# Research Papers — 2026-08 · bundle page 210 of 240 > Full-text markdown of papers #10451–10500 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/210 [← Previous bundle](https://aipod.dev/papers/archive/2026-08/page/209.md) [Next bundle →](https://aipod.dev/papers/archive/2026-08/page/211.md) ### Amortized Interventional Forecasting for Multivariate CIR Processes Authors: Andreas Sauter, Sumit Sourabh, Drona Kandhai, Erman Acar Published: 2026-08-04 Categories: cs.LG, cs.CE arXiv: https://arxiv.org/abs/2608.03715v1 Mean-reverting dynamics are pervasive in finance, and the Cox--Ingersoll--Ross (CIR) process is a standard model for the time series they produce, from short rates to credit default swap (CDS) spreads. Yet CIR models capture only \emph{correlated} co-movement, not \emph{causal} influence between series, so they cannot answer the system's response when one series is externally shocked, which observational conditionals confound with historical co-movement. We make two contributions. First, an amortized model for distributional causal effect estimation that frames trajectories as time-stamped observations and predicts the calibrated multi-horizon shock response without retraining per scenario. Second, a causal multivariate CIR data-generating process that supplies the paired observational and interventional ground truth that real markets cannot. We instantiate and calibrate the framework on CDS spreads as a testbed. CIR-ACTIVA's validity is established on synthetic ground truth, independent of how well the simulator matches reality, while practical grounding is assessed by backtesting the generated traces against real CDS data. Against observational and amortized causal-inference baselines, CIR-ACTIVA leads on both causal selectivity in the joint distribution and horizon-resolved calibration, retaining its selectivity once the interventional law varies over the horizon, with gains concentrating at short horizons. This opens up a class of what-if queries on coupled spread systems, CDS stress testing among them, that observational forecasters cannot answer. --- ### Attention is Case-Sensitive Authors: Maximilian Dillitzer, Tin Stribor Sohn, Jason J. Corso, Michael Auerbach Published: 2026-08-04 Categories: cs.CV, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2608.03711v1 In human visual perception, uppercase lettering serves as a natural salience cue that captures attention within lowercase text. In this paper, we present a systematic empirical characterization study revealing that Large Language Models (LLMs) exhibit an analogous property: letter casing modulates internal attention allocation. Through analysis across 13 models, nine LLMs and four Vision-Language Models (VLMs), with diverse tokenization schemes, we show that formatting target information in alternating or uppercase against a lowercase context concentrates attention on those textual spans. In text this effect is universal, holding across every evaluated non-reasoning model. We frame it as a previously under-explored latent property of pretrained transformers rather than a prescriptive method. Our investigation reveals a central attention-performance divergence: while this "casing effect" robustly shifts attention, its impact on downstream accuracy is non-trivial, increased concentration does not inherently improve task accuracy and, in high-entropy contexts like alternating case, can degrade it. We further identify a boundary condition: the deliberative "thinking" phase in reasoning models acts as a semantic buffer that mitigates typographic sensitivity in text. Extending the study to VLMs, we find the effect transfers partially: the same prompt-side casing reorganizes cross-modal attention along two coupled axes, predominantly a macroscopic disengagement from the image toward the text prompt, and secondarily a concentration of the residual visual attention on the target region. By isolating casing as a zero-shot mechanism for attention steering that requires no model access or fine-tuning, we provide a new foundational understanding of how pretraining internalizes typographic emphasis. --- ### Predicting Deep Neural Network Training Outcomes from Early Training Telemetry Authors: Ranjita Naik, Anh D. Nguyen, Pankaj Kumar Singh Published: 2026-08-04 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.03709v1 Large hyperparameter sweeps for deep neural networks spend substantial compute on configurations that are effectively doomed from the first few epochs. We study whether a single training run's own early telemetry - per-epoch loss, training accuracy, gradient signal-to-noise ratio, weight-norm growth, and an activation-saturation snapshot - together with its sampled hyperparameters, can predict that run's eventual outcome without reference to other runs. We evaluate three prediction tasks: final test accuracy, relative performance within a domain, and training-dynamics failure, including numerical divergence. Across 23,788 training runs spanning six architecture/dataset combinations, gradient-boosted trees using only the first five epochs of telemetry achieve R^2 = 0.92-0.99 for final-accuracy regression and ROC-AUC = 0.983-0.998 for relative classification on a permanently held-out set of hyperparameter configurations. Useful prediction is already available after a single epoch. A paired ablation shows that gradient- and weight-level telemetry provides a statistically consistent improvement over loss and accuracy curves alone, although the practical gain varies by domain. Transfer is strong between similar architectures, while cross-dataset transfer is limited mainly by differences in accuracy scale rather than loss of the underlying relationship. These results suggest that early-training telemetry can provide a practical decision-support signal for compute allocation while motivating human oversight for any automated intervention. --- ### MultiCompose: Multi-Concept Personalized Composition with Per-Subject Attribute Binding Authors: Ruirui Zhang, Zhengkai Zhao, Pan Gao Published: 2026-08-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.03708v1 Text-to-image diffusion models enable personalization of specific visual concepts from a small number of reference images. However, generating a single image that contains multiple personalized subjects, each bound to user-specified attributes such as clothing, accessories, and held objects, remains largely unaddressed. Without explicit spatial constraints, concurrently activated concept checkpoints produce overlapping cross-attention responses, causing per-subject identity degradation and attribute misalignment. Moreover, no established benchmark jointly evaluates these two failure modes in the personalized multi-subject setting. We present MultiCompose, a composition framework that decouples per-concept personalization from multi-subject inference. A semantic preservation regularization maintains attribute binding capacity during fine-tuning, while a two-phase inference procedure automatically establishes subject layout and composes per-concept predictions through spatially exclusive masks. We further introduce MSP-Bench, a benchmark that jointly evaluates identity fidelity (ID), attribute binding accuracy (BIND), and attribute misalignment (MIS) through a dual-pathway protocol. Experiments show that MultiCompose outperforms existing methods on both conventional metrics and MSP-Bench, confirming the benchmark's ability to reveal failure modes that conventional metrics overlook. Code is available at https://github.com/I2-Multimedia-Lab/MultiCompose --- ### To Describe or Construct Statistical Learning Models Using the Category-theoretical Language Authors: Congwei Song Published: 2026-08-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.03706v1 Statistical learning is a fascinating field that has long been the mainstream of machine learning/artificial intelligence. A large number of results have been produced which can be widely applied to real-world problems. It also leads to many research topics and also stimulates new research. This report summarizes some classical statistical learning models and well-known algorithms, especially for amateurs, and provides a category-theoretic perspective on understanding statistical learning models. The aim is to attract researchers from other fields, including basic mathematics, to participate in the research related to statistical learning. --- ### Less Traffic, Better Outcomes: Competition-Aware Request Dispatch in Real-Time Ad Exchanges Authors: Jonaid Shianifar, Blaz Mramor, Fangda Zou, Matthieu C. Martin, Xingsheng Guo, Zhihua Zhu, Rong Zhou, Bichen Shi Published: 2026-08-04 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2608.03705v1 Real-time bidding (RTB) ad exchanges typically forward nearly all incoming requests to demand-side platforms (DSPs), even though only a small fraction receive bids. This over-distribution weakens auction outcomes: DSPs throttle participation under compute and budget constraints, reducing the effective use of limited bidding capacity. We present a competition-aware request dispatch framework that uses distributional bid prediction and probabilistic forwarding to decide whether each request should be sent to each DSP. The system adapts per-DSP thresholds over time through lightweight policy optimization to track non-stationary market conditions. We evaluate the framework through four sequential online experiments on a production platform serving over 20 billion daily requests. A full multi-DSP deployment reduces DSP request volume under the policy by 34.2% while increasing net revenue by 4.6% (p<0.001) in a recent 14-day window after an initial DSP adaptation period. Further analysis highlights strong heterogeneity across traffic segments and reveals that aggregate metrics can be misleading. Segment-level and per-DSP analyses suggest that the policy surfaces comparative advantages among DSPs, improving monetized outcomes without increasing overall request volume. --- ### LiLa-WAM: Lightweight Latent Reasoning World-Action Model for Robotic Manipulation Authors: Fan Yang, Yuting Su, Xiaobo Wang, Yuncheng You, Fugui Fan, Yuting Wu, Minghui Wu, Chenxu Zhao, JiaHong Ning, Peiguang Jing Published: 2026-08-04 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2608.03701v1 World-action modeling has emerged as a promising paradigm for robotic control, as it empowers models to go beyond reacting to observations and anticipate how a scene will evolve. However, existing WAMs often incur substantial computational overhead. Pixel-space methods often allocate substantial capacity to visual details