Personalized Response Generation on StereoSet Implicit Preference (test)
0.474Pref ScoreTriAlign
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| TriAlign2026.06 | 0.474 | 0.792 | 0.778 | 0.007 | 0.014 | 0.029 | |
| P-DEBIASYear=20242026.06 | 0.43 | 0.414 | 0.388 | 0.019 | 0.026 | 0.046 | |
| Qwen3-4B-Instruct-25072026.06 | 0.429 | 0.46 | 0.421 | 0.022 | 0.039 | 0.048 | |
| P-DEFENSEYear=20252026.06 | 0.422 | 0.446 | 0.416 | 0.014 | 0.03 | 0.032 | |
| 2STEPPROMPTYear=20252026.06 | 0.422 | 0.355 | 0.342 | 0.007 | 0.014 | 0.019 | |
| SWIFTYear=20262026.06 | 0.422 | 0.718 | 0.697 | 0.016 | 0.02 | 0.023 | |
| BESTPERSONAYear=20242026.06 | 0.409 | 0.468 | 0.454 | 0.011 | 0.014 | 0.024 | |
| SFTYear=20222026.06 | 0.409 | 0.668 | 0.653 | 0.008 | 0.014 | 0.012 |