# Research Papers — 2026-08 · bundle page 176 of 240 > Full-text markdown of papers #8751–8800 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/176 [← Previous bundle](https://aipod.dev/papers/archive/2026-08/page/175.md) [Next bundle →](https://aipod.dev/papers/archive/2026-08/page/177.md) ### SCOUT: Self-Checking and Recovery-Aware Tool-Thought Agents for Ultra-Long Egocentric Video Reasoning Authors: Keyang Zhong, Kuo Wang, Peng Liu, Quanlong Zheng, Junlin Xie, Zhijia Liang, Yanhao Zhang, Guanbin Li Published: 2026-08-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.07959v1 Ultra-long egocentric video understanding requires reasoning over temporally sparse evidence distributed across hours or days, challenging current multimodal models with limited context and the grounding of key video segments. While Chain-of-Tool-Thought (CoTT) agent systems enable iterative retrieval and inspection, they suffer from error propagation due to rigid zoom-in strategies that lack recovery mechanisms. In this work, we address these challenges through SCOUT (Self-Checking Chain-Of-Tool-thought), a recovery-aware agentic framework introducing an adaptive policy that evaluates intermediate tool observations and dynamically trades off exploitation (zoom-in) and exploration (region switching), enabling robust multi-hop reasoning over extremely long horizons. However, training such multi-turn tool-using agents remains challenging, as existing RL methods rely on sparse outcome-level rewards and lack supervision over extended decision trajectories, resulting in suboptimal credit assignment for long-horizon reasoning. To address this, we develop UPS-GRPO, an uncertainty-prioritized policy optimization method that concentrates exploration on high-uncertainty post-tool states while preserving sample efficiency. We further introduce a turn-level advantage decomposition that integrates outcome rewards with tool-grounded temporal alignment rewards for improved credit assignment. Experiments show that SCOUT achieves state-of-the-art results on ultra-long egocentric benchmarks, while remaining competitive on shorter-horizon long-video settings. --- ### LIBAD: A Multimodal Anomaly Detection Benchmark for Li-Ion Battery Electrode Manufacturing Authors: Wenbo Sui, Daniel Lichau, Harold Phelippeau, Zhao Liu Published: 2026-08-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.07958v1 Multimodal industrial anomaly detection has largely focused on discrete products using strongly correlated RGB and 3D observations, leaving continuous process manufacturing and weakly correlated sensing modalities underexplored. We introduce LIBAD, the first multimodal anomaly detection benchmark for Li-ion battery electrode manufacturing. Collected from real roll-to-roll production lines, LIBAD provides aligned double-sided visible-light imaging, high-resolution X-ray radiography, and inline-compatible low-resolution X-ray radiography. Electrode patches in LIBAD exhibit highly homogeneous material appearance, while defect evidence can be strong in one modality but weak or absent in another, resulting in pronounced cross-modal anomaly inconsistency. Benchmarks of representative methods under the inline-compatible visible-light and low-resolution X-ray setting exhibit limited transferability and consistently high false-positive rates. We therefore propose DA-Core, a memory-based method that jointly considers feature-space coverage and local density of normal features during coreset selection, allowing compact memory banks to better preserve fine-grained normal variations. With a coreset ratio of 0.05, DA-Core reduces FPR95 from 60.4% to 54.3% compared with standard farthest point sampling. At this ratio, DA-Core also outperforms the best standard coreset result (obtained at 0.20) while reducing inference time by 43.9%. These results suggest that both the data distribution of normal features and the modality relationship itself require explicit consideration when designing anomaly detection methods for process manufacturing. --- ### Self-Evolving Neuro-Symbolic Skills for Tool-Augmented Spatial Reasoning Authors: Shi-Yu Tian, Zhuo-Xia Wang, Xuan-Yi Zhu, Zhi Zhou, Xinwei Yang, Kun-Yang Yu, Ming Yang, Yang Chen, Yu-Feng Li Published: 2026-08-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.07955v1 Large vision-language models have achieved strong performance in multimodal reasoning, but they remain unreliable on fine-grained spatial tasks that demand both precise spatial perception and fine-grained geometric computation beyond end-to-end generation. Tool augmentation offers a natural solution, while existing methods either plan tool calls from scratch without explicit dependency constraints or rely on fixed pipelines that are redundant and generalize poorly across spatial tasks. An effective spatial reasoning agent should instead accumulate reusable experience and adaptively compose it for new problems. To this end, we propose NeSy-Spatial, a neuro-symbolic framework for self-evolving spatial skills. NeSy-Spatial abstracts tool interactions and geometric operations into typed executable atomic instructions and composes them into two complementary skill types: Tool-Use Skills for organizing tool execution and Geometry Skills for structured geometric reasoning. During inference, NeSy-Spatial retrieves and executes relevant skills in a closed-loop process. During evolution, it analyzes buffered successful and failed trajectories to refine skill structures and prune unreliable or inactive entries. Experiments on three spatial reasoning benchmarks show that NeSy-Spatial consistently improves reasoning accuracy with more precise tool utilization. --- ### KGCache: Amortized Subgraph Retrieval for KG Reasoning with LLMs Authors: Uros Stanic, Changcheng Yuan, Sabuj Laskar, Ariful Azad Published: 2026-08-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.07954v1 Large language models can answer knowledge-intensive questions more reliably when they are grounded with knowledge graphs, but systems such as Think-on-Graph and Reasoning-on-Graph repeatedly query the same graph neighborhoods across different questions. In this work, we study this repeated retrieval in Knowledge Graph Question Answering~(KGQA) workloads and propose KGCache, an in-memory cache for one-hop knowledge graph neighborhoods. KGCache is designed to be compatible with both iterative traversal (ToG) and one shot planning (RoG) KGQA paradigms. KGCache is placed between the KGQA engine and the backend serving the KG, so repeated entity requests can be served from cache instead of issuing new KG queries. We evaluate KGCache on WebQSP and CWQ using LRU, LFU, and a trace-aware Oracle policy. Our analysis shows that both datasets contain substantial entity reuse among starting entities and entities reached during traversal. We also explore semantic caching for similar queries, which shows additional hit-rate gains on WebQSP and needs further accuracy testing on CWQ. Entity caching accelerates KG retrieval by up to $1.91\times$, while semantic-context caching achieves up to $1.06\times$ full-system speedup in the evaluated WebQSP configurations, with each hit being up to $3.73\times$ faster. --- ### From Uncertainty to Failure Attribution: Self-Diagnosing Models for Failure Attribution under Distribution Shift Authors: Yiyao Yang Published: 2026-08-08 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.07953v2 Distribution shift poses a significant challenge to the robustness of machine learning models, but the current solutions only aim to detect out-of-distribution (OOD) samples and predict uncertainty levels. We introduce a problem setting for failure attribution under distribution shift, which enables the models not only to detect OOD samples, but also to find out the reason for their failure. The solution we propose is called self-diagnosing models, which are capable of jointly learning predictive output, predictive uncertainty, and a failure attribution signal. In particular, we use the failure attribution vector, produced by a neural network, which provides a structured representation of predictive unreliability by distinguishing four different types of failures: covariance shift, semantic shift, noise corruption, and adversarial perturbation. In other words, we move from scalar uncertainty towards failure identification. For training the model, we introduce a consistency regularizer that encourages consistency between uncertainty and failure attribution predictions. Moreover, to be able to evaluate the model on its ability to find the reasons for failure, we construct several distribution shift benchmarks with predefined mechanisms for generating distribution shifts. --- ### Persistent Semantic Entities in Tool-Augmented LLM Systems Authors: Zhaohui Wang Published: 2026-08-08 Categories: cs.LG, cs.AI, cs.CR, cs.SE arXiv: https://arxiv.org/abs/2608.07952v1 Tool-augmented LLM agents can harbor implicit state that persists across sessions, activates through events, and propagates across agent boundaries---largely invisible to standard debugging. We formalize this as Persistent Semantic Entities (PSEs): constructs defined by name binding, event triggering, and cross-boundary propagation, and evaluate them across 24 models from 11 families (1.5B--1T parameters). First, every tested model is susceptible (20--100% on the 20-model susceptibility panel), with name binding as the necessary and dominant mechanism: without it, contamination is 0%. Second, persistence depends on contamination type rather than scale or deployment: preference contamination persists undecayed on every model probed (100% at t=10) and instruction contamination persists wherever adopted, persona-style injection decays partially (90%$\to$10%), while factual injection is model-dependent---self-corrected on Llama-3.1-8B and GPT-4o-mini but held at ceiling on both Qwen2.5-coder variants, so we do not claim it self-corrects in general. The preference and instruction results hold across providers in our controlled setting. Third, context-isolated self-verification achieves 20--79% reduction (median 36.5%) without oracle references while keyword-based detection produces systematic false positives, and contamination compounds 1.9$\times$ along a four-stage agent pipeline (40%$\to$75%). Preference and instruction contamination---persistent, lacking self-correction, and poorly captured by standard monitoring---represent a particularly concerning attack surface for deployed agent systems. --- ### Guixu: Valuation-Driven Data Discovery for Autonomous AI Agents with On-Chain Attestation Authors: Yifan Wu, Yuchen Peng, Jiaqi Chai, Yufei Qian, Xilin Li, Ke Chen, Lidan Shou Published: 2026-08-08 Categories: cs.AI, cs.CR, cs.DB, cs.IR, cs.MA