Policy generalization on MESSENGER (S3)
32,190Win RateCRL
Evaluation Results
| Method | Links | |
|---|---|---|
| CRLCurriculum training=false, Training seeds=52025.11 | 32,190 | |
| EMMA (w/ curriculum)Curriculum training=true, Training seeds=52025.11 | 22,380 | |
| EMMA (w/o curriculum)Curriculum training=false, Training seeds=52025.11 | 10,080 |