MAVERICK Cross-Domain Validation Dataset (N=400): Legal and Clinical Generalization

MAVERICK cross-domain validation dataset (N=400) spanning four sources: US legal statutes (100), China legal statutes (100), US clinical trials (100), and China clinical trials (100). Composition: 200 genuine, 120 manipulated, and 80 in-the-wild hallucinated citation records. Each record contains source, truth label, category, primary key, and input text, validated against a programmatic gold standard. This dataset underpins the cross-domain generalization analysis in the MAVERICK manuscript and complements the main validation dataset (v8, N=5,094; DOI 10.5281/zenodo.21975064).