Math Reasoning on Math Benchmarks (MATH, GSM8K, AMC23, AIME24, Minerva, Gaokao, Olympiad) (test)
75.1MATH ScoreQwen2.5-7B-Instruct
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Qwen2.5-7B-InstructBackbone=Qwen2.5-7B2025.12 | 75.1 | 92.4 | 47.5 | 10 | 34.9 | 48.4 | 40.6 | 49.8 | |
| Qwen2.5-7B-Base + DataFlow-Instruct-10KBackbone=Qwen2.5-7B, Training Data=DataFlow-Instruct-10K2025.12 | 73.8 | 88.2 | 47.5 | 16.7 | 30.9 | 31.9 | 37.6 | 46.7 | |
| Qwen2.5-7B-BaseBackbone=Qwen2.5-7B2025.12 | 62.8 | 67.1 | 45 | 10 | 17.6 | 27.5 | 29.6 | 37.1 | |
| Qwen2-7B-Base + DataFlow-Instruct-10KBackbone=Qwen2-7B, Training Data=DataFlow-Instruct-10K2025.12 | 54 | 83 | 27.5 | 0 | 16.5 | 25.3 | 20.3 | 32.4 | |
| Qwen2-7B-InstructBackbone=Qwen2-7B2025.12 | 53.9 | 86.2 | 22.5 | 3.3 | 17.6 | 35.2 | 19.6 | 34 | |
| Qwen2.5-7B-Base + Inf-1MBackbone=Qwen2.5-7B, Training Data=Inf-1M2025.12 | 50.6 | 82 | 27.5 | 0 | 22.1 | 30.8 | 20 | 33.3 | |
| Qwen2-7B-Base + Inf-10KBackbone=Qwen2-7B, Training Data=Inf-10K2025.12 | 45.6 | 81.7 | 25 | 3.3 | 11.8 | 24.2 | 11.1 | 29 | |
| Qwen2-7B-Base + Inf-1MBackbone=Qwen2-7B, Training Data=Inf-1M2025.12 | 45.4 | 79.2 | 25 | 0 | 13.2 | 22 | 10.4 | 27.9 | |
| Qwen2.5-7B-Base + Inf-10KBackbone=Qwen2.5-7B, Training Data=Inf-10K2025.12 | 40.2 | 30.9 | 25 | 3.3 | 9.2 | 27.5 | 21.8 | 22.6 | |
| Qwen2-7B-BaseBackbone=Qwen2-7B2025.12 | 21.2 | 55.9 | 15 | 0 | 9.9 | 30.8 | 7.7 | 20.1 |