Conservative Offline Robot Policy Learning via Posterior-Transition Reweighting

Benchmark Model Rank Results
robot-manipulation-on-liberoPTR#57Success Rate: 97.8