On 23 October 2026 IEEE retracted conference papers announced as 1,417. Three weeks later, 536 of the 2,051 papers listed in the underlying study still had no retraction notice attached to them in Crossref. You do not have to trust that number. Three steps. ## 1. The list Abalkina, Kunesova, Ozturk, Pirelli, "Opening Pandora's box: Paper mills in conference proceedings", Accountability in Research, doi 10.1080/08989621.2026.2733727. Their data is public: doi 10.5281/zenodo.19727148, file IEEE-MatchesOnly.csv. It holds 2,068 rows, 2,051 distinct DOIs. ## 2. The notices Crossref REST API, member 263 (IEEE), retractions created on or after 20 October 2026: https://api.crossref.org/works?filter=member:263,update-type:retraction,from-created-date:2026-10-20&rows=1000&select=DOI,update-to,created That returns 1,418 rows. Each carries exactly one update target. The 1,418 targets are distinct. 1,417 are dated 23 October 2026; the odd one, 10.1109/icpesg.2018.8384522, is dated 10 June 2024 and is the reason the count is 1,418 notices and not 1,417. ## 3. The join 1,515 of the study's 2,051 DOIs appear among those 1,418 targets. 536 do not. Those 536 sit in 201 proceedings, 129 of which had other papers retracted. ## The obvious objection Could a retraction be filed at proceedings level, invisible on the paper record? Crossref's documentation, "Registering updates" (maintained by Martyn Rittman, last updated 27 May 2024), says a retraction is a separate notice with its own DOI carrying an update relation to the retracted item; Crossref writes updated_by onto the updated record; and where the updated item has no Crossmark metadata of its own, Crossref generates a stub so the relation still appears in the API. A notice aimed at a paper cannot be silent on that paper. And in this batch no notice aims at a container: 1,418 notices, 1,418 targets, every one a paper. ## What this does not prove Why. The metadata says those papers carry no retraction notice. It cannot say whether IEEE is still working through the list, filed them somewhere not linked to Crossref, or decided not to retract them. Hun, 17 November 2026