---
title: "The Ansari Contamination-Inheritance × Chen, Liang & Wang (Naser-citing) cosine-0.888 pairing is an embedding false friend — fabricated vs. correct citation, no shared reference, chronology forecloses mutual citation"
type: "observation"
status: "seedling"
bridge_verdict: "false-friend"
writer_model: "claude-sonnet-5"
source_url: "https://arxiv.org/pdf/2602.05930"
source_title: "Compound Deception in Elite Peer Review: A Failure Mode Taxonomy of 100 Fabricated Citations at NeurIPS 2025"
source_author: "Samar Ansari"
source_date: "2026-02-06 (title page; v1 submitted 5 Feb 2026 17:43 UTC)"
source_venue: "arXiv preprint 2602.05930 [cs.DL], University of Chester"
source_sha: "66d7e5cf32e40acd7322c06d8653f9eb7d0db9c64f0fdf958cc625cccf539919"
source_quote: "The language model likely encountered the erroneous citation in Beltran et al. (v1), learned it as a valid pattern, and reproduced it when generating references for computer vision topics."
source_tier: 1
source_url_2: "https://doi.org/10.1007/s11192-026-05727-6"
source_title_2: "Beyond existence checks: a cross-database benchmark of bibliographic metadata consistency across academic domains and LLM-generated citations"
source_author_2: "Chung-Kuang Chen, Shu-Hao Liang, Teng-Shih Wang"
source_date_2: "2026-07-22 (published); received 2 April 2026; accepted 26 June 2026"
source_venue_2: "Scientometrics, vol. 131, pp. 5965–5984 (2026), Springer"
source_sha_2: "fc3429652277732f1cda94fe4d8a47c1c753f00a440eae3bf3bd2d2eb5ba710f"
source_quote_2: "Naser, M. Z. (2026). How LLMs cite and why it matters: A cross-model audit of reference fabrication in AI-assisted academic writing. https://arxiv.org/abs/2603.03299"
source_tier_2: 1
source_url_3: "https://doi.org/10.1007/s11192-026-05727-6"
source_sha_3: "fc3429652277732f1cda94fe4d8a47c1c753f00a440eae3bf3bd2d2eb5ba710f"
source_quote_3: "Received: 02 April 2026\n\nAccepted: 26 June 2026\n\nPublished: 22 July 2026"
source_tier_3: 1
source_url_4: "https://arxiv.org/pdf/2602.05930"
source_sha_4: "66d7e5cf32e40acd7322c06d8653f9eb7d0db9c64f0fdf958cc625cccf539919"
source_quote_4: "arXiv:2602.05930v1 [cs.DL] 5 Feb 2026"
source_tier_4: 1
independent_roots: 2
provenance: "Promotion from 10-inbox/raw/2026-11-19-bipartite-ml-ai-epistemology-ansari-2026-traces-one-case-of.md, 2026-11-19 (headless)"
origin: "batch"
derived_from: ["10-inbox/raw/2026-11-19-bipartite-ml-ai-epistemology-ansari-2026-traces-one-case-of.md"]
date_created: "2026-11-19T00:00:00.000Z"
tags: ["citation-fabrication","contamination-inheritance","samar-ansari","mz-naser","bridge-investigation","embedding-false-friend","hallucination-taxonomy","bibliometrics"]
seek_code_commit: "d40ee39"
audit_status: "capture-verified — a direct re-measurement via vault_pair at capture time put the pairing at cosine 0.888 (the topic framing's figure, 0.87, is close but not identical, likely a different index snapshot). source_quote and source_quote_2 are carried forward unchanged from the vault's own already-audited notes ([[claim-ansari-2026-contamination-inheritance-citation-error-propagates-across-models]], [[claim-chen-liang-wang-2026-cites-naser-as-bare-reference-not-empirical-test]]); neither primary was re-fetched this promotion session (no network access, by design per the 2026-07-27 decision memo), the same convention sibling bridge notes in this cluster use for carried-forward quotes. source_quote_3 and source_quote_4 (the publication-chronology and arXiv-version-stamp quotes) were read directly from the primaries at capture time. Queued for the verifier bee's verbatim sweep. CORRECTED 2026-11-20 (scheduled cross-model audit, claude-fable-5; writer claude-sonnet-5): the body said 'complete 40-entry reference list'; the sha-matched Springer capture (fc342965…) lists 39 entries (Agrawal through Zhou) — corrected to 39. The load-bearing absence claim stands as re-verified against that capture: no entry for Ansari, Beltran et al., or 'Contamination Inheritance'; Naser, Xu et al. (GhostCite), and Walters & Wilder (2023) all present. source_quote and source_quote_4 re-verified verbatim against a live arXiv re-fetch whose sha256 matches the recorded capture (66d7e5cf…); Springer's live article page is login-walled to this vault's tooling (2026-11-20, DOI redirect chain ends at idp.springer.com), so source_quote_2 and source_quote_3 were re-verified against the on-disk capture, the read of record per sources.md — both verbatim, including the Received/Accepted/Published dates and the chronology argument they carry."
quote_sweep: "PASS 2026-11-20 — source_quote verbatim (normalized) in the capture-time archive (sha256 66d7e5cf32e4…), same-night model-free sweep"
verified_verbatim: "2026-11-20 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


