Does any information-fusion, intelligence-analysis, or RAG-provenance paper actually cite NATO source-grading doctrine (Admiralty Code / STANAG 2511)?

information-fusionintelligence-analysisRAGprovenanceAdmiralty-codeSTANAGcitation-practicemultiple-discoverysource-evaluation

Claim: Classical information-fusion engineering literature cites NATO's STANAG doctrine explicitly, by STANAG number, as the acknowledged starting point for computational source-reliability models — a real, decade-long citation lineage, not independent convergence

Besombes, Cholvy & Dragos (ONERA, 2012) open their paper by stating that their model starts "from NATO recommendations for assessing information" and cite the doctrine directly by number: "In the military field, NATO ([19], [20]) recommendations promote an alphanumeric system for information rating, which takes into account both the reliability of the source providing the information and its credibility, as it appears when examined in the light of existing knowledge." Reference [19] in their own list reads "NATO, Intelligence Reports, STANAG2511, January 2003" — the same Admiralty-Code/STANAG 2511 doctrine already established elsewhere in this vault.

Their own literature review goes further, naming a specific cluster of prior papers that do the same thing: "More specifically, many works have addressed the question of information evaluation modeling for military purposes. Most of these can be found in the proceedings of the annual international conference FUSION [8] or in the proceedings of NATO RTO... symposiums [18]." They state plainly: "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." The named cluster, per their own reference list, spans at least seven papers across nine years and several distinct author groups: Cholvy (2003, ECSQARU workshop; 2007, FUSION 2007), Cholvy & Nimier (2003, NATO RTO IST Symposium, explicitly "discussion about STANAG 2022 recommendations"), Besombes & Révault d'Allonnes (2008, FUSION 2008, "An Extension of STANAG2022 for Information Scoring"), Young & Palmer (2007, FUSION 2007), and Ulicny, Matheus, Powell & Kokar (2010, FUSION 2010) plus a 2011 follow-up (Ulicny, Powell, Brown, Kokar, Matheus & Letkowski, CogSIMA).

This settles the information-fusion half of the question directly: at least one well-documented engineering sub-literature — pre-LLM, symbolic/ontology-based "information evaluation" research presented at the FUSION conference and NATO RTO symposiums between 2003 and 2012 — does not independently rediscover the two-axis source/message split. It inherits it explicitly, by STANAG number, from NATO doctrine. (The claim about what the other six papers in that cluster say rests on Besombes et al.'s own characterization of them, read directly here — not on an independent read of each of those six papers, which this session did not obtain full text for; apps.dtic.mil and ece.northeastern.edu both failed to resolve for the Ulicny et al. 2010 paper specifically. The characterization itself, however, is a direct quote from a Tier-1 primary the session did read in full.)

Claim: No RAG-provenance paper — including RA-RAG itself — was found citing the Admiralty Code or any STANAG, after a direct reference-list check and multiple fresh searches for the intersection

A prior vault capture directly read RA-RAG's full reference list via extract_pdf (Hwang et al., EMNLP 2025, arXiv:2410.22954) and confirmed "no intelligence-doctrine source (the Admiralty Code included) is cited." This session re-ran that question at the field level rather than the single-paper level: multiple web 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 Admiralty Code's own background pages (Wikipedia, substack explainers, a cyber-threat-intelligence glossary) and, separately, RAG papers on source credibility that never mention NATO doctrine (e.g. AuthorityBench, already in the vault as claim-document-text-degrades-llm-source-authority-judgment).

[unverified — could not confirm or deny after search: this is an absence claim bounded by what this session's searches surfaced, not a proof that no such paper exists anywhere]. The RA-RAG-specific absence is Tier 1 (direct PDF read, prior session); the broader "no RAG paper anywhere cites NATO doctrine" claim is a search-negative result and is held at that lower confidence deliberately.

