AI Prompt Engineering and Key Concepts in Machine Learning and NLP Practice Test

RAG combines retrieval with generation: a retriever pulls documents relevant to the query, a ranker re-orders those documents by relevance, and a reader uses the retrieved text to craft an answer with citations. This sequence matters because the generator bases its response on evidence from the most relevant sources, making the answer grounded and traceable. If you try to generate before gathering or rereading the sources, you risk an answer that isn’t anchored in useful material or lacks proper citations. The description that includes retrieving documents, then ranking them by relevance, and finally using the retrieved text to produce an answer with citations captures how RAG typically works and why evidence-backed responses are possible. Other descriptions either skip the ranking step, place retrieval after reading, or mix up the order, which would undermine how RAG ensures the answer is tied to relevant sources.