Module 05 Knowledge Check
5 questions. Pass mark 4 out of 5. Answer every question before checking the answer key below, then retry after reading the feedback.
1. A retriever's job is to…
- A. Generate the answer
- B. Return candidate documents for a query
- C. Validate output
- D. Store embeddings
2. Metadata filtering matters most for…
- A. Speed only
- B. Restricting results by permission, recency, or document type before ranking
- C. Prettier output
- D. Smaller vectors
3. Hybrid (vector plus keyword) retrieval helps because…
- A. It is cheaper
- B. Exact identifiers and rare terms retrieve poorly by embedding alone
- C. It uses less memory
- D. It removes re-ranking
4. Raising k in a retriever tends to…
- A. Always improve answers
- B. Increase recall while adding distracting context that can degrade the answer
- C. Reduce latency
- D. Fix chunking
5. Before blaming the model for a bad RAG answer, check…
- A. The temperature
- B. What the retriever actually returned for that query
- C. The parser
- D. The token count
---
Answer Key and Explanations
Check these only after attempting every question.
1. B - Return candidate documents for a query
Keeping retrieval separate from generation is what lets you debug the two stages independently.
2. B - Restricting results by permission, recency, or document type before ranking
Filtering is how access control and freshness get enforced at retrieval time.
3. B - Exact identifiers and rare terms retrieve poorly by embedding alone
Product codes and error numbers are exactly what semantic similarity blurs.
4. B - Increase recall while adding distracting context that can degrade the answer
More retrieved text is not automatically better - irrelevant passages measurably hurt grounding.
5. B - What the retriever actually returned for that query
If the answer-bearing chunk never arrived, no amount of prompt tuning will fix the output.
Practical Check
Apply this module to your own work: complete the module activity for *Vector Stores and Retrievers*, then write one sentence naming what your result shows and one naming what it does not.
Strong Answer Pattern
A strong answer names the task, the evidence used, the check performed, and the remaining limitation. It avoids "proved", "guaranteed", or "always" unless the evidence genuinely supports it.