Mathematical Reasoning on AIME 25 (P@4, #Tok)
56.7P@4 AccuracyMPD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MPD2026.05 | 56.7 | 12,900 | |
| CRISP2026.05 | 56.6 | 16,300 | |
| Vanilla LLM2026.05 | 53.3 | 17,700 | |
| Direct Comp.2026.05 | 50 | 15,200 | |
| Chain-of-Draft2026.05 | 50 | 13,000 | |
| LiteCoT2026.05 | 46.7 | 14,600 |