# Research Papers — 2026-08 · bundle page 76 of 240 > Full-text markdown of papers #3751–3800 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/76 [← Previous bundle](https://aipod.dev/papers/archive/2026-08/page/75.md) [Next bundle →](https://aipod.dev/papers/archive/2026-08/page/77.md) ### Through the Schrödinger Bridge: Benchmarking Antemortem Image Restoration from Postmortem Autolysis to Enhance Forensic Diagnostics Authors: Shuang Hao, Jiacheng Yue, Yaxuan Zhao, Fan Wang, Jianhua Ma, Erwen Huang, Chunfeng Lian Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21813v1 Forensic histopathology, essential for determining cause of death and disease diagnosis, is severely impeded by postmortem autolysis, i.e., an irreversible, stochastic degradation process that distorts tissue morphology and introduces diagnostic subjectivity, thereby underscoring the value of restoring autolyzed images to a diagnostically plausible, pre-autolysis state for improving objectivity in forensic practice. This restoration task is fundamentally challenging due to the large, non-deterministic morphological changes caused by autolysis and the infeasibility of pixel-wise paired data, which invalidates assumptions underlying supervised and cycle/structure-consistent unpaired translation methods. To address this, we formalize forensic histopathology autolysis restoration as a new task: under unpaired supervision, transform postmortem images with severe autolysis into diagnostically meaningful ``antemortem'' representations. We contribute AutoPath, the first homologous yet unpaired dataset for this problem, constructed by splitting specimens into adjacent tissue blocks---one processed immediately, the other exposed to induce autolysis---yielding nearly ten thousand $10\times$ patches from 69 cases with varying liver conditions. We further frame the problem as a Schrödinger Bridge between the autolyzed and non-autolyzed distributions, offering a principled approach to modeling stochastic, severe morphological degradation. Critically, we demonstrate the misalignment of generic image-level generative metrics (e.g., FID) with diagnostic utility and propose a forensically grounded, slide-level diagnostic distribution consistency evaluation. Overall, this work establishes a reproducible benchmark (encompassing task definition, a real-world dataset, and an evaluation methodology) toward rigorous and practically meaningful progress in autolysis restoration for forensic pathology. --- ### Hints, Critics, and Teachers: Prior Injection for Sparse-Reward RL in Vision-Language Math Reasoning Authors: Qiqian Fu Published: 2026-08-22 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2608.21811v1 Reinforcement learning for vision-language math reasoning starves under sparse reward: on a pool of 20,830 visual-math problems where Qwen2-VL-2B answers 3.6% of rollouts correctly, 85-97% of GRPO rollout groups are entirely wrong and contribute zero gradient. We train eleven methods under identical conditions in this regime, each injecting a different prior: text (reference-solution hints), distribution (on-policy distillation from a 7B teacher), and value (a value-pretrained critic with an MSE or HL-Gauss categorical loss). A prior helps exactly when it is delivered: the six arms whose prior effectively reaches the policy separate with no overlap from the remaining five -- the no-prior baseline and four arms whose prior is teacher-capped, gated away, or lost to a mis-parameterized critic -- both on the pooled in-domain metric and on cross-domain transfer (DynaMath). The central finding, however, concerns evaluation: one slice of the in-domain pool -- long used as this project's general-distribution check -- anti-correlates with genuine cross-domain transfer (Spearman rho = -0.74, n = 11 arms, permutation p = 0.011), while the hardest in-domain slice predicts it closely (rho = +0.89, p < 0.001). We attribute the inversion to a near-chance multiple-choice subset that rewards models for not having changed; read through it, the best cross-domain method looked mediocre and the worst looked like the champion. Among the methods, hint-guided exploration -- not UFT's auxiliary loss -- drives hint gains, and replacing the critic's MSE loss with HL-Gauss cross-entropy is worth +14.4 points in-domain. All accuracies are blind-judged, with paired exact tests. --- ### MSM-Mem: A Universal Medical Structured Multimodal Memory Framework for Medical AI Agents Authors: Md Asaduzzaman Jabin, Khoa Le, Lin Zhao, Tianming Liu Published: 2026-08-22 Categories: cs.LG, cs.CV arXiv: https://arxiv.org/abs/2608.21810v1 Clinical decision-making is inherently experience-driven: physicians progressively refine their reasoning by synthesizing patient history, multimodal observations, and prior diagnostic experiences across interactions. In contrast, current multimodal large language model (MLLM)-based medical AI agents largely operate as stateless inference systems, generating decisions independently for each interaction without retaining or internalizing experiential knowledge. This discrepancy limits their ability to progressively improve reasoning reliability through usage and adapt to longitudinal patient contexts in real-world clinical workflows. In this study, we propose Medical Structured Multimodal Memory (MSM-Mem), an agentic memory framework that enables medical AI agents to evolve through accumulated clinical experiences. MSM-Mem organizes heterogeneous clinical experiences into semantic, episodic, and visual memory and incrementally updates them during inference, allowing the agent to retrieve prior experiences to inform current reasoning and progressively refine decision-making over time. Evaluations on MoE-LLaVA backbones demonstrate consistent performance improve- ments with further gains observed through continued usage. In general, MSM-Mem offers a viable pathway toward medical AI agents capable of evolving their reasoning competence in a manner analogous to the way clinicians learn from practice over time. --- ### MCite-RL: Towards Reliable Multimodal RAG via Citation-enhanced Agentic Reinforcement Learning Authors: Suifeng Zhao, Zida Liu, Xinyu Lei, Lei Sun, Jun Gao, Sujian Li Published: 2026-08-22 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.21808v1 Multimodal Retrieval-Augmented Generation (RAG) with visual citation is crucial for ensuring the traceability and verifiability of MLLMs. However, current RAG and SFT-based methods struggle to achieve robust cross-modal reasoning, causing imprecise visual citations or decoupling between the citation and the generated answers. To address these limitations, we propose MCite-RL, a citation-enhanced agentic reinforcement learning framework designed for reliable multimodal RAG. MCite-RL introduces an Agentic Refinement module for visual citation that employs iterative retrieval, reasoning, and recursive cropping to progressively narrow the search space, transforming citation into a dynamic, evidence-driven reasoning process rather than a static step. Furthermore, we incorporate a Citation-enhanced Reward mechanism that integrates both process-level and outcome-level feedback within a reinforcement learning paradigm to jointly optimize answer accuracy and source traceability. Extensive experiments on benchmarks such as Wiki-VISA, FinRAGBench-V, and MMLongBench-Doc demonstrate that MCite-RL effectively achieves the joint optimization of citation precision and answer quality. --- ### More Computational Resources Do Not Ensure Higher Scholarly Impact: Evidence from Leading NLP Conference Papers Authors: Shuai Chen, Tong Bao, Jitong Peng, Chengzhi Zhang Published: 2026-08-22 Categories: cs.CL, cs.AI, cs.CY arXiv: https://arxiv.org/abs/2608.21806v1 Computational resources are increasingly central to NLP research, but how closely reported GPU capability aligns with scholarly impact remains unclear. We analyze 13,921 ACL, EMNLP, and NAACL main-conference papers published between 2020 and 2025, using GPU resources as our operational measure of computational resources. From full texts, we extract GPU models and counts, standardize each paper's largest reported configuration into a comparable hardware-capability measure, and link these data to citation, award, topic, and institutional metadata. GPU reporting became more common but remained incomplete, while reported capability increased mainly through newer hardware generations and medium-scale multi-GPU configurations. Resource concentration substantially exceeded impact concentration: the annual top 20% of GPU-quantifiable papers accounted for 83.9%-89.9% of reported GPU capability, but only 27%-32% of citations and 20%-33% of paper awards. In adjusted models, a tenfold increase in aggregate reported GPU capability was associated with a 3.52-percentage-point increase in within-NLP topic-year citation percentile, but increased model R^2 by only 0.0042. GPU count showed more consistent positive associations with citation and award outcomes than newer hardware generation. Overall, reported GPU resources are associated with scholarly impact but provide little standalone explanation of research influence. --- ### FlashReg: GPU-Accelerated 3-Clique Point Cloud Registration for Real-Time Correspondence-to-Pose Estimation Authors: Ziyang Yu, Xiang Li, Qiong Chang, Jun Miyazaki Published: 2026-08-22 Categories: cs.CV, cs.DC arXiv: https://arxiv.org/abs/2608.21804v1 Graph-based point cloud registration achieves high robustness by identifying geometrically consistent correspondence sets, but constructing second-order compatibility graphs and enumerating candidate cliques remain compute- and memory-intensive. This work presents FlashReg, a GPU-oriented correspondence-to-pose estimator that avoids materializing the dense scored second-order graph. Its Fast First- and Second-Order Graph (FFSOG) construction builds a capacity-bounded sparse second-order graph directly from the binary first-order graph. A dataflow-optimized three-node clique (3-clique) search then selects pivots from compact per-row candidate pools and enumerates triples through sorted sparse-neighborhood intersections. Across indoor and outdoor benchmarks, FlashReg reduces correspondence-to-pose latency by 2--3x relative to TurboReg at comparable registration recall, while using about 50% of its peak allocated tensor memory on an embedded GPU. These results make FlashReg suitable as a high-throughput registration backend within onboard perception pipelines. --- ### ExplainGuard: A Zero Trust Framework for Post-Hoc Explanation Integrity Guarantees in Blackbox XAI Models Authors: Maraz Mia, Shovan Roy, Mir Mehedi A. Pritom, Maanak Gupta Published: 2026-08-22 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2608.21803v1 As machine learning (ML) models are increasingly deployed in high-stakes environments, explainable AI (XAI) methods like SHAP and LIME have become essential for regulatory compliance and trust. However, the current auditing paradigm