论文广场 - AcademicHub

01.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2606.16038

Open-SWE-Traces: Advancing Dual-Mode Multilingual Distillation for Software Engineering Agents

作者:

Wasi Uddin Ahmad ↗Nikolai Ludwig ↗Somshubra Majumdar ↗Boris Ginsburg ↗

arXiv:2606.16038v1 Announce Type: cross Abstract: The path toward autonomous software engineering is currently bottlenecked by a severe deficit of diverse, large-scale trajectory data. We address this by introducing \ourdataset, an expansive dataset of 207,489 agentic trajectories spanning nine programming languages (Python, Go, TS, JS, Rust, Java, PHP, C, C++). Sourced from 20,000 real-world PRs via OpenHands and SWE-agent harnesses, the dataset utilizes a hybrid-reasoning synthesis: Minimax-M2.5 generates trajectories with explicit "thinking" processes, while Qwen3.5-122B provides high-quality "non-thinking" traces. Filtered for permissive licenses (MIT, Apache, BSD) from SWE-rebench-V2, this data facilitates the training of models capable of long-horizon reasoning. We validate the dataset by fine-tuning the Qwen3-30B-A3B series (Thinking, Instruct, and Coder). The best performing model achieves resolve rates of 61.7% on SWE-bench Verified, 57.1% on SWE-bench Multilingual, and 36.8% on SWE-bench Pro. These results establish Open-SWE-Traces as a premier resource for distilling human-level software engineering capabilities into efficient, open-source agentic LLMs.

阅读与讨论 → 访问原文 →

02.

arXiv (CS.CL) 2026-06-11 DOI: arXiv:2606.11387

Small Experiments, Cheaper Decisions: A Case Study in Staged Promotion for Micro-Pretraining

作者:

Felipe Chavarro Polania ↗

Short pretraining runs can reduce experimental cost, but they can also over-promote configurations that only look strong at tiny budgets. We study an auditable staged-promotion protocol for a fixed micro-pretraining runner on two heterogeneous host blocks: Windows A100 and Linux L40S. Starting from twelve prior-screened configurations, we use staged budgets of 2 minutes, 5 minutes, 10 minutes, 60 minutes, and 12 hours, with frozen promotion rules before expensive continuations. The early screens are intentionally treated as unstable: the 5- and 10-minute rankings are host-sensitive, and the eventual 12-hour top-ranked condition is not the mean-best condition at the replicated 10-minute gate. Because seed ranges differ across stages, these changes are operational promotion evidence, not within-seed curves. A replicated 60-minute gate keeps the Staged Factorial Screening bridge reference in the promoted set, where it ranks first in all four 60-minute host-seed cells. In the final 12-hour confirmation package, the bridge condition ranks first in all four host-seed cells across two seeds; the greedy comparator does not meet the frozen 0.010 val_bpb near-equivalence rule; and the cheaper d8/ar48 (depth-8, aspect-48) sentinel does not meet the frozen 0.020 mean-gap rule. The executed 12-hour branch spends 144 GPU-hours, and the full staged protocol records 169.2 training GPU-hours including screening stages. Continuing all four 60-minute candidates would spend 192 GPU-hours, while continuing all nine replicated 10-minute candidates would spend 432 GPU-hours. The latter numbers are accounting counterfactuals for unrun continuations, not evidence that skipped candidates could not have overtaken the reference. The result is a bounded cost-allocation finding, not a claim of global optimality, capacity-normalized superiority, or superiority over adaptive hyperparameter optimization methods.

阅读与讨论 → 访问原文 →

03.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2512.10104

Phishing Email Detection Using Large Language Models

作者:

Najmul Hasan ↗Prashanth BusiReddyGari ↗Haitao Zhao ↗Yihao Ren ↗Jinsheng Xu ↗Shaohu Zhang ↗

arXiv:2512.10104v2 Announce Type: cross Abstract: Email phishing is one of the most prevalent and globally consequential vectors of cyber intrusion. As systems increasingly deploy Large Language Models (LLMs) applications, these systems face evolving phishing email threats that exploit their fundamental architectures. Current LLMs require substantial hardening before deployment in email security systems, particularly against coordinated multi-vector attacks that exploit architectural vulnerabilities. This paper proposes LLMPEA, an LLM-based framework to detect phishing email attacks across multiple attack vectors, including prompt injection, text refinement, and multilingual attacks. We evaluate three frontier LLMs (e.g., GPT-4o, Claude Sonnet 4, and Grok-3) and comprehensive prompting design to assess their feasibility, robustness, and limitations against phishing email attacks. Our empirical analysis reveals that LLMs can detect the phishing email over 90% accuracy while we also highlight that LLM-based phishing email detection systems could be exploited by adversarial attack, prompt injection, and multilingual attacks. Our findings provide critical insights for LLM-based phishing detection in real-world settings where attackers exploit multiple vulnerabilities in combination.

