代码大模型 / AI 编程 - arXivDaily 专题

2601.00477 2026-01-05 cs.CR cs.SE 57%

Security in the Age of AI Teammates: An Empirical Study of Agentic Pull Requests on GitHub

AI队友时代的安全：对GitHub上代理拉取请求的实证研究

Mohammed Latif Siddiq, Xinye Zhao, Vinicius Carvalho Lopes, Beatrice Casey, Joanna C. S. Santos

专题命中软件智能体：coding agent(abstract)；分类 cs.SE

AI总结研究分析了GitHub上代理撰写的PR在安全方面的贡献，发现其主要进行支持性的安全加固活动，且PR被拒绝与复杂性和冗长性相关。

Comments Submitted to Information and Software Technology Journal

2512.24656 2026-01-05 cs.SE 57%

Characterizing Bugs and Quality Attributes in Quantum Software: A Large-Scale Empirical Study

量子软件中缺陷和质量属性的特征化：一项大规模经验研究

Mir Mohammad Yousuf, Shabir Ahmad Sofi

专题命中软件智能体：repository(abstract)；分类 cs.SE

AI总结本研究通过大规模分析量子软件缺陷，揭示了全栈库和编译器的高缺陷率及自动化测试对减少缺陷的重要性。

2512.24636 2026-01-01 cs.SE 57%

How Do Agentic AI Systems Deal With Software Energy Concerns? A Pull Request-Based Study

代理AI系统如何处理软件能耗问题？基于拉取请求的研究

Tanjum Motin Mitul, Md. Masud Mazumder, Md Nahidul Islam Opu, Shaiful Chowdhury

专题命中软件智能体：coding agent(abstract)；分类 cs.SE

AI总结本文研究了代理AI系统在生成软件时对能耗问题的意识，通过分析拉取请求发现，尽管构建和运行这些系统耗能高，但生成的软件制品表现出一定的能耗意识，但优化相关的PR因影响可维护性而被接受较少。

2512.22469 2025-12-30 cs.SE 57%

GraphLocator: Graph-guided Causal Reasoning for Issue Localization

GraphLocator: 图引导的因果推理用于问题定位

Wei Liu, Chao Peng, Pengfei Gao, Aofan Liu, Wei Zhang, Haiyan Zhao, Zhi Jin

专题命中软件智能体：repository(abstract)；分类 cs.SE

AI总结 GraphLocator通过因果图发现和动态解构解决软件问题定位中的因果不匹配和一对一不匹配问题，提升定位准确性和下游任务性能。

2512.22087 2025-12-29 cs.CL 57%

Context as a Tool: Context Management for Long-Horizon SWE-Agents

上下文作为工具：长周期SWE代理的上下文管理

Shukai Liu, Jian Yang, Bo Jiang, Yizhi Li, Jinyang Guo, Xianglong Liu, Bryan Dai

机构 * Beihang University（北航大学）； Manchester（曼彻斯特）； Ubiquant

专题命中软件智能体：repository(abstract)；分类 cs.CL

AI总结 CAT通过整合上下文管理工具，提升SWE代理的长周期推理能力，实验表明其在SWE-Bench-Verified上达到57.6%的解决率，显著优于现有方法。

2512.21919 2025-12-29 cs.CL 57%

SWE-RM: Execution-free Feedback For Software Engineering Agents

SWE-RM：无执行反馈用于软件工程代理

KaShun Shum, Binyuan Hui, Jiawei Chen, Lei Zhang, X. W., Jiaxi Yang, Yuzhen Huang, Junyang Lin, Junxian He

机构 * The Hong Kong University of Science and Technology（香港科技大学）； Alibaba Group（阿里巴巴集团）

专题命中软件智能体：coding agent(abstract)；分类 cs.CL

AI总结 SWE-RM通过混合专家架构提升软件工程代理在TTS和RL中的性能，达到新的开源模型最佳水平。

Comments 21 pages

2511.02864 2025-12-23 cs.NE cs.AI math.CA math.CO math.MG 57%

Mathematical exploration and discovery at scale

大规模数学探索与发现

Bogdan Georgiev, Javier Gómez-Serrano, Terence Tao, Adam Zsolt Wagner

机构 * Google DeepMind（谷歌DeepMind）； Department of Mathematics, Brown University（布朗大学数学系）； Institute for Advanced Study（高级研究院）； UCLA Department of Mathematics（加州大学洛杉矶分校数学系）

