Discarding or regenerating outputs that fail a quality check, instead of training on everything a model produces.
Continue to AI University →