跳到正文
研踪PaperTrail

论文库

检索计算机领域论文，按研究方向、作者或关键词筛选。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+0新增论文
+0其中 AI 相关
+0已有论文修订
+100新增 AI 导读

arXiv 最近完成扫描：2026-11-15 15:59 · 全站最近入库：2026-11-14 15:09各信源进度 →

统计口径与数据新鲜度

统计 2026-11-14 16:24 — 2026-11-15 16:24（北京时间） · 点击数字查看对应简报；AI 相关包含在新增论文内。

最近完成检查：2026-11-15 15:59最近发现新增版本：2026-11-14 15:09库内论文最新更新时间：2026-11-13 02:59

每分钟刷新已保存的数据。空检查不代表采集异常；新收录可能包含历史论文。AI 导读按新增版本计数。

23,348 已收录论文19,580 AI 相关6,729 有修订112 HF 社区发现

浏览论文库

研究日报 / 周报 →按信源浏览 →

时间范围跟随排序：版本更新、本站收录或首次发表。最新收录可能包含历史论文。部分信源异常，查看状态 →

找到 15 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-15 15:58（北京时间）
新论文 · v1arXiv:2611.16591

Singularity-Bench: Can Agents Build Agents?

Yang Zhou, Krish Agarwal, Zhuoming Chen, Zerui Wang, Zhenyu Zhang, Wenshuang Li, Xinpeng Wei, Zhichen Zeng 等 19 位作者

原文摘要

Frontier AI agents increasingly succeed on realistic, long-horizon tasks and are beginning to improve the systems through which they operate. Yet existing benchmarks often restrict agents to narrow action spaces, use simplified environments, or rely on verification methods that are unreliable for evaluating arbitrary agent-generated systems. We introduce Singularity-Bench, an open-ended benchmark for autonomous AI development with six end-to-end tasks spanning pre-training, mid-training, post-training, inference, the agent harness, and cluster orchestration. Each task requires agents to implement core systems from a minimal starting point and evaluates them against mature human-engineered baselines on identical hardware. Additionally, through extensive red-teaming covering 9 of the 10 categories in the OWASP Top 10 for Agentic Applications, we expose reward-hacking vulnerabilities in the conventional Harbor framework and develop Breakwater, an extension of Harbor that blocks all attacks tested in our study. Across 11 agents built from 7 frontier models and 2 harnesses, Opus 5.5 with Claude Code matches or outperforms the human baseline on all six tasks with a Singularity-Index of 1.25, suggesting it may have the skills needed to reach the singularity. The remaining agents average only 0.58. Our vision is for Singularity-Bench to track how much of AI development agents can already automate and when they may transition from assisting to taking end-to-end responsibility for building stronger successors.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2611.12870

SWAM: Scaling World Action Models with Cross-Modal Alignment

Yu Zhou, Xinyuan Xia, Xuankun Yang, Yinghao Yang, Wendi Chen, Bowen Liu, Jianhan Ma, Weicheng Duan 等 11 位作者

原文摘要

World action models jointly predict video and actions conditioned on language. Existing studies focus primarily on architectures and training mechanisms, typically treating language-video alignment acquired through video pretraining as an established foundation. However, introducing action learning may impair this alignment. Principled guidance on data and supervision design for maintaining language-video alignment during action learning remains limited. To investigate this, we curate a pool of approximately 47,000 hours of general videos, egocentric videos, UMI demonstrations, and robot demonstrations. Building on this, we systematically study how language-video supervision, language-video-action supervision, and the joint use of these two forms affect WAM performance. In particular, we examine how language annotation quality affects pretrained language-video capabilities, and how action supervision from UMI and robot demonstrations extends these capabilities to language-video-action modeling. To better scale the model, we further conduct detailed studies on joint training recipe and model architecture. Under a fixed training budget, mixing egocentric videos and UMI demonstrations improves control generalization. Model architecture affects how effectively these experiences are used, while scaling pretraining under the selected configuration brings further gains. The resulting recipe, post-trained on RoboTwin Clean demonstrations only, achieves success rates of 88.10% and 61.56% under Easy (Clean-to-Clean) and Hard (Clean-to-Random) conditions, respectively. Our model achieves 40.00% SR across three real-world tasks, compared with 23.33% for the strongest baseline, indicating better generalization under unseen conditions. We release the curated data, training code, and model weights.