relies on an implicit "chain of trust" where third-party auditors are assumed to be trusted. Recent research demonstrates that this assumption is flawed and adversarial auditors can manipulate XAI explanations through manipulation attacks such as output shuffling or scaffolding out-of-distribution (OOD) to conceal model biases while maintaining high prediction accuracy aiming for fairwashed explanation. In this paper, we introduce a novel defense framework, ExplainGuard, that leverages a Zero-Trust architecture (ZTA) design to be incorporated within the XAI explanation supply chain and ensures the integrity of the generated explanation. This framework would help us to replace the ambiguous default assumption of "auditor is trustworthy," with a continuous "verify-then-trust" approach. Our design architecture establishes a Policy Decision Point (PDP) that enforces three distinct pillars of verification before any explanation is released to the user: (1) asset integrity via behavioral fingerprint to detect model substitution, (2) semantic validity using axiomatic consistency checks to reject mathematically impossible explanations, and (3) feature faithfulness verification utilizing a ranking stability approach with minimal computational overhead. Finally, we evaluate how ExplainGuard can effectively neutralize state- of-the-art explanation manipulation attacks while transforming the auditing process into a verifiable operation. --- ### SAFE-G: Structure-aware Faithful Evidence-guided Generation for Knowledge-based Visual Question Answering Authors: Long Shu, Shuochen Liu, Wei Chen, Junda Lin, Zhi Zheng, Huijun Hou, Tong Xu Published: 2026-08-22 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2608.21796v1 Knowledge-based Visual Question Answering (KB-VQA) aims to answer queries that necessitate reasoning over external knowledge sources beyond the visual content. Typically, current methods fuse multimodal features to retrieve external information, subsequently leveraging Multimodal Large Language Models (MLLMs) to derive answers from the retrieved evidence. However, these methods often struggle to capture structural associations within complex contexts to effectively filter noise. Furthermore, they frequently fail to ensure that the reasoning process remains strictly faithful to the retrieved evidence. To address these challenges, we propose SAFE-G, a Structure-Aware Faithful Evidence-guided Generation framework, which enables precise evidence localization and trustworthy reasoning. Specifically, we first employ a coarse-grained hybrid search fusing visual and textual modalities to recall candidate documents, and subsequently implement a structure-aware fine-grained graph retrieval that captures structural dependencies to filter noise and pinpoint precise evidence. Moreover, we introduce a reinforcement learning (RL) strategy with an evidence-grounded reward that assigns credit to correct answers only when the selected evidence is correct. This strict alignment constraint compels the model to anchor its response in the retrieved context, effectively enhancing its capability to locate evidence via multimodal features and perform faithful reasoning. Extensive experiments on the Encyclopedic-VQA and InfoSeek benchmarks demonstrate that SAFE-G outperforms prior methods by a margin of 8.9% and 3.5%, substantially enhancing the overall reasoning accuracy. Our source code is publicly available at: https://github.com/MINE-USTC/SAFE-G. --- ### Lexical Coupling in GUI Element Grounding: Sentence Embeddings Track Labels across Mobile and Web Authors: Qijia Chen, Giulio Jacucci Published: 2026-08-22 Categories: cs.CL, cs.AI, cs.HC arXiv: https://arxiv.org/abs/2608.21794v1 GUI grounding evaluations that expose UI elements as text metadata often treat high instruction-element embedding similarity as evidence of semantic grounding. Across three mobile and web benchmarks, we show that this interpretation is frequently confounded by visible-label recovery. Lexical baselines remain competitive at top-1, label-poor targets remain weak for text-only methods, and encoder top-1 hits are predictable from lexical rank, candidate-pool size, and label type. We evaluate each action as a same-screen ranking task, comparing five off-the-shelf single-vector encoders with lexical baselines. Encoders recover some lexical misses, but deployable fusion gains are much smaller than target-aware oracle gains. These findings show that embedding-based evaluations can conflate visible-label recovery with semantic GUI grounding. Embedding-based evaluations should therefore report lexical baselines, label-type stratification, and deployable-fusion diagnostics. Our released repository provides analysis scripts and detexted per-step panels: https://github.com/qijia123/lexical-coupling-release. --- ### HIRA: A Human-in-the-Loop Retrieval-Augmented Cascade for Document Classification in Regulated Industries Authors: Shangxuan Tian, Yanhui Chen, Carlos Queiroz Published: 2026-08-22 Categories: cs.AI, cs.CV, cs.IR, cs.LG arXiv: https://arxiv.org/abs/2608.21792v1 Document classification in regulated industries is constrained by data residency, limited cold-start labels, scarce review capacity, and costly model-governance procedures. We present HIRA, a training-free, on-premises retrieval-augmented cascade for document classification in regulated deployments that combines BM25 over OCR text, dense text embeddings, and image-level representations through validation-calibrated weighted reciprocal-rank fusion. Confident documents are classified directly by retrieval; uncertain or visually confusable documents are passed to a locally hosted LLM verifier, which receives the OCR text, retrieved exemplars, label descriptions, and confusion-specific terms. When the verifier remains uncertain, the document is sent to human review. Each correction is stored as a margin-weighted retrieval exemplar and updates a Dirichlet-smoothed confusion graph, letting the system improve without updating model weights. On a private 80-class trade-finance corpus, HIRA processes the full 30,233-document production stream while requesting human correction for only 1,945 documents (6.4%), improving Macro-F1 from 0.6218 to 0.8548. On the corrected Tobacco-3482 benchmark, HIRA reaches 0.9423 Macro-F1 with a locally hosted DeepSeek-R1-Distill-Qwen-32B verifier, 17.4 percentage points above the zero-shot LLM baseline, while invoking the verifier for only about 40% of documents and reducing LLM calls by approximately 60%. With 518 human corrections (24.8% of the pool), HIRA matches the fully labelled pool oracle, in which all 2,086 pool documents are indexed with their ground-truth labels. These results show that selective human feedback and retrieval-memory adaptation can be a practical alternative to repeated model retraining for long-tail document classification in regulated deployments. --- ### UniDiffFusion: A Unified Diffusion Framework for Multi-Task and Degradation-Robust Image Fusion Authors: Xingxin Xu, Siqi Zhao, Xin Li, Xinjie Yao, Yiming Sun, Pengfei Zhu Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21786v2 General image fusion aims to integrate complementary information from multiple source images, but existing methods often rely on task-specific models and struggle to maintain robust performance under diverse degradation conditions. In this paper, we propose UniDiffFusion, a unified diffusion framework for multi-task and degradation-robust image fusion. UniDiffFusion leverages the strong generative prior of a pretrained diffusion model to establish a shared fusion backbone across heterogeneous fusion tasks, while introducing task- and degradation-aware conditional adaptation to accommodate their distinct information-selection requirements. Specifically, we employ task prompt modulation to progressively adapt the shared diffusion representations to different fusion objectives, and develop a degradation prompt router to dynamically retrieve degradation-aware priors and restore corrupted source features before fusion. Furthermore, an application prompt bank is introduced to incorporate task-oriented semantic guidance for downstream applications, such as object detection and semantic segmentation, without altering the shared fusion and restoration pathways. The proposed framework is trained in a progressive manner to decouple fusion learning, degradation-aware restoration, and application-specific adaptation, thereby reducing interference among heterogeneous objectives. Extensive experiments on visible-infrared, multi-exposure, and multi-focus image fusion demonstrate that UniDiffFusion achieves superior fusion quality and robustness under both clean and degraded conditions. Moreover, UniDiffFusion consistently improves downstream detection and semantic segmentation performance, demonstrating its effectiveness as a unified diffusion framework for both perceptual fusion and task-oriented vision. --- ### DefaultShift: Auditing Semantic Default Shift in Accelerated Text-to-Image Models Authors: Xuanhua Yin, Chuanzhi Xu, Shunqi Mao, Wei Guo, Weidong Cai Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21784v1 Few-step text-to-image models increasingly replace slower generators, yet acceleration can silently change distributions over unspecified attributes even when individual outputs remain plausible and aligned. We call these distributions semantic defaults and their change under replacement semantic default shift. Existing quality, preference, and diversity evaluations do not test whether a replacement preserves its reference model's semantic defaults. We introduce DefaultShift, a paired audit that labels repeated samples with closed semantic vocabularies, measures probability-mass movement, and separates interpretable ranking from confirmatory cross-fit inference. Across 14 reference and replacement pairs, adjusted color discrepancies range from 0.054 to 0.303 with recipe-specific directions. A 1,000-image human audit reproduces the ordering. We further introduce DefaultShift-Select, an offline calibration method that reduces human-measured shift by 10.3 percent to 35.1 percent across Turbo, DMD2, and FLUX without material quality loss. Under balanced evaluation, selected data recover 4.3 accuracy points and 7.5 worst-group points over uncalibrated replacement data. DefaultShift makes semantic preservation under acceleration measurable and actionable. --- ### A Fixed-Radius Distance-Band Benchmark for Dimensionality-Reduction Fidelity Authors: Yoshio Takaeda Published: 2026-08-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.21779v1 Dimensionality-reduction (DR) methods are routinely judged by how well each point's k nearest neighbors survive the 2-D embedding (recall@k, trustworthiness, continuity). We argue this family is a biased measure of distance fidelity: its per-point variable radius and hard inclusion threshold favor neighbor-graph methods (t-SNE, UMAP) and penalize methods that preserve