This submission proposes an unsupervised approach to learning correspondences between 3D shapes of different topologies. It initially received four reviews with all positive scores (7,8,7,9). The reviewers appreciated novelty and originality, as well as quality of empirical evaluation and presentation. The rebuttal addressed some of the remaining concerns, which resulted in an increase of scores to (8,8,7,9). For these reasons, the AC's recommendation is to accept this submission for oral presentation.