Data/ML/AI
Cross-Validation
Also written as Train/Test Split, Holdout Set, Data Leakage
Holding back some of the data during training and testing the model on it, so the score reflects unseen data rather than material the model has already seen.
Think of it like
Marking a student on questions you deliberately kept out of the revision pack.
Junior or senior?
Junior sounds like
Knows to split the data.
Senior sounds like
Knows the ways a split can leak, especially with data ordered by time, where training on the future is an easy accident.
Ask them
“How did you split your data, and could anything from the test set have leaked into training?”