Rag And Retrieval medium
Your assembler fills a 6,000-token budget with the top 8 chunks. For many queries, 5 of the 8 are near-identical copies of the same FAQ, mirrored across several internal sites. Answers miss facts held in other documents. What should you do?
- A Deduplicate near-identical chunks and apply a diversity step such as MMR before filling the token budget.
- B Increase top-k to 30 so the unique documents eventually appear after the duplicated chunks in the ranking.
- C Increase chunk size so each FAQ copy occupies only one chunk, leaving more room for other relevant documents.
- D Add a reranker over the top 8, which will push the duplicate FAQ copies below the unique documents in order.
Answer: A. Deduplicate near-identical chunks and apply a diversity step such as MMR before filling the token budget.
**Deduplication** and **Maximal Marginal Relevance** give up a little raw relevance in exchange for coverage, so the token budget holds distinct evidence. A reranker scores each duplicate about as high as the original, so the duplicates stay on top. A higher k adds tokens without removing the redundancy. Diversity steps matter less when the corpus is deduplicated at ingestion.
One system design question a day, with spaced repetition so the ones you miss come back.
Get the appRelated questions
- You must index 100M chunks using a 3,072-dimension embedding in fp32, served from RAM via HNSW. The raw vectors alone take about 1.2 TB before graph overhead. You need to cut memory several-fold while losing at most a point or two of recall@10. Which approach fits?
- A multi-tenant SaaS app stores all tenants in one vector index with a tenant_id field. Queries fetch the top 20 chunks by similarity and then drop chunks from other tenants. Small tenants often get 0-2 results. Why does this happen, and what is the fix?
- Runbooks are split into fixed 300-token chunks with no overlap. When a procedure step starts near the end of one chunk and finishes in the next, answers come back incomplete. What is the lowest-cost fix?
- Your offline eval shows recall@5 of 92% on a labeled set, so the gold passage is in the context for nearly every question. Yet human graders mark 35% of answers wrong, and the wrong answers often ignore or contradict the gold passage. Where should the team invest next?
- Your product requires every RAG answer to include clickable source citations that users can verify. How should those citations be produced?
- A runbook RAG system uses fixed 500-token chunks with no overlap. Users who ask for a multi-step recovery procedure get answers that stop at step 4 of 9, and retrieval logs show the top hit ends mid-procedure. What is the most targeted fix?