ResearchBenchmarksOffline Goal-Conditioned Reinforcement Learning on cube-octuple-1BOffline Goal-Conditioned Reinforcement Learning on cube-octuple-1BFollow3,400Success RateSHARSA-1367821,7002,618Dec 11, 2025Evaluation resultsMethodMethodLinksSuccess RateSHARSA2025.123,400DQCMethod=Decoupled Q-chu...Method=Decoupled Q-chunking2025.123,400HFBC2025.122,800HIQL2025.122,000NSBackup=n-step return b...Backup=n-step return backup2025.12900DQC-naïveExecution=partial acti...Execution=partial action chunk2025.12300FBC2025.120IQL2025.120OSBackup=1-step TD-backupBackup=1-step TD-backup2025.120QCStrategy=Q-chunkingStrategy=Q-chunking2025.120