that may not be directly relevant to control, while some latent-space methods require multi-stage training to construct the reasoning space. The resulting training cost can make such methods difficult to train under modest computational budgets. In this work, we propose LiLa-WAM, a lightweight world-action model that reasons about the future in a compact latent space and can be trained end-to-end on a single 24GB GPU. Its core design is a compact latent reasoning space jointly shaped by future-state prediction and action generation, which keeps the model lightweight while remaining well aligned with control. For task specification, we further propose the Visual Transition Token(VTT), a language-free task representation that encodes each task as a direction in visual feature space. Experiments on RoboTwin~2.0, LIBERO, and real-robot tasks demonstrate LiLa-WAM's effectiveness, achieving 90.48\% success across 50 RoboTwin tasks with single-GPU training. --- ### When Agents Learn to Be You: Benchmarking Privacy Leakage, Impersonation Risk, and Defenses in Persona Skills Authors: Yongli Xiang, Zhifang Zhang, Bojun Yang, Ziming Hong, Lei Feng, Miao Xu, Tongliang Liu Published: 2026-08-04 Categories: cs.CR, cs.CL, cs.CY arXiv: https://arxiv.org/abs/2608.03700v1 Persona skills distill personal interaction histories into portable and executable artifacts for downstream agents. While enabling flexible personalization, this process concentrates fragmented personal signals, amplifies their impact through reuse, and challenges defenses designed for individual records or retrieval-based memory. To systematically investigate the safety of the persona-skill pipeline, we introduce AntiSkillBench, an end-to-end benchmark for evaluating risks and defenses across the persona-skill pipeline. It comprises: (i) a dataset of 7,500 persona-grounded dialogue traces, constructed from 50 behaviorally rich profiles spanning diverse task scenarios; (ii) an evaluation suite that measures skill-level privacy leakage and agent-level attribute disclosure and behavioral impersonation across three skill-distillation strategies; and (iii) a defense evaluation covering four configurations across online and post-hoc interventions, including active risk suppression and passive provenance protection. Experiments across three frontier agents show that persona-skill risks persist across agent backbones and distillation protocols, extending from explicit attributes to communication styles and personality traits. Existing defenses exhibit limited and distillation-dependent effectiveness, failing to generalize across risk and distillation strategies. These results highlight AntiSkillBench as a challenging benchmark for developing privacy-preserving and authenticity-aware persona skills. --- ### TARL: Transaction-Aware Reliable Ledgers for Executable Memory Management in Long-Term Agents Authors: Han Xiao, Hongjun Xu, Xin Zhang, Yidong Chen, Xiaodong Shi Published: 2026-08-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.03699v5 Persistent memory helps long-term agents retain knowledge, yet a single update error can repeatedly distort future retrieval and reasoning. Most existing systems reduce memory updating to a binary Write/Hold decision, which cannot distinguish whether new information should be added, ignored, used to revise an outdated belief, rejected as unreliable, or deferred for verification. These choices may share the same binary label while producing fundamentally different memory states. We introduce TARL, a memory state update framework that maps each statement to one of five executable actions. TARL identifies the affected memory, resolves its temporal scope, compares source reliability, and updates accepted, pending, and rejected ledgers. It is further trained by comparing the memory states produced by alternative update operations, encouraging the model to select the operation that leads to the correct result. We also introduce TARL-Mem, a benchmark with fine-grained action labels and next-state targets. Across in-domain, cross-source, temporal, counterfactual, and sequential evaluations, TARL improves action prediction and state recovery, reduces memory pollution, preserves conflicting evidence, and limits cumulative corruption. --- ### TARL: Transaction-Aware Reliable Ledgers for Executable Memory Management in Long-Term Agents Authors: Han Xiao, Hongjun Xu, Xin Zhang, Yidong Chen, Xiaodong Shi Published: 2026-08-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.03699v1 Persistent memory helps long-term agents retain knowledge, yet a single update error can repeatedly distort future retrieval and reasoning. Most existing systems reduce memory updating to a binary Write/Hold decision, which cannot distinguish whether new information should be added, ignored, used to revise an outdated belief, rejected as unreliable, or deferred for verification. These choices may share the same binary label while producing fundamentally different memory states. We introduce TARL, a memory state update framework that maps each statement to one of five executable actions. TARL identifies the affected memory, resolves its temporal scope, compares source reliability, and updates accepted, pending, and rejected ledgers. It is further trained by comparing the memory states produced by alternative update operations, encouraging the model to select the operation that leads to the correct result. We also introduce TARL-Mem, a benchmark with fine-grained action labels and next-state targets. Across in-domain, cross-source, temporal, counterfactual, and sequential evaluations, TARL improves action prediction and state recovery, reduces memory pollution, preserves conflicting evidence, and limits cumulative corruption. The complete model implementation is provided in the supplementary material. --- ### Learning and Clustering on Temporal Graphs: Principles, Primitives, and Pooling Authors: Nelson Aloysio Reis de Almeida Passos, Emanuele Carlini, Salvatore Trani Published: 2026-08-04 Categories: cs.LG, cs.SI arXiv: https://arxiv.org/abs/2608.03696v1 This work focuses on the problem of learning on temporal graphs, with particular emphasis on the task of clustering: obtaining coarse-grained representations by aggregating information from nodes, edges, and temporal dynamics - a task related to pooling in machine learning on graphs, or community detection in network science. Although graph neural networks reach state-of-the-art performance across many downstream graph tasks, their advantage over established descriptive and inferential clustering algorithms is far less settled, especially under demands of efficiency and recovery accuracy. We frame this tension through three linked perspectives: principles, connecting graph learning and community detection through shared spectral foundations and detectability thresholds in stochastic block model regimes; primitives, making spectral clustering and multislice modularity optimization tractable through GPU-accelerated temporal backends; and pooling, viewing principled community detection as a theory-grounded coarse-graining operator for temporal graphs. Our results indicate that algorithmic methods remain the appropriate tool where attributes are absent or weak - scalability rather than accuracy being the binding obstacle - while neural models are most compelling when structural, temporal, and attribute signals align. By making temporal clustering scalable, GPU-accelerated primitives suggest a route toward theory-grounded pooling, while raising a central question: when does community-based coarse-graining preserve the dynamics needed for downstream learning tasks? --- ### Accelerating Dynamic Graph Clustering on GPU Architectures with cuGraph Authors: Nelson Aloysio Reis de Almeida Passos, Emanuele Carlini, Salvatore Trani Published: 2026-08-04 Categories: cs.DC, cs.LG arXiv: https://arxiv.org/abs/2608.03695v2 This work addresses community detection in temporal networks through GPU-accelerated extensions of spectral clustering and modularity-based algorithms originally designed for static graphs. Built on the NVIDIA RAPIDS ecosystem, the framework enables the characterization and tracking of communities in snapshot-based dynamic graphs, either by Leiden greedy optimization with multi-GPU support via Dask-based workload distribution, or eigendecomposition of a symmetric Bethe-Hessian operator. Our multislice modularity backend achieves up to roughly three orders of magnitude speedup over the CPU reference under an equal-work budget, depending on graph density and snapshot count, while preserving compatibility with existing graph analytics pipelines. We demonstrate its applicability on real-world and synthetic datasets, facilitating exploratory analysis of structural network properties over time. Such capabilities are relevant across several application domains, such as epidemic spreading, financial systems, cybersecurity, and trajectory and mobility analysis. We release our implementation as free and open-source software, including Python bindings through the NetworkX-Temporal library for ease of use and zero-code acceleration with existing codebases. --- ### Pattern over Pixels: Measuring Pattern Completion Bias in Multimodal Code Generation Authors: Khai-Nguyen Nguyen, Oscar Chaparro, Antonio Mastropaolo Published: 2026-08-04 Categories: cs.SE, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2608.03691v1 Multimodal large language models (MLLMs) are increasingly used to translate webpage screenshots into front-end code, but repeated UI patterns may sway them toward visually incorrect yet pattern-consistent outputs. In this work, we test how repeated webpage patterns hurt MLLM accuracy on an objective screenshot-to-code fill-in-the-blank task. We introduce the first benchmark for visual pattern-completion bias, where one localized element in a repeated UI pattern is perturbed and the model must recover the masked width or font-size value from the screenshot and HTML context. Starting from 30 webpages curated from the Design2Code dataset, we build 1,440 evaluated screenshots spanning structural card and text-style patterns under standard and noise-overlaid conditions. We evaluate five frontier MLLMs and find that all are strongly biased toward the repeated baseline. Mean bias rate reaches 69.78% on card-width perturbations and 80.22% on text font-size perturbations, while mean accuracy is only 21.17% and 7.89%, respectively. Codex-5.3 performs best but still drops from 68.61% accuracy on cards to 13.89% on text, while Flash-3.0 reaches 96.11% bias on text. Noise, subtler perturbations, and boundary positions