ResearchBenchmarksInteractive Imitation Learning on Policy suboptimality guaranteesFollow1Interaction Rounds I(ε)Behavior Cloning0.950.97511.025Sep 26, 2022Evaluation ResultsMethodMethodLinksInteraction Rounds I(ε)Bias TermExpert Annotations A(ε)Oracle Calls C(ε)Behavior Cloning2022.091——1