GitHub ReviewBench code review benchmark preview

GitHub released ReviewBench as a public research preview on October 5. The benchmark tests AI code review agents on 219 real pull requests from public repositories and publishes the corpus, known findings, evaluation method and a leaderboard. Developers can run a reviewer against the same tasks and submit it for evaluation. ReviewBench says its first leaderboard entries were run by its own team and were not verified by the vendors; scores describe this benchmark, not guaranteed performance on other codebases.