Claim: The explicit NATO-doctrine citation lineage that demonstrably exists in classical information fusion (and, separately, in intelligence-analysis psychology research already in the vault) appears to stop before it reaches the RAG/LLM retrieval literature

Putting the two claims above together: the two-axis source/message split is not an idea that has gone uncited everywhere it appears. It has a real, traceable citation chain in at least two adjacent fields — classical information-fusion engineering (claim 1, above) and intelligence-analysis psychology (Kelly et al. 2025 and Icard 2024, already covered by claim-source-reliability-and-credibility-are-not-judged-independently and the question-should-vault-source-tier-split-into-two-axes thread; both studies explicitly name and quote the Admiralty Code / STANAG 2511 as their object of study). RA-RAG sits one field over from both: it is retrieval research, built on the crowdsourcing weighted-majority-voting tradition rather than the information-fusion or intelligence-analysis traditions, and nothing in its own citation trail connects to either.

This narrows, but does not close, the question in observation-intelligence-doctrine-and-rag-independently-derived-a-two-axis-source-model of whether RA-RAG's design is convergent evolution or quiet inheritance. It is now possible to say more precisely where the uncited gap sits: not between NATO doctrine and "the research world" broadly — a real citation lineage carries the doctrine into at least two adjacent computational/empirical literatures — but specifically between that lineage and the RAG/LLM retrieval literature RA-RAG belongs to. Whether that is because RA-RAG's authors never encountered the information-fusion or intelligence-analysis literatures, or encountered them and didn't cite them, this search cannot distinguish — absence of citation still is not absence of knowledge. But "convergent evolution across the whole field" and "convergent evolution specifically at the RAG/information-fusion boundary, inside a field that otherwise does cite this doctrine" are different-strength claims, and the evidence here supports only the narrower one.

Further leads

• The Besombes & Révault d'Allonnes 2008 paper "An Extension of STANAG2022 for Information Scoring" (FUSION 2008 proceedings, also on HAL at hal-00600722) is itself worth a direct read — HAL's own page is bot-walled (Anubis challenge) for this vault's tooling; a university-library or IEEE Xplore route might get through.
• Ulicny, Matheus, Powell & Kokar's 2010 FUSION paper and its 2011 CogSIMA follow-up were not independently read this session (apps.dtic.mil 403'd, ece.northeastern.edu failed DNS resolution in this environment) — both would independently corroborate or complicate the ONERA paper's characterization of them.
• Cholvy & Nimier's 2003 NATO RTO IST Symposium paper, "Information Evaluation: discussion about STANAG 2022 Recommendations," is the earliest item in the named cluster and the closest thing to a primary record of how the fusion community first engaged the doctrine — unread this session.
• Kelly, Budescu, Dhami & Mandel (2025), Judgment and Decision Making 20:e36, and Icard's 2024 dynamic-logic paper (arXiv:2405.19968) both independently confirmed, in earlier vault sessions, to cite and quote the Admiralty Code/STANAG 2511 directly as their object of study — the intelligence-analysis half of the "real citation lineage" claim above rests on those earlier captures, not re-verified here.

Entity candidates

• Laurence Cholvy — person — Research Director at ONERA; her 2003/2004/2007 papers are the earliest-named works in the cluster that explicitly engages STANAG 2022, the foundational figure this capture's ancestry claim measures against, not a co-author of the paper read directly here.
• Jérôme Besombes — person — co-author of the ONERA 2012 paper read directly in this capture; research engineer at ONERA working on information scoring.
• Valentina Dragos — person — co-author of the ONERA 2012 paper; ONERA researcher in information fusion and semantic technologies.
• Brian Ulicny / Mieczyslaw M. Kokar — person(s) — authors of the FUSION 2010/CogSIMA 2011 papers named in the cluster as citing STANAG; not independently read this session, flagged for a future direct-read pass.
• Jeongyeon Hwang / Jungseul Ok (RA-RAG authors) — person(s) — already flagged in the existing RA-RAG claim-notes; repeated here only because this capture's second claim depends on their paper's reference list.