Encoding formulas as deep networks: Reinforcement learning for zero-shot execution of LTL formulas | IEEE Conference Publication | IEEE Xplore