Rag And Retrieval easy
A team raises the number of chunks passed to the LLM from 5 to 40, expecting better answers. Responses become slower and are sometimes less accurate. What is the most likely reason?
- A Approximate nearest-neighbor search becomes far less accurate once k grows past about 10.
- B LLMs can only attend to the first few chunks; the rest are silently truncated by the API.
- C Extra low-relevance chunks add noise and tokens, so the model can latch onto wrong passages.
- D Higher k forces the embedding model to re-encode the query once for every chunk that is returned.
Answer: C. Extra low-relevance chunks add noise and tokens, so the model can latch onto wrong passages.
**A higher k trades precision and cost for recall**: each extra chunk adds tokens, latency, and potential distractors that compete with the right passage. Option B is a common belief, but models do read the whole prompt; they simply get worse at using information buried among irrelevant text. A larger k helps when answers truly span many documents, such as summarizing across reports.
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?