Mathematical Proof Construction on IMOProofBench
90.85Mean ScoreGPT-5.5
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-5.5Search Strategy=Standalone, Generation temperature=1.0, top-p=0.95, Maximum output length=512K, Judger=GPT-5.42026.06 | 90.85 | |
| Gemini 3.1 ProSearch Strategy=Standalone, Generation temperature=1.0, top-p=0.95, Maximum output length=512K, Judger=GPT-5.42026.06 | 75.71 | |
| M3Search Strategy=Standalone, Generation temperature=1.0, top-p=0.95, Maximum output length=512K, Judger=GPT-5.42026.06 | 67.4 | |
| Opus 4.7Search Strategy=Standalone, Generation temperature=1.0, top-p=0.95, Maximum output length=512K, Judger=GPT-5.42026.06 | 65.85 |