Whether a model applies the same reasoning strategy (highlights the same regions) across different instances of the same class.