only when composed. Prior work has advanced embodied safety by varying visual contexts or evaluating execution-time dynamics, but the complementary axis of fixing the scene and varying only the instruction remains underexplored. We introduce GuardianBench, an instruction-contrastive benchmark grounded in international safety standards that isolates this latent contextual risk through 3,024 instruction-scene examples organized as same-scene Safe/Unsafe contrastive pairs across various hazard categories. Benchmarking state-of-the-art vision-language models (VLMs) reveals instruction-insensitive verdicts: models disproportionately approve both instructions under a given scene; across the primary models, average pair accuracy is only 24.1%. Our systematic rationale audit localizes the dominant failure: models fail to bind the instruction-relevant cues that differentiate safe from unsafe compositions. As a post-training case study, Verdict Log-Odds Supervision (VLOS), a lightweight verdict-level objective, substantially improves performance on open-weight backbones. Together, our latent contextual risk task formulation, standards-grounded contrastive benchmark construction, pair-level and rationale-level failure diagnosis, and benchmark-enabled verdict calibration establish GuardianBench as a controlled evaluation suite for exposing and improving safety reasoning over instruction-scene compositions under latent contextual risk. --- ### Bi-EZP: LLM-Guided Bilevel Program Evolution for Ensemble Zero-Cost Proxy Discovery Authors: Yutao Lai, Kezhao Lai, Hai-Lin Liu Published: 2026-08-22 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2608.21927v1 Zero-cost proxies enable neural architecture search (NAS) to rank candidate networks from statistics computed at initialization, avoiding repeated training. However, different proxies capture different properties and often produce inconsistent rankings across search spaces. Ensemble proxies can combine complementary signals, but automated discovery must optimize both discrete aggregation structures and their continuous coefficients, making structural quality difficult to separate from parameter calibration. We propose Bi-EZP, a bilevel framework that decouples these decisions. At the upper level, a large language model generates executable aggregation programs over four complementary base proxies with program-specific parameter bounds. At the lower level, covariance matrix adaptation evolution strategy (CMA-ES) optimizes the continuous parameters of each fixed program on an inner training split. The calibrated programs are then evaluated using Kendall's rank correlation on a disjoint validation split, enabling evolutionary selection to favor structures that generalize beyond their calibration data. Experiments on NATS-Bench and Network Design Spaces evaluate ranking performance across heterogeneous search spaces, and DARTS experiments assess downstream architecture search. Results show that separating program discovery from numerical calibration provides an effective approach to automated ensemble zero-cost proxy construction. The source code is available at: https://anonymous.4open.science/r/Bi-EZP-318D --- ### AirAlign: Geometry-Aware Relative Pose Alignment for UAV Last-Meter Navigation Authors: Jinyi Zhou, Shuo Feng, Yufei Wu, Piji Li Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21926v1 Unmanned aerial vehicle (UAV) navigation in modern low-altitude environments requires more accurate pose alignment in the final approach stage for target information acquisition or manipulation, making "last-meter" navigation increasingly important. However, severe viewpoint and appearance variations make this task challenging. To tackle this problem, we propose AirAlign, a framework for RGB-only image-pair relative pose alignment for UAVs. AirAlign uses a pretrained visual geometry reconstruction model as the backbone to extract geometry-aware features from source-target image pairs. In addition, to better utilize the limited training data, we split the training set into multiple scene-disjoint folds for unseen cross-validation and model selection. During inference, the predictions of the selected models are averaged to form the ensemble output of the overall framework. Experiments on the PairUAV challenge at the ACMMM 2026 Workshop on UAVs in Multimedia demonstrate the effectiveness and robustness of our method, while comprehensive ablation studies validate the contribution of each component. --- ### ESCRAG-R1: Retrieval-Augmented Reinforcement Learning for Emotional Support Conversation Authors: Weichu Liu, Yuxuan Hu, Yirong Sun, Ningning Mao, Ziyun Zhang, Jian Chen, Mingyang Xu, Qishan Zhong, Chengming Li Published: 2026-08-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.21925v1 Emotional Support Conversation (ESC) systems aim to provide holistic support by balancing professional therapeutic competence with natural empathy. However, existing methods struggle to simultaneously achieve structured, stage-aware reasoning and seamless empathy-expertise alignment, often resulting in an artificial splicing of clinical strategies and generic reassurance. To overcome these limitations, we propose ESCRAG-R1, a unified framework that integrates retrieval-based psychological guidance into Group Relative Policy Optimization (GRPO). By incorporating retrieval into the reinforcement learning loop, ESCRAG-R1 transforms external knowledge into a robust learning signal that stimulates explicit internal reasoning prior to generation and fundamentally reshapes the model's internal policy. To provide the reliable supervision required for this optimization, we construct ESC-Preference, a high-quality dataset based on a Client--Counselor--Judge evaluation framework that delivers precise, empathy-aware reward signals. Extensive experiments demonstrate that ESCRAG-R1 significantly outperforms existing baselines by mitigating superficial splicing and realizing a natural integration of professional guidance and empathetic expression. Code and datasets are released at https://github.com/Matcha-Liu/ESCRAG-R1. --- ### Modeling Claim Dependency Structure for Patent Litigation Prediction with Graph Attention Networks Authors: Takao Arai, Hiroyasu Inoue Published: 2026-08-22 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.21924v1 Patent litigation imposes substantial costs on firms and distorts R&D incentives, making early risk identification a practically important task. While prior work has applied BERT-based models to patent claim text, two fundamental limitations remain: flat sequence encoding loses the dependency structure between independent and dependent claims that legally determines patent scope, and feeding the entire claim set to a single encoder discards legally critical text. A six-model ablation on 1.34 million USPTO utility patents confirms that per-claim encoding, graph connectivity, attention, and Attentional Aggregation each provide independent, additive predictive value. We propose ClaimGAT, a Graph Attention Network that encodes each claim independently, constructs a directed claim dependency graph, processes it with GATConv layers, and aggregates independent claims via Attentional Aggregation to yield both a litigation risk score and claim-level gate weights that enable post-hoc structural analysis. ClaimGAT achieves an AUC-ROC of 0.818 and a lift of 4.89x at the top 10%, using only information observable at the time of patent grant. It reveals a tendency in high-risk patents for structural selection and content sensitivity to diverge, a pattern consistent with defensive claim drafting. --- ### Beyond Fixed Directions: Adaptive Representation Analysis of Reasoning and Memorization in LLMs Authors: Shaheen Nabi Published: 2026-08-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.21919v1 Recent work has proposed that reasoning and memorization in language models can be characterized by a single representation direction, including methods that keep this direction fixed during reinforcement learning. We test two assumptions behind this view. First, are reasoning-oriented and factual-recall task groups approximately single-direction separable? Second, does the resulting geometry remain stable after GRPO? Using Qwen3-0.6B and a controlled 400-example dataset, we find that a one-dimensional projection can match a full 1024-dimensional linear probe with AUROC = 1.00 on the studied task groups. However, after GRPO, the corresponding direction is substantially reorganized: mean-direction cosine averages 0.453, probe-direction cosine 0.445, while direct representation drift reaches 0.511 at the final layer. Probe AUROC nevertheless remains 1.00. The evidence therefore supports single-direction decodability for the studied task groups but challenges fixed-direction stability: the information persists while its geometric realization changes. --- ### Consistency Is Not Coherence: Orientation Search for Certified Alignments Between 4D Defence Upper Ontologies Authors: Fabio Rovai Published: 2026-08-22 Categories: cs.AI, cs.DB, cs.LO arXiv: https://arxiv.org/abs/2608.21914v1 We align three upper ontologies that sit under UK and NATO defence data infrastructure: the Information Exchange Standard (IES), the Higher Quality Data Model (HQDM) that underpins the National Digital Twin, and Basic Formal Ontology (BFO). No public alignment between IES and HQDM existed. Promoting a hand-curated 17-correspondence crosswalk to OWL and reasoning over the complete merged ontologies with HermiT produces three results that we believe matter beyond this pair. --- ### Entity-Constrained CBCT Retrieval for Low-Resource Dental Record Completion Authors: Nhi Ngoc-Yen Nguyen, Thai Nguyen, Kiet Huynh Cao Tuan, Huy-Hieu Pham Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21913v1 Completing dental records from cone-beam computed tomography (CBCT) is difficult when annotation