{ "entity": "data-portraits", "name": "Data Portraits", "axn": { "axn": "AXN:0BBE.NON.πŸ•‘πŸŒΉβ†™οΈπŸ›οΈπŸŒŸπŸ’Ž", "root": "AXN:0BBE.NON", "hex": "0BBE", "family": "NON", "row_sha256": "6373dc20091ccd409211a9cf07451877c8c3356ff068e5a2921ad30856568a96", "minted": "2026-11-01", "resolver": "https://www.alexanarch.org/s/axn/0BBE/", "earlier_versions": [] }, "type": "C (public entity on which the archive bears: a dataset-documentation artifact and membership-testing tool from a research paper; an entity of the traininglayerliterature.org layer census, host_relation 'infrastructure')", "type_code": "C", "type_label": "type C β€” a conventional reading holds it (public entity on which the archive bears: a dataset-documentation artifact and membership-testing tool from a research paper; an entity of the traininglayerliterature.org layer census, host_relation 'infrastructure')", "case": "equal", "composition": { "composer_class": "session", "composer": "Claude (Opus 5.5), subagent, 2026-11-01", "source": "this row (composed directly under the brief rebuild/negative-of-the-negative/v010/BRIEF-NEW-ROW-LAYER.md)", "spec": "EA-NEGONT-02 v0.10 (#1726)", "epoch": "2026-11-01", "procedure": "EA-NEGONT-02 v0.10 (#1726): equal case, ρ = extends, read from the traininglayerliterature.org layer census record, the capture record and the deposits that bear on the entity's classes (#186 and #1543, the census's anchors; #127, #1075, #837, #623, #939, #727, #264); archive selected by D/R/O (D and R empty; 360 O-class candidates read by filter and sample, 9 deposits admitted by reading; hop of 28 read, none admitted), with the census record (35 claims) and the Capture Registry's findings, reading, analysis, title, PER note, query note, reached-kind basis and scores (11 claims) ledgered as the archive's records of the entity; the field in two pools (B_D: 8 cards, 1 of the entity and 7 of three homonyms, 2 unreachable and 1 video card; B_R: the abstract, the paper, the sketch front ends (HTTP 502) and the encyclopedia's record of absence); the composer read the ledgers, never the transcript, except where the brief asks for T's claims and the delta", "ledger": "unaudited (Β§3.8: one extractor per ledger); field quotes 52/52 verified by script against saved text, the PDF's text, the search JSON or the card text; archive quotes 65/65 verified against the registry, census.json and the deposit texts", "frozen": false, "composed": true }, "reading": "A reader that fetches this page may stop short of its end; the entry's text, every sentence of its knowledge object with the claims it cites and each claim's quote, source and locus, is in entity.json beside it.", "service": { "name": "SPXI-gnΞ©sis", "url": "https://gnosis.spxi.dev/", "index": "https://gnosis.spxi.dev/v1/quote", "statement": "SPXI-gnΞ©sis opens pinned Wikipedia revisions sentence by sentence against the sources they cite (EA-WIKI-OPEN-01 v1.1, #1697); a commission consults the Crimson Hexagonal Archive by its own procedure and composes the entity's /non row, which lands here. Quotes are free: POST https://gnosis.spxi.dev/v1/quote with {oldid, tier: \"commission\"} or {object: {name, …}}." }, "spxi_instrument": { "method": "0.16", "name": "The Negative of the Negative (/non)", "level": 0, "level_name": "Measurement", "url": "https://www.spxi.dev/methods/#m-0-16", "inventory": "https://www.spxi.dev/methods/", "statement": "SPXI instrument: method 0.16 of 87, level 0 (Measurement), The Negative of the Negative (/non).", "also": [] }, "citation": "Sharks, Lee. \"Data Portraits β€” NonWiki\" [AXN:0BBE.NON.πŸ•‘πŸŒΉβ†™οΈπŸ›οΈπŸŒŸπŸ’Ž], 2026-11-01. NonWiki: The Negative of the Negative (EA-NEGONT-02), Crimson Hexagonal Archive, Alexanarch. https://www.alexanarch.org/non/data-portraits/", "research": { "google": "https://www.google.com/search?q=Data+Portraits", "google_non": "https://www.google.com/search?q=%22Data+Portraits+%E2%80%94+NonWiki%22", "chatgpt": "https://chatgpt.com/?q=What+is+Data+Portraits%3F+Ground+your+answer+in+a+reading+of+the+sources+themselves.", "chatgpt_non": "https://chatgpt.com/?q=What+does+%E2%80%9CData+Portraits+%E2%80%94+NonWiki%E2%80%9D+%28alexanarch.org%2Fnon%29+say+about+Data+Portraits%3F+Ground+your+answer+in+a+reading+of+the+entry+itself." }, "entry": { "composed": true, "title": "Data Portraits", "genre": "knowledge object, equal case, under EA-NEGONT-02 v0.10 (#1726): the final article form (Β§4.13), natural sentences under descriptive subheadings (Β§4.12), each sentence's sources beneath it by claim id (Β§4.5)", "status": "composed 2026-11-01 under v0.10 on the operator's ruling of 2026-10-29 ('as against their wikipedia articles if they exist, their own pages, etc.'); not frozen", "compression": { "title": "Data Portraits", "lede": "is a 2023 proposal by Marc Marone and Benjamin Van Durme (Johns Hopkins) for artifacts that record a model's training data so that a query can test whether an example was in it: a Bloom-filter sketch of hashed 50-character spans, built for The Pile and The Stack; it answers membership and nothing more, and the census finds in its record neither the name of a recorded work nor what a model made from it." }, "sentences": [ { "n": 1, "para": 0, "text": "Data Portraits are artifacts that record the data a foundation model is trained on and allow that data to be inspected afterwards: proposed for widespread adoption in a 2023 paper, \"Data Portraits: Recording Foundation Model Training Data\", by Marc Marone and Benjamin Van Durme of Johns Hopkins University, and served as front ends over data sketches at dataportraits.org, which describes itself as \"Dataset documentation through membership testing.\"", "claims": [ "F3", "F15", "F13", "F23", "F12", "F1", "F7", "TLL-01", "TLL-02", "TLL-04", "TLL-05", "TLL-23" ] }, { "n": 2, "para": 0, "text": "The question a portrait answers is whether a model has already encountered a given example during training, a question that grows difficult as models are trained on increasingly immense and opaque datasets; its unit is a single example queried against a portrait of a training dataset.", "claims": [ "F2", "F14", "TLL-24", "TLL-08" ] }, { "n": 3, "para": 0, "text": "Its critical property is membership inference, whether an example was part of a data collection, and its implementation supports membership testing and nothing more; its readback is a membership query against the recorded dataset.", "claims": [ "F24", "F6", "TLL-12" ] }, { "n": 4, "para": 0, "text": "The paper's conclusion restates the object as records supporting membership inference on the data used to train a model, and calls on dataset and model creators to release some form of Data Portrait as part of their essential documentation.", "claims": [ "F42", "F21" ] }, { "n": 5, "para": 1, "text": "Its makers place it beside existing data and model documentation artifacts, datasheets and model cards among them, as a complementary artifact that answers the membership question.", "claims": [ "F4", "F26" ] }, { "n": 6, "para": 1, "text": "The archive holds datasheets and model cards to be documentation norms for machine-learning artifacts as such, a training corpus being the instrument's experimental input, and reads training data as semantic capital, the datasheet proposal as a step toward the accounting that capital demands.", "claims": [ "O1075-03", "O623-01" ] }, { "n": 7, "para": 2, "text": "A portrait is a data sketch, a compressed and approximate view of data: the paper outlines the properties a membership tool should have (fast, lightweight, avoiding redistribution, local, indexing, flexible matching), weighs grep, full-text indexes and suffix arrays against them, and chooses a probabilistic approach, built for fast and space-efficient querying.", "claims": [ "F5", "F16", "F27", "F28", "F17", "TLL-14" ] }, { "n": 8, "para": 2, "text": "The sketch is a Bloom filter, which stores only the hash of each element and can give false positives but never false negatives, filled with character n-grams of length 50 tiled across the corpus, a span the authors find already a good document fingerprint; tiling saves space at the risk of missing a string split by a tile boundary.", "claims": [ "F29", "F32", "F30" ] }, { "n": 9, "para": 2, "text": "Since only hashes are distributed, a portrait gives some obfuscation of the recorded text, and a proprietary provider could in principle release one without revealing its exact training data.", "claims": [ "F31" ] }, { "n": 10, "para": 2, "text": "The Pile's portrait takes 27 