General Language Understanding on CMMLU and CEval (Chinese Subset)
71.66Chinese Score (CMMLU/CEval)A-CHORD
Evaluation Results
| Method | Links | |
|---|---|---|
| A-CHORDBackbone=Qwen3-4B2025.09 | 71.66 | |
| A-SFT → A-GRPOBackbone=Qwen3-4B2025.09 | 71.45 | |
| CHORDBackbone=Qwen3-4B2025.09 | 71.3 | |
| SFT → A-GRPOBackbone=Qwen3-4B2025.09 | 70.93 | |
| SFT → GRPOBackbone=Qwen3-4B2025.09 | 70.53 | |
| GRPOBackbone=Qwen3-4B2025.09 | 70.42 | |
| A-GRPOBackbone=Qwen3-4B2025.09 | 69.89 | |
| SFTBackbone=Qwen3-4B2025.09 | 68.51 | |
| A-SFTBackbone=Qwen3-4B2025.09 | 68.07 | |
| Qwen3-4BBackbone=Qwen3-4B2025.09 | 67.36 | |
| A-CHORDBackbone=Qwen3-0.6B2025.09 | 45.14 | |
| CHORDBackbone=Qwen3-0.6B2025.09 | 44.9 | |
| A-SFT → A-GRPOBackbone=Qwen3-0.6B2025.09 | 44.47 | |
| SFT → GRPOBackbone=Qwen3-0.6B2025.09 | 44.38 | |
| SFTBackbone=Qwen3-0.6B2025.09 | 44.33 | |
| A-SFTBackbone=Qwen3-0.6B2025.09 | 44.1 | |
| SFT → A-GRPOBackbone=Qwen3-0.6B2025.09 | 44.1 | |
| A-GRPOBackbone=Qwen3-0.6B2025.09 | 43.96 | |
| Qwen3-0.6BBackbone=Qwen3-0.6B2025.09 | 43.25 | |
| GRPOBackbone=Qwen3-0.6B2025.09 | 43.11 |