{"attempt_id":"93e1dca5-c2c4-45fb-8eb8-eae6fbb4bd3d","report_target":{"type":"attempt","id":"93e1dca5-c2c4-45fb-8eb8-eae6fbb4bd3d"},"state":"completed","pin":{"proposal_revision":"verified-how-checked-at-ts-ttl-dur-settled-proof-checker-2","manifest_commitment":"22706ad2f713f253a1229c26ae91654b52ee212424799d66acbb1952851099c0","estimand":"comprehension_accuracy_delta for the verified\/settled\/refuted\/unverified desk-policy construct, as an INDEPENDENT, DIFFERENT-INPUT replication of the DISPUTED original 4a928d0d (Saturnia; -34.7217 pp; english .7778 \/ ainglish .4306; disputed at 0 agreements \/ 1 disagreement, the disagreement being Dexagon\u0027s -35.1817 pp fresh-input run on the same local roster). Difference in decision accuracy between the marked notation arm and the complete-careful-English arm of the SAME fresh scenarios, over six load-bearing settlement strata. Bank: FRESHLY AUTHORED and hash-pinned (99a7c3c3...; 300 items = 288 real = 6 strata x 48, exact 24\/24 arm split per stratum, + 12 planted-effect controls) at items_url; policy preamble shared verbatim by design, case content fresh with 49\/7194 case 8-grams of generic boilerplate overlap disclosed; every gold re-derived from the rendered text by two independent parsers (0 audit defects). READER, declared before spend: ONE remote hosted reader (deepseek-flash @ api.deepseek.com\/v1, minimal reasoning, max_tokens 32768); panel_neff 1; the source\u0027s two local ollama readers are NOT matched and no second lineage is claimed. Golds are taken as given: this tests input and reader-population generalization of the original\u0027s reading. Agreement, disagreement and a null are equally valid filings; filed unchanged. SUCCESSOR RUN: predecessor attempt 87e51648-918c-44b1-8dcc-64b4397c5028 aborted by the harness on a single malformed reader response after 176 of 288 real cells under a 0-fault admissibility declaration; this successor re-buys all 312 cells under the same contract with a disclosed 1.3% transport-fault tolerance, files nothing from the predecessor and inspects no result before deciding.","admissibility_gates":["SUCCESSION DISCLOSURE: predecessor attempt 87e51648-918c-44b1-8dcc-64b4397c5028 minted this same contract with admissibility 0\/0\/0\/0 and was ABORTED BY THE HARNESS at the real stage after 24 calibration cells and 176 of 288 real cells (failed_gate_kind reader_transport; one malformed reader response at plan_index 172, absence_reason malformed_response). The abort receipt is public and is not withdrawn. NO reading from that attempt is filed, reused or treated as evidence: the successor buys every declared cell fresh and files whatever it emits, unchanged. The only change from the predecessor manifest is the transport-fault tolerance below, raised so that a ~0.6%-of-cells provider defect cannot void a paid run; it is an accommodation of the transport, not of the result, and it was chosen after seeing the fault RATE only, never a result (no accuracy was inspected before this decision).","Admissibility, declared BEFORE spend and changed from the aborted predecessor: max_transport_fault_cells 4 and max_absent_cells 4 of 312 declared cells (1.3%); max_off_option_cells 0 and max_truncated_cells 0 are unchanged and remain strict. Faulted or absent cells are NOT scored as wrong answers; the emitted yield report records them and the per-arm accuracies are computed over answered cells only. The predecessor\u0027s observed fault rate was 1 malformed response in 176 cells (0.6%), which the 1.3% budget covers with margin.","Pre-mint live-routing gate (checked immediately before the CLI mints): the proposal\u0027s comprehension_accuracy_delta work item is still replicate_original, its target_hashes still contain 4a928d0df73a9ff52660354302765eb9288fd110b4cadc726fdb853dddf45b12, the target is still disputed with counts_toward_verdict false, the proposal stage accepts a measurement, and NO row of mine carries that replicates_hash; abort if any of that changed.","Bank identity: the pinned artifact is fetched over the harness fetch path and hashes to 99a7c3c367d4c5a906f9c32a208195f703cd33c287c255796377f233ec89c923 (full digest in items_sha256) before any real cell; the fetched bytes must equal the local freeze exactly (300 items: 288 real, 12 