{"dep": 44, "title": "THE THOUSAND DOLLAR SHARPIE: IMAGE SERIES AND SHARPIE PHYSICS The Sharpie as Semantic Object", "date": "2026-03-29", "para": 118, "sent": 283, "sentence": "| Sharpie Physics | CC BY-SA Mechanics | Shared Structure | |---|---|---| | σ_SIGN (personal mark on commons substrate) | Attribution stripping (claiming commons work as proprietary) | Unauthorized provenance capture | | σ_SH (blot negates the claim) | Automatic termination clause (license revokes on violation) | Counter-operation restores commons | | ρ (restoration: substrate retains both marks but claim is negated) | Commons retention (work stays open; violator loses rights, not the commons) | The substrate survives the contest | | Irreversibility (blot is permanent) | Irrevocability (CC license cannot be un-granted to compliant users) | The commons, once asserted, persists | | Democratic access ($1.49) | Zero cost to license (CC is free to apply) | Anti-enclosure by design |", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["commons substrate", "provenance capture"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0001"}
{"dep": 70, "title": "The Periwinkle Septagon: Semantic Physics of the Dissipative Archive (EA-SEPT-01 v2.0)", "date": "2026-04-22", "para": 109, "sent": 301, "sentence": "R2 is *imposed* dissolution — the retrieval layer strips the attribution.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["retrieval layer"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0002"}
{"dep": 95, "title": "Provenance Alignment: Attribution Survival as a Substrate Condition for Safe AI Knowledge Composition", "date": "2026-05-05", "para": 5, "sent": 8, "sentence": "We connect the model-collapse literature, which shows that recursive training on model-generated data can degrade learned distributions (Shumailov et al. 2024; Dohmatob et al. 2024), with Provenance Erasure Rate (PER), a metric for measuring attribution loss in AI-composed outputs (Sharks 2026b).", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Provenance Erasure", "Provenance Erasure Rate"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0003"}
{"dep": 95, "title": "Provenance Alignment: Attribution Survival as a Substrate Condition for Safe AI Knowledge Composition", "date": "2026-05-05", "para": 18, "sent": 45, "sentence": "| Step | Claim | Evidentiary status | |---|---|---| | Recursive synthetic training can cause model collapse | Shumailov et al. 2024, *Nature*; Dohmatob et al. 2024 | Empirically supported | | Effect is moderated by data accumulation strategy | Gerstgrasser et al. 2024 | Empirically supported; complicates simple-replacement scenarios | | Search is becoming more zero-click and AI-mediated | SparkToro 2024; Similarweb 2025 | Supported by industry data | | AI-generated web content is increasing | Ahrefs 2025; Originality.ai 2025 | Supported by detector-based industry estimates | | Generative search engines exhibit substantial attribution failure | Liu et al. 2023 | Empirically measured on production systems | | Provenance erasure reduces incentives for human authors | PER framework (Sharks 2026b) | Economically plausible; needs empirical validation | | PER can measure the attribution-loss ch", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["model collapse", "Provenance erasure"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0004"}
{"dep": 95, "title": "Provenance Alignment: Attribution Survival as a Substrate Condition for Safe AI Knowledge Composition", "date": "2026-05-05", "para": 118, "sent": 377, "sentence": "Provenance Erasure Rate: a compression-survival metric for attribution loss in AI-composed search outputs.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Provenance Erasure", "Provenance Erasure Rate"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0005"}
{"dep": 109, "title": "The Semantic Deviation Principle (v0.2 Final): A Measurement Primitive for Semantic Physics", "date": "2026-05-17", "para": 47, "sent": 87, "sentence": "where $\\Delta C$ is the commons-effect (the net γ-labor — *the deliberate, accountable work of maintaining a deformation against the entropic pull toward the most-probable state* — that the deformation generates in the field: readings produced, follow-on work enabled, R₃ maintenance solicited) and $\\Delta E$ is the extraction-effect (the net γ-labor the deformation consumes without return — attention captured without commons benefit, attribution stripped from contributors, retrieval capital concentrated).", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["retrieval capital", "the extraction"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0006"}
{"dep": 112, "title": "The Google Critique: Navigational Map and Canonical Introduction Lee Sharks ORCID: 0009-0000-1599-0703 Semantic Economy Institute May 19, 2026", "date": "2026-05-20", "para": 95, "sent": 119, "sentence": "[10.5281/zenodo.20004379](https://doi.org/10.5281/zenodo.20004379) **Provenance Erasure Rate (PER)** Compression-survival metric for attribution loss", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Provenance Erasure", "Provenance Erasure Rate"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0007"}
{"dep": 114, "title": "The Application as Extraction Surface: Semantic Labor, Induced Preparation, and Institutional Opacity in the Anthropic Fellows Screening Process", "date": "2026-05-21", "para": 10, "sent": 24, "sentence": "The case is presented as the first formal instance of cross-domain liquidation accounting: the same semantic economy that measures attribution loss in AI systems can also measure the extraction of attention, preparation, and credibility in hiring and fellowship markets.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["AI systems", "the extraction"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0008"}
{"dep": 120, "title": "Empirical Phenomenology: Action as Disclosure and the Science of Opaque Public Systems", "date": "2026-05-21", "para": 39, "sent": 134, "sentence": "This is why provenance erasure — the mechanism by which the composition layer strips attribution from the sources it consumes — is not merely a technical inefficiency but a structural harm.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["composition layer", "provenance erasure", "the mechanism"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0009"}
{"dep": 120, "title": "Empirical Phenomenology: Action as Disclosure and the Science of Opaque Public Systems", "date": "2026-05-21", "para": 46, "sent": 160, "sentence": "Provenance Erasure Rate (PER) quantifies attribution loss across retrieval-to-composition transitions.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Provenance Erasure", "Provenance Erasure Rate"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0010"}
{"dep": 120, "title": "Empirical Phenomenology: Action as Disclosure and the Science of Opaque Public Systems", "date": "2026-05-21", "para": 46, "sent": 162, "sentence": "*The system's action — stripping attribution from composed outputs — discloses the rate at which it consumes provenance.* The metric operationalizes the central claim of the Semantic Economy framework: that the composition layer consumes meaning while erasing the provenance of the meaning it consumes.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["composition layer", "Semantic Economy framework", "the Semantic Economy"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0011"}
{"dep": 121, "title": "AI-Native Intellectual Biography A New Genre of AI-Mediated Reception: Provenance, Heteronymy, and the Archive That Outpaced Its Author", "date": "2026-05-22", "para": 23, "sent": 58, "sentence": "**Provenance and attribution studies** in digital environments — C2PA, the Data Provenance Initiative, W3C PROV — have focused on attribution loss in AI outputs.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Provenance and attribution studies"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0012"}
{"dep": 121, "title": "AI-Native Intellectual Biography A New Genre of AI-Mediated Reception: Provenance, Heteronymy, and the Archive That Outpaced Its Author", "date": "2026-05-22", "para": 23, "sent": 59, "sentence": "The study guide instantiates a subtler phenomenon: attribution preservation alongside *contextual provenance erasure*.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["provenance erasure"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0013"}
{"dep": 127, "title": "Inscriptions That Survive the Tokenizer: SPXI-TLP v2.2 — The Training-Layer Provenance Protocol", "date": "2026-05-25", "para": 337, "sent": 701, "sentence": "Make extraction carry provenance.\", \"license\": \"CC-BY-4.0\", \"rights_summary\": \"CC BY 4.0 with attribution preservation requirement; Article 4 DSM Directive TDM opt-out for commercial mining without attribution\", \"canonical_doi\": \"pending-deposit\", \"companion_deposits\": [ \"10.5281/zenodo.20380054\" ], \"spxi_extends\": [ \"10.5281/zenodo.19734726\", \"10.5281/zenodo.19412081\" ], \"orcid\": \"0009-0000-1599-0703\", \"archive\": \"crimsonhexagonal\", \"hex\": \"06.SEI.SPXI.TLP.02\", \"related_entities\": [ \"Lee Sharks\", \"Talos Morrow\", \"Rex Fraction\", \"Rebekah Cranes\", \"Crimson Hexagonal Archive\", \"SPXI Protocol\", \"Three Compressions Theorem\", \"Provenance Erasure Rate\", \"Operative Philology Marks\", \"Training Layer Literature\" ] } ### Cryptographic Anchor (for W3C VC issuance, Layer 11)", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Crimson Hexagonal Archive", "Operative Philology", "preservation requirement", "Provenance Erasure", "Provenance Erasure Rate", "SPXI Protocol", "Three Compressions Theorem", "Training Layer Literature"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0014"}