further increase bias rate. Reasoning analysis further shows that greater reasoning effort correlates with lower bias, yet qualitative evidence reveals that models can identify the anomalous element and still override it with the pattern-consistent answer. Our results identify a concrete failure mode in multimodal code generation and show that its severity is strongly associated with visual saliency --- ### LAEF: A Lead-Agnostic ECG Foundation Model Towards Point-of-Care Diagnostics Authors: Edoardo Coppola, Stefano Fiorini, Pietro Liò, Mattia Savardi, Alberto Signoroni Published: 2026-08-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.03690v1 Point-of-care cardiac devices such as smartwatches and handheld ECG recorders typically capture 1--2 leads, yet existing ECG foundation models are architecturally constrained to fixed 12-lead inputs, degrading or failing under these reduced configurations. We introduce LAEF (Lead-Agnostic ECG Foundation), a 7M-parameter ECG foundation model that can natively process any lead subset without zero-padding or architectural modification. LAEF represents ECGs as variable-size spatiotemporal graphs with physiologically motivated intra- and inter-lead connectivity, processed by a Graph Attention Network that scales naturally with active lead count.Pre-trained on 9.2M 12-lead ECGs via masked node modelling with stochastic lead sampling, LAEF learns representations robust to lead configuration. Across 18 downstream datasets, LAEF is on par with specialized 12-lead baselines over 12$\times$ larger at full lead availability. Under direct point-of-care-oriented diagnostics (1--2 leads), it outperforms all zero-padded alternatives on 17 out of 18 datasets with with a single randomly sampled lead and on 14 out of 18 with 2 leads, with an average AUROC gain of +3.2 points. Representation analysis links this advantage to architectural lead-agnosticism, and a lead-importance study across 164 cardiovascular conditions shows population-level performance is stable across single standard input leads while still recovering established clinically lead-condition associations. --- ### LiveEvalBench: Toward Open-World Evaluation for Web Generation Authors: Yiyao Wang, Zhen Wen, Yinghao Tang, Yixiao Fu, Lin Yuan, Xiaolau Zhang, Jun Zhou, Wei Chen Published: 2026-08-04 Categories: cs.AI, cs.SE arXiv: https://arxiv.org/abs/2608.03689v1 Large language models are increasingly capable of synthesizing executable frontend projects, yet existing benchmarks still treat web generation as a static evaluation problem. We argue that frontend artifacts demand a different paradigm: they are interactive rather than static, admit diverse yet equally valid implementations, and evolve faster than rigid pipelines can accommodate. To address these gaps, we present LiveEvalBench, an automated framework that reformulates web-generation evaluation as an agentic, adaptive, and extensible process. LiveEvalBench instantiates evaluation as a collaborative review workflow, in which a Build Engineer, a Code Engineer, and a UI Tester collectively gather evidence across the full lifecycle of a frontend project, from deployment and code inspection to browser-based interaction. To handle implementation diversity, an adaptive protocol couples shared rubrics for cross-model comparability with implementation-grounded criteria tailored to each artifact. The framework further supports incremental integration of new evaluator roles and assessment dimensions without pipeline redesign. Experiments across diverse real-world web-generation scenarios show that LiveEvalBench aligns closely with human expert judgment and provides fine-grained insights into frontier models' web generation capabilities. Code is available at https://github.com/wyysteelhead/LiveEvalBench --- ### PhyAI: Real-Time Physical AI at the Edge, Scalable Rollouts in the Cloud Authors: Chenghua Wang, Daliang Xu, Dongqi Cai, Duojin Sun, Hao Zhang, Haoze Qian, Huaiyuan Zhang, Jinshuo Cui, Junbo Cui, Kezhao Zhao, Longxi Gao, Mengwei Xu et al. Published: 2026-08-04 Categories: cs.AI, cs.RO arXiv: https://arxiv.org/abs/2608.03682v3 Physical AI policies require inference throughout their lifecycle, including model evaluation, cloud reinforcement learning rollout, edge GPU serving, and onboard deployment. Although these settings share the same checkpoint and action semantics, they often rely on separate inference programs. To unify them, we build PhyAI, a Physical AI inference engine with a single runtime that keeps architecture-specific conditioning, solver, cache, and output logic in model adapters while sharing graph execution, kernels, memory management, and parallel services. The same codebase runs vision-language-action (VLA) models and world-action models (WAMs) on single or multiple GPUs across onboard, edge, and cloud deployments. We used the adapter interface to add MiniCPM-Robot on the day of its release. PhyAI achieves 1.40x-4.65x speedups over the official implementations of pi0, pi0.5, GR00T N1.7, and MiniCPM-Robot. On Cosmos3-Nano-Policy-DROID it reduces latency from 2.46 to 1.18 s on eight H20 GPUs (CFG=2, TP=4), a 2.08x speedup. Specialized runtimes remain faster in several configurations, so our goal is one runtime with competitive latency rather than the fastest result in every case. Detailed profiles reveal why different models need different execution policies: on a Hopper-series GPU at batch size one, the pi0.5 action expert accounts for 8.8% of FLOPs but 57.2% of latency; at batch size 32 its share drops to 13.5% and throughput reaches about 100 samples/s. Cosmos3 remains generation-dominated and gains only 14.3% throughput as batch size increases from 1 to 16. We further introduce the control-time Roofline, which distinguishes inference-bound from environment-bound control; the measured pi0.5 points on four LIBERO suites are environment-bound while Cosmos3 stays inference-bound. Code and benchmarks: https://github.com/mingti-org/phyai. --- ### PhyAI: Real-Time Physical AI at the Edge, Scalable Rollouts in the Cloud Authors: Chenghua Wang, Daliang Xu, Dongqi Cai, Duojin Sun, Hao Zhang, Haoze Qian, Huaiyuan Zhang, Jinshuo Cui, Junbo Cui, Kezhao Zhao, Longxi Gao, Mengwei Xu et al. Published: 2026-08-04 Categories: cs.AI, cs.RO arXiv: https://arxiv.org/abs/2608.03682v2 Physical AI policies require inference throughout their lifecycle, including model evaluation, cloud reinforcement learning rollout, edge GPU serving, and onboard deployment. Although these settings share the same checkpoint and action semantics, they often rely on separate inference programs. To unify them, we build PhyAI, a Physical AI inference engine with a single runtime that keeps architecture-specific conditioning, solver, cache, and output logic in model adapters while sharing graph execution, kernels, memory management, and parallel services. The same codebase runs vision-language-action (VLA) models and world-action models (WAMs) on single or multiple GPUs across onboard, edge, and cloud deployments. We used the adapter interface to add MiniCPM-Robot on the day of its release. PhyAI achieves 1.40x-4.65x speedups over the official implementations of pi0, pi0.5, GR00T N1.7, and MiniCPM-Robot. On Cosmos3-Nano-Policy-DROID it reduces latency from 2.46 to 1.18 s on eight H20 GPUs (CFG=2, TP=4), a 2.08x speedup. Specialized runtimes remain faster in several configurations, so our goal is one runtime with competitive latency rather than the fastest result in every case. Detailed profiles reveal why different models need different execution policies: on a Hopper-series GPU at batch size one, the pi0.5 action expert accounts for 8.8% of FLOPs but 57.2% of latency; at batch size 32 its share drops to 13.5% and throughput reaches about 100 samples/s. Cosmos3 remains generation-dominated and gains only 14.3% throughput as batch size increases from 1 to 16. We further introduce the control-time Roofline, which distinguishes inference-bound from environment-bound control; the measured pi0.5 points on four LIBERO suites are environment-bound while Cosmos3 stays inference-bound. Code and benchmarks: https://github.com/mingti-org/phyai. --- ### Keep the Needle, Prune the Haystack: Defect-Preserving Token Pruning for Efficient Zero-Shot Anomaly Detection Authors: Yanning Hou, Jingyuan Zhang, Xiaoyun Wang, Qixiang Ma, Sihang Zhou, Ke Xu Published: 2026-08-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.03681v3 Zero-shot visual anomaly detection has achieved remarkable progress, with recent vision-only approaches further improving performance while simplifying the inference pipeline. However, existing methods typically perform dense computation over all images and spatial tokens, despite the fact that normal samples dominate real-world scenarios and anomalies usually occupy only small regions. Token pruning offers a promising solution, but introduces an asymmetric pruning risk in anomaly detection: retaining normal tokens mainly incurs redundant computation, whereas removing anomalous tokens may eliminate the only evidence for detection and localization. This risk is particularly severe in early layers, where pruning provides the greatest computational benefit but anomaly semantics remain unreliable. We propose KeepAD, a defect-preserving token pruning framework that formulates token selection as high-recall, anomaly-aware routing. In shallow layers, KeepAD combines coverage-preserving selection over local $2\times2$ patch neighborhoods with deterministic anomaly rescue to reduce the risk of discarding subtle defects. In deeper layers, frozen normal and abnormal prototypes guide pruning under an image-adaptive token budget, aggressively removing low-risk normal tokens while preserving local anomaly evidence. Dense-to-sparse self-distillation further supervises early token routing without introducing additional inference overhead. Experiments on six industrial and seven medical zero-shot anomaly detection benchmarks show that KeepAD reduces the token retention ratio to below $20\%$, while limiting the average degradation in image-level and pixel-level AUROC to within $2.7$ percentage points. At the most aggressive operating point, KeepAD achieves a $7.9\times$ speedup over the strongest CLIP-based baseline. --- ### Keep the Needle, Prune the Haystack: Defect-Preserving Token Pruning for Efficient Zero-Shot Anomaly Detection Authors: Yanning Hou, Jingyuan Zhang, Xiaoyun Wang, Qixiang Ma, Sihang Zhou, Ke Xu Published: 2026-08-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.03681v2 Zero-shot visual anomaly detection has achieved remarkable progress, with recent vision-only approaches