Formal Proof Generation on Lean 4 (val)
49.8Pass@1Ours
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| OursModel type=Fine-tuned2026.03 | 49.8 | 52.7 | 54.1 | |
| Leanabell2026.03 | 25.4 | 36 | 39.9 | |
| STP2026.03 | 23.7 | 33.2 | 37.5 | |
| Deepseek-v22026.03 | 14.1 | 30 | 36.2 | |
| Goedel-v22026.03 | 14 | 31.1 | 38.2 | |
| Kimina-distill2026.03 | 9.8 | 25.3 | 35.8 |