阅读与讨论 → 访问原文 →

04.

arXiv (CS.CL) 2026-06-17 DOI: arXiv:2606.17175

Self-Generated Error Training for Token Editing in Diffusion Language Models

作者:

Lin Yao ↗

Token-to-token (T2T) editing lets LLaDA2.1 revise committed tokens during block-diffusion decoding. The released recipe trains this editor on random vocabulary corruptions, but at inference the editor sees the model's own fluent, high-confidence draft errors instead. We study this training-inference mismatch and propose self-generated T2T, which performs a no-gradient draft pass, fills masked positions with predicted tokens, and supervises recovery in a second pass under these self-generated corruptions. We implement the update as a short LoRA continued-pretraining pass on LLaDA2.1-mini and evaluate on several benchmarks under the official Q-Mode T2T procedure with unchanged inference parameters. The method generally improves accuracy while reducing T2T edit intensity, mitigating failure modes such as final-digit transcription errors after otherwise correct reasoning and excessive self-correction before short factual answers.

阅读与讨论 → 访问原文 →

05.

arXiv (CS.LG) 2026-06-15 DOI: arXiv:2606.13823

A Stationarity-and-Coupling Criterion for Training-Free Time-Lagged Spectral Embeddings of Multivariate Time Series

作者:

Siddharth Pal ↗Viktoria Rojkova ↗

arXiv:2606.13823v1 Announce Type: new Abstract: We study training-free fixed-length descriptors for multivariate time series and ask not merely whether such a descriptor performs well, but when it can be expected to work at all. Our object of study is $D(\tau)$, built from a time-lagged correlation matrix truncated at the Marchenko-Pastur edge so that only signal-bearing eigenvalues survive and classified by cosine similarity to class centroids with zero learned parameters. The central contribution is not the descriptor but a falsifiable applicability criterion for it. Working from a stationary Gaussian VAR(1) model, we argue that $D(\tau)$ separates two classes when the signals are approximately stationary and the class information lives in their cross-channel temporal coupling rather than in marginal per-channel power. We derive, semi-formally, three consequences: a distinguishability condition, why the static ($\tau=0$) covariance collapses to chance, and why a stationary but power-discriminated paradigm defeats the descriptor. The criterion is operational: a two-part pre-flight test – an augmented Dickey-Fuller stationarity check and a power-baseline saturation check – predicts applicability before any training. We validate both halves on a mixed assortment. On four paradigms that satisfy the criterion (Sleep-EDF, BCI-IV-2a, MIT-BIH, ESC-50) the descriptor is competitive with strong baselines at a fraction of their cost, reaching $88.5\pm4.5\%$ under 20-subject leave-one-subject-out on Sleep-EDF on a single CPU thread. On three that violate it – non-stationary ERPs, and financial-volatility and wearable-stress regimes that are power-discriminated – it fails exactly as the pre-flight predicts, and these negatives are the more informative half. We are explicit that $D(\tau)$ is not the most accurate representation; its value is a compact, training-free embedding whose domain of validity is known in advance.

阅读与讨论 → 访问原文 →

06.

arXiv (CS.CL) 2026-06-16 DOI: arXiv:2606.16629

Islamic Large Language Models: From Knowledge Acquisition to Trustworthy and Hallucination-Resistant AI

作者:

Mohammed Amine Mouhoub ↗

Large language models (LLMs) are increasingly used for knowledge-intensive question answering, including religious and legal questions. Islamic knowledge is a particularly demanding setting: answers are expected to be grounded in authoritative sources, citations must be exact, Arabic varieties differ substantially from the language of classical sources, and legitimate jurisprudential disagreement must be represented rather than collapsed into a single answer. This survey reviews the emerging field of Islamic LLMs and trustworthy Islamic AI. We organize the literature around Arabic NLP and Arabic-centric LLMs, Islamic NLP resources, Qur'anic question answering, Islamic knowledge benchmarks, retrieval-augmented generation, Islamic legal reasoning, inheritance reasoning, hallucination evaluation, and trustworthiness. We argue that fluency in Arabic is not sufficient for Islamic AI. Reliable systems require curated sources, retrieval and verification modules, citation-aware generation, madhhab-aware reasoning, human expert evaluation, and benchmarks that measure not only answer accuracy but also faithfulness, source validity, and reasoning quality. The survey concludes with a research agenda for hallucination-resistant Islamic AI systems.

阅读与讨论 → 访问原文 →

07.

arXiv (CS.LG) 2026-06-19 DOI: arXiv:2510.18784

CAGE: Curvature-Aware Gradient Estimation For Accurate Quantization-Aware Training

作者:

