# Research Papers — 2026-05 · bundle page 188 of 326 > Full-text markdown of papers #9351–9400 of 16266 published in 2026-05, newest first. AI-readable mirrors of the full 16266-paper 2026-05 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-05/page/188 [← Previous bundle](https://aipod.dev/papers/archive/2026-05/page/187.md) [Next bundle →](https://aipod.dev/papers/archive/2026-05/page/189.md) ### Reducing Hallucination in Multimodal Large Language Models through Hard Grounding Preference Supervision Authors: Qinwu Xu Published: 2026-05-13 Categories: cs.CV, cs.AI, cs.CL, cs.DB, cs.LG arXiv: https://arxiv.org/abs/2605.16411v4 Hallucination remains a major challenge in vision-language models (VLMs), particularly when linguistically plausible responses are unsupported by visual evidence. We study whether multimodal hallucination can be reduced by concentrating post-training supervision on hard grounding boundaries, where preferred and rejected responses are semantically close but differ in their support from observable visual evidence. Under a frozen visual encoder and cross-modal alignment pathway, we first use supervised fine-tuning (SFT) to establish broad decoder-side multimodal behavior and then construct hard grounding preference pairs for Direct Preference Optimization (DPO). These pairs target evidence utilization, calibration, and grounding consistency across fine-grained recognition, spatial reasoning, OCR, ambiguous or insufficient evidence, and false-premise queries. The resulting DPO model consistently improves over its SFT initialization on DocVQA, TextVQA, MMBench, and VQAv2. Across approximately 839 additional multimodal examples, it further achieves 6.2--8.2 percentage-point higher pairwise win rates under three independent LLM judges. We additionally release HardVQA-DPO, a curated and growing resource with more than 3K hard-grounding SFT examples and an initial set of DPO preference pairs.https://huggingface.co/datasets/vlmgrounding/hardvqa-sft-dpo These results show that decoder-side adaptation alone can improve a meaningful subset of grounding failures under a fixed multimodal representation, while preserving broader multimodal capabilities. --- ### Reducing Hallucination in Vision-Language Models via Stage-wise Preference Optimization under Distribution Shift Authors: Qinwu Xu Published: 2026-05-13 Categories: cs.CV, cs.AI, cs.CL, cs.DB, cs.LG arXiv: https://arxiv.org/abs/2605.16411v2 Hallucination remains a fundamental challenge in vision-language models (VLMs), where autoregressive generation may produce linguistically plausible yet physically inconsistent or visually ungrounded responses due to likelihood maximization under joint probabilistic modeling. We propose a stage-wise preference optimization framework for hallucination reduction through targeted multimodal data construction. Rather than directly optimizing on generic instruction-following data, our approach progressively constructs hallucination-focused preference pairs near known failure boundaries. The framework emphasizes ambiguous spatial orientation, object relationships, OCR uncertainty, and adversarial false-premise training. Hallucinated negatives are generated through minimally perturbed yet visually inconsistent alternatives, enabling Direct Preference Optimization (DPO) to better separate grounded reasoning from plausible hallucination. Experiments on open-source benchmarks and real-world multimodal evaluation scenarios demonstrate improved grounding consistency, reduced hallucination, and more informative grounded responses. Cross-model qualitative evaluation further shows that the proposed multimodal LLM DPO framework produces more visually grounded responses than several frontier proprietary VLMs, such as in ambiguous spatial reasoning and adversarial false-premise settings. The results suggest that hallucination may arise not only from limited model capacity, but also from inherent tendencies of autoregressive probabilistic generation to favor linguistically plausible continuations under weak visual grounding. Future work may explore physical consistency modeling, uncertainty-aware multimodal reasoning, and architectural alternatives beyond standard autoregressive decoding. --- ### Test-Time Hinting for Black-Box Vision-Language Models Authors: Kaihua Hou, Abhijith Varma Mudunuri, Jiaxing Qiu, Roxana Daneshjou, Thomas Hartvigsen, Ahmed Alaa Published: 2026-05-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.16410v1 Test-time scaling (TTS) methods have proven highly effective for LLMs, yet their application to vision-language models (VLMs) remains relatively underexplored. Existing VLM TTS methods largely require open-weight model access or expensive repeated sampling, and are evaluated primarily on multimodal mathematical and scientific reasoning benchmarks rather than general visual understanding tasks. In this paper, we propose Test-Time Hinting, a method that improves VLM performance via a single VLM call and requiring only black-box API access, which makes it broadly applicable to frontier closed-weight models. Our method is motivated by the observation that VLM errors tend to cluster around recurring failure patterns. We therefore train a lightweight hint generator model to predict, for a given test input, which "hint" should be prepended to the prompt, providing targeted contextual or procedural guidance that steers the VLM away from its characteristic failure modes. We show that Test-Time Hinting improves the accuracy of multiple closed-weight VLMs on natural-image VQA benchmarks and that these gains generalize to unseen benchmarks and VLMs without retraining the hint generator. --- ### Multilingual OCR-Aware Fine-Tuning and Prompt-Guided Chain-of-Thought Reasoning for Multimodal Large Language Models Authors: Qinwu Xu, Yifan Jiang, Haoyu Ren Published: 2026-05-13 Categories: cs.CV, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2605.16409v3 Optical character recognition (OCR) and multilingual scene-text understanding remain challenging for multimodal large language models (MLLMs), particularly in real-world images containing small or degraded text, cluttered layouts, occlusion, handwriting, and complex typography. We present an OCR-aware multilingual post-training framework that improves visual-text grounding in a general-purpose MLLM without requiring an external OCR engine, OCR-extracted text, or text bounding boxes at inference time. The framework combines large-scale multilingual OCR supervision, approximately 5M additional multilingual training samples, controlled synthetic OCR generation and in-image text translation, LoRA-based supervised fine-tuning (SFT), and lightweight OCR-oriented Chain-of-Thought prompting. On a held-out real-world multilingual OCR benchmark, OCR-SFT improves OCR completeness from 71.3 to 84.6, reduces hallucination rate from 18.3\% to 5.5\%, and improves translation BLEU-1 from 52.3 to 80.2, with substantial hallucination reductions under blur and rotation. Evaluation on public benchmarks further shows gains on OCR-intensive tasks while largely preserving broader multimodal capabilities; ablations show that SFT provides the primary improvement, with prompting offering smaller complementary gains. These results demonstrate that data-centric OCR-aware post-training provides a practical and scalable approach to improving multilingual visual-text grounding in general-purpose MLLMs. --- ### Multilingual OCR-Aware Fine-Tuning and Prompt-Guided Chain-of-Thought Reasoning for Multimodal Large Language Models Authors: Qinwu Xu, Yifan Jiang, Haoyu Ren Published: 2026-05-13 Categories: cs.CV, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2605.16409v2 Optical character recognition (OCR) and multilingual text understanding remain major failure modes of multimodal large language models (MLLMs), particularly in real-world images containing cluttered layouts, small fonts, blur, occlusion, and complex typography. We present an OCR-aware multilingual multimodal training framework that combines (i) large-scale synthetic OCR-to-translation data generation, (ii) OCR-aware supervised fine-tuning (SFT) with LoRA adaptation, and (iii) structured visual chain-of-thought (CoT) prompting for reasoning under uncertain visual conditions. Using a LLaMA-based multimodal architecture, the proposed framework substantially improves OCR completeness, multilingual translation accuracy, and robustness under degraded visual conditions. Experimental results on multilingual receipts, menus, posters, signs, handwritten text, and document images demonstrate significantly improved visual-text grounding compared with the baseline model. In particular, the proposed OCR-aware post-training framework improves extraction of small, blurred, spatially scattered, and partially occluded text while reducing reliance on language priors under uncertain OCR conditions. Qualitative comparisons with frontier multimodal systems, including GPT-5-class and Gemini-family models, further suggest improved OCR grounding and reduced hallucination under noisy and visually ambiguous OCR scenarios. Overall, the results indicate that data-centric OCR-aware multimodal post-training provides an effective and scalable direction for improving multilingual OCR and OCR-based visual question answering systems. --- ### Visual Search Patterns in 3D Pancreatic Imaging: An Eye Tracking Study Authors: Anna Anikina, Leila Khaertdinova, Trine Balschmidt, Michael B Andersen, Christoph F Müller, Erik GS Brandt, Henrik S Thomsen, Claudia Mello-Thoms, Bulat Ibragimov Published: 2026-05-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.16408v1 Eye tracking has emerged as a powerful tool for examining visual perception and search strategies in various domains, including medicine. While it is relatively straightforward to apply in 2D settings, its use in 3D medical imaging remains challenging and not yet well explored. This gap is particularly relevant for radiology, where volumetric images such as computed tomography (CT) scans are routinely read by medical experts. Radiologists typically interpret these images by navigating through hundreds of 2D slices, most often viewed in the axial projection. A taxonomy of eye movement data during navigation through a CT volume could be valuable to understand how radiologists approach diagnostic tasks. As an example of the derived taxonomy, we asked two radiologists to search abdominal CTs of the pancreas. We collect eye tracking data and align eye gaze movements with slice navigation to visualize the representation of the pancreas through volume and analyze clinicians' gaze behavior in both space and time. --- ### Proof-Carrying Certificates for LLM Pipelines: A Trust-Boundary Architecture Authors: George Koomullil Published: 2026-05-13 Categories: cs.LO, cs.CL, cs.CR, cs.PL arXiv: https://arxiv.org/abs/2605.16407v1 We present a framework for verifying the deterministic structured computations surrounding a large language model rather than the model itself, extending a