Multi-task Reinforcement Learning on Meta-World MT10 Conditioned v1
71.8Average Success RateSoft Modularization (Shallow)
Evaluation Results
| Method | Links | |
|---|---|---|
| Soft Modularization (Shallow)Network Depth=Shallow2020.03 | 71.8 | |
| Soft Modularization (Deep)Network Depth=Deep2020.03 | 68.4 | |
| MT-MH-SACImplementation=Internal implementation baseline2020.03 | 67.4 | |
| MT-SACImplementation=Internal implementation baseline2020.03 | 42.6 | |
| Mix-ExpertImplementation=Internal implementation baseline2020.03 | 40 | |
| Hard RoutingImplementation=Internal implementation baseline2020.03 | 27 |