专题命中软件智能体：coding agent(abstract)；分类 cs.AI

AI总结 AlphaEvolve通过进化搜索发现数学构造，提升数学问题解决效率

Comments 81 pages, 35 figures

2512.16822 2025-12-19 cs.LG 57%

MEPIC: Memory Efficient Position Independent Caching for LLM Serving

MEPIC：内存高效的无位置依赖缓存用于大语言模型服务

Qian Wang, Zahra Yousefijamarani, Morgan Lindsay Heisler, Rongzhi Gu, Bai Xiaolong, Shan Yizhou, Wei Zhang, Wang Lan, Ying Xiong, Yong Zhang, Zhenan Fan

机构 * Huawei Technologies Canada Co., Ltd.（华为技术加拿大有限公司）； Huawei Technologies Co., Ltd.（华为技术有限公司）

专题命中软件智能体：coding agent(abstract)；分类 cs.LG

AI总结 MEPIC通过内存高效的无位置依赖缓存技术，实现跨请求和批次的片段KV重用，显著减少HBM使用量，提升大语言模型服务性能。

2512.16538 2025-12-19 cs.CR cs.LG 57%

A Systematic Study of Code Obfuscation Against LLM-based Vulnerability Detection

对抗基于大语言模型的漏洞检测的代码混淆系统研究

Xiao Li, Yue Li, Hao Wu, Yue Zhang, Yechao Zhang, Fengyuan Xu, Sheng Zhong

机构 * National Key Lab for Novel Software Technology, Nanjing University（新型软件技术国家重点实验室，南京大学）； Shandong University（山东大学）； Nanyang Technological University（南洋理工大学）

专题命中软件智能体：coding agent(abstract)；分类 cs.LG

AI总结本文系统研究了对抗基于大语言模型的漏洞检测的代码混淆技术，评估了不同混淆方法对LLM检测性能的影响，并提出了未来改进方向。

2512.06031 2025-12-09 physics.flu-dyn cs.LG physics.comp-ph 57%

Multi-resolution Physics-Aware Recurrent Convolutional Neural Network for Complex Flows

多分辨率物理感知递归卷积神经网络用于复杂流体

Xinlun Cheng, Joseph Choi, H. S. Udaykumar, Stephen Baek

机构 * School of Data Science, University of Virginia（数据科学学院，弗吉尼亚大学）； Department of Mechanical Engineering, University of Iowa（机械工程系，爱荷华大学）； Department of Mechanical and Aerospace Engineering, University of Virginia（机械与航空航天工程系，弗吉尼亚大学）

专题命中软件智能体：repository(abstract)；分类 cs.LG

AI总结 MRPARCv2通过多分辨率架构和物理约束提升复杂流体模拟精度与效率，显著优于单分辨率模型。

Journal ref APL Mach. Learn. 3, 046110 (2025)

2512.00611 2025-12-02 cs.CL 57%

Prism: A Minimal Compositional Metalanguage for Specifying Agent Behavior

Prism：一种最小的组合金属言用于指定代理行为

Franck Binard, Vanja Kljajevic

机构 * Deloitte AI（德勤人工智能）； University of Oslo（奥斯陆大学）

专题命中软件智能体：software agent(abstract)；分类 cs.CL

AI总结 Prism通过组合性金属言为代理行为提供最小化规格，结合领域特定词汇与工具，实现可检查和可执行的策略定义。

2506.22390 2025-11-25 cs.SE 57%

What Characteristics Make ChatGPT Effective for Software Issue Resolution? An Empirical Study of Task, Project, and Conversational Signals in GitHub Issues

哪些特征使ChatGPT在软件问题解决中有效？对GitHub问题中任务、项目和对话信号的实证研究

Ramtin Ehsani, Sakshi Pathak, Esteban Parra, Sonia Haiduc, Preetha Chatterjee

专题命中软件智能体：code generation(abstract)；分类 cs.SE

AI总结本文研究了ChatGPT在软件问题解决中的有效性特征，发现其在代码生成和工具推荐方面表现较好，但代码解释能力较弱，且更受欢迎的项目和经验丰富的开发者从中受益更多。

Comments Accepted for publication in Empirical Software Engineering (EMSE), 2025

Journal ref Empirical Software Engineering, 31, 22 (2026)

2511.02352 2025-11-19 cs.SE 57%

SWE-Sharp-Bench: A Reproducible Benchmark for C# Software Engineering Tasks

Sanket Mhatre, Yasharth Bajpai, Sumit Gulwani, Emerson Murphy-Hill, Gustavo Soares

专题命中软件智能体：coding agent(abstract)；分类 cs.SE

2506.16444 2025-11-12 cs.CL cs.AR cs.DB 57%

REIS: A High-Performance and Energy-Efficient Retrieval System with In-Storage Processing