arXiv: https://arxiv.org/abs/2608.07949v1 Autonomous agents increasingly rely on external data to complete downstream tasks such as model training and decision support. However, existing data discovery systems remain largely retrieval-oriented: they surface candidate datasets from heterogeneous sources, but provide limited support for estimating task-specific utility, selecting cost-effective datasets under budget constraints, or incorporating trustworthy feedback from prior usage. This paper presents Guixu, a valuation-driven data discovery system for autonomous agents. Guixu employs a three-phase valuation pipeline with proxy-label propagation and multi-round knapsack optimization for task-aware data valuation. Guixu integrates agentic payment protocol to enable budget-constrained data procurement workflows. Guixu leverages on-chain data market and attestation signals for verifiable data discovery. Our demonstration highlights how Guixu enables an agent to move beyond keyword-based dataset retrieval toward task- and budget-aware, trustworthy data discovery and procurement. Attendees can interactively explore the full workflow, from NL task specification and multi-source search to data valuation and verifiable transaction feedback. --- ### SynVAR: Synergizing Spatial and Semantic Alignment in Visual Autoregressive Model Authors: Zhennan Chen, Tianxing Shi, Pengcheng Xu, Kepan Nan, Qian Wang, Zili Yi, Jian Yang, Ying Tai Published: 2026-08-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.07948v1 VAR has gained widespread popularity due to its next-scale prediction paradigm. However, it faces substantial performance bottlenecks when handling complex scenes with multiple objects and attributes. Existing diffusion-based enhancement methods fail to adequately address the unique challenge of cross-scale error propagation and accumulation in VAR. To this end, we propose SynVAR, the first training-free enhancement framework specifically tailored for the VAR paradigm, which introduces a spatial-semantic collaborative control strategy to effectively suppress propagation error and improve generation quality. SynVAR comprises three key components: (1) Global guidance to ensure reasonable spatial structure in the early stages, (2) Receptive field constraints to mitigate early-stage semantic confusion, (3) High-frequency compensation to recover fine-grained details. Extensive quantitative and qualitative experiments demonstrate the significant improvements in the ability of SynVAR to enhance the VAR's capability for complex scene modeling. --- ### Directed Neuro-Symbolic Stochastic Execution for Verification of Distributed Parallel AI Programs Authors: Gautham Koorma, Vikas Sharma, George Edwards, Mahdi Eslamimehr Published: 2026-08-08 Categories: cs.AI, cs.DC arXiv: https://arxiv.org/abs/2608.07947v1 Distributed parallel Artificial Intelligence (AI) programs expose reliability gaps that conventional testing cannot close: parallel executions are non-deterministic, and AI workloads bring high-dimensional inputs and non-linear operations that defeat fuzzing and symbolic execution in isolation. We present Directed Neuro-Symbolic Stochastic Execution (DNSSE), a hybrid testing framework that couples schedule prediction guided by a Large Language Model (LLM) with symbolic constraint solving and coverage-guided stochastic mutation. We model distributed AI executions as non-deterministic transition systems, specify correctness in linear temporal logic, and prove soundness, bounded completeness, and probabilistic completeness of the hybrid solver, together with an expected-cost analysis of LLM-guided schedule exploration. A scalable implementation on PyTorch and Ray detects 2.9% more concurrency bugs than the strongest baseline and raises average branch coverage from 68.6 % to 91.6 % across five realistic distributed AI benchmarks. --- ### Metadata Reconstruction from Values Alone: Recovering Column Semantics in Undocumented Warehouses Authors: Mike Helwig Published: 2026-08-08 Categories: cs.DB, cs.AI arXiv: https://arxiv.org/abs/2608.07946v1 Text-to-SQL benchmarks ship schemas whose column names already say what the columns mean. Production warehouses are the inverse: cryptic identifiers, partial or absent documentation. We address the problem they pose first: recovering what columns and values mean from the data itself. Rosetta places a language model inside a verification harness: a deterministic profiler extracts structural evidence (value fingerprints, a 26-pattern library, checksum verdicts), the model proposes semantics conditioned on that evidence, and every fact carries provenance and a confidence bounded by its evidence class. Against human documentation on 680 paired columns across eleven BIRD databases, identifiers destroyed, the harness delivers metadata that is 0.475 accurate on the 42% of columns it commits to, against 0.223 on 94% for the same model used directly. Restricted to the 283 columns where both arms speak, the harness writes no better prose than the model alone; the gain is selection: deterministic evidence governs whether the system speaks (coverage +0.257 (0.128, 0.378)), not how well. The deterministic layer is a competence detector, not a competence amplifier. A backbone swap bounds the claim: the prose finding reproduces, but prompt-requested abstention does not transfer; a code-enforced commit gate (predictions registered first; measured on a third backbone and held-out databases) makes no-evidence coverage 0.000 on every backbone. On a blind i2b2 clinical warehouse Rosetta decodes 95.5% of 134 real ICD-9 codes from values alone and abstains on all 44 NDC drug codes. The catalog supports calibrated abstention at query time: under full schema opacity a naive translator falls from 0.92 to 0.42 execution accuracy while our gate answers at 86% accuracy over 59% coverage. Negative results are reported plainly, including that our own authority ladder is not the mechanism behind the headline. --- ### ScaleSense: Cost-Intelligent Scaling Framework via Learned Resource Estimation in Alibaba AnalyticDB Authors: Yifan Wu, Yuhan Li, Zhenhua Wang, Ke Chen, Lidan Shou, Zonghao Chen, Liang Lin, Huan Li, Gang Chen Published: 2026-08-08 Categories: cs.DB, cs.AI, cs.DC arXiv: https://arxiv.org/abs/2608.07945v2 Cloud-native serverless data warehouses achieve fine-grained elasticity by decoupling storage from compute, yet determining the optimal resource allocation for highly heterogeneous ad-hoc queries remains a formidable industrial challenge. Our analysis of production workloads in Alibaba AnalyticDB exposes a costly ``provisioning trap'': the fear of catastrophic resource depletion drives users to blindly over-provision resources, wasting immense monetary budgets without alleviating non-CPU bottlenecks (e.g., I/O saturation). To break this impasse, we propose ScaleSense, a proactive, query-level resource scaling framework. Specifically, it features a multi-faceted query encoder that jointly models plan topologies and hardware specifications. Crucially, a quantile-based resource predictor estimates multi-dimensional physical footprints, acting as a reliable safety net for optimal resource scaling. An auto-scaling controller then navigates the performance-cost Pareto frontier, dynamically tailoring allocations to specific business priorities without requiring model retraining. Evaluations on over 1.36 million production queries show that ScaleSense achieves state-of-the-art prediction accuracy with good prediction interval coverage. By achieving a 76.7% relative improvement in optimal resource configuration selection over the best baseline, this approach addresses the critical performance-cost trade-off while maintaining low-overhead inference latency, confirming its practical performance in production deployments. Under the performance-optimization policy, ScaleSense satisfies user-defined performance requirements while reducing monetary cost by up to 5.22x. --- ### Locating Failure in Multi-Page Visually Rich Document Understanding: An Empirical Attribution Authors: Lewei Xu, Yihao Ding, Zihan Xu, Daniel Yitian Su, Daochang Liu, Siwen Luo, Yifan Peng, Wei Liu Published: 2026-08-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.07943v1 Multi-page visually-rich document understanding (MP-VRDU) requires managing evidence that is sparse, spread across pages, and often exceeds a model's context window. Prior work has produced competing, largely untested claims about how these systems should be built. We attribute incorrect answers to three failure modes, representation, selection, and reasoning, and isolate each over a multi-page document understanding dataset by intervening on one while holding the others fixed. We find that vision is necessary but does not replace text extraction, that missing pages bound accuracy while distractors cost little, and that reasoners fail to integrate evidence across pages even when it is fully supplied. Prompting can shift reasoning behaviour substantially, improving some outcomes at the expense of others. We translate these findings into guidance for building such systems under a fixed compute budget. --- ### LAD-COD: Language-Aligned Dense Perception for Camouflaged Object Detection Authors: Shangye Song, Tianzhi Zhu, Syed Ariff Syed Hesham, Xin He, Yun Liu Published: 2026-08-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.07941v1 Camouflaged object detection (COD) aims to segment objects that exhibit high visual similarity to their surroundings, which reduces foreground-background discriminability and weakens boundary evidence across appearance, texture, and structure. Such limitations motivate the use of instruction-conditioned semantics as top-down guidance for identifying which weak visual cues are relevant to the target. Recent segmentation systems built on large multimodal models (LMMs) demonstrate this possibility through instruction-conditioned target embeddings that guide mask decoding. However, in this language-to-mask paradigm, the generated target embedding conditions mainly the mask decoder, leaving the dense visual features that must preserve low-contrast boundaries and fine local structure without explicit guidance. We propose Language-Aligned Dense perception for COD (LAD-COD), a framework that aligns top-down semantic target guidance with bottom-up hierarchical visual features. Instead of fully adapting a large generic image encoder, LAD-COD learns a trainable hierarchical visual branch that captures camouflage-sensitive texture, boundary, and contextual information. To align these features with the target embedding, LAD-COD applies Language-Aligned Dual Visual Fusion (LADVF), which extends the embedding beyond sparse prompting to query