Your question is Dense vs Sparse Retrieval Trade-offs. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
Discuss the trade-offs between dense retrieval and sparse retrieval when building custom search components for an LLM application.
Compare methods such as BM25 and embedding-based vector search, then explain when you would choose one or combine them. Address semantic relevance, exact-match behavior, indexing and query cost, latency, multilingual or domain-specific data, and operational complexity. Include an evaluation plan for retrieval quality and downstream answer faithfulness.
