Module quiz

17 questions · pass bar 80% · retry as often as you like — your best score counts toward the gate.

  1. 1.Why is chunking called the highest-leverage decision in a RAG pipeline?
  2. 2.Which two query types does dense (embedding) retrieval handle poorly while BM25 handles them well?
  3. 3.How does Reciprocal Rank Fusion (RRF) combine BM25 and dense results, and why ranks instead of scores?
  4. 4.Why are bi-encoders used for first-stage retrieval but cross-encoders for reranking?
  5. 5.Define precision@5, recall@5, and MRR.
  6. 6.Your recall@5 is high but final answers are frequently wrong. Where do you look first?
  7. 7.What's the difference between faithfulness and answer relevance, and how does LLM-as-judge measure faithfulness?
  8. 8.How do you build a retrieval eval set cheaply but credibly?
  9. 9.What is query decomposition, and when is single-shot RAG structurally unable to answer?
  10. 10.Fixed single-shot RAG pipeline vs. agentic retrieval-as-a-tool: what are the trade-offs?
  11. 11.Your RAG system confidently answers questions the corpus can't support. Name three layered mitigations.
  12. 12.How does HyDE improve retrieval, and what does it actually embed?
  13. 13."Context windows are a million tokens — why not paste the whole corpus into every prompt instead of building RAG?" Which answer covers the senior objections?
  14. 14.When should you fine-tune a model versus building RAG, for a product that must answer questions about internal documentation?
  15. 15.What problem do contextual retrieval and small-to-big (parent-document) retrieval both solve, and what's the shared principle?
  16. 16.Your multi-tenant vector search returns almost nothing for small tenants, though their documents are indexed. What's the likely cause?
  17. 17.Before trusting an LLM judge's faithfulness scores, what validation does a senior engineer run?