cs.CV · cs.RO

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.09074

TAP: Efficient Long-Horizon Agent Pruning via Trajectory-Anchored Recovery

Yuanzhe Li, Pengxin Wang, Yuxin Ren, Jianing Deng, Jingtong Hu, Song Wang, Jingdi Chen, Huanrui Yang

原文摘要

Emerging long-horizon agentic tasks require repeated model calls, worsening the inference cost of already-costly language models. While narrow agentic tasks suggest potential for aggressive model pruning without performance drop, empirical results show existing methods proposed for question answering tasks severely degrade task performance when applied to agentic models. We trace this failure to two decisions: what to prune and how to recover. For pruning, one-shot importance estimates fail to track how the pruned model adapts. For recovery, offline distillation covers only teacher prefixes, while full-trajectory on-policy distillation causes student errors to compound across turns. In this work, we propose Trajectory-Anchored Pruning (TAP), the first structural pruning framework for reinforcement learning (RL)-trained agents. TAP couples structural pruning with efficient on-policy recovery, anchoring interactions to teacher trajectories while allowing the student to generate each reasoning-action response. A frozen dense teacher supervises the student's response prefixes, addressing within-response training-inference mismatch while preventing student-induced deviations from propagating across training turns. Instead of one-shot pruning, TAP re-scores channels using gradients of the recovery objective on the recovered student, connecting iterative channel selection to the evolving policy. With 60% of FFN channels removed, TAP retains 99.2% and 88.0% of the dense 7B agents' task success rates on ALFWorld and WebShop, respectively, while reducing GPU time per successful task by approximately 22% and 17%. These results demonstrate effective structural compression of long-horizon agents under a limited recovery budget.

cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2609.34064

Learning Perturbation Robust Policies for LLM Agents with Stable Optimization

Pengxin Wang, Yuanzhe LI, Yuxin Ren, Huanrui Yang, Jingdi Chen

原文摘要

Reinforcement learning (RL) has become an effective post-training paradigm for long-horizon large language model (LLM) agents. However, we find that the resulting policies can be sensitive to various policy perturbations, such as hidden-state noise, pruning, and quantization. In this work, we study how to improve perturbation robustness during policy optimization. We first introduce the notion of a perturbation robust policy and analyze conditions under which perturbed policy updates preserve stable monotonic improvement. Based on this analysis, we introduce Stable Perturbation-Robust Policy Optimization (SPrPO), which applies adaptive and sensitivity-aware perturbations during RL training. We evaluate SPrPO on ALFWorld and WebShop and conduct systematic experiments across multiple perturbation types and scales, showing improved perturbation robustness while maintaining stable policy optimization.

cs.LG · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2611.06858

CompressionBench: How can your coding agents compress LLMs?

Xinrui Zhong, Zhifeng Li, Zhiming Wang, Xinze Feng, Yongqi Liu, Shanglin Yang, Yang Zhou, Zikun Li 等 10 位作者

原文摘要

Recursive self-improvement requires AI systems to improve the models and infrastructure used by subsequent systems. LLM compression offers a concrete efficiency objective: retain a pretrained model's behavior within a smaller model storage budget. We study whether coding agents can carry out this step autonomously. We formulate compression as a terminal task in which an agent receives a source checkpoint, storage limit, inference environment, and time budget, then constructs compressed weights and executable inference code. The agent controls its algorithm and experiments; perplexity feedback guides search, and independent verification checks the final model. Across six agents and six model--size settings, 29 of 36 submissions pass verification. Selecting the lowest-perplexity agent design per task yields an average 3.4% perplexity increase over BF16 and 77.33% downstream accuracy, compared with 79.56% for BF16. Recorded experiments reveal how agents achieve these outcomes and where improvement breaks down. Agents lower PPL in 21 of 24 revisions at nearly fixed storage; complete-model feedback reverses 17 of 79 prior local rankings; and generation and inference tests expose failures that likelihood alone misses. The results identify representation design, experimental judgment, and capability validation as central to agent-driven model compression, a measurable component of efficiency-oriented self-improvement.

cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.10356

SepLRM: A Foundation Model Paradigm for Recommendation

Di Chen, Xijun Xiao, Qin Ren, Weijiang Lai, Zhijian Pan, Zhentao Pan, Pengfei Nie, Cheng Yuan 等 26 位作者

原文摘要

Large recommendation models (LRMs) improve recommendation performance by scaling behavior sequence length, model capacity, and training compute. However, the return on investment (ROI) of continued scaling deteriorates as marginal gains diminish and long-sequence encoding costs rise, especially when this computation is repeatedly executed by distinct models for diverse recommendation stages and business scenarios. To address this issue, we propose SepLRM, a recommendation foundation model (RFM) paradigm that identifies reusable sequence encoding and reorganizes multiple end-to-end (E2E) LRMs into one RFM and multiple independently optimized vertical models (VMs). The RFM adopts an encoder-decoder architecture and exports its final-layer encoder state as the shared key-value (KV) cache that is independent of downstream task-offered candidates. Unlike candidate-specific logits or aggregated user embeddings, the KV cache preserves fine-grained behavioral information. Separation at this native KV interface also allows each VM to construct candidate-conditioned queries while retaining the same query-KV interaction as its E2E counterpart, enabling high-fidelity capability transfer. To ensure both transfer effectiveness and system efficiency, we co-design cache compression, the training and serving workflows, and LG-Store, a distributed storage system for KV caches. Experiments on three large-scale datasets show that SepLRM achieves state-of-the-art offline performance with a transfer rate of approximately 100%. It also demonstrates strong transferability across model iterations, recommendation stages, and business scenarios. Deployment in two large-scale industrial scenarios substantially improves the ROI of sequence scaling and delivers consistent gains in online A/B tests at ByteDance.

cs.IR

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.09004

Sharing the Road: LLM-Based Bargaining in Mixed-Autonomy Traffic

Lingxiao Zhou, Di Chen, Jia Li

原文摘要

Mixed-autonomy traffic involves heterogeneous driving populations competing for shared road space. Previous studies have shown the existence and emergence of collective rationality using game-theoretic and reinforcement learning approaches, which rely on explicitly specified behavioral assumptions or reward functions for driving agents. However, how bargaining behaviors between driving agents evolve and lead to collective outcomes based on richer agent characteristics and language-based reasoning remains less understood. To fill this gap, we develop a bargaining framework based on large language models (LLMs) in which population-level autonomous vehicle (AV) and human-driven vehicle (HV) agents negotiate road-space allocations under heterogeneous objectives and private information. Through controlled experiments varying traffic composition, agent objectives, and traffic-domain knowledge, we examine bargaining behavior, coordination, and resource utilization. Results show that negotiated allocations respond to relative traffic density without strictly following proportional sharing, while density-balanced states exhibit greater coordination difficulty. Fairness-oriented objectives substantially improve agreement rates but do not necessarily yield equally high road-space utilization. Furthermore, domain knowledge affects bargaining outcomes differently across objective configurations, shaping both coordination and the role-dependent effects of fairness. These findings demonstrate the potential of LLM-based bargaining for modeling and designing interactions among driving agents in mixed-autonomy traffic.

cs.MA

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2603.23481

VTAM: Efficient Visuo-Tactile World-Action Models for Contact-Rich Manipulation without Tactile Pretraining

Haoran Yuan, Weigang Yi, Zhenyu Zhang, Wendi Chen, Yuchen Mo, Jiashi Yin, Xinzhuo Li, Yue Luo 等 14 位作者