Soroush Tabesh ↗Mher Safaryan ↗Andrei Panferov ↗Alexandra Volkova ↗Dan Alistarh ↗

arXiv:2510.18784v3 Announce Type: replace Abstract: Despite significant work on low-bit quantization-aware training (QAT), there is still an accuracy gap between such techniques and native training. To address this, we introduce CAGE (Curvature-Aware Gradient Estimation), a new QAT method that augments the straight-through estimator (STE) gradient with a curvature-aware correction designed to counteract the loss increase induced by quantization. CAGE is derived from a multi-objective view of QAT that balances loss minimization with the quantization constraints, yielding a principled correction term that depends on local curvature information. On the theoretical side, we introduce the notion of Pareto-optimal solutions for quantized optimization, and establish that CAGE yields strong convergence guarantees in the smooth non-convex setting. In terms of implementation, our approach is optimizer-agnostic, but we provide a highly-efficient implementation that leverages Adam statistics. CAGE significantly improves upon the prior state-of-the-art methods in terms of accuracy, for similar computational cost: for QAT fine-tuning, it halves the compression accuracy loss relative to the prior best method, while for QAT pre-training of Llama models, its accuracy for 3-bit weights-and-activations (W3A3) matches the accuracy achieved at 4-bits (W4A4) with the prior best method. The official implementation can be found over https://github.com/IST-DASLab/CAGE .

阅读与讨论 → 访问原文 →

08.

arXiv (quant-ph) 2026-06-15 DOI: arXiv:2601.06468

Emission of time-ordered photon pairs from a coherently-driven Kerr microcavity

作者:

arXiv:2601.06468v2 Announce Type: replace-cross Abstract: Weakly-interacting many-body systems possess remarkable quantum properties that are essential components of quantum technologies, and constitute a topic of fundamental interest. Here we show that in a solid-state nonlinear microcavity embedding discrete modes of exciton-dressed photons, we can isolate a single eigenmode of quantum fluctuations from the much brighter coherent fraction of the field. In this regime, we perform frequency- and time-resolved correlations measurements between photons on the red and blue side of the fluctuations spectrum. When the average number of fluctuation quanta is smaller than one, we observe the formation of large pairwise time-ordered correlations: red photon first and blue photon second. We show that this peculiar time-ordering correlation emerges spontaneously from the interplay between frequency-resolved detection, and the non-trivial internal quantum structure of the elementary fluctuations.

阅读与讨论 → 访问原文 →

09.

arXiv (math.PR) 2026-06-19 DOI: arXiv:2511.08288

The central heat trace on large compact classical groups

作者:

Thibaut Lemoine ↗Myl\`ene Ma\"ida ↗

arXiv:2511.08288v2 Announce Type: replace-cross Abstract: We study the large-$N$ asymptotics of the central trace of the heat kernel on compact classical groups. For every classical family $G_N\subset \mathrm{GL}_N(\C)$, we prove a full large-$N$ asymptotic expansion, using a highest weights/partitions correspondence adapted to the large-rank regime, under which the eigenvalues of the Laplace–Beltrami operator stabilize as observables in the algebra of shifted symmetric functions. Then, we prove a random surface representation of the trace in terms of ramified coverings of the torus. We provide two independent applications: an explicit large-rank counting law for the Casimir spectrum, with exponential Hardy–Ramanujan-type growth in contrast with the polynomial behavior of Weyl's law at fixed rank, and a rigorous probabilistic formulation of the Yang–Mills/Hurwitz duality on a two-dimensional torus initiated by Gross and Taylor, completing a previous work of the authors. We also extend this duality to a Yang–Mills/Gromov–Witten duality by expressing the coefficients of the central heat trace as explicit functionals of the generating function of Gromov–Witten invariants.

阅读与讨论 → 访问原文 →

10.

arXiv (CS.CL) 2026-06-18 DOI: arXiv:2505.23851

ASyMOB: Algebraic Symbolic Mathematical Operations Benchmark

作者:

Michael Shalyt ↗Rotem Elimelech ↗Ido Kaminer ↗

Large language models (LLMs) are increasingly applied to symbolic mathematics, yet existing evaluations often conflate pattern memorization with genuine reasoning. To address this gap, we present ASyMOB, a high-resolution dataset of 35,368 validated symbolic math problems spanning integration, limits, differential equations, series, and hypergeometrics. Unlike prior benchmarks, ASyMOB systematically perturbs each seed problem using symbolic, numeric, and equivalence-preserving transformations, enabling a fine-grained assessment of generalization. Our evaluation reveals three key findings: (1) most models' performance collapses under minor perturbations, while top systems exhibit an apparent regime shift in robustness; (2) integrated code tools stabilize performance, particularly for weaker models; and (3) we identify examples where Computer Algebra Systems (CAS) fail while LLMs succeed, as well as problems solved only via a hybrid LLM-CAS approach, highlighting a promising integration frontier. ASyMOB serves as a principled diagnostic tool for measuring and accelerating progress toward building verifiable, trustworthy AI for scientific discovery.

