{ "schema_version": "1.0.0", "package_id": "W05", "title": "Generation, Source Use, Citation, and Absorption", "language": "en", "phase": "Explain", "essential_question": "Was a source merely displayed, or did it shape the answer?", "research_cutoff": "2026-08-24", "status": "structurally_complete_text_first_authored_draft_human_review_and_timed_pilot_pending", "required_artifacts": [ "README.md", "LECTURE_NOTES.md", "SLIDE_SCRIPT.md", "WORKED_CASE.md", "TRANSCRIPT.md", "manifest.json", "validate_package.py" ], "learning_outcomes": [ "Segment generated answers into material atomic claims while preserving quantities, conditions, scope, time, and qualifications.", "Separate source presence, source use, claim absorption, displayed attribution, claim-passage entailment, and answer correctness.", "Resolve citation objects to versioned source identities, passages, and provenance origins before scoring support.", "Compute citation entailment and completeness with declared units, denominators, weights, unresolved states, and eligibility rules.", "Diagnose placement, entailment, qualification, source-fit, completeness, provenance, absorption, and use-identification failures.", "Preserve independent labels and adjudicate disagreements without deleting pre-adjudication evidence.", "Design a controlled source-removal comparison and state its conditional evidence ceiling.", "Write a bounded conclusion for an open synthetic annotation fixture without inventing a closed-platform mechanism." ], "time_design": { "status": "authoring_budgets_not_observed_delivery_times", "synchronous_seminar_minutes": 90, "synchronous_core_studio_minutes": 75, "asynchronous_core_minutes_min": 190, "asynchronous_core_minutes_max": 210, "transcript_equivalent_minutes_min": 35, "transcript_equivalent_minutes_max": 45, "transcript_planned_minutes": 42, "full_lab_effort_hours": 6, "pilot_state": "not_taught_not_rehearsed_not_timed", "media_state": "no_video_or_audio_recording_exists" }, "local_dependencies": [ { "id": "COURSE-W05", "path": "course-site/app/lib/course-data.ts#W05", "role": "Authoritative week title, essential question, outcomes, prerequisites, L04 route, and reading boundary.", "boundary": "A curriculum record does not establish delivery time, accessibility, learning effectiveness, or external validity." }, { "id": "NOTES-GENERATION-ATTRIBUTION", "path": "notes-latex/chapters/05-generation-attribution.tex", "role": "Conceptual alignment for effective context, atomic claims, citation objects, entailment, completeness, absorption, source use, and counterfactual tests.", "boundary": "The package uses independently authored exposition and diagrams; local notes are not copied and do not disclose a current closed platform." }, { "id": "LAB-L04", "path": "course-labs/L04_claim_evidence_source_graph", "role": "Deterministic offline synthetic fixture for claim segmentation, multi-axis annotation, graph construction, disagreement preservation, and bounded diagnosis.", "boundary": "Authored teaching labels and a ten-claim double-coded subset do not establish population reliability, model quality, or causal source use." }, { "id": "PAPER-CATALOG", "path": "sources/paper_catalog_2026-08-24.json", "role": "Frozen local mapping from PAPER-01 through PAPER-42 to source metadata.", "boundary": "Catalog inclusion is not peer review of course interpretation; each paper claim remains version- and design-bounded." }, { "id": "REFERENCE-CATALOG", "path": "course-site/app/data/references.json", "role": "Controlled mapping for PLAT and other reference families, including PLAT-04.", "boundary": "Candidate status and first-party product definitions do not make a reference a scientific mechanism or effect estimate." } ], "external_sources": [ { "id": "PAPER-24", "type": "arxiv_preprint", "title": "Diagnosing and Repairing Citation Failures in Generative Engine Optimization", "authors": "Zhihua Tian, Yuhan Chen, Yao Tang, Jian Liu, and Ruoxi Jia", "doi": "10.48550/arXiv.2603.09296", "url": "https://arxiv.org/abs/2603.09296", "role": "Bounded research route for stage-aware citation-failure diagnosis and targeted repair.", "boundary": "Taxonomy and reported outcomes remain tied to the paper version, benchmark, systems, and intervention design; they are not universal repair effects." }, { "id": "PAPER-33", "type": "arxiv_preprint", "title": "From Citation Selection to Citation Absorption: A Measurement Framework for Generative Engine Optimization Across AI Search Platforms", "authors": "Zhang Kai, He Xinyue, and Yao Jingang", "doi": "10.48550/arXiv.2604.25707", "url": "https://arxiv.org/abs/2604.25707", "role": "Bounded research