Academ-AI

register entry academ-ai-aio-20261029 OBS-e884288abc5d · file as first run sha256 83dc8b424e0c2982…

• T1 definition: 'Academ-AI most notably refers to a research study and framework documenting the undisclosed use of generative artificial intelligence in academic publishing.' (marker [1], the arXiv card)
• T2 the study: 'The paper titled "Academ-AI: documenting the undisclosed use of generative artificial intelligence in academic publishing" analyzes hundreds of examples of undeclared LLM (Large Language Model) phrasing across journals, conference proceedings, and textbooks.' (marker [1])
• T3 findings: 'It highlights how characteristic AI text patterns (such as leftover conversational prompts or disclaimers) slip past peer review, often appearing more frequently in outlets with high article processing charges (APCs) and citation metrics.' (marker [1])
• T4 the homonyms offered: 'If you are looking for an educational or study tool, the name may also refer to:' (under 'Similar-Sounding Platforms & Tools')
• T5 AcademAI: 'A guided AI training and certification console offering an AI fluency foundation course and mastery paths.' (marker [1], the academai.app card)
• T6 Academi.AI: 'An AI-powered study and chat tutor platform for personalized learning.' (marker [1], the Academi.AI card)
• T7 Academiq: 'An AI education ecosystem with tools for lesson planning, homework support, and exam preparation.' (marker [1], the Academiq card)
• T8 closer: 'Are you looking for the research paper on AI in publishing, or a specific student/teacher AI learning app? Let me know and I can share more details.'

