Multilingual Reasoning on MMLU ProX Lite
79Accuracy (en)Full translation
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Full translationModel=Qwen3-4B2025.10 | 79 | 77.8 | 78.3 | 76 | 77.8 | 77.4 | 76.1 | 76.4 | 71.6 | 74.8 | 76.5 | 100 | |
| Selective translationModel=Qwen3-4B2025.10 | 77.3 | 76.9 | 77.2 | 74.1 | 75.9 | 75.9 | 73.9 | 74.1 | 65.1 | 72.8 | 74.3 | 20.8 | |
| BaseModel=Qwen3-4B2025.10 | 77 | 77.8 | 76.5 | 73.3 | 75 | 74.6 | 73.9 | 74.6 | 53.6 | 71.1 | 72.7 | 0 |