arXiv Robotics — research abstracts· Peng Cheng, Yunxian Hou, Zhi Zhou, Qian Zhang, Chang Huang, Xianyuan Zhan·· 1 days agoEditorial score32
Higher-Order Action Supervision Enhances Policy Robustness
Higher-Order Action Supervision Makes A Strong Policy Class
Summary
This study shows that supervising zeroth- and first-order actions improves policy performance and robustness. The method works with existing RL frameworks and boosts out-of-distribution generalization in low-data settings.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
Source：arXiv Robotics — research abstracts · arxiv.org