Progressive Agent Skill Generation via Reinforcement Learning
This paper proposes a reinforcement learning method to generate high-quality agent skills by progressively editing skills in a sequential process. Practitioners may care about this approach because it can improve the performance of agents on downstream tasks.