Scientific Reasoning on SUPER-GPQA (Score, ∆Avg)
10.54ScoreT3S
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| T3STraining Dataset=S1K-200, Teacher Model=QWQ, Student Model=Qwen3-8B2026.01 | 10.54 | 0.4 | |
| Qwen3-8B (Base)Student Model=Qwen3-8B2026.01 | 10.51 | — | |
| T3STraining Dataset=BOBA-200, Teacher Model=QWQ, Student Model=Qwen3-8B2026.01 | 10.4 | 0.04 | |
| T3STraining Dataset=S1K-200, Teacher Model=DeepSeek-R1, Student Model=Qwen3-8B2026.01 | 10.33 | 0 | |
| T3STraining Dataset=BOBA-200, Teacher Model=DeepSeek-R1, Student Model=Qwen3-8B2026.01 | 10.29 | 0.07 | |
| SFTTraining Dataset=S1K-200, Teacher Model=QWQ, Student Model=Qwen3-8B2026.01 | 10.29 | -0.58 | |
| SFTTraining Dataset=BOBA-200, Teacher Model=QWQ, Student Model=Qwen3-8B2026.01 | 10.21 | -1.26 | |
| SFTTraining Dataset=S1K-200, Teacher Model=DeepSeek-R1, Student Model=Qwen3-8B2026.01 | 10.21 | -0.46 | |
| SFTTraining Dataset=BOBA-200, Teacher Model=DeepSeek-R1, Student Model=Qwen3-8B2026.01 | 10.12 | -0.75 |