Tobacco-Misinformation Claims for AI Credibility Assessment

"Tobacco-Misinformation Claims for AI Credibility Assessment” is a curated dataset consisting of 20 representative statements commonly encountered as misinformation in tobacco control discourse. Each claim has been normalized and formatted in a single-column CSV file (UTF-8 encoded) under the header “claim.” Selection criteria prioritized prominence in public-health conversations and balanced representation across four categories—health impact, scientific assertions, policy measures, and statistical statements. These claims serve as the test set for evaluating the accuracy and reliability of a multi-agent AI pipeline for expert-level credibility scoring, as detailed in the associated manuscript. Researchers and practitioners may use this dataset to benchmark misinformation-detection models, reproduce the reported results, and compare system performance.