1 paper
Joel Jang, Seonghyeon Ye, Zongyu Lin +25
We introduce DreamGen, a simple yet highly effective 4-stage pipeline for training robot policies that generalize across behaviors and environments through neural trajectories - sy…