Restricting the model's token choices at each step to only those that keep the output grammatically valid.
Continue to AI University →