RAG Without the Hand-Waving
A practical quiz about how RAG systems split, retrieve, rank, cite, and evaluate information, and how they fail. Every explanation links to supporting research or technical documentation.
Questions
- Not answered. What two kinds of memory did the original RAG model combine?
- Not answered. Which statements describe the original RAG design?
- Not answered. A chunk splits a refund rule from its exception. What should you fix first?
- Not answered. Why can whole-document chunks work poorly for clause-level questions?
- Not answered. What can fixed-window chunk overlap change?
- Not answered. Why are dual encoders useful for first-stage retrieval over a large corpus?
- Not answered. For L2-normalized vectors, which score gives the same ranking as cosine similarity?
- Not answered. Dense retrieval misses an exact error code. What should you add?
- Not answered. What can go wrong with approximate nearest-neighbor search?
- Not answered. Three of four relevant chunks appear in the top five results. What is recall@5?
- Not answered. Why might increasing retrieval
kfrom 5 to 50 make an answer worse? - Not answered. With a rank constant of 60, what RRF score do ranks 1 and 4 produce?
- Not answered. Why can RRF combine BM25 and vector rankings without normalizing their raw scores?
- Not answered. Where should an expensive query-passage reranker run?
- Not answered. The retriever drops the only supporting passage. Can the reranker recover it?
- Not answered. What did the "Lost in the Middle" experiments find?
- Not answered. How many tokens remain in this context budget?
- Not answered. What must a citation evaluator check?
- Not answered. Why score provenance separately from answer accuracy?
- Not answered. The right passage was retrieved. What can still go wrong?
- Not answered. Which parts of a RAG pipeline should be evaluated separately?
- Not answered. Why is one embedding leaderboard score not enough to choose a model?
- Not answered. Retrieval returns an outdated policy after the source changes. What should you fix?
- Not answered. Which RAG security risks have been demonstrated?