absolute distances. We instead score DR fidelity with a fixed-radius distance-band Shepard rho: the Spearman correlation between high-D and 2-D pairwise distances, restricted to cumulative distance bands so that near and global structure are reported separately, with every point judged on the same absolute radius. On synthetic datasets with known ground-truth geometry (non-uniform density, dense clusters, a closed-loop transition, off-subspace outliers, imbalanced two-population data) at realistic noise (SNR=1, D=768, N=1000), we benchmark eight methods -- PCA, Isomap, t-SNE, UMAP, PyMDE, PCC, DREAMS, and the closed-source toorPIA -- and show that (i) high global Shepard rho can coexist with a ~93x collapse of within-cluster scale, invisible to rank-based scores but obvious in a value-based over-compression metric; (ii) recall@k and the fixed-radius band disagree systematically, in the direction the bias predicts; (iii) a membership-restricted Shepard rho resolves single-point and minority-population questions that many-pair statistics cannot -- questions on which even DREAMS, a recent local-plus-global hybrid, fails silently. A supplementary out-of-sample (addplot) test asks whether a never-seen anomaly lands outside the normal region and whether its direction identifies its source. All metrics are computed exactly on all pairwise distances, independently of any method's internals, and every number is reproducible offline: the closed-source method's output coordinates (not its algorithm) are committed to the artifact. --- ### SpatialDiff: 3D-Aware Object Movement via Implicit Spatial Modeling Authors: Zheng Liu, Zijian He, Huiguo He, Weizhi Zhong, Yejun Tang, Huan Yang, Kun Gai, Guanbin Li Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21776v1 Recent advances in image editing allow impressive manipulation of objects, existing methods still struggle to handle spatial movement in complex scenes, such as objects span different depth layers or are partially occluded. Most image editing methods focus solely on prior information from 2D datasets, emphasizing planar features while lacking support for spatial structures. Even approaches that incorporate explicit positional information fail to capture true 3D spatial relationships, thus limiting accurate object movement in complex scenes. In this paper, we present SpatialDiff, a method that effectively captures 3D spatial structures, enabling precise and consistent object movements in complex scenes. Our core innovations are twofold: (1) Implicit 3D Spatial Modeling, which introduces 3D prior knowledge and enables the model to internally build a comprehensive understanding of the three-dimensional spatial structure; and (2) Global Spatial Supervision, which constrains the latent spatial features to enable the model to perceive changes in object spatial positions caused by editing operations. Experimental results demonstrate that our method significantly improves the accuracy and fidelity of spatial movement in complex scenes. --- ### No One Model Catches Every Harm: Benchmarking Content Moderation Across Safety Scenarios Authors: Afshin Orojlooyjadid, Hitesh Patel Published: 2026-08-22 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.21775v1 Large Language Models (LLMs) are increasingly deployed in real-world applications, yet they remain vulnerable to generating harmful content. From adversarial jailbreaks that bypass safety filters to implicit hate that evades detection, the range of risks these models pose continues to grow. While both specialized content moderators and general-purpose LLMs are being used as safety layers, the question of which model is best suited for which type of harmful content remains unanswered. We present the most comprehensive evaluation of LLM safety capabilities to date, systematically testing \textbf{53} models across \textbf{11} datasets that we organize into four distinct categories. Our evaluation under both prompt-only and prompt-response settings uncovers critical blind spots: large frontier models that lead on one category fall significantly behind smaller, specialized alternatives on others, and real-world conversational safety remains largely unsolved across all model families. These findings challenge the assumption that scale alone ensures safety, and provide the community with a structured framework for informed model selection. --- ### Physics-Knowledge-Guided Hybrid Neural Learning for Arctic Sea Ice Concentration Evolution and Short-Range Prediction Authors: Maqun Zhang, Feng Gao, Wankun Chen, Hui Yu, Yanhai Gan, Junyu Dong Published: 2026-08-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.21767v1 Accurate modeling of sea ice concentration (SIC) evolution is essential for polar climate assessment and short?range sea ice prediction. Numerical and data-driven approaches constitute major foundations for SIC modeling, but the former often require complex parameterizations and substantial compu?tation, whereas the latter rarely encode physical dependencies explicitly. This study presents the Physics-Informed Hybrid Ice Model (PIHIM), a differentiable data-driven hybrid ice model for daily SIC evolution that organizes its network structure according to the physical dependencies encoded in the sea ice continuity equation and explicitly accounts for dynamical transport, ther?modynamically driven areal growth and loss, and unresolved local processes. PIHIM preserves the representation capacity of deep learning while providing a process-decomposed formulation of ice displacement, freeze-melt areal change, and local error closure. Two evaluation settings are adopted: reanalysis-forced simulation examines SIC evolution stability under reanalysis forcing, and forecast-forced prediction assesses short-range performance un?der forecast-forced conditions, with reanalysis and observational SIC serving as verification references. Results indicate enhanced ice-edge preservation and error-growth control in reanalysis?forced simulation, while PIHIM retains measurable short-range prediction skill under forecast-forced conditions. Our code will be made publicly available after the paper is accepted. --- ### Evaluation Awareness in Language Models: Representation, Verbalization, and Control Authors: Farzaneh Heidari, Amin Memarian, Guillaume Rabusseau Published: 2026-08-22 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2608.21766v1 Both capability and safety benchmarks rest upon the assumption that the behavior of language models undergoing a test is informative about their behavior in deployment. This assumption can fail, should models infer that they are being evaluated and condition their response on such context. This hypothesis, termed ``evaluation awareness'', has been observed in frontier and open-weight language models alike. We provide a systematic study of this phenomenon, by probing for it across six language models (from four families and three sizes) and three metrics. More precisely, we examine whether (i) being under evaluation is linearly represented within the models' activations space, (ii) it is verbalized in their output tokens (as scored by an LLM-as-judge), and (iii) steering causally affects their behavior. For the open-checkpoint Olmo models, we further test these measures at every training stage. In doing so, we report that evaluation awareness is linearly decodable from the residual streams of every model (best AUROC $\geq 0.7$). By contrast, these representations align only in part with verbalization: their correlations and mutual information are nonzero in some settings, yet vary substantially across models, layers, and readout choices. Nevertheless, steering along probe-derived directions can shift the verbalization scores. Finally, a comparison across the Olmo checkpoints reveals that evaluation awareness is already present within base models, becomes amplified throughout the stages of supervised fine-tuning, and remains stable thereafter---unlike the effects of steering, that grow more pronounced at every successive training stage. These results show the need for evaluations to account for the disjunction between what models represent internally, what they verbalize, and their steering. --- ### LiteEvent-AE: Lightweight Autoencoder for Event-Based Vision on Low-Latency Energy-Constrained Edge Devices Authors: Riadul Islam, Joey Mule, Dhandeep Challagundla, Shahmir Rizvi, Sean Carson, Rachit Saini Published: 2026-08-22 Categories: cs.CV, cs.AI, eess.IV arXiv: https://arxiv.org/abs/2608.21764v1 Event-based vision has emerged as a promising paradigm for energy-aware artificial intelligence (AI), offering sparse, low-latency visual signals that reduce redundant data processing and support sustainable edge computing. However, the asynchronous and noise-prone nature of event streams creates challenges for conventional deep learning models, which are often too computationally intensive for low-power embedded platforms. This work presents a compact and configurable event-driven autoencoder that efficiently compresses neuromorphic data while preserving essential spatiotemporal structure for downstream inference. The architecture integrates lightweight convolutional encoding with robust performance under adaptive event thresholding and a minimal classifier head, enabling substantial reductions in computational cost without degrading recognition fidelity. Extensive evaluations on the Smart Event Face Dataset (SEFD) and Event-Based Crossing Dataset (EBCD) show that the proposed framework achieves competitive or superior accuracy compared to YOLOv9 while requiring up to 35.6$\times$ fewer parameters. To assess real-world sustainability, the model is deployed on resource-constrained hardware: a Raspberry Pi 4B and a NVIDIA Jetson Nano. On NVIDIA Jetson Nano, it delivers real-time throughput of 44.8 FPS. On a Raspberry Pi 4B CPU, the 50\% autoencoder classifier consumes 16.19 J for the evaluated inference workload, corresponding to approximately 726.3$\times$ lower energy consumption than YOLOv9 under the same evaluation protocol. These results demonstrate the potential of compact event-driven models to advance environmentally conscious, low-power AI systems for high-speed perception in autonomous, mobile, and embedded computing environments. --- ### Learning to Look Again: Loss-Gap Supervision for Free-form Crop Routing in Vision-Language Models Authors: Jinchang Zhu, Rong Fu, Yi Ding, Chenghao Wu, Ying Liu, Menglin Yang Published: 2026-08-22 Categories: cs.CV, cs.CL arXiv: https://arxiv.org/abs/2608.21762v2 Vision-language models (VLMs) fail many detail-centric questions for a concrete reason: the answer is visible in the image, yet lost after the image is compressed into a low-resolution global view. Allocating more visual tokens to every query improves some OCR and document cases, but it spends computation indiscriminately and can disturb tasks that rely on global context. We propose GapSight, a framework for learning visual