Expert-level Question Answering on GPQA (test)
34.51GPQA ScoreDPO + OGPSA
Evaluation Results
| Method | Links | |
|---|---|---|
| DPO + OGPSABase Model=Qwen2.5-7B-Instruct, Alignment=DPO, Strategy=OGPSA2026.02 | 34.51 | |
| SFT-DPO + General DataBase Model=Llama3.1-8B-Instruct, Alignment=SFT-DPO, Strategy=General Data2026.02 | 29.46 |