TAIL: Task-specific Adapters for Imitation Learning with Large Pretrained Models

arXiv:2310.05905v2 Announce Type: replace
Abstract: The full potential of large pretrained models remains largely untapped in control domains like robotics. This is mainly because of the scarcity of data and the computational challenges associated with training or fine-tuning these large models for such applications. Prior work mainly emphasizes either effective pretraining of large models for decision-making or single-task adaptation. But real-world problems will require data-efficient, continual adaptation for new control tasks. Recognizing these constraints, we introduce TAIL (Task-specific Adapters for …

abstract applications arxiv challenges computational control cs.ai cs.lg cs.ro data domains fine-tuning imitation learning large models pretrained models prior robotics training type work