Mathematical Reasoning on AIME 2025 (Pass@32, Avg@32)
30.4Avg@32SCOPE (Ours)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SCOPE (Ours)Teacher Model=SkyWork-OR1-Math-7B, Student Model=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 30.4 | 50.9 | |
| w/ OPDTeacher Model=SkyWork-OR1-Math-7B, Student Model=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 28.9 | 48.5 | |
| w/ GRPOTeacher Model=SkyWork-OR1-Math-7B, Student Model=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 24.5 | 45.1 | |
| R1-Distill-Qwen-1.5BTeacher Model=SkyWork-OR1-Math-7B, Student Model=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 23.9 | 46.9 | |
| w/ KDTeacher Model=SkyWork-OR1-Math-7B, Student Model=DeepSeek-R1-Distill-Qwen-1.5B2026.04 | 22.2 | 45.6 | |
| SCOPE (Ours)Teacher Model=Qwen3-8B-Instruct, Student Model=Qwen3-1.7B-Base2026.04 | 12.1 | 35.6 | |
| w/ OPDTeacher Model=Qwen3-8B-Instruct, Student Model=Qwen3-1.7B-Base2026.04 | 10.6 | 29.7 | |
| w/ KDTeacher Model=Qwen3-8B-Instruct, Student Model=Qwen3-1.7B-Base2026.04 | 5.8 | 24.8 | |
| w/ GRPOTeacher Model=Qwen3-8B-Instruct, Student Model=Qwen3-1.7B-Base2026.04 | 4.2 | 24.5 | |
| Qwen3-1.7B-BaseTeacher Model=Qwen3-8B-Instruct, Student Model=Qwen3-1.7B-Base2026.04 | 3.2 | 20.7 |