Math Reasoning on Mean of Six Math Benchmarks (Pass@K)
43.8Pass@1GRPO
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GRPOModel=Qwen2.5-Math-7B2025.10 | 43.8 | 61.8 | 71.6 | 59.1 | |
| GRPOModel=Qwen2.5-7B-Instruct2025.10 | 39.6 | 56.3 | 67.5 | 54.5 | |
| SePTModel=Qwen2.5-Math-7B2025.10 | 39.5 | 57.7 | 67.9 | 55 | |
| GRPOModel=Qwen2.5-7B2025.10 | 39.2 | 56.6 | 66.1 | 54 | |
| BaselineModel=Qwen2.5-7B-Instruct2025.10 | 36.8 | 56.6 | 67.3 | 53.6 | |
| SePTModel=Qwen2.5-7B-Instruct2025.10 | 36.6 | 56.8 | 69 | 54.1 | |
| SePTModel=Qwen2.5-7B2025.10 | 32.3 | 54.6 | 65.3 | 50.7 | |
| BaselineModel=Qwen2.5-Math-7B2025.10 | 22.7 | 47.3 | 61 | 43.7 | |
| BaselineModel=Qwen2.5-7B2025.10 | 21.3 | 48.7 | 62.2 | 44.1 | |
| GRPOModel=DeepSeek-Math-7B-Instruct2025.10 | 18.2 | 37.7 | 51 | 35.6 | |
| SePTModel=DeepSeek-Math-7B-Instruct2025.10 | 15.9 | 35.5 | 50.3 | 33.9 | |
| BaselineModel=DeepSeek-Math-7B-Instruct2025.10 | 15.4 | 34.6 | 48.1 | 32.7 |