Correctness Assessment on ProcessBench (test)
0.24Worst-case Size Distortion (QwenPRM)naive (1)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| naive (1)Score=naive (1)2025.10 | 0.24 | 0.2 | |
| e-score (2)Score=e-score (2)2025.10 | 0.79 | 0.73 | |
| e-score (combined)Score=e-score (combined)2025.10 | 0.94 | 0.91 | |
| e-score (1)Score=e-score (1)2025.10 | 1 | 1.01 | |
| e-score (3)Score=e-score (3)2025.10 | 1.01 | 1.01 | |
| naive (3)Score=naive (3)2025.10 | 1.39 | 0.8 | |
| naive (2)Score=naive (2)2025.10 | 1.89 | 1.3 | |
| p-scoreScore=p-score2025.10 | 7.21 | 7.46 | |
| p-score (randomized)Score=p-score (randomized)2025.10 | 15.8 | 13.96 |