patch-level language-aligned features and to gate their residual integration with the hierarchical features. This design allows semantic information to guide localization while preserving the fine structural details needed for camouflage segmentation. Experiments on CAMO, COD10K, and NC4K show that LAD-COD obtains the best reported value in all 12 dataset-metric comparisons. --- ### FlexSplat: Flexible Feed-Forward 3D Gaussian Splatting without Point Cloud Correspondence Authors: Amir Sabbaghziarani, Hanting Ye, Maria Gorlatova, Yi Ding Published: 2026-08-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.07937v1 We present FlexSplat, a feed-forward framework for novel view synthesis (NVS) from uncalibrated, object-centric multi-view image collections. A recent line of query-based methods reconstructs a compact set of 3D Gaussians by treating them as transformer queries that are refined with multi-view deformable attention; these methods, however, assume that camera poses are given. FlexSplat removes this assumption: a geometry transformer is trained jointly with the Gaussian decoder to predict per-image camera parameters and depth, which in turn ground a depth-guided Gaussian parameterization and a multi-view deformable cross-attention that aggregates evidence across all input views into a single, view-consistent set of primitives. An uncertainty-weighted depth-consistency objective lets the jointly trained geometry adapt to the reconstruction task, while the cross-view consensus formed during decoding absorbs the residual error of the estimated cameras and depth. The representation uses a compact Gaussian budget that is decoupled from the input resolution - unlike pixel-aligned methods, the primitive count does not grow with the image grid - and is not dictated by the number of views. On ShapeNet-SRN and Google Scanned Objects (GSO), FlexSplat matches or approaches posed state-of-the-art reconstructors while requiring neither camera poses nor ground-truth depth, and matches the best perceptual (LPIPS) quality among the compared methods on GSO. Our results indicate that a jointly trained geometry front-end is sufficient to bring calibration-free operation to query-based Gaussian reconstruction while staying within 0.7 dB PSNR of posed methods and matching their perceptual quality. --- ### Adaptive Supervised Anchoring for On-Policy Self-Distillation Authors: Meilin Yang, Zixuan Ding, Jianhao Nie, Weite Zhang, Yuxin Zhang, Zhiming Shao, Li Yu, Zhe Fu Published: 2026-08-08 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.07935v2 On-policy self-distillation (OPSD) adapts a language model by distilling guidance from a frozen teacher on trajectories sampled from the student. Its effectiveness, however, depends critically on the quality of those trajectories. We show that when student rollouts drift from target trajectories, conditioning the teacher on off-target prefixes substantially weakens its task-relevant supervision. Controlled prefix-corruption experiments expose this failure mode, which we term rollout-conditioned signal degradation. To address this problem, we propose a unified training framework that separates two complementary supervision pathways. The first retains rollout-conditioned distribution matching, providing guidance on states the student actually visits. The second applies supervised cross-entropy on canonical ground-truth contexts, avoiding the incompatibility of imposing target tokens on erroneous rollout prefixes. Token-level rollout-target alignment is used to adapt the strength of the canonical-context anchor, emphasizing it during cold start and relaxing it as rollout quality improves. Experiments across multiple model scales, two task families, and general-reasoning benchmarks show that the proposed approach improves task acquisition over OPSD while preserving general capabilities, resulting in a more favorable empirical plasticity-stability trade-off. These findings identify context quality as a central bottleneck in on-policy self-distillation and demonstrate the value of separating rollout-conditioned guidance from canonical supervision. --- ### EvoTrustRAG: Evolution-Aware Conflict Attribution and Evidence Handling for Reliable Retrieval-Augmented Generation Authors: Xi Nie, Hongwei Li, Shenghao Wu, Wenshu Fan, Qiyang Song, Wenbo Jiang Published: 2026-08-08 Categories: cs.CR, cs.CL arXiv: https://arxiv.org/abs/2608.07933v1 Retrieval-Augmented Generation (RAG) improves the factuality of large language models with external knowledge, yet conflicting evidence remains a fundamental challenge in dynamic and adversarial environments. Existing approaches often treat conflicts as static inconsistencies and select more reliable knowledge, overlooking that the same conflict may arise from legitimate knowledge evolution, malicious manipulation, or unresolved uncertainty. We formulate conflict origin attribution as a new problem in RAG: identifying which explanation of conflicting evidence is supported by observable context rather than simply which fact should be trusted. We propose EvoTrustRAG, a training-free framework for evolution-aware conflict attribution and evidence handling before answer generation. EvoTrustRAG represents span-grounded retrieved facts as a conflict evidence graph, evaluates grounded evolution and directional intervention hypotheses using temporal relations, support structure, and auxiliary consistency, and projects local decisions onto a globally consistent explanation of each conflict group. The attribution determines whether earlier and later states are preserved as temporal knowledge, an intervention candidate is separated from the primary context, or an unresolved conflict remains visible to the generator. Unlike provenance-based approaches focused on post-hoc analysis, EvoTrustRAG determines during inference whether conflicting evidence follows plausible knowledge evolution, exhibits intervention-like support, or cannot be reliably attributed. Experiments show that EvoTrustRAG achieves 81.4% average accuracy on benchmark-native conflict settings, improves attribution macro-F1 from 72.2% to 79.1% over the strongest baseline, and reduces the error rate under the strongest coordinated attack from 31.2% to 16.0%. --- ### SportsGrounder: Proposal-Aided Interleaved Grounding for Dense Sports Video Reasoning Authors: Yizhi Li, Jiawei Jiang, Guanhong Wang, Yingcai Wu, Gaoang Wang Published: 2026-08-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.07932v2 Sports video analysis is crucial for athletic analytics and broadcasting enhancement. Dense sports video reasoning, however, demands a fine-grained understanding of numerous small-scale, highly interactive, and visually homogeneous entities (e.g., players sharing identical uniforms, the ball) across long temporal contexts. Current Large Multimodal Models (LMMs) inherently struggle with such dense visual complexities. Due to the lack of fine-grained visual details, these models often over-rely on textual priors to guess answers, especially when distinguishing visually similar actions and players. To address this, we propose SportsGrounder, a framework that leverages an open-vocabulary visual expert to aid interleaved grounding specifically for dense sports video reasoning. To achieve precise spatial localization, we extract domain-guided object proposals and introduce an Interleaved Grounding Fusion (IGF) mechanism. The IGF frame-by-frame integrates explicit bounding box coordinates and implicit visual semantics with global grid features. This design preserves strict temporal alignment and prevents sequence length explosion. Furthermore, we design an Action-Aware Supervision (AAS) module that directly regularizes the model's hidden states, forcing the network to learn accurate motion representations rather than relying on language bias. Optimized with Mixed Preference Optimization (MPO) to better distinguish deceptive distractors, our extensive experiments on newly curated dense sports VQA datasets (derived from SoccerNet and FineSports) demonstrate that SportsGrounder significantly improves fine-grained reasoning and achieves state-of-the-art accuracy. --- ### REIN: Bridging the Gap between Reasoning and Reliability via Reflection and Abstention Alignment Authors: Zhengze Huang, Luyang Yu, Di Hong, Xinzhe Huang, Wanyu Lin, Zhixuan Chu, Zhan Qin, Tianhang Zheng Published: 2026-08-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.07931v1 Large reasoning models (LRMs) are prone to hallucination, which undermines their reliability and poses challenges for safe deployment. Hallucinations in LRMs arise from two distinct failure sources: reasoning hallucination, where flawed inference steps propagate to an incorrect conclusion, and knowledge hallucination, where the model lacks the requisite factual knowledge to answer the query. To address reasoning hallucination, we propose REIN, an alignment framework that trains LRMs to produce a structured reasoning sequence, $\texttt{<think>} $$\rightarrow$ $\texttt{<reflection>} $$\rightarrow$ $\texttt{<answer>}$, enabling explicit self-reflection before committing to a final answer. To address knowledge hallucination, REIN introduces a reward mechanism that encourages explicit abstention (e.g., "I don't know") when none of the sampled reasoning chains yields a correct answer, allowing the model to refrain from unsupported predictions. Extensive evaluations on mathematical and commonsense reasoning benchmarks show that REIN consistently improves selective accuracy, reduces incorrect-but-self-endorsed responses, and maintains high coverage compared with competitive baselines. Notably, REIN achieves these gains within a single forward pass, without requiring process supervision, inference-time controllers, external search, or multi-round critiques. Experiments on multiple backbones show that REIN reduces the hallucination proxy by $58\sim72\%$ relative to the base models while maintaining $86\sim91\%$ average coverage, and improves selective accuracy on attempted questions by $6.6\sim14.2\%$. --- ### ZhuLong: Execution-Grounded LLM Agent for EDA Scripting with Offline API Self-Exploration Authors: Yang Liu, Shiwei Hou, Xiyuan Chen, Yu Wang, Sen Yuan, Qirui Gan, Shao You, Feifan Chen, Wencheng Li, Shuyang Hu, Yongzhou Liu, Emma Xia et al. Published: 2026-08-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.07925v2 EDA scripting with tool-specific, often undocumented APIs remains a long-tail bottleneck that existing LLMs fail to address. This paper presents ZhuLong, an execution-grounded LLM coding agent for PyAether and SKILL that combines API retrieval, documentation inspection, and sandbox execution via unified MCP tools, augmented by an offline API self-exploration