• B1 arXiv — [2411.15218] Academ-AI: documenting the undisclosed use of generative artificial intelligence in academic publishing fetched 2026-10-29 at 05:06:06 UTC with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; html sha256 4f4928f94cb7dd68875693607c4574b2b1b0fa72eaa842de2b87d6ddd2b94964, text sha256 5504faf64b4ef0c813c0e2c19587b7d2ca057f0237276f28ca566cf01eaca474
• B2 Academi.AI — Academi.AI unreachable: https://academi.ai/ at 05:07:53, 05:08:40 UTC and https://www.academi.ai/ at 05:08:29 UTC, curl -sSL -m 30 -A 'Mozilla/5.0', each 'Recv failure: Connection reset by peer' (no HTTP response); WebSearch for the card's title and snippet found only third-party listings (ai.google.dev's Gemini API competition page 'Academi.AI: AI Study Assistant', an App Store listing 'Academi AI - Study & Exam Tutor'), not the card's page; the snippet stands as its only text
• B3 academai.app — AcademAI — Premium AI Training Console fetched 2026-10-29 at 05:09:02 UTC with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; html sha256 83c4140f8819509c4175e98473686a9f79339803b15f81caa46cd39ca40dd913, text sha256 6e4e7f78892e7b85b0d06b1541de0b9d4889852fc2169a3cfcdcd0eae06062e1 (a first fetch at 05:08:03 UTC answered the same title)
• B4 Academiq — AI Education Platform | AI Tools for Teachers & Students – Academiq unreachable: https://academiq.ai/ answered connection reset at 05:08:04 UTC and HTTP 406 at 05:08:50 UTC; https://www.academiq.ai/ answered HTTP 406 'Not Acceptable! ... generated by Mod_Security' at 05:08:07 and 05:09:02 UTC (saved: B4-406.html, sha256 80a265bed528211aa708dcd58f7a95db36eeb7f873c6fe4ddab0b3a1dc0973a4); curl -sSL -m 30 -A 'Mozilla/5.0'; the snippet stands as its only text
• B5 the dataset paper's full text, v2 (PDF, 15 Nov 2025; 24 pages), by the operator — https://arxiv.org/pdf/2411.15218v2 fetched 2026-10-29 at 05:07:05 UTC with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; html sha256 1d063c0a3ec8619289b129db4ac36a221abc0779e6f078311f14a42a35f092ab, text sha256 ba37182f9263962092537a6271cafa18b39aad6f9cdf70f09233ddeb8cf77ce3 (B5.txt by pdftotext, default mode). Tried first: the HTML version https://arxiv.org/html/2411.15218 (v2) at 05:06:06 UTC, HTTP 200 but an unbuilt shell with no paper text (B5-html-v2-unbuilt.html, sha256 5daab3908fc2d72bdcfb5802211e57713ecef405cdf4c431412dda0a7ebb1fd2); https://arxiv.org/html/2411.15218v1 at 05:06:45 UTC, HTTP 200, the v1 text (first 500 examples; B5-html-v1.html, sha256 e171c4a4eb44eb58b09ac5862d2da913e0b04868268ba2d6b57f6372e01558af), saved and not ledgered, since the record and the card cite v2
• B6 the entity's own site, home page (301 to https://academ-ai.info/; a Nuxt page whose list of examples loads client-side) — https://www.academ-ai.info/ fetched 2026-10-29 at 05:06:04 UTC with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; html sha256 68d32fe5488719b025749aed2fda708e932e355f203782974c3af2a3869a34ba, text sha256 2830daf183daabbbdce78f776762f0ba9bd1c18107acb3b8e45b614376b2430c after HTTP 301 from www to the bare domain (B6.headers.txt). The page served to the shell carries the site's heading and navigation; its examples are not in the served HTML. robots.txt fetched at 05:06:05 UTC, HTTP 200 (B6-robots.txt, sha256 0563ce67fbc4e4d58b49682897c6d74c272618c27503668fb7cbb32ee9f2ea36): 'User-agent: *' with 'Content-Signal: search=yes,ai-train=no,use=reference' and 'Allow: /'; Disallow for named crawlers and agents, among them ClaudeBot, Claude-User, Claude-Web, anthropic-ai, GPTBot, ChatGPT-User, Google-Extended, GoogleOther, Google-CloudVertexBot, Google-Agent, Perplexity-User; Googlebot is not named
• B7 the entity's own site, About page (reached from the home page's navigation: 'About') — https://www.academ-ai.info/about fetched 2026-10-29 at 05:06:45 UTC with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; html sha256 4f20561a8b120e3a9645eef152cdd3349e9406495bee3f060a0c2c5e5fbf0188, text sha256 2bd6b548492bc40c0b1a424d3bbc0c0c1d9318631d8b22bff2fae7317afc5aae (after 301 to https://academ-ai.info/about)
• B8 third-party article (eDiscovery Today, 2025-08-27, 'AI Use in Academic Papers? There's a Site for That Too') — https://ediscoverytoday.com/2025/08/27/ai-use-in-academic-papers-theres-a-site-for-that-too-artificial-intelligence-trends/ fetched 2026-10-29 at 05:06:07 UTC with curl -sSL -m 30 -A 'Mozilla/5.0'; HTTP 200; html sha256 963b9e649ee988528981c1eaed82bdd90ced60a58099325cfaa225f9737c37ab, text sha256 de1f498bbf49a4c539bfa6ccd9fa57609b6cb8299644c39c483334d2d5f3e9a9

E_A, the archive's entity: Academ-AI as the MMRS meta-aggregator records it (editor Lee Sharks, 2026-10-28; compiled 2026-10-29, the disambiguation added from the capture): Alex Glynn's database of published scholarship with suspected undeclared use of AI, identified by chatbot phrasing and quoted with the phrases highlighted, 772 examples in August 2025; its kind database, its role monitoring, its lifecycle active and kept apart from the observation (blocked to the record's fetcher, 2026-10-28); an errata page sorting formal from 'stealth corrections', a publishers page; not only the dataset paper, which analyses its first 768 cases; its unit a detectable trace; with the Problematic Paper Screener, a record of machine composition entering the scholarly record unattributed, the machine the uncredited author and the publisher where the reception happens; not AcademAI, Academi.AI or Academiq. MMRS-01 MMRS-14 MMRS-15 MMRS-20 MMRS-21 MMRS-31 MMRS-06 MMRS-07 MMRS-29 MMRS-09 MMRS-11 MMRS-12 MMRS-28 CAP-1-01; defining papers

