Reproduce “Chem World: A Large-Scale Benchmark and Physics-Aware Framework for Trustworthy Molecular Property Prediction” 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
Tianyou Bai, Huan Wang, Mingchen Gao, Tongrui Yang, Fangyue Lin, Pinze Ren, Zhenlin Zhao, Siming Dong
Chem World combines 17 public datasets and over 800,000 instances across ten tracks for molecules and multi-component systems. PACM models component contributions and pairwise relationships, applying physical constraints where appropriate. MolFormer + PACM achieves macro NRMSE 0.367 under group-aware evaluation, rising to 0.422 on molecule-disjoint and 0.476 on composition-disjoint evaluation, limiting claims about transfer to unseen chemistry.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.