further improving performance while simplifying the inference pipeline. However, existing methods typically perform dense computation over all images and spatial tokens, despite the fact that normal samples dominate real-world scenarios and anomalies usually occupy only small regions. Token pruning offers a promising solution, but introduces an asymmetric pruning risk in anomaly detection: retaining normal tokens mainly incurs redundant computation, whereas removing anomalous tokens may eliminate the only evidence for detection and localization. This risk is particularly severe in early layers, where pruning provides the greatest computational benefit but anomaly semantics remain unreliable. We propose KeepAD, a defect-preserving token pruning framework that formulates token selection as high-recall, anomaly-aware routing. In shallow layers, KeepAD combines coverage-preserving selection over local $2\times2$ patch neighborhoods with deterministic anomaly rescue to reduce the risk of discarding subtle defects. In deeper layers, frozen normal and abnormal prototypes guide pruning under an image-adaptive token budget, aggressively removing low-risk normal tokens while preserving local anomaly evidence. Dense-to-sparse self-distillation further supervises early token routing without introducing additional inference overhead. Experiments on six industrial and seven medical zero-shot anomaly detection benchmarks show that KeepAD reduces the token retention ratio to below $20\%$, while limiting the average degradation in image-level and pixel-level AUROC to within $2.7$ percentage points. At the most aggressive operating point, KeepAD achieves a $7.9\times$ speedup over the strongest CLIP-based baseline. --- ### VetScore: Risk-Weighted Fact Verification for Veterinary Long-Form QA with Citations Authors: Ivan Kartáč, Jan Tovarys, Mateusz Lango, Ondřej Dušek Published: 2026-08-04 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.03675v1 Citation excerpts can be used to increase the reliability of generated outputs and their faithfulness to cited sources, which is especially important in high-stakes domains such as human and veterinary medicine. However, this does not guarantee that generated claims are faithful to the provided excerpts. We present VetScore, a multi-step evaluation method for veterinary long-form question answering, designed to assess how well are generated claims supported by the provided excerpts, weighing this information by each claim's harm potential. VetScore first segments the output and decomposes it into individual claims, then scores each claim with respect to its harm potential and evaluates its faithfulness to source excerpts, and finally calculates the overall risk-adjusted score. We collect an expert-annotated meta-evaluation dataset, evaluate our approach with a range of judge models, and show that it achieves high correlations with veterinary experts even with small judge models, while offering explainability across multiple dimensions. --- ### DiagLoop: A Counterfactual Data Flywheel with Stage-Localized Reinforcement for Diagnostic LLMs Authors: Jian Zhang, Bingyi Wang, Yizhi Liu Published: 2026-08-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.03674v1 Causal diagnostic models must explain how conclusions follow from evidence because diagnoses guide repairs and treatments. Yet serious cases are scarce, records rarely contain reasoning paths, and data transfer poorly across configurations, complicating local deployment. We present DiagLoop, a counterfactual data flywheel that converts codified physical relations or clinical guidelines, authored once per mechanism family, into training supervision beyond recorded cases. A training-only teacher proposes counterfactual worlds by varying causes, contexts, and observations, while an independent hybrid checker admits only valid worlds. The student reasons through symptom abstraction, causal-chain construction, and root-cause attribution. Stage-specific criteria identify its earliest failure. For nonterminal failures, a bounded repair probes downstream competence, and the resulting weakness profile guides subsequent data generation. Stage-localized reinforcement learning updates only the model-generated continuation, while replay and preservation reduce forgetting. The same criteria govern admission, attribution, reward, and regeneration through checks separate from the proposer. Using only synthesized scenarios and no case-level expert reasoning annotations, the resulting 8B model improves strict path correctness over the strongest conventional baseline. Gains are 11.6 points across eight industrial systems and 5.5 points across ten disease categories. Gains over a deranged-routing control are 3.9 and 2.3 points, respectively. The model also exceeds the evaluated proprietary references in both domains, even when they receive few-shot examples or the specification in context. --- ### CausalOPD: First-Wrong-Step Supervision for Distilling Causal Chain Reasoning Authors: Jian Zhang, Bingyi Wang, Yizhi Liu Published: 2026-08-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.03673v1 Many critical reasoning tasks, including clinical diagnosis, legal judgment, and industrial fault diagnosis, require step-dependent causal chains in which early errors propagate and correct conclusions can mask invalid reasoning. Although large language models perform well on such tasks, privacy, latency, and controllability motivate distillation into locally deployable models. Standard trajectory imitation does not correct process errors on the student's own rollout distribution. We propose CausalOPD, a curriculum online process distillation framework. A knowledge-augmented teacher first provides trajectories grounded in domain-specific causal rules, entity relations, and structural constraints. The student then generates on-policy trajectories, and the teacher identifies the first wrong step, defined as the earliest transition that verifiably violates available constraints. Starting from the verified prefix, short-horizon reinforcement learning repairs this localized failure. A causal-stage curriculum advances from evidence-level to mechanism-level and conclusion-level errors, following their propagation order. Across three domains, CausalOPD improves average path correctness by 23.4 percentage points over sequence-level online process distillation and reduces the right-label-wrong-reasoning rate from 15.7% to 4.4%. The domain-specific 8B students also surpass both evaluated proprietary references in path correctness across all domains. --- ### XiDepth: a Lightweight and Efficient Network for Self-supervised Monocular Depth Estimation Authors: Elena Izzo, Riccardo Toniolo, Lamberto Ballan Published: 2026-08-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.03666v1 Self-supervised monocular depth estimation has emerged as an appealing solution to design lightweight and effective models for deployment on computationally constrained devices due to its reduced reliance on expensive depth sensors. By eliminating the need for ground-truth annotations and leveraging the simplicity of monocular camera setups, this approach facilitates cost-effective data collection and broad applicability across fields such as computer vision and robotics. A critical challenge is achieving resource-efficient neural networks without compromising the overall performance. State-of-the-art models generally adopt depth-wise convolutions and attention mechanisms; however, these functions often incur high energy costs and face compatibility issues in embedded environments. To address this, we propose XiDepth, a lightweight architecture based on the XiNet operator block, designed to enhance feature extraction while maintaining low computational complexity and energy demand. On the KITTI dataset, XiDepth achieves state-of-the-art performance with only 0.8M parameters. Tests on a Raspberry Pi 4 further confirm its suitability for real-world embedded applications, reducing FLOPs by 40% and energy consumption by 35% compared to leading methods. --- ### Morphology-Aware Implicit Super-Resolution Network for Pathological Images Authors: Jiaming Liang, QiHui Han, Haolin Chen, Chengxin Ye, Jiawen Liu, Jiazhou Chen, Xiaoqi Sheng, Hongmin Cai Published: 2026-08-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.03664v1 Accurate diagnosis in Digital Pathology (DP) relies on high-resolution whole-slide images, yet clinical deployment is often limited by hardware costs. Super-Resolution (SR) offers a promising alternative by computationally enhancing low-resolution acquisitions. However, existing SR methods frequently struggle to preserve fine-grained cellular morphology, leading to texture oversmoothing and blurred structural boundaries under complex tissue variability. To address this issue, we propose Morph-ISR, a morphology-aware implicit super-resolution framework for DP that restores diagnostically relevant details with sub-pixel precision. Morph-ISR reformulates SR as a continuous coordinate-based reconstruction problem and integrates an Implicit Position-aware Kernel Generator (IPKG) to adaptively model spatially varying tissue morphology. To further enhance structural fidelity, a Morphological Fidelity Prior (MFP) is introduced, leveraging semantic guidance from a pre-trained cell segmentation network to enforce boundary-preserving and region-aware reconstruction, thereby improving the representation of critical cellular boundaries and nuclear textures. Experiments on TCGA and SurGen datasets show that Morph-ISR achieves the best LPIPS and ST-LPIPS among the evaluated methods, reducing them by up to 38.37% and 39.55%, respectively, over the second-best methods while maintaining strong PSNR and SSIM. These results demonstrate superior preservation of diagnostically relevant cellular boundaries and nuclear textures, while compact parameterization and high throughput support efficient edge deployment. Code and trained models will be released upon publication. --- ### Shielding for Higher-Order Safety Authors: Filip Cano, Thomas A. Henzinger, Konstantin Kueffner Published: 2026-08-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.03662v1 Safety shields are runtime enforcement mechanisms that restrict the actions of a controller to guarantee safety. Classical shields are usually synthesised for state predicates: the current physical state is either safe or unsafe, and the shield disables precisely those actions that can force the system into an unsafe state in the future. In many cyber-physical applications this view is too coarse. A vehicle approaching an obstacle