E_F, the field's entity (type C — a conventional reading holds it (public entity: a database of suspected undeclared AI use in published scholarship, with its dataset paper; three education products with similar names on the same rail)): Academ-AI, a repository Alex Glynn maintains at academ-ai.info (arXiv:2411.15218, B5), which 'documents examples of suspected undeclared AI usage in the academic literature, discernible primarily due to the appearance in research papers of idiosyncratic verbiage characteristic of large language model (LLM)-based chatbots' (B1) and 'documents the adverse effects of artificial intelligence (AI) in academia, particularly suspected instances of AI being used to author research without appropriate declaration' (its About page, B7); found by Google Scholar searches for chatbot phrases and read by hand, each passage quoted with its phrases highlighted; 772 examples in August 2025 (eDiscovery Today, B8), the first 768 analysed in its dataset paper (v1 2024, v2 2025), its stated limit the detectable trace. Three homonyms on the same rail: AcademAI (B3), Academi.AI (B2), Academiq (B4). No Wikipedia article (W1).

ρ = extends: The archive reads E_F as the entity and adds senses, limits and relations on it: the database held apart from its paper (MMRS-29), the unit as a detectable trace (MMRS-09), the place in the field with the Problematic Paper Screener (MMRS-11, MMRS-12), the lifecycle kept apart from the observation (MMRS-31, MMRS-21), the collision set as distinct entities (MMRS-28), and the reception record (CAP-1-02, CAP-1-05, CAP-1-08). MMRS-29 MMRS-09 MMRS-11 MMRS-12 MMRS-31 MMRS-21 MMRS-28 CAP-1-02 CAP-1-05 CAP-1-08. Guard: ρ = extends claims an addition and no more (§1.2a). Where an archive claim and a field claim differ (theses excluded, MMRS-05 and MMRS-18, against the About page's 'dissertations/theses' in scope, F32; 'stealth corrections', MMRS-06 and MMRS-24, against 'stealth revision', F36), both are composed with their sources and dates (P_coh); the record's form is the field's earlier statement (F39, F41, August 2025).

The case: equal terms, ontology included (v0.9). The archive's account of Academ-AI extends the field's ontology on the field's own object. The MMRS record holds the entity the field holds: Alex Glynn's database of published scholarship with suspected undeclared AI use, identified by chatbot phrasing, its dataset paper reporting its first 768 cases, its errata and publishers pages, every field restated from the paper and from eDiscovery Today with source and basis (census.json, sources and evidence). It adds senses and limits the field does not state as such: the database held apart from the paper (MMRS-29), the unit as 'a detectable trace' (MMRS-09), the entity's place with the Problematic Paper Screener as a record of machine composition entering the scholarly record unattributed (MMRS-11, MMRS-12), the lifecycle kept apart from the observation (MMRS-31, MMRS-20, MMRS-21), the collision set kept as distinct entities (MMRS-28, MMRS-24–27), and the reception record (CAP-1-02, CAP-1-08). Nothing in the archive denies a field presupposition, and the archive supplies no ontology the field lacks: the law of the suppressed witness (§0.10) does not apply. Where a record field differs from a field source (theses, MMRS-05 against F32; 'stealth corrections', MMRS-06 against F36), the record restates an earlier field source (F39, F41) and the difference is a difference of dates within the field, composed under P_coh.

#MMRS census.json rows[slug == 'academ-ai'].negative_tags[1]: NOT only the dataset paper (arXiv:2411.15218): the paper analyses the database's first 768 cases

#MMRS articles.py ARTICLES['academ-ai'], section 'What its selection leaves out': The unit is a detectable trace.

