All terms
Glossary
Fine-tuning
An existing model is further trained on your own examples, to hit style, format or domain language reliably.
Fine-tuning changes the behaviour of a model, not its knowledge in any useful sense. For current facts from your own data, retrieval is the right tool, not retraining.
It pays off when format and tone have to be exact, when the task is narrow and frequent, or when a smaller model suffices once it knows the job.
How you notice it
- Style or format must be hit reliably.
- There are a few hundred good example pairs.
- Prompting and context are already exhausted.
Frequently asked
Fine tuning or knowledge in the context?
Context for knowledge, fine tuning for behaviour. Anyone who needs current facts injects them at runtime, otherwise every change means retraining. Fine tuning is the answer to style, structure and tone.