Preprocessing or feature creation uses information from the test set before evaluation.

The report documents dataset sources and licenses.

The model is evaluated on a hidden test set after training.

Cross-validation is performed only on the training data.