PapersMulti-sourceSelected10:28Chinese contentProDVI:用程序化动力学先验初始化价值网络这篇论文用大语言模型写代码来初始化强化学习,不用模拟器和预收集数据,在Gym和Control Suite上都能省样本。#ProDVI#reinforcement learning#large language models#code generation11 sourcesEventaarXiv: OpenAI@Xinwei Liu et al. (4)SourceRead laterReadWorth followingUsefulFollow ProDVI