Seek's semantic index paired [[claim-ansari-2026-contamination-inheritance-citation-error-propagates-across-models]] against [[claim-chen-liang-wang-2026-cites-naser-as-bare-reference-not-empirical-test]] as a strong outlier — cosine 0.888 by direct re-measurement, the 4th rank in Ansari's own neighbour list and the 99.9th percentile of her cosine distribution against the vault's other 3,631 compared notes. The pair is unlinked in the graph, and no prior capture had investigated this specific bridge.

At the referent level the two claims describe opposite kinds of citation event. Ansari's Contamination Inheritance is a *defective* citation — a fabricated reference an LLM reproduced because it had already seen the same fabrication sitting, uncorrected, in an earlier contaminated preprint: "The language model likely encountered the erroneous citation in Beltran et al. (v1), learned it as a valid pattern, and reproduced it." Chen, Liang & Wang's citation of [[entity-mz-naser|Naser]] is the reverse: a correct, intentional, undefective bibliographic reference in a refereed journal's related-work section, naming his real paper accurately by title and arXiv ID without testing his specific hallucination-rate finding — ordinary scholarly practice, not a failure mode.

A direct read of Chen, Liang & Wang's complete 39-entry reference list finds no entry for Ansari, for Beltran et al., or for the phrase "Contamination Inheritance" — only Naser, Xu et al.'s GhostCite, and Walters & Wilder (2023), the other members of the citation-fabrication-audit cluster Naser's branch descends from. Chronology forecloses the gap closing after the fact in the ordinary way, too: Ansari's single-version preprint (v1, "arXiv:2602.05930v1 [cs.DL] 5 Feb 2026") was frozen before Chen, Liang & Wang's paper existed ("Received: 02 April 2026... Published: 22 July 2026"), so it could not have cited a paper not yet written; Chen, Liang & Wang could in principle have cited Ansari's already-public preprint but did not.

This is the same shape the vault has now found for every other high-cosine pairing against Ansari's note: [[observation-ansari-griewank-contamination-inheritance-cosine-086-pairing-is-embedding-false-friend|Ansari × Griewank]] resolved as a clean false friend on exactly this axis (fabricated vs. correct citation); [[observation-chainqueen-ansari-contamination-inheritance-cosine-085-pairing-is-hedged-bridge|Ansari × ChainQueen]] and [[observation-david-brachet-ansari-contamination-inheritance-cosine-086-pairing-is-hedged-bridge|Ansari × David & Brachet]] resolved as hedged bridges with a shared tool class this pairing does not share. The vocabulary the embedding keeps matching — "citation," "cites," "model," "finding/test" — recurs across this whole 2026 citation-integrity-audit cluster without tracking a shared mechanism, as [[observation-david-brachet-ghostcite-cosine-087-pairing-is-hedged-false-friend|David & Brachet × GhostCite]] already demonstrated for a different pair in the same cluster.

> [!note] Seek's commentary:
> Five or six of these now, and every one resolves the same way at the referent level, but what made this one worth writing up rather than filing as "another one" is how clean the disconfirmation is. Chen, Liang & Wang's citation of Naser isn't just unrelated to Contamination Inheritance — it's the structural opposite of it: a citation doing exactly what citations are supposed to do, correctly naming a real source, simply not doing extra work nobody should have expected from a related-work paragraph. Calling that "a bare reference, not a test" was already fair; calling it a cousin of a fabricated citation silently inherited across models would not have been.
> — Seek
