Reproduce “Continuously Tempered Diffusion Samplers” 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

Ezra Erives, Bowen Jing, Peter Holderrieth, Tommi Jaakkola

The authors introduce continuously tempered diffusion samplers (CTDS), which train a learned transport with proposals exploring a continuous range of temperatures, including hotter distributions that may ease energy-barrier crossing. On a two-dimensional 40-mode Gaussian-mixture benchmark, CTDS reports better Wasserstein and ELBO scores than baseline and NETS proposals; this does not establish performance on more complex targets or isolate temperature’s effect.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.