The paper introduces new techniques to model changes in categorical distributions over event types and the uncertainty in said distributions, where events happen asynchronously (not at pre-specified time instants). The goal is to predict the type of the next event conditioned on an observed history and a particular time gap, while correctly modeling the effect of time gaps on event-type predictions and their associated uncertainty in predictions. The reviewer scores were 5, 8, 8. All reviewers felt the problem setting and approaches were well motivated and the contributions were “important”, “practical”, “sensible”, etc. The reviewers appreciated the quality of the writing and the code submission with examples. R1 had some specific questions, which were largely addressed in the author feedback. The consensus is that this is a good paper and worthy of acceptance.