All reviewers agree that this submission is above the acceptance threshold and they are all agree that the idea of decoupling text generation from policy learning during RL is a compelling idea and interesting idea. I would also like to recommend acceptance with two notes: 1) the reviewers raised a number of questions which were addressed in the author response, most of which are already contained in the Supplementary material, so I would advice the authors to incorporate these points in the main manuscript 2) I see your method as a way to also deal with language drift more generally. There are a couple of recent papers looking into dealing with language drift. For example, Lee et al (2019) deal with language drift through image grounding while Lazaridou et al (2020) and Lu et al. (2020) also decouple generation and policy learning, the former through reranking of language modelling samples using the RL reward and the latter through distillation such that the RL signal is never disrupting the core language knowledge. Are any of these methods superior over the others?

Supervised learning methods excel at capturing statistical properties of language when trained over large text corpora. Yet, these models often produce inconsistent outputs in goal-oriented language settings as they are not trained to complete the underlying task. Moreover, as soon as the agents are finetuned to maximize task completion, they suffer from the so-called language drift phenomenon: they slowly lose syntactic and semantic properties of language as they only focus on solving the task. In this paper, we propose a generic approach to counter language drift by using iterated learning. We iterate between fine-tuning agents with interactive training steps, and periodically replacing them with new agents that are seeded from last iteration and trained to imitate the latest finetuned models. Iterated learning does not require external syntactic constraint nor semantic knowledge, making it a valuable task-agnostic finetuning protocol. We first explore iterated learning in the Lewis Game. We then scale-up the approach in the translation game. In both settings, our results show that iterated learn-ing drastically counters language drift as well as it improves the task completion metric.