P1–P3 Reporting Layer · D1

Reporting Specification v1.0

Signed off · 23 August 2026. Public reporting continues the preservation obligations of P3: compression may simplify presentation, but it may not change claim, comparator, time, scope, deficits or record semantics.

01Six visible fields

Every public run report uses the same six-field grammar.

The fields are fixed before the first case card. They are reporting rules, not case-specific editorial choices.

1 · Claim tested

Preserve the locked claim.

Use the exact locked Cᵛ from the Run Record. Keep the original language. Do not translate, broaden or categorically strengthen the claim.

2 · Run date / Evidence state

Keep t_R visible.

The run date and evidence/practice horizon remain part of the verdict. Later evidence never overwrites an earlier run.

3 · Verdict

PASS · OPEN · LoA

Report the claim-level verdict. If Loss of Applicability occurs, the case page must identify the subtype LoA-T or LoA-K.

4 · Why

Preserve every deduplicated finding.

Use one sentence per deduplicated finding. There is no one-reason-per-card rule. If D_R = {}, state briefly why no applicability-relevant residual deficit remains.

5 · What this does not mean

Protect truth, scope and comparator.

This field is mandatory. It states truth-neutrality and the case-specific scope/comparator boundary: PASS is not truth, OPEN is not falsehood, and LoA is not physical falsity.

6 · What would trigger a new run?

Preserve record semantics.

State the Directive / record rule. New evidence, materially changed practice or a new theory state creates a new run; it does not retrospectively rewrite the old one.

02Language rule

Bilingual reporting is semantic parity, not independent rewriting.

Field 1 stays in the original language.

The locked quotation or claim version is byte-/word-preserved across EN and DE. A reader aid may appear outside Field 1 only if explicitly marked as explanation or paraphrase.

Fields 2–6 are localized.

EN and DE must be semantically equivalent, not necessarily literal. Verdict, date, Run ID, comparator and numerical values remain identical across both language versions.

03Preservation obligations

Seven hard constraints survive every layer of compression.

Claim preservation

No extension, categorical upgrade or silent translation.

Comparator preservation

Benchmark- and model-relative claims remain benchmark- and model-relative.

Temporal preservation

t_R remains part of the verdict; later evidence generates a new run.

Scope preservation

PASS, OPEN and LoA may not be paraphrased beyond the locked claim.

Deficit preservation

Independent deduplicated findings may not be merged into an artificial single “main reason”.

Truth neutrality

PASS ≠ true · OPEN ≠ false · LoA ≠ physical falsity.

Record semantics

“New run” and “audit correction” are distinct operations and must never be conflated.

04Technical metadata

Audit metadata is internal to the card, not a seventh visible field.

Each card carries: Report-Spec version, Run ID, source record, last card revision and language version. These fields support audit and regression testing.

Machine-readable rules P3 v2.0 + Supplement ↗

Gate G1 · SIGNED OFF

Reporting Specification v1.0 is frozen for the public run cards. Case reporting now proceeds under this same six-field grammar.