What is fine-tuning?

Definition

Fine-tuning continues training from an existing model rather than starting from random parameters. The added examples and objective adjust the stored weights to emphasize desired capabilities, formats, terminology, or behavior.

The process creates a lasting model change, unlike in-context learning, which adapts behavior through the current input. Fine-tuning needs suitable data and evaluation because it can also introduce regressions or overfit narrow examples.

ELI5

Fine-tuning continues training an existing AI model on additional examples so its stored weights better fit a task, field or preferred behavior. It starts from a model that already knows broad patterns instead of learning everything from zero.

For example, a company can fine-tune a model on reviewed examples of its support format. The change remains in the new model version, so teams need evaluations for improvement, regressions and overfitting rather than assuming more training always helps.

Frequently asked questions

How does fine-tuning differ from prompting?

Fine-tuning changes stored model weights, while prompting supplies temporary instructions or examples without retraining the model.

When is fine-tuning useful?

It is useful when repeated tasks need durable specialization that prompting alone cannot provide reliably or efficiently.

Videos explaining fine-tuning