Venn — RAG vs. fine-tuning vs. prompt
Where these three customisation strategies actually overlap.
Make it your own.
title "Customising LLM behaviour"
set R "RAG"
set F "Fine-tune"
set P "Prompting"
only "R" "Retrieve fresh docs"
only "F" "Bake in style / format"
only "P" "Few-shot, system msg"
intersect "R,P" "Reranking + structured prompt"
intersect "F,P" "Tuned model + instructions"
intersect "R,F" "Tuned retriever"
intersect "R,F,P" "Domain-grade assistant"