Built and measured two committing readers on held-out (not just diagnosed): - resolve_pooled composer: 2 correct / 87 wrong (17%) - maximally-narrow forced reader: 0 correct / 61 wrong (100% when it fires) Shallow committing CANNOT be sound on real GSM8K (measured). The impasse: - candidate-graph path is SOUND (wrong=0) but FROZEN by ADR-0207 §4, covers 0 real - composer is OPEN (ADR-0207 §5) but UNSOUND (17% wrong, no separating gate) ADR-0207's premise (composer is the wrong=0-safe path) is FALSIFIED by held-out data; 'feed the composer' is a path to more confabulation. Needs a follow-up ADR. |
||
|---|---|---|
| .. | ||
| adr | ||
| analysis | ||
| architecture | ||
| audit | ||
| brief | ||
| briefs | ||
| curriculum | ||
| decisions | ||
| evals | ||
| handoff | ||
| handoffs | ||
| implementation | ||
| plans | ||
| research | ||
| sessions | ||
| workbench | ||
| zig | ||
| admissibility-exemplars.md | ||
| audio_pipeline_overview.md | ||
| capability_roadmap.md | ||
| claims_ledger.md | ||
| ethics_packs.md | ||
| EVAL_AUDIT_2026-05-20.md | ||
| eval_methodology.md | ||
| formation_pipeline_plan.md | ||
| frontier_baselines.md | ||
| gaps.md | ||
| hitl-backpressure.md | ||
| holdout_recipients.txt | ||
| identity_packs.md | ||
| master-plan-post-substrate-audit.md | ||
| model_dependency_size_tally.md | ||
| pack_inventory_2026-05-21.md | ||
| position_paper.md | ||
| PROGRESS.md | ||
| recognizer-registry.md | ||
| refusal-taxonomy.md | ||
| reviewers.yaml | ||
| runtime_contracts.md | ||
| RUST.md | ||
| safety_packs.md | ||
| teaching_order.md | ||
| test-debt-quarantine.md | ||
| truth_seeking_schema.md | ||
| Whitepaper.md | ||
| Yellowpaper.md | ||