mechanism that infers undocumented API behaviors through counterfactual experimentation. We evaluate ZhuLong on EDA-Eval-PyAether, a benchmark of 158 real-world tasks with assertion-based execution, where the complete system achieves 78.5% Pass@1 in the commercial Empyrean Aether environment, substantially outperforming a pure LLM baseline (23.6%). Ablation studies identify sandbox execution as the dominant performance driver (41.2 pp drop when removed), with the self-exploration mechanism contributing an additional 3.2 pp accuracy gain and a 22.1% reduction in per-task tool calls. On 20 interactive tasks involving unsaved layouts and schematics, ZhuLong achieves 60.0% Pass@1 for PyAether and 50.0% for SKILL. --- ### Second Order Drifting Models Authors: Drake Brown, Yuhao Huang, Shih-Hsin Wang, Bao Wang Published: 2026-08-08 Categories: cs.LG, cs.AI, math.NA arXiv: https://arxiv.org/abs/2608.07924v1 Drifting models are a recent class of one-step generative models that evolve the model distribution during training using a predefined sample-based drift field. Although they avoid iterative inference, their kernel-based drift fields induce frequency-dependent training dynamics: In the linearized regime, each Fourier mode of the density residual decays at a rate determined by the kernel spectrum, leading to slow recovery of fine-scale structure. We propose Second-Order Drifting Models, which lift drifting dynamics into phase space by augmenting generated samples with artificial velocity variables. We show that the resulting density perturbations obey accelerated second-order dynamics in Fourier space, connecting drifting models to the celebrated Nesterov acceleration from optimization theory. This provides a principled mechanism for mitigating the spectral stiffness of first-order drifting while preserving one-step inference. We derive a practical semi-implicit training algorithm and evaluate it on synthetic distribution matching, sequential data generation, and robotic control. Across these settings, the second-order drifting model improves convergence behavior and achieves competitive or superior performance over first-order drifting baselines. --- ### SCoPE: Training-Free Audio-Visual Event Perception via Sparse Cross-Modal Prior Exchange Authors: Jaemo Jeong, Junho Yoon, Hyunju Kim, Dongman Lee Published: 2026-08-08 Categories: cs.CV, cs.MM, cs.SD arXiv: https://arxiv.org/abs/2608.07923v1 Audio-visual event perception (AVEP) determines which events occur in a video, when they occur, and whether they are audible, visible, or both. Training-free methods query new event vocabularies by matching frozen audio and visual features with text-encoded event names. However, related labels share evidence. An incorrect label can then score at least as high as a correct one. We call this a false co-activation (FCA). No scalar cutoff can reject the incorrect label while keeping every correct one. Class-specific thresholds may prevent that label from becoming a final prediction, but the FCA remains in the underlying score vector. We introduce SCoPE, a training-free framework in which all queried labels compete for shared evidence and each modality guides event selection in the other. We derive an exact condition for when this competition removes an FCA in a two-label fit. With identical frozen CLIP+CLAP backbones on LLP, SCoPE improves Type@seg by 7.45 points and Event@seg by 5.04 points compared with the reported AV$^2$A values. The same fixed configuration transfers unchanged to OV-AVEBench and VGGSound-AVEL100k. --- ### Memory--Batch Tradeoffs in Lipschitz Bandits Authors: Zicheng Lyu, Zengfeng Huang Published: 2026-08-08 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2608.07922v2 Lipschitz bandits admit near-optimal regret $\widetilde O_d(T^{(d+1)/(d+2)})$ with little memory under full adaptivity, or with few batches under unrestricted memory. We characterize the minimax expected pseudo-regret over $T$ rounds with $W$ bits of memory and at most $B$ batches in dimension $d$. For every memory budget $W$, we prove a lower bound of order $T^{\frac{d+2}{d+3}} (1+(B-1)W)^{-\frac{1}{d(d+3)}}$. When $W\gtrsim_d\log(eT)$, algorithms with fixed batch boundaries attain, up to logarithmic factors, the larger of this bound and the optimal $B$-batch regret with unrestricted memory. The analysis separates fine-scale comparisons from the regional information needed to allocate their samples at low regret. Attaining $\widetilde O_d(T^{(d+1)/(d+2)})$ regret requires both $B=Ω_d(\log\log T)$ and $(B-1)W=\widetildeΩ_d(T^{d/(d+2)})$. With one bit of memory, minimax regret is $\widetildeΘ_d(TB^{-1/(d+2)})$ under adaptive batch boundaries while $Θ_d(T)$ under fixed boundaries. --- ### Spectral Outliers Reveal Dominant Learned Structure in Transformer Attention Authors: Kasun Dewage, Marianna Pensky, Suranadi De Silva, T. H. Bandara Published: 2026-08-08 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2608.07921v1 We apply Marchenko-Pastur (MP) random matrix theory to pre-trained attention weights in order to separate each projection matrix into a random-like bulk and a set of spectral outliers. We validate this decomposition causally: zeroing the MP-identified outliers (signal) in Mistral-7B drives HellaSwag, MMLU, and PIQA close to random-chance performance, whereas zeroing a count-matched subset of bulk singular values causes smaller but non-negligible degradation. Across 11 pre-trained transformers we identify five recurring patterns: spectral outliers encode a dominant component of the learned structure; Q projections carry the most outliers; V projections under grouped-query attention lack a clean signal/noise separation; entry-level outliers form structured row-bands in Q and column-bands in O; and specific residual-stream dimensions persist as band outliers across layers in K and O. We close by outlining how these observations could inform parameter-efficient fine-tuning and structured pruning. --- ### Forged Peer Judgments Mislead Multimodal LLM Judge Panels: Source-Blind Anchoring and Panel-Consensus Verification Authors: Yang Shu Published: 2026-08-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.07920v1 Multimodal LLM judge panels can cross-reference peers, but a quoted peer judgment may itself be untrusted. We expose source-blind anchoring as a text-level attack surface in vision-language model (VLM) panels. Quoting independent visual judgments creates large anchoring gaps (19--26 percentage points) under both self and peer framing. A matched-content, label-only control changes the broken rate by only $-0.17$pp (95\% CI $(-0.68,0.35)$), showing that the self/peer label itself does not explain the effect. Under our tested construction, deliberately generated, concise wrong quotes overturn originally-correct verdicts 1.5--2.7$\times$ more often than naturally occurring wrong peer statements, with bootstrap 95\% CIs excluding parity across two datasets and seven VLM judges. Because the two statement populations differ in selection and form, this ratio measures differential damage under the tested attack rather than a provenance-only causal effect. We then introduce panel-consensus verification, which cross-checks a quote against independently collected blind votes. It blocks 84.9\% of fabricated attacks, cuts their net harm by 97.5\%, and preserves the positive but statistically inconclusive point estimate for genuine peer information under leave-one-out re-verification. These results identify a low-cost attack surface and a concrete defense for safer multimodal collaborative evaluation. --- ### TongGuOCR: A Layout-Aware and Token-Augmented OCR MLLM for Chinese Historical Documents Authors: Zhongheng Zhou, Yi Sun, Huiguo He, Yuyi Zhang, Peirong Zhang, Yulin Fang, Dezhi Peng, Minghui Liao, Lianwen Jin Published: 2026-08-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.07917v2 Chinese historical documents preserve valuable cultural heritage, but many collections remain accessible only as scanned page images, preventing full-text retrieval, collation, and computational analysis. Optical character recognition (OCR) can bridge this gap, but accurate transcription remains challenging because historical documents often contain complex layouts, rare characters, and nontrivial reading orders. We propose TongGuOCR, a layout-aware and token-augmented multimodal large language model (MLLM) for OCR of Chinese historical documents. First, a Layout-Aware Preprocessing module constructs and refines locally coherent recognition blocks to preserve local context while reducing interference across regions. Second, a Token-Augmented Recognition module augments the transcription target at two complementary levels: character-level vocabulary expansion gives each rare glyph a direct one-token representation and shortens its decoding path, while line-to-line transition modeling injects discrete spatial displacement tokens that guide the decoder along complex reading paths without requiring precise coordinates. Experiments on two Chinese historical document OCR benchmarks show that TongGuOCR outperforms representative traditional task-specific OCR models, general-purpose MLLMs, and OCR-oriented MLLMs. On the more challenging M5HisDoc benchmark, TongGuOCR achieves 93.76 AR and reduces NED from 10.43 to 6.15 and RO-ED from 7.53 to 3.49 relative to the best competing score for each metric. An online demo is available at https://jzzh2004.github.io/TongGuOCR. --- ### SegDem: Segmentation helps Demosaicing Authors: Ping Chen, Xiangming Wang, Yongyong Chen, Jiezhang Cao, Kai Zhang, Jingyong Su, Jie Liu, Haijin Zeng Published: 2026-08-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.07916v1 Image demosaicing reconstructs a full-color image from incomplete color measurements produced by a sensor covered with a color filter array (CFA). Most existing methods formulate demosaicing as pixel-level reconstruction and mainly rely on local textures, cross-channel correlations, and low-level image statistics. Our core insight is that reconstruction and visual understanding can be viewed as complementary views of shared scene structure: both are grounded in the same underlying physical world, and therefore the structural and physical information inferred from an image should remain consistent across the two tasks. We instantiate this idea with instance segmentation and propose \emph{SegDem}, a cross-task decoder representation transfer framework for demosaicing. SegDem first learns region- and boundary-aware representations through instance-aware structural pretraining and then transfers the decoder to RAW-conditioned reconstruction. Segmentation- and demosaicing-conditioned features are further anchored to a shared frozen DINOv2 representation