Multi-objective Alignment on Human Values (full evaluation set)
67.64Help ScoreMOPPO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MOPPOObjective=Scalarized with equal weights2025.10 | 67.64 | 29.21 | 14.6 | |
| MAH-DPO EnsembleMode=Ensemble2025.10 | 63.89 | 36.87 | 24.78 | |
| MODPO2025.10 | 61.75 | 34.77 | 23.25 | |
| DPO Soup2025.10 | 61.28 | 32.17 | 21.53 | |
| Base2025.10 | 58 | 30.42 | 18.88 |