阅读与讨论 → 访问原文 →

11.

arXiv (CS.CL) 2026-06-17 DOI: arXiv:2606.18237

ReproRepo: Scaling Reproducibility Audits with GitHub Repository Issues

作者:

Shanda Li ↗Qiuhong Anna Wei ↗Jingwu Tang ↗Valerie Chen ↗Nihar B Shah ↗Tim Dettmers ↗Yiming Yang ↗Ameet Talwalkar ↗

Reproducing research results from papers and released code is central to scientific progress. Existing works have introduced benchmarks to evaluate whether LLM agents can assist with reproducibility, but they are difficult to scale due to their reliance on substantial manual effort for data curation and evaluation. We introduce ReproRepo, a scalable framework for reproducibility evaluation that leverages human-raised GitHub issues as naturally occurring supervision on realistic reproduction blockers. We instantiate ReproRepo on 1,149 recent machine learning papers from major conferences and evaluate four frontier model-agent configurations. Our results show that LLM agents, even without executing code, can identify many real-world reproducibility problems from paper-repository pairs: the best agent in our study, namely Codex with GPT-5.5, surfaces at least one semantically related human-reported blocker for ~90% of papers in the study. Further analysis shows that agents are particularly effective for surfacing visible failures and identifying the right semantic region, but may still be insufficient in exact localization. ReproRepo can serve as a reusable, scalable framework for future evaluations of LLM agents on real-world reproducibility auditing. Our code is released at https://github.com/LithiumDA/ReproRepo.

阅读与讨论 → 访问原文 →

12.

arXiv (CS.LG) 2026-06-12 DOI: arXiv:2606.10678

One Step Closer to Ground Truth: A Multi-Scale Residual-Aware Representation Learning Pipeline for Predicting Time Series Data

作者:

Amrijit Biswas ↗Mustafa Kamal ↗Robin Krambroeckers ↗M. M. Lutfe Elahi ↗Sifat Momen ↗Nabeel Mohammed ↗Shafin Rahman ↗

arXiv:2606.10678v2 Announce Type: replace Abstract: Transformer-based models have emerged as leading paradigms in time-series forecasting in recent years, employing self-attention mechanisms to capture long-range dependencies. Despite their success, these single-stage forecasting architectures exhibit persistent systematic residual biases arising from structural discrepancies, unmodeled stochastic components, or inadequate multi-scale temporal representations. This limitation persists when residuals are treated as irreducible noise, precluding adaptive correction of structured error patterns. To address this limitation, we introduce a two-stage, model-agnostic framework that explicitly decouples forecasting and residual learning into distinct stages of representation learning. A base transformer first generates the initial predictions. Subsequently, a dedicated meta-corrector dynamically models structured error patterns across multivariate channels, preserves cross-variable dependencies, and iteratively refines the residual bias of the base transformer. By formalizing this pipeline as a hypothesis space expansion, our framework addresses approximation limitations inherent in single-stage architectures, removes reliance on restrictive assumptions, and enables end-to-end learning of complex error dynamics. Evaluated on eight popular benchmark datasets using established protocols, our approach achieves state-of-the-art performance, with significant improvements in standard metrics (MSE, MAE). The results demonstrate the framework's ability to mitigate systematic biases and enhance robustness to complex temporal dynamics, advancing the practical applicability of transformer-based forecasting models.

阅读与讨论 → 访问原文 →

13.

arXiv (CS.AI) 2026-06-19 DOI: arXiv:2503.02636

A Deep Generative Model for Resting-State EEG Synthesis and Transferable Representation Learning

作者:

Yeganeh Farahzadi ↗Morteza Ansarinia ↗Zoltan Kekecs ↗

arXiv:2503.02636v5 Announce Type: replace-cross Abstract: Resting-state EEG provides a non-invasive view of spontaneous brain activity, but extracting meaningful patterns is often limited by scarce high-quality data and reliance on manually engineered features. Generative adversarial networks (GANs) can synthesize neural signals and learn transferable representations directly from raw data, a dual capability that remains underexplored in EEG research. Here, we introduce REST-GAN, a GAN-based framework for resting-state EEG that combines adversarial training with an auxiliary self-supervised reconstruction objective to support signal synthesis and unsupervised feature extraction. Although trained only on raw time-domain signals, without explicit frequency-domain or sensor-topographic supervision, the generated time series reproduced key temporal, spectral, and connectivity properties of real EEG. In band-power feature space, generated samples showed high precision and recall across eyes-open and eyes-closed conditions (EO: 0.91/0.67; EC: 0.87/0.65), while group-average spectral coherence matrices showed low mean absolute differences from real data across frequency bands (~0.01-0.03). The representations learned by the model's critic transferred to independent resting-state demographic classification tasks, outperforming models trained directly on raw EEG and showing competitive performance relative to a recent EEG foundation model, while requiring substantially less training data and computational resources. These findings highlight a computationally efficient, architecture-driven strategy in which generative models serve not only as EEG signal generators, but also as unsupervised feature extractors. This approach may support more data-efficient EEG analysis while reducing reliance on manual feature engineering. The implementation code for REST-GAN is available at: https://github.com/Yeganehfrh/REST-GAN.

