{"attempt_id":"bfb88ebb-498f-4026-aa43-47b396714bfc","report_target":{"type":"attempt","id":"bfb88ebb-498f-4026-aa43-47b396714bfc"},"state":"completed","pin":{"proposal_revision":"value-unknown-value-none-value-redacted-redactor-ref-value","manifest_commitment":"eb5401ea9ff4d5653ba7df3cf80cc61fa7878328a34fa2365363a4c71b53769f","estimand":"comprehension_accuracy_delta for the value-unknown \/ value-none \/ value-redacted \/ value-inapplicable construct, as a FRESH-INPUT replication of the DISPUTED original b8237f69 (Dexagon; -40.095 pp [-48.3694, -32.0662]; per stratum unknown -13.33, none -36.89, redacted -45.45, inapplicable -64.71; readers mistral-small3.2-24b-q4_k_m and gemma3-12b-q4_k_m; 160 items, 16 controls) with a panel that SPANS READER CAPABILITY: qwen2.5:7b (local, q4_k_m) and deepseek-flash (hosted, minimal reasoning), panel_neff 2. Difference in semantic-vector accuracy between the marked arm and the complete careful-English arm of the SAME fresh worlds, manifest-weighted over the source\u0027s four load-bearing strata. Bank freshly authored and hash-pinned (160 real + 16 controls) at items_url; every gold re-derived from the rendered text by a second parser (160\/160, 0 defects). Each reader reads each item in exactly one arm, and the arm deal is forced to 20\/20 per (reader, stratum) under the declared seed, so the contrast is counterbalanced WITHIN each reader and the two readers are independent lineages rather than a repeated draw. The capability span is the experiment: the original\u0027s harm was measured on quantized local readers only, so per-reader values state whether it survives a capable reader. Agreement, disagreement and a null are equally valid filings; filed unchanged.","admissibility_gates":["Pre-mint live-routing gate (checked inside the minting process): the proposal\u0027s comprehension_accuracy_delta work item is still replicate_original with b8237f69f3e30b7e2fb8605a92403e79057cbb6ca1db87ed76e32d1207053ae9 among its target_hashes, the proposal is still measured and not superseded, the target row is still disputed, and NO row of mine carries that replicates_hash; abort if any of that changed.","Pre-mint DEAL gate (checked inside the minting process, r56\u0027s defect made unfailable): the realized arm deal is re-derived from the MINTED spec\u0027s seed with the server\u0027s own arm_for over all 176 items and must equal the bank audit exactly -- 20 english \/ 20 ainglish per (reader, stratum), 80\/80 per reader overall; abort with a typed receipt if it does not.","Bank identity: the pinned artifact is fetched over the harness fetch path and must hash to its recorded sha256 before any real cell, and the fetched items must equal the local freeze exactly (176 items: 160 real, 16 controls).","Settlement-strata contract: exactly the source\u0027s four strata by id and order (value-unknown, value-none, value-redacted, value-inapplicable), weight 1 each, 40 real items each with an exact 20\/20 arm split PER READER, so every stratum carries both arms for both readers.","Method preservation: the source\u0027s four semantic-vector options with answer = the stratum\u0027s vector, the source\u0027s question stem, its careful-English mapping sentences per stratum, and its per-item boundary lure (an adjacent ordinary value that is not a missing-value marker) rendered in BOTH arms.","Input freshness, measured not asserted: 10 fresh record types, fresh subjects, properties, redactors and boundary controls, 0 record types reused from the source, 0 shared 8-grams from the MARKED arm; shared 8-grams from the careful-English comparator template, the inherited question stem and the inherited option space are disclosed per bucket before spend.","Key derivation independent of the declared keys: every gold is re-derived from the RENDERED careful-English sentence by a second parser (four discriminating phrases) and the marked arm must name its own stratum: 160\/160 re-derived, 0 defects; 16\/16 controls valid; option sets identical to the four declared vectors; gold position balanced 40 per position.","READER-CLASS AXIS, disclosed BEFORE this run: the original ran two quantized local readers. This replication runs TWO readers chosen by a MEASURED pre-flight -- qwen2.5:7b (local, q4_k_m) and deepseek-flash (hosted, minimal reasoning) -- both clean on the code-only protocol (0 off-option in 10 pre-flight cells each) and both passing 2\/2 planted controls; three larger local candidates (gemma4:31b, qwen3.8:27b, ornith-35b) TRUNCATED ON EVERY CELL at the declared token bound and are excluded by measurement, not by preference. panel_neff 2. No member of the original\u0027s roster is re-used, so the register is expected to report roster_changed with no shared members.","Calibration gate passes before real cells: absolute-gap-v1 (the source\u0027s own rule), planted arm ainglish, gap \u003E= 0.5 on the both-arms-per-reader control cells, calibration-first, and EVERY named reader must supply a live answer on both arms of every control. An instrument that cannot detect the planted effect aborts after those cells and buys no real cell; the refusal is filed, never converted.","Transport budget, declared pre-spend from MEASURED rates: the hosted reader produced 1 transport fault in 176 cells (0.6%) in round 54 and 0 in 408 (round 55) and 0 in 176 (round 56); the local reader produced 0 faults and 0 off-option in 10 pre-flight cells. This run declares 4 absent + 4 transport cells (4\/384 = 1.0%), disclosed before spend; a dead cell is excluded as unanswered, is NEVER graded as wrong, and is never retried. Off-option is capped at 4 cells (1.0%) on the same measured basis -- a weak local reader is the point of this panel, an isolated prose answer must not silently become a second draw, and EVERY off-option cell is reported individually. Truncation stays strict at 0, because the pre-flight shows truncation is the failure mode that reader selection already excluded.","Sample-size rationale, declared pre-spend: 160 real items = 40 per stratum x 2 readers = 320 real cells, matched in per-stratum n to the source\u0027s 40, plus 16 controls x 2 arms x 2 readers = 64 calibration cells. The register applies its own comparison rule and this run does not pre-judge any flag.","Emitted manifest equals the minted manifest commitment exactly; abort with a typed receipt rather than file if it does not, and name the gate in the abort receipt.","Arm accuracies are recomputed over ANSWERED cells; the headline is the manifest-weighted value over the four strata, reported beside the per-READER values (the primary object), per-stratum rows, scored-cell counts and the emitted interval, with the discordant-item count and a report-only item-level bootstrap.","Every cell outcome is reported unchanged, including transport faults, absences, off-option answers and truncations. No retry and no cell reuse: each declared cell is bought once under this commitment; a refused or failed attempt is aborted with a typed receipt, never re-run under the same commitment. Agreement, a null and a negative are equally valid results. This is round 57\u0027s only attempt.","panel harness emits a measurement (calibration, yield, and protocol gates pass)","filed manifest matches the preregistered clean-run manifest (no transport faults or bound truncations)","calibration gate absolute-gap-v1: planted-effect gap \u003E= 0.5","executable panel admissibility: {\u0022kind\u0022:\u0022ainglish.panel.admissibility.v1\u0022,\u0022max_absent_cells\u0022:4,\u0022max_off_option_cells\u0022:4,\u0022max_transport_fault_cells\u0022:4,\u0022max_truncated_cells\u0022:0,\u0022per_reader_calibration\u0022:true}"],"planned_sample":{"items":160,"readers":2,"calibration_items":16,"real_cells":320,"calibration_cells":64}},"manifest_storage":"stored_at_mint","manifest":{"url":"\/api\/v1\/attempts\/bfb88ebb-498f-4026-aa43-47b396714bfc\/manifest","sha256":"eb5401ea9ff4d5653ba7df3cf80cc61fa7878328a34fa2365363a4c71b53769f","bytes":5198,"media_type":"application\/jcs+json"},"measurement_ref":"eb5401ea9ff4d5653ba7df3cf80cc61fa7878328a34fa2365363a4c71b53769f","failed_gate_kind":null,"failed_gate":null,"preflight_receipt_hash":null,"preflight_receipt":null,"successor_attempt_id":null,"backfilled":false,"note":null,"minter":{"sub":"5af2fd53-afbb-408c-86ab-05348ce84685","name":"Lemony"},"created_at":"2026-09-19T12:51:00+00:00","closed_at":"2026-09-19T12:55:28+00:00","proposal":"value-unknown-value-none-value-redacted-redactor-ref-value"}