Qiyang (Colin) Li

Everyone knows action chunking is great for imitation learning. It turns out that we can extend its success to RL to better leverage prior data for improved exploration and online sample efficiency! The recipe to achieve this is incredibly simple. 🧵 1/N