分布式系统bug修复遇上AI智能体，DDBench双条件评测揭示其真实能力边界。
arXiv:2608.14863v1 Announce Type: cross Abstract: LLM-based coding agents have advanced rapidly on single-process SWE tasks, with frontier models now …
这份研究直击LLM修复代理的验证盲区：测试通过并不等于bug真被修复
arXiv:2607.28871v1 Announce Type: cross Abstract: When a repair agent runs a test and sees it pass, the result is treated as evidence about the report…
针对LLM生成C代码的安全漏洞，提出工具引导的检索增强修复方法，兼顾效率与可靠性。
arXiv:2607.21641v1 Announce Type: cross Abstract: Large language models can generate C code from natural-language descriptions, but resulting programs…
执行引导的Linux内核崩溃修复智能体，已被ICML 2026接收，为syzkaller发现的数千个内核崩溃提供快速修复方案。
arXiv:2504.20412v3 Announce Type: replace-cross Abstract: Fuzzing frameworks like syzkaller have uncovered thousands of Linux kernel crashes, many of …
自博弈训练让AI修复代码能力再进一步，锚定策略稳定训练过程，ICML论文值得读。
arXiv:2607.03523v1 Announce Type: cross Abstract: Code repair is an important capability for language models (LMs): given a buggy program and unit tes…
新框架评估文件级问题定位对仓库级LLM修复的影响，为更精准的代码修复提供理论支撑
arXiv:2606.30963v1 Announce Type: cross Abstract: Repository-grounded automated repair is often reported as a single end-to-end capability, which hide…
一套分层Oracle框架，专门识别LLM辅助Terraform修复中的“假修复”，保障基础设施代码可信度。
arXiv:2606.26590v1 Announce Type: new Abstract: Security misconfigurations in Terraform Infrastructure-as-Code are a growing risk in cloud deployments…
结构化诊断定位框架SHERLOC，精准提升代码修复智能体的错误定位效率与准确性，值得关注。
arXiv:2606.24820v1 Announce Type: new Abstract: LLM agents solve repository-level coding tasks through multi-turn tool use, but utilize half their bud…
新方法利用大语言模型结合上下文信息，智能修复软件回归错误，经验评估验证有效性
arXiv:2506.13182v2 Announce Type: replace-cross Abstract: [...] Since then, various APR approaches, especially those leveraging the power of large lan…
可视化AI智能体修复代码的完整轨迹，揭示其决策过程与修复路径。
arXiv:2606.22110v1 Announce Type: cross Abstract: LLM-based automated program repair (APR) agents generate patches to fix software bugs with minimal h…
一款GitHub应用，自动建议代码修复以解决转换失败问题，按影响用户排序，高效定位并修复错误。
Hi Hacker News! I’m Rashid, a first-year student at UT Austin and the person leading development of Rejourney, a web/app analytics and session replay …
让编码智能体读懂仓库的“潜规则”？Probe-and-Refine调优法挖掘AGENTS.md指导价值，显著提升代码修复准确率。
arXiv:2606.20512v1 Announce Type: cross Abstract: LLM-based coding agents need higher-level operational knowledge about a repository (which files hous…
编程竞赛中的LLM不止解题：新基准UOJ-Bench全面评估代码生成、攻击与修复能力
arXiv:2606.12864v1 Announce Type: cross Abstract: Despite strong performance in competitive programming, the role of Large Language Models (LLMs) in s…
AI生成的PR修复请求为何频频遭拒？基于AIDev数据集给出拒绝模式与改进方向。
arXiv:2606.13468v1 Announce Type: cross Abstract: AI coding agents are increasingly used to generate pull requests (PRs) that propose code fixes in so…
SWE-bench Verified上创97%新高！开源Subscription-token Agent超越顶级模型，代码修复能力惊人。
Article URL: https://github.com/kimjune01/swebench-verified Comments URL: https://news.ycombinator.com/item?id=48259575 Points: 2 # Comments: 0
首个评估AI Agent在真实硬件工程中导航、定位、EDA验证与修复能力的基准，揭示软件工程AI迁移至硬件的挑战。
arXiv:2605.15226v1 Announce Type: cross Abstract: We ask whether agentic AI systems built for software engineering transfer to realistic hardware engi…