Audit scientific data semantics, sample and label alignment, leakage, group splits, preprocessing boundaries, and train-to-inference transforms. Use for any result based on datasets, feature matrices, tensors, repeated observations, or learned preprocessing.
