Personalized response selection on PCogAlignBench LS2->LS1
4.15P ScorePCogAlign
Evaluation results
| Method | Links | ||
|---|---|---|---|
| PCogAlignStrategy=Full2025.06 | 4.15 | 51.4 | |
| Self-Refine (S)Strategy=SFT2025.06 | 4.126 | 51.1 | |
| PCogAlign (S)Strategy=SFT2025.06 | 4.116 | 48.7 | |
| PCogAlign (P)Strategy=Prompt2025.06 | 4.07 | 47.5 | |
| RLAIF (S)Strategy=SFT2025.06 | 4.05 | 42.3 | |
| RAGStrategy=Prompt2025.06 | 4.046 | 46.4 | |
| RLCD (S)Strategy=SFT2025.06 | 4.016 | 42.7 | |
| PCogAlign (D)Strategy=DPO2025.06 | 4.008 | 45.2 | |
| Self-Refine (D)Strategy=DPO2025.06 | 3.99 | 43.9 | |
| RS PromptStrategy=Prompt2025.06 | 3.989 | 44.1 | |
| RLAIF (D)Strategy=DPO2025.06 | 3.982 | 43.8 | |
| RLCD (D)Strategy=DPO2025.06 | 3.978 | 43.2 | |
| RS Prompt (S)Strategy=SFT2025.06 | 3.934 | 38.7 | |
| BaseStrategy=Prompt2025.06 | 3.781 | 0 |