Fine-Tuning
Further training a model on your own examples to change its style or format, not to teach it new facts.
Reviewed by the RadarTrek editorial team · June 2026
Fine-tuning continues training a base model on a curated dataset of your own input-output examples, shifting how it responds — tone, format, narrow task performance — without changing what facts it knows. It is commonly confused with RAG: fine-tuning teaches style and format, RAG supplies knowledge. Using fine-tuning to "add facts" usually backfires, producing a model that confabulates confidently about things it wasn't actually trained on.
Why it matters
- —Exhaust prompting and RAG before fine-tuning — most quality problems are prompt problems, not model problems.
- —Fine-tuning is the strongest lever for enforcing a very specific output style or format on every call.
- —At high volume, a fine-tuned small model can match a larger model's quality on a narrow task for a fraction of the cost.
Where to learn this
When to Fine-tune
LLM Evals and Fine-tuning course
This is the exact lesson that covers this term in depth — with examples, diagrams, and a hands-on exercise.