ResearchBenchmarksLong-horizon Robotic Planning on OGBenchFollow72AntMaze Medium ScoreHDFlow12.7228.1143.558.89May 6, 2026Evaluation ResultsMethodMethodLinksAntMaze Medium ScoreAntMaze Large ScoreAntMaze Giant ScoreHumanoidMaze Medium ScoreHumanoidMaze Large ScoreHumanoidMaze Giant ScoreCube Single Manipulation ScoreCube Double Manipulation ScoreCube Triple Manipulation ScoreScene ScorePuzzle 3x3 ScorePuzzle 4x4 ScorePuzzle 4x5 ScorePuzzle 4x6 ScoreHDFlow2026.057265485142254535177062321614DV2026.05605135382916322314564919—6DL2026.0551402730221125171145391160SHD2026.0542311924167201383731550AD2026.0540281521135181163328200HDMI2026.05312211159312742420000DD2026.05221459518421613000Diffuser2026.051582520521108000