Alignment-diversity coverage on HH-RLHF
81.7CoverageEvoPref
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| EvoPref2026.05 | 81.7 | 0.83 | 0.18 | |
| Group DPO2026.05 | 68.4 | 0.84 | 0.24 | |
| MO-RLHF2026.05 | 63.8 | 0.85 | 0.29 | |
| DPO-Ensembleensemble size=282026.05 | 54.3 | 0.86 | 0.41 | |
| DPOtype=single2026.05 | 34.7 | 0.88 | 0.65 |