Mathematical Reasoning on AIME 2025 (avg@8, pass@8)
40.42Avg@8 ScoreMixed
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MixedFramework=TD-Grokking2026.06 | 40.42 | 60 | |
| DynamicFramework=TD-Grokking2026.06 | 39.58 | 63.33 | |
| SFTFramework=GRPO2026.06 | 37.5 | 50 | |
| QuestAFramework=GRPO2026.06 | 37.5 | 60 | |
| VanillaFramework=GRPO2026.06 | 36.67 | 56.67 | |
| Base2026.06 | 35.83 | 53.33 | |
| Scaf-GRPOFramework=GRPO2026.06 | 34.58 | 63.33 | |
| Sub-onlyFramework=TD-Grokking2026.06 | 34.58 | 50 | |
| AblationFramework=GRPO2026.06 | 34.17 | 50 |