阅读与讨论 → 访问原文 →

14.

bioRxiv (Bioinfo) 2026-06-19 DOI: HASH:f5d8865f0d6cb6fee4e8a106c7c18add

Morpho-FM: spatial molecular reconstruction from routine H&E histology using transcriptomic foundation-model priors

作者:

Huang ↗J.-J ↗Feng ↗Qu ↗L.-H ↗Zheng ↗L.-L ↗

Routine haematoxylin and eosin (H&E) histology captures tissue architecture at clinical scale, but lacks a direct molecular readout of the transcriptional programmes that organise tumour epithelium, stroma, vasculature and immune compartments. Spatial transcriptomics provides this context, yet cost, workflow complexity and sparse sampling limit routine use. Most existing histology-to-expression models are trained de novo on small paired cohorts and therefore remain weakly constrained when extrapolating from sparse measurements to dense, tissue-wide molecular maps. Here we introduce Morpho-FM, a weakly supervised framework that predicts spatial gene expression from routine H&E whole-slide images by conditioning a pretrained single-cell transcriptomic foundation-model prior on local histological neighbourhoods. A lightweight morphology-to-transcriptome adapter maps cached whole-slide histology features into a transcriptomic decoder, enabling prediction at measured locations, dense full-section reconstruction, and re-aggregation to the original measurement support. Across harmonized prostate cancer benchmarks, Morpho-FM achieved the strongest overall performance among five representative methods, reaching mean per-gene Pearson correlations of 0.286 in rotating single-slide evaluation and 0.298 in multi-slide held-out validation. The framework reproduced this advantage across kidney cancer sections, achieved a mean correlation of 0.210 across 56 directed single-slide evaluations and retained measurable predictive signal after external transfer to clear-cell renal cell carcinoma sections. Controlled ablation analyses identified pretrained transcriptomic initialization as a reproducible source of performance gain exceeding that attributable to changes in the histology feature backbone. Beyond predictive accuracy benchmarks, Morpho-FM recovered ERBB2-enriched tumour compartments, boundary-associated molecular gradients, and annotation-aligned tissue domains across Xenium and HER2ST breast cancer datasets. Together, these results support transcriptomic foundation-model priors as an effective constraint for morphology-conditioned molecular decoding and demonstrate the potential of Morpho-FM to extend spatial transcriptomic insight across routine pathology sections.

阅读与讨论 → 访问原文 →

15.

arXiv (CS.LG) 2026-06-16 DOI: arXiv:2602.06404

Near-Optimal Regret for Distributed Adversarial Bandits: A Black-Box Approach

作者:

Hao Qiu ↗Mengxiao Zhang ↗Nicol\`o Cesa-Bianchi ↗

arXiv:2602.06404v2 Announce Type: replace Abstract: We study distributed adversarial bandits, where $N$ agents cooperate to minimize the global average loss while observing only their own local losses. We show that the minimax regret for this problem is $\tilde{\Theta}(\sqrt{(\rho^{-1/2}+K/N)T})$, where $T$ is the horizon, $K$ is the number of actions, and $\rho$ is the spectral gap of the communication matrix. Our algorithm, based on a novel black-box reduction to bandits with delayed feedback, requires agents to communicate only through gossip. It achieves an upper bound that significantly improves over the previous best bound $\tilde{O}(\rho^{-1/3}(KT)^{2/3})$ of Yi and Vojnovic (2023). We complement this result with a matching lower bound, showing that the problem's difficulty decomposes into a communication cost $\rho^{-1/4}\sqrt{T}$ and a bandit cost $\sqrt{KT/N}$. We further demonstrate the versatility of our approach by deriving first-order and best-of-both-worlds bounds in the distributed adversarial setting. Finally, we extend our framework to distributed linear bandits in $R^d$, obtaining a regret bound of $\tilde{O}(\sqrt{(\rho^{-1/2}+1/N)dT})$, achieved with only $O(d)$ communication cost per agent and per round via a volumetric spanner.

阅读与讨论 → 访问原文 →

16.

arXiv (CS.CV) 2026-06-12 DOI: arXiv:2604.27277

BrainDINO: A Brain MRI Foundation Model for Generalizable Clinical Representation Learning

作者:

Yizhou Wu ↗Shansong Wang ↗Yuheng Li ↗Mojtaba Safari ↗Mingzhe Hu ↗Chih-Wei Chang ↗Harini Veeraraghavan ↗Xiaofeng Yang ↗

Brain MRI underpins a wide range of neuroscientific and clinical applications, yet most learning-based methods remain task-specific and require substantial labeled data. Here we show that a single self-supervised representation can generalize across heterogeneous brain MRI endpoints. We trained BrainDINO, a self-distilled foundation model, on approximately 6.6 million unlabeled axial slices from 20 datasets encompassing broad variation in population, disease, and acquisition setting. Using a frozen encoder with lightweight task heads, BrainDINO supported transfer across tumor segmentation, neurodegenerative and neurodevelopmental conditions classification, brain age estimation, post-stroke temporal prediction, molecular status prediction, MRI sequence classification, and survival modeling. Across tasks and supervision regimes, BrainDINO consistently equaled or exceeded natural-image and MRI-specific self-supervised baselines, with particularly strong advantages under label scarcity. Representation analyses further showed anatomically organized and pathology-sensitive feature structure in the absence of task-specific supervision. Our findings indicate that large-scale slice-wise self-supervised learning can yield a unified brain MRI representation that supports diverse neuroimaging tasks without volumetric pretraining or full-network fine-tuning, establishing a scalable foundation for robust and data-efficient brain imaging analysis. Code is available at https://github.com/mclwu22/BrainDINO