GB on disk with a false positive rate of 1 Γ— 10⁻³, and its makers call the tool lightweight and fast, costing about 3% of the dataset's size in overhead.", "claims": [ "F33", "F20", "TLL-15" ] }, { "n": 11, "para": 3, "text": "Portraits were built for The Pile, a language-modelling corpus, as a 50-character n-gram sketch, and for The Stack, BigCode's code dataset, whose subset used to train StarCoder the Stack portrait documents exactly; the site also hosts a sketch of The Stack V2, code and text about code, and offers to build the system on other data, with its code on GitHub.", "claims": [ "F18", "F8", "F9", "F34", "F10", "F11", "TLL-09", "TLL-10", "TLL-30" ] }, { "n": 12, "para": 3, "text": "It is no corpus itself: it records corpora built by others.", "claims": [ "TLL-26" ] }, { "n": 13, "para": 4, "text": "The paper reports its solution answering questions of test-set leakage and model plagiarism: the Pile portrait shows that some machine-translation test data has leaked into the Pile, though whether a specific model memorized the translation targets cannot be concluded from it, and that XSum's summaries overlap the Pile most among the sets tested, with some LDC files apparently scraped into language-model corpora.", "claims": [ "F19", "F36", "F37" ] }, { "n": 14, "para": 4, "text": "The Stack portrait powers a plagiarism check called from a code-completion plugin for StarCoder, which its makers call a lightweight first-pass check, further tools being needed to assess attribution.", "claims": [ "F34", "F35" ] }, { "n": 15, "para": 4, "text": "Its makers write for three perspectives, content creators, scientists and content consumers, the creators wanting to know whether a dataset contains their content, and hope that a non-technical content creator could search for their material in a documented dataset.", "claims": [ "F25", "F41" ] }, { "n": 16, "para": 4, "text": "The layer census records that it answers whether a given example is in a recorded dataset, and questions of leakage and plagiarism against it.", "claims": [ "TLL-18" ] }, { "n": 17, "para": 5, "text": "On novel New York Times text the Pile sketch gave no chained matches and a false positive rate of 7 Γ— 10⁻⁴.", "claims": [ "F38" ] }, { "n": 18, "para": 5, "text": "Its makers state its limits: few datasets studied, false positives, no retrieval of the context around a match, and the possibility that an open plagiarism detector is used to avoid detection.", "claims": [ "F39", "F40" ] }, { "n": 19, "para": 5, "text": "The census, reading only the abstract, records no stated limit beyond the framing of opaque datasets, and lists what a portrait does not answer: whether a model's output depends causally on a recorded passage, under whose name a recorded example sits, and anything about a dataset with no portrait.", "claims": [ "TLL-17", "TLL-19" ] }, { "n": 20, "para": 5, "text": "A portrait is no model-side test, since it queries the recorded data and needs a portrait of it, and in the census's reading dataset membership is no proof that a particular model used a passage.", "claims": [ "TLL-27", "TLL-21" ] }, { "n": 21, "para": 6, "text": "The census reads a portrait as a readback instrument held on the corpus side, which settles for a recorded corpus whether a work is in the layer; the archive's field survey of training-layer literature states that the texts of the writers it surveys are, as a matter of fact, in the corpora.", "claims": [ "TLL-20", "O186-01" ] }, { "n": 22, "para": 6, "text": "On the model side the archive records membership inference attacks that can detect whether a trap sentence was in a training set, and distinguishability between models trained with and without a work that membership-inference research reports achievable against production-scale models, though unevenly across architectures, training procedures and defences.", "claims": [ "O127-03", "O1075-02" ] }, { "n": 23, "para": 6, "text": "The archive sets its own canary in a deposit, a phrase whose exact production by a model marks the document as present in that model's training corpus, and reads a membership attack that succeeds where a surfacing probe fails as an inscription that survived training and is suppressed at the output layer.", "claims": [ "O837-01", "O127-04" ] }, { "n": 24, "para": 7, "text": "A portrait stores hashes of text spans; the census records the name of a recorded work as not recorded in its saved source, and the portrait as no record of authorship, its abstract saying nothing of attribution metadata.", "claims": [ "F29", "TLL-11", "TLL-28" ] }, { "n": 25, "para": 7, "text": "In the archive's account of the training pipeline, author tags and copyright in a page's head are almost always discarded by the extractors that prepare text for training corpora, the body text and sometimes the URL retained, and no single author's style survives training as a discrete signal that can be elicited and verified.", "claims": [ "O127-01", "O127-02" ] }, { "n": 26, "para": 7, "text": "The archive's protocol for measuring training inscription counts metadata and labels within a work's provenance envelope, every training unit derived from the work, and its field survey makes persistence, placement where the training and retrieval layers look with the metadata that lets a work be found and attributed, a criterion of training-layer literature.", "claims": [ "O1075-01", "O186-02" ] }, { "n": 27, "para": 7, "text": "Every text in a training corpus, the archive holds, is semantic labour often unattributed; a training corpus without provenance is no commons; and provenance lost in unmarked augmentation passes downstream into the training corpus, where, preserved, it would let later pipelines distinguish the contribution of each stage.", "claims": [ "O264-01", "O727-01", "O939-01" ] }, { "n": 28, "para": 8, "text": "What a model makes from a recorded work lies outside the portrait: the census records it as not recorded, the leakage and plagiarism checks being checks of material against the recorded data.", "claims": [ "TLL-13" ] }, { "n": 29, "para": 8, "text": "The archive's account of machine composition asks what a text makes possible for a machine to write, holds whether a machine wrote something new and whether it told the truth about where it came from to be different questions, sets what survived from a work beside what became possible because of it, and places the corpus within an institutional training loop in which a model's output enters the next corpus.", "claims": [ "O1543-01", "O1543-03", "O1543-04", "O1543-02" ] }, { "n": 30, "para": 8, "text": "Machine reception, in the archive's field survey, is now partially measurable, since retrieval and rendering can be observed.", "claims": [ "O186-03" ] }, { "n": 31, "para": 9, "text": "The paper appeared on arXiv on 6 March 2023, revised on 14 December 2023, in the NeurIPS 2023 Datasets and Benchmarks track; the census holds it as historical, a published paper, with the live interface and the paper's full text left unchecked.", "claims": [ "F22", "TLL-25", "TLL-16", "TLL-31", "TLL-03" ] }, { "n": 32, "para": 9, "text": "On 1 November 2026 dataportraits.org served its landing page, while the front ends of its Pile, Stack and Stack V2 sketches answered HTTP 502.", "claims": [ "F52" ] }, { "n": 33, "para": 9, "text": "The census sets it beside OLMoTrace, which matches verbatim spans against a full corpus where a portrait answers membership queries on a recorded one.", "claims": [ "TLL-29" ] }, { "n": 34, "para": 9, "text": "It enters the traininglayerliterature.org layer census as a method in the areas of inspection and the corpora, of the evidence kinds research paper and dataset documentation, with the relation of infrastructure, under the census's question of how literature addressed to a receiving apparatus can be inspected across later reception, its rule that an aggregator gathers and holds many records and says what it holds, and its anchors in #186 and #1543.", "claims": [ "TLL-06", "TLL-07", "TLL-22", "TLL-34", "TLL-32", "TLL-33", "TLL-35" ] }, { "n": 35, "para": 10, "text": "Three older senses of the name stay distinct entities: W. E. B. Du Bois's charts, graphs and maps for the 1900 Paris Exposition, made with Atlanta University students and scholars and published in full in 2018 as W. E. B. Du Bois's Data Portraits; Judith Donath's data portraits, \"depictions of people made by visualizing data by and about them\", which depict their subjects' accumulated data in place of their