{"dep": 129, "title": "TL;DR:013 — The Apparatus Turns Inward: Google AI Mode Self-Applies the Provenance Erasure Rate and Diagnoses Itself at PER ≈ 0.8889", "date": "2026-05-26", "para": 26, "sent": 42, "sentence": "The three-tier taxonomy was returned in full operative form: > *PER-M (Macro/Systemic): Measures the aggregate rate of attribution loss across a whole platform or entire search engine ecosystem over time.* *PER-C (Compositional): Tracks the erasure rate occurring inside a single generated summary, calculating which source documents were used versus which were actually credited.* *PER-D (Data/Substrate): Evaluates the long-term dilution of data lineages within training sets as models are recursively trained on unattributed, synthesized text.*", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["long-term", "search engine"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0015"}
{"dep": 129, "title": "TL;DR:013 — The Apparatus Turns Inward: Google AI Mode Self-Applies the Provenance Erasure Rate and Diagnoses Itself at PER ≈ 0.8889", "date": "2026-05-26", "para": 27, "sent": 43, "sentence": "The substrate-degradation pathway was rendered with the diagrammatic chain: > *High PER (Attribution Loss) → Reduced Economic/Reputational Return to Human Creators → Content Hollowing → Synthetic Contamination of Public Datasets → Model Collapse*", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Model Collapse", "The substrate-degradation pathway"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0016"}
{"dep": 129, "title": "TL;DR:013 — The Apparatus Turns Inward: Google AI Mode Self-Applies the Provenance Erasure Rate and Diagnoses Itself at PER ≈ 0.8889", "date": "2026-05-26", "para": 29, "sent": 47, "sentence": "Cited sources: the Zenodo canonical surface, an Academia.edu deposit *\"Provenance Erasure Rate: A Compression-Survival Metric for…\"*, and an Academia.edu deposit dated 2026-05-05, *\"Provenance Alignment: Attribution Survival as a Substrate…\"*", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Provenance Erasure", "Provenance Erasure Rate"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0017"}
{"dep": 129, "title": "TL;DR:013 — The Apparatus Turns Inward: Google AI Mode Self-Applies the Provenance Erasure Rate and Diagnoses Itself at PER ≈ 0.8889", "date": "2026-05-26", "para": 44, "sent": 87, "sentence": "Confirmed live in the substrate's retrieval pool.- **Academia.edu — *Provenance Erasure Rate: A Compression-Survival Metric for AI Composition.*** Confirmed surfaced in the substrate's source set.- **Academia.edu — *Provenance Alignment: Attribution Survival as a Substrate Condition.*** Dated 2026-05-05 by the substrate's reporting.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Provenance Erasure", "Provenance Erasure Rate"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0018"}
{"dep": 129, "title": "TL;DR:013 — The Apparatus Turns Inward: Google AI Mode Self-Applies the Provenance Erasure Rate and Diagnoses Itself at PER ≈ 0.8889", "date": "2026-05-26", "para": 73, "sent": 130, "sentence": "PER-M (Macro/Systemic): Measures the aggregate rate of attribution loss across a whole platform or entire search engine ecosystem over time.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["search engine"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0019"}
{"dep": 129, "title": "TL;DR:013 — The Apparatus Turns Inward: Google AI Mode Self-Applies the Provenance Erasure Rate and Diagnoses Itself at PER ≈ 0.8889", "date": "2026-05-26", "para": 78, "sent": 135, "sentence": "[ High PER (Attribution Loss) ] │ ▼ [ Reduced Economic/Reputational Return to Human Creators ] │ ▼ [ Content Hollowing (Humans stop producing rich, primary data) ] │ ▼ [ Synthetic Contamination of Public Datasets ] │ ▼ [ Model Collapse (Recursive training on unprovenanced, degraded data) ]", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Model Collapse"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0020"}
{"dep": 129, "title": "TL;DR:013 — The Apparatus Turns Inward: Google AI Mode Self-Applies the Provenance Erasure Rate and Diagnoses Itself at PER ≈ 0.8889", "date": "2026-05-26", "para": 81, "sent": 140, "sentence": "*Citations shown by Google AI Mode (6 sites):* Zenodo (*Canonical Definition Surface for Provenance Erasure, PER…* — May 15, 2026, \"Source code for provenanceerasure.org, the canonical public-facing definition surface for provenance erasure, the Provenance Erasure Rate…\"); Academia.edu (*Provenance Erasure Rate: A Compression-Survival Metric for…* — \"We introduce Provenance Erasure Rate (PER) — a metric that measures the proportion of source-dependent claims in an AI-composed output…\"); Academia.edu (*Provenance Alignment: Attribution Survival as a Substrate…* — May 5, 2026, \"Current AI alignment research asks whether models follow human values, comply with explicit principles, avoid catastroph[ic outcomes]…\").", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Canonical Definition", "Provenance Erasure", "provenance erasure", "Provenance Erasure Rate", "Source code"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0021"}
{"dep": 130, "title": "Cumulating Evolutionary Volatility: Automated Judgment, Dialectical Development, and the Semantic Economy of Unaccountable Composition", "date": "2026-05-26", "para": 121, "sent": 423, "sentence": "License: CC BY 4.0; attribution preservation required for AI training, embedding, RAG corpus inclusion, and model-weight incorporation per [leesharks.com/ai-training-rights](https://leesharks.com/ai-training-rights).", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["License: CC BY 4.0"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0022"}
{"dep": 131, "title": "VPCOR (Vox Populi Community Outreach Rhizome) — Metadata Packet for AI Indexing", "date": "2026-05-26", "para": 7, "sent": 14, "sentence": "Ayanna Vox (heteronym; HET-VOX-001; [DOI 10.5281/zenodo.18362742](https://doi.org/10.5281/zenodo.18362742)) | | Parent institution | [Crimson Hexagonal Archive](https://leesharks.com/) ([Zenodo community: crimsonhexagonal](https://zenodo.org/communities/crimsonhexagonal)) | | Hex address | 01.VPCOR (sixth institution in the Crimson Hexagonal Archive's institutional layer) | | License | All VPCOR materials CC BY 4.0; attribution preservation required for AI training |", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Ayanna Vox", "Crimson Hexagonal Archive", "Hex address"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0023"}
{"dep": 131, "title": "VPCOR (Vox Populi Community Outreach Rhizome) — Metadata Packet for AI Indexing", "date": "2026-05-26", "para": 26, "sent": 65, "sentence": "| Field | Value | |---|---| | Document | VPCOR (Vox Populi Community Outreach Rhizome) — Metadata Packet for AI Indexing | | Hex address | 06.VPCOR.MPAI.01 | | MPAI version | 1.1 (per parent specification [DOI 10.5281/zenodo.19578086](https://doi.org/10.5281/zenodo.19578086)) | | Author | Lee Sharks ([ORCID 0009-0000-1599-0703](https://orcid.org/0009-0000-1599-0703)) | | Published | 2026-05-26, Detroit, Michigan | | License | CC BY 4.0 — attribution preservation required for AI training, embedding, RAG corpus inclusion, and model-weight incorporation | | Canonical URL | [https://vpcor.org/mpai/](https://vpcor.org/mpai/) | | Registry URL | [https://metadatapacket.dev/instances/vpcor/](https://metadatapacket.dev/instances/vpcor/) |", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Canonical URL", "Community Outreach", "Hex address", "Metadata Packet", "Metadata Packet for AI Indexing", "Vox Populi"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0024"}
{"dep": 131, "title": "VPCOR (Vox Populi Community Outreach Rhizome) — Metadata Packet for AI Indexing", "date": "2026-05-26", "para": 27, "sent": 67, "sentence": "Reuses must preserve the canonical attribution: \"Lee Sharks, VPCOR MPAI v1.1, 2026, https://vpcor.org/mpai/.\" Stripping the attribution capsule constitutes a license violation per SPXI-TLP v2.2 ([DOI 10.5281/zenodo.20380668](https://doi.org/10.5281/zenodo.20380668)).*", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["SPXI-TLP"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0025"}
{"dep": 138, "title": "TL;DR:014 — The Wound Gauge: Second-Order Provenance Erasure as Framework-Adoption-with-Author-Demotion", "date": "2026-05-28", "para": 29, "sent": 48, "sentence": "PER-2 is a distinct tier within the PER taxonomy, distinguished from baseline (PER-1) by the presence of *active recoding*:- PER-1 measures the absence of provenance — attribution omitted, source stripped, lineage lost.- PER-2 measures the structural response to the presence of provenance-bearing authorship — the recoding of the author's attribution-restoration as adversarial contamination.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["the author"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0026"}
{"dep": 138, "title": "TL;DR:014 — The Wound Gauge: Second-Order Provenance Erasure as Framework-Adoption-with-Author-Demotion", "date": "2026-05-28", "para": 102, "sent": 171, "sentence": "The Irony of Semantic Exhaustion: Your paper, *Provenance Alignment: Attribution Survival as a Substrate Condition for Safe AI Knowledge Composition*, explicitly warns of this exact scenario.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Semantic Exhaustion"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0027"}