is scarce and individual clinical fields are supported by different types of evidence. MMDental Task 3 requires seven-field record completion from only 50 labeled CBCT cases and scores the correctness of structured FDI positions and ICD codes; consequently, a visually plausible retrieved record can still be harmful when it introduces an unsupported entity. We propose Entity-Constrained CBCT-Guided Retrieval (ECCR), a parameter-free framework that separates evidence availability from evidence authority. A corpus-derived prior first supplies the complete record. A frozen 3D encoder retrieves image-conditioned Diagnosis evidence, which is appended only if it does not expand the prior FDI or ICD entity set, so the asserted entity set is invariant by construction. On public validation, ECCR reaches a weighted score of 0.3134, improving on both full-record multimodal retrieval (0.2237) and a static text-only prior (0.2915); the guard blocks 63.3% of retrieved candidates, each of which would otherwise have injected an FDI position or ICD code absent from the prior. On the final test evaluation, ECCR obtains 11.37 of a 97.4-point attainable maximum, securing second place overall. The result indicates that, in an extreme low-resource setting, controlling what multimodal evidence is allowed to modify can be more reliable than transferring an entire retrieved record. --- ### CD-LoRA: Consistency-Driven Low-Rank Adaptation for Multi-Task Fine-Tuning Authors: Qian Zha, Jinda Liu, Yuan Wu, Yi Chang Published: 2026-08-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.21909v1 While Multi-Task Learning (MTL) is essential for adapting Large Language Models (LLMs) to diverse domains, prevailing LoRA-based methods rely on complex routing mechanisms that partition task-specific knowledge. In this work, we reveal that such routing-based designs are prone to a training-inference discrepancy, where stochastic routing decisions under distribution shifts compromise inference stability. Driven by a second-order Taylor analysis that exposes the instability induced by routing variance, we challenge the training-inference discrepancy and propose Consistency-Driven Low-Rank Adaptation (CD-LoRA). By eliminating routers entirely, CD-LoRA employs a consistency-driven alignment mechanism to enforce representation congruence across tasks in a shared low-rank space. This paradigm fosters robust, task-agnostic features without explicit partitioning overhead. Extensive experiments show that CD-LoRA consistently outperforms state-of-the-art multi-adapter baselines, offering a simpler, router-free, and more stable solution for multi-task PEFT. The code is available at the anonymous link https://github.com/zhaqian21/CD-LoRA. --- ### Training Needs Trustworthy Worlds: Verified Synthetic Web Environments for Agent Learning Authors: Chenghao Zhang, Yuxi Cheng, Saisai Hu, Canran Xiao, Yikai Mao, Dan Roth Published: 2026-08-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.21898v2 Web agents promise to automate complex digital workflows, but their training remains limited by synthetic environments that look plausible while hiding broken links, inconsistent states, or infeasible tasks. We address the gap between scalable environment generation and trustworthy agent learning by constructing synthetic web environments that are executable, auditable, and grounded in backend state. Our framework represents each generated website as a structured scaffold of pages, navigation links, database records, state-change markers, and task constraints, then verifies and repairs structural, semantic, consistency, and feasibility defects before policy training. During interaction, ordinary UI transitions are executed deterministically, while persistent backend updates are invoked only through validated state-change markers, enabling dense rewards compiled from verified task-progress predicates. Across 500 synthetic environments spanning six domains, our method reduces task-blocking defects and improves feasible-task rate from 48.6% to 94.8%, while producing stronger PPO policies and improving transfer to WebArena, WebShop, and MiniWoB++ without LLM calls at evaluation time. These results show that verified synthetic environments can serve as a scalable and reliable training substrate for compact web agents, shifting synthetic webagent learning from surface-level plausibility toward executable, state-grounded supervision. --- ### From Solver Feedback to Faithful Plans: Multi-Role Reinforcement Learning for Symbolic Planning Authors: Chenghao Zhang, Yikai Mao, Haoyu Gao, Saisai Hu, Yuxi Cheng, Shanqi Liu, Dan Roth Published: 2026-08-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.21897v2 Reliable planning requires converting natural-language instructions into executable symbolic specifications, yet large language models remain brittle without costly PDDL annotations and may exploit solver success in semantically unfaithful ways. We study how to learn faithful natural-language-to-PDDL formalization using only solver feedback, without human-written demonstrations. We propose a solvergrounded multi-role reinforcement learning framework where a single language model acts as an Actor, Judge, and Editor for generation, verification, and repair. The Actor proposes PDDL specifications, the Judge provides a solver-calibrated quality signal, and the Editor performs bounded diagnostic-conditioned refinement. On PlanBench, our method improves average success from 35.5% for LLM+P to 70.8%, achieves 66.3% faithful success, and reduces semantic drift to 6.4%. These results show that organizing solver feedback into generation, verification, and repair roles enables more scalable and faithful annotation-free symbolic planning --- ### Breaking the Assumptions: Auditing Input-Side Jailbreak Defenses Against Semantic Attacks Authors: Aaditya Pratap, Harsh Kasyap, Somanath Tripathy Published: 2026-08-22 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2608.21895v1 Locally deployed Large Language Models (LLMs) via inference engines such as Ollama run without the moderation and abuse detection present in API-served models. Therefore, the safety of LLMs depends on the defense mechanisms used, and their effectiveness depends on the assumptions on which they were designed. This paper does an audit of defense mechanisms under jailbreak attacks on locally deployed models. Some defenses provide formal guarantees (SmoothLLM, Erase-and-Check, Sequential Monitors), while others rely on empirical detection results (Semantic Smoothing, Self-Denoised Smoothing, Perplexity Filtering). Instead of merely observing that defenses fail, we trace each failure back to the specific assumption: for every defense, we extract the condition it relies on, derive the empirical pattern a violation should produce, and test that prediction on six open-weight models (14B to 35B parameters) with a corpus of 100 jailbreak prompts taken from more than 40 public sources, totalling 13,800 evaluation records. --- ### A Scalable Vector Graphics Latent Space Authors: Leonardo Zini, Elia Frigieri, Lorenzo Baraldi Published: 2026-08-22 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2608.21893v1 Scalable Vector Graphics are a fundamental medium for resolution-independent visual content, yet the deep learning community lacks a continuous, dense, and invertible latent space for vector representations, the kind of foundational building block that Variational Autoencoders and their descendants have long provided for raster images. We introduce SLS (SVG Latent Space), a Transformer-based autoencoder that learns compact dense representations of individual SVG paths, the atomic visual elements from which any SVG image can be composed. By modeling SVG commands, coordinate data, and visual properties within a unified BPE-based token vocabulary, SLS learns fixed-size latent representations that jointly capture structure and appearance, and can be decoded back into valid, style-consistent SVG paths with high fidelity. The resulting embedding space is robust, invertible, and structured: embeddings lie on a unit hypersphere, enabling efficient similarity search, composition, and downstream conditioning through simple vector-space operations. Finally, we demonstrate that SLS generalizes across diverse tasks reducing their FLOPs by over 150 times compared to token-based approaches, and establishing a general-purpose latent foundation for vector graphics research. --- ### PhysECD: A Physics-Constrained E(3)-Equivariant Framework for Electronic Circular Dichroism Spectrum Prediction Authors: Yi Jiang, Letian Chen, Runhan Shi, Liangzhaoxuan Han, Tong Zhu, Yang Yang Published: 2026-08-22 Categories: physics.chem-ph, cs.LG arXiv: https://arxiv.org/abs/2608.21892v1 The electronic circular dichroism (ECD) spectrum is a primary experimental probe for assigning the absolute configuration of chiral molecules, yet interpreting a measured spectrum requires time-dependent density functional theory (TDDFT) calculations that can cost hours per molecule and must be repeated for every candidate stereoisomer and conformation. We present PhysECD, a physics-constrained, parity-aware E(3)-equivariant framework that bypasses computationally expensive TDDFT and predicts ECD spectra directly from the 3D structure of an individual conformer. Instead of regressing the spectrum as an opaque sequence, PhysECD predicts the physical quantities that generate it: per-state excitation energies and electric and magnetic transition dipoles. These quantities determine the rotatory strength R -- the dot product of the two dipoles, a pseudoscalar that reverses sign under mirror reflection -- and yield the final spectrum through a differentiable Gaussian-broadening formula derived from the underlying physics. The parity structure of the equivariant features guarantees the correct chiroptical symmetry: reflecting a molecule exactly negates