Balancing methods used to address imbalanced datasets can inflate the predictive multiplicity in the Rashomon set, leading to increased risks in model selection.
