Chem World: A Large-Scale Benchmark and Physics-Aware Framework for Trustworthy Molecular Property Prediction
Chem World is a large-scale benchmark that integrates 17 public datasets covering over 800,000 molecular and chemical-system instances across 10 property tracks, spanning single-molecule and multi-component systems including drug solubility. The work aims to unify fragmented evaluation by providing a single testbed for molecular representations and compositional modeling. Building on the benchmark, the authors propose PACM, a physics-aware compositional modeling approach that combines molecular representations with composition-aware modeling and domain-specific physical constraints. Experiments report that molecular language models provide strong, transferable representations while PACM improves prediction across diverse chemical systems, and the authors release a drug-property prediction agent built on PACM.
It gives AI-for-science researchers a consolidated, large-scale molecular property benchmark plus a physics-aware modeling framework, which is directly useful for comparing and improving chemistry models.
Comments & Questions (0)
No comments yet
Be the first to comment!