Mathematical Reasoning on GSM8K (test) (Hit@1, TOK (x10^3), AC)
82.9Hit@1 AccuracyMASPRM-7B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MASPRM-7BSearch=MCTS(10,3), Agent=Qwen2.5-1.5B-Instruct2025.10 | 82.9 | 5.5 | 34.4 | |
| MASPRM-7BSearch=SBS(5,3), Agent=Qwen2.5-1.5B-Instruct2025.10 | 77.5 | 7.9 | 50 | |
| MASPRM-7BSearch=SBS(5,1), Agent=Qwen2.5-1.5B-Instruct2025.10 | 73.4 | 3.2 | 20 | |
| MASPRM-1.5BSearch=MCTS(10,3), Agent=Qwen2.5-1.5B-Instruct2025.10 | 69.8 | 5.6 | 34.7 | |
| MASPRM-1.5BSearch=SBS(5,3), Agent=Qwen2.5-1.5B-Instruct2025.10 | 69.1 | 8.4 | 50 | |
| MASPRM-7BSearch=SC@5, Agent=Qwen2.5-1.5B-Instruct2025.10 | 67.8 | 3.6 | 20 | |
| ORM-1.5BSearch=MCTS(10,3), Agent=Qwen2.5-1.5B-Instruct2025.10 | 66.8 | 5.8 | 35.2 | |
| ORM-1.5BSearch=SC@5, Agent=Qwen2.5-1.5B-Instruct2025.10 | 65.4 | 3.6 | 20 | |
| MASPRM-1.5BSearch=SC@5, Agent=Qwen2.5-1.5B-Instruct2025.10 | 64.2 | 3.6 | 20 | |
| Qwen2.5-1.5B-InstructSearch=SC@5, Scorer=None2025.10 | 63.4 | 3.6 | 20 | |
| MASPRM-1.5BSearch=SBS(5,1), Agent=Qwen2.5-1.5B-Instruct2025.10 | 63.4 | 3.4 | 20 | |
| Qwen2.5-1.5B-InstructSearch=MCTS(10,3), Scorer=Policy Likelihood2025.10 | 62.5 | 5.7 | 38.2 | |
| Qwen2.5-1.5B-InstructSearch=SC@5, Scorer=Policy Likelihood2025.10 | 61.2 | 3.6 | 20 | |
| Qwen2.5-1.5B-InstructSearch=SBS(5,3), Scorer=Policy Likelihood2025.10 | 60.3 | 8.4 | 50 | |
| Qwen2.5-1.5B-InstructSearch=SBS(5,1), Scorer=Policy Likelihood2025.10 | 55 | 3.6 | 20 | |
| Qwen2.5-1.5B-InstructSearch=Single-Pass, Scorer=None2025.10 | 47.2 | 0.7 | 4 |