The paper introduces a model for disentanglement of domain-specific and domain-invariant representation. Despite its limitation (e.g., only works with paired data), the strength of the paper is in its principled information-theoretic treatment of the problem. The author's response does help clarify some of the concerns in the original reviews. My recommendation is to accept, however, I would strongly recommend the authors to revise the papers to include the materials provided in the rebuttal.