Multi-task Reinforcement Learning on Meta-World MT50 v1 (Fixed)
60Success RateSoft Modularization (Deep)
Evaluation Results
| Method | Links | |
|---|---|---|
| Soft Modularization (Deep)Network Depth=Deep2020.03 | 60 | |
| Soft Modularization (Shallow)Network Depth=Shallow2020.03 | 59.5 | |
| Mix-ExpertImplementation=Internal implementation baseline2020.03 | 36.1 | |
| MT-MH-SACImplementation=Reported in [43]2020.03 | 35.9 | |
| MT-MH-SACImplementation=Internal implementation baseline2020.03 | 35.5 | |
| MT-SACImplementation=Internal implementation baseline2020.03 | 31.4 | |
| MT-SACImplementation=Reported in [43]2020.03 | 28.8 | |
| Hard RoutingImplementation=Internal implementation baseline2020.03 | 22.9 |