Zero-shot Reasoning on Advanced Reasoning Suite (MMLU-Pro, GPQA, AIME)
74.8MMLU-Pro AccuracyBase
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| BaseModel=Qwen3-8B, Bit=162025.10 | 74.8 | 58.6 | 73.3 | 73.3 | 70 | |
| QTIPModel=Qwen3-8B, Bit=42025.10 | 74 | 57.7 | 70 | 68.9 | 67.7 | |
| NWCModel=Qwen3-8B, Bit=3.942025.10 | 73.8 | 58.8 | 71.1 | 71.2 | 69 | |
| BaseModel=Qwen3-4B, Bit=162025.10 | 70.7 | 54 | 73.3 | 60 | 64.5 | |
| QTIPModel=Qwen3-4B, Bit=42025.10 | 69.8 | 55.2 | 71.1 | 57.8 | 63.5 | |
| NWCModel=Qwen3-4B, Bit=3.942025.10 | 69.4 | 53.2 | 73.3 | 61.1 | 63.7 |