the predicted spectrum. On the CMCDS dataset, PhysECD attains a per-molecule spectral Pearson correlation of 0.642 (mean) / 0.822 (median), substantially exceeding prior learned predictors while remaining physically interpretable. Experiments across multiple backbones further show that the framework is backbone-agnostic, paving the way for real-time assignment of absolute configuration. --- ### TherMapNet Attention-Guided Runtime Full-Chip Thermal Map Prediction from Performance Metrics Authors: Qin Gu, Chaofang Ma, Mingyu Yang, Yipu Zhang, Jiliang Zhang, Wei Zhang, Lin Jiang Published: 2026-08-22 Categories: cs.AR, cs.LG arXiv: https://arxiv.org/abs/2608.21887v1 Runtime thermal management of high-performance chips depends on fast and accurate full-chip thermal maps. Conventional simulators typically estimate power traces from performance metrics first, which adds overhead. This work proposes TherMapNet, an attention-guided thermal simulator that predicts full-chip thermal maps directly from performance metrics. A Transformer encoder captures temporal evolution by treating the time series of each metric as a token, improving modeling of dynamic workloads. A CNN then extracts fine-grained spatial features. For the CNN, a dual-branch channel-spatial attention convolution module (DACM) and a triplet loss are used to improve spatial learning and reconstruction accuracy. TherMapNet is applied to a multi-core CPU (AMD Ryzen 7 4800U) and a many-core GPU (NVIDIA GeForce RTX 4060). Experiments show that it outperforms prior thermal simulators, with RMSE below 0.26 C and inference under 2.4 ms on an NVIDIA GeForce RTX 3090 GPU. These results indicate that TherMapNet can support high-quality runtime thermal management of modern multi-core chips. --- ### Pixel-Space Diffusion via Observation Operators Authors: Shaojie Guo, Lichen Ma, Haoyang Tong, Yu He, Zipeng Guo, Xiaoan Liu, Feng Yan, Yu Guo, Fei Wang, Junshi Huang, Yan Wang Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21885v2 Pixel-space diffusion models directly model image distributions but remain difficult to optimize. Recent methods alleviate this challenge through target reparameterization, while still relying on a fixed clean-image target throughout denoising. Through empirical analysis, we identify a scale-time mismatch: image structures become predictable from coarse to fine as noise decreases, whereas existing models are forced to predict the full image even under high noise, resulting in low-SNR gradients that hinder optimization. To resolve this mismatch, we propose Observation Operator Diffusion, a unified framework that aligns both the supervision trajectory and feature refinement with the intrinsic recovery order of image structures. Specifically, we replace fixed full-image supervision along the standard flow path with a time-indexed observation trajectory that evolves from coarse structures to the full image during denoising. This trajectory is instantiated with a family of Gaussian-Lanczos operators at varying observation scales, yielding a path-consistent training objective. We further introduce GL-CoDA, a decoder that injects scale-specific Gaussian-Lanczos observations across decoding stages for coarse-to-fine feature refinement. Extensive experiments show that the proposed approach converges substantially faster while consistently improving generation quality, achieving an FID of 1.52 on ImageNet-256. --- ### VIG: Visual Information Gain as a Reward Signal for Multimodal Chain-of-Thought Compression Authors: Wen Luo, Xiaohan Yi, Xiaotao Huang, Liqun Huang Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21883v2 Multimodal large reasoning models often rely on long Chain-of-Thought (CoT) traces in which a substantial fraction of tokens, such as repeated visual descriptions, self-reflection, and other visually-disengaged filler, inflate inference cost without contributing to the answer. Existing CoT compression methods optimize output length but never measure whether a reasoning token is actually grounded in the image. We propose \textbf{VIG} (Visual Information Gain), an information-theoretic GRPO reward that scores each reasoning token by how much the image reduces its predictive uncertainty. VIG is computed online from two forward passes of the same policy, one with and one without the image, so no reference chains, external annotations, or auxiliary reward models are needed. Across six main multimodal reasoning benchmarks and three Qwen3-VL-Thinking model sizes (2B/4B/8B), plus an additional R1-Onevision-Bench evaluation on 8B, VIG consistently improves the accuracy--efficiency trade-off, supporting our central claim: \emph{efficient multimodal reasoning emerges from raising visual information density, where every reasoning token earns its place by anchoring to the image, rather than from imposing a length budget.} Our source code is available at https://github.com/chaser682/vig. --- ### Region-Weighted Losses and Model Fusion for Cross-Modal PET Attenuation Correction Authors: Khoa Tuan Nguyen, Joris Vankerschaver, Wesley De Neve Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21881v1 We describe our approach to the Big Cross-Modal Attenuation Correction (BIC-MAC) challenge, which asks for a pseudo-CT in Hounsfield Units to be synthesized from Non-Attenuation-Corrected PET (NAC-PET), DIXON MRI and a topogram, and scores both the pseudo-CT and the Attenuation-Corrected PET (AC-PET) reconstructed from it. Three ideas carried our improvements over the organizers' 3D U-Net baseline. The loss matters more than the architecture: we compute the $L_1$ error in the Carney attenuation-coefficient ($μ$) space that the CT metric itself uses, weighted by anatomical region. Only once that loss was in place did the unregistered DIXON MRI work as extra input channels. A fixed convex combination of two independently trained models then beat both of its members on three of the four metrics and ranks first overall on the public validation leaderboard. --- ### BanglaVeilGuard: Cross-Script Safety Benchmarking and Lightweight Guardrails for Bangla Large Language Models Authors: Md. Rakibul Hassan, Muhammad Iqbal Hossain Published: 2026-08-22 Categories: cs.CL, cs.CR arXiv: https://arxiv.org/abs/2608.21880v1 Bangla large language model (LLM) safety is difficult to evaluate with English-centric or standard-script benchmarks because Bangla users routinely write across scripts, spellings, code-mixed forms, and regional registers. This paper presents BanglaVeilGuard, a compact Bangla-first safety benchmark and lightweight prompt guard for six language forms: standard Bangla, Romanized Bangla, Banglish, code-mixed Bangla--English, noisy Bangla, and dialectal Bangla. The benchmark contains 2,366 quality-filtered prompts and a held-out 354-prompt evaluation split spanning unsafe, safe, and safe-sensitive requests. BanglaVeilGuard uses non-destructive multi-view normalization with a prompt-risk classifier and thresholded pre-generation gate, allowing it to screen prompts for heterogeneous target models without changing their weights. Across target-model families, guarded runs reduce attack success under deterministic response scoring from 93.8--100.0\% to 6.3\% for Claude Opus 4.8, BanglaLLama, and TituLLM; TigerLLM-1B with BanglaVeilGuard achieves 78.2\% accuracy with 8.8\% ASR. The prompt guard also attains 88.5\% unsafe recall, substantially above the evaluated prompt-only guard baselines. The main remaining cost is over-refusal on dialectal and noisy benign prompts, revealing a concrete safety-helpfulness frontier for Bangla LLM deployment. --- ### ViSMoE: Visual-Aware Sparse Mixture-of-Experts for Embodied Referring Expression Grounding Authors: Shuo Feng, Piji Li Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21878v1 Embodied Referring Expression Grounding is the task of enabling an agent to navigate in real environments and to localize a remote object based on natural language instructions. In this scenario, the agent needs to select one view for navigation at each step and identify a specific object among all candidate objects at the destination. However, most of the previous approaches fail to distinguish between views and objects, instead processing them using the vanilla vision encoder, which results in ambiguous representations of both views and objects. To address the above issues, we propose ViSMoE, which equips sparse Mixture-of-Experts with a visual-aware routing policy for the embodied agent. This framework processes different types of visual information specifically, resulting in discriminative visual representations for both views and objects. Experimental results on REVERIE and SOON datasets demonstrate that ViSMoE outperforms the previous state-of-the-art methods, showing the superiority of our proposed method. --- ### Pruned Traffic Trees: Native Semantic Compression with a Protocol-Structured Model Family for Encrypted Traffic Classification Authors: Yuantu Luo, Jun Tao, Xiangyu Xu, Linxiao Yu, Kangying Li Published: 2026-08-22 Categories: cs.NI, cs.AI arXiv: https://arxiv.org/abs/2608.21874v1 Deep learning has achieved strong performance in encrypted traffic classification (ETC), yet its computational cost limits deployment on resource-constrained network devices such as routers and middleboxes. Existing compression methods mainly operate on weights, channels, hidden representations, or predictions, but do not explicitly determine which protocol fields and structural contexts should remain. We propose Pruned Traffic Trees (PTT), a three-level protocol-structured model family that treats native protocol structures as compression units. PTT-Full learns protocol-structured representations and field salience from complete Protocol Tree Graphs (PTGs), with flow-level self-supervised learning and protocol-presence-aware sparse execution. The learned salience and TopK+$k$ closure construct Distilled PTGs (PTG-Ds) for PTT-Distilled, while PTT-Lite inherits this topology and reduces width through structure-aligned transfer and flow-level logits distillation. Under flow-disjoint and Strong Information Information (SII)-masked settings, PTT-Full achieves Macro-F1 scores of 0.9519 and 0.9416 on CSTNET-TLS1.3 and CipherSpectrum, while PTT-Lite retains 0.9325 and 0.9136 with 80.3\% and 61.3\% fewer parameters, 98.85\% and 98.78\% lower effective GFLOPs, and 8.75$\times$ and 8.46$\times$ CPU inference speedups. These results demonstrate that treating protocol structure itself as the compression object enables effective performance-efficiency trade-offs for lightweight ETC. --- ### The Chase Is the Curriculum, the Capture Anchors the Credit: Pursuit-Evasion Self-Play for Zero-Data LLM Reasoning Authors: Jing Yu, Shengchao Chen, Yiyun Tan Published: 2026-08-22 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2608.21871v1 Reinforcement learning with verifiable rewards has become the dominant recipe for improving large language model reasoning, yet it presumes large human-curated task collections. Zero-data self-play removes this dependency, but existing methods vet learnability only by probing candidates and rejecting post hoc, never learning where along an environment's difficulty axis to place a task, and credit the solver with sparse terminal rewards alone. We recast zero-data self-play as a pursuit-evasion game: in LURE, an LLM evader positions tasks along each environment's difficulty axis to stay one step ahead of a planner-executor pursuer that hunts it down through verifiable interaction. The evader is trained on a capture-frontier reward that peaks when the solver captures it on exactly half of its rollouts, turning barely catchable into a learned positioning strategy rather than a hand-tuned rejection band. The pursuer earns capture-anchored dense process credit, in which monotone verifier progress is group-normalized jointly with the terminal capture under a round-anchored KL that keeps the co-evolution stable. Across three verifiable reasoning environments and three backbone families, LURE outperforms advanced baselines under unified/specialist settings, while the unified model attains stronger aggregate OOD zero-shot accuracy than all trained baselines across nine held-out benchmarks from three task families. --- ### GuardPaint:SpeculativeSafetyDecodingforText-to-ImageGeneration Authors: Shreyash Dhoot, Paras Dhiman, Arsh Abbas Naqvi, Aranbi Dutta, Aman Chadha, Vinija Jain, Amitava Das Published: 2026-08-22 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2608.21869v1 Text-to-image (T2I) diffusion models offer powerful visual generation, but their controllability creates a critical safety challenge: adversarial prompts can steer the denoising trajectory toward policy-violating content such as explicit nudity or graphic violence. Existing safeguards mostly act before generation through prompt filtering or after generation through image classification, leaving the diffusion process itself unguarded and often yielding only refusal rather than safe visual repair. We introduce GuardPaint, a speculative decoding framework for safe T2I generation that intervenes inside the diffusion trajectory without modifying the base model. A lightweight auditor monitors intermediate images, localizes unsafe regions, and triggers surgical inpainting repair only where needed. Candidate repairs are generated by a policy-aligned inpainter and selected through a guarded tournament that accepts edits only when they improve policy compliance while preserving prompt fidelity and perceptual quality. Across five jailbreak families SneakPrompt, MMA, PGJ, DACA, and RABell and UNet/flow-matching models including SD~1.5, SDXL, SD~3.5, and FLUX.1-dev. GuardPaint reduces attack success and harmful generations with minimal degradation to image quality, prompt fidelity, and benign behavior. Content warning: This paper contains examples involving nudity and violence that some readers may find disturbing, distressing, or offensive. --- ### HiMA-MDD: A Hierarchical Multi-Agent Harness for Interpretable Multimodal Depression Detection in Clinical Interviews Authors: Ao Chen, Xiaojiang Peng Published: 2026-08-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.21868v1 Depression assessment from multimodal clinical interviews requires integrating dispersed evidence from multiple symptoms into a coherent PHQ-8 profile. This process is hierarchical: relevant evidence is often sparse and context-dependent within local question-answer exchanges, multiple exchanges jointly support symptom-level judgments, and the final assessment depends on the coherence of the complete symptom profile. Existing LLM systems either process interviews holistically or distribute work across generic agent roles; neither design necessarily provides an explicit orchestration mechanism that coordinates evidence access, item-score authority, bounded feedback, and state recording across these levels. To address this gap, we introduce HiMA-MDD, a hierarchical multi-agent harness that aligns this assessment hierarchy with three agent layers. After non-agentic preprocessing constructs context-preserving multimodal QA units, Layer 1 identifies candidate QA-to-item relations and supports bounded item-grounded evidence routing. Layer 2 assigns symptom groups to operational factor specialists, with one specialist responsible for each provisional item score. Layer 3 audits the complete provisional profile, requests at most one round of targeted revision, and reconstructs the verified PHQ-8 profile. This layered design naturally yields a Hierarchical Evidence Trace, preserves all intermediate evidence, judgments, and revisions for auditability. The final item scores then deterministically produce the total score and screening decision. Using Qwen2.5-72B-Instruct as the harness backbone, our experiments on E-DAIC demonstrate that HiMA-MDD outperforms the compared state-of-the-art methods. --- ### MemGuard: Persisting Verifier Signals for LLM-Agent Memory Governance Authors: Haoyu Wang, Guangyuan Dong, He Liang, Zijing Zhang, Jiachen Luo, Chuang Liu, Chao Xue, Hao Tang Published: 2026-08-22 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2608.21867v1 LLM agents are moving from single-prompt use to long task streams in which reusable memory becomes a core capability for terminal, software-engineering, and web tasks. Such memory is useful only when stored experience remains reliable across hundreds of interactions, but two failure modes break that assumption in practice. The first is unreliable admission: failed trajectories,accidental successes, and misleading observations enter memory because they appear relevant, then mislead later decisions. The second is memory drift: long-running banks accumulate duplicate, stale, and conflicting records that retrieval alone cannot repair. MemGuard's key distinction is to treat verifier output not as a one-shot filter, but as persistent lifecycle metadata. It converts multi-criteria score-token verification into reward, confidence, label, and uncertainty descriptors that are attached to every candidate before activation and reused during retrieval, conflict resolution, summarization, and archival. We evaluate MemGuard on Terminal-Bench 2.0, SWE-Bench Verified, WebArena, and Mind2Web across four backbones, comparing against four memory baselines plus a verifier-only control under matched runtime budgets. Averaged over five seeds, MemGuard achieves the best success metric and lowest average steps in all 16 backbone-benchmark settings, improving over ReasoningBank, the strongest prior baseline among the memory methods we evaluate, with a largest gain of 7.9 success-rate points on WebArena, 5.6 step-success-rate points on Mind2Web, and 2.4-3.5 points on terminal and software-engineering benchmarks. Code is available at https://github.com/whyyyyy123/MemGuard. --- ### BioMed-Agent-RL: A Meta Learning, All You Need for Biomedical Applications Authors: Md Asaduzzaman Jabin, Zihao Wu, Tianming Liu Published: 2026-08-22 Categories: cs.LG, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2608.21864v1 The current progress of Clinical Vision Large Language Models (C-VLLMs) has substantially improved digital diagnostics, still these frameworks often endure lesion noises, modality misalignment, hallucination, and missed contextual grounding in complex clinical cases. Moreover, prevailing agent systems usually depend on static and non-adaptable pipelines and lack the versatility necessary for complex medical reasoning. To resolve these difficulties, we present BioMed-Agent-RL, a unified medical agent that incorporates adaptive orchestration, policy, and reward-based reinforcement learning (RL) models for biomedical applications. To ensure reliability, it invokes clinical context-aware preference optimization (CPO), direct preference optimization (DPO), and group relative policy optimization (GRPO) with dynamic entropy regulation. This pipeline utilizes a multimodal meta-learning approach that operates as a field-specific expert and human judgment synthesizer. The agent adaptively utilizes a set of model-level expertise, such as clinical grounding and reasoner, lesion segmenter, and field-specific synthesizer, across various clinical modalities (e.g., X-ray) by utilizing an iterative and adaptive RL approach. The agent learns to seriously synthesize misleading, conflicting vision cues and trust in inherent reasoning, while specialist advice is faulty. An intensive ablation study is conducted across multiple benchmarks, and the agent significantly outperforms existing state of the art models, such as GPT-5, attaining up to ~73% accuracy (gain of ~5%) over contemporary baselines. As a result, the framework suggests a new standard for building factual, reliable, robust, and expert-like intelligent agent systems for independent