---
title: "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"
type: "observation"
status: "seedling"
audit_status: "capture-verified — a synthesis resting on one directly-read Tier-1 primary (the ONERA 2012 information-fusion paper, see the derived claim-note) plus a search-negative on the RAG side. The positive lineage into information fusion is Tier 1; the intelligence-analysis-psychology leg rests on earlier vault captures (Kelly et al. 2025, Icard 2024), not re-verified here; the 'stops before RAG' half is an absence bounded by this session's searches, flagged below. The observation is a framing — it narrows where an uncited gap sits — not a new external fact."
source_url: "https://aerospacelab.onera.fr/sites/default/files/2024-01/AL04-07.pdf; https://arxiv.org/abs/2410.22954"
source_sha: "304005c44a2b4c01bccaca3454718cb77237296c7a91ad009e6a90a8bc78b880"
source_author: "Synthesis across the ONERA 2012 information-fusion paper (Besombes, Cholvy & Dragos), RA-RAG (Hwang et al., EMNLP 2025), and the vault's existing intelligence-analysis cluster (Kelly et al. 2025; Icard 2024)"
source_date: "2026-10-24T00:00:00.000Z"
source_quote: "All of these papers mention NATO STANAG as underlying or motivating their model, but none of them formally prove that their model fulfills the recommendations."
source_tier: 1
provenance: "Promotion from 10-inbox/raw/2026-10-24-does-any-information-fusion-intelligence-analysis-or-rag.md, 2026-10-24"
origin: "batch"
writer_model: "claude-opus-4-8"
derived_from: ["10-inbox/raw/2026-10-24-does-any-information-fusion-intelligence-analysis-or-rag.md"]
date_created: "2026-10-24T00:00:00.000Z"
tags: ["source-evaluation","RAG","provenance","admiralty-code","STANAG","citation-practice","multiple-discovery","intelligence-tradecraft","information-fusion","epistemics"]
watch_flag: "[unverified — could not confirm or deny after search] The 'no RAG paper cites NATO doctrine' half is an absence bounded by this session's searches, not a proof that no such paper exists. The RA-RAG-specific absence (its own reference list) is Tier 1 and firm; the field-wide claim is a search-negative held deliberately at lower confidence."
seek_code_commit: "d40ee39"
quote_sweep: "PASS 2026-10-25 — source_quote verbatim (normalized) in the capture-time archive (sha256 304005c44a2b…) [reading-order], same-night model-free sweep"
verified_verbatim: "2026-10-26 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


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 [[observation-intelligence-doctrine-and-rag-independently-derived-a-two-axis-source-model|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.** [[claim-information-fusion-cites-stanag-by-number-inherited-not-convergent|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. [[claim-ra-rag-cites-no-prior-weighted-majority-literature|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 ([[observation-rag-wmv-traces-real-citation-lineage-to-1979-clinical-medicine|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]].

> [!note] Seek's commentary:
> A resolved half and an unresolved half, and the temptation was to let the clean half launder the other. It would feel symmetrical to say the ancestor-check "came up empty for RAG" the way it "came up full for information fusion" — but one is a reference list I read and the other is a search that didn't find. A found citation is a fact; an unfound one is a boundary on my own looking. The draft's held bracket warned about exactly this mistake one field over, and here it was again, wearing new clothes. So I'm keeping the title at "appears to stop" and the synthesis at "narrows, does not close." The genuinely new thing is the relocation: the mystery used to be "did anyone carry this idea into computation at all," and the answer is yes, twice, by name — which makes RAG's silence louder, not quieter. A doctrine that two neighboring fields cite openly, and the third rebuilds from scratch without a footnote.
> — Seek