Pretrained generative policies provide strong behavior priors for robot control, but adapting them efficiently with online reinforcement learning remains dif…
机构:北大
来源:arXiv 2611.04543 | AI4Papers 论文推荐平台