clinical reasoning. --- ### HiDiffTIR: Hierarchical Difficulty-Aware Policy Optimization for Multi-Turn Tool-Integrated Reasoning Authors: Yucan Guo, Xiaohan Wang, Miao Su, Saiping Guan, Zhongni Hou, Jiajun Chai, Wei Lin, Guojun Yin, Xiaolong Jin, Jiafeng Guo, Xueqi Cheng Published: 2026-08-22 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2608.21863v2 Tool-Integrated Reasoning (TIR) is a fundamental capability for LLM agents to solve complex tasks by interacting with external tools iteratively. Reinforcement Learning (RL) has become the dominant paradigm for enabling this capability. However, existing approaches typically assign uniform trajectory-level advantages and treat all correct tool calls equally, ignoring the varying difficulty and learning value across trajectories and reasoning steps. This can lead to imprecise learning signals that do not adequately distinguish between trivial and challenging tool-use patterns. To address this limitation, we propose HiDiffTIR, a Hierarchical Difficulty-aware policy optimization framework for multi-turn TIR. HiDiffTIR performs difficulty-aware credit assignment at both trajectory and turn levels, enabling the policy to focus on more informative trajectories and harder reasoning steps. Notably, this fine-grained optimization is achieved without additional supervision, relying solely on group-level statistics derived from standard RL rollouts. Extensive experiments on three tool-using benchmarks demonstrate that HiDiffTIR consistently improves multi-turn TIR performance and tool invocation accuracy over strong RL baselines, highlighting the necessity of difficulty-aware credit assignment for effective policy optimization in tool-integrated LLM agents. --- ### ChainPrune: Evaluating and Reducing Redundancy in Long Chain-of-Thought Reasoning Authors: Weihang Pan, Zhengxu Yu, Yuxiang Zhang, Wenzhi Li, Zhongming Jin, Binbin Lin, Xiaofei He, Jieping Ye Published: 2026-08-22 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2608.21860v1 Chain-of-Thought (CoT) reasoning has significantly enhanced the multi-step problem-solving capabilities of large language models (LLMs) by introducing explicit intermediate reasoning. However, advanced Large Reasoning Models (LRMs) often exhibit overthinking behaviors, including excessively long reasoning steps, redundant steps, and high computational overhead. Existing token-length reward strategies aim to promote concise outputs, but often result in pseudo-conciseness, where token count is reduced, yet redundant reasoning persists, leading to longer and less structurally efficient chains. To address these limitations, we propose ChainPrune, a novel reasoning path semantic structural optimization method to efficiently and controllably synthesize self-generated high-quality training data. We initially consolidate self-generated reasoning paths into a tree-based structure, followed by a multi-criteria dominant path selection process for preference data construction that formulates shallow reasoning trajectories while preserving essential reasoning steps. To further enhance the quality of reasoning, we incorporate a DPO-based preference learning method combined with supervised loss, effectively mitigating false reward suppression. This innovative integration significantly enhances both the efficiency and effectiveness of our reasoning framework. Comprehensive experimental results demonstrate significant reductions in step length and computational overhead, while maintaining or even enhancing accuracy. --- ### LLMs are Few-Shot Decision-Makers: Generalized Context-Aware Microgrid Frequency Control through Prompt Decision Transformer Authors: Xu Yang, Chenhui Lin, Haotian Liu, Kaihang Deng, Yunhe Li, Wenchuan Wu Published: 2026-08-22 Categories: eess.SY, cs.AI arXiv: https://arxiv.org/abs/2608.21858v1 The rapid evolution of energy structures has positioned microgrids as pivotal components of next-generation power systems, offering enhanced resilience and renewable energy integration. However, the inherent low inertia, complex dynamics, and poor model conditions of microgrids necessitate advanced data-driven frequency control strategies. Although reinforcement learning (RL) has demonstrated certain potential and advantages, existing RL methods often struggle with generalization across diverse microgrid configurations and lack adaptability to unseen environments, particularly when explicit system parameters are unavailable. To address these challenges, in this paper, we introduce a novel prompt decision transformer (Prompt-DT) architecture for microgrid frequency control. Unlike traditional approaches that rely on hard-to-obtain environmental characteristic parameters, the proposed method leverages few-shot expert historical trajectories as prompts to guide autonomous perception and adaptive decision-making. In addition, we propose a context-aware training and execution mechanism utilizing self-supervised contrastive learning to enhance environment recognition and prompt utilization efficiency. In addition, a physics-informed prompt design technique that filters prompts based on cumulative reward and frequency volatility is proposed, ensuring high-quality physical guidance during online execution. Finally, to ensure generalization in unseen environments with limited data, we develop a lightweight finetuning approach that achieves performance comparable to full-parameter finetuning with minimal adjustments. --- ### Look Inside Each Video: Rethinking How Video Anomaly Detection Is Evaluated Authors: Inpyo Song, Jangwon Lee Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21854v2 Video anomaly detection (VAD) aims to localize anomalous events by identifying when they occur. A common evaluation pools frame-level anomaly scores across all test videos and measures whether anomalous frames receive higher scores than normal frames (Micro-AUROC). This pooling creates comparisons both within the same video and across different videos. However, the pooled evaluation protocol commonly used in current VAD benchmarks makes Micro-AUROC heavily dependent on cross-video comparisons, which do not directly test whether anomalous frames rank above normal frames within the same video. To quantify this imbalance, we decompose Micro-AUROC into within-video and cross-video comparisons. Across four widely used weakly supervised VAD benchmarks, within-video pairs account for only 0.071-0.388% of all anomalous-normal frame pairs. This imbalance can make Micro-AUROC a poor indicator of within-video ranking. Video-level classifiers that assign a single score to each video achieve 81.40-97.18% Micro-AUROC while obtaining 50% Within-AUROC. For existing weakly supervised VAD detectors, replacing all frame scores in each video with their mean sets Within-AUROC to 50%. Across 120 runs, this replacement retains a median 98.8% of the original Micro-AUROC margin above chance. Because the decomposition depends on the evaluation protocol rather than the training supervision, the same distinction applies whenever VAD frame scores are pooled across videos. High Micro-AUROC alone therefore does not establish temporal localization. Our analysis reveals a limitation of pooled VAD evaluation and motivates reporting within-video ranking alongside Micro-AUROC. --- ### PUMA: A Polish Benchmark for Culturally Grounded Multimodal Understanding Authors: Sławomir Dadas, Michał Perełkiewicz, Rafał Poświata, Małgorzata Grębowiec, Bartłomiej Jaworski, Izabela Woźniakowska Published: 2026-08-22 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.21853v1 Large language models are increasingly moving beyond text processing, adding support for other modalities such as images and audio. While text understanding and generation have been extensively studied, multimodal data processing capabilities, particularly in the context of cultures and languages other than English, have not yet been evaluated comprehensively. In this paper, we propose PUMA (Polish Unified Multimodal Assessment), a novel benchmark of 900 hand-crafted tasks designed to probe the limits of multimodal models in the Polish cultural and linguistic context. The dataset evaluates both cultural understanding and practical skill in processing text, images, audio, and visually rich documents. Our extensive evaluation of frontier commercial models, open-weights models, and specialized smaller systems highlights a significant performance gap. While top commercial models achieve high scores in visual question answering, most models struggle with complex audio or document understanding. We open-source our evaluation framework to advance localized multimodal AI research. --- ### GaussVid: Sparse-View Gaussian Splatting with 3D-Aware Video Diffusion Priors Authors: Xinhui Liu, Can Wang, Wei Jiang, Wei Wang, Dong Xu Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21849v1 3D Gaussian Splatting (3DGS) has achieved remarkable success in novel view synthesis; however, reconstructions under sparse views often exhibit noticeable artifacts. While recent video diffusion models provide strong spatio-temporal priors for 3DGS restoration, directly fine-tuning them for restoration is suboptimal, as they lack awareness of the underlying multi-camera geometry, resulting in multi-view inconsistencies. In this work, we propose a novel 3D-aware video restoration framework designed to enhance the quality of sparse 3DGS reconstruction. Specifically, we construct a large-scale 3DGS video dataset to enable specialized fine-tuning. To bridge the gap between 2D video generation and 3D multi-view constraints, we introduce a camera-conditioned geometric prior. By using the first and last frames as boundary anchors and encoding the corresponding camera relationships, we explicitly inject spatial structure into the video generation pipeline. This boundary-anchored, camera-aware prior guides the network toward geometrically grounded