Simplicity Bias

An inherent tendency of machine learning models to prefer simpler hypotheses when fitting training data. This bias can cause models to learn shortcuts rather than genuine reasoning rules, potentially leading to poor generalization when the underlying data distribution shifts.