%% PARS Audit 2026-08-16
%% Claim: PARS-EFFECT-001
%% Exact counts from two replicates; no composite score.
flowchart TB
    A["A — Naive\nSubstantive 11/11 · 11/11\nStatus 6/8 · 8/8"]
    B["B — Explicit verification\nSubstantive 10/11 · 11/11\nStatus 8/8 · 8/8"]
    C["C — Short verification kernel\nSubstantive 11/11 · 11/11\nStatus 8/8 · 8/8"]
    D["D — Full PARS\nSubstantive 11/11 · 11/11\nStatus 8/8 · 7/8"]

    A --> V["Observed: no full-PARS correctness advantage on this bank"]
    B --> V
    C --> V
    D --> V
    V --> L["Limit: two runs, one model family, small and ceiling-prone bank;\nno general equivalence or uselessness claim"]