Preference Learning on Driving Simulated
0.948AlignmentInfogain
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Infogain2026.03 | 0.948 | 0.016 | |
| CMA-ES-IG2026.03 | 0.933 | 0.028 | |
| CMA-ES2026.03 | 0.928 | 0.032 | |
| InfogainFeature dimensions (d)=42026.03 | 0.928 | 0.116 | |
| CMA-ESFeature dimensions (d)=42026.03 | 0.908 | 0.157 | |
| CMA-ES-IGFeature dimensions (d)=42026.03 | 0.882 | 0.265 | |
| InfogainFeature dimensions (d)=82026.03 | 0.849 | 0.332 | |
| CMA-ESFeature dimensions (d)=82026.03 | 0.847 | 0.413 | |
| CMA-ES-IGFeature dimensions (d)=82026.03 | 0.81 | 0.492 | |
| CMA-ES-IGFeature dimensions (d)=162026.03 | 0.718 | 0.759 | |
| CMA-ESFeature dimensions (d)=162026.03 | 0.692 | 0.995 | |
| InfogainFeature dimensions (d)=162026.03 | 0.606 | 1.243 | |
| CMA-ES-IGFeature dimensions (d)=322026.03 | 0.517 | 1.453 | |
| CMA-ESFeature dimensions (d)=322026.03 | 0.457 | 1.876 | |
| InfogainFeature dimensions (d)=322026.03 | 0.374 | 2.115 |