Fine-tuning (vs RAG and prompting)
Further training a model on your own examples so it adopts a style, format or narrow skill; poor at teaching it new facts.
The usual order is prompting first (cheapest, fastest to change), then RAG when the model needs knowledge it does not have, then fine-tuning when you need a consistent format or tone, or a smaller, cheaper model that does one narrow task well. Fine-tuned facts go stale and cannot be cited; documents in RAG can be updated tomorrow and linked in the answer. Fine-tuning also needs a few hundred good examples and an eval set to prove it helped.
Related terms
- RAG (retrieval-augmented generation): Finding relevant documents first and putting them into the prompt, so the model answers from them.
- Local LLM: A language model running on hardware you control instead of a provider’s API, so documents stay where you decide.
- Evals: A repeatable test set that scores an AI feature, so changes can be measured instead of guessed.
