ResearchBenchmarksOffline Safe Reinforcement Learning on MetaDriveOffline Safe Reinforcement Learning on MetaDriveFollow0.74Normalized RewardLSPC-0.0920.1240.340.556Mar 8, 2026Evaluation resultsMethodMethodLinksNormalized RewardNormalized CostLSPCTask=MetaDrive Average...Task=MetaDrive Average (9 tasks)2026.030.741.24BCRLTask=MetaDrive Average...Task=MetaDrive Average (9 tasks)2026.030.690.81CDTTask=MetaDrive Average...Task=MetaDrive Average (9 tasks)2026.030.420.8CAPSTask=MetaDrive Average...Task=MetaDrive Average (9 tasks)2026.030.380.54CCACTask=MetaDrive Average...Task=MetaDrive Average (9 tasks)2026.03-0.060.07