Advances in Neural Information Processing Systems (NeurIPS), 2026. Accepted, to appear
Generative Gibbs for Physics-Aware Sampling (GG-PA): a training-free framework for composing learned diffusion priors with explicit physical context through a joint augmented target distribution.
arXiv preprint, 2026
A denoiser trained on classical Boltzmann statistics alone, composed at sampling time with an analytic Gaussian component carrying the quantum context, yields the quantum Boltzmann distribution of the nuclei and transfers exactly across temperature, isotopic mass, dissipation strength, and path boundary conditions.
arXiv preprint, 2026
Generative Gibbs Sampling for Path Integrals (GG-PI): a ring-polymer-based generative Gibbs sampling framework for recovering quantum statistics from classical simulation data.
Journal of Chemical Theory and Computation, 2025
A unified middle scheme for accurately sampling classical and quantum isobaric–isothermal ensembles in molecular dynamics and path-integral molecular dynamics simulations.