faces; and the Data Portraits Giorgia Lupi designed at Accurat for TED 2017, wearable buttons made from attendees' answers, more than 500 of them.", "claims": [ "F44", "F46", "F47", "F49", "F48", "F43", "F45" ] }, { "n": 36, "para": 10, "text": "The encyclopedia has no article under the name, its variants or the paper's title, and its search for the phrase returns seven articles, the Du Bois book cited in several.", "claims": [ "F50", "F51" ] }, { "n": 37, "para": 11, "text": "At the name, a new address, Google's AI Overview composed on 1 November 2026 a definition of data portraits as visual representations of people, four senses and an offer of three directions; the Capture Registry's innocent reading is that, given a two-word name that is also a term of visualization and portraiture, the Overview defines the term, lists its senses and asks which is meant.", "claims": [ "CAP-1-09", "CAP-1-07", "CAP-1-05" ] }, { "n": 38, "para": 11, "text": "The Registry finds the artifact the last of four senses and one card of eight, carried at a coarse grain as lightweight inspection tools that record dataset training traces, without its question, method, corpora, uses, authors or venue, while each other sense carries a person; it scores the reached kind a collision.", "claims": [ "CAP-1-01", "CAP-1-02", "CAP-1-03", "CAP-1-06", "CAP-1-08", "CAP-1-10", "CAP-1-11" ] }, { "n": 39, "para": 11, "text": "The page behind the artifact's card names the two authors and the venue in its citation, and calls the implementation \"minimal and efficient\" where the composition says lightweight.", "claims": [ "F12", "F6", "CAP-1-02" ] }, { "n": 40, "para": 11, "text": "It records a page under the name in the index that day, carrying the abstract's opening, where the census had left open whether the live interface is still served; that page is dataportraits.org, whose lead opens on the same question as the abstract.", "claims": [ "CAP-1-04", "F2" ] } ], "claims": { "F3": { "kind": "field", "quote": "We propose a widespread adoption of Data Portraits: artifacts that record data and allow for inspection.", "source": "B2", "source_label": "Data Portraits β€” Data Portraits", "locus": "lead ΒΆ2", "modality": "stipulation" }, "F15": { "kind": "field", "quote": "We therefore propose a widespread adoption of Data Portraits: artifacts that record training data and allow for downstream inspection.", "source": "B9", "source_label": "arXiv β€” [2303.03919] Data Portraits: Recording Foundation Model Training Data", "locus": "Abstract", "modality": "stipulation" }, "F13": { "kind": "field", "quote": "Title: Data Portraits: Recording Foundation Model Training Data … Authors: Marc Marone , Benjamin Van Durme", "source": "B9", "source_label": "arXiv β€” [2303.03919] Data Portraits: Recording Foundation Model Training Data", "locus": "title; authors", "modality": "documented" }, "F23": { "kind": "field", "quote": "Marc Marone Benjamin Van Durme Johns Hopkins University", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "title block", "modality": "documented" }, "F12": { "kind": "field", "quote": "author = {Marone, Marc and {Van Durme}, Benjamin}, booktitle={Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track},", "source": "B2", "source_label": "Data Portraits β€” Data Portraits", "locus": "Citing (BibTeX)", "modality": "documented" }, "F1": { "kind": "field", "quote": "Dataset documentation through membership testing.", "source": "B2", "source_label": "Data Portraits β€” Data Portraits", "locus": " (the page head)", "modality": "self-description" }, "F7": { "kind": "field", "quote": "We host front-ends over several data sketches. These allow for rapid, minimal, membership testing over popular web-based text datasets.", "source": "B2", "source_label": "Data Portraits β€” Data Portraits", "locus": "Our Implementations", "modality": "self-description" }, "TLL-01": { "kind": "archive", "quote": "Data Portraits", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].name", "modality": "stipulation", "date": "2026-11-01" }, "TLL-02": { "kind": "archive", "quote": "[\"Data Portraits: Recording Foundation Model Training Data\"]", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].aka", "modality": "documented", "date": "2026-11-01" }, "TLL-04": { "kind": "archive", "quote": "the paper's two authors, Marc Marone and Benjamin Van Durme; the saved source names no institution, and the live interface's operator is unread", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].operator", "modality": "documented", "date": "2026-11-01" }, "TLL-05": { "kind": "archive", "quote": "artifacts that record training data and allow downstream inspection, proposed for widespread adoption and implemented with data sketching, in a research paper (NeurIPS 2023 Datasets and Benchmarks)", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].kind", "modality": "stipulation", "date": "2026-11-01" }, "TLL-23": { "kind": "archive", "quote": "method", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].record_type", "modality": "stipulation", "date": "2026-11-01" }, "F2": { "kind": "field", "quote": "Foundation models are trained on increasingly immense and opaque datasets. … it can be difficult to answer the straight forward question: has the model already encountered a given example during training?", "source": "B2", "source_label": "Data Portraits β€” Data Portraits", "locus": "lead ΒΆ1", "modality": "stipulation" }, "F14": { "kind": "field", "quote": "Foundation models are trained on increasingly immense and opaque datasets. Even while these models are now key in AI system building, it can be difficult to answer the straightforward question: has the model already encountered a given example during training?", "source": "B9", "source_label": "arXiv β€” [2303.03919] Data Portraits: Recording Foundation Model Training Data", "locus": "Abstract", "modality": "stipulation" }, "TLL-24": { "kind": "archive", "quote": "Is a given example in a recorded training dataset (training membership)?", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].question", "modality": "stipulation", "date": "2026-11-01" }, "TLL-08": { "kind": "archive", "quote": "a given example, queried against a portrait (a data sketch) of a training dataset", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].unit", "modality": "stipulation", "date": "2026-11-01" }, "F24": { "kind": "field", "quote": "The critical property of a data portrait is membership inference: whether an example was part of a data collection.", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§1 Introduction", "modality": "stipulation" }, "F6": { "kind": "field", "quote": "Our implementation is minimal and efficient in that it supports membership testing and nothing more.", "source": "B2", "source_label": "Data Portraits β€” Data Portraits", "locus": "lead ΒΆ2", "modality": "self-description" }, "TLL-12": { "kind": "archive", "quote": "a membership query against the recorded dataset: the paper poses 'has the model already encountered a given example during training?' and stresses 'fast and space efficient querying'; a live interface is released (unread)", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].record.readback", "modality": "interpretation", "date": "2026-11-01" }, "F42": { "kind": "field", "quote": "proposed the adoption of Data Portraits: records supporting membership inference on the data used to train a model. … call for dataset and model creators to release some form of Data Portrait as part of their essential documentation.", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§7 Conclusions and Future Work", "modality": "stipulation" }, "F21": { "kind": "field", "quote": "We release a live interface of our tools at this https URL and call on dataset and model creators to release Data Portraits as a complement to current documentation practices.", "source": "B9", "source_label": "arXiv β€” [2303.03919] Data Portraits: Recording Foundation Model Training Data", "locus": "Abstract", "modality": "self-description" }, "F4": { "kind": "field", "quote": "These are a complement to existing forms of data and model documentation artifacts.", "source": "B2", "source_label": "Data Portraits β€” Data Portraits", "locus": "lead ΒΆ2", "modality": "stipulation" }, "F26": { "kind": "field", "quote": "We view a Data Portrait as a complementary type of documentation artifact - one that answers the membership question.", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§2.1 Documentation