Mathematical Reasoning on MATH-500 (Accuracy, TFLOPs)
81.6Accuracy (MATH-500)Ours-context-unaware
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Ours-context-unawarePair=8B → 4B, Context Setting=context-unaware2026.06 | 81.6 | 6.56 | |
| Ours-context-unawarePair=4B → 8B, Context Setting=context-unaware2026.06 | 78.8 | 9.42 | |
| T2T-context-unawarePair=4B → 14B, Context Setting=context-unaware2026.06 | 75.2 | 60.83 | |
| T2T-context-unawarePair=4B → 8B, Context Setting=context-unaware2026.06 | 74.4 | 37.86 | |
| Ours-context-unawarePair=4B → 14B, Context Setting=context-unaware2026.06 | 70.6 | 17.15 | |
| T2T-context-unawarePair=8B → 14B, Context Setting=context-unaware2026.06 | 70.4 | 72.06 | |
| T2T-context-unawarePair=8B → 4B, Context Setting=context-unaware2026.06 | 69.4 | 32.75 | |
| Ours-context-unawarePair=14B → 8B, Context Setting=context-unaware2026.06 | 68.4 | 13.28 | |
| T2T-context-unawarePair=14B → 4B, Context Setting=context-unaware2026.06 | 66.6 | 41.95 | |
| T2T-context-unawarePair=14B → 8B, Context Setting=context-unaware2026.06 | 66.4 | 53.8 | |
| Ours-context-unawarePair=8B → 14B, Context Setting=context-unaware2026.06 | 65 | 15.85 | |
| Ours-context-unawarePair=14B → 4B, Context Setting=context-unaware2026.06 | 64 | 8.9 | |
| C2C-context-unawarePair=4B → 8B, Context Setting=context-unaware2026.06 | 3 | 14.94 | |
| C2C-context-unawarePair=8B → 4B, Context Setting=context-unaware2026.06 | 2.4 | 14.76 | |
| C2C-context-unawarePair=14B → 4B, Context Setting=context-unaware2026.06 | 0.4 | 17.93 | |
| C2C-context-unawarePair=14B → 8B, Context Setting=context-unaware2026.06 | 0.2 | 24.77 | |
| C2C-context-unawarePair=4B → 14B, Context Setting=context-unaware2026.06 | 0 | 20.05 | |
| C2C-context-unawarePair=8B → 14B, Context Setting=context-unaware2026.06 | 0 | 9.84 |