RAG Interview Questions
Dashboard
0%
1
Curious builder0 XP earned · 300 to level 2
0 daysFinish a lesson to begin
Badge collection0 of 6 unlocked
51 small wins to finish your pathNext question →

Q32IntermediateScenario

Retrieved documents contradict each other (an old vs new policy, two teams' docs). How should the system handle it?

30-second answerSay your answer out loud first, then reveal.

Layers

  1. Data governance (best fix):
    • Archive or delete superseded documents; mark status: superseded with superseded_by.
    • Define authority: the official policy portal outranks a team wiki page, which outranks Slack messages.
  2. Retrieval:
    • Filter to status = current by default.
    • Boost by authority and recency.
    • Deduplicate near-identical chunks across versions.
  3. Generation:
    • Include metadata in context: <doc id="D3" source="HR Policy Portal" effective="2026-04-01">.
    • Instructions: "If sources conflict, prefer the most recent official policy, and mention that an older or other version says otherwise, with citations."
    • For high-stakes questions, ask the user to confirm or escalate to a human.
  4. Feedback loop: detect conflicts (an LLM flags contradictions among the retrieved chunks), log them, and send a report to content owners.
Example answer. "As of the April 2026 HR policy, casual leave is 12 days [D3]. Note: an older team wiki page says 10 days [D9]; the HR policy is the authoritative source."
PreviousExplain vector quantization and Matryoshka embeddings. What are the trade-offs?

You understood something today that you didn't yesterday.