Abhay Kumar Pathak

CNNs for ECG classification may achieve high accuracy by exploiting misleading visual cues, raising serious concerns about their clinical reliability.

ReMoDEx reveals that deep learning classifiers often rely on shortcut associations, exposing critical blind spots in conventional evaluation metrics.