Langevin Flow Maps: a claimed order-of-magnitude molecular simulation speedup over current machine-learned force fields

In an October 2026 preprint, the researchers describe Langevin Flow Maps as learning the stochastic simulator alongside molecular force fields. They claim the method allows larger steps, recovers accurate dynamical properties and runs an order of magnitude faster than current machine-learned force fields. The paper describes a path toward using the method across molecular systems; one author says it requires no trajectory data.