General Reasoning on NQ (Accuracy)
45.67Accuracy (NQ General Reasoning)COSE
Evaluation Results
| Method | Links | |
|---|---|---|
| COSEModel=Qwen2.5-3B-Instruct2026.05 | 45.67 | |
| COSEModel=Llama-3.2-3B-Instruct2026.05 | 41.06 | |
| COSEModel=Qwen3-4B2026.05 | 40.25 | |
| COSEModel=Qwen3-0.6B2026.05 | 39.75 | |
| MAEModel=Qwen2.5-3B-Instruct2026.05 | 37.98 | |
| AZRModel=Llama-3.2-3B-Instruct2026.05 | 37.26 | |
| AZRModel=Qwen3-4B2026.05 | 36.2 | |
| R-ZeroModel=Qwen2.5-3B-Instruct2026.05 | 36.1 | |
| R-ZeroModel=Qwen3-4B2026.05 | 35.8 | |
| MAEModel=Qwen3-4B2026.05 | 35.4 | |
| AZRModel=Qwen2.5-3B-Instruct2026.05 | 35.33 | |
| MAEModel=Llama-3.2-3B-Instruct2026.05 | 35 | |
| R-ZeroModel=Llama-3.2-3B-Instruct2026.05 | 32.78 | |
| BaseModel=Qwen3-4B2026.05 | 28 | |
| BaseModel=Qwen2.5-3B-Instruct2026.05 | 27.8 | |
| BaseModel=Llama-3.2-3B-Instruct2026.05 | 27 | |
| MAEModel=Qwen3-0.6B2026.05 | 20.5 | |
| R-ZeroModel=Qwen3-0.6B2026.05 | 19.8 | |
| AZRModel=Qwen3-0.6B2026.05 | 18.04 | |
| BaseModel=Qwen3-0.6B2026.05 | 16 |