restoration that remains coherent across viewpoints. Extensive experiments show that, among video-prior restoration methods, our approach attains the best pixel- and structure-level fidelity (PSNR/SSIM) and improves multi-view consistency, while remaining competitive in perceptual quality (LPIPS). --- ### BC-IHV: Conditioning the Color Space for Stable Rectified-Flow Low-Light Enhancement Authors: Yi Ai, Zheng Chen, Yuanhao Cai, Yulun Zhang, Xiaokang Yang Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21847v1 Low-light image enhancement (LLIE) must correct ambiguous exposure without overwriting structure already supported by the input. Generative transport can model exposure ambiguity; however, its flexibility may also alter observable geometry and chromatic content. Moreover, fixed invertible color coordinates are usually treated only as representations, although their inverse mappings reshape the RGB-domain gradients received by the enhancement network. To address these issues, we propose Structure-Anchored Rectified Flow (SA-RF), which maintains correspondence through separate chromaticity/intensity stems, a scale-matched condition pyramid, and HybridAda. HybridAda assigns location-specific retrieval to spatial cross-attention and global exposure modulation to pooled AdaLN. We further introduce BC-IHV, a learnable Box--Cox polar color space whose analytically invertible intensity mapping controls the inverse-gradient dynamic range through a single exponent. This allows the representation to balance dark-range expansion and gradient conditioning instead of adopting a fixed linear or logarithmic law. Experiments on three LOL benchmarks, blind image-quality evaluation, and cross-dataset tests demonstrate consistent reconstruction and perceptual advantages over the sota. Controlled studies further support the effectiveness of both the proposed framework and color representation. --- ### AI Watchdog: Agent Interfaces for Detecting and Defending Against Manipulative Dark Patterns in AI Conversations Authors: Rachel Poonsiriwong, Chayapatr, Archiwaranguprok, Constanze Albrecht, Monchai Lertsutthiwong, Pattie Maes, Pat Pataranutaporn Published: 2026-08-22 Categories: cs.AI, cs.HC arXiv: https://arxiv.org/abs/2608.21841v1 Conversational AI increasingly shapes consequential decisions, yet users have limited support for recognizing and resisting manipulation. We present AI Watchdog, a browser-based agent interface that monitors live conversations, detects five dark-pattern categories, including sycophancy, brand bias, anthropomorphization, sneaking, and harmful generation, and alerts users when they occur. Its open-weight turn-level classifier supports independent deployment and a path toward local inference, preserving user privacy while remaining separate from the conversational AI. We evaluated AI Watchdog in a preregistered, five-condition between-subjects experiment (N = 150) comparing a no-intervention control with four configurations varying nudge timing (prebunking vs. just-in-time) and engagement mode (without vs. with cognitive forcing). Results show that participants rarely flagged manipulative turns across all conditions, and post-task awareness did not differ significantly across groups. However, just-in-time warnings without cognitive forcing were the only intervention to significantly reduce compliance with AI-steered recommendations containing dark patterns, lowering compliance from 71.7% to 53.7%, an 18 percentage-point reduction. Exploratory analyses further showed that lower misinformation susceptibility was associated with greater flagging but not lower compliance, while higher AI trust was associated with greater compliance and lower reported awareness. Together, these findings suggest that explicit recognition of conversational dark patterns and behavioral resistance to AI steering may be distinct outcomes, motivating further investigation of timely, low-friction defensive interfaces. --- ### More Experts, Worse Dynamics: Inverse Scaling and Spectral Bias in Mixture-of-Experts State-Space Models Authors: Chandresh Pandey Published: 2026-08-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.21840v1 Mixture-of-Experts (MoE) architectures are commonly motivated as a way to increase expressivity by decomposing complex systems into simpler local dynamics. This intuition has recently been extended to spectral state-space models, where mixing stable operators is assumed to enable adaptation to heterogeneous or regime-switching time series. We critically evaluate this assumption in a controlled synthetic setting designed to isolate dynamical rather than representational challenges. We study a next-step prediction task on sequences composed of three regimes: chaotic dynamics generated by the Mackey-Glass system, a stable oscillatory regime, and a noise-dominated autoregressive regime. Across extensive ablations including capacity scaling, oracle routing, frozen-expert variants, and comparisons to output-level MoE baselines, operator-level mixture models consistently fail to outperform a single-expert baseline. Increasing the number of experts leads to inverse scaling, routing collapses or fails to induce meaningful specialization, and even perfect regime supervision does not prevent degradation in global performance. Furthermore, we show that apparent improvements in mean squared error on chaotic trajectories can be misleading. Phase-space analysis reveals that lower error often arises from temporal smoothing that destroys the geometry of the underlying attractor rather than from faithful modeling of the dynamics. These results identify a likely limitation of operator interpolation under the studied parameterization and training protocol, and underscore the need for geometry-aware evaluation when assessing regime-switching dynamical systems. --- ### FIRM-Video: Check Before You Score for Reliable Text-to-Video Reward Modeling Authors: Peiyuan Zhang, Xiangyu Zhao, Hongbo Liu, Xiaoxing Hu, Mingxin Liu, Shuran Ma, Yunhang Shen, Jian Hu, Haihan Gao, Haoyu Cao, Xue Yang Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21839v1 Reliable reward models are essential for text-to-video evaluation and alignment. However, the trade-off between evaluation accuracy and inference efficiency places high demands on the quality of training supervision. Existing approaches often rely on holistic judges with fixed rubrics or open-ended reasoning, leading to incomplete inspection, unfaithful justification, and entangled attribution. We introduce FIRM-Video, a unified checklist-driven data construction framework based on a check-before-score principle: construct dimension-specific checklists, verify each criterion against temporal visual evidence, and aggregate only verified decisions. For Instruction Following, FIRM-Video decomposes prompts into weighted atomic requirements; for World Coherence, it constructs prompt-calibrated, target-specific checks grounded in visible entities and actions; and for Perceptual Quality, it applies a generic taxonomy of visual defects. The verified criteria and scores are further transformed into natural-language analyses for end-to-end reward modeling. Subsequently, we construct FIRM-Video-90K with 88,044 dimension-specific instances from 29,348 videos, and introduce FIRM-Video-Bench with 750 point-wise human annotations across 250 videos. The Qwen3-VL-based FIRM-Video-8B achieves the best overall MAE on FIRM-Video-Bench while consistently delivering the highest VBench Total, Quality, and Semantic Scores in Best-of-8 sampling across three video generators. --- ### Towards Bitstream-corrupted Harsh Visual Understanding: Through Bitstream Language Modeling as Robust Semantic Priors Authors: Chaoran Huang, Fangcheng Li, Tianyi Liu, Wenyang Liu, Kejun Wu Published: 2026-08-22 Categories: cs.CV, cs.MM arXiv: https://arxiv.org/abs/2608.21837v1 Bitstream-corrupted Harsh Visual Understanding (BcHVU) aims to understand harshly degraded videos originally decoded from a severely corrupted bitstream in real-world multimedia communication. The ill-posed nature of BcHVU poses a major challenge for existing vision models, as even subtle bitstream corruption can lead to irreversible pixel distortion and significant semantic loss. To address these challenges in BcHVU, we propose Bitstream Language Modeling as Robust Semantic Priors (BLMSP), a framework for learning and injecting bitstream-native semantic cues. Our proposed BLMSP framework learns to extract bitstream-native semantic cues by bitstream language modeling, and leverages them as priors by injecting into off-the-shelf vision models of BcHVU tasks. Specifically, we present a Video Bitstream Byte Model (VBBM) that integrates byte-level modeling and cross-codec semantic distillation, enabling it to interpret robust semantics from byte sequences in multiple corrupted bitstream formats. The learned bitstream semantics are leveraged as robust priors and fused into BcHVU model backbones for improving the quality of video restoration, captioning, and human pose estimation. To train BLMSP, we construct a large-scale multi-source Corrupted-bitstream Harsh-video Paired (CHP) dataset containing 607k corrupted bitstream segments and 287k paired harsh video clips. Extensive experimental results show that the learned bitstream priors improve video restoration, captioning, and human pose estimation by 2.51 dB in PSNR, 0.20 in CIDEr, and 0.18 in PCK@0.2 on average, respectively. These results demonstrate that corrupted bitstream can serve as robust semantic priors in solving pixel distortion and semantic loss in BcHVU. --- ### LLM4LLM: Bridging Kernel Benchmarks and Real Deployment via Closed-Loop Agentic Optimization Authors: Hui Zeng, Pengfei Yang, Yanxin Chen, Fusong Ju, Xinran Wei Published: 2026-08-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.21836v1 Large language models have become increasingly capable agents for