Session 11· 8 questions

End-of-session Quiz

What you'll learn

▸Confirm you understand the full RAG pipeline end-to-end
▸Identify any gaps in chunking, retrieval, or generation

8 questions covering the full RAG pipeline: chunking, embeddings, vector search, BM25, hybrid retrieval, and grounded answers. Aim for 6+ to confirm you understand the full pipeline.

Phase 1 — Indexing

Knowledge Check

Why do we chunk documents instead of pasting the whole thing into a prompt?

Knowledge Check

When is size-based chunking the best choice?

Code Check

What does cosine_similarity return for two identical embedding vectors?

Phase 2 — Retrieval

Knowledge Check

What kind of query does BM25 handle better than semantic search?

Knowledge Check

Why can you not simply add raw BM25 scores and vector similarity scores?

Code Check

What is the Reciprocal Rank Fusion formula?

Phase 3 — Generation

Knowledge Check

Why does the RAG prompt use XML tags like <context> and <excerpt>?

Knowledge Check

Why is refusal ("I don't have enough information") a feature, not a bug?

Scored 6+ out of 8?

Congratulations — you have completed the RAG pipeline from chunking to grounded answers! You now have the tools to build production RAG systems.