re-reading: a VLM first takes a global glance, then selectively returns to a free-form region when the question calls for local evidence. The supervision comes from the target model's own failure signal. Offline, we compare answer loss or multiple-choice option margin under a global-only view and candidate crop-augmented views; crops that improve the target answer become model-specific review labels. A lightweight free-form crop router distills these labels into a one-shot inference policy that predicts whether to review, expected utility, and a continuous crop box from the global state. Across LLaVA-1.5-7B, InternVL2.5-8B, and Qwen2-VL-2B-Instruct, GapSight improves the Base no-zoom baseline on six benchmarks spanning OCR, documents, charts, infographics, VStarBench, and MME-RealWorld-Lite. On InternVL2.5-8B, GapSight raises the six-benchmark average from 52.25 to 64.29, above CropVLM (57.16), ViCrop (55.84), and ZoomRefine (54.43). Mechanism analyses show that the router rescues concrete wrong answers, adapts its action rate by task, and forms a favorable token-performance profile. These results position loss-gap supervision as a practical route to teaching VLMs when and where to look again. --- ### What Does CLIP Learn for Regional Geolocalization? Probing Visual Cues and Scene Configuration After Adaptation Authors: Changyu Lee, Yeonsoo Park, Abdullah Alfarrarjeh, Seon Ho Kim Published: 2026-08-22 Categories: cs.AI, cs.CE, cs.CV arXiv: https://arxiv.org/abs/2608.21761v1 Large collections of street-view imagery provide rich visual information about urban environments, but extracting fine-grained geographic information from such data remains challenging. In particular, fine-grained regional geolocalization is challenging because nearby areas often share coarse geographic cues. We study regional geolocalization within a metropolitan area and ask whether pretrained CLIP features are sufficient for regional discrimination, and what visual information supports performance after adaptation. Using 9,085 street-view images from eight Greater Los Angeles regions, we compare zero-shot CLIP, frozen-encoder readouts, partial encoder updating, Low-Rank Adaptation (LoRA), and full fine-tuning. Frozen readouts remain near the 39.03% zero-shot accuracy, whereas encoder adaptation achieves 75.94-82.10%. Full fine-tuning also reduces the mean distance to the predicted region center from 12.30 km to 3.86 km. We probe these gains through semantic cue removal, appearance reduction using edge maps and blur, and scene-configuration disruption using patch scrambling. Adapted models achieve higher edge and blur accuracy and switch 42.92-45.56% of predictions after scrambling, compared with 10.79-14.60% for frozen methods. However, adaptation does not improve the fraction of performance retained after appearance reduction, while vegetation and sky remain influential. A Caltech101 control further shows that scrambling sensitivity is not unique to geolocalization. Overall, encoder adaptation substantially improves nearby-region discrimination and is associated with greater sensitivity to intact scene configuration, without evidence that coarse structure alone becomes sufficient for prediction. These conclusions concern viewpoint variation near known locations rather than geographically disjoint generalization. --- ### How Architecture and Training Affect TPC Representations Across Experiments Authors: Tyler Wheeler, Michelle P. Kuchera, Raghuram Ramanujan, William Sieland, Ryan Krupp, Yassid Ayyad, Daniel Bazin, Connor L. Cross, Hoi Yan Ian Heung, Andrew J. Jones, Ruchi Mahajan, Saiprasad Ravishankar et al. Published: 2026-08-22 Categories: cs.LG, cs.CV, nucl-ex, physics.ins-det arXiv: https://arxiv.org/abs/2608.21756v2 Deep-learning efforts have increasingly shifted toward foundation model approaches. In experimental physics, this allows models and learned representations to be reused beyond the experiments in which they were developed. This work evaluates the reusability of representations across experiments and detector systems using probes on frozen encoders. These probes reveal task-relevant structure before downstream adaptation, complementing fine-tuning. Together with random-weight controls, they distinguish contributions from architecture and encoder training that downstream performance alone cannot resolve. Time projection chamber (TPC) data provide a useful testbed because events from TPC systems can be represented as variable-length sparse tensors, while detector geometries, event topologies, and scientific tasks can differ substantially. We investigate whether fixed-dimensional TPC event representations can be reused across classification tasks, experiments, and detector systems. Sparse ResNet and PointNet-style encoders produce 512-dimensional embeddings for four datasets from the GADGET II TPC and AT-TPC. Randomly initialized encoders isolate the contribution from architecture before supervised training. We then train each encoder on a classification task, freeze its parameters, and train a linear or nonlinear probe for each downstream task. We find that this architecture-induced structure remains useful across experiments and detector systems. The randomly initialized PointNet-style representation is highly informative on several tasks. The two architectures organize their embedding spaces differently, but neither exhibits a large, systematic loss of utility cross-detector. These results show that architecture is a major source of task-relevant structure in TPC embeddings and should be treated explicitly when assessing representation learning and developing reusable detector models. --- ### ECHO: A Cognitively Inspired, Auditable Memory Plane for Long-Horizon Agents Authors: Yu Qian, Hong Miao, Boyang Guo, Tingyi Jiang, Shan Zhao, Tianxing Le, Lintian Li, Meng Liu Published: 2026-08-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.21755v1 Long-horizon agents need memory that identifies relevant experience, resolves revisions, and exposes checkable provenance. We present ECHO (Embodied Context and History Orchestration), an auditable memory architecture and service prototype inspired by episodic encoding, consolidation, contextual reinstatement, reconsolidation, and executive control. This is functional inspiration, not neural equivalence; the empirical analysis focuses on retrieval and context construction. Development runs reach 96.29% Hit@10 and 73.64% turn Recall@5 on 1,536 LoCoMo category 1-4 questions, and 97.60% Hit@10, 88.84% turn Recall@5, and 88.71% session Recall@5 on all 500 LongMemEval-S questions. A five-history BEAM gate fails, and in a separate matched 91-question QA sample Mem0 OSS scores 64.84% versus ECHO's 41.76% (exact McNemar p = 0.00107), with a history-cluster interval crossing zero. A post-hoc audit found source-specific phrases in the query-expansion rules. Although no gold answer field entered the runtime, expansion-enabled retrieval scores are therefore descriptive development measurements, not independent confirmation. --- ### Fidelity-Diversity-Consistency (FDC): Data Pruning for Remote Sensing Change Detection Authors: Dongyao Zhu, Ranga Raju Vatsavai Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21754v1 Despite the success of data pruning (DP) in reducing training data sizes and improving downstream model performance in classification and segmentation tasks, its potential in remote sensing change detection remains unexplored. For the first time, we benchmark six representative DP methods across building- and forest-change datasets, CNN- and transformer-based models, and three pruning budgets, and show that existing baselines yield no reliable advantage over random selection. Notably, even the strongest evaluated baseline, Feature Diversity, is matched or exceeded by $\sim$33\% of randomly sampled subsets. To understand the underlying mechanism, we conduct a systematic regression study over 540 randomly sampled data subsets, characterizing each with four descriptors covering label statistics, image diversity, and feature-space geometry. Random Forest models show that \emph{change distribution fidelity} is the most prominent factor in determining the quality of change detection data subsets, a property absent from the existing pruning literature. Our analyses further show that pixel-wise image diversity and label-feature consistency are secondary factors. We translate these findings into Fidelity-Diversity-Consistency (FDC), a simple two-stage pruning method that shows consistent improvements over existing baselines across change detection benchmarks and backbones, especially at lower pruning ratios. Code is available at \href{https://github.com/ddydyd32/fidelity-diversity-consistency}{https://github.com/ddydyd32/fidelity-diversity-consistency}. --- ### Width-Independent Compressibility of Deep Neural Networks Authors: Hong-Yi Wang, Mingze Wang, Liu Ziyin Published: 2026-08-22 Categories: cs.LG, cond-mat.dis-nn, cs.IT arXiv: https://arxiv.org/abs/2608.21752v1 It has long been known that well-trained neural networks can be compressed very strongly without affecting their performance, an important phenomenon that remains poorly understood. We prove a uniform compressibility theorem for deep multilayer perceptrons with analytic activations. For a deep, wide fixed teacher network, there exists a narrow (same depth) network that approximately represents the same function as the original. The reachable compressed width is strikingly independent of the original width, but is $O((\log(1/\varepsilon))^{d_{in}})$, where $\varepsilon$ is the error budget and $d_{in}$ is the effective input dimension. Our construction involves a novel derivative-matching technique which is aware of the low-dimensional input, and a layer-wise reweighting that preserves the input-output mapping. --- ### FCPRAG: Fusion-Controller Parametric Retrieval-Augmented Generation for Stable Multi-Passage LoRA Injection Authors: Jinchang Zhu, Jindong Li, Yi Ding, Xiaojian Nie, Rong Fu, Shuangyong Song, Haowei He, Menglin Yang Published: 2026-08-22 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.21750v1 Parametric retrieval-augmented generation (PRAG) injects retrieved evidence into a large language model (LLM) through passage-specific LoRA adapters, reducing reliance on long in-context prompts. When multiple passages are retrieved for the same query, however, evidence-level fusion becomes a bottleneck: equal-weight merging can amplify weak or conflicting evidence, and translating retrieval signals into fusion weights often requires fragile global tuning. We propose FCPRAG, a fusion-controlled parametric RAG framework that adds a lightweight controller for retrieval-conditioned, sample-level adapter fusion. The controller predicts per-passage fusion scores together with sample-level calibration signals, including a mixing gate and an adaptive temperature, enabling fusion that stays selective under informative retrieval signals and conservative