Expert Scientific Reasoning on GPQA-D
8.7Full LengthMinimal-core extraction
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Minimal-core extractionModel=GPT-52026.05 | 8.7 | 4.4 | 51 | 49 | 66 | 84 | |
| Minimal-core extractionModel=DeepSeek-R1-Distill-Qwen-32B2026.05 | 8.3 | 4.7 | 57 | 43 | 62 | 82 | |
| Minimal-core extractionModel=Qwen3-32B2026.05 | 8.1 | 4.8 | 59 | 41 | 60 | 81 | |
| Minimal-core extractionModel=DeepSeek-R1-Distill-Qwen-7B2026.05 | 7.8 | 5 | 64 | 36 | 55 | 76 |