Mathematical Reasoning on ProcessBench (OlympiaBench) 1.0 (test)
79.8AccuracyQwen2.5-Math-PRM
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen2.5-Math-PRMModel Size=72B2025.03 | 79.8 | 77.9 | 86 | 71.2 | |
| Greedy EDU PRMModel Size=72B2025.03 | 77.2 | 76.2 | 80.1 | 72.6 | |
| Qwen2.5-Math-PRMModel Size=7B2025.03 | 74.1 | 72.1 | 78.5 | 66.6 | |
| Math-Shepherd PRMModel Size=72B2025.03 | 71 | 74 | 69.1 | 79.6 | |
| Sample EDU PRMModel Size=72B2025.03 | 69.7 | 72.3 | 67 | 78.6 | |
| Omega PRMModel Size=72B2025.03 | 66.1 | 55.3 | 81.6 | 41.8 | |
| Sample EDU PRMModel Size=7B2025.03 | 53.8 | 63.6 | 52.8 | 79.8 | |
| Math-Shepherd PRMModel Size=7B2025.03 | 53.6 | 53.9 | 54.1 | 53.6 | |
| Greedy EDU PRMModel Size=7B2025.03 | 51.7 | 8.3 | 81.5 | 0.4 | |
| Omega PRMModel Size=7B2025.03 | 51.3 | 7.9 | 72.4 | 4.2 |