Mathematical Problem Solving on AIME'24 (test)
26.67Average@16 AccuracyPRM-CoT (Process-Aware)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| PRM-CoT (Process-Aware)Base Policy=Qwen2.5-Math-7B2025.12 | 26.67 | 30 | 40 | 50 | |
| RLVR (Ground Truth)Base Policy=Qwen2.5-Math-7B2025.12 | 22.71 | 26.67 | 36.67 | 50 | |
| PRM (Process-Aware)Base Policy=Qwen2.5-Math-7B2025.12 | 18.12 | 26.67 | 40 | 46.67 | |
| SFT (Baseline)Base Policy=Qwen2.5-Math-7B2025.12 | 9.58 | 3.3 | 23.33 | 26.67 |