Mathematical Reasoning on OlympiadBench 675 (Accuracy)
56.85AccuracyDAPO w/ Forking Tokens
Evaluation Results
| Method | Links | |
|---|---|---|
| DAPO w/ Forking TokensBackbone=Qwen3-14B2025.09 | 56.85 | |
| HAPOBackbone=Qwen3-14B2025.09 | 56.43 | |
| Entropy EnvBackbone=Qwen3-14B2025.09 | 54.12 | |
| ArcherBackbone=Qwen3-14B2025.09 | 53.19 | |
| EDGE-GRPOBackbone=Qwen3-14B2025.09 | 52.45 | |
| Vanilla DAPOBackbone=Qwen3-14B2025.09 | 52.28 | |
| Vanilla GRPOBackbone=Qwen3-14B2025.09 | 50.14 | |
| DAPO w/ Forking TokensBackbone=Qwen3-8B2025.09 | 46.27 | |
| HAPOBackbone=Qwen3-8B2025.09 | 45.26 | |
| Vanilla DAPOBackbone=Qwen3-8B2025.09 | 44.87 | |
| EDGE-GRPOBackbone=Qwen3-8B2025.09 | 43.35 | |
| Vanilla GRPOBackbone=Qwen3-8B2025.09 | 42.93 | |
| Entropy EnvBackbone=Qwen3-8B2025.09 | 42.88 | |
| ArcherBackbone=Qwen3-8B2025.09 | 42.69 | |
| HAPOBackbone=Qwen2.5-Math-7B2025.09 | 36.94 | |
| DAPO w/ Forking TokensBackbone=Qwen2.5-Math-7B2025.09 | 35.52 | |
| Vanilla DAPOBackbone=Qwen2.5-Math-7B2025.09 | 34.35 | |
| HAPOBackbone=Qwen2.5-Math-1.5B2025.09 | 33.86 | |
| Entropy EnvBackbone=Qwen2.5-Math-7B2025.09 | 33.68 | |
| ArcherBackbone=Qwen2.5-Math-7B2025.09 | 33.04 | |
| DAPO w/ Forking TokensBackbone=Qwen2.5-Math-1.5B2025.09 | 32.81 | |
| Vanilla GRPOBackbone=Qwen2.5-Math-7B2025.09 | 32.76 | |
| EDGE-GRPOBackbone=Qwen2.5-Math-7B2025.09 | 32.43 | |
| Vanilla DAPOBackbone=Qwen2.5-Math-1.5B2025.09 | 31.47 | |
| EDGE-GRPOBackbone=Qwen2.5-Math-1.5B2025.09 | 30.67 | |
| ArcherBackbone=Qwen2.5-Math-1.5B2025.09 | 29.98 | |
| Vanilla GRPOBackbone=Qwen2.5-Math-1.5B2025.09 | 29.84 | |
| Entropy EnvBackbone=Qwen2.5-Math-1.5B2025.09 | 29.56 |