原文摘要

Tactile vision-language-action models (VLAs) and tactile world-action models (WAMs) are a promising new paradigm for contact-rich manipulation, but existing models either underutilize touch, rely on large-scale tactile pretraining, or incur high inference latency due to iterative video generation. We present VTAM, an efficient visuo-tactile world-action model for contact-rich manipulation that requires no tactile or action pretraining. RGB and tactile observations share a frozen video VAE and the pretrained video transformer weights. VTAM first adapts the video backbone to jointly predict future visual and tactile latents, then fine-tunes it end-to-end with a randomly initialized action expert that predicts actions, proprioceptive states, and a light flow-force proxy computed from tactile optical flow. Trained only on task demonstrations, VTAM outperforms all baselines on four real-robot tasks across two platforms with different tactile sensors, outperforms all baselines without tactile pretraining in average success on eight UniVTAC simulation tasks, and is comparable to tactile VLAs and WAMs pretrained on over 30K hours of largely visuo-tactile robot data (83.4% vs. 84.5% for N0-TWAM). Without full video denoising at inference, VTAM is approximately 6x and 16x faster than N0-TWAM and LingBot-VA without compilation, with compiled latency comparable to Fast-WAM. Ablations show that feeding frozen-VAE tactile features directly to the action expert, bypassing tactile world modeling, fails on every tested task, that Stage I visuo-tactile world-model fine-tuning improves UniVTAC success on most tasks, and that flow-force proxy keeps the policy from fully ignoring touch. The tactile encoder transfers zero-shot across GelSight Mini and Xense Photon sensors.

cs.RO · cs.AI · cs.CV · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.06841

ServerBench：你的编码智能体如何设计专用 LLM 服务系统？

ServerBench: How can your coding agents design specialized LLM serving systems?

Xinrui Zhong, Zhiming Wang, Yongqi Liu, Zhaohui Wang, Shanglin Yang, Zikun Li, Zhuoming Chen, Beidi Chen

AI 速读 基于摘要

提出 ServerBench 基准，用于评估编码智能体能否端到端设计并交付经过验证的专用 LLM 服务系统，并以“每美元延迟合格 token 价值”作为统一目标。

cs.AI

arXiv ↗PDF ↗
查看原文摘要

Large language model (LLM) coding agents develop software through experimentation, offering an opportunity to design the serving systems supporting their own operation. Effective serving requires coordinating implementation and operating decisions under session dependencies, latency constraints, and correctness requirements. We introduce ServerBench, a benchmark for agentic design of complete, specialized LLM serving systems. Agents experiment against a common objective, dollars per dollar: latency-qualified token value per dollar of GPU cost. Stateful replay preserves session dependencies, and independent verification checks the submitted system. Across 11 coding-agent models and three deployments, Claude Sonnet 5.5 achieves 2.97 dollars per dollar on Gemma E4B, $4.87\times$ the vLLM reference; 22 of 33 runs yield verified systems, and 13 exceed the strongest baseline. Our analysis identifies three findings: high-value designs coordinate capacity improvements with load selection; agents interpret performance feedback through workload-relevant experiments; and reliable delivery requires verification of the execution paths introduced by optimization. Overall, ServerBench supports systematic study of how agents translate service objectives and feedback into effective, verified serving systems for given task specifications.

新论文 · v1arXiv:2611.02797

SelfImprove-Bench: Evaluating Generalization in Self-Improving Agents

Bangrui Xu, Jie Yang, Sidi Chen, Xuanhe Zhou, Jiantao Qiu, Qianyu Yang, Zijie Wang, Fan Wu 等 9 位作者

原文摘要

Recursively self-improving (RSI) systems are increasingly viewed as a promising means of accelerating progress toward artificial general intelligence. They aim to autonomously improve their capabilities by refining their working pipelines and recursively improving the strategies that guide subsequent refinements, with gains generalizing to unseen scenarios.However, existing evaluations lack reliable metrics to distinguish capability gains from scenario fitting, sustainable and extensible protocols for unseen tasks and difficulties, and fine-grained attribution of improvement to evolution traces. To this end, we propose SelfImprove-Bench,an RSI evaluation framework that organizes tasks into three primary capabilities (i.e., Exploring, Planning, Reasoning) as scenarios for guiding model evolution. We instantiated the tasks on games for controllability and reproducibility, and pairs them with scalable generation and orchestration driven by task templates and difficulty coefficients. Beyond task-level scores, SelfImprove-Bench organizes evaluation into generalization ladder levels G0 to G4 that measure how far gains transfer from training conditions. Experiments reveal that the evolutionary gains of existing RSI methods are inconsistent across generalization levels.

