<ul data-eligibleForWebStory="false"><li>Offline multi-task reinforcement learning faces challenges in sharing knowledge across tasks.</li><li>Goal-Oriented Skill Abstraction (GO-Skill) proposed to enhance knowledge transfer and task performance.</li><li>GO-Skill extracts reusable skills through a goal-oriented process and constructs a discrete skill library using vector quantization.</li><li>Experiments on robotic manipulation tasks show the effectiveness and versatility of GO-Skill in MetaWorld benchmark.</li></ul>

Goal-Oriented Skill Abstraction for Offline Multi-Task Reinforcement Learning

Discover more