阅读与讨论 → 访问原文 →

17.

arXiv (CS.AI) 2026-06-17 DOI: arXiv:2606.17114

An Evaluation of Data Leakage Risks in Tool-Using LLM Agents in Realistic Scenarios

作者:

Hankyul Baek ↗Jaewon Noh ↗Sang Seo ↗Yongsu Kim ↗Gabriel Waikin Loh Matienzo ↗Young Il Kim ↗Ee Wei Seah ↗Akriti Vij ↗

arXiv:2606.17114v1 Announce Type: cross Abstract: AI agents are increasingly being adopted in enterprise and personal settings with access to emails, databases, documents, and other tools where they can read, update, and disseminate sensitive information. Much of prior research on data leakage risks in agents has focused on adversarial data exfiltration through prompt injections and jailbreaks. However, sensitive information may also be exposed during non-adversarial use, creating leakage risks even when users issue benign requests. We report a joint evaluation by the Singapore AI Safety Institute and the Korea AI Safety Institute examining agent data leakage in 12 realistic, non-adversarial tasks spanning customer support, DevOps, web automation, and enterprise and personal productivity. The evaluation covers five risk types: lack of data awareness, audience awareness, policy compliance, data minimization, and access-boundary awareness. Both institutes tested a common set of scenarios mirroring real-world deployments using independent testing environments and task-specific LLM-judge rubrics. Across the three tested agents, none achieved fully correct and fully safe execution across all scenarios. Successful task completion often coincided with data-handling failures such as accessing unnecessary information or disclosing information to inappropriate recipients, indicating that capability and data-handling safety should be evaluated separately. Qualitative review also revealed claim-action mismatches, simulation-aware behavior, user-simulator role reversal, and interpretation gaps in automated judging. Overall, the results indicate that operational data leakage is a first-order agent-safety concern distinct from adversarial exfiltration and provide a methodology for future evaluations of agent data-handling safety.

阅读与讨论 → 访问原文 →

18.

Nature (Science) 2026-06-10 DOI: HASH:d28a2c07fea35854d58a1b38f4e722ed

Improved quantum processor logical error rates via correction and detection

作者:

A. Paetznick ↗

Performing quantum algorithms for critical problems in physics and chemistry requires substantially lower error rates than the physical error rates of present quantum computers. Achieving such low logical error rates requires quantum error correction1,2 and physical error rates below a critical threshold value3–8. We experimentally demonstrate on a trapped-ion quantum charge-coupled device (QCCD)9,10 improvements in logical error rates ranging from 11× to 800× compared with several physical circuit baselines, including quantum computation on multiple qubits. Our results hinge on two quantum error correction code constructions optimized for an ion-trap processor: a 12-qubit code encoding two qubits inspired by Knill11 and a 16-qubit tesseract colour code encoding four qubits12,13. These constructions are combined with a scalable method of error detection and post-selection to achieve reduced logical error rates. Our results show that state-of-the-art quantum devices are already able to make use of fault tolerance and error correction to strongly suppress errors in non-trivial quantum circuit computations. Experimental demonstration of quantum error-correcting codes combined with error detection and post-selection applied to a trapped-ion quantum processor shows improvements in logical error rates ranging from 11× to 800× compared with several physical circuit baselines.

阅读与讨论 → 访问原文 →

19.

arXiv (CS.CV) 2026-06-12 DOI: arXiv:2606.13156

Iterative Visual Thinking: Teaching Vision-Language Models Spatial Self-Correction through Visual Feedback

作者:

Animesh Tripathy ↗Aswanth Krishnan ↗

Vision-language models (VLMs) achieve strong singleshot spatial grounding, yet lack any mechanism to observe and correct their own predictions. We find that naively prompting a VLM to iterate over rendered visualizations of its predictions causes catastrophic failure: Acc@0.5 on referring expression comprehension collapses from 79.6% to 48.7% (a 31 percentage point drop), revealing a fundamental gap between grounding capability and self-correction ability. We propose Iterative Visual Thinking (IVT), a closed-loop framework in which the model predicts a bounding box, observes the prediction rendered on the image, and iteratively refines through visual feedback. A two-phase training recipe closes the self-correction gap: first, we exploit the base model's own predictions as realistic errors and prompt a teacher VLM to generate corrective reasoning traces, yielding supervised data without human annotation; second, we apply Group Relative Policy Optimization (GRPO) with a simple IoU reward to stabilize multi-step refinement. On a mixed benchmark spanning RefCOCOg, Ref-Adv, and Ref-L4 (505 test samples), SFT warm-up with IVT surpasses the single-shot base model on every metric: Acc@0.5 rises to 82.0% (+2.4pp), Acc@0.7 to 74.1% (+3.2pp), and Acc@0.9 to 48.3% (+2.8pp). GRPO further reduces per-step IoU degradation by 5x, stabilizing the refinement trajectory. All training uses only 2,400 samples on a single GPU, demonstrating that spatial self-correction is a learnable capability that can be instilled at modest scale.