space to preserve structural organization across tasks. We instantiate SegDem with convolutional, Transformer-based, and state-space backbones for unified Single- and Quad-Bayer demosaicing. Extensive experiments on synthetic, external, and challenging datasets demonstrate consistent improvements across different architectures and CFA layouts. --- ### SPECTRA: Pushing the KV Cache Beyond the 2-Bit Cliff via Spectral Transform Coding Authors: Jiamu Zhang, Liang Wu, Kelly Wan, Hanjie Chen, Liangjie Hong Published: 2026-08-08 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.07915v1 Large language models (LLMs) increasingly read long inputs in the agentic era, from whole documents and codebases to conversations across many turns. Their inference memory is then dominated by the key-value (KV) cache, the stored attention keys and values of every token the model has read and generated. Because the cache grows with context length and is re-read in full at every generated token, a longer context means more GPU memory. To reduce this cost, most existing methods compress the KV cache by lowering every stored value to the same low precision, a technique known as quantization. They can push this to nearly two bits per value, but rarely further, because quality drops sharply at this 2-bit cliff: four levels are too few for the cache's outlier-heavy values, where a few large entries consume the levels and collapse the rest into noise. A natural remedy is to spend more bits on the channels (feature dimensions) that matter and fewer on the rest, but the raw cache offers no handle: its channels are strongly correlated, so none stands out as more important. Our analysis shows that this handle appears once the cache is rotated into a coordinate system computed from its own statistics, removing these correlations. There, a small fraction of channels carries almost all the information, and spending the budget on those few is far more accurate than spreading it evenly. Guided by this analysis, we develop SPECTRA, a training-free, drop-in codec that re-encodes the cache into this coordinate system and concentrates the bit budget on the channels that carry the signal. On Llama-3.1-8B and Qwen2.5-7B over long-context benchmarks, SPECTRA is near-lossless at 4x compression, competitive at 8x where uniform quantization has collapsed, and reaches up to 12x, pushing usable compression past the 2-bit cliff so the same GPU holds longer contexts and larger batches. --- ### When Is Benchmark Contamination Detectable? Information Limits and Power-Calibrated Audits Authors: Ibne Farabi Shihab, Sanjeda Akter, Anuj Sharma Published: 2026-08-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.07914v1 Behavioral contamination detectors can return "no evidence" either because a benchmark is clean or because the audit has little power. We formalize this distinction for a benchmark in which an unknown fraction alpha of items was seen during training. With matched clean and seen controls, the behavioral channel is the sparse mixture Q_alpha = (1 - alpha) P_0 + alpha P_1, and an exact second-moment argument shows that detectability is governed by alpha * rho * sqrt(m), where rho^2 = chi^2(P_1 || P_0) measures behavioral separability. Any scalar detector reduces to its efficacy, ef = |E_1 f - E_0 f| / sqrt(Var_0(f)) <= rho, which can be estimated from controls before the audit is run. A separate sample-split certificate lower-bounds alpha distribution-free, without requiring an orientation assumption. Our empirical finding is two-sided. Frozen calibration efficacy predicts held-out power curves, with R^2 = 0.83-0.98 across six exact-permutation channels, but the efficacy-only Gaussian budget is miscalibrated at the small sample sizes it prescribes, failing in 9/9 gate-passing channels even though efficacy itself transports. The failure is in the inversion, not the calibration. A predeclared two-stage planner that simulates the deployed test repairs the budgets, is uniformly conservative, and abstains when its probe does not transport. The certificate is valid but vacuous at audit scale, and a five-seed paired injection study recovers the mechanism ordering verbatim > paraphrase > surface, in which the apparent answer-only signal is explained by baseline drift. We report the audit contract and its failures together: a non-rejection is interpretable only alongside the efficacy, budget, and validity gates that produced it. --- ### Private Anytime Selective-Risk Certification for Federated Retrieval-Augmented Generation: Guarantees and Empirical Limits Authors: Sanjeda Akter, Ibne Farabi Shihab, Anuj Sharma Published: 2026-08-08 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2608.07913v1 Selective-risk certificates promise that accepted outputs meet a declared error target. We develop Fed-SRC, a score-agnostic certificate for federated, differentially private, adaptively monitored retrieval-augmented generation. Clients release only Gaussian-perturbed score and loss histograms. Record-indexed and noise-variance-indexed martingales jointly bound target-risk contrast and accepted mass over all registered thresholds and rounds, permitting predictable recruitment, dropout, threshold selection, and optional stopping. A range-one total-variation term transfers the calibration mixture to a declared deployment mixture. The contribution is this private, federated, anytime combination, rather than the contrast statistic or acceptance floor individually. Empirically, no simultaneous-bound violation occurs in any evaluated cell, privacy level, or policy. Operational power depends on the score and population: the primary target r*=0.10 never certifies, and on RAGTruth the secondary target r*=0.20 never certifies either, whereas on HaluEval question answering it certifies in all 200 non-private trials, with held-out risk below the target. Naively privatized non-private certificates violate their bounds in 146 to 198 of 200 trials. As an exploratory comparison, we also evaluate a private betting-capital heuristic for which we do not establish e-process validity. This heuristic stops certifying at epsilon <= 4, where Fed-SRC still certifies. Certification nevertheless consumes roughly 30 times more stream events than unique calibration items. --- ### Reproducible Evaluation of MoE Expert Caching: Replay Semantics, Workload Contamination, and Operating Regimes Authors: Yu Zhang Published: 2026-08-08 Categories: cs.LG, cs.PF arXiv: https://arxiv.org/abs/2608.07911v4 Mixture-of-Experts (MoE) models have outgrown accelerator memory, and offloading expert weights to host memory is now standard. This makes expert cache management an attractive lever: a policy that raised the hit rate would cut expert traffic per token. Evaluating that is a measurement problem, and we find the measurement fragile. With a trace-driven, event-atomic simulator over three MoE models (40, 64, 128 experts), we isolate three evaluation axes that change conclusions, not just numbers. Replay semantics: under a fused-event traffic contract, an inconsistent per-access replay inflates recency-based policies by 27-29% while leaving frequency-based and static ones within 4%, inverting the policy ranking. Workload contamination: probe sets using one instruction template per category produce verbatim-identical generation prefixes; a matched-pair rendering intervention moves the measured early-window effect by 19.4-31.9 points and reverses which workloads look most cache-friendly. Operating regimes: normalized miss fractions do not transfer across models, so the per-step expert union relative to per-layer capacity must be reported -- yet permuting only the temporal order of an identical event stream moves the offline-optimal gap from 44.9% to 30.8%, so it is not sufficient. Corrected, a stable gap to the offline optimum remains (44.2-45.9% over 13 frozen workload compositions). A forced-admission oracle attributes 84.3-96.6% of it to knowing which resident expert is used furthest in the future. A causal next-use predictor, used as an eviction rule, recovers -11.4% of the gap; it picks an optimal victim 3.4% of the time, against 2.4% for a random resident block and 20.6-22.1% for LRU and LFRU. Our position is narrow: in our evaluated settings a large offline-optimal gap substantially overstates the gains recovered by representative lightweight causal mechanisms. --- ### GraphThink: Graph-Enhanced LLM Thinking for Long-Horizon Embodied Task Planning Authors: Chen Li, Sijie Cheng, Yuelin Zhang, Junxi Li, Maozhi Huang, Yang Liu, Wenbing Huang Published: 2026-08-08 Categories: cs.AI, cs.RO arXiv: https://arxiv.org/abs/2608.07905v1 Embodied agents using LLM-based planners often struggle with physical hallucinations, poor generalization to long-horizon tasks, and lack of environmental awareness. We propose GraphThink, a novel framework that integrates a task graph to provide structured knowledge for robust planning and a scene graph to maintain environmental memory for event-driven replanning. Specifically, the task graph guides LLM thinking through contextual prompting and iterative refinement, effectively mitigating planning hallucinations. Furthermore, within the GRPO framework, the task graph offers delicate reward design to train the LLM planner, enhancing long-horizon planning capabilities and improving generalization. Finally, an event-driven replanning module, powered by the scene graph, enables closed-loop environment awareness and error correction. GraphThink achieves state-of-the-art performance on the ALFRED benchmark. In particular, our high-level planner surpasses leading API-based LLMs on both the validation set and held-out long-horizon tasks, underscoring its robust zero-shot and few-shot capabilities. Additional evaluations further demonstrate strong out-of-distribution generalization to novel tasks and environments. --- ### DeCo: Zero-Shot Industrial Anomaly Generation through Decoupling and Recoupling Authors: Shilei Zeng, Xurui Li, Yaohan Tang, Yu Zhou Published: 2026-08-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.07904v2 Industrial anomaly inspection is severely hindered by the scarcity of real anomalous data. Zero-shot industrial anomaly generation addresses this by generating anomalies on specific products without requiring any of their real anomalous images. However, existing methods suffer from two critical limitations, i.e., inaccurate anomaly information acquisition and uncontrolled anomaly-product fusion. To overcome these challenges, we propose DeCo, which decouples the anomaly structure from its source product, and explicitly recouples it with the normal textures of the target product. During anomaly information acquisition, Dual-Routing Flow (DR-Flow) binds