Fine-tuning takes a general-purpose language model and continues training it on a curated dataset of domain-relevant examples. The result is a model that understands the specific terminology, conventions, and patterns of a particular domain or use case.
A bank that fine-tunes a language model on thousands of past loan assessment reports gets a model that understands credit terminology and produces output that requires less human editing. But fine-tuning is not always the right choice — it is more expensive and harder to update than RAG. For knowledge that changes frequently, RAG is typically more practical. For capabilities that need to be deeply embedded in the model's behavior, fine-tuning is the right approach.