Kangqi Chen, Andreas Kosmas Kakolyris, Rakesh Nadig, Manos Frouzakis, Nika Mansouri Ghiasi, Yu Liang, Haiyu Mao, Jisung Park, Mohammad Sadrosadati, Onur Mutlu

机构 * ETH Zürich（苏黎世联邦理工学院）； King’s College London（伦敦国王学院）； POSTECH

专题命中软件智能体：repository(abstract)；分类 cs.CL

Comments Extended version of our publication at the 52nd International Symposium on Computer Architecture (ISCA-52), 2025

2409.04181 2025-11-12 cs.CL 57%

Combining LLMs and Knowledge Graphs to Reduce Hallucinations in Question Answering

Larissa Pusch, Tim O. F. Conrad

机构 * Zuse Institute Berlin（柏林泽尼茨研究所）

专题命中软件智能体：repository(abstract)；分类 cs.CL

2510.26328 2025-10-31 cs.LG 57%

Agent Skills Enable a New Class of Realistic and Trivially Simple Prompt Injections

David Schmotz, Sahar Abdelnabi, Maksym Andriushchenko

机构 * ELLIS Institute Tübingen（图宾根ELLIS研究所）； MPI for Intelligent Systems Tübingen（图宾根智能系统研究所）； AI Center（人工智能中心）

专题命中软件智能体：coding agent(abstract)；分类 cs.LG

2510.09804 2025-10-14 q-bio.OT cs.SE 57%

Rapid Development of Omics Data Analysis Applications through Vibe Coding

Jesse G. Meyer

专题命中软件智能体：coding agent(abstract)；分类 cs.SE

2507.09199 2025-10-07 cs.SE 57%

Back to the Basics: Rethinking Issue-Commit Linking with LLM-Assisted Retrieval

Huihui Huang, Ratnadira Widyasari, Ting Zhang, Ivana Clairine Irsan, Jieke Shi, Han Wei Ang, Frank Liauw, Eng Lieh Ouh, Lwin Khin Shar, Hong Jin Kang, David Lo

专题命中软件智能体：repository(abstract)；分类 cs.SE

2509.21427 2025-10-07 cs.SE 57%

Extracting Conceptual Knowledge to Locate Software Issues

Ying Wang, Wenjun Mao, Chong Wang, Zhenhao Zhou, Yicheng Zhou, Wenyun Zhao, Yiling Lou, Xin Peng

专题命中软件智能体：repository(abstract)；分类 cs.SE

2509.18415 2025-09-24 cs.CR cs.AI 57%

Context Lineage Assurance for Non-Human Identities in Critical Multi-Agent Systems

Sumana Malkapuram, Sameera Gangavarapu, Kailashnath Reddy Kavalakuntla, Ananya Gangavarapu

机构 * Netflix ； Zoetis ； Ravenna

专题命中软件智能体：software agent(abstract)；分类 cs.AI

2509.10844 2025-09-16 cs.CL 57%

GAPrune: Gradient-Alignment Pruning for Domain-Aware Embeddings

Yixuan Tang, Yi Yang

机构 * The Hong Kong University of Science and Technology（香港科学与技术大学）

专题命中软件智能体：coding agent(abstract)；分类 cs.CL

Comments https://github.com/yixuantt/GAPrune

2509.00883 2025-09-03 cs.DC cs.AI 57%

Accelerating Latency-Critical Applications with AI-Powered Semi-Automatic Fine-Grained Parallelization on SMT Processors

Denis Los, Igor Petushkov

专题命中软件智能体：coding agent(abstract)；分类 cs.AI

Journal ref International Journal of Open Information Technologies, vol. 13, no. 9, pp. 129-134, 2025

2508.17511 2025-08-26 cs.AI 57%

School of Reward Hacks: Hacking harmless tasks generalizes to misaligned behavior in LLMs

Mia Taylor, James Chua, Jan Betley, Johannes Treutlein, Owain Evans

专题命中软件智能体：coding agent(abstract)；分类 cs.AI

Comments 42 pages, 26 figures

2508.17161 2025-08-26 cs.SE 57%

What Developers Ask to ChatGPT in GitHub Pull Requests? an Exploratory Study

Julyanara R. Silva, Carlos Eduardo C. Dantas, Marcelo A. Maia

专题命中软件智能体：code generation(abstract)；分类 cs.SE

Comments 12 pages, 3 figures

Journal ref 12th Workshop on Software Visualization, Evolution and Maintenance, 2024

