note · evergreen
Why RAG? Hallucination
A model’s knowledge is frozen: not current, out of domain, unsourced — and confident anyway.
Everything a model knows is frozen in its training data. So it cannot do three things at once: be current, know your domain, cite a source.
Rendering diagram…
Not currentknowledge frozen at training time
Out of domainnever saw your documents
No sourcescannot say where it came from
Confident anywaythe tone never changes
RAG in one sentence: put what the model does not know in front of it before it answers. Retrieval-augmented generation — retrieve first, then generate.
Fine-tuning is also an answer, just not the first one: every new fact means labelling, cleaning and GPU time again. For a small corpus that is overkill.