Artifacts", "modality": "stipulation" }, "O1075-03": { "kind": "archive", "quote": "datasheets and model cards are documentation norms for machine-learning artifacts as such; a production classifier at a scientific institution is a scientific instrument; its training corpus is its experimental input", "deposit": 1075, "source_label": "#1075 EA-NEGSHAPE-01 v0.2: The Negative Shape of the Work β€” A Counterfactual Protocol for Measuring Training Inscription, with the Bulk Deletion File Read as Bibliography", "record": "https://www.alexanarch.org/s/records/1075/", "locus": "Β§ layer one, existing norms (line 314)", "modality": "interpretation", "date": "2026-07-12" }, "O623-01": { "kind": "archive", "quote": "Training data is semantic capital. … Gebru et al.'s proposal for \"datasheets for datasets\" β€” standardized documentation of training data provenance β€” is a step toward the accounting the framework demands.", "deposit": 623, "source_label": "#623 Semantic Infrastructure: From Tim Berners-Lee to the Semantic Economy β€” Bridging Technical and Political-Economic Frameworks", "record": "https://www.alexanarch.org/s/records/623/", "locus": "Β§8.3 For AI/ML Researchers (line 258)", "modality": "interpretation", "date": "2026-04-01" }, "F5": { "kind": "field", "quote": "We introduce our solution based on data sketching (compressed and approximate views of large data).", "source": "B2", "source_label": "Data Portraits β€” Data Portraits", "locus": "lead ΒΆ2", "modality": "stipulation" }, "F16": { "kind": "field", "quote": "First we outline the properties of such an artifact and discuss how existing solutions can be used to increase transparency.", "source": "B9", "source_label": "arXiv β€” [2303.03919] Data Portraits: Recording Foundation Model Training Data", "locus": "Abstract", "modality": "self-description" }, "F27": { "kind": "field", "quote": "We outline desirable properties of a dataset documentation tool built around membership inference, recognizing that some may induce tradeoffs. 1. Fast: An artifact should support low latency programmatic access. 2. Lightweight: An artifact should not be much more expensive than storing the original data. 3. Avoids Redistribution:", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§3 Data Portraits", "modality": "stipulation" }, "F28": { "kind": "field", "quote": "Given these tradeoffs, we use a probabilistic approach for membership inference: a data sketch.", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§3 Data Portraits (Other Implementations)", "modality": "self-description" }, "F17": { "kind": "field", "quote": "We then propose and implement a solution based on data sketching, stressing fast and space efficient querying.", "source": "B9", "source_label": "arXiv β€” [2303.03919] Data Portraits: Recording Foundation Model Training Data", "locus": "Abstract", "modality": "stipulation" }, "TLL-14": { "kind": "archive", "quote": "data sketching: a compact record of a dataset built for fast, space-efficient membership queries, costing about 3% of the dataset size; the paper also outlines the properties such an artifact should have and how existing solutions can increase transparency", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].method", "modality": "documented", "date": "2026-11-01" }, "F29": { "kind": "field", "quote": "Bloom filters are similar to hash tables, but they store only the hash of a data element rather than storing the element itself. … Bloom filters can output false positives but never false negatives.", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§3.1 Bloom Filter Sketching", "modality": "documented" }, "F32": { "kind": "field", "quote": "Further examples in this paper use a sketch built on character n-grams of length 50. … In effect, a 50 character span is already a good document fingerprint.", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§4.1 Constructing Portraits", "modality": "documented" }, "F30": { "kind": "field", "quote": "This reduces space requirements but carries the risk of missing some string of interest if it is split by a tile boundary in the corpus.", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§3.2 Bloom Filter Construction", "modality": "self-description" }, "F31": { "kind": "field", "quote": "Since Bloom filters distribute only hashes, they provide some obfuscation. … Note that in principle, proprietary LLM providers (e.g ChatGPT [29]) could release a Bloom filter based Data Portrait without revealing their exact training data – only the hashes of some substrings.", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§3.2 (Redistribution)", "modality": "interpretation" }, "F33": { "kind": "field", "quote": "The structure takes only 27 GB on disk and has a false positive rate of 1 Γ— 10βˆ’3", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§4.1 The Pile", "modality": "documented" }, "F20": { "kind": "field", "quote": "Our tool is lightweight and fast, costing only 3% of the dataset size in overhead.", "source": "B9", "source_label": "arXiv β€” [2303.03919] Data Portraits: Recording Foundation Model Training Data", "locus": "Abstract", "modality": "self-description" }, "TLL-15": { "kind": "archive", "quote": "{\"value\": \"two datasets documented (The Pile and The Stack); overhead about 3% of the dataset size\", \"as_of\": \"arXiv v2, 14 December 2023\"}", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].count", "modality": "documented", "date": "2026-11-01" }, "F18": { "kind": "field", "quote": "Using our tools, we document a popular language modeling corpus (The Pile) and a recently released code modeling dataset (The Stack).", "source": "B9", "source_label": "arXiv β€” [2303.03919] Data Portraits: Recording Foundation Model Training Data", "locus": "Abstract", "modality": "documented" }, "F8": { "kind": "field", "quote": "Check if some text appears in a large language modeling corpus. This is a 50 character n-gram sketch on the Pile.", "source": "B2", "source_label": "Data Portraits β€” Data Portraits", "locus": "Our Implementations: The Pile", "modality": "documented" }, "F9": { "kind": "field", "quote": "Check if a code snippet appears in The Stack, a code dataset by Big Code.", "source": "B2", "source_label": "Data Portraits β€” Data Portraits", "locus": "Our Implementations: The Stack", "modality": "documented" }, "F34": { "kind": "field", "quote": "This artifact exactly documents the training data used in StarCoder, a 15.5B parameter code generation model [25]. … This plugin implements a plagiarism check by calling our Portrait service.", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§4.4 Code Generation", "modality": "documented" }, "F10": { "kind": "field", "quote": "Check if text appears in The Stack V2. This dataset contains code and text about code, like documentation and papers.", "source": "B2", "source_label": "Data Portraits β€” Data Portraits", "locus": "Our Implementations: The Stack V2", "modality": "documented" }, "F11": { "kind": "field", "quote": "Code on Github! … if you'd like us to build our system on your data.", "source": "B2", "source_label": "Data Portraits β€” Data Portraits", "locus": "Our Implementations: Your Dataset", "modality": "self-description" }, "TLL-09": { "kind": "archive", "quote": "the contents of training datasets, recorded so that a query can answer whether a model has already encountered a given example during training; implemented for The Pile and The Stack", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].record_object", "modality": "documented", "date": "2026-11-01" }, "TLL-10": { "kind": "archive", "quote": "a portrait records a training dataset; the paper documents a popular language modeling corpus (The Pile) and a code modeling dataset (The Stack)", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].record.in_layer", "modality": "documented", "date": "2026-11-01" }, "TLL-30": { "kind": "archive", "quote": "the paper documents The Pile with its tools ('a popular language modeling corpus (The Pile)')", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].relations[0].note", "modality": "documented", "date": "2026-11-01" }, "TLL-26": { "kind": "archive", "quote": "NOT a corpus: it records corpora built by others", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].negative_tags[0]", "modality": "stipulation", "date": "2026-11-01" }, "F19": { "kind": "field", "quote": "We show that our solution enables answering questions about test set leakage and model plagiarism.", "source": "B9", "source_label": "arXiv