Overfitting occurs when a machine learning model learns not only the underlying patterns in the training dataset but also the noise and idiosyncrasies unique to that data, causing it to perform excellently on the training set but poorly on unseen data. This typically happens when the model is excessively complex relative to the volume and variability of training examples, leading to low bias but very high variance. Techniques to diagnose overfitting include monitoring training versus validation error curves and applying cross-validation. Common mitigation strategies involve simplifying the model, gathering more training data, applying regularization methods, or using data augmentation to increase dataset diversity. -edit