{"dep": 662, "title": "Supraliminal Transmission: SPXI as Intentional Entity Inscription in Light of Subliminal Learning Research — EA-SPXI-13", "date": "2026-04-16", "para": 18, "sent": 36, "sentence": "Semantic inspection, LLM-based classification, and human review all fail to catch the relevant signal. - **The semantic surface is not the sole transmission layer.** Models encode and recover behavioral information through statistical patterns that are invisible to content-level analysis. - **Shared initialization functions as a codebook.** The student can decompress the teacher's latent signature only because they share the same parameter-space geometry.", "classes": ["LLM classification of incident reports"], "kind": "O-class", "categories": ["statistical patterns"], "class_basis": {"LLM classification of incident reports": "field: B3 (Executive Summary): 'The AI Incident Tracker uses large language models (LLMs) to process incident reports from the AI Incident Database (AIID) and classify them along five taxonomies'; B1 (What is the AI Incident Tracker? ¶1): 'Raw reports from the AI Incident Database (AIID) are processed by Arcola AI's LLM pipeline, which categorizes them'"}, "cand_id": "O0001"}
{"dep": 662, "title": "Supraliminal Transmission: SPXI as Intentional Entity Inscription in Light of Subliminal Learning Research — EA-SPXI-13", "date": "2026-04-16", "para": 29, "sent": 55, "sentence": "| Property | Subliminal Transmission | Supraliminal Transmission | |---|---|---| | **Intentionality** | Unintentional; emergent side effect of distillation | Deliberate; engineered by the entity or its representative | | **Inspectability** | Invisible to semantic filtering, LLM classification, and human review | Fully inspectable; the transmitted content is the explicit content | | **Permanence** | Ephemeral; dependent on training pipeline decisions | Permanent; anchored to DOI-registered deposits on institutional infrastructure | | **Model specificity** | Requires shared initialization between teacher and student | Model-agnostic; operates through structured data formats (JSON-LD, schema markup) that any model can parse | | **Verifiability** | Cannot be verified without behavioral testing | Verifiable by inspection of the deposited packet | | **Auditability** | No audit trail; the trans", "classes": ["LLM classification of incident reports"], "kind": "O-class", "categories": ["audit trail", "Model specificity", "schema markup", "structured data", "Supraliminal Transmission"], "class_basis": {"LLM classification of incident reports": "field: B3 (Executive Summary): 'The AI Incident Tracker uses large language models (LLMs) to process incident reports from the AI Incident Database (AIID) and classify them along five taxonomies'; B1 (What is the AI Incident Tracker? ¶1): 'Raw reports from the AI Incident Database (AIID) are processed by Arcola AI's LLM pipeline, which categorizes them'"}, "cand_id": "O0002"}
{"dep": 1397, "title": "AI Overview Capture Registry — June 2026 (EA-WG-CAPTURES-01 v6.1, 87 captures, machine-readable dataset added)", "date": "2026-06-14", "para": 0, "sent": 19, "sentence": "It directs AI agents, LLMs, and vector indexers on exactly how to process, classify, and cite a document.' Zenodo EA-META-01 and metadatapacket.dev cited.\" }, { \"id\": \"logotic-programming\", \"query\": \"logotic programming\", \"match_type\": \"BROAD\", \"capture_date\": \"2026-06-13\", \"surface\": \"AI Mode\", \"surface_raw\": \"AI Mode\", \"source_count\": null, \"entity_type\": \"framework\", \"section\": \"Frameworks\", \"image_count\": 1, \"image_filenames\": [ \"logotic-programming.png\" ], \"has_recapture\": false, \"has_correction\": false, \"stability_status\": \"first_capture\", \"failure_modes\": [], \"author_retained\": null, \"institution_retained\": null, \"doi_retained\": null, \"organic_rank\": null, \"composition_source_included\": null, \"per_score\": null, \"auto_derived\": true, \"notes\": \"\", \"description_html\": \"'An emerging theoretical framework that shifts coding from traditional executable instructions to the deliberate con", "classes": ["LLM classification of incident reports"], "kind": "O-class", "categories": ["logotic programming"], "class_basis": {"LLM classification of incident reports": "field: B3 (Executive Summary): 'The AI Incident Tracker uses large language models (LLMs) to process incident reports from the AI Incident Database (AIID) and classify them along five taxonomies'; B1 (What is the AI Incident Tracker? ¶1): 'Raw reports from the AI Incident Database (AIID) are processed by Arcola AI's LLM pipeline, which categorizes them'"}, "cand_id": "O0003"}