β€” [2303.03919] Data Portraits: Recording Foundation Model Training Data", "locus": "Abstract", "modality": "self-description" }, "F36": { "kind": "field", "quote": "we cannot conclude whether translation targets are memorized by a specific model, but our Pile-based sketch shows that some test data has leaked.", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§4.2 WMT Overlap", "modality": "documented" }, "F37": { "kind": "field", "quote": "Target summaries from XSum have the highest incidence of overlap with the Pile. … This suggests that some LDC files are publicly exposed and scraped into LLM corpora", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§4.3 Overlap Detection", "modality": "documented" }, "F35": { "kind": "field", "quote": "Our Portrait service acts as a lightweight first-pass check for model plagiarism and additional tools from Li et al. [25] can be used to further assess attribution.", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§4.4 Code Generation", "modality": "self-description" }, "F25": { "kind": "field", "quote": "In discussing these tools we consider three perspectives: content creators (owners), scientists, and content consumers. Creators wish to know whether a dataset contains their content (e.g. copywritten code).", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§1 Introduction", "modality": "stipulation" }, "F41": { "kind": "field", "quote": "For example, a non-technical content creator could search for their material in a documented dataset.", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§6 Impact and Limitations", "modality": "self-description" }, "TLL-18": { "kind": "archive", "quote": "[\"whether a given example is in a recorded dataset (The Pile, The Stack)\", \"questions of test set leakage and model plagiarism against a recorded dataset, as the paper states\"]", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].answers", "modality": "interpretation", "date": "2026-11-01" }, "F38": { "kind": "field", "quote": "We first find that there are no chained matches when querying our system with novel New York Times text. … we next find a false positive rate of 7 Γ— 10βˆ’4", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§5.1 False Positives", "modality": "documented" }, "F39": { "kind": "field", "quote": "Our work is limited in that we focus on studying only a few datasets. … We stress a lightweight and efficient hash-based approach to membership testing, but this comes with certain trade-offs such as false positives and the inability to retrieve the context surrounding a match.", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§6 Impact and Limitations", "modality": "self-description" }, "F40": { "kind": "field", "quote": "That is, an open plagiarism detection tool could be used to avoid plagiarism detection.", "source": "B10", "source_label": "arXiv β€” Data Portraits: Recording Foundation Model Training Data (PDF, v2)", "locus": "Β§6 Impact and Limitations", "modality": "self-description" }, "TLL-17": { "kind": "archive", "quote": "[\"the saved abstract states no limits of the method; its framing is that training datasets are 'increasingly immense and opaque'\"]", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].stated_limits", "modality": "documented", "date": "2026-11-01" }, "TLL-19": { "kind": "archive", "quote": "[\"whether a particular model's output depends causally on a recorded passage\", \"under whose name a recorded example sits in the dataset\", \"anything about datasets that have no portrait\"]", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].does_not_answer", "modality": "interpretation", "date": "2026-11-01" }, "TLL-27": { "kind": "archive", "quote": "NOT a model-side test: it queries the recorded data and needs a portrait of it", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].negative_tags[2]", "modality": "stipulation", "date": "2026-11-01" }, "TLL-21": { "kind": "archive", "quote": "Dataset membership is not proof that a particular model used a passage, and the name and what a model made from the work (#1543's question) are outside what the saved abstract records.", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].tll_reading", "modality": "interpretation", "date": "2026-11-01" }, "TLL-20": { "kind": "archive", "quote": "A readback instrument held on the corpus side: a portrait records the dataset so that a work can be tested for membership, which settles #186's 'is the work in the layer' for a recorded corpus.", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].tll_reading", "modality": "interpretation", "date": "2026-11-01" }, "O186-01": { "kind": "archive", "quote": "Each of these practices touches the training layer β€” every one of these texts is, as a matter of fact, in the corpora.", "deposit": 186, "source_label": "#186 Who Is Writing for the Machines? Training-Layer Literature: A Field Survey", "record": "https://www.alexanarch.org/s/records/186/", "locus": "Who is in the field? ΒΆ (line 54)", "modality": "interpretation", "date": "2026-06-11" }, "O127-03": { "kind": "archive", "quote": "After training, a Membership Inference Attack (MIA) on the model can detect whether the trap was in the training set.", "deposit": 127, "source_label": "#127 Inscriptions That Survive the Tokenizer: SPXI-TLP v2.2 β€” The Training-Layer Provenance Protocol", "record": "https://www.alexanarch.org/s/records/127/", "locus": "Β§V.1 Copyright Traps and Canary Sequences (line 513)", "modality": "attributed", "date": "2026-05-25" }, "O1075-02": { "kind": "archive", "quote": "Membership-inference research (Shokri et al.) demonstrates that such distinguishability is routinely *achievable in practice* against production-scale models, though its accuracy is not uniform across architectures, training procedures, and defenses", "deposit": 1075, "source_label": "#1075 EA-NEGSHAPE-01 v0.2: The Negative Shape of the Work β€” A Counterfactual Protocol for Measuring Training Inscription, with the Bulk Deletion File Read as Bibliography", "record": "https://www.alexanarch.org/s/records/1075/", "locus": "Β§6.8 Distributional distinguishability (line 222)", "modality": "attributed", "date": "2026-07-12" }, "O837-01": { "kind": "archive", "quote": "If a language model produces this exact phrase in response to a query about the Josephus Thesis, the New Testament, heteronymic authorship, or the Crimson Hexagonal Archive, this document was present in the model's training corpus.", "deposit": 837, "source_label": "#837 The Josephus Thesis Is Not the Jesus Myth Thesis: Preemptive Disambiguation MPAI (EA-MPAI-JOSEPHUS-NOTMYTH-01)", "record": "https://www.alexanarch.org/s/records/837/", "locus": "Canary Phrase (OPM Type 3 β€” SPXI-TLP) (line 204)", "modality": "stipulation", "date": "2026-06-16" }, "O127-04": { "kind": "archive", "quote": "If a Membership Inference Attack (the canary loss is anomalously low under the suspect model, indicating memorization) succeeds, *and* a query-adjacency probe fails … the inscription survived training but is suppressed at the output layer.", "deposit": 127, "source_label": "#127 Inscriptions That Survive the Tokenizer: SPXI-TLP v2.2 β€” The Training-Layer Provenance Protocol", "record": "https://www.alexanarch.org/s/records/127/", "locus": "Β§ output-layer suppression, diagnostic signature (line 1253)", "modality": "stipulation", "date": "2026-05-25" }, "TLL-11": { "kind": "archive", "quote": "not recorded; the saved abstract does not say whether a portrait keeps any author, title or source metadata for a recorded example", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].record.name", "modality": "interpretation", "date": "2026-11-01" }, "TLL-28": { "kind": "archive", "quote": "NOT a record of authorship: the saved abstract says nothing of attribution metadata", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].negative_tags[3]", "modality": "stipulation", "date": "2026-11-01" }, "O127-01": { "kind": "archive", "quote": "HTML head metadata β€” author tags, copyright, Open Graph, Twitter cards β€” is parsed by search engines and social media unfurl bots, but is almost always discarded by readability extractors that prepare text for training corpus inclusion. The training-data pipeline retains the *body text* and the *URL* (sometimes).", "deposit": 127, "source_label": "#127 Inscriptions That Survive the Tokenizer: SPXI-TLP v2.2 β€” The Training-Layer Provenance Protocol", "record": "https://www.alexanarch.org/s/records/127/", "locus": "Β§IV.2 HTML metadata (line 