Which of the following statements accurately describe the RAG process? Note: There are 2 correct ans-wers to this question.

Question

Options

• AThe user's questi on is used to search a knowledge base or a set of documents.
• BThe embedding model stores the generated ans wers for future reference.
• CThe retrieved content is combined with the LLM's capabilities to generate a response.
• DThe LLM directly ans wers the user's question without accessing external information.

• A
  73% (22)
• B
  10% (3)
• D
  17% (5)

Explanation

RAG (Retrieval-Augmented Generation) works in two key stages: first, the user's query is used to search an external knowledge base for relevant documents (A), and then the retrieved content is fed into the LLM alongside the query so it can generate a grounded, informed response (C).

Why the distractors fail:

• B is wrong because embedding models convert text into vector representations for similarity search - they don't store generated answers. That's a database's job, not the embedding model's.
• D describes a plain LLM call with no retrieval at all, which is the opposite of RAG - the entire point of RAG is to augment generation with external information.

Memory tip: Think of RAG as "Search, then Speak" - the model retrieves before it generates. If an answer choice describes storing outputs or skipping retrieval entirely, it's breaking the RAG pipeline.

No community discussion yet for this question.