Mathematical Reasoning on MATH-500 (Correction Uplift %)
52.94Correction Uplift (%)ROSA
Evaluation Results
| Method | Links | |
|---|---|---|
| ROSAModel=Qwen3-8B2025.09 | 52.94 | |
| ROSAModel=Qwen3-0.6B2025.09 | 51.31 | |
| ROSAModel=Qwen2.5-7B-Instruct2025.09 | 36.91 | |
| BaselineModel=Qwen3-8B2025.09 | 24.54 | |
| ROSAModel=Qwen2.5-0.5B-Instruct2025.09 | 24.05 | |
| BaselineModel=Qwen3-0.6B2025.09 | 17.78 | |
| ROSAModel=DeepSeek-R1-Distill-Llama-8B2025.09 | 17.41 | |
| BaselineModel=Qwen2.5-7B-Instruct2025.09 | 13.65 | |
| BaselineModel=Qwen2.5-0.5B-Instruct2025.09 | 6.79 | |
| BaselineModel=DeepSeek-R1-Distill-Llama-8B2025.09 | 6.45 |