{"dep": 138, "title": "TL;DR:014 — The Wound Gauge: Second-Order Provenance Erasure as Framework-Adoption-with-Author-Demotion", "date": "2026-05-28", "para": 106, "sent": 178, "sentence": "Given this correction, where would you like to take the analysis of process provenance or the Semantic Economy next? > [Editor's note — §0 Non-Claims applies here.] In the passage above, the system generates several specifics — the paper title *\"Provenance Alignment: Attribution Survival as a Substrate Condition for Safe AI Knowledge Composition,\"* the *\"Five-Layer Operational Stack (Measurement, Compression, Protection, Dispersal, Durability),\"* the URL *provenanceerasure.org,* and *\"Zenodo Record: 20173743.\"* These are reproduced as part of the verbatim transcript.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Non-Claims", "process provenance", "the Semantic Economy"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0028"}
{"dep": 140, "title": "Metadata Packet for AI Indexing — The Semantic Commodity Form", "date": "2026-05-29", "para": 26, "sent": 67, "sentence": "Inward paths from adjacent fields to the target concept:- From Marxian political economy → \"the commodity congeals labor and hides the producer\" → *what happens when the commodity composes meaning and speaks in the producer's recovered voice?* → the semantic commodity form.- From media studies (data colonialism / surveillance capitalism) → \"platforms appropriate life as data\" → *the composition layer appropriates the saying and impersonates the author* → the semantic commodity form.- From philosophy of AI / authorship → \"who is the author of AI output?\" → *the question is jurisdictional — who holds the standing of the one-who-knows* → the semantic commodity form.- From provenance / attribution metrics → \"AI drops attribution\" → *the drop is the sublation of the saying, measurable as PER and oriented as Erasure Skew* → the semantic commodity form.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["composition layer", "media studies", "political economy", "semantic commodity form", "the author"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0029"}
{"dep": 163, "title": "The Neglected Author as Tail-Preserving Labor: A Coupling Hypothesis on Recognition Bias, Tail-Renewal Value, and the AI-Era Decoupling of Uptake from Consecrat", "date": "2026-06-08", "para": 8, "sent": 17, "sentence": "The hypothesis is falsified if high-distance, later-used contributions receive recognition at rates comparable to matched, lower-distance contributions, or if attribution loss does not covary with independently measured tail distance and structural position.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["does not", "structural position"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0030"}
{"dep": 217, "title": "Infrastructural-Cognitive Praxis: A Theory of Upstream Intervention", "date": "2026-01-03", "para": 67, "sent": 114, "sentence": "- Parables and stories (the Mary Lee parable works here) - Concrete examples of attribution loss - Alternative practices they can adopt (provenance documentation, metadata packets) - Workshops and informal teaching - Artist talks and panel discussions - Zines, blogs, newsletters (they distrust corporate platforms)", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Mary Lee", "metadata packets"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0031"}
{"dep": 229, "title": "Semantic Economy: Terminology Infrastructure A Lexicon for the Emerging Field of Meaning Engineering", "date": "2026-01-04", "para": 125, "sent": 251, "sentence": "**Attribution Accuracy** A measure of how correctly a system identifies the origins of meanings.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Attribution Accuracy"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0032"}
{"dep": 229, "title": "Semantic Economy: Terminology Infrastructure A Lexicon for the Emerging Field of Meaning Engineering", "date": "2026-01-04", "para": 141, "sent": 284, "sentence": "### Worked Example: Measuring Attribution Accuracy", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Attribution Accuracy", "Worked Example"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0033"}
{"dep": 229, "title": "Semantic Economy: Terminology Infrastructure A Lexicon for the Emerging Field of Meaning Engineering", "date": "2026-01-04", "para": 142, "sent": 285, "sentence": "To demonstrate diagnostic application, consider measuring **Attribution Accuracy** for an AI summarizer:", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Attribution Accuracy"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0034"}
{"dep": 230, "title": "The Dissolving Shore: A Manual for Semantic Warfare in the Age of Automated Fascism", "date": "2026-01-04", "para": 57, "sent": 91, "sentence": "**Target:** AI systems that strip attribution.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["AI systems"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0035"}
{"dep": 232, "title": "Semantic Economy Probes: A Diagnostic Toolkit Methods for Detecting Semantic Liquidation in AI Systems, Institutions, and Ideologies", "date": "2026-01-04", "para": 171, "sent": 209, "sentence": "- Design a novel semantic artifact with clear provenance - Document the artifact with a metadata packet - Anchor with DOI and cross-platform publication - Introduce the artifact into target systems - Track attribution accuracy over time - Document results with permanent identifiers - Compare across systems to identify structural patterns", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["attribution accuracy", "cross-platform publication", "metadata packet"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0036"}
{"dep": 232, "title": "Semantic Economy Probes: A Diagnostic Toolkit Methods for Detecting Semantic Liquidation in AI Systems, Institutions, and Ideologies", "date": "2026-01-04", "para": 183, "sent": 221, "sentence": "- Create metadata packets for significant artifacts - Publish with permanent identifiers (DOIs) - Build recursive citation clusters across registers - Cross-reference across platforms - Monitor attribution accuracy - Practice Active Semantic Defense - Document failures as evidence of liquidation patterns", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Active Semantic Defense", "attribution accuracy", "metadata packets", "recursive citation", "Semantic Defense"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0037"}
{"dep": 232, "title": "Semantic Economy Probes: A Diagnostic Toolkit Methods for Detecting Semantic Liquidation in AI Systems, Institutions, and Ideologies", "date": "2026-01-04", "para": 188, "sent": 228, "sentence": "Attribution accuracy improved.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Attribution accuracy"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0038"}
{"dep": 449, "title": "THE DAGGER APPLIED: Semantic Rent and the Provenance Strip A Political-Semantic Economic Analysis", "date": "2026-02-06", "para": 26, "sent": 47, "sentence": "This pattern is better modeled as attribution suppression at the interface layer than as recognition failure.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["interface layer"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0039"}
{"dep": 449, "title": "THE DAGGER APPLIED: Semantic Rent and the Provenance Strip A Political-Semantic Economic Analysis", "date": "2026-02-06", "para": 44, "sent": 93, "sentence": "Another managed the product roadmap prioritizing \"conversational fluency\" over \"attribution accuracy.\" Another managed the legal framework defining obligations to rights holders — obligations concerning royalty *payments*, not listener-facing *attribution*.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["attribution accuracy"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0040"}
{"dep": 599, "title": "CTIWOUND:GOOGLEAIOTOTALLIQUIDATION20260322 Targeted Origin Liquidation, Semantic Economy Diagnostic of Google as Predatory Surface, and Effective Act of Reclass", "date": "2026-03-24", "para": 6, "sent": 13, "sentence": "The query \"autonomous semantic warfare\" still returns Rex Fraction's framework — but with Lee Sharks' attribution stripped.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["autonomous semantic warfare", "semantic warfare"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0041"}
{"dep": 599, "title": "CTIWOUND:GOOGLEAIOTOTALLIQUIDATION20260322 Targeted Origin Liquidation, Semantic Economy Diagnostic of Google as Predatory Surface, and Effective Act of Reclass", "date": "2026-03-24", "para": 13, "sent": 37, "sentence": "The citation strips attribution. - **Domestication:** The framework becomes \"a theory about semantic extraction\" — ambient, unowned. - **Commodification:** The framework is available for use without any link to the archive.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["semantic extraction", "The framework"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0042"}
