What is Fine-tuning?

A base model knows language; fine-tuning adjusts how it behaves — your support-ticket tone, your house contract format, your medical-report style. You supply hundreds to thousands of good examples, and training nudges the model's parameters toward that pattern.

The classic mistake: fine-tuning to teach facts. Facts change, and a tuned model states them confidently with no sources — fresh hallucination risk. Facts belong in retrieval, where they stay current and citable. Style and format belong in fine-tuning.