Mathematical Reasoning on Minerva 272
42.53Accuracy (Minerva 272)HAPO
Evaluation Results
| Method | Links | |
|---|---|---|
| HAPOBackbone=Qwen3-14B2025.09 | 42.53 | |
| DAPO w/ Forking TokensBackbone=Qwen3-14B2025.09 | 40.84 | |
| ArcherBackbone=Qwen3-14B2025.09 | 40.62 | |
| Entropy EnvBackbone=Qwen3-14B2025.09 | 40.56 | |
| EDGE-GRPOBackbone=Qwen3-14B2025.09 | 40.31 | |
| Vanilla DAPOBackbone=Qwen3-14B2025.09 | 39.97 | |
| Vanilla GRPOBackbone=Qwen3-14B2025.09 | 37.52 | |
| HAPOBackbone=Qwen3-8B2025.09 | 34.74 | |
| EDGE-GRPOBackbone=Qwen3-8B2025.09 | 33.91 | |
| HAPOBackbone=Qwen2.5-Math-7B2025.09 | 33.73 | |
| ArcherBackbone=Qwen3-8B2025.09 | 33.22 | |
| Entropy EnvBackbone=Qwen3-8B2025.09 | 32.95 | |
| DAPO w/ Forking TokensBackbone=Qwen2.5-Math-7B2025.09 | 32.79 | |
| Vanilla DAPOBackbone=Qwen3-8B2025.09 | 32.13 | |
| ArcherBackbone=Qwen2.5-Math-7B2025.09 | 31.61 | |
| Entropy EnvBackbone=Qwen2.5-Math-7B2025.09 | 31.43 | |
| Vanilla DAPOBackbone=Qwen2.5-Math-7B2025.09 | 31.28 | |
| Vanilla GRPOBackbone=Qwen3-8B2025.09 | 30.47 | |
| DAPO w/ Forking TokensBackbone=Qwen3-8B2025.09 | 29.91 | |
| Vanilla GRPOBackbone=Qwen2.5-Math-7B2025.09 | 29.85 | |
| EDGE-GRPOBackbone=Qwen2.5-Math-7B2025.09 | 29.24 | |
| HAPOBackbone=Qwen2.5-Math-1.5B2025.09 | 21.91 | |
| EDGE-GRPOBackbone=Qwen2.5-Math-1.5B2025.09 | 19.93 | |
| Entropy EnvBackbone=Qwen2.5-Math-1.5B2025.09 | 19.67 | |
| Vanilla DAPOBackbone=Qwen2.5-Math-1.5B2025.09 | 19.52 | |
| ArcherBackbone=Qwen2.5-Math-1.5B2025.09 | 19.14 | |
| DAPO w/ Forking TokensBackbone=Qwen2.5-Math-1.5B2025.09 | 18.74 | |
| Vanilla GRPOBackbone=Qwen2.5-Math-1.5B2025.09 | 18.23 |