Math Proof Reward Modeling on OPC (test)
83.1Accuracyo3
Evaluation Results
| Method | Links | |
|---|---|---|
| o3Access=Closed-source2026.02 | 83.1 | |
| Gemini-2.5-FlashAccess=Closed-source2026.02 | 82.7 | |
| ProofRMSize=32B2026.02 | 81.3 | |
| Deepseek-R12026.02 | 80.9 | |
| ProofRMSize=14B2026.02 | 77.4 | |
| ProofRMSize=8B2026.02 | 75.7 | |
| Distill-Qwen3Size=8B2026.02 | 70.8 | |
| Distill-Qwen3Size=32B2026.02 | 70.2 | |
| Distill-Qwen3Size=14B2026.02 | 69.4 |