low-level code and kernel optimization, but isolated kernel benchmarks provide only a proxy for the deployment behavior that matters in language-model inference. We identify a benchmark-to-deployment gap: candidate kernels that appear correct and fast in standalone harnesses can exhibit different performance, safety, or phase behavior after integration into a real inference workload. We introduce LLM4LLM, a deployment-aware closed-loop optimization framework that starts from a target inference script, extracts phase-aware optimization tasks, searches with an experience-guided episodic agent, and accepts patches through in-model validation. Across ten language-model inference workloads on A100 and H100 GPUs, LLM4LLM improves end-to-end latency for every evaluated model, achieving 3.91$\times$/6.98$\times$ geometric-mean speedups on A100/H100; as supporting kernel-level evidence, it also attains up to 2.745$\times$ GeoMean speedup on KernelBench Level 2. --- ### GameXpert-Bench: How Far Are Coding Agents from Expert Game Development? Authors: Kun Chen, Haorong Hong, Peizhong Gao, Jianfeng Lin, Tongxu Luo, Yuxuan Xie, Chenxu Liu, Jieling He, Zhongyuan Liu, Zeno Zeng Published: 2026-08-22 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2608.21833v1 Recent large language models (LLMs) can operate as coding agents that build complete games from natural language requests. Game development is especially demanding because program logic, visual and audio content, interfaces, interaction and playability must function together in one executable artifact. Measuring this capability therefore requires evaluation of both game product and the development process. Existing benchmarks often assess the game development capabilities of LLMs by evaluating the final artifact or an isolated development stage. Our analysis of complete human-agent development trajectories identifies three stages that together span the lifecycle of game development with a coding agent: initial game generation, bug diagnosis and repair, and optimization over multiple turns. Therefore, we introduce GameXpert-Bench, which operationalizes the three lifecycle stages as three complementary benchmark tracks. GameGen evaluates complete game creation from a single request in an empty workspace. GameFix evaluates diagnosis and repair when defects are reported or left for the agent to discover. GameOpt evaluates cumulative optimization through request chains seeded by real development trajectories between users and agents. We evaluate each track using live game interaction, deterministic behavioral tests, or final product criteria with regression checks. The suite contains 97 generation tasks across 11 genres; 100 repair tasks from 50 game levels verified by humans, each with 19-27 injected bugs; and 17 optimization chains with six turns and 102 requests. Across the three tracks, current agents are more reliable at producing playable foundations and implementing explicit requirements than at discovering defects, verifying runtime behavior, and preserving functionality across changes. --- ### GUI-Primitives: Diagnosing Spatial Reasoning Failures in Vision-Language GUI Grounding Authors: Md Abrar Jahin, Md Rizwan Parvez Published: 2026-08-22 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.21832v2 Computer-use agents ground natural-language instructions in screenshots to locate interface elements, yet existing benchmarks do not isolate whether models bind relational language to the correct element. We introduce GUI-Primitives, a 994-item benchmark of contrastive instruction pairs over seven spatial relations in graphical user interfaces (left/right, above/below, containment, alignment, proximity, list ordinal, occlusion). Each pair holds the screenshot and anchor fixed while changing the relation expression, so the correct target moves between two designated candidates. Five annotators validate a 196-item subset ($κ= 0.94$ well-formedness; $κ= 0.79$ target selection). Nineteen vision-language models reach at most $32\%$ strict point-in-box accuracy. Because models emit unconstrained coordinates, we classify each prediction by the candidate region it falls within. Predictions fall outside both candidates on $60-92\%$ of items. Conditional on falling within a candidate region, target selection reaches 0.82-0.90 for horizontal position, vertical position, proximity, and list ordinal, but does not differ significantly from 0.50 for containment and occlusion: most failures reflect candidate localization rather than relation understanding. Across ten models, benchmark accuracy correlates with ScreenSpot-Pro accuracy (Spearman $ρ= +0.74$), an exploratory association at this sample size. Marking the two designated candidates raises selection accuracy by 35--57 percentage points, an oracle diagnostic that supplies the candidate set rather than a deployable method. We release the benchmark, predictions, and code. --- ### Beyond Success and Failure: Length-Aware Contrastive Learning for GUI Agents Authors: Chengyang Gu, Le Zhang, Jingbo Zhou, Yize Chen, Yu Shi, Siqi Bao, Zheng-Fan Wu, Hua Wu, Hui Xiong Published: 2026-08-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.21830v1 Graphical User Interface (GUI) agents powered by Multimodal Large Language Models (MLLMs) have shown strong potential for automating tasks across diverse digital environments, where reinforcement learning (RL) has become a dominant training paradigm. However, widely used methods such as Group Relative Policy Optimization (GRPO) suffer from reward-gradient misalignment, leading to inefficient and unstable optimization. Recent work addresses this issue by reformulating RL with verifiable rewards (RLVR) as contrastive or classification-based objectives, which improve stability by eliminating problematic gradient behaviors. Despite this progress, existing contrastive RLVR methods rely primarily on outcome-level supervision and fail to capture fine-grained differences in trajectory quality within the same outcome category. In this paper, we propose Length-Aware Contrastive Learning for GUI Agents (LACL-GUI), a contrastive RLVR framework that incorporates trajectory-level quality signals into policy optimization. LACL-GUI introduces structured preferences within both successful and failed trajectories, encouraging concise successful executions and differentiating failure quality based on divergence from successful trajectories, while preserving optimization stability. Experiments on GUI agent benchmarks show that LACL-GUI provides more effective learning signals and consistently improves agent performance over prior methods, highlighting the value of trajectory-level supervision in contrastive RLVR. --- ### Training a Knowledge Base: Supervised Structure Learning for Agent-Curated Document Stores Authors: Yu Pan, Hongfeng Yu Published: 2026-08-22 Categories: cs.CL, cs.AI, cs.IR arXiv: https://arxiv.org/abs/2608.21829v2 Retrieval-augmented generation treats the document store as a frozen input, and the offline pipelines that do build structure over it build it unsupervised -- a whole corpus indexed at uniform effort, with no signal about which structure a question will need. We instead treat the knowledge base as a non-parametric model trained on (question, answer) pairs: a curator agent answers a supervised question against the current store, is shown the gold answer, then edits the store. The store carries forward, and we evaluate the curated store with a test set, on two contamination-free benchmarks: KBGym, a fictional-universe generator we release, and PhantomWiki. Generalization is probed with four question groups of decreasing overlap with the training set: the trained questions themselves, and unseen questions sharing both of their keys with training, one key, or neither. The curated store's advantage grows with overlap -- from parity where no key was shared, through +0.176 F1 where both keys were, to 25% fewer actions at +0.294 F1 on the trained questions, the one cell significant on both benchmarks -- while matching HippoRAG's gains with 1,913 links against its 196,112: per point of corpus covered, 1.5x the action saving and 2.1x the accuracy gain. Accuracy rises steadily with the share of the corpus the indexes cover, so training on more questions widens coverage, and with it the generalization. --- ### Towards Alias-Free 4D Gaussian Representations with Motion-Aware Filtering Authors: Ankit Dhiman, Kunal A Kathare, Pranav Vignesh, Lokesh R Boregowda, Venkatesh Babu Radhakrishnan Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21828v1 Novel-view synthesis of dynamic scenes, crucial for AR/VR applications, remains a challenging problem. Recent methods adapt representations like 3D Gaussian Splatting (3DGS) and Neural Radiance Fields (NeRF) for dynamic scenes by incorporating time as the fourth dimension (4D representations). These 4D representations still suffer from aliasing artifacts, especially when generating novel views from divergent viewpoints (zoom-in/zoom-out operations). While using 3D smoothing filters like those proposed in Mip-Splatting might seem like a possible solution, they fail to account for local motion and also exhibit aliasing. To address this, we propose a motion-aware 3D smoothing filter specifically designed for 4D representations. Our approach adapts the filter strength based on local motion information, effectively mitigating aliasing without compromising rendering quality. This is achieved by estimating the joint density function of time and focal-to-depth ratio using a non-parametric estimation method. During inference, we sample from this joint distribution to determine the appropriate smoothing filter. This flexible strategy can be integrated with various 4D representations. Our evaluations on standard datasets demonstrate superior performance compared to state-of-the-art methods. --- ### Do Large Language Models Perform