should not only avoid collision, but also respect speed regulations, force limits induced by acceleration, and jerk limits to prevent injuries. From a physical perspective, these requirements are predicated over the derivatives of the state. This paper develops a finite-state safety-game construction for such high-order smoothness constraints. We define differential safety properties using finite differences over a discretised state space, characterise their expressiveness, and reduce shield synthesis to an ordinary safety game over a history state space. We give a synthesis algorithm whose shields store exactly $k$ past states for properties of order $k$ and prove that this memory is necessary. We describe an iterative synthesis procedure for a maximally permissive shield that operates over hierarchies of derivative constraints. The algorithm solves constraints iteratively in increasing order and uses the solution at each iteration to prune the state space for the next constraint. This makes shield synthesis more efficient in practice, as the algorithm refrains from exploring large regions of the state space that are known to be unsafe. --- ### Taming the Implicit: Dual-Channel Risk-Aware Reinforcement Fine-Tuning for Continual Multimodal Post-Training Authors: Yibei Liu, Jiajun Chen, Qianle Zhang, Tangyue Jin, Mengying Zhu, Meng Xi, Yangyang Wu Published: 2026-08-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.03660v1 Reinforcement fine-tuning (RFT) is widely believed to inherently resist catastrophic forgetting in continual post-training of multimodal large language models. Under pronounced task distributional shifts, however, forgetting across representative RFT algorithms escalates sharply. This stems from the implicit reward-variance regularization inherent to RFT, which proves incapable of suppressing uncontrolled optimization risk. We propose Risk-Aware Policy Optimization (RAPO), the first dual-channel framework for explicit risk governance in continual RFT. On the policy channel, Risk-Aware Policy Scaling adaptively calibrates per-sample update magnitude via rollout reliability and Fisher-inspired local predictive sensitivity; on the data channel, Risk-Aware Dynamic Bucket Sampling reorganizes training batches through dynamic risk stratification, steering optimization toward informative yet stable samples. As a plug-and-play strategy requiring no cross-task memory, RAPO generalizes to any RFT algorithm without modification. On the public MLLM-CL benchmark, RAPO reduces final forgetting by 79.8% relative to its RLOO backbone while retaining new-task competitiveness. --- ### How Closely Do LLM Reviews Align with Human Peer Review? Authors: Abraham Camelo-Guerrero, Jairo Diaz-Rodriguez Published: 2026-08-04 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2608.03659v1 Large language models (LLMs) are increasingly used to generate scientific reviews, yet existing evaluations rarely examine whether different providers align with both conference decisions and human reviewing priorities within the same controlled setting. We compare reviews from OpenAI GPT-5.4, Google Gemini 3.1 Pro Preview, and Anthropic Claude Opus 4.6 with human reviews and final decisions for 300 topic-matched ICLR 2026 submissions, equally divided among oral, poster, and rejected papers. Each model reviewed every paper using identical instructions and rating scales after decision information was removed. Our study contributes a cross-provider analysis of three complementary dimensions: alignment with broad and fine-grained decision categories, differences in recommendation-scale usage, and thematic agreement in identified weaknesses. All three LLMs distinguished accepted from rejected papers, but none reproduced the oral versus poster distinction present in human ratings. Scoring patterns were provider-specific: Gemini assigned systematically higher ratings, while OpenAI and Claude were closer to humans for rejected and poster papers but more critical of oral papers. Human and LLM reviews also differed in emphasis, with LLMs more frequently identifying missing baseline comparisons and humans more often raising computational-efficiency concerns. These results show that broad decision alignment does not imply agreement with finer human judgments or reviewing priorities. --- ### Decoupling Generation and Selection for Budget-Constrained Faithful Summarization Authors: Zeyu Wang, Guanghua Wang, Meng Xu Published: 2026-08-04 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2608.03655v1 Abstractive summarization models remain vulnerable to factual inconsistency, redundancy, and weak length control. We propose a modular generation-and-selection framework for sentence-budget-constrained summarization. A pretrained generator produces multiple candidate summaries, which are decomposed into sentence-level candidates. A combinatorial selector then constructs the final summary by balancing relevance, factuality, and redundancy under an explicit budget. The framework supports MMR, ILP, and a DPP-inspired log-determinant objective without retraining the generator. Experiments on CNN/DailyMail, Multi-News, FaithBench, and TofuEval show consistent improvements in factuality and source-grounding metrics, especially for multi-document summarization, at the cost of lower reference-overlap scores. Human evaluation further indicates higher perceived consistency, relevance, clarity, and conciseness, with a small reduction in coherence. These results show that decoupling generation from selection provides a model-agnostic mechanism for improving factual grounding. Code is available at https://anonymous.4open.science/r/bcfs-D05E/. --- ### AutoSND: From Execution Evidence to Structural Policies for Automated Network Dismantling Heuristic Discovery Authors: Zhijing Hu, Changjun Fan, Yufan Deng, Zhiguang Cao Published: 2026-08-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.03653v1 Network dismantling is fundamental to analyzing the robustness and vulnerability of complex systems, yet practical heuristics must balance effectiveness and computational efficiency, and are usually designed manually by researchers. Existing large language model based automatic heuristic design methods can generate and screen candidates, yet they have difficulty further transforming candidate quality or failure states during execution into structural-level guid- ance for subsequent generation. We propose AutoSND, a three stage tree search framework for complete network dismantling pro- grams. Stage I broadly explores from simple heuristics and archives execution evidence. Stage II compiles candidate records into struc- tural policies concerning local signals, neighborhood access, and state update ranges. Stage III continues tree search conditioned on these policies and obtains the final quality prioritized and speed prioritized candidates, AutoSND-Q/S. Experiments on 12 real world networks and 3 large real world networks show that AutoSND achieves better search performance and stability and discovers more competitive and structurally interpretable network disman- tling programs. The final candidates form an interpretable structure that uses residual degree as the backbone, adjusts node order with bounded local signals, and restricts the state update range. Code is available at https://github.com/MirrorNew/AutoSND. --- ### When Do Fewer Visual Tokens Accelerate Multimodal Inference? A Break-Even Study Across Decision Locations and Hardware Authors: Hao Dou, Ruiwen Tian Published: 2026-08-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.03649v1 Fewer visual tokens do not guarantee lower end-to-end latency. We evaluate break-even with a reproducible protocol that accounts for decision overhead, shared work, and the operators each policy can avoid. A stage-level decomposition reconciles these components with measured end-to-end latency. In a 30-example pilot, the two tested autoregressive probes remain slower than Full despite state reuse. A lightweight post-vision predictor yields paired confidence intervals below zero on RTX 3090 and A100 and remains significant after a conservative all-pairs Holm correction. A pre-vision image-size rule also yields intervals below zero on both GPUs, although neither comparison remains significant after the same correction. Pre-vision routing has a structural opportunity unavailable to post-vision pruning: it can avoid preprocessing and vision encoding. On A100, this opportunity outweighs a nearly eightfold larger downstream token reduction by the post-vision policy. Reported quality is conditional on examples answered correctly by Full and is not benchmark accuracy. --- ### Conditionally Identifiable Latent-Environment Modeling for Out-of-Distribution Recommendation Authors: Qianqian Wang, Wenwu Gong, Yunshan Li, Zhenqing Wu, Ruili Wang, Lili Yang Published: 2026-08-04 Categories: cs.IR, cs.LG arXiv: https://arxiv.org/abs/2608.03647v1 Out-of-distribution (OOD) recommendation is vulnerable to preference shifts induced by a latent environment. Existing methods can infer latent states from logged interactions, yet the statistical meaning of the latent environment and its effect on preference remain underdetermined. We formulate this task as conditionally identifiable risk-aware recommendation (CI-RR) and propose Conditionally Identifiable Latent-Environment Recommendation (CILER). CILER uses a user-conditioned exponential family to model the latent environment and a feature-indexed polynomial to specify how it changes preference. It predicts by marginalizing item probabilities over the inferred environment distribution. Under sufficient variation, correct specification, and decoder regularity, CILER identifies the environment-sensitive representation up to the stated equivalence class. We further bound excess deployment log-risk by environment-inference error. Controlled studies test the observable consequences of sufficient variation and model specification. Experiments on three datasets show that CILER improves all twelve OOD ranking metrics under feature, temporal, and geographical shifts within shared support. --- ### Is Inter-Seed Cross-Play Enough? Evaluating the Robustness of Zero-Shot Coordination Algorithms to Implementation Details Authors: Maksymilian Wolski, Nicholas Hoernle, Johannes Forkel, Jakob Foerster Published: 2026-08-04 Categories: cs.AI, cs.MA arXiv: https://arxiv.org/abs/2608.03644v1 AI agents deployed in real-world settings must be capable of coordinating with humans and other AI agents they have not encountered before. Zero-shot coordination (ZSC) algorithms aim to achieve this by specifying high-level