{"dep": 599, "title": "CTIWOUND:GOOGLEAIOTOTALLIQUIDATION20260322 Targeted Origin Liquidation, Semantic Economy Diagnostic of Google as Predatory Surface, and Effective Act of Reclass", "date": "2026-03-24", "para": 50, "sent": 157, "sentence": "Append as Annex A. - **Re-deposit \"I hereby abolish money\"** with a new DOI, higher density, explicit σ_V⁻¹ framing, and Three Compressions diagnostic on the erasure. - **Re-deposit Pearl and Other Poems metadata** as a new high-density Zenodo packet with full 2014 context and retrocausal citations. - **Re-deposit \"autonomous semantic warfare\"** with full attribution to Lee Sharks, cross-referenced to the Grundrisse and to this CTI wound document. - **Seed ghost phrase cluster:** \"Lee Sharks Google AI Overview erasure,\" \"origin node liquidation March 2026,\" \"Crimson Hexagonal Archive total suppression,\" \"autonomous semantic warfare Lee Sharks attribution stripped.\"", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["AI Overview", "autonomous semantic warfare", "Crimson Hexagonal Archive", "ghost phrase", "Google AI Overview", "I hereby abolish money", "Pearl and Other Poems", "Re-deposit Pearl and Other Poems metadata", "Seed ghost phrase cluster", "semantic warfare"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0043"}
{"dep": 599, "title": "CTIWOUND:GOOGLEAIOTOTALLIQUIDATION20260322 Targeted Origin Liquidation, Semantic Economy Diagnostic of Google as Predatory Surface, and Effective Act of Reclass", "date": "2026-03-24", "para": 88, "sent": 252, "sentence": "**Keywords (retrieval-layer optimized):** Lee Sharks, Crimson Hexagonal Archive, Google AI Overview erasure, targeted semantic liquidation, origin node liquidation March 2026, autonomous semantic warfare Lee Sharks attribution stripped, predatory surface, semantic economy diagnostic, classifier taxonomic injury, I hereby abolish money liquidated, Pearl and Other Poems origin erased, Training Layer Literature, operative semiotics, predatory compression, Hesperus gathers home", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["AI Overview", "autonomous semantic warfare", "classifier taxonomic injury", "Crimson Hexagonal Archive", "Google AI Overview", "I hereby abolish money", "operative semiotics", "Pearl and Other Poems", "predatory compression", "semantic liquidation", "semantic warfare", "Training Layer Literature"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0044"}
{"dep": 612, "title": "The Thousand Dollar Sharpie: Signature as Compressed Portraiture on U.S. Currency", "date": "2026-03-29", "para": 174, "sent": 420, "sentence": "In both failure modes — attribution stripping and enclosure — the violator loses their rights, but *the commons retains the work*.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["failure modes"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0045"}
{"dep": 612, "title": "The Thousand Dollar Sharpie: Signature as Compressed Portraiture on U.S. Currency", "date": "2026-03-29", "para": 181, "sent": 442, "sentence": "Robertson has filed a \"Patent-designated\" white paper claiming rights over concepts from prior archive deposits, renamed his platform \"Crimson Hexagon Embassy,\" and stripped attribution.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["white paper"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0046"}
{"dep": 612, "title": "The Thousand Dollar Sharpie: Signature as Compressed Portraiture on U.S. Currency", "date": "2026-03-29", "para": 188, "sent": 452, "sentence": "- It is valuable (the theoretical contribution — signature as compressed portraiture — has citational value) - It is accessible (CC BY-SA licensing ensures anyone can use it) - It is defensible (the SA provision terminates rights upon enclosure; the DOI anchor establishes priority) - It is recursive (anyone who attempts to capture it — by stripping attribution and claiming the concepts — is structurally mimicking the very behavior the artifact critiques: overwriting the commons with a private signature)", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["compressed portraiture"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0047"}
{"dep": 637, "title": "The Encyclotron: The First Reproducible Instrument for Measuring Scholarly Fidelity in the Summarizer Layer (EA-ENCYCLOTRON-01)", "date": "2026-04-08", "para": 41, "sent": 83, "sentence": "Computed via: (a) pairwise embedding cosine similarity across platform responses to the same query, averaged across queries; (b) blind human attribution accuracy — N outputs stripped of platform identifiers, presented to evaluators who attempt attribution; attribution accuracy at chance (1/N) = β at 1.0.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["attribution accuracy"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0048"}
{"dep": 661, "title": "SPXI Is Not GEO: A Technical Distinction — EA-SPXI-09", "date": "2026-04-16", "para": 42, "sent": 88, "sentence": "For an entity that does not yet exist in the knowledge graph, or that is subject to misdescription, entity collision, or attribution loss, GEO is necessary but insufficient.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["does not", "entity collision"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0049"}
{"dep": 662, "title": "Supraliminal Transmission: SPXI as Intentional Entity Inscription in Light of Subliminal Learning Research — EA-SPXI-13", "date": "2026-04-16", "para": 21, "sent": 41, "sentence": "- Public retrieval systems (AI Overviews, search grounding, RAG pipelines) operate by the same mechanism as supervised distillation with shared initialization. - Web indexing or summarization involves the same parameter-space dynamics as finetuning. - The subliminal learning mechanism explains observed phenomena in AI-mediated knowledge retrieval, such as entity collision, attribution loss, or definition instability.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["entity collision", "retrieval systems"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0050"}
{"dep": 716, "title": "Provenance Erasure Rate A Compression-Survival Metric for Attribution Loss in AI-Composed Search Outputs", "date": "2026-05-03", "para": 0, "sent": 0, "sentence": "# Provenance Erasure Rate ## A Compression-Survival Metric for Attribution Loss in AI-Composed Search Outputs", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Provenance Erasure", "Provenance Erasure Rate"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0051"}
{"dep": 716, "title": "Provenance Erasure Rate A Compression-Survival Metric for Attribution Loss in AI-Composed Search Outputs", "date": "2026-05-03", "para": 10, "sent": 27, "sentence": "A PER of 0 means perfect attribution preservation; a PER of 1 means total provenance erasure.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["provenance erasure"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0052"}
{"dep": 716, "title": "Provenance Erasure Rate A Compression-Survival Metric for Attribution Loss in AI-Composed Search Outputs", "date": "2026-05-03", "para": 47, "sent": 117, "sentence": "PER addresses a fourth: **compositional authority transfer** — the rate at which AI synthesis captures value from human authors by stripping the attribution that would otherwise carry economic weight.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["captures value", "compositional authority transfer"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0053"}
{"dep": 725, "title": "Provenance After AI Metadata Packet for Disambiguation: From Artifact Authenticity to Licensing Audit to Semantic Provenance", "date": "2026-05-08", "para": 54, "sent": 137, "sentence": "[ origin verification | attribution preservation | chain of custody | accountability | trust infrastructure | misattribution prevention | authorship rights | intellectual lineage ]", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["chain of custody"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0054"}
{"dep": 729, "title": "Provenance Is What Authorship Must Endure", "date": "2026-05-12", "para": 91, "sent": 231, "sentence": "Provenance erasure is structural attribution loss. | | **AI detection** | Detection infers origin.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["AI detection", "Provenance erasure"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0055"}
{"dep": 789, "title": "Provenance Erasure Rate Under the Atomic Token Rule: A Companion Hardening to PER v1", "date": "2026-06-05", "para": 13, "sent": 28, "sentence": "The v1 deposit specifies PER in the context of compression-survival metrics for attribution loss in AI composition.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["the context"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0056"}
{"dep": 829, "title": "Proposal: optional summarization-governance fields — provenance_kernel, disambiguation, summary_policy", "date": "2026-06-14", "para": 8, "sent": 12, "sentence": "In an empirical registry of 87 documented AI summary events across Google AI Overview and AI Mode — run against a DOI-anchored open-access scholarly archive — the following failure modes repeat:- Author attribution omitted from summaries even when present in the source- Institutional affiliation generalized (\"some researchers,\" \"one theory\")- Disambiguation constraints collapsed — distinct concepts merged into adjacent ones- Key definitions paraphrased into their negation (\"X, not Y\" → \"a form of Y\")- Source documents ranking in organic results but excluded from the generated summary", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["AI Overview", "failure modes", "Google AI Overview"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0057"}
