Offline Meta-Reinforcement Learning on Half-Cheetah-Vel sampled 10 unseen (test)
-48.4Average ReturnCSRO
Evaluation Results
| Method | Links | |
|---|---|---|
| CSROnon-prior context collection strategy (Np)=true, offline RL algorithm=BRAC2023.11 | -48.4 | |
| BORELnon-prior context collection strategy (Np)=false, offline RL algorithm=BRAC2023.11 | -56.1 | |
| FOCALnon-prior context collection strategy (Np)=true, offline RL algorithm=BRAC2023.11 | -60.9 | |
| CORROnon-prior context collection strategy (Np)=true, offline RL algorithm=BRAC2023.11 | -65.6 | |
| CSROnon-prior context collection strategy (Np)=false, offline RL algorithm=BRAC2023.11 | -68.5 | |
| FOCALnon-prior context collection strategy (Np)=false, offline RL algorithm=BRAC2023.11 | -69.5 | |
| BORELnon-prior context collection strategy (Np)=true, offline RL algorithm=BRAC2023.11 | -90.1 | |
| CORROnon-prior context collection strategy (Np)=false, offline RL algorithm=BRAC2023.11 | -92.1 | |
| OffPearlnon-prior context collection strategy (Np)=true, offline RL algorithm=BRAC2023.11 | -133.7 | |
| OffPearlnon-prior context collection strategy (Np)=false, offline RL algorithm=BRAC2023.11 | -162.8 |