Reproduce “Efficient and Unbiased Sampling from Boltzmann Distributions via Variance-Tuned Diffusion Models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Fengzhe Zhang, Laurence I. Midgley, José Miguel Hernández-Lobato
VT-DIS improves importance sampling for molecular simulation by tuning reverse-diffusion noise covariances while keeping the pretrained score network fixed, then weighting trajectories. This avoids computing endpoint likelihoods with a probability-flow ODE. On LJ-13 at 400 steps, it raises forward and reverse effective sample size from virtually zero to about 35%; on alanine dipeptide, forward ESS remains nearly zero.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.