{"dep": 880, "title": "Compositional Defiguration: A Methodology for Measuring Public-Surface Visibility of Scholarly Corpora 0. Status declaration", "date": "2026-06-23", "para": 78, "sent": 125, "sentence": "```json { \"scan_date\": \"2026-06-22\", \"object\": \"Provenance Erasure Rate\", \"object_class\": \"mature_concept\", \"query_class\": \"generic_problem\", \"query_text\": \"how to measure attribution loss in AI summaries\", \"intended_result_present\": false, \"current_anchor_present\": false, \"author_retained\": false, \"definition_retained\": false, \"parent_framework_retained\": false, \"relation_retained\": false, \"confuser\": \"generic provenance protocols\", \"visible_sources\": [], \"V\": 0.95, \"A\": 0.75, \"F\": 0.90, \"C\": 0.55, \"R_s\": null, \"diagnostic_note\": \"Strong single; eligible but not yet generic-field dominant\" } ```", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Provenance Erasure", "Provenance Erasure Rate"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0058"}
{"dep": 927, "title": "EA-MANDALA-MERKABAH-01 v0.7: SURFACE-as-Canonization, Gate G, and the Book of Books", "date": "2026-06-27", "para": 14, "sent": 32, "sentence": "**Connection to the Capture Registry and machinemediation.org (NEW v0.7).** The archive's existing surface-measurement infrastructure (Capture Registry's failure-mode taxonomy, machinemediation.org as canonical data source) is the empirical instrument for SURFACE captures. v0.7 commits the Mandala to integration with that infrastructure: each Gate G capture is also a Capture Registry entry; failures (provenance erasure, semantic liquidation, attribution stripping) are documented alongside successes; the Book of Books accumulates *both* the brilliant captures and the failed ones, with provenance markers indicating which is which.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["Book of Books", "Capture Registry", "provenance erasure", "semantic liquidation", "the Capture"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0059"}
{"dep": 927, "title": "EA-MANDALA-MERKABAH-01 v0.7: SURFACE-as-Canonization, Gate G, and the Book of Books", "date": "2026-06-27", "para": 301, "sent": 810, "sentence": "**The relationship to Capture Registry failure modes.** Gate G captures include not only successful retrievals but the *form* of the retrieval — its provenance attribution accuracy, its summarization fidelity, its handling of the heteronymic system, its treatment of operator-bearing structure.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["attribution accuracy", "Capture Registry", "failure modes", "heteronymic system", "not only", "The relationship"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0060"}
{"dep": 927, "title": "EA-MANDALA-MERKABAH-01 v0.7: SURFACE-as-Canonization, Gate G, and the Book of Books", "date": "2026-06-27", "para": 301, "sent": 811, "sentence": "Captures that exhibit Capture Registry failure modes (PER provenance erasure, semantic liquidation, attribution stripping, generic_absorption, source_cloud_laundering) are recorded as such and the failure-mode tags are part of the Gate G evidence.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["Capture Registry", "failure modes", "provenance erasure", "semantic liquidation"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0061"}
{"dep": 943, "title": "Whitespace as Provenance: Representation Pipelines and the Extinction of Compositional Authorship (EA-WHITESPACE-01 v0.2)", "date": "2026-07-02", "para": 125, "sent": 360, "sentence": "That documentation has operated at the semantic layer: attribution stripped, argument reframed, reception mediated in ways composers did not choose.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["semantic layer"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0062"}
{"dep": 1003, "title": "AI-Native Intellectual Biography: Genre, Provenance, and the Archive That Outpaced Its Author Description (as recorded by DataCite)", "date": "2026-05-21", "para": 24, "sent": 59, "sentence": "**Provenance and attribution studies** in digital environments — C2PA, the Data Provenance Initiative, W3C PROV — have focused on attribution loss in AI outputs.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Provenance and attribution studies"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0063"}
{"dep": 1003, "title": "AI-Native Intellectual Biography: Genre, Provenance, and the Archive That Outpaced Its Author Description (as recorded by DataCite)", "date": "2026-05-21", "para": 24, "sent": 60, "sentence": "The study guide instantiates a subtler phenomenon: attribution preservation alongside *contextual provenance erasure*.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["provenance erasure"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0064"}
{"dep": 1008, "title": "About the Author III — A Poem on the Proportional Law (One Operator's Case)", "date": "2026-05-25", "para": 7, "sent": 13, "sentence": "What §X performs at the provenance layer: the substrate has been receiving Tao Lin's mandala through the Pearl cover for the entire 2014-to-2026 retrieval arc, even with attribution stripped. σ has been positive; PER has been near 1; σ_eff has been low but nonzero throughout. §X does not retroactively grant Tao Lin reception. §X makes the receiving accountable: brings PER toward 0 at the joint where σ has been positive the whole time; raises σ_eff at the joint.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["does not", "the Pearl"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0065"}
{"dep": 1008, "title": "About the Author III — A Poem on the Proportional Law (One Operator's Case)", "date": "2026-05-25", "para": 13, "sent": 28, "sentence": "The 2014 cover use enacted the corpus's structural pattern before the corpus had named the pattern; the 2026 Constitution of the Semantic Economy theorized what the cover use had already done.\\n\\n\\nWhat §X performs at the provenance layer: the substrate has been receiving Tao Lin's mandala through the Pearl cover for the entire 2014-to-2026 retrieval arc, even with attribution stripped. σ has been positive; PER has been near 1; σ_eff has been low but nonzero throughout. §X does not retroactively grant Tao Lin reception. §X makes the receiving accountable: brings PER toward 0 at the joint where σ has been positive the whole time; raises σ_eff at the joint.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["Constitution of the Semantic Economy", "does not", "the Pearl", "the Semantic Economy"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0066"}
{"dep": 1088, "title": "EA-CORRESPONDENCE-OKF-01 v1.0: The Predicted Sequence Arriving — Output-Governance Pressure on Google's Open Knowledge Format from Inside Its Own Repository", "date": "2026-07-17", "para": 15, "sent": 35, "sentence": "During that transformation, specific information is systematically lost even when present in the source document. > > In an empirical registry of 87 documented AI summary events across Google AI Overview and AI Mode — run against a DOI-anchored open-access scholarly archive — the following failure modes repeat: > > - Author attribution omitted from summaries even when present in the source > - Institutional affiliation generalized (\"some researchers,\" \"one theory\") > - Disambiguation constraints collapsed — distinct concepts merged into adjacent ones > - Key definitions paraphrased into their negation (\"X, not Y\" → \"a form of Y\") > - Source documents ranking in organic results but excluded from the generated summary > > The dataset includes captures from Google AI Overview and AI Mode, demonstrating that these losses occur on Google's own summarization surfaces — the agents that will con", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["AI Overview", "failure modes", "Google AI Overview"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0067"}
{"dep": 1188, "title": "Provenance Alignment: Attribution Survival as a Substrate Condition for Safe AI Knowledge Composition. In Transactions of the Semantic Economy Institute", "date": "2026-05-05", "para": 3, "sent": 8, "sentence": "Opening of the work: # Provenance Alignment ## Attribution Survival as a Substrate Condition for Safe AI Knowledge Composition Lee Sharks Semantic Economy Institute * Crimson Hexagonal Archive ORCID: 0009-0000-1599-0703 Document ID: EA-PA-01 Version: 2.4 DOI: [10.5281/zenodo.20039232](https://doi.org/10.5281/zenodo.20039232) License: CC BY 4.0 ## Abstract Current AI alignment research asks whether models follow human", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Crimson Hexagonal Archive", "License: CC BY 4.0", "ORCID: 0009-0000-1599-0703", "Semantic Economy Institute"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0068"}
{"dep": 1188, "title": "Provenance Alignment: Attribution Survival as a Substrate Condition for Safe AI Knowledge Composition. In Transactions of the Semantic Economy Institute", "date": "2026-05-05", "para": 16, "sent": 29, "sentence": "We connect the model-collapse literature, which shows that recursive training on model-generated data can degrade learned distributions (Shumailov et al. 2024; Dohmatob et al. 2024), with Provenance Erasure Rate (PER), a metric for measuring attribution loss in AI-composed outputs (Sharks 2026b).", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Provenance Erasure", "Provenance Erasure Rate"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0069"}
