ResearchBenchmarksMulti-task Reinforcement Learning on Meta-World MT10 5 x 10^6 stepsMulti-task Reinforcement Learning on Meta-World MT10 5 x 10^6 stepsFollow79IQMGIRL60.2865.147074.86Apr 8, 2026Evaluation resultsMethodMethodLinksIQMPIGIRLvariant=fullvariant=full2026.0479—TD-MPC22026.047235GIRLvariant=NoGroundvariant=NoGround2026.046938GIRLvariant=FixedBetavariant=FixedBeta2026.046736DreamerV32026.046124