Fine-Tuning
Fine-tuning is the further training of an already trained language model on a smaller, specific dataset, so that it performs a narrower task better or adopts a particular tone or specialist vocabulary.
In practice
For most companies, RAG – feeding relevant documents to the model at the moment of the query – or good prompt engineering is a more practical alternative to fine-tuning, because both are cheaper and easier to keep up to date. Fine-tuning pays off mainly for very specific, recurring tasks with stable requirements and enough training data.