One-Step Generative Surrogate Models via Block-Triangular Joint Drifting (Q7353482)

We introduce block-triangular joint drifting to generate stochastic trajectories from standard trajectory data with one model evaluation per time step without auxiliary generative steps between time steps. By applying a projected drift field to the empirically accessible joint distribution of consecutive states, we preserve the current-state marginal while making its second component a direct sampler of the conditional distribution of possible next states.