Reproduce “Solvating machine-learning interatomic potentials for solution-phase chemistry” 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
Anton Z. Ni, Joonho Lee
AnySolv learns a solvent-dependent energy-and-force correction that can be added to a gas-phase model, using seven solvent descriptors and training against continuum-solvent calculations. Its MoE model predicts solvation energies with 0.0027 eV error on 4,510 pairs in Solv@TUM. The correction supports screening 120 reactions across six conditions, but accuracy remains bounded by the base model and SMD reference.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.