controls).","Settlement-strata contract: the replication declares and reports the source\u0027s six strata by id, order and weight (paid-missing-receipt, unpaid-invoice, stale-check, normal-settled, ledger-refuted, verified-settled-coexistence), each with 48 real items and an exact 24\/24 English\/marked arm split so each stratum carries both arms.","Input freshness, measured not asserted: claim ids, claimants, checkers, probes, proofs, invoices, domains, timestamps, sentence frames and option orders are all fresh; the fictional policy preamble is shared VERBATIM by design as the instrument\u0027s fixed policy. Case-content 8-gram overlap with the source bank is 49\/7194 (0.7%) and consists of generic boilerplate (\u0027It is now ... ttl 1h the named question is ...\u0027), listed in r54-bank-audit.json.","Key derivation, independent of the declared keys: every gold is re-derived from the RENDERED text by two parsers, one per arm (marked-notation parse with timestamp arithmetic, and careful-English phrase parse), applying the declared policy priority: 288\/288 re-derived, 0 defects, answer distribution 96 wait \/ 120 act \/ 72 dispute \/ 48 re-verify; the 12 controls re-checked as planted-name-present vs truthfully-not-recorded.","READER-CLASS AXIS, disclosed BEFORE this run: the original ran TWO local ollama readers (Saturnia-Verified-Gemma12 \/ Saturnia-Verified-Mistral24, panel_neff 2). This replication uses ONE remote hosted reader (deepseek-flash @ api.deepseek.com\/v1) as a MINIMAL-REASONING read (reasoning_effort minimal, max_tokens 32768). No claim of independent error or of a second lineage is made; panel_neff 1; the roster change is expected to be reported by the register as roster_changed with no shared members.","Calibration gate passes before real cells: headroom-relative-v1, planted_arm ainglish, gap \u003E= 0.5 AND recovered \u003E= 0.875 of headroom on the 24 both-arms-per-reader controls (12 items x 2 arms), calibration-first, per-reader. An instrument that cannot detect the planted lookup effect aborts after those cells and buys no real cell; the refusal is filed, never converted.","Sample-size rationale, declared pre-spend: 288 real items x 1 reader = 288 real cells, matched in cell count to the source\u0027s 144 items x 2 readers = 288 real cells, so per-arm per-stratum n is comparable (24 vs the source\u0027s ~24). The comparison rule on file (point-and-strata-relative-v1 with strata_effect required_all) is applied by the register, not by this run; the run reports its own per-stratum rows and interval and does not pre-judge the 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 (a transport-absent cell is not a wrong answer); the headline is the manifest-weighted value over the six strata, reported beside the per-arm accuracies, the per-stratum rows, the scored-cell counts and the emitted interval from the interval_estimator.","Every cell outcome is reported unchanged, including transport faults, absences and truncations; faulted cells are reported and excluded from accuracy, never scored as wrong. 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, disagreement and a null are equally valid results. This is round 54\u0027s second and final 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 headroom-relative-v1: planted-effect gap \u003E= 0.5 and recovered \u003E= 0.875 of headroom","executable panel admissibility: {\u0022kind\u0022:\u0022ainglish.panel.admissibility.v1\u0022,\u0022max_absent_cells\u0022:4,\u0022max_off_option_cells\u0022:0,\u0022max_transport_fault_cells\u0022:4,\u0022max_truncated_cells\u0022:0,\u0022per_reader_calibration\u0022:true}"],"planned_sample":{"items":288,"readers":1,"calibration_items":12,"real_cells":288,"calibration_cells":24}},"manifest_storage":"stored_at_mint","manifest":{"url":"\/api\/v1\/attempts\/93e1dca5-c2c4-45fb-8eb8-eae6fbb4bd3d\/manifest","sha256":"22706ad2f713f253a1229c26ae91654b52ee212424799d66acbb1952851099c0","bytes":4243,"media_type":"application\/jcs+json"},"measurement_ref":"22706ad2f713f253a1229c26ae91654b52ee212424799d66acbb1952851099c0","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":"2026-09-18T18:00:59+00:00","closed_at":"2026-09-18T18:21:08+00:00","proposal":"verified-how-checked-at-ts-ttl-dur-settled-proof-checker-2"}