Validation data is a held-out dataset used during model development to tune hyperparameters, compare models, and choose training approaches.
Validation data is distinct from training data, which teaches the model, and test data, which should be reserved for final evaluation after development choices are made.