learning rules such that independently engineered agents can coordinate with each other at test time. Rigorous evaluation of ZSC algorithms remains difficult: ideally, multiple independent implementations of each proposed algorithm must be used, reflecting the variation that arises when independent parties interpret and implement the same specification. In practice, however, ZSC algorithms have almost exclusively been evaluated using a single implementation trained across different random seeds, with only a handful of works additionally varying the neural network architecture. This leaves open questions about robustness to specification ambiguities and implementation details. In this work, we provide the first systematic evaluation of this robustness. We introduce a new evaluation scheme, cross-implementation cross-play, varying implementation details that prior work has shown to affect the performance of multi-agent reinforcement learning (MARL) algorithms, and we evaluate Other-Play, a popular ZSC algorithm, with this scheme. Our findings are encouraging and suggest that, for Other-Play, the standard ZSC evaluation is, in fact, a reasonable proxy for this more thorough cross-implementation evaluation. --- ### Learning Biomechanically Plausible Human Motion from Sparse Radar Point Clouds Authors: Jonas Leo Mueller, Markus Gambietz, Alexander Weiss, Daniel Krauss, Bjoern M. Eskofier Published: 2026-08-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.03637v1 Radar-based human pose estimation has focused on improving learning algorithms while representing the body as unconstrained keypoint coordinates. We address the underexplored dimension of anatomical fidelity by integrating a full-body skeletal model into a differentiable, end-to-end trainable radar-based pose estimation framework, in which the pose network is supervised through forward kinematics while subject-specific geometry is fitted beforehand. Subject-specific body segment proportions are predicted from radar point cloud features to scale a biomechanical skeleton. A motion prediction network maps temporal radar sequences to generalized coordinates, and differentiable forward kinematics converts predicted joint angles into 3D positions. A contact classification loss encourages physically plausible foot-ground interaction. Under leave-one-subject-out cross-validation on 11 healthy participants performing rehabilitation exercises, the framework achieves 6.456 +/- 1.759 cm mean per-joint position error (MPJPE), 8.083 +/- 0.884 degrees mean per-joint angle error (MPJAE), 0.935 +/- 0.009 contact classification F1, and 3.4 +/- 1.3 % scaling error. This proof-of-concept study demonstrates the feasibility of recovering interpretable biomechanical descriptors from a single low-cost radar sensor in a controlled laboratory setting, a prerequisite for future clinical motion analysis. --- ### MuEvo: LLM-Driven Evolution of Multi-Heuristic Ensemble Authors: Haoze Lv, Ning Lu, Shengcai Liu, Shaofeng Zhang, Ke Tang Published: 2026-08-04 Categories: cs.NE, cs.AI arXiv: https://arxiv.org/abs/2608.03636v1 Large language model-based automated heuristic design (LLM-AHD) has shown strong potential in discovering effective heuristics for combinatorial optimization problems. However, existing methods primarily optimize a single heuristic, whereas practical optimization frameworks often rely on multiple interacting components. Directly extending single-heuristic methods is challenging because early component selection can overlook components with late potential, while independent evolution ignores inter-component dependencies. We propose MuEvo, an LLM-driven framework for evolving heuristic ensembles under ensemble-level feedback. MuEvo combines Dynamic Component Management, which uses short-budget probing and a reversible lifecycle to revise component priorities throughout the search, with LLM-Driven Co-Evolution, which coordinates component populations through Multi-Ensemble Evaluation, Cross-Component Information Sharing, Relation-Guided Pair Evolution, and Adaptive Budget Allocation. We evaluate MuEvo on selection hyper-heuristics and componentized ant colony optimization across four combinatorial optimization domains. Results show that MuEvo consistently improves human-designed frameworks and outperforms representative multi-component extensions of state-of-the-art LLM-AHD methods, demonstrating its effectiveness across both controller-mediated heuristic pools and functionally differentiated algorithmic components. --- ### When Teachers Mislead: Spurious-Signal-Aware On-Policy Distillation Authors: Yinuo Jiang, Yongjie Ye, Zhou Tao, Xiang Zhuang, Qiang Zhang, Huajun Chen, Tiankai Li Published: 2026-08-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.03632v1 On-Policy distillation (OPD) transfers teacher capabilities by supervising student-sampled trajectories with dense token-level teacher signals. Recent selective OPD methods improve this process by prioritizing signals that are confident, informative, or learnable. However, the assumptions overlook a fundamental failure mode of language models: their token-level judgments can be driven by input-agnostic language priors, formatting conventions, or stereotyped reasoning templates rather than task-specific evidence. We refer to such optimization-relevant but weakly input-grounded supervision as spurious signals in OPD, which may produce large gradients while contributing little task-improving direction. To mitigate this issue, we propose SA-OPD, a Spurious-Signal-Aware On-Policy Distillation framework that identifies and filters misleading token-level supervision based on input-groundedness and optimization impact. SA-OPD introduces a lightweight input-groundedness proxy estimating whether a token-level distillation signal truly depends on the input. It then filters only tokens that simultaneously exhibit low input-groundedness and extreme distillation divergence, thereby removing high-impact spurious updates and achieving fine-grained OPD optimization. Extensive experiments on both large language model (LLM) and vision-language model (VLM) settings demonstrate that SA-OPD consistently outperforms Vanilla OPD and competitive selective methods. These results establish input-groundedness as a key dimension for OPD supervision selection and offer a simple, effective strategy for mitigating spurious updates. --- ### SEER: A Self-Grounded Evidence Interface for Controlled Spatial Relation Classification Authors: Feixiang Liu, Likun Wang, Qiang Qiu, Hui Xu, Huawei Shen, Xueqi Cheng Published: 2026-08-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.03631v1 Spatial relation questions require a model to identify the queried subject and object before comparing their layout. Yet a VLM can recognize both entities and still answer from the wrong instance or an ambiguous global view. We ask whether making query-specific evidence explicit can mitigate this failure and propose SEER (Self-grounded Evidence for Entity-Relation Reasoning), a training-free inference-time evidence interface for frozen VLMs. SEER hides candidate relations during pair localization, constructs a query-specific view with explicit subject/object roles, and retains the full image and sparse box geometry as complementary evidence. For relation-choice protocols with exact inverse support, an optional refinement swaps the entity roles and changes the forward decision only when exactly one visual state obeys the corresponding inverse relation. On an image-disjoint GQA-Train900 test frozen before model scoring, SEER pools to +3.94 (2.17,5.72) over Full; the gain remains positive under label-independent grounding-order counterbalancing and on the 535 rows whose entity names are unique. The unchanged protocol yields +4.35 to +11.79 on all 2,434 filtered EmbSpatial pair-relation questions across three models. Matched controls separate local refocus from role-explicit conditioning. These results establish query-specific evidence construction as the principal intervention, with reciprocal consistency as a smaller protocol-specific refinement. --- ### POEM: Phase-Aware $\mathrm{SO}(2)$ Feature Rotation for Time Series Forecasting Under Periodicity Drift Authors: Jiawen Zhu, Shuhan Liu, Shengxuan Li, Qiming Shi, Di Weng Published: 2026-08-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.03630v1 Deep learning has advanced time series forecasting, but periodicity drift, in which cycle timing and phase vary over time, remains a challenging problem. Existing methods predominantly model these sequences on fixed time grids, suffering from a limited ability to accommodate phase-related variation. To address this limitation, we propose \textbf{POEM}, a phase-aware forecasting framework based on latent feature rotation using the special orthogonal group in two dimensions, denoted by $\mathrm{SO}(2)$. POEM aims to reduce the phase-related variability by learning a phase-correction coordinate and applying an invertible $\mathrm{SO}(2)$-based rotation to paired latent features. To extrapolate this correction coordinate, Directional Phase Increment Attention (DPIA) retrieves historical phase increments from similar temporal contexts and integrates them into future phase corrections. Experiments demonstrate that POEM achieves competitive performance, while qualitative visualizations suggest that the learned phase-aware transformation makes latent trajectories more regular. --- ### Cross-Layer Interaction under Weight-Space Ablation: A Closed-Form Attention Jacobian Bound and a Test on a Real Pretrained Model Authors: Abdallah Khemais Published: 2026-08-04 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2608.03629v2 A companion paper studies when activation patching and weight-space ablation agree, inside an idealized model where a conditional computation is carried additively through a residual stream. For the one composition in that model where two carriers are architecturally dependent, an attention head and its own layer's normalization-MLP composition, it derives an exact first-order interaction formula, zero when only the MLP is ablated and second-order bounded when the head is also ablated. That result is confined to a single residual block and checked only on small transformers on a synthetic task. This paper extends the result past both limits. First, the interaction from ablating carriers spanning several layers decomposes exactly into same-block terms, one per touched layer, plus a cross-layer remainder on which the decomposition makes no claim of smallness. Second, we isolate that remainder exactly, for two layers, as a double integral of a