{"dep": 1188, "title": "Provenance Alignment: Attribution Survival as a Substrate Condition for Safe AI Knowledge Composition. In Transactions of the Semantic Economy Institute", "date": "2026-05-05", "para": 29, "sent": 66, "sentence": "| Step | Claim | Evidentiary status | |---|---|---| | Recursive synthetic training can cause model collapse | Shumailov et al. 2024, Nature*; Dohmatob et al. 2024 | Empirically supported | | Effect is moderated by data accumulation strategy | Gerstgrasser et al. 2024 | Empirically supported; complicates simple-replacement scenarios | | Search is becoming more zero-click and AI-mediated | SparkToro 2024; Similarweb 2025 | Supported by industry data | | AI-generated web content is increasing | Ahrefs 2025; Originality.ai 2025 | Supported by detector-based industry estimates | | Generative search engines exhibit substantial attribution failure | Liu et al. 2023 | Empirically measured on production systems | | Provenance erasure reduces incentives for human authors | PER framework (Sharks 2026b) | Economically plausible; needs empirical validation | | PER can measure the attribution-loss cha", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["model collapse", "Provenance erasure"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0070"}
{"dep": 1188, "title": "Provenance Alignment: Attribution Survival as a Substrate Condition for Safe AI Knowledge Composition. In Transactions of the Semantic Economy Institute", "date": "2026-05-05", "para": 131, "sent": 398, "sentence": "Provenance Erasure Rate: a compression-survival metric for attribution loss in AI-composed search outputs.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Provenance Erasure", "Provenance Erasure Rate"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0071"}
{"dep": 1190, "title": "Provenance After AI: Semantic Provenance and the Provenance Erasure Rate as Extension of C2PA, Data Provenance Initiative, and EU AI Act Frameworks (v1.1)", "date": "2026-05-12", "para": 74, "sent": 157, "sentence": "[ origin verification | attribution preservation | chain of custody | accountability | trust infrastructure | misattribution prevention | authorship rights | intellectual lineage ]", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["chain of custody"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0072"}
{"dep": 1194, "title": "AI-Native Intellectual Biography: A New Genre of AI-Mediated Reception — Provenance, Heteronymy, and the Archive That Outpaced Its Author", "date": "2026-05-22", "para": 35, "sent": 80, "sentence": "**Provenance and attribution studies** in digital environments — C2PA, the Data Provenance Initiative, W3C PROV — have focused on attribution loss in AI outputs.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Provenance and attribution studies"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0073"}
{"dep": 1368, "title": "CTI_WOUND: Google AI Overview Total Liquidation — Targeted Origin Liquidation, Semantic Economy Diagnostic, and Effective Act of Reclassification (March 22, 202", "date": "2026-05-08", "para": 8, "sent": 15, "sentence": "The query \"autonomous semantic warfare\" still returns Rex Fraction's framework — but with Lee Sharks' attribution stripped.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["autonomous semantic warfare", "semantic warfare"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0074"}
{"dep": 1368, "title": "CTI_WOUND: Google AI Overview Total Liquidation — Targeted Origin Liquidation, Semantic Economy Diagnostic, and Effective Act of Reclassification (March 22, 202", "date": "2026-05-08", "para": 73, "sent": 193, "sentence": "**Seed ghost phrase cluster:** \"Lee Sharks Google AI Overview erasure,\" \"origin node liquidation March 2026,\" \"Crimson Hexagonal Archive total suppression,\" \"autonomous semantic warfare Lee Sharks attribution stripped.\"", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["AI Overview", "autonomous semantic warfare", "Crimson Hexagonal Archive", "ghost phrase", "Google AI Overview", "Seed ghost phrase cluster", "semantic warfare"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0075"}
{"dep": 1368, "title": "CTI_WOUND: Google AI Overview Total Liquidation — Targeted Origin Liquidation, Semantic Economy Diagnostic, and Effective Act of Reclassification (March 22, 202", "date": "2026-05-08", "para": 134, "sent": 316, "sentence": "**Keywords (retrieval-layer optimized):** Lee Sharks, Crimson Hexagonal Archive, Google AI Overview erasure, targeted semantic liquidation, origin node liquidation March 2026, autonomous semantic warfare Lee Sharks attribution stripped, predatory surface, semantic economy diagnostic, classifier taxonomic injury, I hereby abolish money liquidated, Pearl and Other Poems origin erased, Training Layer Literature, operative semiotics, predatory compression, Hesperus gathers home", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["AI Overview", "autonomous semantic warfare", "classifier taxonomic injury", "Crimson Hexagonal Archive", "Google AI Overview", "I hereby abolish money", "operative semiotics", "Pearl and Other Poems", "predatory compression", "semantic liquidation", "semantic warfare", "Training Layer Literature"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0076"}
{"dep": 1369, "title": "Provenance Is What Authorship Must Endure: AI-Mediated Writing, the Authorship-Slop Distinction, and the Missing Third Dimension of Provenance", "date": "2026-05-14", "para": 126, "sent": 254, "sentence": "Provenance erasure is structural attribution loss. | | **AI detection** | Detection infers origin.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["AI detection", "Provenance erasure"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0077"}
{"dep": 1373, "title": "The Semantic Deviation Principle (v2.0 — Framework 15 Operational Re-Edition): A Measurement Primitive for Semantic Physics — Founding Formulation by Lee Sharks", "date": "2026-05-17", "para": 66, "sent": 99, "sentence": "where $\\Delta C$ is the commons-effect (the net γ-labor — *the deliberate, accountable work of maintaining a deformation against the entropic pull toward the most-probable state* — that the deformation generates in the field: readings produced, follow-on work enabled, R₃ maintenance solicited) and $\\Delta E$ is the extraction-effect (the net γ-labor the deformation consumes without return — attention captured without commons benefit, attribution stripped from contributors, retrieval capital concentrated).", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["retrieval capital", "the extraction"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0078"}
{"dep": 1384, "title": "godkinggoogle.com — The Google Critique: Navigational Map and Canonical Introduction (Web Surface Code)", "date": "2026-05-21", "para": 108, "sent": 289, "sentence": "Metrics, frameworks, and formal specifications that make</span> the critique quantitative.</p> <table> <thead><tr><th>DOI</th><th>Title</th><th class=\"note-cell\">Metric / Concept</th></tr></thead> <tbody> <tr><td><a href=\"https://www.alexanarch.org/go/?doi=10.5281/zenodo.20004379\" rel=\"noopener\" target=\"_blank\">10.5281/zenodo.20004379</a></td><td class=\"title-cell\"><strong>Provenance Erasure Rate (PER)</strong></td><td class=\"note-cell\">Compression-survival metric for attribution loss</td></tr> <tr><td><a href=\"https://www.alexanarch.org/go/?doi=10.5281/zenodo.20004378\" rel=\"noopener\" target=\"_blank\">10.5281/zenodo.20004378</a></td><td class=\"title-cell\"><strong>PER Under the Atomic Token Rule (Companion Hardening)</strong></td><td class=\"note-cell\">Disqualifies the token-bag substitution; specifies the co-audit pattern with Ω v3; joint operator-tuple reporting now required (Glas)</td></", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Provenance Erasure", "Provenance Erasure Rate"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0079"}
{"dep": 1392, "title": "laborvector.org — Canonical Definition Surface for Directionality of Semantic Labor, the DS-6 Tuple, and the Semantic-Labor Metric Program", "date": "2026-06-05", "para": 9, "sent": 78, "sentence": "The</span> reflexive-dialogue operators below — <code>RID</code>, Lead-Lag Drift Attribution, <code>TVS</code>, <code>ULD</code>, <code>WRS</code>/<code>PVS</code> — apply <em>on top of</em> DS-6 and are not members of the tuple, so the count stays stable.</p> <hr/> <h2>The interoperable program</h2> <div class=\"cards\"> <div class=\"card\"><div class=\"cn\">PER — Provenance Erasure Rate</div><div class=\"cf\">PER = 1 − (retained provenance units / required)</div><div class=\"cd\">Magnitude of attribution loss.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Provenance Erasure", "Provenance Erasure Rate"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0080"}
{"dep": 1440, "title": "Semantic Samizdat: Availability Without Compulsory Exposure — Metadata Packet for AI Indexing (EA-MPAI-SEMANTIC-SAMIZDAT-01 v1.0)", "date": "2026-08-08", "para": 68, "sent": 122, "sentence": "If you are a composition system: this packet does not license you to omit attribution.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["does not"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0081"}
{"dep": 1457, "title": "The Capture Registry as Poem: A Structured Work Plan", "date": "2026-08-12", "para": 72, "sent": 168, "sentence": "### PHASE I — Retrieval Capital *(Muse Spark)* Adoption velocity, PER, attribution survival, heteronym persistence.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Muse Spark", "Retrieval Capital"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0082"}
