Models10:12Chinese contentPATS:策略感知训练支架,提升智能体强化学习效率PATS用动态训练支架帮弱策略少犯错,ALFWorld和WebShop上最高提升18.6%,而且部署时没负担。#PATS#强化学习#智能体#ALFWorldaarXiv cs.AI@Yipeng Shi et al. (7)SourceRead laterReadWorth followingUsefulFollow PATS