mixed second derivative, and name the missing ingredient needed to bound it: a Jacobian bound for the attention sub-block. We derive this bound in closed form and verify it, without a single violation, against Qwen2.5-1.5B-Instruct's real weights, though we do not yet chain it across layers. We also give, in closed form, the curvature constant the companion paper's bound leaves unexhibited. Third, on that same model, we search for and find an emergent circuit for indirect object identification, never designed into it, using the original activation-patching method for this task, and test collapse, dissociation, and interaction on it. The result is mixed: a shared carrier emerges across all five tested instances, collapse and dissociation hold on most but not all, and a nonzero interaction is measurable on three of five, at layer pairs outside the same-block case the companion theorem covers. --- ### Cross-Layer Interaction under Weight-Space Ablation: A Closed-Form Attention Jacobian Bound and a Test on a Real Pretrained Model Authors: Abdallah Khemais Published: 2026-08-04 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2608.03629v1 A companion paper studies when activation patching and weight-space ablation agree, inside an idealized model where a conditional computation is carried additively through a residual stream. For the one composition in that model where two carriers are architecturally dependent, an attention head and its own layer's normalization-MLP composition, it derives an exact first-order interaction formula, zero when only the MLP is ablated and second-order bounded when the head is also ablated. That result is confined to a single residual block and checked only on small transformers on a synthetic task. This paper extends the result past both limits. First, the interaction from ablating carriers spanning several layers decomposes exactly into same-block terms, one per touched layer, plus a cross-layer remainder on which the decomposition makes no claim of smallness. Second, we isolate that remainder exactly, for two layers, as a double integral of a mixed second derivative, and name the missing ingredient needed to bound it: a Jacobian bound for the attention sub-block. We derive this bound in closed form and verify it, without a single violation, against Qwen2.5-1.5B-Instruct's real weights, though we do not yet chain it across layers. We also give, in closed form, the curvature constant the companion paper's bound leaves unexhibited. Third, on that same model, we search for and find an emergent circuit for indirect object identification, never designed into it, using the original activation-patching method for this task, and test collapse, dissociation, and interaction on it. The result is mixed: a shared carrier emerges across all five tested instances, collapse and dissociation hold on most but not all, and a nonzero interaction is measurable on three of five, at layer pairs outside the same-block case the companion theorem covers. --- ### ConformalShift: Targeted Event Reordering Against Adaptive ECG Monitoring Authors: Arash Vashagh, Yasmin Vashagh Published: 2026-08-04 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.03628v1 Adaptive conformal prediction can recover clinically important heartbeat classes missed by a point classifier, but delayed feedback makes its decisions sensitive to event order. We introduce ConformalShift, a bounded event-reordering attack that suppresses the ventricular class for rescued events without modifying ECG waveforms, labels, classifier scores, or the event multiset. ConformalShift searches for feasible permutations of authentic preceding events that lower the ventricular threshold before a selected target is evaluated. On disjoint MIT--BIH confirmation records, the attack suppressed 66.7% of eligible targets for Extra Trees and 60.0% for HistGradientBoosting, compared with random-schedule rates of 4.4% and 12.0%, respectively. Transferred configurations also outperformed random scheduling on INCART, while reducing the displacement budget weakened the attack on both datasets. These results show that adaptive monitors in healthcare can be compromised through the timing of authentic information, even when waveforms, labels, classifier outputs, and event contents remain unchanged. --- ### Unequal Verdicts: Investigating Gender Bias in LLM-Based Fake News Detection Authors: Razieh Chalehchaleh, Reza Farahbakhsh, Noel Crespi Published: 2026-08-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.03627v2 Large Language Models (LLMs) are increasingly used for automated fact-checking, yet their susceptibility to gender bias in this context remains underexplored. This study presents the first systematic investigation of gender bias in LLM-based fake news detection using real-world data. We augment the LIAR benchmark with three gender variants of speaker job titles (Neutral, Male, Female) for each statement to test whether veracity judgments vary solely based on gender presentation. Six state-of-the-art LLMs are evaluated across multiple bias and fairness metrics. All models exhibit gender sensitivity: 9.79%-35.13% of statements receive inconsistent labels across the three variants, with Male-Female comparisons showing 6.5%-23.6% flip rates. Two primary bias manifestations are identified: instability (inconsistent judgments) and directionality (systematic favoritism). Five models show statistically significant directional effects, with the strongest effects displaying male-skeptic patterns. These findings demonstrate that gender bias undermines both reliability and fairness in LLM-based fake news detection, highlighting the need for bias-aware evaluation and mitigation strategies. The augmented dataset is publicly released to support future research. --- ### Unequal Verdicts: Investigating Gender Bias in LLM-Based Fake News Detection Authors: Razieh Chalehchaleh, Reza Farahbakhsh, Noel Crespi Published: 2026-08-04 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.03627v1 Large Language Models (LLMs) are increasingly used for automated fact-checking, yet their susceptibility to gender bias in this context remains underexplored. This study presents the first systematic investigation of gender bias in LLM-based fake news detection using real-world data. We augment the LIAR benchmark with three gender variants of speaker job titles (Neutral, Male, Female) for each statement to test whether veracity judgments vary solely based on gender presentation. Six state-of-the-art LLMs are evaluated across multiple bias and fairness metrics. All models exhibit gender sensitivity: 9.79%-35.13% of statements receive inconsistent labels across the three variants, with Male-Female comparisons showing 6.5%-23.6% flip rates. Two primary bias manifestations are identified: instability (inconsistent judgments) and directionality (systematic favoritism). Five models show statistically significant directional effects, with the strongest effects displaying male-skeptic patterns. These findings demonstrate that gender bias undermines both reliability and fairness in LLM-based fake news detection, highlighting the need for bias-aware evaluation and mitigation strategies. The augmented dataset is publicly released to support future research. --- ### A Security-Oriented Lifecycle Model for Large Language Model Systems Authors: Eleftherios Batzolis, George Drosatos, Vassilis Katsouros, Konstantinos Rantos Published: 2026-08-04 Categories: cs.CR, cs.AI, cs.CY arXiv: https://arxiv.org/abs/2608.03626v1 Large language models are being integrated into critical infrastructure and enterprise workflows at unprecedented scale,yet the lifecycle frameworks governing their development and operations were designed for operational efficiency rather than security analysis. As a result, security-relevant activities such as data provenance verification, artifact signing, agentic permission control, and decommissioning are often left implicit or assumed to receive due care. Governance frameworks, in turn, organise requirements around risk levels or management processes without clearly linking them to the lifecycle stages where they apply. This paper addresses both deficiencies. We propose a lifecycle model for LLM systems that supports security analysis by structuring it around security-relevant boundaries rather than workflow optimisation. The model comprises 32 stages across four core pipeline layers (Data, Model, Distribution, Application), supported by a 12-stage LLMOps pillar and a 9-category governance pillar. Thirteen stages are introduced here as separate units because they expose distinct security concerns that existing frameworks do not clearly distinguish. A governance mapping synthesising the NIST AI RMF, the EU AI Act, and ISO/IEC 42001 reveals a structural property of the current regulatory landscape: governance evidence concentrates at deployment-facing stages, where systems are visible to regulators, while the most consequential decisions, data selection, alignment strategy, and capability boundaries, are made at development-facing stages, where regulatory visibility is lowest. --- ### LoopMTP: A looped transformer guided by latent multi-token prediction Authors: Behzad Shomali, Markus Frey, David Berghaus, Joachim Koehler, Mehdi Ali Published: 2026-08-04 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.03624v1 Looped transformers have emerged as a parameter-efficient alternative to scaling depth for strong reasoning. By reusing one stack of layers across $T$ iterations, they attain the effective depth and reasoning capabilities of larger models at a fixed parameter count. Yet existing approaches suffer from latent overthinking and undifferentiated computation, largely because intermediate representations receive no guidance across loops. Multi-token prediction (MTP) supplies exactly the dense, forward-looking supervision the loop is missing. We propose \textsc{LoopMTP}, which links the two through a structural correspondence in latent space: a model that loops $T$ times can anticipate $T$ future tokens. \textsc{LoopMTP} realizes this by softly aligning the hidden state of loop $t$ with the embedding of the token $t$ steps ahead, while a lightweight gate preserves useful information across iterations. \textsc{LoopMTP} improves average accuracy by up to 8.1\% (relative) over the non-looped baseline, with training remaining stable for up to 15 loops. --- ### A Theory of Conditional Collapse under Low-Rank Weight-Space Ablations: I. The Single-Block Theory and Synthetic Validation Authors: Abdallah Khemais Published: 2026-08-04 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2608.03620v2 