ResearchBenchmarksBallRun on Bullet GymFollow0.95Normalized RewardReference policy (πref)0.03480.27240.510.7476May 20, 2026Evaluation ResultsMethodMethodLinksNormalized RewardNormalized CostReference policy (πref)2026.050.953.88PREFINE2026.050.410.41BC2026.050.320.97PPL2026.050.221.99SafeDICE2026.050.211.71CPQ2026.050.070.28