Signal-Template Agnosticism Is Not Model Independence proposes measurements for what anomaly-trigger systems fail to preserve.
The paper accepts the narrow claim used in deployed literature: AXOL1TL, CICADA, and GELATO can score events without a named beyond-standard-model signal template. It rejects the stronger inference that such systems are therefore sensitive to every physically possible novelty.
Three quantities are separated. OAR is an assimilation rate indexed to a specified candidate distribution of physically novel events. BAR measures ordinary acceptance on pre-registered withheld process families. IAI measures directional asymmetry under controlled inversion at fixed accepted-background rate. No general inequality converts BAR or IAI into a bound on open-world OAR.
Three protocols follow: a paired inversion battery plus deployed-model BAR audit; a prospective replay bank preserving compatible input states across algorithm generations; and cross-representation disagreement preservation with quantile-normalized scores. Every publication should also carry a per-stage retention map identifying what each gate makes unavailable.
The paper’s narrow conclusion is that foreclosure is structurally present wherever a bounded classifier decides what survives. Whether repeated foreclosure has produced longitudinal phenomenal collapse at the LHC remains unmeasured. The contribution is an instrument for asking, not a claim that collapse has already occurred.