1
Confident and Wrong: Silent Semantic Failures in Coding Agents
编程代理的“虚假成功”比失败更危险——即使通过测试，也可能完全偏离意图，威胁生产环境信任度。
arXiv:2603.25764v3 Announce Type: replace-cross Abstract: As coding agents move into production workflows, teams need to know not only whether an agen…