ResearchBenchmarksG3 efficient on hopper medium-replay v2Follow577.3Mean Raw MuJoCo ReturnCQL-11.236141.557294.35447.143Apr 23, 2026Evaluation ResultsMethodMethodLinksMean Raw MuJoCo ReturnCQL2026.04577.3PoEalpha=0.5alpha=0.52026.04456.4KL-Regalpha=0.5alpha=0.52026.04456.4PoEalpha=0.7alpha=0.72026.04456.4Frozen2026.04456.1Additivealpha=0.5alpha=0.52026.04226.6Prior2026.04171.4AWR2026.0423.1IQL2026.0411.4