General Reasoning on MMLU Pro OOD official (test)
73.43Overall AccuracyQwen3-4B-2507
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Qwen3-4B-2507Model Scale=4B-2507, Optimization Strategy=AWPO2025.12 | 73.43 | 78.9 | 71.29 | 68.64 | 68.78 | 79.5 | |
| Qwen3-8BModel Scale=8B, Optimization Strategy=AWPO2025.12 | 72.67 | 80.46 | 68.51 | 67.49 | 67.8 | 79.08 | |
| Qwen3-4B-2507Model Scale=4B-2507, Optimization Strategy=Base2025.12 | 72.37 | 79.79 | 70.13 | 67.05 | 67.07 | 77.82 | |
| Qwen3-8BModel Scale=8B, Optimization Strategy=Base2025.12 | 72.1 | 80.01 | 67.9 | 68.63 | 65.85 | 78.1 | |
| Qwen3-1.7BModel Scale=1.7B, Optimization Strategy=AWPO2025.12 | 50.07 | 57.81 | 43.19 | 41.34 | 46.34 | 61.65 | |
| Qwen3-1.7BModel Scale=1.7B, Optimization Strategy=Base2025.12 | 48.6 | 56.77 | 41.88 | 43.64 | 41.71 | 59 |