RAG or fine-tune?

RAG feeds a model your documents at question time. Fine-tuning changes its weights. Three questions point you at the right one.

1. What do you want to change about the model?It should answer from my documents or dataIt should write differently. Style, format, or domain habitsBoth: my data and my style

2. Do you have training examples (prompt to ideal answer pairs)?Yes. Hundreds or moreNo, or only a handful

3. Do answers need to point back to their sources?Yes. Citations matterNo. Answers alone are fine

Answer all 3 questions to get a recommendation.

← Back to the Field Guide