Mathematical Problem Solving on Minerva 272 samples
17.11AccuracyHAPO
Evaluation Results
| Method | Links | |
|---|---|---|
| HAPOBackbone=LLaMA3.1-8B-Instruct2025.09 | 17.11 | |
| DAPO w/ Forking TokensBackbone=LLaMA3.1-8B-Instruct2025.09 | 16.39 | |
| HAPOBackbone=LLaMA3.2-3B Instruct2025.09 | 16.05 | |
| Vanilla DAPOBackbone=LLaMA3.1-8B-Instruct2025.09 | 14.76 | |
| DAPO w/ Forking TokensBackbone=LLaMA3.2-3B Instruct2025.09 | 14.54 | |
| Vanilla DAPOBackbone=LLaMA3.2-3B Instruct2025.09 | 13.32 | |
| Vanilla GRPOBackbone=LLaMA3.1-8B-Instruct2025.09 | 12.89 | |
| Vanilla GRPOBackbone=LLaMA3.2-3B Instruct2025.09 | 11.94 |