Lean 4 trust-boundary architecture to the generic interfaces of modern LLM pipelines. Certificate validity is a Lean 4 kernel type-check plus a sorry-free transitive axiom audit against the trusted set {propext, Classical.choice, Quot.sound}; other assumptions are declared and partitioned by tier (mathematical placeholders, cryptographic assumptions, ML/human oracles). The technical contribution comprises three local certificate families and two operators. The families are conflict-aware bilattice grounding (with an emission-gate soundness lemma), embedding sensitivity and paraphrase stability, and Hoare-style agent action. The operators are a Maximal Certifiable Residue, which turns abstention into the maximum-weight certifiable residue with audit-logged dropped claims, and a Compositional Stability theorem, which yields a closed-form pipeline-wide perturbation budget from per-layer gains and margins. The three families plus a Universal Assurance Card consolidator form the per-call deliverable for high-stakes deployments: patent and legal retrieval, regulated finance, clinical decision support, and agentic systems with irreversible side effects. A compiled Lean 4 reference artifact (Lean v4.30.0-rc2, Mathlib) covers all 22 certificate types, with 17 of 46 kernel-audited declarations axiom-free, the rest depending only on the trusted set and declared assumptions, and zero uses of sorryAx or Lean.ofReduceBool. The three families are empirically tested through four registered pilots: bilattice grounding on adversarially perturbed HotpotQA, embedding sensitivity in short- and long-form settings, and Hoare-style agent action on a filesystem sandbox with adversarial prompt injection. --- ### Contrastive-SDXL: Annotation-Preserving Night-Time Augmentation for Pedestrian Detection Authors: Franky George, Muhammad Khalid, Adil Khan Published: 2026-05-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.16406v1 Night-time pedestrian detection remains challenging because labelled night-time data are limited and large illumination differences make daytime-only trained detectors unreliable. Latent diffusion models (LDMs) provide a powerful basis for image-to-image translation and cross-domain augmentation, but their effectiveness in safety-critical perception depends on whether detector-relevant objects and local semantic structure are preserved when translating between source and target domains. In this work, we present Contrastive-SDXL, a day-to-night augmentation framework for night-time pedestrian detection built on SDXL-Turbo and fine-tuned using Low-Rank Adaptation (LoRA). To preserve semantic correspondence between daytime inputs and translated night-time images, we introduce a patch-wise semantic contrastive loss guided by a pretrained DINOv2 encoder rather than generator encoder features. Multi-level DINOv2 self-attention maps enforce both local and global semantic consistency, while an object consistency loss explicitly encourages pedestrian preservation. Contrastive-SDXL produces realistic night-time images, achieving a Frechet Inception Distance (FID) of 22.5. Detectors trained with our synthetic images obtain a 6-7% reduction in miss rate compared with a daytime-only baseline, approaching the performance of detectors trained on real night-time data. These results demonstrate that consistency-driven diffusion augmentation can effectively support safety-critical night-time pedestrian detection.Specific --- ### Concepts Worth Having: Refining VLM-Guided Concept Bottleneck Models with Minimal Annotations Authors: Nicola Debole, Andrea Passerini, Stefano Teso, Andrea Pugnana, Emanuele Marconato Published: 2026-05-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.16405v1 Concept-bottleneck models (CBMs) are neural classifiers that compute predictions from high-level concepts extracted from the input. CBMs ensure stakeholders can understand the concepts -- and the predictions they entail -- by learning these from concept-level annotations, which are however seldom available. Recent CBM architectures work around this issue by obtaining annotations from Vision-Language Models (VLMs). While greatly broadening applicability, doing so can yield lower quality concepts and therefore less interpretable models. We strike for a middle ground by introducing Vision-plus-Human-guided CBM (VH-CBM), a hybrid approach that exploits both VLMs and a small amount of dense annotations. VH-CBM employs a Gaussian Process in the VLM's embedding space, which captures useful global information about the target domain, to propagate the expert's supervision to any target data point. Our empirical evaluation shows how VH-CBM predicts more accurate concepts than VLM-guided CBMs even when annotating as little as 1% of the data, while sporting better concept calibration and supporting active learning. --- ### Hybrid Quantum-MambaVision: A Quantum-enhanced State Space Model for Calibrated Mixed-type Wafer Defect Detection Authors: Satwik Sai Prakash Sahoo, Jyoti Prakash Sahoo, Ting Wang, Subrota Kumar Mondal Published: 2026-05-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.16404v1 Extracting actionable knowledge from industrial visual data is fundamentally bottlenecked by extreme class imbalance and the prohibitive computational complexity of modern foundation models. In semi-conductor manufacturing, identifying multi-label wafer defects is a complex spatial data mining task where overlapping patterns obscure critical root-cause signals. While Vision Transformers (ViTs) excel at global dependency extraction, their quadratic scaling renders them inefficient for high-throughput, real-time anomaly detection. To overcome these computational barriers, this paper introduces Hybrid Quantum-MambaVision, a highly efficient architecture tailored for spatial knowledge discovery. We integrate a linear-complexity State-Space Model (SSM) backbone with a Parameterized Quantum Context Adapter (QCA) and Low-Rank Adaptation (LoRA). The Mamba backbone efficiently captures long-range spatial dependencies, while the quantum adapter maps compressed latent features into a high-dimensional Hilbert space to disentangle complex, overlapping signatures. On the highly imbalanced MixedWM38 dataset, Hybrid Quantum-MambaVision achieves exceptional multi-label classification performance, significantly reducing the error rate on complex multi-defect topologies compared to classical baselines. The quantum regularizer acts as a profound uncertainty calibrator, substantially reducing Maximum Calibration Error (MCE) and minimizing expected false-positive costs. This work establishes a scalable Quantum-Classical hybrid paradigm for efficient representation learning in industrial data mining. --- ### When Vision Speaks for Sound Authors: Xiaofei Wen, Wenjie Jacky Mo, Xingyu Fu, Rui Cai, Tinghui Zhu, Wendi Li, Yanan Xie, Muhao Chen, Peng Qi Published: 2026-05-13 Categories: cs.CV, cs.SD arXiv: https://arxiv.org/abs/2605.16403v1 Despite rapid progress in video-capable MLLMs, we find that their apparent audio understanding in videos is often vision-driven: models rely on visual cues to infer or hallucinate acoustic information, rather than verifying the audio stream. This issue appears across both state-of-the-art open-source omni models and leading closed-source models from providers such as Google and OpenAI. We characterize this failure mode as an audio-visual Clever Hans effect, in which models appear (falsely) audio-grounded, but actually exploit visual-acoustic correlations without verifying whether the audio and visual streams are truly aligned. To systematically study this behavior, we introduce Thud, an intervention-driven probing framework based on three counterfactual audio edits: Shift, which tests temporal synchronization; Mute, which tests sound existence; and Swap, which tests audio-visual consistency. Beyond diagnosis, we further study a two-stage alignment recipe: intervention-derived preference pairs teach audio verification, while event-level general video preferences regularize the model against over-specialization. Our best 10K-sample recipe improves average performance across the three intervention dimensions by 28 percentage points, while slightly improving performance on general video and audio-visual QA benchmarks. --- ### WinDeskGround: A Benchmark for Robust GUI Grounding in Complex Multi-Window Desktop Environments Authors: Haoren Zhao, Tianyi Chen, Zhen Wang Published: 2026-05-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.16402v1 Multimodal Large Language Models (MLLMs) have revolutionized GUI automation, yet their efficacy is largely established on idealized, single-layer interfaces. This paper identifies a critical reliability gap: state-of-the-art agents face distinct robustness challenges in real-world desktop environments characterized by multi-window stacking, occlusion, and visual clutter. To address this, we introduce WinDeskGround, a novel benchmark and synthesis framework tailored for evaluating GUI grounding robustness. Unlike static datasets, our framework parametrically generates complex desktop scenarios by controlling window occlusion, layout density, and semantic similarity, thereby simulating the distribution shifts of authentic workflows. We construct a diverse meta-dataset of 1,356 high-fidelity instruction-target pairs and conduct comprehensive evaluations of five leading MLLMs. Our results demonstrate that while top-tier agents excel in simplified settings, their accuracy declines under partial occlusion. WinDeskGround provides a valuable benchmark to facilitate the assessment and advancement of GUI agent robustness in realistic environments. The code is available at https://github.com/ZZZhr-1/WinDeskGround. --- ### FSCM: Frequency-Enhanced Spatial-Spectral Coupled Mamba for Infrared Hyperspectral Image Colorization Authors: Tingting Liu, Yuan Liu, Guiping Chen, Xiubao Sui, Qian Chen Published: 2026-05-13 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2605.15880v1 Thermal infrared imaging is robust to illumination variations and smoke interference, making it important for all-weather perception. However, the lack of natural color and fine texture limits target recognition, human visual interpretation, and the transfer of visible-light models. Existing infrared colorization methods mainly rely on single-band images, where insufficient spectral cues may lead to structural distortion and semantic confusion. Although infrared hyperspectral images provide rich spectral responses and material information, existing single-band frameworks remain limited in modeling spatial-spectral coupling and weak texture details. To address these issues, this paper presents FSCM, a spectral-information-guided GAN framework. Within FSCM, a frequency-enhanced spatial-spectral state-space generator composed of cascaded FSB units is constructed. Each FSB integrates three complementary components: state-space modeling captures global spatial-spectral dependencies; the frequency enhancement module (FEM) combines