Automated Theorem Proving on FATE-M
98Pass RateAxProverBase
Evaluation Results
| Method | Links | |
|---|---|---|
| AxProverBaseunderlying LLM=Opus 4.5, thinking budget=32k, iterations=50, pass@k=pass@12026.02 | 98 | |
| DeepSeek V2pass@k=pass@642026.02 | 62.7 | |
| REAL-Proverpass@k=pass@642026.02 | 56.7 | |
| Goedel Prover V2pass@k=pass@642026.02 | 48.7 | |
| Kimina Proverpass@k=pass@642026.02 | 36 |