Mathematical Reasoning on MATH 500 (Pass@4, Pass@16, Pass@64)
66.8Pass@4Q-RM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Q-RMBase Model=Qwen3-8B Base2026.04 | 66.8 | 69 | 68.8 | |
| IPVRMBase Model=Qwen3-8B Base2026.04 | 66.8 | 67.6 | 69.2 | |
| DPO-RMBase Model=Qwen3-8B Base2026.04 | 66.6 | 68 | 69.8 | |
| Implicit PRMBase Model=Qwen3-8B Base2026.04 | 66.4 | 68 | 70.2 | |
| EndoRMBase Model=Qwen3-8B Base2026.04 | 60.6 | 61 | 57.6 | |
| IPVRMBase Model=Qwen3-0.6B Base2026.04 | 42 | 46 | 47.6 | |
| DPO-RMBase Model=Qwen3-0.6B Base2026.04 | 41.6 | 42.6 | 42.2 | |
| Q-RMBase Model=Qwen3-0.6B Base2026.04 | 41.4 | 44.4 | 44.6 | |
| Implicit PRMBase Model=Qwen3-0.6B Base2026.04 | 39.6 | 39.6 | 38.6 | |
| EndoRMBase Model=Qwen3-0.6B Base2026.04 | 32 | 33.8 | 38.8 | |
| IPVRMBase Model=Llama3.2-3B2026.04 | 30.6 | 34.6 | 35.8 | |
| Implicit PRMBase Model=Llama3.2-3B2026.04 | 29.8 | 32.8 | 35.4 | |
| DPO-RMBase Model=Llama3.2-3B2026.04 | 29.2 | 31.6 | 34.6 | |
| Q-RMBase Model=Llama3.2-3B2026.04 | 29 | 31.4 | 33.8 | |
| EndoRMBase Model=Llama3.2-3B2026.04 | 25.4 | 26.8 | 27 | |
| IPVRMBase Model=Llama3.2-1B2026.04 | 13.9 | 14.2 | 14.8 | |
| Q-RMBase Model=Llama3.2-1B2026.04 | 13.6 | 14.6 | 13.4 | |
| Implicit PRMBase Model=Llama3.2-1B2026.04 | 13 | 14 | 14.2 | |
| DPO-RMBase Model=Llama3.2-1B2026.04 | 11.2 | 12.6 | 12.6 | |
| EndoRMBase Model=Llama3.2-1B2026.04 | 10.4 | 10.6 | 10 |