EventFlow: Forecasting Temporal Point Processes with Flow Matching

Gavin Kerrigan, Kai Nelson, Padhraic Smyth
Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, PMLR 300:3124-3132, 2026.

Abstract

Continuous-time event sequences, in which events occur at irregular intervals, are ubiquitous across a wide range of industrial and scientific domains. The contemporary modeling paradigm is to treat such data as realizations of a temporal point process, and in machine learning it is common to model temporal point processes in an autoregressive fashion using a neural network. While autoregressive models are successful in predicting the time of a single subsequent event, their performance can degrade when forecasting longer horizons due to cascading errors and myopic predictions. We propose EventFlow, a non-autoregressive generative model for temporal point processes. The model builds on the flow matching framework in order to directly learn joint distributions over event times, side-stepping the autoregressive process. EventFlow is simple to implement and achieves a 20%-53% lower forecast error than the nearest baseline on standard TPP benchmarks while simultaneously using fewer model calls at sampling time.

Cite this Paper

BibTeX
@InProceedings{pmlr-v300-kerrigan26a, title = { EventFlow: Forecasting Temporal Point Processes with Flow Matching }, author = {Kerrigan, Gavin and Nelson, Kai and Smyth, Padhraic}, booktitle = {Proceedings of The 29th International Conference on Artificial Intelligence and Statistics}, pages = {3124--3132}, year = {2026}, editor = {Khan, Emtiyaz and Li, Yingzhen and Solin, Arno and Ramdas, Aaditya}, volume = {300}, series = {Proceedings of Machine Learning Research}, month = {02--05 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v300/main/assets/kerrigan26a/kerrigan26a.pdf}, url = {https://proceedings.mlr.press/v300/kerrigan26a.html}, abstract = { Continuous-time event sequences, in which events occur at irregular intervals, are ubiquitous across a wide range of industrial and scientific domains. The contemporary modeling paradigm is to treat such data as realizations of a temporal point process, and in machine learning it is common to model temporal point processes in an autoregressive fashion using a neural network. While autoregressive models are successful in predicting the time of a single subsequent event, their performance can degrade when forecasting longer horizons due to cascading errors and myopic predictions. We propose EventFlow, a non-autoregressive generative model for temporal point processes. The model builds on the flow matching framework in order to directly learn joint distributions over event times, side-stepping the autoregressive process. EventFlow is simple to implement and achieves a 20%-53% lower forecast error than the nearest baseline on standard TPP benchmarks while simultaneously using fewer model calls at sampling time. } }
Endnote
%0 Conference Paper %T EventFlow: Forecasting Temporal Point Processes with Flow Matching %A Gavin Kerrigan %A Kai Nelson %A Padhraic Smyth %B Proceedings of The 29th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2026 %E Emtiyaz Khan %E Yingzhen Li %E Arno Solin %E Aaditya Ramdas %F pmlr-v300-kerrigan26a %I PMLR %P 3124--3132 %U https://proceedings.mlr.press/v300/kerrigan26a.html %V 300 %X Continuous-time event sequences, in which events occur at irregular intervals, are ubiquitous across a wide range of industrial and scientific domains. The contemporary modeling paradigm is to treat such data as realizations of a temporal point process, and in machine learning it is common to model temporal point processes in an autoregressive fashion using a neural network. While autoregressive models are successful in predicting the time of a single subsequent event, their performance can degrade when forecasting longer horizons due to cascading errors and myopic predictions. We propose EventFlow, a non-autoregressive generative model for temporal point processes. The model builds on the flow matching framework in order to directly learn joint distributions over event times, side-stepping the autoregressive process. EventFlow is simple to implement and achieves a 20%-53% lower forecast error than the nearest baseline on standard TPP benchmarks while simultaneously using fewer model calls at sampling time.
APA
Kerrigan, G., Nelson, K. & Smyth, P.. (2026). EventFlow: Forecasting Temporal Point Processes with Flow Matching . Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 300:3124-3132 Available from https://proceedings.mlr.press/v300/kerrigan26a.html.

Related Material