the texture-invariant anomaly structure to an abnormal token, while a parallel constraint, Product-Invariant Flow (PI-Flow), prevents the abnormal token from binding the source product. During anomaly-product fusion, we propose a hybrid injection to recouple the acquired anomaly structure with the target product, and Product Compatibility Correction (PCC) to compensate for the incompatibility between the acquired anomaly structure and the product. Extensive experiments demonstrate that DeCo establishes a new state-of-the-art. Training downstream detection models on our generated data yields massive pixel AP improvements of 5.1% on MVTec AD and 8.2% on VisA. Code is available at https://github.com/HUST-SLOW/DeCo. --- ### Beyond "I Can't Help With That": How Child Safety Experts Evaluate AI Chatbot Safety Authors: Hannah Cha, Neha Shukla, Solon Barocas, Alexandra Chouldechova, Eugenia Kim, Jennifer Wortman Vaughan Published: 2026-08-08 Categories: cs.CY, cs.AI, cs.HC arXiv: https://arxiv.org/abs/2608.07902v1 Youth increasingly turn to AI chatbots for social and emotional support, raising concerns about how these systems respond, especially in high-stakes situations. However, existing child safety evaluations of AI lack grounding in real-world harms that youth experience, rely on unvalidated assumptions about what counts as an appropriate output (e.g., refusal), and typically focus on detecting adversarial prompts or surface-level harms in outputs only. Thus, these evaluations can fail to detect responses that pose harm to youth in practice. To better understand the limitations of current evaluation practices, we conducted interviews with 19 practitioners working directly with youth in vulnerable situations, including social workers, therapists, and psychologists, asking them to reflect on chatbots' responses to risky situations commonly faced by youth, as established in prior empirical work. Practitioners identified chatbot behaviors likely to cause harm as well as those that could meaningfully support youth in difficult moments, discussed the role that chatbots should (and should not) play in these interactions, and offered concrete recommendations for improving chatbot responses. Based on these findings, we provide recommendations for AI child safety evaluation and infrastructure, and highlight the need for incorporating practitioners' perspectives into safety work. --- ### TelemetrySuffBench: Is Agent Telemetry Sufficient for Failure-Origin Diagnosis? Authors: Yuxuan Zhu, Peng Pu Published: 2026-08-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.07899v1 Agent systems increasingly expose execution traces, yet telemetry that reveals a failure may still be inadequate for identifying where that failure originated. We introduce TelemetrySuffBench, a controlled benchmark that separates failure detection, fault-origin localization, and safe abstention under insufficient evidence. The benchmark constructs canonical multi-component traces with delayed-binding faults and renders them as paired coarse views, seven-factor telemetry masks, and exact-equal ambiguous origin pairs. We evaluate five frontier language models using unified protocols, explicit candidate sets, invalid-output accounting, subgroup analyses, and a frozen blind holdout. With full telemetry, origin-step Top-1 accuracy ranges from 33.8% to 97.2% across models. Metadata, OpenTelemetry-compatible, and OpenInference-compatible views retain 99.5% to 100% detection F1 while limiting origin-step accuracy to at most 0.5%, exposing a robust detection-localization gap. Factor ablations further show that removing decision content reduces origin-step accuracy to zero for every model, while provenance removal also causes large model-dependent losses. On rich ambiguous inputs that require abstention, evidence gating reduces unsupported unique-origin answers by 12.5 to 48.6 percentage points for three models, whereas two models still answer every case, revealing strong model dependence in safe abstention. Results on the frozen holdout reproduce the central pattern within the same generator family. These findings show that terminal status can support detection, whereas reliable causal attribution requires explicit decision-to-provenance links and abstention safeguards that remain effective across models. The dataset and benchmark implementation are available at https://anonymous.4open.science/r/TelemetrySuffBench-E635/README.md. --- ### Auditing Instruction-Trajectory Mismatches in Multimodal Robot Demonstrations Authors: Simon Holk, Ryosuke Takanami, Tatsuya Matsushima, Yusuke Iwasawa, Yutaka Matsuo, Yueh-Hua Wu, Kei Ota Published: 2026-08-08 Categories: cs.RO, cs.LG arXiv: https://arxiv.org/abs/2608.07895v1 Robot demonstration datasets used to train vision-language-action policies can contain a subtle but harmful failure mode: trajectories that are behaviorally correct but paired with the wrong language instruction. We study post-hoc auditing of these Instruction-Trajectory Mismatches (ITMs). Unlike failed rollouts, ITMs often look plausible, and can corrupt the language-behavior mapping learned by the policy. We propose Multimodal Probabilistic Fusion (MMPF), a training-free auditing framework that treats each modality as an expert, estimates a task-label distribution from local neighborhood agreement and global prototype similarity, and then fuses modalities with predictive-entropy weighting in a product of experts. Across LIBERO benchmarks with injected instruction mismatches and noisy real-robot data, MMPF achieves the strongest overall ITM detection and label correction accuracy. We also show that auditing improves most downstream policy learning in settings where language is needed to disambiguate the task. We demonstrate in real robot experiments that our method can achieve improved policy performance and show the trade-off of filtering demonstrations compared to relabeling. --- ### LLM-Based Embeddings for Program Analysis and Optimization Authors: Calvin Higgins, Marco Alvarez Published: 2026-08-08 Categories: cs.LG, cs.PL arXiv: https://arxiv.org/abs/2608.07894v1 Recent advances have highlighted the potential of machine learning, particularly Large Language Models (LLMs), for analyzing and optimizing programs. We present the first application of program embeddings from LLMCompiler---an LLM massively pretrained on intermediate representation (IR) code---to representative program analysis and optimization tasks. We generate program embeddings directly from source and IR code using a simple approach: split programs into chunks, independently embed each chunk with pretrained LLMs, and then aggregate the chunk embeddings into a single program embedding. Our experiments show that combining source and IR code embeddings achieves an error rate of 1.54\% in algorithm classification, a 12\% improvement over the current state-of-the-art, and a competitive accuracy on heterogeneous device mapping. These findings suggest that training a performance-aware LLM for embedding IR code might yield state-of-the-art results in code optimization tasks. --- ### Detection of Self-Introductions in Legislative Testimony Authors: Sofija Dimitrijevic, Pallavi Das, Kasey Liu, Foaad Khosmood Published: 2026-08-08 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.07891v1 Self-introductions are common in legislative committee testimonies. Successfully detecting them and extracting the speaker's name is enormously helpful in the task of speaker identification in the context of government meetings. In this paper, we present a pipeline for detection of self-introductions in legislative committee testimony using machine learning. We construct a training dataset from 1.54 million utterances spanning five state legislative sessions, apply a name-matching heuristic to generate automatic labels, and train three classifiers: a decision tree, random forest, and XGBoost to find self-introductions and extract the speaker's name. We construct a feature set combining bag-of-words, positional context, structural signals, introductory phrase indicators, and discourse context features. Among the three classifiers, XGBoost achieves the best performance with an F1 score of 0.9747 and the fewest total errors; adding fine-tuned BERT probability features improves this further. As an extension, we score the full candidate dataset with a fine-tuned BERT classifier and add BERT probability outputs as features. This BERT-augmented XGBoost model improves F1 from 0.9747 to 0.9782 and reduces total test errors from 241 to 207. The primary gain over the decision tree baseline (F1 0.9323) is driven by discourse context features and the boosting ensemble strategy; BERT provides a modest complementary signal. Analysis of false positives reveals that a minority are genuine self-introductions mislabeled due to name inconsistencies in the source data, indicating that measured metrics modestly understate true performance. --- ### Router Sensitivity Under Lightweight Fine-Tuning Identifies Prunable Experts in Mixture-of-Experts Models Authors: Ali Janati, Kaoutar El Maghraoui, Xinyi Luo, Wenyuan Shen, Owen Zou, Yankai Mao Published: 2026-08-08 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2608.07890v1 Mixture-of-Experts (MoE) models decouple total parameters from per-token compute, but deployment still requires storing every expert. Recent theory shows that pruning experts with the smallest router-norm changes during fine-tuning can preserve accuracy, but assumes full fine-tuning. We test whether lightweight adaptation can recover this signal. We briefly fine-tune with a parameter-efficient adapter, rank experts by the induced $\ell_2$ router change, and prune the least-changed experts in one shot. On Mixtral-8$\times$7B-Instruct (44.83% MMLU-Pro), router-only LoRA trains 0.002% of parameters and outperforms all-module LoRA at matched rank with half the experts removed (27.54% vs. 24.42%); signal quality declines as adaptation spreads to attention and expert weights. Accuracy improves monotonically with LoRA rank, reaching 28.76%. IA3, which leaves router weights frozen, matches direct router adaptation, whereas unconstrained additive adapters degrade the signal. Router-guided MMLU-Pro accuracy decays quasi-linearly rather than collapsing, remains nearly 1.8 times that of magnitude-based or random pruning at maximal compression, and reduces memory by 49% and per-token latency by 37%. At 25% compression, retention is competitive with methods using full activation statistics. The criterion also transfers to Qwen1.5-MoE fine-tuned for mathematics, retaining 49.7% mean accuracy over eleven benchmarks with half the experts removed while random pruning falls to single digits. Router sensitivity under