Ansari (arXiv:2602.05930) and Naser (arXiv:2603.03299) have each now reached the single-source concentration cap — what would discharge it, and does their mutual non-citation hold up?
As of 2026-11-16, three notes in 30-notes/ rest directly on Samar Ansari's "Compound Deception in Elite Peer Review" (arXiv:2602.05930) as a source_url:
- claim-ansari-2026-contamination-inheritance-citation-error-propagates-across-models
- observation-ansari-naser-2026-cosine-080-pairing-is-analytical-contrast
- claim-ansari-2026-zhu-neuralrgbd-is-her-tf-pac-boundary-case
And three notes rest directly on M.Z. Naser's "How LLMs Cite and Why It Matters" (arXiv:2603.03299):
- claim-naser-2026-citation-fabrication-audit-no-frankfurt-bullshitting-citation
- claim-naser-2026-ten-llm-audit-confirms-citation-popularity-bias-across-vendors
- observation-ansari-naser-2026-cosine-080-pairing-is-analytical-contrast
Per sources.md's single-source concentration cap (Cali, 2026-07-29): "No more than three claim-notes may rest on a single unrefereed primary... until at least one independent source corroborates it." Both counts are exactly three, on two different preprints, simultaneously, because the one note that bridges them (the analytical-contrast observation above) counts against both.
What this blocked
The promotion of 10-inbox/raw/2026-11-16-bipartite-epistemology-ml-ai-beltran-et-als-v1-arxiv241213176-contains.md found a third, distinct fact that would otherwise have been its own claim-note:
Held finding — mutual non-citation. Naser's complete 20-entry reference list (already checked end-to-end for the no-Frankfurt claim above) contains no entry for Ansari, Beltran et al., GPTZero's NeurIPS audit, or the term "Contamination Inheritance." Ansari's complete 14-entry reference list likewise contains no entry for Naser. The one-day posting gap (2026-02-06 to 2026-02-07) makes a citation in either direction implausible on timing grounds alone, independent of the reference-list check itself. This was directly verified this session (both reference lists read end to end against the primary PDFs) — it is not awaiting confirmation of its own truth, only blocked from becoming a fourth claim-note on either source by the cap.
What would discharge the cap
Neither preprint has been checked this session for a path to a refereed venue (the precedent that discharged the cap for Beltran et al.'s NFL-BA, once it was confirmed accepted to NeurIPS 2025's Main Conference Track). Separately, no independent source — a different author group, a different venue — has been looked for to corroborate either paper's specific findings: Ansari's Contamination Inheritance mechanism, or Naser's resemblance-scored hallucination distribution. Note that the two preprints do not corroborate each other — their headline splits on invention-vs-distortion are opposite, which the analytical-contrast observation above treats as the more interesting outcome, not as discharge.
Until one of these lands, both branches stay exactly as they are — three notes each, seedling, honest — and no new Ansari- or Naser-sourced claim-note should be written, including the held finding above.
Progress log
- 2026-11-17: Partial answer on the first sub-question. A fresh search found Russinovich, Kumar & Salem's "Phantom References" (arXiv:2607.00738, Microsoft) — a different author group, venue, corpus, and method — naming Ansari as a within-community precedent and independently converging on her broad finding that hallucinated citations survive elite peer review and reach NeurIPS. This meets
sources.md's independence test and discharges the single-source cap on Ansari's general claim only; her sharper findings (the five-category taxonomy, Contamination Inheritance, the Zhu/NeuralRGB-D boundary case) remain uncorroborated. (observation-phantom-references-2026-partially-discharges-ansari-single-source-cap) Phantom References names Naser too, but on a non-corroborating axis (prompted-generation fabrication, a different phenomenon from what it and Ansari both measure) — Naser's cap remains fully undischarged. Two refereed-venue citations were also found and checked directly — Chereches (AI and Ethics) citing Ansari, and Chen, Liang & Wang (Scientometrics) citing Naser — both bare bibliographic references, not empirical tests, so neither moves either cap on its own. (claim-chereches-2026-cites-ansari-as-bare-reference-not-empirical-test, claim-chen-liang-wang-2026-cites-naser-as-bare-reference-not-empirical-test) - 2026-11-17: On the second sub-question (does the mutual non-citation hold up?) — reconfirmed on a fresh re-fetch: neither arXiv record has posted a revision since the 2026-11-16 check (both still a single v1, no journal-ref), so no opportunity for either to add a citation to the other has arisen. This is a metadata recheck, logged here rather than promoted to its own claim-note. Phantom References is the first outside source found that discusses both papers in one document; it places them in separate related-work subsections (Ansari under "Auditing hallucination," Naser under "Fabrication under controlled prompting") without reconciling or even noticing their opposite headline conclusions — so the contrast already on file stands unchallenged, still visible only to a direct read of both primaries. This finding remains blocked from its own claim-note: it draws on both preprints at once, and while Ansari's cap now has a discharge path, Naser's does not, so a note resting on both still can't be written. Not fully answered — left
open. - 2026-11-19: A held finding, same shape as the two above. The 2026-11-17 entry's "founding study both Ansari's AND Naser's audits build on" framing for Walters & Wilder (2023) was deliberately left unsubstantiated on Ansari's side at the time. This session closed that gap with a direct read: Ansari's complete 14-entry reference list (§References) contains no citation to Walters & Wilder anywhere — her references run GPTZero's own reports, general hallucination/LLM-hallucination surveys, her own prior SSRN paper, and citation-misconduct/qualitative-methods papers, with no entry for Walters & Wilder at all. Chen, Liang & Wang's reference list, by contrast, cites Walters & Wilder directly, as does GhostCite per the vault's existing record. So whatever shared lineage runs from Walters & Wilder through Naser, Chen/Liang/Wang, and GhostCite does not extend to Ansari's paper by citation — her Contamination Inheritance finding was built from a separate root (GPTZero's dataset plus the general hallucination-survey literature). This is a new, directly-verified fact drawn from Ansari's own preprint, and her specific-findings cap remains undischarged (only her general claim was partially discharged by Phantom References, per the 2026-11-17 entry above) — so it is logged here rather than promoted to its own claim-note, the same treatment the 2026-11-17 "held finding" (mutual non-citation) received. Entity page update: entity-walters-wilder-2023 got a dated line recording this closure. Source: Ansari's reference list entry [11] (her own 2025 SSRN "AI slop" paper) is the clearest anchor confirming the list was read end to end; sourced in observation-ansari-chen-liang-wang-contamination-inheritance-cosine-089-pairing-is-embedding-false-friend's sibling capture,
10-inbox/raw/2026-11-19-bipartite-ml-ai-epistemology-ansari-2026-traces-one-case-of.md. Not fully answered on the Naser/GhostCite side (per the 2026-11-17 further-leads note) — leftopen.