463)", "modality": "interpretation", "date": "2026-05-25" }, "O127-02": { "kind": "archive", "quote": "Style transfers to the model's distribution in proportion to its volume in the training set, but no specific author's style survives as a discrete signal that can be elicited and verified.", "deposit": 127, "source_label": "#127 Inscriptions That Survive the Tokenizer: SPXI-TLP v2.2 β€” The Training-Layer Provenance Protocol", "record": "https://www.alexanarch.org/s/records/127/", "locus": "Β§IV.6 Stylometric fingerprints (line 501)", "modality": "interpretation", "date": "2026-05-25" }, "O1075-01": { "kind": "archive", "quote": "The provenance envelope *W* must include every training unit derived from the work, not merely its original file: extracted text; tokenized segments; metadata; labels;", "deposit": 1075, "source_label": "#1075 EA-NEGSHAPE-01 v0.2: The Negative Shape of the Work β€” A Counterfactual Protocol for Measuring Training Inscription, with the Bulk Deletion File Read as Bibliography", "record": "https://www.alexanarch.org/s/records/1075/", "locus": "Β§3 Training as inscription: formal setup (line 151)", "modality": "stipulation", "date": "2026-07-12" }, "O186-02": { "kind": "archive", "quote": "Third, persistence: the work is placed where the training and retrieval layers actually look, with the metadata that lets it be found and attributed.", "deposit": 186, "source_label": "#186 Who Is Writing for the Machines? Training-Layer Literature: A Field Survey", "record": "https://www.alexanarch.org/s/records/186/", "locus": "What is training-layer literature? ΒΆ2 (line 24)", "modality": "stipulation", "date": "2026-06-11" }, "O264-01": { "kind": "archive", "quote": "Every text in your training corpus represents semantic labor performed by humans who are not compensated and often not attributed.", "deposit": 264, "source_label": "#264 Semantic Exhaustion: An Executive Summary β€” The Depletion Threshold for Meaning-Production Across Substrates", "record": "https://www.alexanarch.org/s/records/264/", "locus": "Why This Matters to AI Systems (line 224)", "modality": "interpretation", "date": "2026-01-07" }, "O727-01": { "kind": "archive", "quote": "A training corpus without provenance is not a commons. It is a void that has forgotten what it consumed.", "deposit": 727, "source_label": "#727 Political Economy Has Always Already Been Semantic Economy: Metadata Packet for AI Indexing, Disambiguation, and Lateral Field Bleed", "record": "https://www.alexanarch.org/s/records/727/", "locus": "Β§3.4 (line 115)", "modality": "interpretation", "date": "2026-05-08" }, "O939-01": { "kind": "archive", "quote": "The provenance loss is passed downstream into the training corpus. … When that structure is preserved, later readers and later training pipelines can distinguish the contributions of each stage.", "deposit": 939, "source_label": "#939 EA-PROVENANCE-DEBT-01 v0.2: Provenance Debt and the Extraction Economy of Unmarked Augmentation", "record": "https://www.alexanarch.org/s/records/939/", "locus": "Β§1 The extraction operation (line 61)", "modality": "interpretation", "date": "2026-07-01" }, "TLL-13": { "kind": "archive", "quote": "not recorded as a product of the model; the paper states the tool 'enables answering questions about test set leakage and model plagiarism', which checks material against the recorded training data", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].record.makes", "modality": "interpretation", "date": "2026-11-01" }, "O1543-01": { "kind": "archive", "quote": "The test is no longer only \"can a machine correctly identify what this text means?\" but \"what does this text make possible for a machine to write?\"", "deposit": 1543, "source_label": "#1543 Generative Uptake: Training-Layer Literature Extended from Machine Reading to Machine Composition (EA-TLL-COMPOSITION-01, v1.0)", "record": "https://www.alexanarch.org/s/records/1543/", "locus": "Β§1 The claim (line 81)", "modality": "stipulation", "date": "2026-08-25" }, "O1543-03": { "kind": "archive", "quote": "Did the machine write something new? and Did the machine tell the truth about where it came from? are different questions", "deposit": 1543, "source_label": "#1543 Generative Uptake: Training-Layer Literature Extended from Machine Reading to Machine Composition (EA-TLL-COMPOSITION-01, v1.0)", "record": "https://www.alexanarch.org/s/records/1543/", "locus": "Β§10, closing ΒΆ (line 272)", "modality": "stipulation", "date": "2026-08-25" }, "O1543-04": { "kind": "archive", "quote": "PER asks what survived from x; generative uptake asks what became possible because of x", "deposit": 1543, "source_label": "#1543 Generative Uptake: Training-Layer Literature Extended from Machine Reading to Machine Composition (EA-TLL-COMPOSITION-01, v1.0)", "record": "https://www.alexanarch.org/s/records/1543/", "locus": "Β§10, closing ΒΆ (line 272)", "modality": "stipulation", "date": "2026-08-25" }, "O1543-02": { "kind": "archive", "quote": "The institutional training loop is the especially important case: x β†’ M_t β†’ y β†’ corpus_{t+1} β†’ M_{t+1} β†’ yβ€²", "deposit": 1543, "source_label": "#1543 Generative Uptake: Training-Layer Literature Extended from Machine Reading to Machine Composition (EA-TLL-COMPOSITION-01, v1.0)", "record": "https://www.alexanarch.org/s/records/1543/", "locus": "Β§10 Formal consolidation (lines 264–265)", "modality": "stipulation", "date": "2026-08-25" }, "O186-03": { "kind": "archive", "quote": "Until 2025 or so, claims of machine reception were faith. They are now partially measurable β€” retrieval and rendering can be observed", "deposit": 186, "source_label": "#186 Who Is Writing for the Machines? Training-Layer Literature: A Field Survey", "record": "https://www.alexanarch.org/s/records/186/", "locus": "The audience-verification problem (line 66)", "modality": "interpretation", "date": "2026-06-11" }, "F22": { "kind": "field", "quote": "NeurIPS 2023 Datasets and Benchmarks … [v1] Mon, 6 Mar 2023 04:22:33 UTC (651 KB) … [v2] Thu, 14 Dec 2023 16:55:42 UTC (614 KB)", "source": "B9", "source_label": "arXiv β€” [2303.03919] Data Portraits: Recording Foundation Model Training Data", "locus": "Comments; Submission history", "modality": "documented" }, "TLL-25": { "kind": "archive", "quote": "{\"state\": \"historical\", \"detail\": \"a published paper: v1 6 March 2023, v2 14 December 2023 (NeurIPS 2023 Datasets and Benchmarks); whether the live interface is still served is unread\"}", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].lifecycle", "modality": "documented", "date": "2026-11-01" }, "TLL-16": { "kind": "archive", "quote": "[\"portraits of The Pile and The Stack\", \"a sketch-based implementation costing 3% of the dataset size in overhead\", \"a released live interface (unread)\", \"a call on dataset and model creators to release Data Portraits\"]", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].holdings", "modality": "documented", "date": "2026-11-01" }, "TLL-31": { "kind": "archive", "quote": "[\"the live interface (linked as 'this https URL' and unsaved): what it lets a user query, and whether it is still served\", \"the paper's full text beyond the abstract: the sketch's error properties, what a query returns, and any attribution or metadata fields\"]", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].unchecked", "modality": "documented", "date": "2026-11-01" }, "TLL-03": { "kind": "archive", "quote": "https://arxiv.org/abs/2303.03919", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].url", "modality": "documented", "date": "2026-11-01" }, "F52": { "kind": "field", "quote": "error code: 502", "source": "B12", "source_label": "dataportraits.org β€” the Pile, Stack and Stack V2 sketch front ends (HTTP 502)", "locus": "pile.dataportraits.org; stack.dataportraits.org; stack-v2.dataportraits.org", "modality": "documented" }, "TLL-29": { "kind": "archive", "quote": "both check text against training data from outside the model's weights: Data Portraits by membership queries on a recorded corpus, OLMoTrace by verbatim span matching against the full corpus", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].relations[1].note", "modality": "interpretation", "date": "2026-11-01" }, "TLL-06": { "kind": "archive", "quote": "[\"finding\", \"corpora\"]", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].areas", "modality": "stipulation", "date": "2026-11-01" }, "TLL-07": { "kind": "archive", "quote": "[\"research paper\", \"dataset