A solver-free approach for modelling stochastic differential equations using conditional normalising flows that enables direct sampling between arbitrary time points within trained horizon.

Structured Linear Controlled Differential Equations: A unifying framework for sequence models with structured, input-dependent state-transition matrices