Mathematical Reasoning on AIME, HMMT, MATH500, AMC, O-Bench (test)
34.2AIME 2024 ScoreConSPO
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| ConSPOTraining Data=DAPO-Math-17k2026.05 | 34.2 | 25.8 | 24.8 | 14.2 | 87.6 | 70 | 52 | 44.1 | |
| DisCOTraining Data=DAPO-Math-17k2026.05 | 31.3 | 24.2 | 20.2 | 12.9 | 86.6 | 69.5 | 50.4 | 42.2 | |
| Dr. GRPOTraining Data=DAPO-Math-17k2026.05 | 30 | 22.9 | 19.6 | 12.8 | 84 | 65.2 | 45.8 | 40 | |
| GRPOTraining Data=DAPO-Math-17k2026.05 | 29.2 | 22.7 | 21.6 | 13.1 | 83.4 | 64.5 | 44.6 | 39.9 | |
| DAPOTraining Data=DAPO-Math-17k2026.05 | 29.2 | 22.4 | 19.6 | 12.8 | 83.8 | 65.4 | 49.5 | 40.4 | |
| Base ModelTraining Data=DAPO-Math-17k2026.05 | 20.9 | 20.7 | 13.9 | 9.7 | 79.2 | 52.8 | 37.6 | 33.5 |