Offline-to-online reinforcement learning reduces interaction cost for real-world robot learning but suffers from persistent value estimation instability. Exi…
机构:武汉大学
来源:arXiv 2609.12749 | AI4Papers 论文推荐平台