rag-inference-vs-training-distinction
IN premise — summaries/2026/08/24/wiki-Retrieval-augmented_generation.md
Created 2026-08-25T02:58:52+00:00
RAG retrieves external documents at inference time, while fine-tuning bakes knowledge into model weights at training time; RAG avoids retraining costs.
Summary
This distinction matters because it determines how knowledge gets updated and at what cost. If the system relies on pulling in external documents at answer time, you can swap or refresh that knowledge cheaply; if knowledge is baked into the model itself, every update requires an expensive retraining cycle.