Mathematical Reasoning on AIME 2024 (Avg@32 and Pass@32)
42.7Avg@32SCOPE (Ours)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SCOPE (Ours)Teacher Model=SkyWork-OR1-Math-7B, Student Model=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 42.7 | 77.9 | |
| w/ OPDTeacher Model=SkyWork-OR1-Math-7B, Student Model=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 40.2 | 75 | |
| w/ GRPOTeacher Model=SkyWork-OR1-Math-7B, Student Model=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 35.5 | 68.3 | |
| R1-Distill-Qwen-1.5BTeacher Model=SkyWork-OR1-Math-7B, Student Model=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 29.4 | 76.5 | |
| w/ KDTeacher Model=SkyWork-OR1-Math-7B, Student Model=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 26.6 | 71.4 | |
| SCOPE (Ours)Teacher Model=Qwen3-8B-Instruct, Student Model=Qwen3-1.7B-Base2026.04 | 13.3 | 31.5 | |
| w/ OPDTeacher Model=Qwen3-8B-Instruct, Student Model=Qwen3-1.7B-Base2026.04 | 12.2 | 31.5 | |
| w/ GRPOTeacher Model=Qwen3-8B-Instruct, Student Model=Qwen3-1.7B-Base2026.04 | 7.5 | 27.2 | |
| w/ KDTeacher Model=Qwen3-8B-Instruct, Student Model=Qwen3-1.7B-Base2026.04 | 4.6 | 18.9 | |
| Qwen3-1.7B-BaseTeacher Model=Qwen3-8B-Instruct, Student Model=Qwen3-1.7B-Base2026.04 | 3.7 | 26.7 |