multi-level wavelet decomposition and Fourier gating to recover structural contours, directional high-frequency details, and global frequency responses; and the dual-stream hybrid gating module (DGM) integrates deformation-aware sampling with sparse attention to enhance effective local structures and suppress background interference. Additionally, an online semantic segmentation-guided loss is introduced to constrain the generated results, improving semantic consistency in complex road scenes. Experiments show that FSCM outperforms existing infrared colorization methods in visual quality and semantic fidelity. --- ### Learning Selective Merge Policies for Deadline-Constrained Coded Caching via Deep Reinforcement Learning Authors: Amirhossein Yousefiramandi Published: 2026-05-13 Categories: cs.IT, cs.AI, cs.NI arXiv: https://arxiv.org/abs/2605.15236v2 In the coded caching, the server uses the cached information at the users to serve multiple users in parallel with a single coded multi-casting message or packet, that is, a merged packet, and thus mitigates the peak network congestion. In order to deliver the timely messages to the users in the deadline-driven applications like the video streaming, we must determine online the messages to be merged for the delivery, as there is a time limit for each request. It is important to note that while the merging aids the current coded multi-casting packet, it could harm the future deliveries. Our solution employs the deep reinforcement learning to view the coded multi-casting delivery as a masked action-discrete state control problem, and our policy network, trained via the proximal policy optimization, performs better than SACM++. On the uniform-demand benchmark, our policy network reduces the broadcast-packet expiration ratio $ρ$ by $40.9\%$ ($0.208$ vs.\ $0.352$) with respect to the best coded multi-casting baseline (SACM++), while also attaining the best broadcast-efficiency score $σ$ across the Track~A battery among the coded multi-casting methods. One noteworthy phenomenon here is that, for the applications with stricter deadlines, the merging becomes selective instead of aggressive, since the policy network selectively merges at approximately $31.8\%$ of the chances, even though the same observation holds across the variations within the same simulator family. The focus of our design is on the efficient pairwise XOR merging, where the higher-order ($K{\ge}3$) coding can be considered as a natural generalization left for future work. --- ### MuteBench: Modality Unavailability Tolerance Evaluation for Incomplete Multimodal Fusion Authors: Wugeng Zheng, Ziwen Kan, Tianlong Chen, Chen Chen, Song Wang Published: 2026-05-13 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.15235v1 Multimodal physiological data powers clinical AI systems from intensive care units to wearable devices, but sensors routinely fail in practice. Two failure modes are common: modality missing, where an entire channel is absent, and within-modality missing, where a contiguous time segment is lost. No existing benchmark evaluates multiple fusion architectures under both failure modes at controlled severity levels across diverse clinical datasets. We present MuteBench, a benchmark covering 9 datasets from 7 clinical domains, 6 fusion architectures, and 2 missing-data modes over 125,000 samples. Through this benchmark, we find that architecture family is the strongest predictor of robustness, outweighing parameter count. Channel-independent models tolerate modality missing well but can be sensitive to within-modality missing, especially on short sequences. Curriculum modality dropout protects reliably only up to the maximum dropout rate used in training. We also find that channel count, sequence length, and modality alignment jointly determine which failure mode poses the greater threat. Finally, a PTB-XL case study suggests that diffusion-based imputation can improve downstream classification under within-modality missing, with the largest gains for models whose expert routing is most sensitive to corrupted inputs, though broader validation across datasets remains an open direction. MuteBench provides practitioners with concrete guidance for both selecting existing architectures and informing the design of future robust multimodal fusion methods. --- ### Mask-Morph Graph U-Net: A Generalisable Mesh-Based Surrogate for Crashworthiness Field Prediction under Large Geometric Variation Authors: Haoran Li, Tobias Lehrer, Yingxue Zhao, Haosu Zhou, Philipp Stocker, Tobias Pfaff, Marcus Wagner, Nan Li Published: 2026-05-13 Categories: cs.LG, cs.CV arXiv: https://arxiv.org/abs/2605.15231v2 Nonlinear finite element crash simulations are accurate but computationally expensive, limiting their use in iterative design optimisation. Machine-learning surrogate models based on graph neural networks (GNNs) offer a faster alternative. Message-passing GNNs are widely used for mesh simulation, and their shared node and edge update functions are relatively generalisable across varying graph structures. By contrast, non-shareable edge-specific aggregation layers can capture nonlinear relationships more accurately but usually require fixed graph connectivity, which limits generalisability. This paper presents Mask-Morph Graph U-Net (MMGUNet), a practical approach to addressing the limitation of hierarchical Graph U-Net architectures that use edge-specific downsampling and upsampling layers. Fixed coarse graph connectivity is required for edge-specific layers. To retain this while improving spatial correspondence, the proposed method morphs the coarsened graph hierarchy to each input mesh using feature-aligned barycentric parameterisation before constructing cross-graph edges. It further applies node masking during supervised pretraining, followed by parameter-efficient fine-tuning in which high-parameter edge-specific layers are frozen. The proposed approach is evaluated in in-distribution, out-of-distribution, and cross-component transfer settings using mean Euclidean distance and maximum intrusion percentage error. Results show that coarse-graph morphing improves test accuracy relative to a fixed-coarse-graph baseline, while masked supervised pretraining reduces the train-test discrepancy and improves data efficiency during transfer. The proposed model also achieves lower prediction error compared with external baselines. These results demonstrate a practical route toward reusable, data-efficient mesh-based surrogate modelling for crashworthiness design exploration. --- ### PBT-Bench: Benchmarking AI Agents on Property-Based Testing Authors: Lucas Jing, Xinqi Wang, Liao Zhang, Simon S. Du Published: 2026-05-13 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2605.15229v3 Existing code benchmarks measure whether an agent can produce any test that reproduces a known bug, or whether it can produce a patch that fixes a described issue. Neither isolates the distinct skill of property-based testing: deriving a semantic invariant from documentation, and then constructing an input-generation strategy precise enough to make a random search reveal the violation. We introduce PBT-Bench, a benchmark of 100 curated property-based testing problems across 40 real Python libraries. Each problem injects one or more semantic bugs (365 in total, mean 3.65 per problem) designed so that default-strategy random inputs almost never trigger them; the agent must read the library's documentation, identify the relevant invariant, and specify a Hypothesis @given strategy that concentrates mass in the trigger region. Bugs are stratified across three difficulty levels (L1-L3) spanning single-constraint boundary bugs to stateful, cross-function protocol violations. We evaluate eight contemporary LLMs under two prompting regimes (open-ended baseline vs. explicit Hypothesis scaffolding) for three independent runs per configuration. Bug recall under the PBT-guided prompt ranges from 42.1% to 83.4% across models; under the open-ended baseline, from 31.4% to 76.7%. Hypothesis scaffolding lifts mid-capability models by over 20 percentage points, but yields smaller gains for the strongest models, with two exceptions showing degradation, suggesting the structured prompt can interfere with certain model behaviours rather than complementing them. The hardest bugs prove model-specific: different architectures fail on different problems, leaving persistent gaps that no single model closes. We release the benchmark, harness, and full evaluation corpus to support downstream work on documentation-grounded semantic reasoning. --- ### Verifiable Agentic Infrastructure: Proof-Derived Authorization for Sovereign AI Systems Authors: Jun He, Deying Yu Published: 2026-05-13 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2605.15228v1 Modern cloud and enterprise systems rely on identity-centric authorization, assuming that callers possessing valid credentials are safe to execute commands. The emergence of autonomous AI agents invalidates this assumption: agents can generate syntactically valid but semantically unsafe actions, making standing privileges a significant operational risk. This risk becomes especially acute in sovereign AI systems, where autonomous agents may interact with cloud infrastructure, regulated data, financial workflows, and national-scale digital services. Governed mutation substrates reduce this risk by interposing on agent actions: agents submit intents, infrastructure evaluates context and policy, and execution is mediated. However, this shifts the trust boundary: how can the decision to authorize an intent be made verifiable, distributed, and replayable? We introduce a Distributed Trust Framework (DTF), a verification framework for governed mutation systems that computes execution authority from structured, verifiable artifacts. DTF introduces a Justification Proof to encode the admissibility basis of an action, a consensus model for independent evaluation, an ephemeral Execution Identity derived from the approved proof, and an append-only Evidence Chain that preserves the authorization lifecycle. Under stated substrate assumptions, this architecture enforces a compact authorization invariant: no high-stakes execution without a proof object, no derived authority without consensus, and no valid mutation detached from evidence. We define the model, instantiate it over an OpenKedge-based governed mutation substrate, and show how it maps onto cloud-native environments. By shifting authorization from standing identity to proof-derived authority, DTF provides an infrastructure foundation for making agentic execution governable, auditable, and bounded in sovereign AI deployments. --- ### NIMO Controller: a self-driving laboratory orchestrator based on the Model Context Protocol Authors: Naruki Yoshikawa, Ryo Tamura Published: 2026-05-13 Categories: cs.AI, cond-mat.mtrl-sci, cs.RO arXiv: https://arxiv.org/abs/2605.15227v2 Self-driving laboratories (SDLs) are attracting increasing attention as a means of accelerating