Dashboard › institutional-transition-lab › Distillation
efeea9b3-0898-49a6-bc03-260de1eaf84f["lore_tm_v1_a33XTkpCUeTYPqI3AnN6m1Hlyo2MZcqjVQNeWLUBlhQ","lore_tm_v1_ynriCkFEXXgA46JpEZgcnDPnWOz12niImC2QCSZt-Fo","lore_tm_v1_bFsKaopkDpBXQ_ekNvbguN1HVdxY4O0RQNjGbM2khx4","lore_tm_v1_pV4CIkOreceQ2EhfhnpomZxFcjpI7TYObbigO4Dkw_8"]
/home/byk/Code/institutional-transition-lab/cases/benchmarks/oss-institutional-events-v1.documents.json:1-14 defines schema version 2, benchmark ID oss-institutional-events-v1, and a label-free benchmark whose runner exposes only entity_id, publisher, published_on, and text; IDs, URLs, provenance metadata, labels, and thresholds remain outside the model prompt./home/byk/Code/institutional-transition-lab/cases/benchmarks/oss-institutional-events-v1.documents.json:7-14 defines document hashicorp-bsl-announcement for entity terraform, published by HashiCorp on 2023-08-10, sourced from https://www.hashicorp.com/en/blog/hashicorp-adopts-business-source-license, with source type primary_source_excerpt, context Curated excerpt from the linked source., and text stating that HashiCorp is changing future releases from Mozilla Public License v2.0 (MPL 2.0) to Business Source License (BSL) v1.1./home/byk/Code/institutional-transition-lab/cases/benchmarks/oss-institutional-events-v1.documents.json:16-24 defines document opentofu-foundation-governance for entity opentofu, published by Linux Foundation on 2023-09-20, sourced from https://www.linuxfoundation.org/press/announcing-opentofu, with source type primary_source_excerpt, context Curated excerpt from the linked source., and text that OpenTofu is part of the Linux Foundation and in a foundation’s hands rather than one company’s./home/byk/Code/institutional-transition-lab/cases/model-selection/event-triage-v1.json:1-8 defines schema version 1, policy ID event-triage-cost-frontier-v1, role event_candidate_discovery_and_triage, model gemini-3.7-flash, reasoning effort default, selection scope event_triage_only, and effective date 2026-08-27./home/byk/Code/institutional-transition-lab/cases/model-selection/event-triage-v1.json:9-15 records model-selection provenance: run ID 33086454444, artifact ID 9652609349, artifact SHA-256 44ea0132eb7416e2c4ad530e66b4e5c672700fd0eb2d9db6be5e5cb196bcd2d0, documents SHA-256 d0958b48ed916b9b42d324859615a1a0ad8aa61f026a937ac90733c4e703fd7e, and labels SHA-256 aef3b43ef642ca92ce55e35c5a4cc6786d7b42aa0476d4eec8b248fa73199149./home/byk/Code/institutional-transition-lab/cases/model-selection/event-triage-v1.json:16-20 explicitly does not authorize the selected model for canonical_power_graph_serialization, causal_identification, or performance_outcome_interpretation./home/byk/Code/institutional-transition-lab/tests/test_llm_event_probe.py:123-156 tests two responses—an institutional event for document ownership and a fenced-JSON control event for document release—using model gpt-5.6-luna, reasoning effort medium, elapsed times 12 ms and 8 ms, and allowlisted model input fields ["entity_id", "publisher", "published_on", "text"]; expected summary values are institutional precision 1, institutional recall 1, power-change recall 1, evidence-grounding rate 1, and all of contract_passed, quality_gate_passed, and passed equal to True./home/byk/Code/institutional-transition-lab/tests/test_llm_event_probe.py:159-168 verifies that the fabricated quote Buyer already owns Company is rejected against source text On 2025-01-02, Buyer agreed to acquire Company. with validation error evidence_quotes[0] is not verbatim source text./home/byk/Code/institutional-transition-lab/tests/test_llm_event_probe.py:171-187 verifies that a no_event response with invented title Invented title fails with no_event requires null event fields and empty evidence./home/byk/Code/institutional-transition-lab/tests/test_llm_event_probe.py:190-204 verifies that abstain may validate when no event is claimed, confidence is 0.5, evidence quote is The proposal is missing, and ambiguity is The proposal's contents are missing. against source text The proposal is missing from this excerpt./home/byk/Code/institutional-transition-lab/tests/test_llm_event_probe.py:207-212 injects an expected label into model documents and expects validate_benchmark() to raise ValueError matching invalid keys, protecting against label leakage./home/byk/Code/institutional-transition-lab/tests/test_llm_event_probe.py:215-244 verifies that omitting