Mathematical Reasoning on MATH500 (pass@4)
93.6pass@4ThinkTwice
Evaluation Results
| Method | Links | |
|---|---|---|
| ThinkTwiceBackbone=Qwen3-4B, Evaluation Protocol=Direct prompting, Sampling count (n)=322026.04 | 93.6 | |
| DAPOBackbone=Qwen3-4B, Evaluation Protocol=Direct prompting, Sampling count (n)=322026.04 | 93.55 | |
| ThinkTwiceBackbone=OLMo3-7B, Evaluation Protocol=Direct prompting, Sampling count (n)=322026.04 | 92.74 | |
| GRPOBackbone=OLMo3-7B, Evaluation Protocol=Direct prompting, Sampling count (n)=322026.04 | 92.28 | |
| GRPOBackbone=Qwen3-4B, Evaluation Protocol=Direct prompting, Sampling count (n)=322026.04 | 91.86 | |
| DrGRPOBackbone=Qwen3-4B, Evaluation Protocol=Direct prompting, Sampling count (n)=322026.04 | 91.83 | |
| DrGRPOBackbone=OLMo3-7B, Evaluation Protocol=Direct prompting, Sampling count (n)=322026.04 | 91.65 | |
| DAPOBackbone=OLMo3-7B, Evaluation Protocol=Direct prompting, Sampling count (n)=322026.04 | 91.56 | |
| Base ModelBackbone=OLMo3-7B, Evaluation Protocol=Direct prompting, Sampling count (n)=322026.04 | 89.87 | |
| Base ModelBackbone=Qwen3-4B, Evaluation Protocol=Direct prompting, Sampling count (n)=322026.04 | 88.47 |