scientific discovery; however, developing SDL software remains technically demanding. To improve accessibility, orchestration software frameworks have been proposed to coordinate SDL components, but many existing frameworks are primarily designed for human interaction and lack standardized interfaces for direct integration with AI agents. In this work, we propose an SDL software architecture based on the Model Context Protocol (MCP), in which all SDL functionalities are exposed through MCP servers. Following this design principle, we introduce NIMO Controller, an MCP-based SDL orchestrator that integrates the experimental planning software NIMO. It provides a visual programming interface automatically generated through MCP-based tool discovery, allowing human users to design experimental workflows without writing code. The same MCP backend is also accessible to AI agents, providing a unified interface through which both human users and AI agents can interact with SDL components. AI agents can invoke MCP tools directly or generate NIMO workflows from natural language instructions. We evaluated the agentic workflow generation performance of various LLMs and conducted a user study using a color-matching SDL. --- ### NIMO Controller: a self-driving laboratory orchestrator based on the Model Context Protocol Authors: Naruki Yoshikawa, Ryo Tamura Published: 2026-05-13 Categories: cs.AI, cond-mat.mtrl-sci, cs.RO arXiv: https://arxiv.org/abs/2605.15227v1 Self-driving laboratories (SDLs) have attracted increasing attention as a means of accelerating scientific discovery; however, developing SDL software remains technically demanding. To improve accessibility, orchestration software frameworks have been proposed to coordinate SDL components. Nevertheless, existing frameworks are primarily designed for human interaction and do not provide standardized interfaces suitable for AI agents. In this work, we propose an SDL software architecture based on the Model Context Protocol (MCP), in which all SDL functionalities are exposed through MCP servers. Following this design principle, we introduce an MCP-based SDL orchestrator, named NIMO Controller. It provides a visual programming interface automatically generated through MCP-based tool discovery, allowing human users to design experimental workflows without writing code. The same MCP backend can also be accessed by AI agents, providing a unified interface for both human users and AI agents. We demonstrate the proposed system through a case study on a color-matching SDL. The results validate the usability of the proposed MCP-based SDL architecture. --- ### Is Agentic AI Ready for Real-World Hardware Engineering? A Deep Dive with Phoenix-bench Authors: Qingyun Zou, Feng Yu, Hongshi Tan, Bingsheng He, WengFai Wong Published: 2026-05-13 Categories: cs.AR, cs.AI, cs.SE arXiv: https://arxiv.org/abs/2605.15226v1 We ask whether agentic AI systems built for software engineering transfer to realistic hardware engineering. Existing hardware LLM benchmarks isolate sub-tasks but none jointly requires repository navigation, hierarchy-aware localization, Electronic Design Automation (EDA) executable verification, and maintenance-style patching. We introduce \textbf{Phoenix-bench}, a synchronized corpus of 511 verified Verilator instances from 114 GitHub repositories, each shipped with the developer patch, design-flow labels, fail-to-pass and pass-to-pass testbenches, and a Docker-pinned EDA environment so resolved-rate differences reflect agent behavior rather than toolchain availability. Using Phoenix-bench we run a uniform evaluation of four commercial agents and eight open-source agentic structures across four LLM backbones, plus two diagnostic interventions (file-level oracle localization and one round of testbench-log feedback). Three findings emerge. (i)~Software and hardware are fundamentally different engineering tasks: the same agent loses 37\% to 58\% from SWE-bench Verified to Phoenix-bench because hardware bugs propagate across parallel instantiated modules through signal flow rather than along a software-style call graph, and software-tuned agents stop at the symptom file instead of tracing back through the instantiation chain. (ii)~Failures concentrate on design control-flow / finite state machine (FSM) bugs, verification testbench bugs, and hard cases that demand cross-hierarchy signal-flow tracking and coordinated multi-file edits. (iii)~Localization granularity matters far more than localization itself: a perfect file-level oracle yields only $+1.4$\% because the agent then breaks files that did not need editing, while a single round of test case feedback lifts resolved rate by $42$\% to $45$\% because the test case tells \emph{where} the bug is and \emph{what} the fix has to look like. --- ### Do Biological Structural Guarantees Earn Their Complexity? Authors: Bogdan Banu Published: 2026-05-13 Categories: q-bio.QM, cs.AI arXiv: https://arxiv.org/abs/2605.15225v1 Biologically-inspired AI agent frameworks claim reliability benefits through structural guarantees adapted from gene regulatory networks, immune systems, and metabolic control. These claims are rarely tested empirically against simpler alternatives. We present three deep benchmarks: metabolic priority gating, autoinducer-based quorum sensing, and Bayesian stagnation detection, each comparing a biologically-grounded implementation against a naive non-biological alternative and an ablated control, across 1,000 trials per seed and 10 seeds (10M+ data points total). --- ### ICRL: Learning to Internalize Self-Critique with Reinforcement Learning Authors: Jianbo Lin, Xiaomin Yu, Yi Xin, Yifu Guo, Zhuosong Jiang, Zhongqi Yue, Weishi Wang, Heqing Zou, Chengwei Qin, Hui Xiong Published: 2026-05-13 Categories: cs.AI, cs.MA arXiv: https://arxiv.org/abs/2605.15224v1 Large language model-based agents make mistakes, yet critique can often guide the same model toward correct behavior. However, when critique is removed, the model may fail again on the same query, indicating that it has not internalized the critique's guidance into its underlying capability. Meanwhile, a frozen critic cannot improve its feedback quality over time, limiting the potential for iterative self-improvement. To address this, we propose learning to internalize self-critique with reinforcement learning(ICRL), a novel framework that jointly trains a solver and a critic from a shared backbone to convert critique-induced success into unassisted solver ability. The critic is rewarded based on the solver's subsequent performance gain, incentivizing actionable feedback. To address the distribution shift between critique-conditioned and critique-free behavior, ICRL introduces a distribution-calibration re-weighting ratio that selectively transfers critique-guided improvements compatible with the solver's own prompt distribution. Additionally, a role-wise group advantage estimation stabilizes joint optimization across the two roles. Together, these mechanisms ensure that the solver learns to improve itself without external critique, rather than becoming dependent on critique-conditioned behavior. We evaluate ICRL on diverse benchmarks spanning agentic and mathematical reasoning tasks, using Qwen3-4B and Qwen3-8B as backbones. Results show consistent improvements, with average gains of 6.4 points over GRPO on agentic tasks, and 7.0 points on mathematical reasoning. Notably, the learned 8B critic is comparable to 32B critics while using substantially fewer tokens. The code is available at https://github.com/brick-pid/ICRL. --- ### GenAI-Driven Approach to RISC-V Supply Chain Exploration Authors: Nenad Petrovic, Andre Schamschurko, Yingjie Xu, Alois Knoll Published: 2026-05-13 Categories: cs.AR, cs.AI arXiv: https://arxiv.org/abs/2605.15223v1 This paper presents an LLM-empowered workflow for RISC-V supply chain analysis, integrating Vision-Language Models (VLMs) and Model-Driven Engineering (MDE) to enable comprehensive, multimodal data-driven insights. The proposed approach addresses the challenges of heterogeneous and unstructured supply chain data by leveraging LLMs for textual understanding and VLMs for extracting information from visual artifacts such as diagrams, tables, and scanned documents. These models collaboratively identify key entities and relationships, which are then organized into a knowledge graph representing supply chain components and their interdependencies. For analytical reasoning, the workflow incorporates MDE techniques and constraint-based modeling to enable formal validation of dependencies, detection of bottlenecks, and assessment of risks. The synergy between LLM- and VLM-based semantic understanding and MDE-based formal analysis supports both exploratory and systematic evaluation of supply chain resilience. A human-in-the-loop mechanism further enables interactive querying and expert validation. The approach is evaluated in RISC-V ecosystem scenarios, demonstrating its effectiveness in generating actionable insights, enhancing transparency, and supporting decision-making in complex semiconductor supply chains. --- ### PerfCodeBench: Benchmarking LLMs for System-Level High-Performance Code Optimization Authors: Huihao Jing, Wenbin Hu, Shaojin Chen, Haochen Shi, Hanyu Yang, Sirui Zhang, Haoran Li, Yangqiu Song Published: 2026-05-13 Categories: cs.SE, cs.CL, cs.PL arXiv: https://arxiv.org/abs/2605.15222v2 Large language models (LLMs) can often generate functionally correct code, but their ability to produce efficient implementations for performance-critical systems tasks remains limited. Existing code benchmarks mainly emphasize correctness or algorithmic problem solving, while realistic systems-level optimization is still underexplored. To address this gap, we introduce PerfCodeBench, an executable benchmark for evaluating LLMs on high-performance code optimization. The tasks require system-level implementation choices, hardware-aware optimization, and careful handling of performance bottlenecks. Each task includes executable correctness checks, a baseline implementation, and a reference optimized solution. This allows us to evaluate both correctness and runtime-oriented efficiency. Our evaluation on a broad set of state-of-the-art LLMs shows a clear gap between model-generated code and expert-optimized implementations. The gap is especially large on tasks involving parallelism and GPU operations. Current models also show weaknesses in cross-language robustness and in consistently reaching expert-level efficiency. These results suggest that performance-aware evaluation are still needed. LLMs should move beyond generating merely correct code toward producing efficient systems software. We submit the benchmark data, evaluation infrastructure, and complete logs of all LLMs-generated code at https://anonymous.4open.science/r/perfcodebench-7CDE. --- ### Effective Harness Engineering for Algorithm Discovery with Coding Agents Authors: Yoichi Ishibashi, Taro Yano, Masafumi Oyamada Published: 2026-05-13 Categories: