Fine-tuning

Fine-tuning means retraining an existing language model on your own data to specialize its behavior or style. It's powerful but often less relevant than RAG: use it only when prompting and context are no longer enough.

Rule of thumb: try prompting and RAG first; fine-tuning is a last resort.