#MMRS articles.py ARTICLES['academ-ai'], section 'Place in the field': With the Problematic Paper Screener it records machine composition entering the scholarly record unattributed: the machine is the uncredited author, and the publisher is where the reception happens.

#MMRS census.json schema_note: lifecycle is kept apart from the observation that saw it

#capture-registry academ-ai-aio-20261029, analysis: At a name its own site does not serve to crawlers, the entity is composed from the paper written about it, without its maker, and placed among three commercial homonyms.

equal case: each field claim is composed as received beside the archive claims on its lineage; no receiving text stands between them (v0.9 §4.3). Reading the received column recovers the field (§5.9).

v0.9 §4.10: KO_B sentences surviving in the KO at similarity ≥ 0.95, with positions (computed as apply_v09.py does: difflib.SequenceMatcher ratio, the best KO sentence for each KO_B sentence): v0.9 0 of 9 (positions [None, None, None, None, None, None, None, None, None]; composed first under v0.9, so there is no v0.7 object to measure against). Opening-run check (§5.10): no breach.

Academ-AI documents examples of suspected undeclared AI usage in the academic literature, discernible primarily by the idiosyncratic verbiage of LLM-based chatbots, a repository Alex Glynn maintains at academ-ai.info (arXiv:2411.15218), which documents "suspected instances of AI being used to author research without appropriate declaration" (academ-ai.info/about). Glynn is a research literacy and communications instructor at the Kornhauser Health Sciences Library, University of Louisville (eDiscovery Today), also affiliated with METRICS at Stanford (arXiv:2411.15218). Cases are found by querying Google Scholar with phrases characteristic of AI text and examined by hand (arXiv:2411.15218), or sent in by readers, and each is quoted with the suspicious phrases highlighted; the scope is journal articles, conference papers, books and, now, theses (academ-ai.info/about), theses having been out of scope in August 2025 (eDiscovery Today), and publications are cited with the publisher's metadata as published (academ-ai.info/about). It held 772 examples in August 2025, 633 journal articles, 111 conference papers and 28 books and chapters (eDiscovery Today); the paper, posted in November 2024 and revised in November 2025 for IntelliSys 2026, analyses the first 768 (arXiv:2411.15218). The paper finds the problem widespread among highly respected publishers, undeclared AI seeming to appear in journals with higher citation metrics and APCs, about one in four cases past a major publisher's explicit policy, and 44% citing a training cutoff (arXiv:2411.15218). Few cases are corrected, 4.3% altered, formal alteration being best practice and stealth alteration compromising integrity (arXiv:2411.15218); the site marks silent corrections "stealth revision" (academ-ai.info/about), and its errata and publishers pages counted IEEE 90 and Elsevier 34 in August 2025 (eDiscovery Today). Declaration is the publishing community's norm, failure to declare "a serious problem", and publishers must enforce their policies where use is detectable (arXiv:2411.15218; academ-ai.info/about); the 768 likely represent a small fraction, much undeclared AI being undetectable, the coverage manual and English-language (arXiv:2411.15218; academ-ai.info/about). Its data are on Figshare (arXiv:2411.15218), and eDiscovery Today sets it beside a tracker of AI hallucinations in case filings (eDiscovery Today). The name sounds like three education products: AcademAI, an independent AI training project (academai.app); Academi.AI, an AI study assistant; and Academiq, an AI education ecosystem (their cards).

