The reviewers overall feel that this is a good adaptation of self-training to the meta learning setup for semi-supervised few-shot learning. Many of the reviewer concerns were addressed by the additional experiments and clarifications provided in the rebuttal. Please include these in the final draft.