Policy generalization on MESSENGER (S1)
100,000Win RateLED-WM
Evaluation Results
| Method | Links | |
|---|---|---|
| LED-WMCurriculum training=false, Training seeds=52025.11 | 100,000 | |
| CRLCurriculum training=false, Training seeds=52025.11 | 88,250 | |
| EMMA (w/ curriculum)Curriculum training=true, Training seeds=52025.11 | 88,230 | |
| EMMA (w/o curriculum)Curriculum training=false, Training seeds=52025.11 | 85,140 |