A subset of data used to tune model hyperparameters and prevent overfitting during training. Separate from both training and test data to ensure unbiased model evaluation.