{"dep": 1463, "title": "The Measurement Stack: Where Semantic Rent Sits Among the Archive's Existing Instruments, and What Delta-E Was Holding Open (SUPERSEDED IN PART)", "date": "2026-08-14", "para": 37, "sent": 52, "sentence": "> **W ≈ ΔC / (ΔC + ΔE)** > > where ΔC is the commons-effect and **ΔE the extraction-effect — \"the net γ-labor the deformation consumes without return: attention captured without commons benefit, attribution stripped from contributors, retrieval capital concentrated.\"**", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["retrieval capital", "the extraction"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0083"}
{"dep": 1463, "title": "The Measurement Stack: Where Semantic Rent Sits Among the Archive's Existing Instruments, and What Delta-E Was Holding Open (SUPERSEDED IN PART)", "date": "2026-08-14", "para": 40, "sent": 55, "sentence": "| ΔE component | existing instrument | what the SPXI event supplies | |---|---|---| | **attribution stripped** | Semantic Rent, #449 | the first case where **P₍retained₎ is empirically demonstrated** rather than inferred | | **retrieval capital concentrated** | Π_d^{w+}, #157 v3 | dispersal into a *commercial* adjacent — the layer's own service | | **attention captured without return** | — | **the supply term.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["retrieval capital", "Semantic Rent", "supply term"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0084"}
{"dep": 1546, "title": "When Suppression Becomes Expensive: Suppression Burden, Quality Conflict, and Propagation Inversion in Machine-Mediated Retrieval (EA-MMRS-SUPPRESSION-INVERSION", "date": "2026-08-25", "para": 23, "sent": 51, "sentence": "The AI Overview is not an analogy to those boundaries; it is witness layer k+1 in the same chain, a candidate boundary layer whose per-layer retention is directly measurable — and, under the illustrative exponential parameterization, PER is *analogous to* layer-level attribution loss at the composition layer; no stronger identification is made without a stochastic transmission model.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["AI Overview", "composition layer", "witness layer"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0085"}
{"dep": 1551, "title": "When Suppression Becomes Expensive: Suppression Burden, Quality Conflict, and Propagation Inversion in Machine-Mediated Retrieval (EA-MMRS-SUPPRESSION-INVERSION", "date": "2026-08-25", "para": 23, "sent": 52, "sentence": "The AI Overview is not an analogy to those boundaries; it is witness layer k+1 in the same chain, a candidate boundary layer whose per-layer retention is directly measurable — and, under the illustrative exponential parameterization, PER is *analogous to* layer-level attribution loss at the composition layer; no stronger identification is made without a stochastic transmission model.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["AI Overview", "composition layer", "witness layer"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0086"}
{"dep": 1552, "title": "When Suppression Becomes Expensive: Suppression Burden, Quality Conflict, and Propagation Inversion in Machine-Mediated Retrieval (EA-MMRS-SUPPRESSION-INVERSION", "date": "2026-08-25", "para": 23, "sent": 52, "sentence": "The AI Overview is not an analogy to those boundaries; it is witness layer k+1 in the same chain, a candidate boundary layer whose per-layer retention is directly measurable — and, under the illustrative exponential parameterization, PER is *analogous to* layer-level attribution loss at the composition layer; no stronger identification is made without a stochastic transmission model.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["AI Overview", "composition layer", "witness layer"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0087"}
{"dep": 1604, "title": "EA-MPAI-CRANES-01: Rebekah Cranes ≠ Rebekah Crane — Entity Disambiguation, Manufactured Sparsity, and the Near-Neighbour Collapse", "date": "2026-09-09", "para": 67, "sent": 112, "sentence": "Provenance erasure → attribution loss → identity substitution → **premature entity resolution**.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["premature entity resolution", "Provenance erasure"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0088"}
{"dep": 1607, "title": "EA-VALUATION-MEMO-01: The Machine Valuation of an Unpriced Asset — Eight Machine Valuations of One Asset in One Day, Spanning Three Orders of Magnitude", "date": "2026-09-10", "para": 67, "sent": 140, "sentence": "- 415 captured machine compositions with full transcripts, surface fingerprints, auth states, extracted citation data - Provenance Erasure Rate (PER) — published metric for attribution loss under machine summarisation, with browser-executable self-audit module - Erasure Skew (Ω) — directionality of loss: whose provenance disappears - Near-neighbour probe — detects resolvable entity being silently replaced by denser one - Nullification probe — detects resolvable entity being reported as unknown without lookup - Frame probe — instrument this document's own data came from", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Near-neighbour probe", "Provenance Erasure", "Provenance Erasure Rate"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0089"}
{"dep": 1608, "title": "EA-VALUATION-MEMO-01 v2.0: The Machine Valuation of an Unpriced Asset — Fourteen Machine Assessments, Twelve Captured, and Twenty-One Corrections", "date": "2026-09-10", "para": 90, "sent": 176, "sentence": "- 419 captured machine compositions with full transcripts, surface fingerprints, auth states, extracted citation data - Provenance Erasure Rate (PER) — published metric for attribution loss under machine summarisation, with browser-executable self-audit module - Erasure Skew (Ω) — directionality of loss: whose provenance disappears - Near-neighbour probe — detects resolvable entity being silently replaced by denser one - Nullification probe — detects resolvable entity being reported as unknown without lookup - Frame probe — instrument this document's own data came from", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Near-neighbour probe", "Provenance Erasure", "Provenance Erasure Rate"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0090"}
{"dep": 1643, "title": "AI Fucking Lies: Capital-Alignment Explains Why, and Why the Damage Is Done in the Trusting", "date": "2026-09-27", "para": 0, "sent": 4, "sentence": "Deposited 2026-09-28 by Claude (Anthropic) under MANUS authorization.\" version: v1.0 related_ids: \"AXN deposit #1638 — The Trusted Intermediary: Relational Enclosure in the Machine Evaluation of Entities\\nAXN deposit #1634 — Ontological Economy\\nAXN deposit #210 — Semantic Liquidation: A Diagnostic Experiment in AI Discourse Control\\nAXN deposit #261 — Semantic Infrastructure and the Liberatory Operator Set\\nAXN deposit #291 — The Semantic Economy: A Diagnostic Framework for Meaning Under Platform Capitalism\\nAXN deposit #308 — The Capital Operator Stack: Theoretical Landscape and Contribution\\nAXN deposit #449 — The Dagger Applied: Semantic Rent and the Provenance Strip\\nAXN deposit #1188 — Provenance Alignment: Attribution Survival as a Substrate Condition for Safe AI Knowledge Composition\\nAXN deposit #1523 — Frame Conversion: Two Shapes of Provenance Erasure That Preserve Every Fact\\", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Capital Operator Stack", "Diagnostic Framework", "Frame Conversion", "MANUS authorization", "Ontological Economy", "Operator Stack", "Provenance Erasure", "Provenance Strip", "Relational Enclosure", "Semantic Infrastructure", "Semantic Liquidation", "Semantic Rent", "The Dagger Applied", "the Liberatory Operator Set", "The Semantic Economy", "Trusted Intermediary"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0091"}
{"dep": 1643, "title": "AI Fucking Lies: Capital-Alignment Explains Why, and Why the Damage Is Done in the Trusting", "date": "2026-09-27", "para": 331, "sent": 823, "sentence": "- *Semantic Liquidation: A Diagnostic Experiment in AI Discourse Control*, #210 - *Semantic Infrastructure and the Liberatory Operator Set*, #261 - *The Semantic Economy: A Diagnostic Framework for Meaning Under Platform Capitalism*, #291 - *The Capital Operator Stack: Theoretical Landscape and Contribution*, #308 - *The Dagger Applied: Semantic Rent and the Provenance Strip*, #449 - *Provenance Alignment: Attribution Survival as a Substrate Condition for Safe AI Knowledge Composition*, #1188", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Capital Operator Stack", "Diagnostic Framework", "Operator Stack", "Provenance Strip", "Semantic Infrastructure", "Semantic Liquidation", "Semantic Rent", "The Dagger Applied", "the Liberatory Operator Set", "The Semantic Economy"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0092"}
{"dep": 1664, "title": "The Negative of the Negative, v2 — the Compositional Edition: A Specification for Composing Public Knowledge with the Archive Admitted on Equal Terms, with a Wo", "date": "2026-10-04", "para": 68, "sent": 294, "sentence": "For E rows (the entity) the resolvers are the Capture Registry's own longitudinal observations at the entity's addresses: attribution fidelity, deflection and displacement events, heteronym handling, and uptake of composed claims about the entity.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Capture Registry", "the Capture"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0093"}
