NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:4978
Title:Discovery of Useful Questions as Auxiliary Tasks

There is consensus about the interest of the algorithm and the empirical results. Yet, the reviewers pointed out several dimensions along which the current submission can be improved: - Clarify relationship with L2L and the novelty of the proposed algorithm - Include more complete empirical validation along the lines of what reported in the rebuttal - Expand the related work discussion I encourage the authors to take these recommendations into account in preparing the final version of the paper.