A school-lunch allergen classifier won on 120 notes after one lucky shuffle. What should happen before the kitchen changes process?
Select an answer to reveal the explanation.
Short Explanation
One lucky shuffle on 120 notes can flatter a weak run. Repeat with different seeds or folds before the kitchen changes process. More layers and a fluency swap do not cancel shuffle luck.
Full Explanation
Small NLP datasets are noisy. A single shuffle can put the easy notes in validation and make a weak run look like a win. Repeated seeds or cross-validation show whether the gap holds. Changing a kitchen process, or stacking more layers, on one lucky split is not an experiment.