SWE-Refactor (Java) - Compile and Test Success: leaderboard
Metric: Compile-and-test success rate (%) over the 1,099 instances (the project compiles and its tests pass with the generated code; the printed count times 100/1099); SWE-Refactor: 1,099 pure developer-written method-level refactorings (Extract, Inline, Move, Extract and Move, Move and Rename, Move and Inline Method) mined from 18 Java projects; the model receives the target method, the refactoring type, class and repository context and caller-callee relations; its refactored code is integrated into the project at the original commit; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o Mini (2024-07-18) | 48.86 | #546 |
| 2 | GPT-3.5 Turbo (0125) | 18.11 | #958 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=swe-refactor-java-compile-and-test-success · How It Works · Data refreshed daily, snapshot 2026-10-11.