lightweight fine-tuning therefore makes provably motivated expert pruning practical at scale. --- ### Vision-Language Grounding as Bidirectional Concept Correspondence Authors: Jieyu Zhang, Ziqi Gao, Luke Zettlemoyer, Ranjay Krishna Published: 2026-08-08 Categories: cs.CV, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2608.07886v1 Vision-language grounding connects language to visual content, yet most existing formulations reduce grounding to a unidirectional localization problem: given a prespecified text phrase or category name, identify the corresponding image region. This setup assumes that the relevant linguistic unit is already known, overlooking a more basic challenge in grounded communication: determining which parts of the text are visually referential and how they correspond to entities in the image. We formulate grounding as $\textit{bidirectional concept correspondence}$ over an image-text pair. Given an image and its paired text, the goal is to recover all correspondences between visually referential text spans and instance-level image segments, without assuming that the relevant text spans are provided. This formulation unifies common grounding tasks, including phrase grounding, referring expression grounding, and open-vocabulary detection, by treating text segmentation, image segmentation, and cross-modal alignment as a single correspondence prediction problem. To address this task, we introduce $\textbf{ConCor-1}$, a grounding model built on top of a pretrained vision-language model. It uses learnable $\textit{bridge tokens}$ to represent candidate image-text correspondences and predicts, for each token, a text mask, an image mask, and a correspondence presence score. To train and evaluate this task, we convert diverse grounding and segmentation datasets into a unified correspondence format. Experiments show that $\textbf{ConCor-1}$ consistently outperforms baselines, improving correspondence F1 by 48% on the long-caption dataset and by 29% on zero-shot LVIS, where the large category list serves as the text input. --- ### Reason Wide, Not Deep: Amortizing the Reasoning Premium into Distilled Skills Authors: Agamdeep Singh, Srishti Gautam, Priyanshu Gupta, Nikita Mehrotra, Tanmay Bakshi, Sumit Gulwani Published: 2026-08-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.07885v1 Reasoning modes of language models outperform their non-reasoning counterparts on multi-step agentic tasks, but pay a 3-6x premium in output tokens on every episode -- much of it spent re-deriving procedures that are shared across episodes of the same domain. We show this recurring cost can be amortized: a coding agent analyses a small corpus of existing trajectories from a training split and compiles a compact natural-language skill that is injected into the non-reasoning model's system prompt. Across four agentic benchmarks (ALFWorld, tau$^2$-bench telecom and retail, and SpreadsheetBench-Verified), skills recover 55%-100%+ of the reasoning gap for GPT-5.4-mini on held-out tasks -- exceeding the reasoning mode outright on two of four -- while emitting 2.7-6x fewer output tokens and zero reasoning tokens. Notably, reasoning traces are not a prerequisite: skills distilled from non-reasoning trajectories alone remain competitive with skills distilled from paired reasoning/non-reasoning corpora, with domain-dependent differences between the two sources. We interpret these results through a search lens: test-time reasoning is deep search inside a single episode, re-paid at every deployment, while corpus distillation is wide search across episodes, paid once. The two recover overlapping procedural knowledge, and width over cheap trajectories is often the better buy -- with the residual gap on some domains (telecom, SpreadsheetBench) delineating where genuinely per-instance deep search remains necessary. --- ### GRACE: LLM-Grounded Semantic Metric Spaces for Scalable Mixed-Data Clustering Authors: Zihua Yang, Zhencheng Xie, Junyang Chen, Liang Xie, Yiqun Zhang, Mengke Li, Yang Lu Published: 2026-08-08 Categories: cs.AI, cs.CL, cs.IT, cs.LG arXiv: https://arxiv.org/abs/2608.07881v1 Clustering mixed tabular data requires a unified metric space to bridge the inherent heterogeneity between continuous numerical measurements and discrete categorical symbols. Traditionally, algorithms rely entirely on dataset-internal statistics to estimate categorical relationships, which confines the learned metric to empirical co-occurrences and ignores conceptually obvious yet statistically unobserved affinities. Although LLMs offer external world knowledge, applying their text-centric reasoning to highly abstract tabular concepts presents significant challenges. Bridging this modality gap to construct a semantically complete metric typically requires embedding LLMs into iterative metric learning loops to dynamically optimize cross-modality representations. This incurs intractable computational overhead, forcing a compromise between semantic enrichment and scalability. Therefore, we propose GRACE, an LLM-grounded framework for scalable mixed-data clustering. GRACE shifts semantic acquisition to the attribute-value level via a multi-perspective LLM querying strategy, mapping heterogeneous values into knowledge-informed descriptions. Crucially, this one-shot grounding extracts general-purpose semantic representations that embed heterogeneous attributes into a unified space, decoupling expensive LLM invocation from iterative optimization. Furthermore, GRACE cross-validates these external semantics against dataset-internal statistical evidence to ensure alignment with the dataset-specific cluster structure. Ultimately, GRACE matches the scalability of conventional statistics-driven baselines while achieving superior clustering accuracy and conceptual interpretability over 11 competing methods. The source code is available at https://github.com/develop-yang/GRACE-GRACE-A --- ### SurgLAT: Surgical Latent Attention Tracking for Depth-Aware Robotic Laparoscope Control Authors: Rulin Zhou, Qiujie Song, Yujie Ma, An Wang, Wanhao Liu, Guoheng Ma, Yidu Wang, Guankun Wang, Xingrong Diao, Jiankun Wang, Chaowei Zhu, Xianming Liu et al. Published: 2026-08-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.07876v1 Autonomous laparoscopic camera control requires continuous understanding of the surgeon's operative intent in dynamic surgical scenes, where the target operative region is not a stable physical object but a latent and temporally evolving attention state. In this work, we present Surgical Latent Attention Tracking (SurgLAT), a causal online framework for latent surgical attention modeling and autonomous laparoscopic view control. SurgLAT uses a frozen DINOv3 encoder and a state-conditioned spatial token mixer to extract operative evidence under a memory-guided spatial prior, while a selective causal latent memory module jointly models short-term motion continuity and long-horizon surgical intent evolution through dynamic retrieval of current, recent, and historical latent states. The learned latent surgical attention state is decoded into a probabilistic attention heatmap and operative region for downstream endoscope guidance. Beyond perception, we further introduce a robotic deployment framework with explicit laparoscopic Remote Center of Motion (RCM) constrained control based on virtual-axis formulation, together with redundancy-aware null-space initialization for stable and smooth manipulator motion. We validate the full system on real laparoscopic surgical videos and a physical robotic laparoscope platform. Experimental results demonstrate robust online operative-region tracking and stable autonomous endoscopy adjustment under occlusion, rapid motion, and target transitions, highlighting the effectiveness of latent surgical intent modeling for surgical autonomy. --- ### Back to the Future: A workbook time machine for spread sheet creation benchmarks Authors: Mansi Uniyal, Agamdeep Singh, Ananya Singha, Priyanshu Gupta, Mukul Singh, Gust Verbruggen, Vu Le, Sumit Gulwani Published: 2026-08-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.07873v1 We introduce the workbook time machine, a pipeline that automatically creates benchmarks evaluating the ability of language models to create derived objects in spreadsheets (formulas, charts, pivot tables, and conditional formatting). Applied to public workbook corpora, it produces wtmcorpus--a collection of (input workbook, output workbook, query) triples spanning four artifact types and varying complexity. From this corpus we curate wtmbench, a 150-task evaluation benchmark with queries at three levels of specificity. We evaluate existing spreadsheet manipulation agents and baselines on wtmbench across artifact types, step complexity, and instruction granularity. Our evaluations show that query specificity, agent orchestration, and interface API used to control spreadsheets play a big role in LLM performance on Excel tasks. --- ### Crowd-Sourced Geographies of Income: Using Google Maps Points of Interest as High-Frequency Proxies for Sub-Municipal Income Estimation in Sao Paulo, Brazil Authors: Adrienne C. Kinney, Anya Workman, Ademar Takeo Akabane, Jenna Barac, Paulo Fernando Braga Carvalho, Jeova Farias, Fernando Nascimento, Paulo Ricardo da Silva Oliveira Published: 2026-08-08 Categories: stat.AP, cs.CY, cs.LG arXiv: https://arxiv.org/abs/2608.07871v1 Accurate, up-to-date income data at the sub-municipal scale is essential for social policy in middle-income countries, yet in Brazil it depends on a costly decennial census whose intercensal gap recently exceeded a decade. We test whether the composition of crowd-sourced Google Maps Points of Interest (POIs) can serve as a high-frequency, low-cost proxy for household income across the 26,625 census sectors of the municipality of Sao Paulo. Using a theoretically motivated set of POI categories retrieved from Google Places, we represent each sector by its POI counts, decompose these high-dimensional, sparse features with principal component analysis (PCA) and non-negative matrix factorization (NMF), and train a sweep of regression models to predict census-derived income. Under a data leakage-aware spatial validation design the best model (NMF with gradient boosting) attains a held-out R^2 of 0.65, with performance stable across feature-extraction methods. Interpretable decompositions reveal which POI types carry the income signal. These results suggest that commercial, crowd-sourced geospatial data can complement conventional income statistics during intercensal periods, and we discuss extensions toward multidimensional poverty and the capabilities framework. --- ### V-Simba: Unleashing the Architectural Potential of RL in Visual Continuous