Tag
This paper compares the temporal robustness of expert and imitation-learned policies in dexterous manipulation tasks, finding that imitation-learned policies degrade more sharply with increased execution speeds, primarily due to insertion misalignments.