Fine-tuning
In short: What is Fine-tuning?
Fine-tuning is the further training of an already trained model on targeted examples so that it performs a specific task or style more reliably.
Explanation
A general-purpose model can handle many tasks, but it does not know a company's domain terminology, tone of voice or strict output schemas. During fine-tuning, it continues training on a few hundred or thousand curated examples and internalises their patterns.
It is worth reaching for when careful prompting and retrieval (RAG) can no longer improve quality further. Fine-tuning is not the right tool for teaching fresh facts (RAG handles that better) but for locking in consistent behaviour, tone and format.
Example: a model is fine-tuned on previously approved document classifications so that it routes incoming records into the company's own categories consistently.