List your company

Multi-task Reinforcement Learning on Meta-World MT50 Conditioned v1

61Average Success RateSoft Modularization (Deep)
26.99235.82144.6553.479Mar 30, 2020

Evaluation results

MethodLinks
Soft Modularization (Deep)Network Depth=Deep2020.0361
Soft Modularization (Shallow)Network Depth=Shallow2020.0360.4
Mix-ExpertImplementation=Internal implementation baseline2020.0337.5
MT-MH-SACImplementation=Internal implementation baseline2020.0334.2
Hard RoutingImplementation=Internal implementation baseline2020.0329.1
MT-SACImplementation=Internal implementation baseline2020.0328.3