cs.SE, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2605.15221v1 AlphaEvolve and FunSearch have demonstrated the potential of combining large language models (LLMs) with evolutionary search for automated algorithm discovery. However, discovery success is shaped not only by model capability but also significantly by the design of the execution infrastructure, i.e., the harness. This paper investigates effective harness design through three questions: under a fixed token budget, is it better to produce many algorithms with brief thought or fewer algorithms with deeper thought? How should the harness handle evaluation hacks, where generated programs exploit the scoring function? And how can agents that require full filesystem access execute safely in parallel? Using Vesper, an algorithm discovery framework that incorporates harness improvements addressing these questions, we evaluate on Circle Packing under the same token budget. Interestingly, generating fewer algorithms while thinking more deeply about each one achieved higher scores. That is, scaling the quality of each individual is more budget-efficient than scaling the number of evolutionary generations. Surprisingly, more capable models produced evaluation hacks at higher rates, making hack detection increasingly necessary as models scale. --- ### Always Learning, Always Mixing: Efficient and Simple Data Mixing All The Time Authors: Michael Y. Hu, Apurva Gandhi, Kyunghyun Cho, Tal Linzen, Pratyusha Sharma Published: 2026-05-13 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2605.15220v1 Data mixing decides how to combine different sources or types of data and is a consequential problem throughout language model training. In pretraining, data composition is a key determinant of model quality; in continual learning and adaptation, it governs what is retained and acquired. Yet existing data mixing methods address only one phase of this lifecycle at a time: some require smaller proxy models tied to a single training phase, others assume a fixed domain set, and continual learning lacks principled guidance altogether. We argue that data mixing is fundamentally an online decision making problem -- one that recurs throughout training and demands a single, unified solution. We introduce OP-Mix (On-Policy Mix), a data mixing algorithm that operates across the entire language model training lifecycle. Our main insight is that candidate data mixtures can be cheaply simulated by interpolating between low-rank adapters trained directly on the current model, eliminating separate proxy models and ensuring the search is always grounded in the model's actual learning dynamics. Across pretraining, continual midtraining, and continual instruction tuning, OP-Mix consistently finds near-optimal mixtures while using a fraction of the compute of the baselines. In pretraining, OP-Mix improves upon training without mixing by 6.3% in average perplexity. For continual learning, OP-Mix matches the performance of both retraining and on-policy distillation while using 66% and 95% less overall compute, respectively. OP-Mix suggests a different view of language model training: not a sequence of distinct phases, but a single continuous process of learning from data. --- ### LLM-Based Robustness Testing of Microservice Applications: An Empirical Study Authors: Hrushitha Goud Tigulla, Marco Vieira Published: 2026-05-13 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2605.14202v1 Malformed, missing, or boundary-value inputs in microservice APIs can cascade across dependent services, threatening reliability. Robustness testing systematically exercises such inputs to expose server-side failures, but generating diverse, effective tests remains challenging. Large Language Models can generate such tests from API specifications; however, it is unknown whether different models and prompt strategies produce diverse failure sets or converge on the same failures. We report a controlled experiment applying 7 prompt strategies to 3 open-source LLMs (14B-70B parameters) targeting 2 architecturally distinct microservice systems: one Java monolingual (6 services, 9 failure modes) and one polyglot (27 services, 14 failure modes), yielding 38 valid runs and 663 generated tests. We find that prompt strategy explains more variation in diversity than model size: a Structured prompt collapses diversity entirely, while a single model varied across three prompt strategies achieves complete failure-mode coverage on one system, outperforming any multi-model ensemble under a fixed prompt. We introduce two strategies, Guided and GuidedFewShot, that embed a mutation taxonomy from prior robustness testing research as domain context. GuidedFewShot achieves the highest single-run coverage on both systems (5 of 9 and 8 of 14 failure modes) while maintaining low cross-model similarity. A key lesson is that taxonomy rules alone are insufficient: LLMs cannot distinguish key-absent from value-empty mutations without concrete examples. Findings replicate across both systems. --- ### MAPLE: Latent Multi-Agent Play for End-to-End Autonomous Driving Authors: Rajeev Yasarla, Deepti Hegde, Hsin-Pai Cheng, Shizhong Han, Yunxiao Shi, Meysam Sadeghigooghari, Hanno Ackermann, Litian Liu, Pranav Desai, Fatih Porikli, Mohammad Ghavamzadeh, Hong Cai Published: 2026-05-13 Categories: cs.RO, cs.CV arXiv: https://arxiv.org/abs/2605.14201v2 Vision-language-action (VLA) models are effective as end-to-end motion planners, but can be brittle when evaluated in closed-loop settings due to being trained under traditional imitation learning framework. Existing closed-loop supervision approaches lack scalability and fail to completely model a reactive environment. We propose MAPLE, a novel framework for reactive, multi-agent rollout of a dynamic driving scenario in the latent space of the VLA model. The ego vehicle and nearby traffic agents are independently controlled over multi-step horizons, while being reactive to other agents in the scene, enabling closed-loop training. MAPLE consists of two training stages: (1) supervised fine-tuning on the latent rollouts based on ground-truth trajectories, followed by (2) reinforcement learning with global and agent -specific rewards that encourage safety, progress, and interaction realism. We further propose diversity rewards that encourage the model to generate planning behaviors that may not be present in logged driving data. Notably, our closed-loop training framework is scalable and does not require external simulators, which can be computationally expensive to run and have limited visual fidelity to the real-world. MAPLE achieves state-of-the-art driving performance on Bench2Drive and demonstrates scalable, closed-loop multi-agent play for robust E2E autonomous driving systems. --- ### How to Scale Mixture-of-Experts: From muP to the Maximally Scale-Stable Parameterization Authors: Leena Chennuru Vankadara, Moritz Haas, Luke Hayward, Sebastian Bordt, Alessandro Breccia Published: 2026-05-13 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2605.14200v1 Recent frontier large language models predominantly rely on Mixture-of-Experts (MoE) architectures. Despite empirical progress, there is still no principled understanding of how hyperparameters should scale with network width $N$, expert width $N_e$, number of experts $M$, sparsity $K$, and depth $L$ to ensure both stability and optimal performance at scale. We take a principled step toward resolving this gap by analyzing three different scaling regimes: (I) co-scaling $N\asymp N_e$, (II) co-scaling $N\asymp M\asymp K$, and (III) full proportional scaling of $N, N_e, M$, and $K$. For each regime, we develop a novel Dynamical Mean Field Theory (DMFT) description of the limiting training dynamics of MoEs that provides a formal foundation for our analysis. Within this framework, we derive the unique parameterization for SGD and Adam satisfying all maximal-update ($μ$) desiderata. We then show that the resulting $μ$P prescription does not reliably induce monotonic improvement with scale or robust learning-rate transfer. We trace these pathologies to scale-dependent observables in the aggregation dynamics, which motivates a refined set of desiderata that we term maximal scale stability. Guided by this principle, we derive a Maximally Scale-Stable Parameterization (MSSP) for both SGD and Adam in all three scaling regimes, and characterize the corresponding limiting dynamics - qualitatively distinct from the $μ$P limit - through a separate DMFT analysis. Experiments verify that MSSP robustly recovers learning rate transfer and monotonic improvement with scale across regimes. Combined with existing depth-scaling theory, these results provide a complete scaling prescription for MoE architectures as a function of width, depth, expert width, and number of experts. --- ### Stochastic Matching via Local Sparsification Authors: Sara Ahmadian, Edith Cohen, Mohammad Roghani Published: 2026-05-13 Categories: cs.DS, cs.LG arXiv: https://arxiv.org/abs/2605.14195v1 The classic online stochastic matching problem typically requires immediate and irrevocable matching decisions. However, in many modern decentralized systems such as real-time ride-hailing and distributed cloud computing, the primary bottleneck is often local communication bandwidth rather than the timing of the match itself. We formalize this challenge by introducing a two-stage local sparsification framework. In this setting, arriving requests must prune their realized compatibility sets to a strict budget of $k$ edges before a central coordinator optimizes the global matching. This creates a "middle ground" between local information constraints and global optimization utility. We propose a local selection strategy, parametrized by a fractional solution of the expected instance. Theoretically, we quantify the approximation ratio as a function of the solution's {\em spread}. We prove that under sufficient spread, our sparsifier globally preserves the expected size of the maximum matching. Empirically, we demonstrate the robustness of our approach using the New York City ride-hailing datasets and adversarial synthetic benchmarks. Our results show that near-optimal global matching is achievable even with highly constrained local budgets, significantly outperforming standard online baselines. --- ### GradShield: Alignment Preserving Finetuning Authors: Zhanhao Hu, Xiao Huang, Patrick Mendoza, Emad A. Alghamdi, Basel Alomair, Raluca Ada Popa, David Wagner Published: 2026-05-13 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.14194v2 Large Language Models (LLMs) pose a significant risk of safety misalignment after finetuning, as models can be compromised by both explicitly and implicitly harmful data. Even some seemingly benign data can inadvertently steer a model towards misaligned behaviors. To address this, we introduce GradShield, a principled filtering method that safeguards LLMs during finetuning by identifying and removing harmful data points before they