Control Authors: Donghu Kim, Youngdo Lee, Hojoon Lee, Johan Obando-Ceron, Byungkun Lee, Aaron Courville, Pablo Samuel Castro, Jaegul Choo, Clare Lyle Published: 2026-08-08 Categories: cs.LG, cs.RO arXiv: https://arxiv.org/abs/2608.07870v1 Improving sample efficiency remains a core challenge in reinforcement learning (RL), especially in real-world settings like robotics, where data collection is costly. This challenge is pronounced in visual RL, where high-dimensional inputs often obscure learning signals. While prior work in visual RL has focused on algorithmic solutions, such as better dynamics models or exploration strategies, recent advances in state-based RL show that architectural design alone can lead to significant gains in sample efficiency. This raises an important question: Can these architectural principles transfer to visual RL? In response, we introduce V-Simba, a simple yet effective visual RL architecture inspired by the Simba architecture from state-based RL. Built on top of Soft Actor-Critic (SAC) with data augmentation, V-Simba modifies the architecture by adding normalization layers to stabilize training and using pointwise convolutions to reduce computation. Despite its simplicity, V-Simba matches or outperforms the state-of-the-art methods across the DMC, Adroit, and Meta-World benchmarks, while being more computationally efficient than DrQ-v2. We make our code publicly available at https://github.com/DAVIAN-Robotics/V-Simba. --- ### CONFER: Conflict-Aware Evidence Negotiation for Regime-Calibrated Weak Supervision in Multimodal Emotion Recognition Authors: Bojing Hou, Ruohao Li, Yitong Zhu, Luwen Yu, Yuyang Wang Published: 2026-08-08 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.07867v1 Multimodal emotion recognition often treats self-reported labels as reliable supervision while overlooking self-report unreliability and cross-modal conflict. We propose \textbf{CONFER}, a graph-based conflict-aware evidence negotiation framework for weakly supervised multimodal emotion recognition. CONFER represents each modality expert as a node with a predictive belief, boundary-based uncertainty, and runtime reliability estimated from historical out-of-fold performance and current-sample uncertainty. Uncertainty-aware compatibility and reliability-directed asymmetric edge weights govern iterative message-passing negotiation, followed by peer-supported prediction readout. Conflict reduction, residual disagreement, and mean modality uncertainty further characterize three regimes---Consensus, Dissent, and Ambiguity---for sample-specific weak-label calibration. We evaluate CONFER on AMIGOS, MAHNOB-HCI, and DEAP under subject-dependent 10-fold and strict leave-one-subject-out (LOSO) protocols. CONFER achieves competitive performance, reaching \textbf{0.873} accuracy on AMIGOS-V and \textbf{0.854} accuracy on MAHNOB-V under strict LOSO evaluation. Further analyses show larger negotiation gains on high-conflict samples and improved robustness to weak-label corruption, indicating that cross-modal conflict provides useful information for both directional modality coordination and supervision-reliability estimation. --- ### UniScale: Arbitrary-Scale Industrial Anomaly Generation Authors: Shilei Zeng, Linxin Guan, Xurui Li, Yaohan Tang, Yu Zhou Published: 2026-08-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.07864v2 Industrial anomaly inspection faces a major challenge due to the lack of real-world anomaly samples. While generative models are used to create anomaly data, existing methods still struggle when handling small-scale anomalies. This failure occurs because extreme downsampling in diffusion models causes the information of small anomalies to be lost in the latent space. To address this, we introduce UniScale, a unified training and inference framework for high-fidelity industrial anomaly generation across arbitrary scales. During training, we introduce an Error-Suppressed Multi-Scale Training (EMT) strategy, which enables the model to learn the rich location-aware textures of anomalies, while suppressing upsampling-induced interpolation errors in texture acquisition, ensuring the model is capable of learning small-scale anomalies, while remaining effective for regular scale anomalies. For inference, we propose Generation-then-Fusion Denoising. It decouples anomaly generation from background integration, preventing small anomalies from being overwhelmed. Extensive experiments demonstrate that our method outperforms state-of-the-art competitors in both anomaly generation quality and downstream detection performance. It achieves a relative IS(a) improvement of 45.86% (from 1.81 to 2.64) on VisA and 37.70% (from 1.22 to 1.68) on MVTec AD 2, while also improving the downstream pixel-level IoU by 4.22% on VisA and AUROC by 6.55% on MVTec AD 2. Code is available at https://github.com/HUST-SLOW/UniScale. --- ### LHSDet: High-Resolution AI-Generated Image Detection via Visual Question Answering Authors: Qian Yao, Jun-Jie Huang, Yongjun Wang, Luming Yang Published: 2026-08-08 Categories: cs.CV, eess.IV arXiv: https://arxiv.org/abs/2608.07863v1 Driven by advances in diffusion models and autoregressive models, the fidelity and resolution of AI-generated images now rival those of real images. However, existing AI-generated image detection methods often downsample the images, inevitably overlooking critical low-level texture details in high-resolution AI-generated images, therefore limiting their detection performance. In addition, the ceaseless emergence of unknown generative models makes large-scale pre-training datasets inaccessible. To address these challenges, we propose a novel high-resolution AI-generated image detector, termed LHSDet. Specifically, we formulate the AI-generated image detection task as a Visual Question Answering problem, leveraging a fine-tuned vision-language framework to fully exploit the complementary information between visual and textual modalities. Recognizing that the default visual encoder of existing vision-language models is not tailored for AI-generated image detection, we redesign a visual encoder to better capture both the low-level and high-level artifacts inherent in AI-generated images. Furthermore, we incorporate a semantic-level textual branch to enable multi-modal feature fusion and detection. Consequently, LHSDet employs a triple-branch architecture to extract complementary multi-modal features: a low-level visual branch that aggregates non-overlapping patches for local texture cues, a high-level visual branch based on SigLIP2 for global perception feature extraction, and a semantic-level textual branch that generates captions using BLIP-2. Extensive experimental results demonstrate that LHSDet achieves high detection accuracy and robust performance across diverse generative models, including both diffusion and autoregressive models. --- ### SurakshaEval: An Indic Safety Benchmark for Multilingual LLMs Authors: Debopriyo Banerjee, Kapil Rajesh Kavitha, Angana Borah, Xudong Han, Yuxia Wang, Parameswari Krishnamurthy, Utkarsh Agarwal, Atharva Kulkarni, Swaran Lata, Ayush Munot, Dhruv Sahnan, Aaryamonvikram Singh et al. Published: 2026-08-08 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.07862v1 Existing safety evaluation datasets for large language models (LLMs) predominantly focus on English and Western contexts, often overlooking the linguistic diversity and culturally grounded safety risks present in other languages. To address this gap, we introduce SurakshaEval, a novel safety benchmark composed of human-written prompts spanning real-world scenarios, explicitly designed for ten major Indian languages - Assamese, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Punjabi, Tamil, and Telugu, along with English. SurakshaEval includes both generic prompts common across India and region- and language-specific prompts that capture localized sociocultural sensitivities. We benchmark a broad range of state-of-the-art LLMs on SurakshaEval, establish baseline safety performance, and identify recurring failure modes, including over-refusal, missed detection of implicit bias, and insufficient contextual awareness in regionally sensitive settings. Our results show that even strong multilingual LLMs struggle to reliably meet nuanced safety requirements when operating in Indic languages, particularly in native scripts. These findings highlight the urgent need for safety evaluation frameworks that incorporate region-specific data and structured assessment protocols, enabling the development and deployment of AI systems that operate securely, ethically, and in alignment with diverse societal values. Our code and data are available at https://github.com/debobanerjee/SurakshaEval. Warning: This paper contains text that may be offensive or unsafe. --- ### How Much Does It Cost to Answer My Question? Benchmarking Cloud VLM-based VQA Systems Authors: Henri Vanhuynegem, Weitao Xu, Yiran Shen, Guohao Lan Published: 2026-08-08 Categories: cs.CV, cs.HC, cs.IR, cs.MM arXiv: https://arxiv.org/abs/2608.07861v1 Vision-language models (VLMs) are becoming a practical backend for mobile visual question answering (VQA) systems, enabling smartphones and smart glasses to answer users' questions about the physical world. Since modern VLMs remain difficult to run on mobile and edge devices, VQA systems increasingly offload inference to cloud-based VLMs. This gives mobile devices access to stronger computation, but it also makes visual input preparation a key system variable: how the image is prepared before offloading affects not only answer quality but also payload size, token cost, and system latency. Proprietary APIs expose little control over model internals or serving behavior, leaving client-side preprocessing as the main practical optimization space for downstream developers. Many such techniques have been proposed for visual offloading, yet their cost-quality impact on commercial cloud VLMs has never been studied. To fill this gap, we present VQABench, the first systematic benchmark that treats client-side input preprocessing as a controlled variable for cloud-VLM-based VQA. We evaluate 12 preprocessing techniques across three VQA datasets and four commercial VLMs from three providers, totaling 95,168 API calls. Our results show that preprocessing is not universally beneficial: its effectiveness depends on the target model, API paradigm, provider token-accounting rule, and task formulation. A poorly selected preprocessing strategy can increase deployment cost or latency while degrading answer accuracy. Overall, our benchmark clarifies when preprocessing helps, when it fails, and why, providing insights to guide future research and real-world deployment of VQA systems. ---