Edit-Based Flow Matching for Temporal Point Processes

Temporal point processes (TPPs) are a fundamental tool for modeling event sequences in continuous time, but most existing approaches rely on autoregressive parameterizations that are limited by their sequential sampling. Recent non-autoregressive, diffusion-style models mitigate these issues by jointly interpolating between noise and data through event insertions and deletions in a discrete Markov chain. In this work, we generalize this perspective and introduce an Edit Flow process for TPPs that transports noise to data via insert, delete, and substitute edit operations. By learning the instantaneous edit rates within a continuous-time Markov chain framework, we attain a flexible and efficient model that effectively reduces the total number of necessary edit operations during generation. Empirical results demonstrate the generative flexibility of our unconditionally trained model in a wide range of unconditional and conditional generation tasks on benchmark TPPs.

💡 Research Summary

This paper introduces EDIT‑PP, a novel non‑autoregressive generative framework for temporal point processes (TPPs) that leverages an “edit flow” defined by three atomic operations: insertion, deletion, and substitution. Traditional TPP models rely on autoregressive intensity functions, which require sequential sampling and suffer from error accumulation and linear time complexity with respect to sequence length. Recent diffusion‑style approaches such as ADDTHIN and PSDIFF mitigate some issues by jointly inserting and deleting events, but they still need many discrete steps to transform a noise sequence into a data sequence.

EDIT‑PP formulates the transformation as a continuous‑time Markov chain (CTMC) over the space of event sequences. At any pseudo‑time s∈