What Is Fine-Tuning?

Fine-tuning is a way of taking an existing AI model and giving it additional training so it becomes better at a particular task or type of response.

Rather than building an AI model from scratch, you start with one that already understands language and can perform a wide range of tasks. You then train it further using carefully chosen examples, which can help it become more specialised in the way you need.

For example, an organisation might fine-tune a model using examples of the types of customer enquiries it deals with and the responses it would expect, helping the model learn how to handle that particular kind of task.

How Is Fine-Tuning Different From RAG?

The difference is in what happens to the model. RAG gives an AI relevant information to use when answering a question, whereas fine-tuning changes the model itself by giving it additional training.

If you want an AI to answer questions using your latest company policies, RAG may be more useful because those policies can be updated regularly. If you want the model to become better at performing a particular task or producing a particular type of response, fine-tuning may be more appropriate.

The two can also be used together, so it isn’t necessarily a choice between one or the other.