TestEvo-Bench - Test Update: leaderboard
TestEvo-Bench (2026) packages 1,255 test and code co-evolution tasks mined from the commit histories of 152 Java Maven projects, each with an executable environment, and runs coding agents on them. In the 509 test-update tasks the agent adapts existing tests that the code change broke; score is the macro-averaged share that pass on the new revision, in percent, higher is better.
Metric: Success rate (%; per-task share of failing existing test methods that the agent adapts so they pass on the new code revision, macro-averaged over 509 test-update tasks from 152 Java projects; per-task budget of 3 US dollars). Source: arxiv.org. Saturation forecast: Around February 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 73.9 | #94 |
| 2 | Claude Opus 4.7 | 65.6 | #79 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=testevo-bench-test-update · How It Works · Data refreshed daily, snapshot 2026-11-13.