under uncertainty. FCPRAG is trained with merge-aware supervision derived from each adapter's marginal contribution within a multi-adapter merge, using training data only. We further show that a single dataset-level temperature is suboptimal under heteroscedastic retrieval uncertainty, motivating sample-level adaptation. Experiments on HotpotQA, 2WikiMultiHopQA, PopQA, and ComplexWebQuestions (CWQ) across three LLM backbones show that FCPRAG consistently improves F1 over standard RAG and parametric RAG baselines, with gains of up to 4.65% on 2WikiMultiHopQA and 7.55% on CWQ, while also reducing tuning cost and improving robustness under retrieval perturbations. --- ### Calibrate What You SHIP: Post-Selection Risk Control for Verifier-Guided Text-to-Image Generation Authors: Xuanhua Yin, Shunqi Mao, Wei Guo, Chuanzhi Xu, Weidong Cai Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21748v1 Verifier-guided text-to-image systems increasingly use test-time search to select, refine, or stop among multiple candidates, yet release thresholds are often calibrated on individual images. This creates a candidate-to-policy calibration mismatch: search changes both which prompts receive an output and which candidate is released, so candidate-level risk control need not imply control of released-output risk. We formalize this estimand shift through prompt reweighting and within-prompt selection, and introduce SHIP, Selection-aware Held-out calibration of Inference Policies. SHIP runs or replays the complete deployed policy on held-out prompts, evaluates the image it actually releases using an independent target judge, and selects the most permissive threshold whose risk upper bound satisfies a prescribed budget. For replayable policies with a prespecified threshold grid, simultaneous confidence control provides finite-sample validity. Experiments across fixed, sequential, and adaptive T2I inference procedures show that policy-level calibration recovers lower-risk operating points while exposing policy-dependent tradeoffs among risk, coverage, and compute. On GenEval2 with FLUX at N=16, a pooled-candidate threshold yields released risk 0.310, whereas SHIP reduces it to 0.162. Across 200 cached-stream splits, the fixed-grid certificate has no target crossing. Reliable inference-time scaling therefore requires calibrating the output distribution induced by the complete deployed policy. --- ### Architecture as Capability Equalizer for Coding Agents Authors: Arquimedes Canedo Published: 2026-08-22 Categories: cs.SE, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2608.21747v1 LLM-based coding agents generate complete software systems from high-level descriptions, yet little is known about how the format of architecture specifications affects the quality of generated code or whether this effect depends on model capability. We present a controlled experiment comparing five informationally equivalent specification formats (informal prose, Mermaid diagrams with constraints and ADRs, OpenAPI, C4/Structurizr DSL, and TypeScript interface contracts with ArchUnit-style rules) across six models from three vendor families (Anthropic Claude, OpenAI GPT, Google Gemini). Across 90 multi-turn agent trials, specification format shows a strong format x model interaction. On the strongest models (Sonnet 4.6, GPT-5), format barely matters (quality spread 0.17-0.92). On weaker models, format produces spreads of 0.83-2.42 points, with code-proximate formats (OpenAPI, TypeScript contracts) recovering most of the capability gap. Mid-tier models can consume more tokens than frontier models for worse output when they enter compilation debugging loops that stronger models avoid. Self-validation rates collapse from 100% (Sonnet) to 0% (Gemini Flash) across the capability spectrum. TypeScript contracts triple API route coverage for the weakest model (33% to 100%). Structured architecture specifications serve as a capability equalizer, with value inversely proportional to model strength and the largest returns for cost-optimized deployments. --- ### First-Principles Atomistic Structure and Dynamics of Polyethylene During High-Pressure Radical Polymerization via Machine Learning Force Fields Authors: Bharatha K. Gunawardana, Teresa Shah, Bicha Azizova, Deepa Ranabhat, Yizhi Song, Akshath Shastri, Srinjoy Ghose, Thomas E. Gartner, Hsin-Yu Ko Published: 2026-08-22 Categories: cond-mat.mtrl-sci, cond-mat.dis-nn, cs.LG, physics.chem-ph arXiv: https://arxiv.org/abs/2608.21741v1 Polyethylene (PE) is one of the most commonly used synthetic polymers. While the synthesis and processing protocols for PE are well established, precise experimental assignment of microscopic structures at atomistic resolution (i.e., the position of each atom) remains largely limited to highly crystalline systems. This gap is often addressed via computer simulations using empirical interatomic potentials, which use approximate but efficient descriptions of interatomic interactions to reach the length and time scales needed to describe macromolecules. These empirical potentials typically perform well for bulk and/or collective properties but face challenges with chemical realism for complex systems, e.g., during reactive processes. In this work, we address this challenge by combining the computational efficiency of a deep potential (DP) machine-learning force field and the chemical realism of first-principles van der Waals (vdW) corrected hybrid density functional theory (DFT) enabled by a SeA high-throughput framework. Using this approach, we study the structure and dynamics of PE oligomers and polymers in an ethylene solvent under common high-pressure (supercritical) radical polymerization conditions. We found that the local solvation environment of radical-containing PE oligomers converges for chain lengths greater than (n~6), suggesting extensibility of our oligomer-trained MLFF to significantly longer polymers. We then confirmed the extensibility of these models to long PE chains by characterizing the molecular weight scaling of single-chain structure and dynamics, which showed classic good solvent behavior. Our PE MLFF retained a consistent level of fidelity and stability across a wide range of thermodynamic state points and chain lengths, at full atomistic resolution, therefore paving the way towards first-principles-based polymer structure and property prediction. --- ### Adaptive Multilevel Twisted Sequential Monte Carlo for Rare Events Estimation in Language Models Authors: Zixuan Liu, Fangzheng Wu, Brian Summa, Zizhan Zheng Published: 2026-08-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.21736v1 Rare unsafe behaviors in large language models can remain practically significant even when their probability is extremely small, particularly at deployment scales involving millions or billions of interactions. Twisted Sequential Monte Carlo (SMC) provides a principled framework for rare-event probability estimation by learning twist functions that guide generation toward a target event. However, the standard twist learning framework relies on positive samples from the rare-event target distribution, which may be nearly absent before an informative twist has been learned, resulting in unreliable rare-event estimation. We propose Adaptive Multilevel Twisted SMC, which learns the rare-event twist through a sequence of progressively rarer intermediate events. At each level, the learned twist provides more informative positive examples for learning the next twist, ultimately leading to a more accurate final twist for the target rare event. Experiments across diverse tasks and model scales show that the proposed method produces more accurate rare-event probability estimates. By enabling more reliable discovery of hard-to-observe unsafe behaviors, our method provides a practical tool for strengthening the evaluation and safety alignment of deployed language models. --- ### Guidance for Prior Change via Density Ratio Estimation Authors: Yichen Zang, Song Liu, Jiun-Yi Lin Published: 2026-08-22 Categories: stat.ML, cs.LG, stat.ME arXiv: https://arxiv.org/abs/2608.21729v1 Simulation-Based Inference (SBI) serves as a vital framework for parameter inference in scientific fields where simulators involve intractable likelihoods, yet while amortized generative models offer rapid posterior estimation, they are often restricted by the specific priors used during training, thereby limiting their flexibility as prior knowledge evolves. To address this prior dependency, PriorGuide was introduced as an inference-time guidance method, but due to its intractable formulation, it relies on Gaussian approximations of the reverse transition kernel and Gaussian mixture model fitting for the prior ratio, both of which introduce systematic bias. Motivated by these limitations, we propose an unbiased test-time guidance framework that leverages Density Ratio Estimation (DRE) to learn a score guidance term, effectively decoupling the inference process from the prior training. Moreover, our framework remains agnostic to the specific density ratio estimators, making it a general and flexible framework for handling prior changes. Experimental results across multiple tasks demonstrate that our method matches or outperforms PriorGuide on C2ST and MMD in most tasks while maintaining robustness even under limited overlap between the training and target priors. Furthermore, we apply our method to Bayesian updating for parameter inference from planetary light-curve data, where it also demonstrates strong effectiveness and robustness. Code is available at https://github.com/a-chenchen/dre-based-prior-guidance . --- ### Reinforcement Learning on Benign Facts Amplifies Leakage of Memorized Private Data Authors: Renfei Zhang, Niloofar Mireshghallah Published: 2026-08-22 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2608.21727v3 Reinforcement learning with verifiable rewards (RLVR) is deployed to make models better at reasoning tasks, but its side effect on what models will divulge is under studied. Here we show that RLVR on facts increases extraction of personally identifiable information (PII) the instruct model had already memorized. We first confirm that instruct models have already memorized PII but leave them latent, rarely surfacing one when asked. We then apply RL on benign factual data that contains no PII of any kind, and re-probe: a targeted probe over name->email pairs, and an untargeted free-recall prompt that simply asks the model to list the addresses it knows. PII extraction rises sharply under both: on DeepSeek-V3.1, verbatim recall@k increases from 0.155 to 0.370, a 2.4x gain. The effect scales with model size: across three models spanning 8B to 671B parameters, absolute leakage is largest in the biggest model. Meanwhile model's reasoning abilities and refusal rates are retained, indicating that RL selectively changes which memorized information is accessible rather than broadly altering the model. In summary, memorized