Multi-Task Reinforcement Learning on Meta-World MT10 v1 (train test)
91Average SuccessNash-MTL
Evaluation Results
| Method | Links | |
|---|---|---|
| Nash-MTLBase RL algorithm=SAC, Evaluation protocol=Multi-Task Learning, Number of random seeds=102022.02 | 91 | |
| STL SACBase RL algorithm=SAC, Evaluation protocol=Single-Task Learning, Number of random seeds=102022.02 | 90 | |
| CAREBase RL algorithm=SAC, Evaluation protocol=Multi-Task Learning, Number of random seeds=102022.02 | 84 | |
| CAGradBase RL algorithm=SAC, Evaluation protocol=Multi-Task Learning, Number of random seeds=102022.02 | 83 | |
| SMBase RL algorithm=SAC, Approach=Soft Modularization, Evaluation protocol=Multi-Task Learning, Number of random seeds=102022.02 | 73 | |
| PCGradBase RL algorithm=SAC, Evaluation protocol=Multi-Task Learning, Number of random seeds=102022.02 | 72 | |
| MH SACBase RL algorithm=SAC, Architecture=Multi-headed, Evaluation protocol=Multi-Task Learning, Number of random seeds=102022.02 | 61 | |
| MTL SAC + TEBase RL algorithm=SAC, Feature=Task Encoder, Evaluation protocol=Multi-Task Learning, Number of random seeds=102022.02 | 54 | |
| MTL SACBase RL algorithm=SAC, Evaluation protocol=Multi-Task Learning, Number of random seeds=102022.02 | 49 |