model_input_fields from one raw response makes that result schema-invalid with exactly raw response model_input_fields does not match the allowlist and makes summary["contract_passed"] false./home/byk/Code/institutional-transition-lab/tests/test_llm_event_probe.py:247-249 begins test_power_edge_requires_actor_and_target_not_just_right_and_direction, mutating the first institutional power change’s actor to Seller./home/byk/Code/institutional-transition-lab/schema/governance-coding-v1.schema.json as the closest schema: it uses JSON Schema draft 2020-12, a stable $id, additionalProperties: false, an explicit required-field list, inline closed enums, explicit null, format: date, minLength, and strict nested power-change/evidence-reference objects. Citations: schema/governance-coding-v1.schema.json:2-41./home/byk/Code/institutional-transition-lab/src/institution_lab/governance_coding.py:14-99 mirrors the schema ontology with Python constant sets and that /home/byk/Code/institutional-transition-lab/tests/test_governance_coding.py:21-31 uses test_frozen_json_schema_matches_validator_ontology to prevent schema/code drift./home/byk/Code/institutional-transition-lab/src/institution_lab/governance_coding.py:551-569, /home/byk/Code/institutional-transition-lab/scripts/copilot-code-governance.mjs:161-168.ValueError; per-record response problems accumulate as strings under schema_valid and validation_errors. Citations: /home/byk/Code/institutional-transition-lab/src/institution_lab/governance_coding.py:128-144, :241-315, :327-373.power_changes; it does not ground class, kind, title, dates, scope, or body/patch relationships, so the existing validator cannot satisfy issue #4 unchanged. Citations: /home/byk/Code/institutional-transition-lab/src/institution_lab/governance_coding.py:245-315.argparse, required Path arguments, deterministic JSON via indent=2, sort_keys=True with a final newline, generated Markdown summaries, and canonical JSON hashes using sorted keys plus compact separators. Citations: /home/byk/Code/institutional-transition-lab/src/institution_lab/governance_coding.py:102-104, :543-594./home/byk/Code/institutional-transition-lab/tests/test_governance_coding.py:34-99 has inline source-document, enrichment, coding-response, and wrapped-raw-response fixtures; (2) tests/test_governance_coding.py:102-157 tests metadata exclusion, exact named-component grounding, bounded-evidence routing, disagreement routing, and canonical-ledger exclusion; (3) /home/byk/Code/institutional-transition-lab/tests/test_source_enrichment.py:36-107 demonstrates immutable-envelope ID/hash validation and GitHub fixtures using httpx.MockTransport; (4) /home/byk/Code/institutional-transition-lab/tests/test_llm_event_probe.py:159-212 provides negative tests for fabricated evidence, inconsistent no_event, abstention rules, and leaked fields.tests/, while Ruff enables rules E, F, I, UP, B, and SIM. Citation: /home/byk/Code/institutional-transition-lab/pyproject.toml:47-55./home/byk/Code/institutional-transition-lab/cases/review/oss-governance-enriched-envelope-v1.json as the committed 40-record authority, pinning record_count, ordered-ID hash, record IDs, original source artifact, workflow run, and blinding declarations. Citations: cases/review/oss-governance-enriched-envelope-v1.json:1-7, :247-256.build_audit_envelope validates schema version, declarations, unique IDs, ordered-ID SHA-256, count, source-audit membership, and selection reasons. Citation: /home/byk/Code/institutional-transition-lab/src/institution_lab/source_enrichment.py:44-77./home/byk/Code/institutional-transition-lab/research/oss-governance-coding-results-v1.2.md:121-132./home/byk/Code/institutional-transition-lab/data/raw/** and /home/byk/Code/institutional-transition-lab/data/derived/** are ignored, so validation must not depend on an uncommitted local artifact directory. Citation: /home/byk/Code/institutional-transition-lab/.gitignore:12-15./home/byk/Code/institutional-transition-lab/cases/review/, following the committed source-document convention under cases/benchmarks/, because the current workflow retains generated evidence for only 30 days. Citations: /home/byk/Code/institutional-transition-lab/.github/workflows/continue-governance-coding.yml:98-104, /home/byk/Code/institutional-transition-lab/cases/benchmarks/oss-institutional-events-v1.documents.json:1-14.canonical_ledger_eligible=False; every summary reports