Data and bounded reanalysis code for Observable-Dependent Intermolecular-Force Diagnostics of Pretrained Machine-Learned Potentials for DES-Relevant Molecular Dimers

Version 1.1 updates the archive for the manuscript “Observable-Dependent Intermolecular-Force Diagnostics of Pretrained Machine-Learned Potentials for DES-Relevant Molecular Dimers.” It adds recovered checkpoint and environment records for MACE-MP-0, MACE-OFF23, and CHGNet; a non-destructive two-system MLIP identity check; leave-Me–Da-out sensitivity; and primary-fragment relative-error denominator sensitivity. The archive contains machine-readable source tables, portable dimer geometries, fragment records, provenance manifests and fault-injection tests, bounded reanalysis code and outputs, selection history, and replay-readiness documentation. It supports verification of stored products and strong post-hoc identification of the three MLIP artifacts. It does not claim a clean one-command rerun of every original production calculation, array-by-array cryptographic checkpoint binding, a complete raw reference-input bundle, or training-set independence. Journal-submission PDFs and cover correspondence are not included.