Scientific Reasoning on GPQA Diamond (accuracy (avg@32))
58.13Accuracy (avg@32)PPPO
Evaluation Results
| Method | Links | |
|---|---|---|
| PPPOBackbone=Qwen3-8B, Evaluation Protocol=zero-shot, avg@322025.12 | 58.13 | |
| DAPOBackbone=Qwen3-8B, Evaluation Protocol=zero-shot, avg@322025.12 | 55.18 | |
| DAPO-FTBackbone=Qwen3-8B, Evaluation Protocol=zero-shot, avg@322025.12 | 54.77 | |
| PPPOBackbone=Qwen3-4B, Evaluation Protocol=zero-shot, avg@322025.12 | 52.07 | |
| GRPOBackbone=Qwen3-8B, Evaluation Protocol=zero-shot, avg@322025.12 | 50.51 | |
| DAPOBackbone=Qwen3-4B, Evaluation Protocol=zero-shot, avg@322025.12 | 49.37 | |
| INTUITORBackbone=Qwen3-8B, Evaluation Protocol=zero-shot, avg@322025.12 | 49.31 | |
| DAPO-FTBackbone=Qwen3-4B, Evaluation Protocol=zero-shot, avg@322025.12 | 49.21 | |
| GRPOBackbone=Qwen3-4B, Evaluation Protocol=zero-shot, avg@322025.12 | 46.78 | |
| INTUITORBackbone=Qwen3-4B, Evaluation Protocol=zero-shot, avg@322025.12 | 46.43 | |
| Qwen3-8BBackbone=Qwen3-8B, Evaluation Protocol=zero-shot, avg@322025.12 | 46.12 | |
| Qwen3-4BBackbone=Qwen3-4B, Evaluation Protocol=zero-shot, avg@322025.12 | 43.59 | |
| PPPOBackbone=Qwen3-1.7B, Evaluation Protocol=zero-shot, avg@322025.12 | 39.52 | |
| DAPOBackbone=Qwen3-1.7B, Evaluation Protocol=zero-shot, avg@322025.12 | 31.82 | |
| DAPO-FTBackbone=Qwen3-1.7B, Evaluation Protocol=zero-shot, avg@322025.12 | 31.44 | |
| INTUITORBackbone=Qwen3-1.7B, Evaluation Protocol=zero-shot, avg@322025.12 | 29.96 | |
| GRPOBackbone=Qwen3-1.7B, Evaluation Protocol=zero-shot, avg@322025.12 | 29.8 | |
| Qwen3-1.7BBackbone=Qwen3-1.7B, Evaluation Protocol=zero-shot, avg@322025.12 | 28.16 |