The paper studies a problem of predicting the molecular structured given its NMR spectrum and the molecular formula, through deep imitation learning. The reviewers find the topic important for cheminformatics and the proposed method relevant and potentially impactful. The write-up of the paper should be improved.