ResearchBenchmarksMeta-Reinforcement Learning on MetaWorld ML-10 (test)Follow2,144.3Drawer OpenSDVT1,567.3081,717.1041,866.92,016.696Dec 16, 2025Evaluation ResultsMethodMethodLinksDrawer OpenDoor CloseShelf PlaceSweep IntoLever PullTest AverageTotal AverageSDVT2025.122,144.31,185.7354.3754.8292.6946.42,256CRAFT2025.122,143.4791409.9893.2346.7916.822,416.8CRAFTState Decoder=IncludedState Decoder=Included2025.122,043.6318.3232504.5324.6684.62,053.6VariBAD2025.121,589.5261.861.1359337.5521.81,547.3