Long-horizon tasks are important for modern AI and robotics, yet pose a challenge for reinforcement learning. Historically, a popular approach has been to de…
机构:CMU
来源:arXiv 2610.26629 | AI4Papers 论文推荐平台