Model disagreement helps spot tricky inputs by measuring uncertainty through...

https://seo.edu.rs/blog/counterfactual-augmentation-for-disputed-inputs-how-does-it-work-11189

Model disagreement helps spot tricky inputs by measuring uncertainty through metrics like ensemble variance or prediction margin. When models differ on cases, it flags potential risk or data gaps