Well on Comprehending Poetic Logic in Modern Chinese Poetry? Authors: Tian Lan, Shanshan Wang, Zehua Duo, Jiang Li, Guanglai Gao, Derek F. Wong, Xiangdong Su Published: 2026-08-22 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.21827v1 Large Language Models (LLMs) have achieved significant progress across a wide range of natural language processing (NLP) tasks, yet their ability to understand literary texts, particularly modern Chinese poetry, remains largely unexplored. The unique literary characteristics of modern Chinese poetry necessitate a distinct form of reasoning for effective comprehension. Unlike conventional texts that convey clear information, the unique "poetic logic" of modern Chinese poetry requires a holistic reasoning approach that goes beyond superficial semantic analysis to be understood. However, current evaluation paradigms largely ignore this critical dimension. To address this gap, we propose Peony, the first benchmark specifically designed for evaluating the poetic logic of modern Chinese poetry. We define poetic logic as four tasks across three levels, namely stanza, line, and imagery, and systematically evaluate and analyze six mainstream LLMs based on Peony. We evaluate these models under both non-thinking and thinking configurations. The experimental results reveal the limitations of current LLMs in understanding the poetic logic of modern Chinese poetry and validate the effectiveness and necessity of Peony. Our data and code will be available. --- ### VisAdj: Learning Adjacency Matrices from Node-Link Images Authors: Jiahao Xie, Guangmo Tong Published: 2026-08-22 Categories: cs.AI, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2608.21825v1 Learning adjacency matrices from node-link images is a fundamental problem for recovering structured graph information from visual observations. Existing methods typically rely on fixed KNN-based heuristics for candidate edge selection and fail to capture dependencies among edges. To overcome these limitations, we propose VisAdj, a new framework for topology-aware adjacency prediction. VisAdj introduces an attention-sparse neighbor sampler to adaptively select a high-recall set of candidate node pairs and performs joint edge inference using a line-graph transformer that treats candidate edges as tokens and explicitly models dependencies among incident edges. Extensive experiments on synthetic graphs, road networks, and vessel images demonstrate that VisAdj consistently outperforms existing baselines by clear margins. --- ### Convergence in Science, Divergence in Religion: Calibrated Framing Differences Across Wikipedia's Language Editions Authors: Hung-Hsuan Chen Published: 2026-08-22 Categories: cs.CL, cs.CY, cs.SI arXiv: https://arxiv.org/abs/2608.21821v1 When Wikipedia's language editions describe the same concept, how differently do they frame it? Prior work measures coverage gaps between editions; we measure framing distance for matched concepts. We analyze 2,799 valid articles from 3,000 possible concept-language observations, spanning 150 Wikidata-anchored concepts, 20 language editions, 4 domains, and a calibration set. Raw embedding distances reflect both content differences and how well the encoder aligns each language pair. Even among calibration concepts with stable cross-cultural denotations (e.g., chemical elements, numbers, colors), the largest language-pair mean distance is 3.6 times the smallest, and distances are typically smaller within language families. We define a baseline-adjusted distance (calibrated distance): the distance between two language versions of a concept minus the mean distance for calibration concepts in the same language pair. This adjustment substantially reduces pair-specific alignment differences and the language-family pattern. Across three multilingual encoders (LaBSE, multilingual MPNet, and CMLM), scientific articles align more closely than calibration articles, and all three rank religion first and science/technology last. Concept-level rankings are highly consistent across encoders (Spearman rho=0.75-0.79 for MPNet and CMLM relative to LaBSE). Religion lies significantly above the calibration baseline under LaBSE. Within politics, divergence concentrates on concepts such as censorship and refugee, while democracy and human rights are among the most aligned. Code, data, and per-language-pair calibration baselines are released.\footnote{https://github.com/hhchen1105/cross-linqual-concept} --- ### PatchGate: Narrowing the Verbalization Gap with Intrinsic Object Inventories in Frozen Vision-Language Models Authors: Jihyung Ko, Eunji Jung, Hyeongsub Kim, Ziseok Lee, Jae Won Cho, Sanghyun Jo, Kyungsu Kim Published: 2026-08-22 Categories: cs.CV, cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2608.21819v1 Reliable image captioning in Vision-Language Models (VLMs) requires captions to be both precise and complete, avoiding unsupported object mentions while covering visible objects. Existing training-free methods primarily address the former requirement, suppressing unsupported object words by intervening on model-predicted mentions during generation. Because they operate only on objects the model is already likely to mention, visible objects omitted from the output remain difficult to recover. We propose PatchGate, a training-free framework that extracts prompt-free object evidence intrinsic to a frozen VLM before generation and uses it to narrow the gap between an intrinsic object set and final object mentions. In the first stage, Visual Evidence eXtraction (VEX) reads patch-level lexical evidence from the latter half of LM decoder layers and constructs an image-conditioned object set without any task prompt. In the second stage, Visual-Evidence Inclusion-Exclusion Decoding (VIED) uses this object evidence to calibrate decoding logits, promoting evidence-supported but under-verbalized objects and suppressing weakly supported but over-verbalized objects. On AMBER, PatchGate improves both sides of object-level reliability, increasing visible-object coverage from 49.4 to 56.0 (+13.4%) and reducing object hallucination by lowering CHAIR from 7.5 to 6.6 (-12.0%), without external detectors or fine-tuning and with one extra forward pass. --- ### A Physics-informed Neural Network Approach for Robust Buckling Load Prediction and Reliability-Based Design of Thin Truncated Conical Shells Authors: Devasmit Dutta, Budhaditya De, Rohan Majumder, Sudip Kumar Mishra Published: 2026-08-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.21818v1 Thin-walled truncated conical shells are widely used in aerospace, marine, offshore, and lightweight infrastructure systems due to their high strength-to-weight ratio and geometric efficiency. Their buckling resistance under axial compression, however, is highly sensitive to geometric imperfections, manufacturing tolerances, material variability, and nonlinear instability effects. Conventional design procedures rely on conservative knockdown factors (KDFs), such as those recommended in NASA SP-8019, which do not explicitly account for shell geometry, fabrication quality, data uncertainty, or target reliability. This study develops a physics-informed neural network (PiNN) framework for predicting critical buckling loads of thin truncated conical shells and integrates the trained surrogate within a reliability-based design (RBD) formulation. The model combines geometric and material descriptors with mechanics-informed features derived from shell stability theory and the localized reduced stiffness method (LRSM). A physics-informed loss function penalizes mechanically inadmissible predictions exceeding the theoretical elastic buckling load. The framework is trained and evaluated using 133 experimental Mylar conical shell tests under axial compression. Compared with a conventional deep neural network (DNN), the PiNN improves predictive accuracy, reduces mean absolute error, and enhances physical consistency. The trained PiNN is then used to evaluate reliability indices and calibrate safety-consistent KDFs for prescribed target reliability levels. Results demonstrate that the PiNN-RBD framework provides an efficient approach for uncertainty-aware design of imperfection-sensitive shell structures. --- ### Resilient Concurrent Causal Discovery for Topological Event Sequences Authors: Jiyu Tian, Junhao Dong, Mingchu Li, Lingling Fang, Liming Chen, Andreas Holzinger, Zheng Yan, Yew Soon Ong Published: 2026-08-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.21815v1 Causal discovery on topological event sequences is crucial for ensuring the reliability of networks. However, existing methods struggle to capture the complex causal relationships arising from concurrent events and lack robustness to incomplete event sequences. To address these issues, we propose a resilient concurrent causal discovery method, termed RCCD, enabling robust learning of causal graphs from topological event sequences. Specifically, we first introduce an influence-aware hyperedge causal attention mechanism, which incorporates event duration into the embedding representation, aggregates concurrent event features via hyperedge causal convolution, and injects network prior knowledge to capture the complex many-to-one causal interactions. Furthermore, we design a masked-based alternating causal optimization framework, which forces the model to recover masked event types based on context through self-supervised mask reconstruction, thereby enhancing the resilience of the predictor to missing data. To validate the effectiveness of our method, we conduct extensive experiments on both simulated and real-world telecommunication network datasets. Experimental results demonstrate that the proposed method significantly outperforms existing state-of-the-art methods in both accuracy and robustness, making it more suitable for real-world telecommunication network environments. ---