Vol II · Section 3
Retrieval & Ranking
The section where interviewers check whether you've read an IR textbook or just a vector-DB landing page. Use the vocabulary: recall, precision, nDCG, MRR, ANN recall, candidate generation.
1When would you use dense retrieval vs sparse retrieval?2How does hybrid search improve RAG?3When should you use a reranker after vector search?4How would you troubleshoot irrelevant documents being retrieved?5What is the difference between similarity search and semantic search?6How would metadata filtering improve retrieval?7How would you retrieve information from millions of documents efficiently?8What would you do if the correct document exists but consistently ranks too low?9How would you choose the value of top-k?10How would you design retrieval for queries containing multiple intents?