Alignment Evaluation on Socratic Mind
71.6AccuracyMAH-DPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MAH-DPOtest-time guidance=Accuracy PRM2025.10 | 71.6 | 38 | |
| MAH-DPOtest-time guidance=Engaging PRM2025.10 | 71.2 | 54.2 | |
| MODPO2025.10 | 70.47 | 36 | |
| Single-Head DPO2025.10 | 70.4 | 44.6 | |
| MAH-DPOtest-time guidance=none2025.10 | 68.93 | 45.13 |