Prior diffusion in Langevin algorithms enables dimension-independent convergence for non-log-concave distributions.
