Fine-tuning

Fine-tuning is further training a pretrained language model on a smaller, task- or domain-specific dataset to adapt its behavior for a particular use.

Rather than training from scratch, fine-tuning updates an existing model’s weights on curated examples, which can improve accuracy, style, or format adherence on a narrow task.

It adds training and maintenance cost and can age as data shifts, so whether it pays off is best settled by evaluating outcomes against a benchmark rather than assumed.