What is Fine-tuning?

Fine-tuning continues training a pre-trained model on a smaller, task-specific dataset, baking in a style, format, or domain skill. It excels at consistent formatting and tone; it is poor at adding factual knowledge, where RAG usually wins. Parameter-efficient variants like LoRA make it affordable.