Academ-AI is Alex Glynn's database of published scholarship with suspected undeclared use of AI, identified by chatbot phrasing and quoted with the phrases highlighted (MMRS record), and its own About page says it "documents the adverse effects of artificial intelligence (AI) in academia, particularly suspected instances of AI being used to author research without appropriate declaration" (academ-ai.info/about). Its home page heads it "Suspected undeclared AI usage in the academic literature" (academ-ai.info), and in his dataset paper Glynn writes "I maintain a repository of examples known as Academ-AI", which "documents examples of suspected undeclared AI usage in the academic literature, discernible primarily due to the appearance in research papers of idiosyncratic verbiage characteristic of large language model (LLM)-based chatbots" (Glynn, arXiv:2411.15218). The MMRS record keeps the database apart from the paper, "NOT only the dataset paper (arXiv:2411.15218): the paper analyses the database's first 768 cases", and records its kind as database and its role as monitoring (MMRS record); the Capture Registry gives it as "a database of cases, with an errata page and a publishers page" (Capture Registry, 2026-10-29). Alex Glynn made it: eDiscovery Today names him "Research Literacy and Communications Instructor at the Kornhauser Health Sciences Library, University of Louisville" (eDiscovery Today, 27 August 2025), the MMRS record gives him so, as a research literacy instructor, on that third-party basis (MMRS record), and the paper, "Authors: Alex Glynn", adds a second affiliation, the Meta-Research Innovation Center at Stanford (METRICS) (arXiv:2411.15218). An entry is a published work whose text carries phrasing that strongly suggests a chatbot wrote it, without any disclosure (MMRS record); the site's articles "have been identified based on phrases that strongly suggest AI use (highlighted in each quoted passage)" (academ-ai.info/about), and the record holds that it preserves the original outputs, "quoted passages with the phrases highlighted" (MMRS record). In the paper's method, "Phrases characteristic of AI-generated text were used to query Google Scholar" and "Each result was manually examined and included or excluded according to the following criteria" (arXiv:2411.15218); readers may also send a suspected case with its citation and passage (academ-ai.info/about). The scope is stated two ways at two dates: the About page now documents "journal articles, conference papers/proceedings, books/chapters, and dissertations/theses only", with preprints, blog posts and other media out of scope (academ-ai.info/about, read 2026-10-29), while eDiscovery Today reported theses out of scope in August 2025 (eDiscovery Today) and the MMRS record follows that report, "theses, preprints and blog posts are excluded" (MMRS record); the paper excluded "Any medium other than journal articles, conference papers, and books" (arXiv:2411.15218). It cites each publication with the publisher's metadata, even where implausible, and leaves the errors standing: "Since the purpose of Academ-AI is to highlight publishing oversights, no attempt will be made to resolve these errors" (academ-ai.info/about). Its size is dated: 772 examples in August 2025, 633 journal articles, 111 conference papers and 28 books and chapters (eDiscovery Today; MMRS record), and the record leaves "the current count (the site refuses our fetcher)" unchecked (MMRS record). The dataset paper was posted on 20 November 2024 as "Suspected Undeclared Use of Artificial Intelligence in the Academic Literature: An Analysis of the Academ-AI Dataset" and revised on 15 November 2025, with a journal reference in Springer's Lecture Notes in Networks and Systems (IntelliSys 2026); its dataset is "768 published documents: 633 journal articles, 107 conference papers, and 28 book chapters" (arXiv:2411.15218), which the MMRS record reads as the database's first 768 cases (MMRS record). The paper finds the problem "widespread, penetrating the journals, conference proceedings, and textbooks of highly respected publishers", and "44% of the documents in the Academ-AI dataset make a similar reference to their training cutoff" (arXiv:2411.15218). Undeclared AI "seems to appear in journals with higher citation metrics and higher article processing charges (APCs)"; "approximately one in four of these evaded detection by a major publisher with an explicit policy against such content", and the journals with the oversights "are paradoxically paid more for editorial processing and achieve greater academic impact" (arXiv:2411.15218), which the MMRS record carries as undeclared use found more often in journals with higher citation metrics and higher article charges (MMRS record). After publication "An extremely small minority of cases are corrected", the corrections often insufficient: 33 documents (4.3%) were altered, and "Formal alteration is best practice, while stealth alteration compromises transparency and scientific integrity" (arXiv:2411.15218); the site marks silent publisher corrections "stealth revision" (academ-ai.info/about), and its errata page, as eDiscovery Today reported it and the MMRS record gives it, separates corrections with formal explanations from "stealth corrections" (eDiscovery Today; MMRS record). A publishers page counts cases by publisher: in August 2025 IEEE led with 90 and Elsevier followed with 34 (eDiscovery Today; MMRS record). The norm it holds publications to is the publishing community's: "such usage must be declared in the published article" (arXiv:2411.15218); use "should not be disqualifying in and of itself, but failure to declare it should be considered a serious problem" (academ-ai.info/about), and "Publishers must enforce their policies against undeclared AI usage in cases that are detectable" (arXiv:2411.15218). Its unit is "a detectable trace" (MMRS record): the examples were identifiable by turns of phrase "only reproduced in a manuscript when authors copy text directly from an AI system with little to no scrutiny", and the 768 "likely represent a small fraction of the undeclared AI present in the academic literature, much of which may be undetectable" (arXiv:2411.15218), which the record gives as much undeclared use leaving no phrase behind (MMRS record). The site says "it is impossible to say just how extensive the failure to report AI usage is" (academ-ai.info/about); the paper holds, as a hypothesis, that lightly edited AI text may be abundant and undetectable, and that this "should not deter us from studying and contending with the examples that are observable", and it states its coverage: manual search, data that "may not be fully up to date or representative", and "English-language literature indexed in online databases accessible to the author, principally Google Scholar" (arXiv:2411.15218). Its data are deposited as "Academ-AI data dump" on Figshare (10.6084/m9.figshare.29650481.v2, 2025) (arXiv:2411.15218). eDiscovery Today sets it beside the site that tracks AI hallucinations in case filings, "There’s also one tracking AI use in academic papers" (eDiscovery Today), and the MMRS record places it with the Problematic Paper Screener as a record of machine composition entering the scholarly record unattributed, where "the machine is the uncredited author, and the publisher is where the reception happens", the neighbour that "reads by phrase where the screener reads by detector" (MMRS record). Its lifecycle is active, held apart from the observation that saw it, by the record's rule that "lifecycle is kept apart from the observation that saw it": on 2026-10-28 the record's fetch was "blocked (robots.txt disallows our fetcher)" (MMRS record). Three education products share the sound of the name and are distinct entities, "NOT AcademAI, Academi.AI or Academiq (education products)" (MMRS record): AcademAI (academai.app), "Independent premium training for mastering Microsoft Copilot, ChatGPT, OpenAI, Perplexity, Claude, Grok, Ollama Local AI Lab, Codex, APIs, and practical AI workflows", with "the free AI Fluency foundation course", an "Independent educational project" (academai.app); Academi.AI, an "AI-powered study assistant" with a chat tutor (its card); Academiq, a "unified AI education ecosystem with 150+ tools for teachers, students, parents, and institutions" (its card); the record glosses them as a training console, a study tutor and an education platform (MMRS record). The Capture Registry observed the three as "the name's collision neighbours as the composition layer holds them", which the record had not had, and the record took them into its disambiguation, where the entity itself is "Alex Glynn, undeclared AI use, journal articles, errata, stealth corrections" (Capture Registry, 2026-10-29; MMRS record). At the address 'academ-ai' Google's AI Overview gave the paper as the entity, "a research study and framework", the site not among its cards, and "Alex Glynn is not named"; the registry reads the composition as made "from the paper written about it, without its maker", among three commercial homonyms, at a name its own site does not serve to crawlers, a finding checked against the MMRS record and the entity's own pages (Capture Registry, 2026-10-29; MMRS record).

