What does RAG (Retrieval-Augmented Generation) do?
RAG: augments a language model by retrieving relevant external knowledge and including it in the prompt, grounding answers without retraining the model.
2 / 5
What is the role of chunking in a RAG pipeline?
Chunking: documents are divided into manageable passages so each can be embedded and retrieved independently; chunk size and overlap strongly affect retrieval quality.
3 / 5
Why does RAG help reduce hallucinations?
Grounding: by supplying authoritative passages in context, RAG steers the model toward factual, attributable answers rather than fabricated ones, though it does not eliminate hallucination entirely.
4 / 5
What does a reranker add to a RAG pipeline?
Reranking: a second-stage model scores the top-k retrieved chunks for finer relevance, improving the quality of context passed to the generator over raw vector similarity alone.
5 / 5
What is a key limitation of RAG to manage?
Context quality: RAG is only as good as its retrieval; poor chunking or low-precision retrieval injects noise that can mislead the model, so retrieval evaluation matters.