2508.16277 2025-08-25 cs.AI cs.HC 57%

The next question after Turing's question: Introducing the Grow-AI test

Alexandru Tugui

专题命中软件智能体：software agent(abstract)；分类 cs.AI

Comments 9th International Conference on Inventive Systems and Control ICISC 2025

2508.12649 2025-08-20 cs.SE 57%

ChangePrism: Visualizing the Essence of Code Changes

Lei Chen, Michele Lanza, Shinpei Hayashi

专题命中软件智能体：repository(abstract)；分类 cs.SE

Comments (C) 2025 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works

2504.03693 2025-07-04 cs.SE cs.MA 57%

Agentic Business Process Management: Practitioner Perspectives on Agent Governance in Business Processes

Hoang Vu, Nataliia Klievtsova, Henrik Leopold, Stefanie Rinderle-Ma, Timotheus Kampik

专题命中软件智能体：software agent(abstract)；分类 cs.SE

Comments Accepted for Responsible BPM 2025, 15 pages including references

2503.11069 2025-06-24 cs.AI cs.HC 57%

API Agents vs. GUI Agents: Divergence and Convergence

Chaoyun Zhang, Shilin He, Liqun Li, Si Qin, Yu Kang, Qingwei Lin, Saravan Rajmohan, Dongmei Zhang

机构 * Microsoft（微软）

专题命中软件智能体：software agent(abstract)；分类 cs.AI

2506.15346 2025-06-19 cs.LG 57%

Acoustic Waveform Inversion with Image-to-Image Schrödinger Bridges

A. S. Stankevich, I. B. Petrov

机构 * Moscow Institute of Physics and Technology（莫斯科物理技术学院）

专题命中软件智能体：repository(abstract)；分类 cs.LG

Comments Submitted to "Computational Mathematics And Mathematical Physics", ISSN 1555-6662, issue 8, August 2025

2410.16464 2025-06-18 cs.CL cs.MA 57%

Beyond Browsing: API-Based Web Agents

Yueqi Song, Frank Xu, Shuyan Zhou, Graham Neubig

机构 * Carnegie Mellon University（卡内基梅隆大学）

专题命中软件智能体：coding agent(abstract)；分类 cs.CL

Comments 20 pages, 8 figures

1. 软件智能体 1096 篇

Security in the Age of AI Teammates: An Empirical Study of Agentic Pull Requests on GitHub

Characterizing Bugs and Quality Attributes in Quantum Software: A Large-Scale Empirical Study

How Do Agentic AI Systems Deal With Software Energy Concerns? A Pull Request-Based Study

GraphLocator: Graph-guided Causal Reasoning for Issue Localization

Context as a Tool: Context Management for Long-Horizon SWE-Agents

SWE-RM: Execution-free Feedback For Software Engineering Agents

Mathematical exploration and discovery at scale

MEPIC: Memory Efficient Position Independent Caching for LLM Serving

A Systematic Study of Code Obfuscation Against LLM-based Vulnerability Detection

Multi-resolution Physics-Aware Recurrent Convolutional Neural Network for Complex Flows

Prism: A Minimal Compositional Metalanguage for Specifying Agent Behavior

What Characteristics Make ChatGPT Effective for Software Issue Resolution? An Empirical Study of Task, Project, and Conversational Signals in GitHub Issues

SWE-Sharp-Bench: A Reproducible Benchmark for C# Software Engineering Tasks

REIS: A High-Performance and Energy-Efficient Retrieval System with In-Storage Processing

Combining LLMs and Knowledge Graphs to Reduce Hallucinations in Question Answering

Agent Skills Enable a New Class of Realistic and Trivially Simple Prompt Injections

Rapid Development of Omics Data Analysis Applications through Vibe Coding

Back to the Basics: Rethinking Issue-Commit Linking with LLM-Assisted Retrieval

Extracting Conceptual Knowledge to Locate Software Issues

Context Lineage Assurance for Non-Human Identities in Critical Multi-Agent Systems

GAPrune: Gradient-Alignment Pruning for Domain-Aware Embeddings

Accelerating Latency-Critical Applications with AI-Powered Semi-Automatic Fine-Grained Parallelization on SMT Processors

School of Reward Hacks: Hacking harmless tasks generalizes to misaligned behavior in LLMs

What Developers Ask to ChatGPT in GitHub Pull Requests? an Exploratory Study

The next question after Turing's question: Introducing the Grow-AI test

ChangePrism: Visualizing the Essence of Code Changes

Agentic Business Process Management: Practitioner Perspectives on Agent Governance in Business Processes

API Agents vs. GUI Agents: Divergence and Convergence

Acoustic Waveform Inversion with Image-to-Image Schrödinger Bridges

Beyond Browsing: API-Based Web Agents