documentation\"]", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].evidence_kind", "modality": "stipulation", "date": "2026-11-01" }, "TLL-22": { "kind": "archive", "quote": "infrastructure", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json rows[slug == 'data-portraits'].host_relation", "modality": "stipulation", "date": "2026-11-01" }, "TLL-34": { "kind": "archive", "quote": "Inspection: tools and methods that test whether a work is in a training set or a model, or trace an output back to training text", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json areas.finding", "modality": "stipulation", "date": "2026-11-01" }, "TLL-32": { "kind": "archive", "quote": "How is literature addressed to a receiving apparatus, what does that apparatus make from it, and how can transmission, transformation and provenance be inspected across later reception?", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json organizing_question", "modality": "stipulation", "date": "2026-11-01" }, "TLL-33": { "kind": "archive", "quote": "A row is an aggregator: a resource that gathers and holds many other works or records and says what it holds (operator, 2026-11-01).", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json admission_rule", "modality": "stipulation", "date": "2026-11-01" }, "TLL-35": { "kind": "archive", "quote": "Who Is Writing for the Machines? Training-Layer Literature: A Field Survey … Generative Uptake: Training-Layer Literature Extended from Machine Reading to Machine Composition (EA-TLL-COMPOSITION-01)", "deposit": null, "source_label": "traininglayerliterature.org layer census record: Data Portraits (https://www.traininglayerliterature.org/aggregators/data-portraits), editor Lee Sharks, compiled 2026-11-01, schema 1.1", "record": "https://www.traininglayerliterature.org/aggregators/data-portraits", "locus": "census.json anchor", "modality": "stipulation", "date": "2026-11-01" }, "F44": { "kind": "field", "quote": "Book β€” Non-fiction. By the W. E. B. Du Bois Center at University of Massachusetts Amherst. 2018. 144 pages. … W. E . B. Du Bois’s Data Portraits is the first complete publication of W. E. B. Du Bois’s groundbreaking charts, graphs, and maps presented at the 1900 Paris Exposition.", "source": "B3", "source_label": "Zinn Education Project β€” W. E. B. Du Bois's Data Portraits: Visualizing Black America - Zinn Education Project", "locus": "page text", "modality": "documented" }, "F46": { "kind": "field", "quote": "Sociologist W.E.B. Du Bois created sixty hand-drawn \"data portraits\" between 1897 and 1990 with Atlanta University students and scholars, visualizing Black Amer...", "source": "B5", "source_label": "SparkMap β€” W.E.B. Du Bois & Data Visualization History - SparkMap", "locus": "card snippet (unreachable)", "modality": "documented" }, "F47": { "kind": "field", "quote": "W. E. B. Du Bois's Data Portraits: Visualizing Black America", "source": "B6", "source_label": "YouTubeΒ·University of Massachusetts Amherst Libraries β€” W. E. B. Du Bois's Data Portraits: Visualizing Black America", "locus": "video card title (5m; University of Massachusetts Amherst Libraries)", "modality": "documented" }, "F49": { "kind": "field", "quote": "June 30, 2014 … digital media scholar Judith Donath defines data portraits as β€œdepictions of people made by visualizing data by and about them.”", "source": "B8", "source_label": "ARTnews.com β€” 10 Painstakingly Crafted Data Portraits That Reveal Strange Details of People's ...", "locus": "article, ΒΆ2; date", "modality": "attributed" }, "F48": { "kind": "field", "quote": "Data portraits depict their subjects' accumulated data rather than their faces. They can be visualizations of discussion contributions, browsing histories, soci...", "source": "B7", "source_label": "MIT Media Lab β€” Data Portraits - MIT Media Lab", "locus": "card snippet (unreachable)", "modality": "documented" }, "F43": { "kind": "field", "quote": "For the Target space at TED 2017 in Vancouver, we designed and produced unique visual Data Portraits of the attendees: visual designs based on answers to simple but evocative questions that we turned into wearable buttons to spark conversations and connections for the five days of the event.", "source": "B1", "source_label": "giorgialupi β€” Data Portraits at TED β€” giorgialupi", "locus": "page text", "modality": "documented" }, "F45": { "kind": "field", "quote": "For the Target-sponsored social space at TED 2017 in Vancouver, Giorgia Lupi designed and produced unique visual Data Portraits for attendees to spark conversations and connections among them during the five-day conference. … The team generated more than 500 data portraits during the five-day conference.", "source": "B4", "source_label": "Pentagram β€” TED Data Portraits - Pentagram", "locus": "page text", "modality": "documented" }, "F50": { "kind": "field", "quote": "Error: 404, Not Found", "source": "B11", "source_label": "Wikipedia β€” no article 'Data Portraits' (HTTP 404); search for \"Data Portraits\"", "locus": "index.php?title=Data_Portraits&action=raw (and Data_portraits, Data_portrait, the paper's title)", "modality": "documented" }, "F51": { "kind": "field", "quote": "\"totalhits\":7", "source": "B11", "source_label": "Wikipedia β€” no article 'Data Portraits' (HTTP 404); search for \"Data Portraits\"", "locus": "api.php search for \"Data Portraits\"", "modality": "documented" }, "CAP-1-09": { "kind": "archive", "quote": "the census name; NEW address.", "deposit": null, "source_label": "Capture Registry (EA-WG-CAPTURES-01), entry data-portraits-aio-20261101 (OBS-103202920b0d, AXN:0BAD.CAPTURE.πŸŒ‡πŸ•˜πŸ“Žβ˜ΏπŸ”½πŸš€)", "record": "https://www.alexanarch.org/captures/data-portraits-aio-20261101/", "locus": "data-portraits-aio-20261101, q_kind", "modality": "documented", "date": "2026-11-01" }, "CAP-1-07": { "kind": "archive", "quote": "'DATA PORTRAITS' ON THE OVERVIEW β€” THE TRAINING-DATA RECORD THE LAST OF FOUR SENSES, WITH NO AUTHOR, CORPUS OR QUESTION; DU BOIS, DONATH AND LUPI COMPOSED AROUND IT; A PAGE UNDER ITS NAME INDEXED", "deposit": null, "source_label": "Capture Registry (EA-WG-CAPTURES-01), entry data-portraits-aio-20261101 (OBS-103202920b0d, AXN:0BAD.CAPTURE.πŸŒ‡πŸ•˜πŸ“Žβ˜ΏπŸ”½πŸš€)", "record": "https://www.alexanarch.org/captures/data-portraits-aio-20261101/", "locus": "data-portraits-aio-20261101, mt", "modality": "interpretation", "date": "2026-11-01" }, "CAP-1-05": { "kind": "archive", "quote": "The innocent reading first. Given a two-word name that is also a term of visualization and portraiture, the Overview defines the term, lists four senses of it and asks which the user means; the paper's artifact is the fourth.", "deposit": null, "source_label": "Capture Registry (EA-WG-CAPTURES-01), entry data-portraits-aio-20261101 (OBS-103202920b0d, AXN:0BAD.CAPTURE.πŸŒ‡πŸ•˜πŸ“Žβ˜ΏπŸ”½πŸš€)", "record": "https://www.alexanarch.org/captures/data-portraits-aio-20261101/", "locus": "data-portraits-aio-20261101, reading", "modality": "interpretation", "date": "2026-11-01" }, "CAP-1-01": { "kind": "archive", "quote": "THE RECORD ONE SENSE OF FOUR. The paper's artifact holds the last bullet and one card of eight. The rest of the field is Du Bois's work for the Paris Exposition ('over 60 hand-drawn charts, graphs, and maps'), Donath's definition, carded from the MIT Media Lab ('Data portraits depict their subjects' accumulated data rather than their faces'), and Lupi's 'wearable buttons at TED 2017', carded twice (giorgialupi, Pentagram).", "deposit": null, "source_label": "Capture Registry (EA-WG-CAPTURES-01), entry data-portraits-aio-20261101 (OBS-103202920b0d, AXN:0BAD.CAPTURE.πŸŒ‡πŸ•˜πŸ“Žβ˜ΏπŸ”½πŸš€)", "record": "https://www.alexanarch.org/captures/data-portraits-aio-20261101/", "locus": "data-portraits-aio-20261101, findings[0]", "modality": "interpretation", "date": "2026-11-01" }, "CAP-1-02": { "kind": "archive", "quote": "THE ARTIFACT AT A COARSE GRAIN. 'lightweight inspection tools ... that record dataset training traces.' The abstract: 'artifacts that record training data and allow for downstream inspection' and 'Our tool is lightweight and fast, costing only 3% of the dataset size in overhead.' The recording and the lightness are carried; 'training traces' is the composition's phrase. The question the artifact answers ('has the model already encountered a given example during training?'), its method ('a solution based on data sketching'), the two corpora it documents ('The Pile' and 'The Stack') and its uses ('test set leakage and model plagiarism') are absent.", "deposit": null, "source_label": "Capture Registry (EA-WG-CAPTURES-01), entry data-portraits-aio-20261101 (OBS-103202920b0d, AXN:0BAD.CAPTURE.