Reproduce “Restrained Diffusion Bridges: Targeted Denoiser Training and Refinement” 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
Xuanxi Zhang, Kevin S. Cheng, Weizhou Wang, Jonathan Weare
RDB refines diffusion models to improve sampling of rare molecular states while targeting the original distribution. It restrains samples near a state and trains on diffusion bridges, whose population-optimal denoising target is the original score. In a latent-space chignolin test, refinement estimated the misfolded population at 0.3665%, versus 0.3215% for baseline and 0.3747% reference; global fidelity depended on bridge time.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.