ReviewBench: An open benchmark for AI code review
GitHub has launched ReviewBench, an open-source benchmark specifically designed to evaluate AI agents performing code reviews. The tool utilizes real-world GitHub pull request data and multi-source ground truth to provide standardized performance metrics.
Verified State Diff
Impact & Verification Analysis
AI researchers, developers building autonomous coding agents, and enterprise engineering teams.
It establishes a common industry standard for measuring the efficacy of AI in the code review lifecycle, accelerating the development of reliable autonomous software engineering tools.
Full Fact Overview
ReviewBench addresses the lack of standardized evaluation frameworks for autonomous code review agents. By leveraging production-aligned metrics and multi-source ground truth, it allows developers and researchers to quantify the accuracy and utility of AI-generated code feedback. This benchmark moves beyond generic coding benchmarks by focusing on the specific context of pull request workflows, including diff analysis and repository-level awareness.