route for separating visible citation selection from a reported absorption or influence construct.", "boundary": "The platform snapshot, fetch-conditioned pages, engineered features, systems, and date constrain interpretation; the measure is not hidden attention or causal source use." }, { "id": "PAPER-13", "type": "arxiv_preprint", "title": "Exposing Citation Vulnerabilities in Generative Engines", "authors": "Riku Mochizuki, Shusuke Komatsu, Souta Noguchi, and Kazuto Ataka", "doi": "10.48550/arXiv.2510.06823", "url": "https://arxiv.org/abs/2510.06823", "role": "Adversarial extension for publisher attributes, citation-source scrutiny, and poisoning-risk questions.", "boundary": "Reported topic, country, system, and time conditions limit security conclusions; the paper does not define a universal platform vulnerability level." }, { "id": "PLAT-04", "type": "official_platform_documentation", "title": "Introducing AI Performance in Bing Webmaster Tools Public Preview", "authors": "Microsoft Bing Webmaster Team", "url": "https://blogs.bing.com/webmaster/February-2026/Introducing-AI-Performance-in-Bing-Webmaster-Tools-Public-Preview", "role": "Platform-interface case for distinguishing citations, cited pages, sampled grounding queries, and trends from unsupported influence interpretations.", "boundary": "Official and changeable preview documentation is authoritative only for the named Microsoft surfaces and checked date; its counts do not establish rank, authority, entailment, absorption, or causal use." } ], "fixture_contract": { "lab_id": "L04", "synthetic": true, "offline": true, "deterministic": true, "response_count": 5, "claim_count": 25, "source_count": 8, "double_coded_count": 10, "relation_counts": { "support": 14, "contradiction": 4, "merely_cited": 3, "unsupported": 4 }, "dimension_level_disagreement_rows": 6, "input_sha256": { "responses.jsonl": "fd6ab45e5b54048c278a677c12397c4b23c0a70d44989f3e4771d11d03739502", "annotations_coder_a.csv": "394006259c7ad7c53e1ac020c9652729b7d35bc079db788862da9ee06979355c", "annotations_coder_b.csv": "dcc7df747689a6faec24eb6d318cea5abb2ab6137db54d1a94a7b5fd4908a58b", "source_registry.csv": "f8594d39b231365f26de6dfbbac8a6e2a2356934141e0f912e4fba0432ecacc8", "source_snapshots.jsonl": "caa73ad46da65973b90ca153f5ecea9956ce978cf43f2fe2ce6fc3cabe60c81d" }, "agreement_boundary": "descriptive on a ten-claim synthetic subset; not a general reliability estimate", "evidence_ceiling": "CES labels describe frozen authored annotations; they do not establish hidden retrieval, context selection, generator attention, or causal source use" }, "visual_policy": { "external_figure_reuse": false, "external_screenshot_reuse": false, "visuals": "original_instructional_diagram_specifications_only", "required_encodings": [ "label", "shape_or_position", "line_style_or_pattern", "alt_text", "table_equivalent_for_quantitative_visuals" ], "license_review_required_before_any_later_external_asset_reuse": true }, "claim_boundary": { "separate_objects": [ "source presence in a named set", "source use under a controlled contribution design", "faithful absorption of preregistered source-distinctive content", "displayed attribution to a resolved source", "claim-passage entailment", "answer correctness against appropriate evidence" ], "open_fixture_observable": [ "frozen synthetic response text and visible-source lists", "primary and second-coder annotation rows", "source registry and snapshot identity", "graph-builder invariants and deterministic outputs", "relation counts and dimension-specific disagreements", "input and package hashes" ], "closed_platform_unknown_without_separate_evidence": [ "crawl and retrieval state", "candidate set and ranking", "effective generator context", "model and decoding state when not exposed", "causal source contribution", "citation attachment mechanism", "post-processing and display suppression rule", "alternative unseen source origins" ], "not_evaluated": [ "public crawling", "production indexing", "retrieval recall or rank", "context packing", "public-platform answer generation", "causal source use", "training-data provenance", "population annotation reliability", "multilingual validity", "user referral or action", "commercial outcome", "legal or standards compliance" ] }, "quality_status": { "structural_validator": "passed_2026-08-24", "deterministic_fixture_alignment": "passed_2026-08-24", "scientific_human_review": "pending", "independent_reannotation": "pending", "accessibility_human_review": "pending", "timed_rehearsal": "pending", "classroom_pilot": "pending", "external_peer_review": "pending", "editorial_and_legal_review": "pending" } }