Rag And Retrieval medium
A compliance assistant must cite sources. Today the model writes its answer and is then prompted to add references. Auditors find citations to section numbers that do not exist. What is the best fix?
- A Lower the sampling temperature to zero so the model produces deterministic and therefore accurate references.
- B Fine-tune the model on the compliance corpus so it memorizes section numbers and can cite them reliably.
- C Tag each retrieved chunk with an ID, have the model cite only those IDs, and verify them against the retrieved set.
- D Run a second LLM pass that searches the public web for sources supporting each sentence the first model wrote.
Answer: C. Tag each retrieved chunk with an ID, have the model cite only those IDs, and verify them against the retrieved set.
Citations must point at **retrieved spans**, not at the model's memory. Constraining the model to chunk IDs and validating every ID makes each citation traceable. The cost is extra latency and more 'cannot answer' responses when the evidence is thin. Fine-tuning still has the model cite from memory, so it can hallucinate section numbers and goes stale when documents change.
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?