ResearchBenchmarksPointMass on PlaygroundFollow860.3Average RewardFlow533.74618.52703.3788.08Mar 5, 2026Evaluation ResultsMethodMethodLinksAverage RewardFlowPolicy variant=diffusi...Policy variant=diffusion-only2026.03860.3Flow+ResPolicy variant=combine...Policy variant=combined optimal2026.03853.1FPO2026.03712.6DPPO2026.03546.3