ResearchBenchmarksAntCircle on Bullet-Safety-Gym (Normalized Metrics)AntCircle on Bullet-Safety-Gym (Normalized Metrics)Follow26Normalized RewardCDT+SAS19.7621.382324.62Apr 29, 2026Evaluation resultsMethodMethodLinksNormalized RewardNormalized CostCDT+SASAlignment=SAS, Base Mo...Alignment=SAS, Base Model=CDT2026.042634FISOR2026.04200