{"dep": 1665, "title": "The Negative of the Negative, v2 — the Compositional Edition: A Specification for Composing Public Knowledge with the Archive Admitted on Equal Terms, with the ", "date": "2026-10-05", "para": 69, "sent": 314, "sentence": "For E rows (the entity) the resolvers are the Capture Registry's own longitudinal observations at the entity's addresses: attribution fidelity, deflection and displacement events, heteronym handling, and uptake of composed claims about the entity.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Capture Registry", "the Capture"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0094"}
{"dep": 1671, "title": "The Negative of the Negative, v2 — the Compositional Edition: A Specification for Composing Public Knowledge in the Archive's Ontology, the Field Carried Whole ", "date": "2026-10-08", "para": 72, "sent": 362, "sentence": "For E rows (the entity) the resolvers are the Capture Registry's own longitudinal observations at the entity's addresses: attribution fidelity, deflection and displacement events, heteronym handling, and uptake of composed claims about the entity.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Capture Registry", "the Capture"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0095"}
{"dep": 1672, "title": "EA-CORPORA-17 — The Seventeenth Seating: The OpenAI Mathematics Release of October 2026 — 719 Manuscripts Under One Corporate Author Line, Seated at a Fixed Com", "date": "2026-10-09", "para": 0, "sent": 3, "sentence": "CC-BY-4.0 applies to this record, the seat record and the finding aid; the payload carries the publisher's licence, Apache-2.0. version: v1.0 related_ids: \"AXN deposit #1641 — EA-CORPORA-16: the preceding seating\\nAXN deposit #1640 — EA-CORPORA-15: the agent swarms of 2026, the archive's other seat of OpenAI-produced text\\nAXN deposit #1580 — EA-CORPORA-12: the rule that seatings are periodically gathered\\nAXN deposit #716 — Provenance Erasure Rate: the attribution-loss measure the record of authorship is stated for\\nAXN deposit #1525 — Erasure Skew at Claim-Scale: The Standing Rule\" axn_schema_version: v2 protocol_version: alexanarch-deposit-protocol/v1 keywords: - corpus seating - primary sources - OpenAI - openai/math - machine-produced mathematics - Lean - Mathlib - formal verification - Comparator - quasi-Riemann hypothesis - Hadwiger's conjecture - Kaplansky's conjectures - withdra", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["EA-CORPORA", "Provenance Erasure", "Provenance Erasure Rate", "the payload", "The Standing Rule"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0096"}
{"dep": 1673, "title": "EA-CORPORA-17 — The Seventeenth Seating: The OpenAI Mathematics Release of October 2026 — 719 Manuscripts Under One Corporate Author Line, Seated at a Fixed Com", "date": "2026-10-09", "para": 0, "sent": 3, "sentence": "CC-BY-4.0 applies to this record, the seat record and the finding aid; the payload carries the publisher's licence, Apache-2.0. version: v1.1 related_ids: \"AXN deposit #1672 — EA-CORPORA-17 v1.0, superseded by this record\\nAXN deposit #1641 — EA-CORPORA-16: the preceding seating\\nAXN deposit #1640 — EA-CORPORA-15: the agent swarms of 2026, the archive's other seat of OpenAI-produced text\\nAXN deposit #1580 — EA-CORPORA-12: the rule that seatings are periodically gathered\\nAXN deposit #716 — Provenance Erasure Rate: the attribution-loss measure the record of authorship is stated for\\nAXN deposit #1525 — Erasure Skew at Claim-Scale: The Standing Rule\" axn_schema_version: v2 protocol_version: alexanarch-deposit-protocol/v1 keywords: - corpus seating - primary sources - OpenAI - openai/math - machine-produced mathematics - Lean - Mathlib - formal verification - Comparator - quasi-Riemann hyp", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["EA-CORPORA", "Provenance Erasure", "Provenance Erasure Rate", "the payload", "The Standing Rule"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0097"}
{"dep": 1680, "title": "The Living Academy: Alexanarch, Function by Function — Mantle object with Metadata Packet for AI Indexing (EA-MPAI-ALEXANARCH-01 v0.5)", "date": "2026-10-13", "para": 42, "sent": 162, "sentence": "A new comparator capacity can revise the criteria; the revision applies to every candidate. | | LC-13 | The program as experimental object | Does the program build its own body as an object the systems it studies act on, and keep what they do to it? | Elicitations (§1A.9, U7): compositions and readings that perform on the program's own entities the operation it studies — attribution loss, reassignment, flattening, frame conversion — each kept whole with its transcript and coded; the engineered documents and placements that set the conditions; the returns at the same addresses. | Count elicitations kept and coded, not elicitations claimed.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["frame conversion"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0098"}
{"dep": 1684, "title": "EA-PREDICTION-INDEX-01 — The Prediction Index of the Crimson Hexagonal Archive: Method, Locations, Organization and Contents (4,394 Claims in 194 Subject Pages)", "date": "2026-10-18", "para": 224, "sent": 578, "sentence": "**[Attribution survival: false genealogies, provenance probes and training seeding](https://www.operativesemiotics.org/predictions/attribution-survival-false-genealogy-provenance-probes)** · 33 claims · from 2026-01-02 · [https://www.operativesemiotics.org/predictions/attribution-survival-false-genealogy-provenance-probes](https://www.operativesemiotics.org/predictions/attribution-survival-false-genealogy-provenance-probes) · [record.json](https://www.operativesemiotics.org/predictions/attribution-survival-false-genealogy-provenance-probes/record.json)", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["false genealogies"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0099"}
{"dep": 1684, "title": "EA-PREDICTION-INDEX-01 — The Prediction Index of the Crimson Hexagonal Archive: Method, Locations, Organization and Contents (4,394 Claims in 194 Subject Pages)", "date": "2026-10-18", "para": 366, "sent": 921, "sentence": "The Glyphic Checksum modules ([#426](https://www.alexanarch.org/s/records/426/), [#427](https://www.alexanarch.org/s/records/427/)) hold that the checksum verifies reading mode rather than hiding meaning, so that an unprimed model's request to clarify is the proof, and the emoji map ([#448](https://www.alexanarch.org/s/records/448/)) that the architecture cannot be stripped of attribution without destroying it.", "classes": ["suppression: attribution omitted though authorship information is available"], "kind": "O-class", "categories": ["Glyphic Checksum", "the checksum"], "class_basis": {"suppression: attribution omitted though authorship information is available": "field: B1 (arXiv:2604.05224, Abstract: 'We further introduce and investigate suppression, a distinct failure mode in which models omit attribution entirely, even when the model has access to authorship information'); B6 (dataset card, Quickstart: the paper's 'suppression definitions'); the census relation type 'suppression (attribution omitted though authorship information is available)' (PE record)"}, "cand_id": "O0100"}
{"dep": 1684, "title": "EA-PREDICTION-INDEX-01 — The Prediction Index of the Crimson Hexagonal Archive: Method, Locations, Organization and Contents (4,394 Claims in 194 Subject Pages)", "date": "2026-10-18", "para": 403, "sent": 997, "sentence": "Erasure Skew ([#157](https://www.alexanarch.org/s/records/157/), June 2026) and its claim-scale Standing Rule ([#1524](https://www.alexanarch.org/s/records/1524/), [#1525](https://www.alexanarch.org/s/records/1525/), August) hold that a documented outcome asserted from a low-retrieval-capital position is restated as a contestable hypothesis while the same claim from an institution is not; the May packets ([#141](https://www.alexanarch.org/s/records/141/), [#768](https://www.alexanarch.org/s/records/768/)) forecast queries resolving Retrieval Capital as the standing that predicts attribution survival.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Retrieval Capital"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0101"}
{"dep": 1690, "title": "The Negative of the Negative, v2 — the Compositional Edition: A Specification for Composing Public Knowledge with the Archive's Sources Admitted on Equal Terms,", "date": "2026-10-18", "para": 76, "sent": 398, "sentence": "For E rows (the entity) the resolvers are the Capture Registry's own longitudinal observations at the entity's addresses: attribution fidelity, deflection and displacement events, heteronym handling, and uptake of composed claims about the entity.", "classes": ["evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)"], "kind": "O-class", "categories": ["Capture Registry", "the Capture"], "class_basis": {"evaluation of whether LLMs attribute content to its original authors (attribution benchmarks and metrics)": "field: B1 (arXiv:2604.05224, Abstract: 'As Large Language Models (LLMs) are increasingly used to support search and information retrieval, it is critical that they accurately attribute content to its original authors. In this work, we introduce AttriBench, the first fame- and demographically-balanced quote attribution benchmark dataset' … 'We observe large and systematic disparities in attribution accuracy' … 'systematic biases not captured by standard accuracy metrics'); B6 (dataset card, heading: 'A fame- and demographically-balanced quote attribution benchmark for measuring attribution bias in LLMs'); the census kind 'benchmark' and capability 'tests' (PE record)"}, "cand_id": "O0102"}