arXivCharles Xu, Jacob Berg, Liyiming Ke, Sergey LevineTue, Nov 3, 2026, 10:08 AM PST
score 17.0
New RL method lets robots learn beyond their training data
Original: GSBR: Geometric Support Behavior Regularization for Online RL
Source: arxiv.org ↗
Writing ELI5 summary…