coarse-graining

This paper develops a generative flow matching method to capture non-Markovian dynamics in non-equilibrium stochastic systems, demonstrating improved predictions for the Kramers first passage time problem compared to Markovian baselines.