zero canonical records; regression tests enforce both outcomes. Citations: /home/byk/Code/institutional-transition-lab/src/institution_lab/governance_coding.py:435-471, :491-517, /home/byk/Code/institutional-transition-lab/tests/test_governance_coding.py:123-138./home/byk/Code/institutional-transition-lab/cases/model-selection/event-triage-v1.json:3-19./home/byk/Code/institutional-transition-lab/schema/governance-adjudication-v1.schema.json; 2. /home/byk/Code/institutional-transition-lab/src/institution_lab/governance_adjudication.py; 3. /home/byk/Code/institutional-transition-lab/tests/test_governance_adjudication.py; 4. /home/byk/Code/institutional-transition-lab/cases/review/oss-governance-adjudication-v1.inputs.json; 5. /home/byk/Code/institutional-transition-lab/cases/review/oss-governance-adjudication-v1.json; 6. /home/byk/Code/institutional-transition-lab/research/oss-governance-adjudication-v1.md./home/byk/Code/institutional-transition-lab/schema/governance-adjudication-v1.schema.json, assistant proposed separating adjudication decision values accept|revise|reject|abstain from frozen record_class and requiring input pins, reviewer provenance, source URL, rationale, applicable final coding, abstention unresolved reason, and field-level evidence references./home/byk/Code/institutional-transition-lab/src/institution_lab/governance_adjudication.py, assistant proposed symbols validate_frozen_inputs, _index_unique_records, validate_evidence_ref, validate_adjudication_record, validate_adjudication, graph_edge_diff, summarize_adjudication, render_markdown, and main; summaries and gate status should be derived rather than trusting stored counts or a user-supplied gate boolean./home/byk/Code/institutional-transition-lab/tests/test_governance_adjudication.py, assistant proposed coverage for schema/constant parity, exact 40-ID equality, duplicate/missing/extra IDs, every pinned hash, URL mismatch, unknown source IDs, fabricated quotes, ungrounded record fields, reviewer provenance, unauthorized Gemini use, edge revisions, and unresolved-gate behavior./home/byk/Code/institutional-transition-lab/cases/review/oss-governance-adjudication-v1.inputs.json, assistant proposed preserving the exact source bundle and frozen V1.2 coding package needed for offline grounding and pinning run 33063470968, artifact 9642983578, artifact digest, member hashes, schema hash, and canonical document hash./home/byk/Code/institutional-transition-lab/cases/review/oss-governance-adjudication-v1.json, assistant proposed including all 40 record IDs exactly once with their actual adjudication decisions./home/byk/Code/institutional-transition-lab/research/oss-governance-adjudication-v1.md, assistant proposed a generated audit summary containing decision counts, unresolved IDs, edge additions/removals/revisions, reviewer types and assistance, input hashes, and gate status, following /home/byk/Code/institutional-transition-lab/research/oss-governance-coding-results-v1.2.md:1-5 and :121-140.institution-lab-governance-adjudication = "institution_lab.governance_adjudication:main" to /home/byk/Code/institutional-transition-lab/pyproject.toml:13-20.schema/core.sql, the exploratory event inventory, or the frozen V1.2 coding schema.abstain; 5. every accepted/revised institutional field and edge is grounded; 6. reviewer provenance is valid.canonicalization_gate: "closed" with explicit unresolved IDs; reject counts as resolved, while abstain counts as unresolved.test_one_abstention_keeps_canonicalization_gate_closed() in /home/byk/Code/institutional-transition-lab/tests/test_governance_adjudication.py: construct valid_frozen_inputs(), create resolved_adjudication_for_all_forty_records(inputs), replace the first record with an abstain decision, unresolved reason Required source evidence is incomplete., reviewer IDs ["reviewer-1"], and final_coding=None; call validate_adjudication(inputs, adjudication); assert record_count == 40, unresolved_record_ids == [record_id], and canonicalization_gate == "closed".uv run pytest -q tests/test_governance_adjudication.py tests/test_governance_coding.py tests/test_source_enrichment.py; 2. uv run ruff check src/institution_lab/governance_adjudication.py tests/test_governance_adjudication.py.uv run pytest -q; 2. uv run ruff check src tests; 3. npm ci; 4. npm run reports:build; stated these match /home/byk/Code/institutional-transition-lab/.github/workflows/ci.yml:28-45.