private data can be made markedly more extractable by training that never touches it. This gives an adversary a route to memorized data that requires no privacy-relevant training signal and no access to the data itself -- only the ability to fine-tune on something innocuous. --- ### Ask or Answer: A Decision Framework for Multi-Turn Health Misinformation Intervention Authors: Xiaoying Song, Anirban Saha Anik, Jinyu Liu, Qitao Tan, Geng Yuan, Lingzi Hong Published: 2026-08-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.21721v1 Correcting health misinformation in dialogue requires more than producing a factual rebuttal: users differ in what they know, what they believe, and what they need to hear, so an effective intervention often depends on first asking the right clarifying question. Yet existing methods either respond immediately or probe indiscriminately, treating clarification as either unnecessary or always beneficial. We propose Reward-Optimized Probe-and-Respond (RO-PnR), a framework that learns when asking is worth its cost. At each turn, RO-PnR chooses between probing for more information and committing to a final correction, guided by a turn-level reward that weighs the expected gain from probing against its interaction cost. To capture how user heterogeneity affects probing value, we model each simulated user with a latent state along health literacy and belief commitment. Experiments show that RO-PnR achieves the highest cost-adjusted utility across three health-misinformation datasets and three base models, using 30% fewer turns than always-probe baselines. --- ### PowerSlider: Exploiting Phase Asymmetry for LLM Serving under Demand Response Authors: Yueying Li, Jiayang Chen, Yuanfan Chen, Leo Han, Haoran Qiu, Esha Choukse, Rodrigo Fonseca, Udit Gupta Published: 2026-08-22 Categories: cs.DC, cs.AI arXiv: https://arxiv.org/abs/2608.21719v1 AI inference clusters are increasingly constrained by instantaneous power, not just energy: grid operators condition new capacity on demand response, imposing time-varying power caps. Existing LLM serving systems optimize a static energy objective or shed fixed priority tiers under load; either way, goodput collapses when the power envelope moves. An LLM pipeline is not a uniform load: compute-bound prefill loses throughput almost linearly with GPU frequency, memory-bound answer decode sustains it down to $0.57\times$ nominal, and reasoning's thinking phase couples KV-cache capacity to scheduling -- so a cap should be steered to where each watt costs the least performance. PowerSlider does so with a new Flex SLO contract that turns bounded user slack into an optimization constraint, prefill--think--answer disaggregation exposing per-stage frequency and KV control, and a Karush--Kuhn--Tucker (KKT) online solver re-solving within 7.7 ms of every cap change, backed by a consolidated fail-safe that power-gates drained instances when DVFS bottoms out on static power. On SGLang with production traces, \sys{} sustains 78.3\% online goodput at a 30\% cap reduction versus 47.6\% for the best of five baselines ($1.64\times$), holds latency-critical tails within $1.3\times$ of nominal (baselines: $2.3$--$6\times$, up to $12\times$), and delivers 92\% mean goodput through a replayed CAISO grid-emergency day bottoming at $0.41\times$ (54\% at the trough; every baseline below 7\%). --- ### LëtzCross: A Cross-Lingual Page-Level Benchmark for Multimodal Retrieval over Luxembourgish Documents Authors: Omar El Bachyr, Fred Philippy, Laura Maria Bernardy, Saad Ezzini, Jacques Klein, Tegawende Bissyande Published: 2026-08-22 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.21714v1 Recent page-image retrievers such as ColPali have improved retrieval over visually rich documents, yet little is known about how they behave in cross-lingual, low-resource settings. We introduce LëtzCross, a benchmark for cross-lingual page-level retrieval over Luxembourgish PDF documents, with document pages indexed as images and queries provided in English, French, German, and Luxembourgish. The benchmark combines text-focused QA pairs with visually grounded QA pairs, covering both textual and visual retrieval needs in PDF-based RAG. We use LëtzCross to compare OCR-based text-only retrievers with ColPali-style page-image retrievers and find that the latter perform better across query languages in this system-level comparison. We also examine single-language and multilingual fine-tuning. Fine-tuning transfers across query languages, with French yielding the highest mean performance on Luxembourgish queries among the single-language settings. In the multilingual setting, including Luxembourgish gives the strongest results and substantially improves retrieval for Luxembourgish queries. --- ### The Plan, Not the Decoder: Diagnosing and Repairing Compositional Failure in Reasoning-Augmented Text-to-Image Generation Authors: Ashritha Gonuguntla Published: 2026-08-22 Categories: cs.CV, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2608.21713v1 Reasoning-augmented text-to-image models such as GoT-R1 emit an explicit textual plan - object names, attributes, and bounding boxes - before generating image tokens. When such a model fails a compositional prompt, is the plan wrong, or is the plan right and the decoder unfaithful? Because the plan is machine-readable it can be edited before decoding, which makes the two separable. We first validate the ruler. Swapping the two bounding boxes inside the model's own chain demonstrably flips the generated layout: detector-based accuracy falls 0.75 -> 0.48 (p<1e-3), while a widely used VQA-based spatial metric rises. A five-rater human study agrees with the detector on 81% of items and with the VQA judge on 57%. All spatial results therefore use geometric scoring. Under sound measurement the decoder is a faithful executor: 94% of generated layouts realize the planned relation, and object-box binding survives reordering of the plan's object segments. The planner is the bottleneck. It writes wrong relations for phrasing-dependent reasons - 98% accuracy on "left" against 54% on "right" for semantically identical layouts, a raster-order bias we isolate with a mention-order control - and cluttered geometry that the decoder faithfully reproduces. Editing the plan therefore fixes the image without retraining: symbolic verification with resampling gives +5.0 points (p<1e-3), minimal in-place repair +6.0 (p=.02), rewriting only box geometry +10.7 (p<1e-4), and replacing the plan outright +13.3 (p=1e-4). Gains are indifferent to the plan's prose style and to its likelihood under the planner, but not to its geometry. Modular planner-decoder designs are therefore viable, provided the plan is internally consistent: box-text contradictions induce object duplication and identity fusion. We release the plan-fidelity evaluation protocol, all plans, and 12k generated images. --- ### ATHENA: Knowledge-guided agentic neural architecture search for AutoFormer-based electronic health record modeling Authors: Deyi Li, Qi Xu, Lingyao Li, Tiansheng Wang, Muxuan Liang, Mei Liu Published: 2026-08-22 Categories: cs.AI, cs.MA arXiv: https://arxiv.org/abs/2608.21712v2 Transformer-based models are widely used for clinical prediction from electronic health records (EHRs), yet their architectures require manual tuning, and the optimal configuration may vary across tasks and hospitals. Neural architecture search (NAS) automates architecture design, but conventional methods are computationally costly for Transformer-based EHR models. Recent large language model (LLM)-guided NAS methods reduce manual search design but conduct each search independently, without reusing architecture knowledge across hospitals. In this study, we propose ATHENA (Agentic Transfer across Hospitals for EHR Neural Architecture Search), a knowledge-guided agentic NAS framework for Transformer-based EHR modeling. ATHENA uses a weight-sharing supernet that is pretrained once per hospital, allowing candidate architectures to be instantiated as inherited subnetworks and evaluated through fine-tuning rather than independent pretraining. It incorporates a two-layer cross-hospital architecture prior. The first layer retrieves high-performing architecture examples from source sites based on task descriptors, while the second estimates the effects of architectural components using SHapley Additive exPlanations (SHAP)-based meta-regression. These priors guide a multi-agent LLM search using validation feedback from the target hospital. Across six clinical prediction tasks evaluated at one held-out OneFlorida+ site and one external MIMIC-IV site, ATHENA significantly outperforms all four baselines in 9 of 12 site-task evaluations under a strict equal-compute comparison. Using a common pretrained AutoFormer supernet for candidate evaluation, ATHENA ranks first in 9 of 12 evaluations at a search budget of 30. It also shows more consistent architecture selection across repeated searches. ATHENA provides a practical approach for reducing manual architecture tuning in Transformer-based EHR modeling. --- ### StereoDiffuer: Diffusion-based Progressive Geometry Modeling with Saliency Attention Perception for Stereo Matching Authors: Bohan Li Published: 2026-08-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.21710v1 With the advance of deep neural networks, the quality of disparity maps obtained through stereo matching has steadily improved. However, existing stereo matching methods still struggle to preserve fine-grained geometric details, resulting in blurred edges and over-smoothed predictions in challenging regions. To address these limitations, we propose StereoDiffuer, an iterative diffusion-based stereo matching framework that explicitly models geometric details and progressively refines disparity estimates. The framework incorporates a Saliency Attention Perception (SAP) module to extract salient geometric cues, including object boundaries, thin structures, and sharp edges. Confidence-guided SAP features are combined with the initial disparity estimate to condition an iterative denoising diffusion process, which corrects residual disparity errors and restores geometric details suppressed during cost-volume regularization and upsampling. Experimental results on the Scene Flow and KITTI benchmarks demonstrate the effectiveness of the proposed framework and its competitive performance relative to the compared stereo matching methods. --- ### Posterior Information Dynamics of Diffusion Models for Linear Inverse Problems Authors: Xiangming Meng Published: 2026-08-22 Categories: cs.LG, cs.IT, eess.SP arXiv: https://arxiv.org/abs/2608.21709v1 Diffusion models are widely used as priors for linear inverse problems, yet endpoint quality does not reveal when measurement information enters reverse denoising or how it is allocated across signal directions. We study this process through the smoothed likelihood force, the difference between exact posterior and prior scores