cs.AI · cs.CL · cs.LG · cs.MA

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38389

Visually Irrelevant Hallucinations in Medical Vision-Language Models

Ke Zou, Nan Zhou, Tao Yang, Meng Liu, Linchao He, Zehui Liao, Ling Huang, Guangren Huang 等 12 位作者

原文摘要

Medical vision-language models (VLMs) show strong performance in radiology report generation (RRG) and medical visual question answering (Med-VQA), producing free-form clinical descriptions and answers from medical images. However, they may generate clinically meaningful statements that lack visual evidence or contradict image content. We refer to these errors as visually irrelevant hallucinations and investigate them at the phrase level. We propose CLEAR (Counterfactual Localized Evaluation of Alignment and Response), a model-agnostic post-hoc framework for phrase-level hallucination risk estimation. CLEAR decomposes generated responses into visually assessable clinical phrases. For each phrase, phrase-conditioned grounding identifies a candidate visual region, which is selectively perturbed by a localized counterfactual intervention. CLEAR characterizes the intervention using two complementary auditing cues: response cues, capturing semantic and clinically meaningful changes in the generated response, and alignment cues, measuring changes in image-phrase compatibility. These cues are normalized and jointly modeled to estimate phrase-level hallucination risk. Across four datasets spanning RRG and Med-VQA, CLEAR outperforms state-of-the-art hallucination detection baselines across multiple medical and general-purpose VLMs. On MIMIC-CXR, CLEAR improves AUPRC from 85.10% to 90.23%. Its effectiveness is further validated on a GT-derived controlled VIH benchmark. These results establish visually irrelevant hallucination as an important fine-grained reliability failure mode in medical VLMs and demonstrate the value of model-agnostic visual-support auditing for more trustworthy multimodal clinical generation.

cs.CV

查看论文
arXiv ↗PDF ↗
版本更新 · v4arXiv:2606.13102Berkeley BAIR

FTP-1: A Generalist Foundation Tactile Policy Across Tactile Sensors for Contact-Rich Manipulation

Chengbo Yuan, Zicheng Zhang, Mingjie Zhou, Wendi Chen, Yi Wang, Zhuoyang Liu, Dantong Niu, Shuo Wang 等 17 位作者

原文摘要

Despite the success of vision-based generalist robotic policies, existing tactile-based policies remain tied to fixed embodiments and sensor setups. This is because tactile signals are highly heterogeneous across hardware, making cross-sensor generalization difficult. We present FTP-1,the first generalist foundation tactile policy pretrained to acquire transferable tactile manipulation abilities across diverse sensors and embodiments. FTP-1 supports varied tactile inputs, including image-, array-, and state-based signals, by using heterogeneous encoders to project them into unified morphology-aware latent tokens that are jointly modeled by a shared tactile Transformer expert. Pretrained on around 3,000 hours of tactile manipulation data aggregated from 26 data sources, spanning human and robot demonstrations across 21 sensors, FTP-1 learns tactile skills that transfer beyond the sensors seen during pretraining. Across downstream finetuning experiments spanning 5 hardware configurations, FTP-1 improves contact-rich manipulation on seen sensor setups by +17.2% and, surprisingly, transfers to two previously unseen tactile-sensor setups, achieving a +31% gain in success rate. FTP-1 establishes the first unified foundation baseline for tactile manipulation, providing future tactile policies with a shared model-level starting point. Pretrained models, datasets, training code and more visualization at https://ftp1-policy.github.io.

cs.RO

查看论文
arXiv ↗PDF ↗
首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1447ACL Anthology · ACL 2026

A Survey of Inductive Reasoning for Large Language Models

Kedi Chen, Dezhao Ruan East China Normal University, Yuhao Dan, Yaoting Wang Fudan University, Siyu Yan, Xuecheng Wu, Yinqi Zhang Xiaohongshu, Qin Chen East China Normal University 等 15 位作者

