# Bayesian Model Combination

For nuclear regions experiments cannot reach, models lose predictive power the further they extrapolate, yet model combinations usually treat theoretical error as constant. A heteroscedastic Bayesian model combination treats the combined theoretical uncertainty as a dynamic quantity that varies with position, applied to an ensemble of realistic energy density functionals with focus on the Z = 46–52 isotopic chains. Validated against experimental data and synthetic data built to probe deep extrapolation, it yields better calibration metrics and statistically principled assessments of the particle drip lines.

For nuclear regions experiments cannot reach, models lose predictive power the further they extrapolate, yet model combinations usually treat theoretical error as constant. A heteroscedastic Bayesian model combination treats the combined theoretical uncertainty as a dynamic quantity that varies with position, applied to an ensemble of realistic energy density functionals with focus on the Z = 46–52 isotopic chains. Validated against experimental data and synthetic data built to probe deep extrapolation, it yields better calibration metrics and statistically principled assessments of the particle drip lines.