Datasets of the Instrument

Cumulative epoch captures across five lanes. Each epoch has its own JSON at alexanarch.org, listed in the epochs table on the instrument's main page. Deltas between epochs are the public severance signal this instrument reports.

The two Zenodo bulk deletion exports, the CHA-cohort extraction from the June 19 event, and the tombstone census. The raw substrate the audit datasets are derived from. Every unpreserved month of the export is permanent substrate loss — Zenodo retains only the last three snapshots.

33-day between-event delta across two Zenodo bulk deletion exports (2026-06-07 to 2026-07-10). Three findings at full evidentiary strength: (1) programmed bibliographic suppression — Zenodo's exporter source (zenodo/zenodo-rdm, commit 8339871aff) is written to strip citation_text under the spam label and to physically destroy older public export snapshots via _remove_old_object_versions at EXPORTER_NUMBER_VERSIONS_TO_KEEP = 3; (2) silent restoration — Wu Shaoyuan withdrawal cascade of 2026-06-26, 67 records restored in a 13-second window, detectable only by set-differencing successive exports; (3) corpus-level LLM contamination — +1.2 to +1.4pp post-2022 marker-union prevalence rise on the Zenodo indexed record-text surface (Kobak et al. 2024 methodology, document-level union implementation; robust across three counterfactual specifications; placebo break order-of-magnitude smaller); descriptively-measured signal-to-disclosure ratio ≈60× against a rebuilt 31-phrase three-family disclosure panel. One exploratory finding (institutional/independent AI-signal differential 0.20% vs ~31%) marked as directional observation, not survival ratio; common-cohort test preregistered in §12a P1.

Appendix A + B of EA-NEGSHAPE-01 v0.2. 1,834 identifier entries enumerated by the 2026-06-19 bulk deletion of the Crimson Hexagonal Archive; 1,621 membership-confirmed; 6,484 formal citations rendered across four formats (APA 7, MLA 9, Chicago 17, BibTeX); container-object citation for CHA-DELETION-CORPUS-20260619; rejected-candidate ledger preserving 68 collisions. Establishes the citation discipline the audit datasets above follow.