原文摘要

Reasoning is an important task for large language models (LLMs). Among all the reasoning paradigms, inductive reasoning is one of the basic types, which is characterized by its particular-to-general thinking process and the non-uniqueness of its answers. The inductive mode is crucial for knowledge generalization and aligns better with human cognition, so it is a fundamental mode of learning, hence attracting increasing interest. Despite the importance of inductive reasoning, there is no systematic summary of it. Therefore, this paper presents the first comprehensive survey of inductive reasoning for LLMs. First, methods for improving inductive reasoning are categorized into three main areas: post-training enhancement, test-time exploration, and data augmentation. Then, current benchmarks of inductive reasoning are summarized, and a unified sandbox-based evaluation approach with the observation coverage metric is derived. Finally, we offer some analyses regarding the source of inductive ability and how simple model architectures and data help with inductive tasks, providing a solid foundation for future research.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.699ACL Anthology · ACL 2026

MedVerse: Efficient and Reliable Medical Reasoning via DAG-Structured Parallel Execution

Jianwen Chen University of North Carolina at Chapel Hill, Xinyu Yang Carnegie Mellon University, Peng Xia Department of Computer Science, University of North Carolina at Chapel Hill, Arian Azarang University of North Carolina at Chapel Hill, Yueh Z Lee University of North Carolina at Chapel Hill, Gang Li, Hongtu Zhu, Yun Li University of North Carolina at Chapel Hill 等 10 位作者

原文摘要

Large language models (LLMs) have demonstrated strong performance and rapid progress in a wide range of medical reasoning tasks.However, their sequential autoregressive decoding forces inherently parallel clinical reasoning, such as differential diagnosis, into a single linear reasoning path, limiting both efficiency and reliability for complex medical problems.To address this, we propose MedVerse, a reasoning framework for complex medical inference that reformulates medical reasoning as a parallelizable directed acyclic graph (DAG) process based on Petri Net theory.The framework adopts a full-stack design across data, model architecture, and system execution.For data creation, we introduce the MedVerse Curator, an automated pipeline that synthesizes knowledge-grounded medical reasoning path and transforms them into Petri Net–structured representations.At the architectural level, we propose a topology-aware attention mechanism with adaptive position indices that supports parallel reasoning while preserving logical consistency.Systematically, we develop a customized inference engine that supports parallel execution without additional overhead.Empirical evaluations show that MedVerse improves strong general-purpose LLMs by up to 8.9%. Compared to specialized medical LLMs, MedVerse achieves comparable performance with improved clinical reliability, while delivering a 1.3\times reduction in inference latency and a 1.7\times increase in generation throughput, enabled by its parallel decoding capability.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.707ACL Anthology · ACL 2026

When "Correct" Is Not Safe: Can We Trust Functionally Correct Patches Generated by Code Agents?

Yibo Peng, James Song, Lei Li, Xinyu Yang Carnegie Mellon University, Mihai Christodorescu Google, Ravi Mangal Colorado State University, Corina Pasareanu, Haizhong Zheng CMU, Carnegie Mellon University 等 9 位作者

原文摘要

Code agents are increasingly trusted to autonomously fix bugs on platforms such as GitHub, yet their security evaluation focuses almost exclusively on functional correctness. In this paper, we reveal a novel type of threat to real-world code-agents: functionally correct yet vulnerable (FCV) patches, which pass all test cases but contain vulnerable code. With our proposed FCV-Attack, we demonstrate that SOTA LLMs (e.g., ChatGPT and Claude) and agent scaffolds (e.g., SWE-agent and OpenHands) are all vulnerable to this FCV threat; across 12 agent-model combinations on SWE-Bench, the attack only requires black-box access and a single query to the code agent to perform the attack. For example, for CWE-538 (information exposure vulnerability), the FCV-Attack attains an attack success rate of 40.7% on GPT-5 Mini + OpenHands. Our results reveal an important security threat overlooked by current evaluation paradigms and urge the development of security-aware defenses for code agents.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表