Groundwater ML credibility audit: public data and reproducibility package

Public data and reproducibility package for the manuscript "Groundwater machine learning studies rarely report the validation and baselines their deployment claims require." The package contains DOI-level literature-audit labels, manual-audit summaries, fold-level model and baseline results, the prespecified Assam arsenic spatial-scale sensitivity analysis at 0.10, 0.15 and 0.20 degrees, figure inputs, verification scripts, and provenance metadata. Publisher full texts, Web of Science exports, raw third-party well observations, and manuscript submission files are excluded.