GitHub Blog · AI & ML· Michelle Zhou·· 11 d ago
ReviewBench: An open benchmark for AI code review
ReviewBench: An open benchmark for AI code review
AI summary
We’re launching ReviewBench, a benchmark for code review agents built on representative GitHub pull requests, multi-source ground truth, calibrated evaluation, and production-aligned metrics.
Selection record
Threshold 60Official, first-handFirst 58Second 58
Not admittedSum of both 116 < twice the threshold 120
- Source tier
- Official, first-hand; this tier's threshold is 60
- Pre-filter
- passed：AI代码审查基准与评测方法
A model scores each item twice, independently, against one written standard, out of 100. An item is admitted only when the two scores add up to twice the threshold. The threshold is set per source tier.
Source: GitHub Blog · AI & ML · github.blog