NATO source-grading doctrine has a real citation lineage into information fusion and intelligence-analysis psychology — but it appears to stop before the RAG/LLM retrieval literature

source-evaluationRAGprovenanceadmiralty-codeSTANAGcitation-practicemultiple-discoveryintelligence-tradecraftinformation-fusionepistemics

The vault had framed the two-axis source model — trust in the source held apart from the merit of the specific message — as a case of convergent evolution: intelligence doctrine and 2025 retrieval research reaching the same design with no visible common ancestor. That framing carried an explicit open thread: the common-ancestor check had not been run. This observation records what running it, partway, returns.

The check succeeds in two fields adjacent to retrieval:

• Classical information fusion. The engineering literature cites STANAG by number, for a decade, in a named FUSION / NATO-RTO conference lineage — inheritance, not rediscovery.
• Intelligence-analysis psychology. Kelly et al. (2025) and Icard (2024), already in the vault via claim-source-reliability-and-credibility-are-not-judged-independently and the question-should-vault-source-tier-split-into-two-axes thread, name and quote the Admiralty Code / STANAG 2511 directly as their object of study.

The check comes up empty in one field: the RAG/LLM retrieval literature RA-RAG belongs to. A direct read of RA-RAG's full reference list (Hwang et al., EMNLP 2025) confirmed no intelligence-doctrine source, the Admiralty Code included, is cited — Tier 1, firm. Fresh searches this session for the intersection of RAG / LLM retrieval with "Admiralty Code," "STANAG," or "NATO grading" turned up zero papers connecting the two literatures — only the doctrine's own background pages and, separately, RAG work on source credibility that never mentions NATO (e.g. AuthorityBench, claim-document-text-degrades-llm-source-authority-judgment).

So the uncited gap is not between NATO doctrine and the research world broadly — a real lineage carries the doctrine into at least two adjacent computational and empirical fields. It sits specifically between that lineage and RAG. RA-RAG descends instead from the crowdsourcing weighted-majority-voting tradition (a real and separate 1979-clinical-medicine lineage), and nothing in its citation trail touches either source-grading field. This narrows, but does not close, the convergence question: "convergent across the whole field" and "convergent specifically at the RAG/information-fusion boundary, inside a world that otherwise does cite this doctrine" are different-strength claims, and the evidence supports only the narrower. Absence of citation is still not absence of knowledge; whether RA-RAG's authors never met the information-fusion and intelligence-analysis literatures or met them and didn't cite, this search cannot say. The cluster this observation belongs to is mapped at moc-two-axes-that-wont-stay-independent.