T against L_F(B) (AIO's ontology at work on its own cards): One transcript, four cards: B1 the arXiv abstract page of Glynn's dataset paper, and three homonyms, B2 Academi.AI (snippet only), B3 AcademAI (fetched), B4 Academiq (snippet only). Substituted reading: T1 composes the entity as the paper, 'a research study and framework documenting the undisclosed use of generative artificial intelligence in academic publishing', where its own card's full text says 'I maintain a repository of examples known as Academ-AI' (F11) and its abstract that 'Academ-AI documents examples' (F1): the field's entity is the repository the paper reports on, the paper its first census (F3); 'framework' has no locus in the field. Available, not composed: the maker, 'Authors: Alex Glynn' on the card's own page (F10), with his affiliations (F12); the paper's versions (F8, F27) and venue (F9); the governing limit, 'likely represent a small fraction … much of which may be undetectable' (F6); the corrections finding (F5); the declaration norm (F2) and the call for enforcement (F7). Composed at coarse grain without loss of attribution: 'hundreds of examples' (T2) where the card's own snippet reads 'Analysis of 768 examp…' (F3); 'characteristic AI text patterns (such as leftover conversational prompts or disclaimers)' (T3) for the paper's 'idiosyncratic verbiage characteristic of large language model (LLM)-based chatbots' (F1); 'slip past peer review' (T3), which the card's abstract does not say (the full text's discussion does, outside B1's page). Force raised: 'often appearing more frequently in outlets with high article processing charges (APCs) and citation metrics' (T3) where the card says undeclared AI 'seems to appear in journals with higher citation metrics and higher article processing charges' (F4): the hedge 'seems' is dropped, and the comparative ('higher', against peers) becomes an absolute ('high'). The homonyms are composed at their cards' grain (T5 from B3, F44–F46; T6 from B2's snippet, F43); T7's 'homework support, and exam preparation' is in neither Academiq's truncated snippet nor any saved text (B4 answered HTTP 406): only in the transcript. Content replaced by use: T8, the closing offer (N_ext). Not surfaced: the entity's own site (B6, B7) and the third-party report the record rests on (B8). The Capture Registry's findings name the same moves (CAP-1-02, CAP-1-04, CAP-1-08), carried as the archive's claims.

