main-group-cc: coupled-cluster molecular properties dataset for nine main-group elements

Held-out test set and source structures for "Coupled-cluster molecular properties across the main group that extrapolate beyond training size". Contents: (1) main-group-cc_test-set.tar.gz — the 959-molecule held-out test set (3D structures; composite CCSD(T)/cc-pVTZ, finite-field CCSD/cc-pVDZ and EOM-CCSD/cc-pVDZ labels for seven properties), train/validation/test and scaffold split definitions, and a dataset-index-to-PubChem-CID provenance map (97.7% resolved; every remaining entry classified with coordinate-level evidence); (2) main-group-cc_source-structures.tar.gz — 3D structures of all 44,412 source molecules, dataset-indexed, with per-molecule CID provenance; (3) main-group-cc_test-set-full-inputs.tar.gz — per-molecule Fock/overlap inputs and B3LYP/BP86/DSD-PBEP86 baseline observables enabling turnkey reproduction of the paper’s Table 1. Code: https://github.com/He-Wenhao/ML_electronic_Wenhao