Reproduce “Generating Physically Consistent Molecules with Energy-Based 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
Christoph Griesbacher, Lea Bogensperger, Andreas Habring, Thomas Pock
EBMol learns a persistent scalar score for molecular structures, using restoring field matching without energy labels. It uses the score both to generate molecules and rank candidates. On 10,000 EDM-generated molecules, its score correlates with GFN2-xTB relaxation energy decrease (r = 0.65), but is not a calibrated physical energy or a guarantee for individual molecules.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.