Glossary
Fine-tuning further trains an existing model on your own examples so it adopts a particular style, format or task behaviour. It teaches the model how to respond, not what is currently true.
Retrieval supplies facts at question time and updates the moment the underlying documents change. Fine-tuning bakes patterns into the model's weights, where they are fixed until it is trained again.
Consistent output structure, a particular tone, or a specialised classification task are reasonable uses. Product details, prices and policies are not: they change, and a fine-tuned model will keep confidently reciting the old version.
For most business applications, retrieval over current documents is the better first architecture.
Usually not for factual content. Facts change, and fine-tuned facts go stale invisibly. Use retrieval for what is true and fine-tuning only for how output should be shaped.
When you need consistent format or tone at volume, or a specialised classification task where examples teach the behaviour better than instructions do, and where the maintenance cost is understood.
Fine-tuning further trains an existing model on your own examples so it adopts a particular style, format or task behaviour. It teaches the model how to respond, not what is currently true.
