arXiv · Software Engineering· Chenyang Yang, Hao He, Ti-Chung Cheng, Jenna Butler, Emerson Murphy-Hill, Carmen Badea, Christian Bird·· 3 hr agoAI score54
CRBugBench：代码审查 Agent 缺陷检出基准
(When) Should a Software Factory be Lights-off? Benchmarking Code Review Agents on Bug Detection
AI brief
研究者构建 CRBugBench，用 1,000 个开源 GitHub 仓库审查任务和 800 个某大型公司内部仓库任务评估代码审查 Agent 能否发现人工确认的缺陷。
Source: arXiv · Software Engineering · arxiv.org