ResearchBenchmarksOffline Goal-conditioned Reinforcement Learning on HumanoidMaze Large Navigate (oraclerep v0)Follow49Task 1 ScoreCRL-1.9611.2724.537.73Oct 26, 2025Evaluation ResultsMethodMethodLinksTask 1 ScoreTask 2 ScoreTask 3 ScoreTask 4 ScoreTask 5 ScoreOverall ScoreCRL2025.104916392028IQL2025.1010011205IVL2025.10903123TRL2025.109029408QRL2025.10309013BC2025.10103322FBC2025.10002121TDP2025.10003201COE2025.10004422