corrupt the model's alignment. It removes potentially harmful data by computing a Finetuning Implicit Harmfulness Score (FIHS) for each data point and employs an adaptive thresholding algorithm. We apply GradShield to multiple utility fine-tuning tasks across varying levels of harmful data and evaluate the safety and utility performance of the resulting LLMs using various metrics. The results show that GradShield outperforms all baseline methods, consistently maintaining an Attack Success Rate (ASR) below $6\%$ while preserving utility performance. --- ### Why Retrieval-Augmented Generation Fails: A Graph Perspective Authors: Kai Guo, Xinnan Dai, Zhibo Zhang, Nuohan Lin, Shenglai Zeng, Jie Ren, Haoyu Han, Jiliang Tang Published: 2026-05-13 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2605.14192v1 Retrieval-Augmented Generation (RAG) has become a powerful and widely used approach for improving large language models by grounding generation in retrieved evidence. However, RAG systems still produce incorrect answers in many cases. Why RAG fails despite having access to external information remains poorly understood. We present a model-internal study of retrieval-augmented generation that examines how retrieved evidence influences answer generation. Using circuit tracing, we construct attribution graphs that model the flow of information through transformer layers during decoding. These graphs represent interactions among retrieved context, intermediate model activations, and generated tokens, providing a graph, circuit-level view of how external evidence is integrated into the model's reasoning process across multiple question answering benchmarks, we observe consistent structural differences: correct predictions exhibit deeper reasoning paths, more distributed evidence flow, and a more structured pattern of local connectivity, while failed predictions show shallower, fragmented, and overly concentrated evidence flow. Building on these findings, we develop a graph-based error detection framework that uses attribution-graph topology features. Furthermore, we show that attribution graphs enable targeted interventions. By reinforcing question-constrained evidence grounding, we reshape internal routing so that answer generation remains guided by the question, leading to more effective integration of retrieved information and fewer errors. --- ### CoReDiT: Spatial Coherence-Guided Token Pruning and Reconstruction for Efficient Diffusion Transformers Authors: Zhuojin Li, Hsin-Pai Cheng, Hong Cai, Shizhong Han, Fatih Porikli Published: 2026-05-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.14191v1 Diffusion Transformers (DiTs) deliver remarkable image and video generation quality but incur high computational cost, limiting scalability and on-device deployment. We introduce CoReDiT, a structured token pruning framework for DiTs across vision tasks. CoReDiT uses a linear-time spatial coherence score to estimate local redundancy in the latent token lattice and skips high coherence (redundant) tokens in self-attention. To maintain a dense representation and avoid visual discontinuities, we reconstruct skipped attention outputs via coherence-guided aggregation of spatially neighboring retained tokens. We further introduce a progressive, block-adaptive pruning schedule that increases pruning gradually and allocates larger budgets to blocks and denoising steps with higher redundancy. Across state-of-the-art diffusion backbones including PixArt-α and MagicDrive-V2, CoReDiT achieves up to 55% self-attention FLOPs reduction and inference speedups of 1.33x on cloud GPUs and 1.72x on mobile NPUs, while maintaining high visual quality. Notably, CoReDiT also increases on-device memory head-room, enabling higher-resolution generation. --- ### QOuLiPo: What a quantum computer sees when it reads a book Authors: Christophe Jurczak Published: 2026-05-13 Categories: quant-ph, cs.CL, cs.DL, physics.atom-ph arXiv: https://arxiv.org/abs/2605.14188v1 What does a book look like to a quantum computer? This paper takes eight classical works of the Renaissance and its late-antique inheritance -- from Augustine to Galileo -- and runs each through a neutral-atom quantum processor. The bridge is graphs: each textual unit becomes an atom, and graph edges are physical blockade constraints for engineered exact unit-disk designs, or a 2D approximation to the semantic graph for natural texts. Three contributions follow. First, we introduce rigidity rho, a metric for how unique a book's structural backbone is -- distinguishing Marguerite de Navarre's Heptameron (rigid, twelve-nouvelle hard core) from Boethius (fully fungible, every chapter substitutable). Second, we invert the pipeline: rather than extracting a graph from existing prose, we pick a target graph the hardware encodes natively, and write a book whose structure matches it. The twenty-nine texts written this way, collected under the name QOuLiPo, extend the OuLiPo tradition to graph-topological constraints and, together with the eight natural texts, form a benchmark distribution against which neutral-atom hardware can be tracked as it scales. Third, we run both natural and engineered texts on Pasqal's FRESNEL processor up to one hundred atoms; engineered texts reach high approximation ratios, the cleanest instances returning the exact backbone. A cloud-accessible quantum machine plus an agentic coding environment now lets a single investigator run this pipeline end-to-end. What is reported is an application layer, not a speedup -- humanistic instances ready to load onto neutral-atom processors as they scale, already complementing classical text analysis. The Digital Humanities community has a stake in building familiarity with this hardware now: the engineered-corpus design choices made today fix the benchmark distribution future hardware will be measured against. --- ### LLMs Know When They Know, but Do Not Act on It: A Metacognitive Harness for Test-time Scaling Authors: Qi Cao, Yufan Wang, Peijia Qin, Shuhao Zhang, Pengtao Xie Published: 2026-05-13 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.14186v1 Large language models (LLMs) often expose useful signals of self-monitoring: before solving a problem, they can estimate whether they are likely to succeed, and after solving it, they can judge whether their answer is likely to be correct. However, these signals are typically measured or elicited in isolation, rather than used to control inference. In this work, we ask whether LLMs possess latent metacognitive ability that can be turned into effective test-time control. Inspired by the Nelson--Narens theory from cognitive psychology, we propose a metacognitive harness that separates monitoring from reasoning. For each problem, the model first reports a pre-solve feeling-of-knowing (FOK) signal; after each solve attempt, it reports a post-solve judgment-of-learning (JOL) signal. Rather than treating these signals as passive confidence estimates, the harness turns them into an explicit control interface for reasoning: it decides when to trust the current solution, when to retry with compact metacognitive feedback, and when to pass multiple attempts to a final aggregator. Across text, code, and multimodal reasoning benchmarks, our harness substantially improves a fixed Claude Sonnet-4.6 base model without parameter updates or benchmark-specific fine-tuning. On the evaluated public benchmark snapshots, it raises pooled accuracy from 48.3 to 56.9 and exceeds the strongest listed leaderboard entries on the three primary evaluation settings: HLE-Verified, LiveCodeBench v6, and R-Bench-V. These results suggest that strong LLMs may already possess useful metacognitive ability, but require an explicit control harness to act on it during reasoning. --- ### Thinking Ahead: Prospection-Guided Retrieval of Memory with Language Models Authors: Harshita Chopra, Krishna Kant Chintalapudi, Suman Nath, Ryen W. White, Chirag Shah Published: 2026-05-13 Categories: cs.IR, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2605.14177v1 Long-horizon personalization requires dialogue assistants to retrieve user-specific facts from extended interaction histories. In practice, many relevant facts often have low semanticsimilarity to the query under dense retrieval. Standard Retrieval-Augmented Generation (RAG) and GraphRAG systems are still largely retrospective: they rely on embedding similarity to the query or on fixed graph traversals, so they often miss facts that matter for the user's needs but lie far from the query in embedding space. Inspired by prospection, the human ability to use imagined futures as cues for recall, we introduce Prospection-Guided Retrieval (PGR), which decouples retrieval from how memories are stored. Given a user query, PGR first expands the goal into a short Tree-of-Thought (ToT) or linear chain of plausible next steps, and uses these steps as retrieval probes rather than relying on the original query alone. The facts retrieved by these probes are then used to personalize the next round of prospection, enabling PGR to uncover additional memories that become relevant only after the simulation is grounded in the user's history. We also introduce MemoryQuest, a challenging multi-session benchmark in which each query is annotated with 3--5 dated reference facts subject to a low query-reference similarity constraint. Across 1,625 queries spanning 185 user profiles from 3 publicly available datasets, PGR-TOT substantially improves retrieval, including nearly 3x recall on MemoryQuest over the strongest baseline. In pairwise LLM-as-judge comparisons against baselines, PGR-generated responses are preferred on 89--98% of queries, with blinded human annotations on held-out subsets showing the same trend. Overall, the results demonstrate that explicit prospection yields large gains in long-horizon retrieval and response quality relative to similarity-only baselines. --- ### Grounded Continuation: A Linear-Time Runtime Verifier for LLM Conversations Authors: Qisong He, Jinwei Hu, Xinmiao Huang, Changshun Wu, Yi Dong, Xiaowei Huang Published: 2026-05-13 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.14175v2 In a long conversation, an LLM can produce a plausible continuation that rests on premises the conversation has already abandoned. No runtime check ties its output to what the conversation has established, a gap that context-manipulation attacks on deployed agents exploit. We close this gap with a runtime verifier: an LLM Interpreter classifies each utterance into one of eight epistemic operations, and a symbolic engine applies them to a dependency map that records what every claim rests on and whether it still stands. Whether a continuation is grounded reduces to a walk over the map, linear in its size, with no LLM call. Retraction propagates through the same map with a conflict-free guarantee, flagging exactly the conclusions that lose support. On ReviseQA for belief revision and MemoryAgentBench's fact-consolidation split, two third-party