Generalization is a model's ability to predict well on new data, never seen during training. It is the true goal of machine learning: a good score on the training data is not enough. It is estimated with a test set or cross-validation.
Example
Train score 0.98 but test 0.71: the model generalizes poorly.