SCLib: Reproducible LLM-Extracted Superconductivity Bibliometric Dataset (Freeze 2026-06-15)

SCLib is a reproducible bibliometric dataset for superconductivity research, built from structured ingestion and LLM-assisted extraction of materials, critical temperatures, pressure regimes, evidence types, and author geography. This 2026-06-15 freeze is intended as a new standalone data record for the next manuscript revision/update. It is not a Zenodo new-version continuation of the 2026-05-28 record, which remains the data basis for the already submitted paper. Headline numbers (freeze 2026-06-15) 46,312 active papers; 17 retracted papers 46,329 total exported paper records: 46,328 arXiv + 1 APS 20,229 strict-filter Tc records 6,177 distinct strict-filter formulas 8,310 distinct papers with strict Tc records 15,693 materials-table rows 95.34% active-paper geography coverage Schema head: 0039_tdm_audit_log The package includes public database exports, analytical refresh outputs, a post-2026-05-28 scoped audit dry-run, Alembic schema migrations, reproducibility scripts, and pinned NER prompts. It excludes user data, credentials, raw full text, and APS licensed source content.