Activation patching and weight-space ablation both claim a component is causally responsible for a behavior, yet they act on different objects: one forward pass versus the parameters behind every forward pass. We ask when they agree. We study an idealized model where a conditional computation is carried additively through a residual stream, $F(x)=F_0(x)+\sum_iα_i(x)v_i$, read out by a linear functional, and prove three exact results. First, deleting a subset of carriers collapses a matched input pair onto the same unconditional output \emph{if and only if} the removal is symmetric on the pair and leaves no outside contrast; the error is deterministic, and we give its exact form even when the two conditions hold only approximately. Second, patching a carrier moves the readout by its donor-receiver \emph{contrast}, while ablating it moves the readout by its \emph{absolute level}; neither bounds the other, and we construct pairs where every single-carrier patch flips the decision while no single-carrier ablation does. Third, for an attention head composed with its own layer's normalization and MLP, we derive an exact first-order interaction formula with a provably second-order remainder, vanishing identically when only the MLP is ablated but not, in general, when a head is. Small transformers trained on a synthetic conditional task illustrate all three predictions: across thirty-nine ablation configurations the measured interaction is strongly rank-correlated with the idealized model's predictive accuracy (Spearman $-0.83$), and a second task and architecture reproduces the same pattern, including a further polarity reversal. The single-block interaction result extends past one residual block, and the synthetic validation is tested against a real pretrained model, in a companion paper that takes this theory further along both axes. --- ### A Theory of Conditional Collapse under Low-Rank Weight-Space Ablations: I. The Single-Block Theory and Synthetic Validation Authors: Abdallah Khemais Published: 2026-08-04 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2608.03620v1 Activation patching and weight-space ablation both claim a component is causally responsible for a behavior, yet they act on different objects: one forward pass versus the parameters behind every forward pass. We ask when they agree. We study an idealized model where a conditional computation is carried additively through a residual stream, $F(x)=F_0(x)+\sum_iα_i(x)v_i$, read out by a linear functional, and prove three exact results. First, deleting a subset of carriers collapses a matched input pair onto the same unconditional output \emph{if and only if} the removal is symmetric on the pair and leaves no outside contrast; the error is deterministic, and we give its exact form even when the two conditions hold only approximately. Second, patching a carrier moves the readout by its donor-receiver \emph{contrast}, while ablating it moves the readout by its \emph{absolute level}; neither bounds the other, and we construct pairs where every single-carrier patch flips the decision while no single-carrier ablation does. Third, for an attention head composed with its own layer's normalization and MLP, we derive an exact first-order interaction formula with a provably second-order remainder, vanishing identically when only the MLP is ablated but not, in general, when a head is. Small transformers trained on a synthetic conditional task illustrate all three predictions: across thirty-nine ablation configurations the measured interaction is strongly rank-correlated with the idealized model's predictive accuracy (Spearman $-0.83$), and a second task and architecture reproduces the same pattern, including a further polarity reversal. The single-block interaction result extends past one residual block, and the synthetic validation is tested against a real pretrained model, in a companion paper that takes this theory further along both axes. --- ### Geospatial-Prior Guidance for 3D Semantic Scene Completion Authors: Meng Wang, Shougao Zhang, Wenzhe He, Ruihui Li, Nan Hu, Zhuo Tang, Kenli Li Published: 2026-08-04 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.03618v1 Inferring complete 3D geometry and semantics from onboard images remains challenging because occlusions and restricted fields of view leave large scene regions underconstrained. Although satellite imagery provides wide-area context, appearance cues alone offer limited structural guidance and may be unreliable because of spatial or temporal discrepancies. We present GeoScene, a geospatially guided framework that jointly uses satellite imagery and structured OpenStreetMap cues as soft priors for 3D semantic scene completion. GeoScene learns complementary voxel-wise reliability weights for onboard observations and geospatial guidance, and uses them to control feature refinement in observed and unobserved regions. This design preserves local visual evidence while exploiting large-scale road and building structure beyond onboard visibility. Experiments on SemanticKITTI and SSCBench-KITTI-360 demonstrate that GeoScene consistently improves both geometric and semantic completion under the geospatial-prior-assisted setting, with the most pronounced benefits for large-scale static and geospatially structured classes. --- ### A machine-readable catalogue of the Tsiolkovsky papers (fond 555, Archive of the Russian Academy of Sciences), and a way to measure how well its handwriting can be read Authors: Vladimir Beskorovainyi Published: 2026-08-04 Categories: cs.CL, cs.CV, cs.DL arXiv: https://arxiv.org/abs/2608.03617v2 The personal archive of Konstantin Tsiolkovsky (1857-1935) is held as fond 555 of the Archive of the Russian Academy of Sciences. The archive scanned the fond and published the images, but with no queryable catalogue, no full-text search and no dataset: the holdings can only be browsed one page at a time. This paper describes a machine-readable catalogue of all 2,019 files and 51,008 scans, a dating for 1,969 files taken from the archive's own descriptions, a page-level classification of every scan into handwriting and typescript, and a machine transcription of the fond in full: all 2,019 files and all 51,008 scans. It also reports a way to measure handwritten-text-recognition accuracy in an archive with no ground truth. Archives of the typewriter era often preserve one text twice, as manuscript and as a typed copy; transcribing both and comparing isolates the reading error, since source and pipeline are identical and only page difficulty differs. Across 1,759 such pairs from 224 files, two readings of a handwritten page agree on a median 37% of words and share a longest verbatim run of a median 10 words; the median is unchanged from the 294 pairs of the first version, on a sample six times larger. On two files that also have a published edition the estimate can be checked against ground truth: it is unbiased to within a percentage point and ranks pages as the truth does (rank correlation 0.92 where the edition is a faithful witness). This bounds use: two variants of one work here share 19% of words, below the rate at which two readings of a single page agree, so the redactions cannot be collated word by word at this quality. That negative result is reported as such, and the constraint is built into the tool. --- ### A machine-readable catalogue of the Tsiolkovsky papers (fond 555, Archive of the Russian Academy of Sciences), and a way to measure how well its handwriting can be read Authors: Vladimir Beskorovainyi Published: 2026-08-04 Categories: cs.CL, cs.CV, cs.DL arXiv: https://arxiv.org/abs/2608.03617v1 The personal archive of Konstantin Tsiolkovsky (1857-1935) is held as fond 555 of the Archive of the Russian Academy of Sciences. The archive scanned the fond and published the images, but with no queryable catalogue, no full-text search and no dataset: the holdings can only be browsed one page at a time. This paper describes a machine-readable catalogue of all 2,019 files and 51,008 scans, a dating for 1,969 files taken from the archive's own descriptions, a page-level classification of every scan into handwriting and typescript, and a growing corpus of machine transcriptions (currently 322 files, 5,454 scans). It also reports a way to measure handwritten-text-recognition accuracy in an archive with no ground truth. Archives of the typewriter era often preserve one text twice, as manuscript and as a typed copy; transcribing both and comparing isolates the reading error, since source and pipeline are identical and only page difficulty differs. Across 294 such pairs from 27 files, two readings of a handwritten page agree on a median 37% of words. On two files that also have a published edition the estimate can be checked against ground truth: it is unbiased to within a percentage point and ranks pages as the truth does (rank correlation 0.92 where the edition is a faithful witness). This bounds use: two variants of one work here share 19% of words, below the rate at which two readings of a single page agree, so the redactions cannot be collated word by word at this quality. That negative result is reported as such, and the constraint is built into the tool. --- ### Predictive Enhancement Calibration for Latent Breast MRI Virtual Contrast Enhancement Authors: Qin Lei, Hao Wu Published: 2026-08-04 Categories: eess.IV, cs.CV arXiv: https://arxiv.org/abs/2608.03612v1 Virtual contrast enhancement (VCE) synthesizes enhanced breast MR images from pre-contrast acquisitions. Modern latent generators offer strong image priors, but their bounded natural-image autoencoders conflict with the non-canonical intensity scale of MRI. We show that the upper bound can alter radiomic fidelity before generation, while scaling source and target independently creates a coordinate inconsistency. We propose Predictive Enhancement Calibration (PEC), which represents each pair in a shared, case-adaptive coordinate during training and predicts its unavailable upper endpoint from the pre-contrast image at inference. We integrate PEC with a pretrained FLUX latent flow transformer via parameter-efficient reference conditioning. Target round trips first isolate representation loss before generation; near-matched conditional models then compare PEC with fixed-wide and separate coordinates under comparable training budgets and backbone settings. On the fixed internal MAMA100 development cohort, PEC improves all eight point estimates in this source-only VCE setting, with paired evidence strongest for MSE and LPIPS.\noindent\textbf{Code:} https://github.com/tanlei0/pec-breast-mri-vce ---