Mathematical Reasoning on AIME 25 (Correction Uplift %)
27.59Correction UpliftROSA
Evaluation Results
| Method | Links | |
|---|---|---|
| ROSAModel=Qwen3-8B2025.09 | 27.59 | |
| ROSAModel=Qwen2.5-7B-Instruct2025.09 | 20.69 | |
| ROSAModel=Qwen3-0.6B2025.09 | 16.67 | |
| ROSAModel=DeepSeek-R1-Distill-Llama-8B2025.09 | 13.79 | |
| BaselineModel=Qwen3-8B2025.09 | 7.41 | |
| ROSAModel=Qwen2.5-0.5B-Instruct2025.09 | 6.67 | |
| BaselineModel=Qwen3-0.6B2025.09 | 3.57 | |
| BaselineModel=Qwen2.5-7B-Instruct2025.09 | 3.57 | |
| BaselineModel=DeepSeek-R1-Distill-Llama-8B2025.09 | 3.57 | |
| BaselineModel=Qwen2.5-0.5B-Instruct2025.09 | 0 |