Dashboard › institutional-transition-lab › Distillation
a13b7840-fb1e-4fb0-b28f-a4520a6ea3a7["lore_tm_v1_a_t_y4xGMrGjG1bZaVThVp4BB51lN-Vziob3OE7Q8Xg","lore_tm_v1_h27vTDshKgHEh-UWr42OzsOgLIY6WPt99zuh2xY5K1s","lore_tm_v1_Tg4bw7srNc_mbTPN5FhXXDoaLtWChLLpa84B8usLBRI","lore_tm_v1_APZTf0mY14juygs0v0xRfYYSoImMdtXIndz-UGbC6zc","lore_tm_v1_kJVvwDmbI6KDxx2hGkbqQ2pd_CV13ovdOA4fAPWWd60"]
/home/byk/Code/institutional-transition-lab/src/institution_lab/governance_coding.py defines parse_response(text: str): strips whitespace, accepts an optional complete triple-backtick code fence, rejects an unterminated fence, parses with json.loads(), and requires the parsed value to be a dictionary.validate_response(response: dict[str, Any], document: dict[str, Any]) -> list[str] in /home/byk/Code/institutional-transition-lab/src/institution_lab/governance_coding.py requires the response keys to equal RESPONSE_KEYS; validates record_class against RECORD_CLASSES, non-null event_kind against EVENT_KINDS, body_patch_relation against BODY_PATCH_RELATIONS, and confidence as a finite non-boolean number in the inclusive range 0–1.validate_response() requires title, affected_scope, and ambiguity to be non-empty strings or null, and validates announced_on and effective_on as ISO dates or null via _valid_date().validate_response() constructs source_map from each document source’s source_id and text; requires power_changes to be an array; and requires each power-change object to have exactly POWER_KEYS, non-empty actor, target, and scope, and allowed right_kind, direction, and change_status values.power_changes[*].evidence_refs value must be a non-empty list. Each reference must have exactly REF_KEYS, identify a source containing string text, provide a non-empty quote, and have that quote occur literally in the referenced source text; otherwise validate_response() reports the reference as “not grounded.”validate_response() treats effective_institutional_change, announced_institutional_change, and proposal_only as institutional classes requiring non-empty power_changes; all other classes must have no power changes. A response with record_class == "abstain" must provide non-null ambiguity._validate_run() in /home/byk/Code/institutional-transition-lab/src/institution_lab/governance_coding.py requires a non-empty document list, requires each raw response to be a dictionary with a non-empty record_id, rejects duplicate record IDs, and requires the raw-response ID set to exactly match the coding-document ID set.foundation_transfer in 14 locations: /home/byk/Code/institutional-transition-lab/schema/governance-adjudication-v1.schema.json line 162; /home/byk/Code/institutional-transition-lab/schema/governance-coding-v1.schema.json line 10; /home/byk/Code/institutional-transition-lab/cases/review/oss-governance-adjudication-v1/v1.2-coding-package.json lines 2817 and 2851; and /home/byk/Code/institutional-transition-lab/cases/benchmarks/oss-institutional-events-v1.labels.json lines 36–37, 87–88, 105–106, 132, 167, and 185–186.no_event: /home/byk/Code/institutional-transition-lab/research/oss-governance-retrieval-development-v2.md lines 64 and 86; /home/byk/Code/institutional-transition-lab/research/oss-governance-enriched-audit-envelope-v1.md line 56; /home/byk/Code/institutional-transition-lab/research/oss-governance-coding-results-v1.2.md lines 28 and 108; /home/byk/Code/institutional-transition-lab/research/luna-terra-hard-case-comparison.md line 56; /home/byk/Code/institutional-transition-lab/research/luna-event-extraction-development-benchmark.md lines 80 and 104; and /home/byk/Code/institutional-transition-lab/research/llm-event-triage-cost-frontier-protocol.md lines 116 and 118.no_event; Valkey #345 was again labeled no_event by Luna despite entering the audit envelope; the five provisional core agreements comprised three no_event records and two others; Terra called a reviewed case no_event and was deemed unreliable as a tiebreaker; hard-case predictions varied among no_event, abstain, and once control_event; both clean benchmark runs classified the Valkey governance excerpt as no_event; Run B used no_event for an ambiguous case; and a triage protocol observed Luna labeling the Valkey governance document no_event, while another ambiguous proposal produced two abstain decisions and one no_event.