Statistical consistency of sign-switching vacuum energy with cosmological observations

Abstract

We assess dataset agreement and late-time predictive adequacy in $\Lambda$CDM and its sign-switching extension, $\Lambda_{\rm s}$CDM, using a suite of Gaussian and exact non-Gaussian consistency diagnostics. Both models are constrained with cosmic microwave background measurements from Planck, ACT, and SPT, baryon acoustic oscillation data from DESI DR2, and low-redshift Type Ia supernova data from PantheonPlus+SH0ES. We find that commonly used Gaussian tension metrics can significantly overstate inconsistencies when broad, non-Gaussian posteriors are combined with tightly constrained datasets. In contrast, the exact non-Gaussian parameter shift indicates excellent consistency between CMB and BAO observations in both models. The $\Lambda_{\rm s}$CDM extension modestly improves geometric compatibility at intermediate redshifts, although reductions in parameter-level tension do not necessarily imply improved predictive consistency. These results highlight the importance of exact, non-Gaussian, and predictive diagnostics for robust assessments of cosmological model consistency.