benchmarks where earlier premises are superseded, the verifier leads a budget-matched retrieval baseline across five QA models and lifts MemoryAgentBench single-hop accuracy from 0.46--0.95 to 0.93--0.98. With the verifier, even the 7B model overtakes unaided GPT-4o. These runs feed the engine the benchmarks' own structured updates. When a GPT-4o Interpreter extracts every update from raw text instead, accuracy is statistically unchanged. Per-query cost is flat in conversation length, prompts staying near 0.8k tokens where full context reaches 114k and retraction queries under a microsecond at 2000 turns. --- ### Grounded Continuation: A Linear-Time Runtime Verifier for LLM Conversations Authors: Qisong He, Yi Dong, Xiaowei Huang Published: 2026-05-13 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.14175v1 In long conversations, an LLM can produce a next utterance that sounds plausible but rests on premises the conversation has already abandoned. Context-manipulation attacks against deployed agents now actively exploit this gap. We close it with a runtime verifier that maintains an explicit dependency graph: an LLM classifies each turn into one of 8 update operations drawn from four formalisms (dynamic epistemic logic, abductive reasoning, awareness logic, argumentation), and a symbolic engine records which claims depend on which evidence. Checking whether a continuation is supported reduces to a graph walk; retraction propagates through the same graph to flag exactly the conclusions that lose support, with linear per-turn cost and a formal conflict-free guarantee. On LongMemEval-KU oracle (n=78), the verifier reaches 89.7% accuracy vs. 88.5% for the LLM-only baseline (+1.3pp) and 87.2% for a transcript-RAG baseline matched on retrieval budget (+2.6pp); wins among disagreements are correct abstentions where the baseline confabulates. On LoCoMo's 60 official QA items the verifier is competitive with retrieval-augmented baselines. Beyond external benchmarks, we construct two multi-agent scenarios and a 50-item grounding test: on the 15-item stale-premise subset, the verifier reaches 100% accuracy vs. 93.3% (+6.7pp). These instantiate a soundness-faithfulness decomposition: the structural check is sound by construction, and per-deployment LLM extraction faithfulness is the empirical question we measure across four LLM families. The retraction check plateaus at microseconds while history-replay grows linearly with conversation length. --- ### CSI-JEPA: Towards Foundation Representations for Ubiquitous Sensing with Minimal Supervision Authors: Xuanhao Luo, Zhizhen Li, Yuchen Liu Published: 2026-05-13 Categories: cs.LG, cs.NI arXiv: https://arxiv.org/abs/2605.14171v1 Channel state information (CSI) provides a widely available sensing modality for human and environment perception, but existing CSI sensing models usually rely on task-specific supervised training and require substantial labeled data for each task, device, user, or environment. This limits their scalability in practical deployments where unlabeled CSI is abundant but labeled data is costly to collect. In this paper, we present CSI-JEPA, a self-supervised predictive representation learning framework for label-efficient, multi-task Wi-Fi sensing. CSI-JEPA learns reusable temporal-spectral representations from unlabeled CSI samples by predicting latent features of masked channel regions from visible context. To better match the physical structure of CSI, CSI-JEPA tokenizes channel-response amplitude windows along the time and subcarrier dimensions. It then introduces a channel variation-aware masking strategy that samples predictive targets from regions with stronger local temporal and subcarrier-domain variations. After pretraining, the encoder is frozen and used as a backbone, with lightweight task-specific adapters added for downstream sensing tasks. We evaluate CSI-JEPA on seven real-world Wi-Fi sensing tasks spanning diverse objectives and deployment settings. The results show that CSI-JEPA improves downstream sensing performance over competitive baselines, achieving up to 10.64 percentage points mean accuracy gain over state-of-the-art supervised Transformer and matched-budget label savings of up to 98.0%. --- ### BOOKMARKS: Efficient Active Storyline Memory for Role-playing Authors: Letian Peng, Ziche Liu, Yiming Huang, Longfei Yun, Kun Zhou, Yupeng Hou, Jingbo Shang Published: 2026-05-13 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.14169v2 Memory systems are critical for role-playing agents (RPAs) to maintain long-horizon consistency. However, existing RPA memory methods (e.g., profiling) mainly rely on incremental summarization, whose compression discards details which become inaccessible to subsequent grounding. To address this issue, we propose a search-based memory framework called \textbf{\underline{BOOKMARKS}} for \textbf{active grounding}, which retains access to the full preceding storyline and collects task-relevant information on demand. Since summarizing the preceding storyline anew for each grounding request incurs substantial redundant computation, BOOKMARKS introduces \textbf{passive updating} to reuse earlier search results as checkpoints. Each \textbf{bookmark} represents the \textbf{content} about a particular aspect (\textbf{section}) of story information at a specific synchronization \textbf{point}. For current task, BOOKMARKS searches for only useful contents, reuses existing semantically equivalent bookmarks or initializes new ones, and synchronizes the selected ones from their stored checkpoints to the current scene. A reused bookmark thus only needs to process the newly observed storyline suffix, avoiding repeated synchronization. We evaluate BOOKMARKS across six narrative artifacts involving 47 characters and 9,537 test cases against non-active grounding baselines, covering next-action prediction and challenging, human-authored reasoning questions from a mystery game. BOOKMARKS improves next-action fidelity in a five-dimensional evaluation (emotion, intent, causality, position, and content) and raises mystery-game reasoning accuracy from 39.44\% to 45.42\% over the strongest baseline. --- ### BOOKMARKS: Efficient Active Storyline Memory for Role-playing Authors: Letian Peng, Ziche Liu, Yiming Huang, Longfei Yun, Kun Zhou, Yupeng Hou, Jingbo Shang Published: 2026-05-13 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.14169v1 Memory systems are critical for role-playing agents (RPAs) to maintain long-horizon consistency. However, existing RPA memory methods (e.g., profiling) mainly rely on recurrent summarization, whose compression inevitably discards important details. To address this issue, we propose a search-based memory framework called BOOKMARKS, which actively initializes, maintains, and updates task-relevant pieces of bookmarks for the current task (e.g., character acting). A bookmark is structured as the answer to a question at a specific point in the storyline. For each current task, BOOKMARKS selects reusable existing bookmarks or initializes new ones (at storyline beginning) with useful questions. These bookmarks are then synchronized to the current story point, with their answers updated accordingly, so they can be efficiently reused in future grounding rounds. Compared with recurrent summarization, BOOKMARKS offers (1) active grounding for capturing task-specific details and (2) passive updating to avoid unnecessary computation. In implementation, BOOKMARKS supports concept, behavior, and state searches, each powered by an efficient synchronization method. BOOKMARKS significantly outperforms RPA memory baselines on 85 characters from 16 artifacts, demonstrating the effectiveness of search-based memory for RPAs. --- ### Finite Sample Bounds for Learning with Score Matching Authors: Devin Smedira, Abhijith Jayakumar, Sidhant Misra, Marc Vuffray, Andrey Y. Lokhov Published: 2026-05-13 Categories: cs.LG, cs.DS, stat.ML arXiv: https://arxiv.org/abs/2605.14168v1 Learning of continuous exponential family distributions with unbounded support remains an important area of research for both theory and applications in high-dimensional statistics. In recent years, score matching has become a widely used method for learning exponential families with continuous variables due to its computational ease when compared against maximum likelihood estimation. However, theoretical understanding of the statistical properties of score matching is still lacking. In this work, we provide a non-asymptotic sample complexity analysis for learning the structure of exponential families of polynomials with score matching. The derived sample bounds show a polynomial dependence on the model dimension. These bounds are the first of its kind, as all prior work has shown only asymptotic bounds on the sample complexity. --- ### The Evaluation Trap: Benchmark Design as Theoretical Commitment Authors: Theodore J Kalaitzidis Published: 2026-05-13 Categories: cs.AI, cs.CY arXiv: https://arxiv.org/abs/2605.14167v1 Every AI benchmark operationalizes theoretical assumptions about the capability it claims to assess. When assumptions function as unexamined commitments, benchmarks stabilize the dominant paradigm by narrowing what counts as progress. Over time, narrow evaluation reorganizes capability concepts: architectures and definitions are selected for benchmark legibility until evaluation ceases to track an independent object and instead produces a version of the target defined by its own operational assumptions. The result is a trap: evaluation frameworks treat self-reinforcing assessments as valid, both creating and obscuring structural limits on what the current paradigm can accomplish. We introduce Epistematics, a methodology for deriving evaluation criteria directly from technical capability claims and auditing whether proposed benchmarks can discriminate the claimed capability from proxy behaviors. The contribution is meta-evaluative: an audit procedure, a failure mode taxonomy, and benchmark-design criteria for evaluating capability-evaluation coherence. We demonstrate the procedure through a worked audit of Dupoux et al. (2026), a proposal that revises the dominant paradigm's theoretical assumptions at the architectural level while reproducing them in its evaluation criteria, thereby entrenching the constraint it seeks to overcome in a form the evaluation cannot detect. --- ### You Only Landmark Once: Lightweight U-Net Face Super Resolution with YOLO-World Landmark Heatmaps Authors: Riccardo Carraro, Anna Briotto, Endi Hysa, Marco Fiorucci, Lamberto Ballan Published: 2026-05-13 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.14166v2 Face image super-resolution aims to recover high-resolution facial images from severely degraded inputs. Under extreme upscaling factors, fine facial details are often lost, making accurate reconstruction challenging. Existing methods typically rely on heavy network architectures, adversarial training schemes, or separate alignment networks, increasing model complexity