Mathematical Reasoning on MATH (Correction Uplift %)
48.87Correction UpliftROSA
Evaluation Results
| Method | Links | |
|---|---|---|
| ROSAModel=Qwen3-0.6B2025.09 | 48.87 | |
| ROSAModel=Qwen2.5-7B-Instruct2025.09 | 41.53 | |
| ROSAModel=Qwen3-8B2025.09 | 40.42 | |
| ROSAModel=Qwen2.5-0.5B-Instruct2025.09 | 25.48 | |
| BaselineModel=Qwen3-8B2025.09 | 23 | |
| BaselineModel=Qwen3-0.6B2025.09 | 17.4 | |
| BaselineModel=Qwen2.5-7B-Instruct2025.09 | 12.54 | |
| BaselineModel=Qwen2.5-0.5B-Instruct2025.09 | 6.88 | |
| ROSAModel=DeepSeek-R1-Distill-Llama-8B2025.09 | 6.3 | |
| BaselineModel=DeepSeek-R1-Distill-Llama-8B2025.09 | 4.05 |