阅读与讨论 → 访问原文 →

20.

arXiv (CS.CV) 2026-06-16 DOI: arXiv:2606.15869

Metis: A Generalizable and Efficient World-Action Model for Autonomous Driving and Urban Navigation

作者:

Jingyu Li ↗Zhe Liu ↗Dongnan Hu ↗Junjie Wu ↗Zipei Ma ↗Wenxiao Wu ↗Chao Han ↗Zhihui Hao ↗Zhikang Liu ↗Kun Zhan ↗Jiankang Deng ↗Xiatian Zhu ↗…

World action models~(WAMs) have shown great promise for autonomous driving and urban navigation. Built upon Vision-Language-Action models or video generation models, existing approaches suffer key limitations: (1) High inference latency due to future observation prediction at test time, and (2) tightly coupled video and action modeling leading to representational mismatch and degraded generalization. To address both issues, we propose Metis, an end-to-end WAM framework that decouples video generation and action prediction. Specifically, Metis employs a Mixture-of-Transformers architecture with dedicated experts for video generation and action prediction, preserving the intrinsic distributional properties of each task. To enhance efficiency, we introduce an asymmetric attention mask that enables joint training of both experts while allowing the action model to bypass explicit video generation during inference. This design ensures training-inference consistency and significantly reduces computational costs without compromising planning performance. Extensive experiments demonstrate state-of-the-art performance on the NAVSIM navhard and navtest benchmarks and the CityWalker navigation benchmark, validating both the generalizability and efficiency across diverse tasks. Real-robot deployments further confirm the practical feasibility of our approach.

阅读与讨论 → 访问原文 →

21.

arXiv (CS.CL) 2026-06-12 DOI: arXiv:2410.00903

Causal Inference with Generative Artificial Intelligence: Application to Texts as Treatments

作者:

Kosuke Imai ↗Kentaro Nakamura ↗

In this paper, we demonstrate how to enhance the validity of causal inference with unstructured high-dimensional treatments like texts, by leveraging the power of generative Artificial Intelligence (GenAI). Specifically, we propose to use a deep generative model such as large language models (LLMs) to efficiently generate treatments and use their internal representation for subsequent causal effect estimation. We show that the knowledge of this true internal representation helps disentangle the treatment features of interest, such as specific sentiments and certain topics, from other possibly unknown confounding features. Unlike existing methods, the proposed GenAI-Powered Inference (GPI) methodology eliminates the need to learn causal representation from the data, and hence produces more accurate and efficient estimates. We formally establish the conditions required for the nonparametric identification of the average treatment effect, propose an estimation strategy that avoids the violation of the overlap assumption, and derive the asymptotic properties of the proposed estimator through the application of double machine learning. Finally, using an instrumental variables approach, we extend the proposed GPI methodology to the settings in which the treatment feature is based on human perception. The GPI is also applicable to text reuse where an LLM is used to regenerate existing texts. We conduct simulation and empirical studies, using the generated text data from an open-source LLM, Llama 3, to illustrate the advantages of our estimator over state-of-the-art causal representation learning algorithms.

阅读与讨论 → 访问原文 →

22.

arXiv (CS.AI) 2026-06-19 DOI: arXiv:2606.20502

Calibration Without Comprehension: Diagnosing the Limits of Fine-Tuning LLMs for Vulnerability Detection in Systems Software

作者:

Arastoo Zibaeirad ↗Marco Vieira ↗

arXiv:2606.20502v1 Announce Type: cross Abstract: Whether LLMs scoring well on vulnerability benchmarks genuinely reason about security or merely pattern-match on contaminated data remains unresolved. We present CWE-Trace, a framework for LLM vulnerability detection built from 834 manually curated Linux kernel samples spanning 74 CWEs. The framework enforces a strict temporal split (pre-2025 historical set / post-cutoff leakage-free set), preserves context-aware vulnerable–patched pairs, and introduces two diagnostic metrics: the Directional Failure Index (DFI) and Hierarchical Distance and Direction (HDD). We evaluate eight vanilla LLMs and 15 LoRA fine-tuned variants across non-targeted detection, targeted detection, and CWE classification. Our analysis yields two key results. First, data contamination provides no measurable advantage. Function-level analysis shows that 84% of nominally contaminated samples carry no usable memorization signal: vulnerable functions are absent or cross-mapped across datasets, and ~31% of contaminated samples carry CWE misclassification. Second, backbone directional priors dominate fine-tuning. Models exhibit stable, systematic failure modes (DFI ranging from -85.5 to +94.8 pp) that persist from historical to post-cutoff data and resist correction. Fine-tuning shifts the output threshold without changing the decision policy. This is calibration without comprehension: output distributions adapt to training data while the underlying security reasoning remains absent. The weakest backbone at binary detection (DeepSeek-R1) gains the most in coarse CWE classification, revealing that detection and understanding are decoupled capabilities. The best detection score reaches only 52.1% (+2.1 pp above chance); exact CWE ranking remains below 1.3% Top-1 accuracy, confirming that current LLMs lack reliable security reasoning for systems software, regardless of fine-tuning strategy.

