Mathematical Reasoning on ProcessBench MATH 1.0 (test)
88.4AccuracyGreedy EDU PRM
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Greedy EDU PRMModel Size=72B2025.03 | 88.4 | 88.2 | 90.4 | 86.2 | |
| Qwen2.5-Math-PRMModel Size=72B2025.03 | 87.8 | 87.2 | 91.8 | 83 | |
| Qwen2.5-Math-PRMModel Size=7B2025.03 | 82.4 | 82 | 83.9 | 80.2 | |
| Omega PRMModel Size=72B2025.03 | 79.8 | 76.3 | 92.3 | 65 | |
| Math-Shepherd PRMModel Size=72B2025.03 | 77.8 | 80.5 | 72.7 | 90.2 | |
| Sample EDU PRMModel Size=72B2025.03 | 76.4 | 79.5 | 70.9 | 90.6 | |
| Math-Shepherd PRMModel Size=7B2025.03 | 62.9 | 65.9 | 61.5 | 71 | |
| Sample EDU PRMModel Size=7B2025.03 | 59.2 | 68.9 | 55.9 | 89.8 | |
| Omega PRMModel Size=7B2025.03 | 58 | 29.5 | 91.7 | 17.6 | |
| Greedy EDU PRMModel Size=7B2025.03 | 56.2 | 22.9 | 95.6 | 13 |