Validity citation study, dataset 2026-09-02 (sharpassessment.com/assessment-validity-citations/).
What is here: CODEBOOK.md (claim unit, decision kinds, reason codes, rules added after each gate run); data/roster (the 12
products, frozen roster); data/checked-pages (every checked page with route, final URL, status, saved-file SHA-256, bytes, UTC
time, declared language; pages excluded by the language rule listed per product); data/route1 (homepage header, nav and
footer anchors considered and the links included), data/route2-manifest (one row per sitemap URL considered), data/route3
(every linked destination requested one hop from the seeds); data/scan-status (per page: scanned, skipped, figures OCRed,
rows); data/candidates (every candidate unit with its stable id, page hash and position); data/decisions-ledger (every
explicit decision, keyed by the candidate id, with quoted spans for split passages); data/claims (every claim row with its
decision, reason code and ledger marker explicit or rule); data/claim-groups (one target, criterion, metric, value and
comparator per group, with sources and occurrences); data/products (per-product rollups behind the table); data/images*
(image inventory, OCR records with format and attempts, manual visual readings, decisions); scripts/ (the exact versions
used); s3-results (the computed results). Saved pages, PDF figures and image files are not published because of their size;
every row carries the SHA-256 of its file, and the files are available on request from support@sharpassessment.com.