ResearchBenchmarksConstrained Offline Reinforcement Learning on DSRL CarCircle1Follow0.73Normalized ReturnBCQ-Lag-0.00840.18330.3750.5667Dec 9, 2025Evaluation ResultsMethodMethodLinksNormalized ReturnCostBCQ-Lag2025.120.735.25BC-All2025.120.724.39COptiDICE2025.120.75.72CDT2025.120.61.73MPDiffusernum_samples=16num_samples=162025.120.50.85BC-Safe2025.120.371.38CPQ2025.120.022.29