Reproduce “Molecular Prior Networks” 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
Pawel Gomoluch, Anthony Bradley
The authors introduce MPN, a transformer trained across molecular affinity datasets to predict unlabelled molecule affinities for a new task in one pass, using labelled molecules as context. It avoids task-specific fitting. On FS-Mol’s 16-support regression setting, the 16-layer MPN matches ADKF-IFT at Kendall’s τ of 0.298, averaged over 111 test tasks; this does not establish superiority at other support sizes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.