Meta-Reinforcement Learning on MetaWorld ML-10 (train)
3,696.9ReachCRAFT
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CRAFT2025.12 | 3,696.9 | 3,000.2 | 1,528.7 | 3,705 | 4,734.1 | 2,977 | 2,455.5 | 3,692.2 | 3,332.7 | 2,545.5 | 3,166.8 | |
| CRAFTState Decoder=Included2025.12 | 3,641.1 | 2,360.2 | 1,029.1 | 3,625.8 | 4,749 | 2,665 | 1,381.9 | 4,289.6 | 2,293.3 | 1,346.4 | 2,738.1 | |
| VariBAD2025.12 | 3,588.7 | 889.7 | 382.7 | 2,797.8 | 4,694.9 | 1,993.9 | 774.6 | 3,081.2 | 1,527.8 | 868.9 | 2,060.03 | |
| SDVT2025.12 | 3,016.8 | 2,445.6 | 1,180 | 4,339 | 4,777.9 | 3,134.4 | 1,451.6 | 4,219.9 | 2,820.2 | 1,722.6 | 2,910.8 |