AI coding agents pass tests but miss review standards, new benchmark shows

Original: SWE-Gate: Passing Functional Tests Is Not Enough for Software Engineering Agents

Source: arxiv.org ↗