fine-tuning

Definition

Fine-tuning takes a pretrained base LLM and trains it further on a smaller, task-specific dataset of prompt-and-ideal-response pairs. The result is a model that follows the dataset's style and conventions out of the box. For most knowledge-base questions, RAG is cheaper and easier than fine-tuning; fine-tuning shines when output format or tone matters more than recall.

When to use

See also

• RAG — Retrieval-Augmented Generation — fetch relevant documents at query time and feed them to the LLM as context.
• LLM — Large Language Model — a transformer-based model trained on internet-scale text to generate and reason.