Review for NeurIPS paper: Finding Second-Order Stationary Points Efficiently in Smooth Nonconvex Linearly Constrained Optimization Problems

All the reviewers were positive towards the paper. Originally, the paper got 687, with relatively high confidences. The major concern about the paper was that it may not fit for large scale problems. During discussion, Reviewer #3 deemed that the rebuttal is "clear and makes sense", hence raised his/her score from 7 to 8. The AC deemed that the theoretical contribution of the paper is good and agreed to forgive the weakness in experiments. Thus the AC recommended acceptance.