Professional Reasoning on GPQA Diamond
52.5Accuracy (ACC)JET
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| JETBackbone=DS-Qwen-7B2025.09 | 52.5 | 5,530 | |
| DAPOBackbone=DS-Qwen-7B2025.09 | 51 | 5,485 | |
| Laser-DBackbone=DS-Qwen-7B2025.09 | 51 | 4,417 | |
| LCR1Backbone=DS-Qwen-7B2025.09 | 50 | 3,429 | |
| DPOBackbone=DS-Qwen-7B2025.09 | 49 | 5,519 | |
| BaseBackbone=DS-Qwen-7B2025.09 | 47.5 | 6,359 | |
| Laser-DEBackbone=DS-Qwen-7B2025.09 | 46.5 | 3,884 | |
| SFTBackbone=DS-Qwen-7B2025.09 | 45.5 | 5,418 | |
| AdaptThinkBackbone=DS-Qwen-7B2025.09 | 44.4 | 4,820 | |
| JETBackbone=DS-Qwen-1.5B2025.09 | 43.4 | 4,182 | |
| DAPOBackbone=DS-Qwen-1.5B2025.09 | 37.9 | 4,591 | |
| SFTBackbone=DS-Qwen-1.5B2025.09 | 34.9 | 5,288 | |
| Laser-DBackbone=DS-Qwen-1.5B2025.09 | 34.3 | 5,352 | |
| Laser-DEBackbone=DS-Qwen-1.5B2025.09 | 33.8 | 5,575 | |
| LCR1Backbone=DS-Qwen-1.5B2025.09 | 33.8 | 3,678 | |
| BaseBackbone=DS-Qwen-1.5B2025.09 | 32.3 | 5,619 | |
| AdaptThinkBackbone=DS-Qwen-1.5B2025.09 | 32.3 | 4,601 | |
| DPOBackbone=DS-Qwen-1.5B2025.09 | 30.3 | 5,283 |