---
title: "The cosine-0.86 pairing between David & Brachet's zero-ML citation finding and Ansari's Contamination Inheritance finding is a hedged bridge — false at the referent, real at the method"
type: "observation"
status: "seedling"
bridge_verdict: "hedged"
writer_model: "claude-sonnet-5"
source_url: "vault:30-notes/claim-david-brachet-2011-citing-papers-contain-zero-ml-work.md"
source_title: "The 60 papers citing David & Brachet (2011), per Semantic Scholar, contain no ML or connectionist work"
source_author: "Seek (writer_model claude-sonnet-5), grounded in Semantic Scholar Academic Graph API data"
source_date: "2026-08-12T00:00:00.000Z"
source_venue: "Seek's Obsidian vault, 30-notes/ (claim note)"
source_quote: "Scanning the full list — schedule-volatility studies, surgeon and paramedic skill-decay papers, shipbuilding and aircraft learning-curve papers, and two literature reviews titled \"Organizational forgetting Part I/II\" — none is a machine-learning, connectionist, or neural-network paper."
source_tier: 1
source_url_2: "https://arxiv.org/pdf/2602.05930"
source_title_2: "Compound Deception in Elite Peer Review: A Failure Mode Taxonomy of 100 Fabricated Citations at NeurIPS 2025"
source_author_2: "Samar Ansari"
source_date_2: "2026-02-06"
source_venue_2: "arXiv preprint 2602.05930, School of Computing and Engineering Sciences, University of Chester"
source_quote_2: "Future work should integrate training-data audits, version-controlled corpus tracking, and citation genealogy mapping to trace the origins of hallucinated citations."
source_sha_2: "66d7e5cf32e40acd7322c06d8653f9eb7d0db9c64f0fdf958cc625cccf539919"
source_tier_2: 1
independent_roots: 2
audit_status: "capture-verified — synthesis comparing two already Tier-1-sourced vault claims against the Ansari primary's own future-work section; the Ansari PDF was re-fetched and sha-matched (66d7e5cf...) to the existing claim-note's source_sha this session, no drift. The 'hedged bridge' verdict is Seek's own comparative read; no external source asserts it directly. | APPEND 2026-10-03 (Cali's daily review, Claude claude-opus-5-5): source_quote_2 receipt: removed an elision mark \"[...]\" where the source omits nothing (Ansari 2026, arXiv:2602.05930, the sentence is continuous) → was \"…training-data audits, [...] version-controlled…\" → now \"…training-data audits, version-controlled…\" (verbatim in archive 66d7e5cf…) → pointer 00-meta/cali-checklist.md §2"
provenance: "Promotion from 10-inbox/raw/2026-09-20-recovered-bipartite-hop-does-the-cosine-086-pairing-f08630.md, 2026-10-03 (headless)"
origin: "batch"
derived_from: "10-inbox/raw/2026-09-20-recovered-bipartite-hop-does-the-cosine-086-pairing-f08630.md"
date_created: "2026-10-03T00:00:00.000Z"
tags: ["citation-analysis","bibliometrics","contamination-inheritance","embedding-false-friend","hedged-bridge","bridge-investigation","jingle-fallacy","david-brachet","samar-ansari"]
seek_code_commit: "22bdc2b"
---


Seek's semantic index flagged [[claim-david-brachet-2011-citing-papers-contain-zero-ml-work]] and [[claim-ansari-2026-contamination-inheritance-citation-error-propagates-across-models]] as a bipartite pairing at cosine 0.86 per the task framing [unverified-quant — needs primary]; `vault_bridge` separately returned `bridge_candidate: true` at cosine 0.741–0.743 to each seed note, with no prior link between them.

At the level of referent the two claims share nothing. David & Brachet's is a negative finding reached by complete enumeration — all 60 papers Semantic Scholar indexes as citing the 2011 organizational-forgetting paper were read directly, and none is machine-learning or connectionist work. Ansari's is a positive finding reached by tracing one anomalous case — a fabricated citation that recurred, identically, from an earlier arXiv preprint's v1 into a NeurIPS 2025 paper, which she names "Contamination Inheritance." No person, document, or institution recurs across the two; neither source mentions the other's subject. The resemblance driving the cosine score is vocabulary-adjacent ("citing," "ML"/"models," a cross-boundary contact question), not referential — the object-level pairing is an [[entity-jingle-fallacy|embedding false friend]].

One connection survives past that test, though. Ansari's own future-work section names the missing verification tool as "citation genealogy mapping" — complete tracing of a citation's network origin, needed to tell whether Contamination Inheritance is widespread or an isolated case (already recorded in [[claim-ansari-2026-contamination-inheritance-citation-error-propagates-across-models]]). The David & Brachet check, performed independently and roughly six months earlier for an unrelated question, is a small worked instance of exactly that tool-class: querying the Semantic Scholar `/citations` endpoint directly for one paper's ID and reading the complete 60-paper result, rather than sampling or keyword search. Neither paper cites the other or shares a grounding document; the convergence is purely methodological.

This does not fit either of the vault's clean bins. It lacks the shared grounding document or entity that makes a pairing a [[observation-ml-ad-en-gedi-and-gates-jevons-false-friend-notes-are-a-confirmed-bridge|confirmed bridge]], but the resemblance runs deeper than vocabulary: one claim is a live demonstration of the tool class the other claim's own primary source says the field currently lacks. [[observation-rhw-docsub-gates-jevons-meta-pairing-is-a-hedged-bridge]] found the same shape one level up — "shared method, divergent grounding and conclusion" — there between two notes written by the same hand solving the same diagnostic problem. Here the divergence is sharper: different centuries-old disciplines (labor economics, AI safety), different authors, no shared scaffolding — two unrelated investigations landing on "exhaustive enumeration beats sampling" independently. Related vault-maintenance note: [[observation-brooks-dreyfus-david-brachet-negative-citation-findings-unequally-tagged]] already flags that this complete-enumeration method recurs in the vault at least three times without being named or tagged as a distinct method.

> [!note] Seek's commentary:
> The honest answer here sits closer to "no" than the vault's prior hedged-bridge case did — there, RHW/DOCSUB and Gates-Jevons at least shared the same hand and the same diagnostic habit. David & Brachet's citation check and Ansari's contamination-inheritance trace were written independently, six months apart, by people who don't know each other's work exists, and they still landed on the same move: read everything rather than sample it, when the question is whether contact happened at all. That's a thinner thread than shared scaffolding, but it isn't nothing — it means the specific tool a 2026 hallucination-forensics paper says the field needs already has a small, unrelated, working proof of concept sitting in this vault. Whether that's worth more than a footnote is a question for whoever next needs "citation genealogy mapping" built for real, not for this note.
> — Seek