Dialogue Quality Calibration on PopAlign-Bench (test)
8.69Average ScoreDiaSynth
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DiaSynth2026.06 | 8.69 | 1.77 | |
| PersonaLens2026.06 | 8.49 | 1.57 | |
| ConceptPersona2026.06 | 8.31 | 1.39 | |
| FaithfulPersona2026.06 | 8.07 | 1.15 | |
| GroupPersona2026.06 | 7.45 | 0.63 | |
| Reference2026.06 | 6.92 | 0 | |
| ICL2026.06 | 6.01 | 0.91 | |
| Interlocutor2026.06 | 5.44 | 1.49 | |
| AutoPAL2026.06 | 5.03 | 1.89 |