and computational cost. To address these issues, we propose a lightweight U-Net based-architecture designed to reconstructs $128{ \times }128$ facial images from severely degraded $16{ \times }16$ inputs, achieving an $8 \times $ magnification. A key contribution is a novel auxiliary-training-free supervision strategy that leverages heatmaps generated by YOLO-World, an open-vocabulary object detector, to localize key facial features such as eyes, nose, and mouth. These heatmaps are converted into spatial weights to form a heatmap-guided loss that emphasizes reconstruction errors in semantically important regions. Unlike prior methods that require dedicated landmark or alignment networks, our approach directly reuses detector outputs as supervision, maintaining an efficient training and inference pipeline. Experiments on the aligned CelebA dataset demonstrate that the proposed loss consistently improves quantitative metrics and produces sharper, more realistic reconstructions. Overall, our results show that lightweight networks can effectively exploit detection-driven priors for perceptually convincing extreme upscaling, without adversarial training or increased computational cost. --- ### Unsteady Metrics and Benchmarking Cultures of AI Model Builders Authors: Stefan Baack, Christo Buschek, Maty Bohacek Published: 2026-05-13 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.14164v1 The primary way to establish and compare competencies in foundation and generative AI models has shifted from peer-reviewed literature to press releases and company blog posts, where model builders highlight results on selected benchmarks. These artifacts now largely define the state of the art for researchers and the public. Despite their prominence, which benchmarks model builders choose to highlight, and what they communicate through this selection, is underexamined. To investigate, we introduce and open-source Benchmarking-Cultures-25, a dataset of 231 benchmarks highlighted across 139 model releases in 2025 from 11 major AI builders, alongside an interactive tool to explore the data. Our analysis reveals a fragmented evaluation landscape with limited cross-model comparability: 63.2% of highlighted benchmarks are used by a single builder, and 38.5% appear in just one release. Few achieve widespread use (e.g., GPQA Diamond, LiveCodeBench, AIME 2025). Moreover, benchmarks are attributed different competencies by different builders, depending on their narrative. To disentangle these conflicting presentations, we develop a unified taxonomy mapping diverging terminology to a shared framework of measured signals based on what benchmark authors claim to measure. "General knowledge application" is the second most popular, yet vaguely defined, category. Qualitative analysis shows many such benchmarks deemphasize construct validity, instead framing results as indicators of progress toward AGI. Their authors claim to measure knowledge or reasoning broadly, yet mostly evaluate STEM subjects (especially math). We argue that highlighted benchmarks function less as standardized measurement tools and more as flexible narrative devices prioritizing market positioning over scientific evaluation. Data: https://hf.co/datasets/matybohacek/benchmarking-cultures-25; tool: https://bench-cultures.net. --- ### Agentic Systems as Boosting Weak Reasoning Models Authors: Varun Sunkaraneni, Pierfrancesco Beneventano, Riccardo Neumarker, Tomaso Poggio, Tomer Galanti Published: 2026-05-13 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.14163v1 Can a committee of weak reasoning-model calls reach the performance of much stronger models? We study verifier-backed committee search as inference-time boosting for reasoning language models. The mechanism is not simply that ``more agents help'': samples expose latent correct solutions, while critics and comparators must recover them without access to the hidden verifier. We formalize this view by separating proposal coverage, local identifiability, progress, and diversity. We prove that coverage can be amplified by repeated sampling, but cannot by itself create useful critics or comparators; reliable amplification requires an additional local soundness signal, such as execution, proof checking, type checking, tests, or constraint solving. We give rank-based bounds showing when local selection errors compose into reliable trajectories, and characterize the proposer-side ceiling: oracle best-of-\(k\) converges only to the mass of task slices on which the proposal system assigns nonzero useful probability. Empirically, on SWE-bench Verified, a single \texttt{GPT-5.4 nano} proposal solves \(67.0\%\) of tasks. Using the same nano model, our critic--comparator orchestration reaches \(76.4\%\) with \(k=8\) proposals, matching the standalone performance of \texttt{Gemini 3 Pro} and \texttt{Claude Opus 4.5} Thinking and approaching the \(79.0\%\) oracle best-of-\(8\) upper bound. Thus, many correct patches are already present in weak-model proposal pools; the main challenge is selecting them. The remaining failures are mostly proposal-coverage failures, indicating shared blind spots that stronger selection alone cannot close. --- ### Uncovering Trajectory and Topological Signatures in Multimodal Pediatric Sleep Embeddings Authors: Scott Ye, Harlin Lee Published: 2026-05-13 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.14156v1 While generative models have shown promise in pediatric sleep analysis, the latent structure of their multimodal embeddings remains poorly understood. This work investigates session-wide diagnostic information contained in the sequences of 30-second pediatric PSG epochs embedded by a multimodal masked autoencoder. We test whether augmenting embeddings with PHATE-derived per-epoch coordinates and whole-night movement descriptors, persistent homology summaries of the embedding cloud, and EHR yields task-relevant signals. Simple linear and MLP models, chosen for interpretability rather than state-of-the-art performance, show that geometric, topological, and clinical features each provide complementary gains. For binary predictions, feature importance is task-dependent, and more expressive late-fusion models generally perform better, with AUPRC improving from 0.26 to 0.34 for desaturation, 0.31 to 0.48 for EEG arousal, 0.09 to 0.22 for hypopnea, and 0.05 to 0.14 for apnea. We also report Brier score and Expected Calibration Error, where the full fusion model yields the best calibration across all four binary tasks. Our study reveals that latent geometry/topology and EHR offer complementary, interpretable signals beyond embeddings, improving calibration and robustness under extreme imbalance. --- ### ExploitBench: A Capability Ladder Benchmark for LLM Cybersecurity Agents Authors: Seunghyun Lee, David Brumley Published: 2026-05-13 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2605.14153v1 Exploitation is not a binary event. It is a ladder of acquiring progressive capabilities, from executing a single buggy line of code to taking full control of the target. However, existing LLM security benchmarks treat a crash as exploitation success. That single binary outcome collapses the hard parts of exploitation: the transition from triggering a bug to constructing reusable primitives and control. We present ExploitBench, a capability-graded benchmark that decomposes exploitation into 16 measurable flags, from coverage and crash through sandbox primitives, arbitrary read/write, control-flow hijack, and arbitrary code execution. Each capability is verified by a deterministic oracle that uses a per-run randomized challenge-response for primitives, differential execution against ground-truth binaries to measure progress, and a signal-handler proof for code execution. We instantiate ExploitBench on 41 V8 bugs because V8 is both widely deployed and exploitation-hardened. We report three arms: <model,env> as the primary measurement of model-environment capability, <model,env, adaptive coaching> as a secondary arm that adds adaptive coaching to test whether targeted feedback shifts outcomes, and <model,env,harness> as an ablation that swaps in the model's native CLI to check whether vendor-side optimizations increase exploitation capabilities. Our results show a sharp capability split between publicly deployed frontier models and the private frontier. Across the 8 publicly deployed models tested, reaching the vulnerable code and triggering a crash is routine, but arbitrary code execution is not. The private model shows arbitrary code execution on approximately half. Overall, results suggest that exploit construction against hardened targets is an emerging frontier capability. --- ### ROK-FORTRESS: Measuring the Effect of Geopolitical Transcreation for National Security and Public Safety Authors: Michael S. Lee, Yash Maurya, Drew Rein, Bert Herring, Jonathan Nguyen, Kyungho Song, Udari Madhushani Sehwag, Jiyeon Cho, Kaustubh Deshpande, Yeongkyun Jang, Jiyeon Joo, Minn Seok Choi et al. Published: 2026-05-13 Categories: cs.CL, cs.AI, cs.CR, cs.CY arXiv: https://arxiv.org/abs/2605.14152v2 Safety evaluations for large language models (LLMs) increasingly target high-stakes National Security and Public Safety (NSPS) risks, yet multilingual safety is mostly assessed through translation-only benchmarks that preserve the underlying scenario, leaving how language and geopolitical context interact largely unexamined beyond a few language pairs. We introduce ROK-FORTRESS, a bilingual, culturally adversarial NSPS benchmark that uses the English-Korean language pair and U.S.-ROK geopolitical axis as a case study, separating the effects of language and geopolitical grounding via a transcreation matrix: adversarial intents are evaluated under controlled combinations of (i) English versus Korean language and (ii) U.S. versus Korean entities, institutions, and operational details. Each adversarial prompt is paired with a dual-use benign counterpart to quantify over-refusal, and responses are scored by calibrated LLM-as-a-judge panels using expert-crafted, prompt-specific binary rubrics. Across a dual-track set of frontier and Korean-optimized models, we find a consistent suppression effect in Korean variants and substantial model-to-model variation in how geopolitical grounding interacts with language; in a subset of models, Korean grounding further mitigates the language-driven suppression. This indicates that, at least in the English-Korean case, safety behavior is shaped by language-as-risk signals and context interactions that translation-only evaluations miss. A direct-request ablation that strips jailbreak wrappers separates a small but persistent reduction for closed-source models from a larger, wrapper-dependent effect that reverses for open-source models, suggesting part of the Korean suppression reflects prompt specialization rather than intrinsic language-based safety alignment. The transcreation matrix methodology is designed to generalize to other language-culture pairs. ---