Fine-tuning

Training a base model on your own data so it specializes. Less common now that prompting and RAG cover most cases.