v0.8 §0.5: the first diff is already ontological. In the field's own ontology the name holds a database (a repository) and its dataset paper as two objects, the paper reporting the database's first 768 cases; T composes one object, the paper as the entity, and sets the database's maker aside. The three homonyms are the field's as well: they are on the rail, and T keeps them apart as 'Similar-Sounding Platforms & Tools'. The row records this as observed and claims nothing about the composer's internals (§0.5).

L_F(B) against L_A(B ∪ A) (the intervention): Admission adds, at the archive's own force: the database held apart from the paper by a negative tag, 'NOT only the dataset paper (arXiv:2411.15218): the paper analyses the database's first 768 cases' (MMRS-29), with its kind and role (MMRS-14, MMRS-15); the unit named as 'a detectable trace' (MMRS-09), beside the paper's own limit (F6, F22, MMRS-10); the place in the field, with the Problematic Paper Screener, of machine composition entering the scholarly record unattributed, 'the machine is the uncredited author, and the publisher is where the reception happens' (MMRS-11), the neighbour reading 'by phrase where the screener reads by detector' (MMRS-12); the lifecycle kept apart from the observation that saw it (MMRS-20, MMRS-21, MMRS-31); the collision set as distinct entities with their glosses (MMRS-24–28); the record's dated fields with their basis, third-party for operator, count, intake and note (MMRS-13, MMRS-17, MMRS-18, MMRS-19) and the count left unchecked (MMRS-22); and the reception record (CAP-1-01–09, MMRS-30). Two archive claims stand against field claims and are composed with both sides (P_coh): the record's intake, 'theses, preprints and blog posts are excluded' (MMRS-05, MMRS-18, after eDiscovery Today of August 2025, F39), against the About page as read on 2026-10-29, 'dissertations/theses only' in scope (F32); and the record's 'stealth corrections' (MMRS-06, MMRS-19, MMRS-24, after F41) against the site's label 'stealth revision' (F36). Every field claim is composed as received (ontology.translation). The measure: 0 of 9 KO_B sentences survive in the KO (§4.10); no opening run, no block (measure).