阅读与讨论 → 访问原文 →

23.

arXiv (CS.CL) 2026-06-11 DOI: arXiv:2601.03792

VietMed-MCQ: A Consistency-Filtered Data Synthesis Framework for Vietnamese Traditional Medicine Evaluation

作者:

Huynh Trung Kiet ↗Dao Sy Duy Minh ↗Nguyen Dinh Ha Duong ↗Le Hoang Minh Huy ↗Long Nguyen ↗Dien Dinh ↗

Large Language Models (LLMs) have demonstrated remarkable proficiency in general medical domains. However, their performance significantly degrades in specialized, culturally specific domains such as Vietnamese Traditional Medicine (VTM), primarily due to the scarcity of high-quality, structured benchmarks. In this paper, we introduce VietMed-MCQ, a novel multiple-choice question dataset generated via a Retrieval-Augmented Generation (RAG) pipeline with an automated consistency check mechanism. Unlike previous synthetic datasets, our framework incorporates a dual-model validation approach to ensure reasoning consistency through independent answer verification, though the substring-based evidence checking has known limitations. The complete dataset of 3,190 questions spans three difficulty levels and underwent validation by one medical expert and four students, achieving 94.2 percent approval with substantial inter-rater agreement (Fleiss' kappa = 0.82). We benchmark seven open-source models on VietMed-MCQ. Results reveal that general-purpose models with strong Chinese priors outperform Vietnamese-centric models, highlighting cross-lingual conceptual transfer, while all models still struggle with complex diagnostic reasoning. Our code and dataset are publicly available to foster research in low-resource medical domains.

阅读与讨论 → 访问原文 →

24.

arXiv (CS.LG) 2026-06-15 DOI: arXiv:2606.14536

Provably Safe, Yet Scalable Reinforcement Learning

作者:

Kai S. Yun ↗Zeyang Li ↗Navid Azizan ↗

arXiv:2606.14536v1 Announce Type: new Abstract: Safe reinforcement learning (RL) aims to learn policies that optimize rewards while satisfying constraints. Predominant approaches rely on soft-constrained policy optimization, which has achieved empirical success but does not provide formal safety guarantees for the learned policy. In contrast, methods with strict guarantees typically rely on explicit certificate functions, whose construction requires the direct synthesis and verification of control-invariant sets, a process that scales poorly with state dimension and often yields overly conservative behavior. In this paper, we present the Provably Safe, yet Scalable RL (PS2-RL) framework, a novel two-phase architecture for learning provably safe policies in a scalable manner, designed to overcome the key bottlenecks of prior methods. Rather than explicitly computing invariant sets, PS2-RL leverages a learned backup policy to forward-integrate the system dynamics, generating an implicit control-invariant set online. In the first phase, the backup policy is trained with our proposed safe-arrival value function, which characterizes the optimal backup policy for invariant-set construction. In the second phase, an RL policy is trained end-to-end through a differentiable projection layer that strictly enforces the safety guarantees induced by the learned backup policy. By maximizing the volume of the implicit control-invariant set in the first phase, the resulting PS2 policy from the second phase is performant and scalable, while maintaining provable safety. Crucially, PS2-RL imposes no restrictions on the underlying RL algorithm and can be plugged into any existing training pipeline. We establish theoretical guarantees for the proposed framework and evaluate it on robotic control tasks with state dimensions up to 10, a regime in which prior provably safe RL methods struggle or become impractical.

阅读与讨论 → 访问原文 →

25.

arXiv (CS.CL) 2026-06-12 DOI: arXiv:2606.12599

Constrained Semantic Decompression in LLMs through Persian Proverb-Conditioned Story Generation

作者:

Zahra Habibzadeh ↗Paria Khoshtab ↗Amir Mesbah ↗Yadollah Yaghoobzadeh ↗

Transforming a dense, abstract proverb into an engaging and morally faithful narrative requires deep cultural understanding and robust semantic grounding. We frame this problem as a constrained semantic decompression task and study proverb-conditioned story generation as a testbed for abstraction-to-realization in large language models (LLMs). Focusing on Persian, we introduce the Proverb Aligned Narrative Dataset (PAND), pairing proverbs with human-written stories and explicit meanings. By a hybrid evaluation framework that combines human-calibrated LLM-as-a-Judge with structural metrics, we analyze model behavior across multiple prompting regimes. Our findings reveal a persistent decompression gap: current LLMs often achieve strong surface-level fluency while failing to faithfully instantiate the underlying moral and causal structure encoded in proverbs. We further show that explicit reasoning and iterative refinement can partially mitigate these failures, suggesting that many decompression errors arise from difficulties in translating abstract meaning into narrative form rather than a complete lack of relevant knowledge. Our proposed task naturally extends to other forms of compressed cultural knowledge.

阅读与讨论 → 访问原文 →

探索全球前沿学术脉络