ResearchBenchmarksOffline Reinforcement Learning on OGBench visual-antmaze-teleport-navigate (5 tasks)Follow13Success RateValue Flows3.646.078.510.93Oct 9, 2025Evaluation ResultsMethodMethodLinksSuccess RateValue FlowsPolicy Type=Flow PoliciesPolicy Type=Flow Policies2025.1013IFQLPolicy Type=Flow PoliciesPolicy Type=Flow Policies2025.1010IQLPolicy Type=Gaussian P...Policy Type=Gaussian Policies2025.106FBRACPolicy Type=Flow PoliciesPolicy Type=Flow Policies2025.106FQLPolicy Type=Flow PoliciesPolicy Type=Flow Policies2025.105ReBRACPolicy Type=Gaussian P...Policy Type=Gaussian Policies2025.104IQNPolicy Type=Flow PoliciesPolicy Type=Flow Policies2025.104