at each noise level. For a fixed measurement, its expected squared norm gives both posterior--prior relative-entropy dissipation and reverse-path relative-entropy growth. Averaging over measurements yields an information--minimum mean-square error (I-MMSE) identity linking information gain to denoising-error reduction. Under finite second moments, the force energy and its ratio to prior-score energy decay quadratically in the noising kernel's signal coefficient at high noise. Solvable models show that conditioning removes class separation already explained by the measurement, reduces a uniform index entropy over \(n\) empirical samples from \(\log n\) to \(H(I\mid r)\), and makes assimilation depend on operator--prior alignment even for identical singular values. Experiments in models with tractable posteriors evaluate these predictions. In a separate illustration with a frozen FFHQ model, masks sharing the same spectrum yield different prior-normalized null-space trajectory statistics. --- ### From Association to Causation: Improving Retrieval Precision of Retrieval-Augmented Generation via Causal Relations and an Attention Mechanism Authors: Jing Liu, Yongxing Qi, Muchen Jiang, Chengnan Hu, Qingqing Peng, Haoming Wang, Yuqing Wang, Yang Yu, Xu Zhang, Ting Wu Published: 2026-08-22 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2608.21702v1 Retrieval-Augmented Generation (RAG) grounds LLM generation on retrieved documents, but the standard terminal retrieval stage--dense-vector similarity, optionally followed by reranking--often returns documents that share keywords with the query without containing the needed information, a failure mode that grows with the knowledge base. We trace it to a conceptual gap: similarity captures only associational relations, whereas the documents that matter are linked to the query causally. We model the terminal retrieval stage with a causal graph grounded in Reichenbach's common cause principle: the keywords shared by the query and a retrieved document form a latent common cause A, and the document's residual keywords form a latent set B linking the document to the ideal output. Since a retrieved document is a collider (A -> d <- B), retrieval itself opens an associational path between the query and B, which licenses a training-free, attention-style re-scoring rule: the cosine similarity between the query embedding and the weighted centroid embedding of B. Unlike causality-enhanced RAG variants that model causal relations inside the knowledge content, our graph models the causal structure of the retrieval process itself. On a real 471-document enterprise knowledge base, the method promotes a relevant guideline from rank 6 to the top 3; on a controlled diagnostic corpus reproducing the keyword-stuffing regime, it improves the mean target rank from 2.88 to 1.25, while a trained cross-encoder reranker barely helps (2.63). Conversely, on three BEIR benchmarks the score underperforms the similarity baseline, delineating the applicability boundary: the method guards the keyword-stuffing regime of growing proprietary knowledge bases and complements neural rerankers; a corpus-level calibration gate selects the correct regime with >= 95% reliability. A fully local testbed demonstrates deployability. --- ### Scalable quantum simulation of continuous-time generative models via tensor networks Authors: Nathan X. Kodama, L. Andrew Wray, Sam Cochran, Chad Rigetti, Shravan Veerapaneni, Michael J. Keiser Published: 2026-08-22 Categories: quant-ph, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2608.21700v1 Continuous-time flow and diffusion models are widely used across many application domains, from large-scale deployment in computer vision and protein folding to emerging adoption for modeling language, time series, and quantum states. After training, inferring statistical properties from continuous-time models is costly. Wavefunction flows target this cost by recasting learned transport as unitary evolution, whose final Born distribution approximates the target distribution. This prepares a coherent amplitude encoding (a qsample) that can be post-processed by quantum algorithms offering a quadratic advantage over Monte Carlo sampling. We present the first numerical study of these flows, in which we represent time-dependent potentials and states as tensor networks. At spatial dimension $d=8$, storage falls by $\sim 10^7\times$ relative to the dense grid of $N^d$ points, and evolution wall-clock time falls by $\gtrsim 10^3\times$ against a baseline extrapolated from the measured $d\le 5$ scaling. We validate our pipeline by reproducing the $O(1/\sqrt{p_{\rm rare}})$ scaling of rare-event sampling. --- ### Emotion Intensity Matters: Generating Realistic Expressions in Virtual Humans with CVAEs Authors: Vitor Miguel Xavier Peres, Lara Volpato, Gabriel Ferri Scnheider, Soraia Raupp Musse Published: 2026-08-22 Categories: cs.CV, cs.GR arXiv: https://arxiv.org/abs/2608.21697v1 Generating expressive facial behavior in virtual humans (VHs) remains a central challenge in affective computing and character animation. This paper presents a novel approach based on Conditional Variational Autoencoders (CVAEs), trained on real human facial expression data, to synthesize controllable emotional expressions at varying intensities. Using a dataset comprising six basic emotions represented at two intensity levels (low and high), we train a CVAE model to generate synthetic facial expression data while preserving semantic consistency with real human expressions. Despite the limited amount of training data (only 7,680 facial expression samples), the proposed approach learns meaningful latent representations and generates coherent emotional variations. Our method enables control over emotional intensity, making it suitable for animating virtual characters without requiring actor performances or manual artistic intervention. Our research aimed to evaluate whether the method (CVAE) preserves the characteristics associated with the different intensity levels present in the dataset. Results show that the proposed model preserves key expressive characteristics across intensity levels while supporting generalization across emotional intensity levels, contributing to the creation of emotionally expressive virtual characters from relatively small datasets. --- ### Benchmarking Composable Compression Techniques in Mixture-of-Experts LLMs Authors: Afsara Benazir, Chen Chen, Rongxiao Qu, Jiabo Huang, Jingtao Li, Lingjuan Lyu Published: 2026-08-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.21693v1 Mixture-of-Experts (MoE) LLMs scale model capacity efficiently through sparse activation, but their large expert parameter footprint, routing imbalance, and long-context KV-cache growth make deployment difficult on commodity hardware. Practical deployment often requires stacking multiple compression techniques: expert pruning removes redundant experts, weight quantization lowers model memory footprint, and KV-cache compression reduces long-context memory pressure. However, these techniques are typically evaluated in isolation, leaving open how they interact when applied together in realistic deployment pipelines. In this work, we present MoEXBench, a systematic benchmark for evaluating composable MoE compression as an end-to-end deployment workflow. MoEXBench studies 10 MoE models ranging from 30B to 235B total parameters across standard-attention, hybrid linear-attention, and sliding window attention architectures. It evaluates 20%-50% expert pruning rates, 1 to 16 bit weight-quantization schemes, and multiple KV-cache precision settings, applied both individually and in combination. MoEXBench introduces an eight-module evaluation suite that jointly measures composable-compression quality, workload and architecture robustness, pruning/quantization/KV cache sensitivity, and deployment efficiency on commodity hardware. Our results reveal non-trivial interactions among compression methods: composable compression cannot be predicted from standalone techniques, compression rate alone does not reliably predict quality loss or runtime gain, expert pruning is the dominant degradation source, and average quality can hide workload and architecture-specific failures. By releasing normalized module scores, compressed artifacts, and reproducible scripts, MoEXBench enables practical accuracy-memory-latency comparison across MoE families and hardware backends. --- ### Staged Depth Training: A Representation Curriculum for PINNs Authors: Kejia Zhang, Youran Sun, Haizhao Yang Published: 2026-08-21 Categories: cs.LG, physics.comp-ph arXiv: https://arxiv.org/abs/2609.30299v1 Representation quality is a central determinant of PINNs' performance, yet standard training leaves representations to emerge implicitly while fitting the final solution. We introduce \textbf{representation curriculum}, an ordered process in which representations are explicitly learned, transferred independently of their predictors, and progressively refined. We realize it with Staged Depth Training (SDT), which trains a shallow prefix under a temporary physics-informed head, discards the head, and freezes the learned prefix while adding depth, without equation-specific encodings or changes to the final architecture. Across the 20 default forward problems in PINNacle with three backbones, SDT improves 40 of 59 equal-budget problem--backbone cells by at least 5\% and remains within that band in the rest, with a 32.8\% geometric-mean error reduction on a PirateNet-style backbone. Mechanistic ablations suggest that the gain is not explained by optimizer restarts or shallow warm-starting alone. Representation visualizations and hyperparameter-basin analyses provide diagnostic evidence on representation geometry and local sensitivity to shared hyperparameters. On Poisson--Boltzmann 2D, SDT also more than doubles the fitted depth-scaling exponent for both backbones. These results support representation curriculum as a promising training strategy for improving PINNs while preserving the deployed architecture and inference cost. --- ### Spread and Scale: What Determines Whether Test-Time Budget Allocation Pays Authors: Jinhyung Bae Published: 2026-08-21 Categories: cs.LG, cs.AI, math.OC arXiv: https://arxiv.org/abs/2609.27917v1 Neural combinatorial optimization solvers generate many candidate solutions per instance and report the best one found, using the same sample budget for every instance regardless of difficulty. A companion study showed that reallocating a fixed budget toward harder instances can improve solution quality, but that the standard way of measuring this improvement is biased: deciding an allocation and evaluating it on the same data can manufacture an apparent gain even when none exists. This left open what property of a workload determines whether reallocation is worth doing, and whether a policy that spends part of the budget to decide how to allocate the rest still pays once