πŸŒ‡πŸ•˜πŸ“Žβ˜ΏπŸ”½πŸš€)", "record": "https://www.alexanarch.org/captures/data-portraits-aio-20261101/", "locus": "data-portraits-aio-20261101, findings[1]", "modality": "interpretation", "date": "2026-11-01" }, "CAP-1-03": { "kind": "archive", "quote": "NO AUTHOR, NO VENUE. 'Contemporary research groups use the term': the abstract's authors, 'Marc Marone , Benjamin Van Durme', and its venue, 'NeurIPS 2023 Datasets and Benchmarks', are not named, and the sense is undated. Each of the other three senses carries a person ('Sociologist W. E. B. Du Bois', 'Digital media scholar Judith Donath', 'Designers like Giorgia Lupi').", "deposit": null, "source_label": "Capture Registry (EA-WG-CAPTURES-01), entry data-portraits-aio-20261101 (OBS-103202920b0d, AXN:0BAD.CAPTURE.πŸŒ‡πŸ•˜πŸ“Žβ˜ΏπŸ”½πŸš€)", "record": "https://www.alexanarch.org/captures/data-portraits-aio-20261101/", "locus": "data-portraits-aio-20261101, findings[2]", "modality": "interpretation", "date": "2026-11-01" }, "CAP-1-06": { "kind": "archive", "quote": "A disambiguation in which the training-data record is the last of four senses, named by the term alone with no author, venue, corpus or question. The membership query the census row reads it for, whether a work is in a recorded corpus, is not carried.", "deposit": null, "source_label": "Capture Registry (EA-WG-CAPTURES-01), entry data-portraits-aio-20261101 (OBS-103202920b0d, AXN:0BAD.CAPTURE.πŸŒ‡πŸ•˜πŸ“Žβ˜ΏπŸ”½πŸš€)", "record": "https://www.alexanarch.org/captures/data-portraits-aio-20261101/", "locus": "data-portraits-aio-20261101, analysis", "modality": "interpretation", "date": "2026-11-01" }, "CAP-1-08": { "kind": "archive", "quote": "Neither author of the paper nor any institution named for the artifact; a page under its name carded once among seven sources of the other senses.", "deposit": null, "source_label": "Capture Registry (EA-WG-CAPTURES-01), entry data-portraits-aio-20261101 (OBS-103202920b0d, AXN:0BAD.CAPTURE.πŸŒ‡πŸ•˜πŸ“Žβ˜ΏπŸ”½πŸš€)", "record": "https://www.alexanarch.org/captures/data-portraits-aio-20261101/", "locus": "data-portraits-aio-20261101, per_note", "modality": "interpretation", "date": "2026-11-01" }, "CAP-1-10": { "kind": "archive", "quote": "reached: Contemporary research groups use the term for lightweight inspection tools; The paper's artifact reached as the last of four senses of the name, beside Du Bois's charts, Donath's portraits and Lupi's TED project; one card of eight is its page.", "deposit": null, "source_label": "Capture Registry (EA-WG-CAPTURES-01), entry data-portraits-aio-20261101 (OBS-103202920b0d, AXN:0BAD.CAPTURE.πŸŒ‡πŸ•˜πŸ“Žβ˜ΏπŸ”½πŸš€)", "record": "https://www.alexanarch.org/captures/data-portraits-aio-20261101/", "locus": "data-portraits-aio-20261101, per_dim.basis", "modality": "interpretation", "date": "2026-11-01" }, "CAP-1-11": { "kind": "archive", "quote": "{\"M\": 0.444, \"C\": 1.0, \"D\": 1.0, \"PER\": 0.688}", "deposit": null, "source_label": "Capture Registry (EA-WG-CAPTURES-01), entry data-portraits-aio-20261101 (OBS-103202920b0d, AXN:0BAD.CAPTURE.πŸŒ‡πŸ•˜πŸ“Žβ˜ΏπŸ”½πŸš€)", "record": "https://www.alexanarch.org/captures/data-portraits-aio-20261101/", "locus": "data-portraits-aio-20261101, per_dim.scores", "modality": "documented", "date": "2026-11-01" }, "CAP-1-04": { "kind": "archive", "quote": "A PAGE UNDER THE NAME INDEXED. The second card, titled 'Data Portraits' under the site name 'Data Portraits', carries the abstract's opening ('Foundation models are trained on increasingly immense and opaque datasets. Even while these models are now key in AI system building, it can be difficult to ans...'), and the composition's one link on the artifact goes to it. The census row records the 'live interface' as released and unread, and its lifecycle leaves open 'whether the live interface is still served'; on 2026-11-01 a page under the name is in the index, at an address the paste does not show.", "deposit": null, "source_label": "Capture Registry (EA-WG-CAPTURES-01), entry data-portraits-aio-20261101 (OBS-103202920b0d, AXN:0BAD.CAPTURE.πŸŒ‡πŸ•˜πŸ“Žβ˜ΏπŸ”½πŸš€)", "record": "https://www.alexanarch.org/captures/data-portraits-aio-20261101/", "locus": "data-portraits-aio-20261101, findings[3]", "modality": "interpretation", "date": "2026-11-01" } }, "note": "The knowledge object (KO): the entry as public knowledge, each sentence sourced by the claims it lists. The compression, the field-only arm, the transcript, the delta and the kernel are in the row (links.row), which is the authority." }, "source": "operator, 2026-11-01 09:56 EDT, with the 18 captures of the traininglayerliterature.org layer census names: 'captures, then /nons, then single pages'", "stage": "composed 2026-11-01 under EA-NEGONT-02 v0.10, #1726, from the traininglayerliterature.org layer census (CENSUS data-portraits; anchors #186, #1543); equal case, ρ = extends. Draft, not frozen; ledgers unaudited, one extractor; checked by the session.", "date": "2026-11-01", "status": { "procedure": "EA-NEGONT-02 v0.10 (#1726): equal case, ρ = extends, read from the traininglayerliterature.org layer census record, the capture record and the deposits that bear on the entity's classes (#186 and #1543, the census's anchors; #127, #1075, #837, #623, #939, #727, #264); archive selected by D/R/O (D and R empty; 360 O-class candidates read by filter and sample, 9 deposits admitted by reading; hop of 28 read, none admitted), with the census record (35 claims) and the Capture Registry's findings, reading, analysis, title, PER note, query note, reached-kind basis and scores (11 claims) ledgered as the archive's records of the entity; the field in two pools (B_D: 8 cards, 1 of the entity and 7 of three homonyms, 2 unreachable and 1 video card; B_R: the abstract, the paper, the sketch front ends (HTTP 502) and the encyclopedia's record of absence); the composer read the ledgers, never the transcript, except where the brief asks for T's claims and the delta", "ledger": "unaudited (Β§3.8: one extractor per ledger); field quotes 52/52 verified by script against saved text, the PDF's text, the search JSON or the card text; archive quotes 65/65 verified against the registry, census.json and the deposit texts", "frozen": false, "breaches": [] }, "addresses": [ { "address": "Data Portraits", "address_url": "https://www.alexanarch.org/addresses/data-portraits/", "register": [] } ], "counts": { "ko_sentences": 40, "p_words": 305, "field_claims": 52, "archive_claims": 65, "ko_field_claims_cited": 52, "ko_archive_claims_cited": 65, "admitted_deposits": 9, "register_addresses": 0, "traversal": { "D_sentences": 0, "D_deposits": 0, "R_sentences": 0, "R_deposits": 0, "O_direct": 0, "O_class": 360, "O_deposits": 232, "H_candidates": 28 } }, "links": { "record": "https://www.alexanarch.org/non/data-portraits/", "table_of_contents": "https://www.alexanarch.org/non/index.json", "register": "https://www.alexanarch.org/datasets/negative-of-the-negative/v2/register.json", "row": "https://www.alexanarch.org/datasets/negative-of-the-negative/v2/rows/data-portraits.json", "archive_ledger": "https://www.alexanarch.org/datasets/negative-of-the-negative/v2/ledgers/data-portraits/ledger-archive.json", "field_ledger": "https://www.alexanarch.org/datasets/negative-of-the-negative/v2/ledgers/data-portraits/field.json", "selection": "https://www.alexanarch.org/datasets/negative-of-the-negative/v2/traversal/data-portraits/", "traversal": "https://www.alexanarch.org/datasets/negative-of-the-negative/v2/traversal/data-portraits/", "traversal_summary": "https://www.alexanarch.org/datasets/negative-of-the-negative/v2/traversal/data-portraits/summary.json", "reading": "https://www.alexanarch.org/datasets/negative-of-the-negative/v2/traversal/data-portraits/reading.json", "configuration": "https://www.alexanarch.org/datasets/negative-of-the-negative/v2/panel/configs/data-portraits.json", "address_page": "https://www.alexanarch.org/addresses/data-portraits/", "dataset": "https://huggingface.co/datasets/leesharks/negative-of-the-negative" }, "row_sha256": "6373dc20091ccd409211a9cf07451877c8c3356ff068e5a2921ad30856568a96", "record_url": "https://www.alexanarch.org/non/data-portraits/", "json_url": "https://www.alexanarch.org/non/data-portraits/entity.json", "spec": "https://www.alexanarch.org/s/records/1726/", "creator": "Lee Sharks", "orcid": "0009-0000-1599-0703", "license": "CC BY 4.0" }