Can you answer your original question now? If not, you should think about the similarities between humans and LLMs more rather than get lost in technical pedantry.
Or think about an extreme case: an LLM that is trained on almost nothing combined with great RAG and context.
Or think about an extreme case: an LLM that is trained on almost nothing combined with great RAG and context.