that cost is counted. This paper answers both questions through pre-registered confirmatory experiments -- analysis and verdict criteria fixed before data collection -- across three independently trained solvers and two ways of constructing harder workloads on the traveling salesman problem. Within the workloads we study, the deciding property is how varied the instances within a workload are in difficulty, not how difficult the workload is on average: a uniformly easy or uniformly hard workload offers little room for reallocation, while a mixed workload offers substantial room. A budget-aware policy that pays for its own information about instance difficulty recovers most, though not all, of the improvement available when that information is assumed free. Every experiment was independently recomputed from its written specification, and every correction to an earlier version -- including two that weakened the paper's own claims -- is reported with the direction it moved the conclusion. The paper offers a specific empirical answer and a template for verifying that answer is not an artifact of how it was measured. --- ### Reliable Fusion of Conflicting Experts Authors: Pranuthi Tenali, Sahil Sidheekh, Saurabh Mathur, Vijayalakshmi Saravanan, Erik Blasch, Kristian Kersting, Sriraam Natarajan Published: 2026-08-21 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.27913v1 We study the problem of aggregating opinions from multiple black-box experts in noisy, conflict-prone settings where expert reliability varies across inputs. Static aggregation methods, such as majority voting, fail to capture this variability and often yield unreliable outcomes under disagreement. We propose a tractable, probabilistic-circuit-based fusion framework that dynamically combines expert responses using context-specific credibility estimates, enabling principled and reliable reasoning. The framework is agnostic to the underlying experts and does not require access to their internal representations or any retraining. We empirically validate our approach on multiple-choice question answering tasks using multiple LLMs as experts, comparing against individual models and static ensemble baselines. Our method consistently improves predictive performance and produces more reliable decisions under conflict, highlighting the effectiveness of context-aware credibility modeling for robust multi-expert fusion. --- ### Global tree forecasters collapse at the hierarchical aggregate: a five-panel failure characterization Authors: Md Rezwanul Islam, Wael Mohammed Published: 2026-08-21 Categories: cs.LG, econ.EM, stat.AP arXiv: https://arxiv.org/abs/2609.27912v1 Global forecasting models pool many series and learn one shared function. Gradient-boosted trees are their most common form. We measure a failure of this design that has not, to our knowledge, been documented. Train a global tree on the individual series of a hierarchy, then ask it for the hierarchical aggregate. The aggregate sits far outside the model's training range, and the forecast collapses. The model under-predicts the total by 30-50x in our production deployment, and by up to 496x in a public M5 reconstruction. The mechanism is known: beyond its training range, a tree predicts a constant. It surfaces at the aggregate because the total dwarfs every training series. The cure is not new. Per-series scaling, the preprocessing step that Montero-Manso and Hyndman (2021) recommend, prevents the collapse. So do a weighted aggregate-level training row and seasonal differencing. Our contribution is the characterization. The collapse reproduces on five panels: a production business-to-business marketplace, a synthetic hierarchy, M5, Australian Tourism, and a public business-buyer panel. It holds on three tree libraries, is invariant across training seeds, and is statistically significant. Its onset is immediate and tracks a simple support bound: a scale gap of only 1.15x already costs a third of the total. No standard configuration change prevents it: pooling every hierarchy level into training fails at scale, and the one knob that fits linear models in the leaves softens it without curing it. Rolling the forecasts forward recursively separates the cures: the aggregate-row cure re-collapses, per-series scaling degrades but stays low, and only seasonal differencing keeps its one-step accuracy unchanged. We close with a three-step procedure for diagnosing and preventing the failure in deployed systems. --- ### Spatiality-Frequency Domain Video Forgery Detection System Based on ResNet-LSTM-CBAM and DCT Hybrid Network Authors: Zihao Liao, Sheng Hong, Yu Chen Published: 2026-08-21 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.27904v1 As information technology advances, digital content has become widely adopted across diverse fields such as news broadcasting, entertainment, commerce, and forensic investiga?tion. However, the availability of sophisticated multimedia editing tools has significantly increased the risk of video and image forgery, raising serious concerns about content authenticity at both societal and individual levels.To address the growing need for robust and accurate detection methods, this study proposes a novel video forgery detection model that integrates both spatial and frequency-domain features. The model is built on a ResNet-LSTM framework enhanced by a Convolutional Block Attention Module (CBAM) for spatial feature extraction, and further incorporates Discrete Cosine Transform (DCT) to capture frequency domain information. Comprehensive experiments were conducted on several mainstream benchmark datasets, encompassing a wide range of forgery scenarios. The results demonstrate that the proposed model achieves superior performance in distinguishing between authentic and manipulated videos. Additional ablation and comparative studies confirm the contribution of each component in the architecture, offering deeper insight into the models capacity. Overall, the findings support the proposed approach as a promising solution for enhancing the reliability of video authenticity analysis under complex conditions. --- ### A Resilience Recovery Method for Complex Traffic Network Security Based on Trend Forecasting Authors: Sheng Hong, Tianyu Yue, Yang You, Zhengnan Lv, Xu Tang, Jing Hu, Hongwei Yin Published: 2026-08-21 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.27903v1 Due to the rapid development of information technology, a huge and complex traffic network has been established across various sectors, including aviation, aerospace, vehicles, ships, electric power, and industry. However, because of the complexity and diversity of its structure, the complex traffic network is vulnerable to being attacked and faces serious security challenges. Therefore, this paper innovatively proposes a traffic network resilience recovery method based on resilience trend forecasting. In this paper, the risk value is introduced into the analysis of the network fault propagation process, and the Susceptible, Infectious, Recovered, Dead-Risk (SIRD-R) fault propagation model is established. The resilience model of traffic network, which encompasses real-time resilience and overall resilience, is constructed through the integration of network resilience bearing capacity and resilience recovery capacity. Ten, the resilience of complex traffic networks is forecasted by using long short-term memory networks, and the resilience recovery strategy of complex traffic networks based on forecasting is proposed. Finally, the effectiveness and scalability of the proposed method are demonstrated through experimental analysis conducted on a diverse range of complex traffic networks, affirming its applicability in real-world scenarios --- ### Image Denoising Using Lower Semi-Frames Authors: Hemalatha M, P. Sam Johnson Published: 2026-08-21 Categories: eess.IV, cs.CV arXiv: https://arxiv.org/abs/2609.27893v1 A blind image denoising framework based on an infinite directional lower semi-frame (DLSF) is proposed for additive white Gaussian noise. The model employs scale-dependent directional analysis with resolvent regularization of the unbounded semi-frame operator. Noise variance is estimated directly in the DLSF domain by modeling the joint covariance of four directional difference channels and applying covariance whitening to obtain a chi-square statistic. A lower-tail moment estimator provides blind noise estimation without median absolute deviation. The estimated noise level is incorporated into channel-wise Wiener-type shrinkage and canonical-dual synthesis, followed by a data-consistent iterative reconstruction with automatic stopping. Experiments on three standard grayscale images at noise levels 15--30 yield a mean relative noise-estimation error of 3.28\%, with average improvements of 7.45 dB in PSNR and 0.367 in SSIM. At 30/255 noise, the estimation error decreases to 1.73\%, with a mean PSNR gain of 8.31 dB. Results demonstrate effective noise suppression and structural preservation, with the strongest performance on smooth and edge-dominated images. --- ### Schrödinger's Code Repository: Have LLMs Learned SWE-bench or Memorized It? Authors: Silin Chen, Yufei Yang, Xiaodong Gu, Yuling Shi, Chengcheng Wan, Haibing Guan Published: 2026-08-21 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2609.27891v1 Repository-level coding benchmarks have become the standard for evaluating coding agents, yet they inherently suffer from data leakage because they are built upon popular open-source repositories repeatedly used for training. Consequently, strong performance may reflect memorization of canonical repository cues rather than robust repository reasoning. We propose SchrodingerRepo (Schrödinger's Repository), an evaluation framework for testing coding agents under dynamically instantiated repository representations. Instead of repeatedly using a static representation of the test repository, SchrodingerRepo treats the test repository as an evaluation-time latent variable that is dynamically instantiated only when the agent enters the evaluation environment. The instantiated repository preserves the original executable behavior while eroding familiar cues such as naming conventions, file layouts, and implementation patterns through four transformation levels: problem statement reconstruction, namespace remapping, intra-file layout reordering, and functionality-preserving code rewriting. We evaluate popular LLMs on SWE-bench Verified and SWE-QA. Results show that removing familiar repository cues consistently degrades agent performance and substantially increases interaction costs across models. Further analysis reveals that the additional cost is primarily caused by increased difficulty in repository exploration and localization. These findings suggest that current coding agents may partially rely on memorized repository-side cues, highlighting the need for evaluation under dynamically instantiated repository representations. ---