Classical reinforcement learning (RL) assumes that a reward is observed for every visited state-action pair. However, in real-world applications such as auto…
机构：NVIDIA
来源：arXiv 2610.08271 | AI4Papers 论文推荐平台