A retrieval-backed agent starts anchoring on stale memory instead of fresh evidence. How would you resolve the conflict?

Instruction: Describe how you would handle disagreement between memory and current retrieved evidence.

Context: Tests how the candidate diagnoses the problem, chooses the safest next step, and reasons through recovery. Describe how you would handle disagreement between memory and current retrieved evidence.

Official answer available

Preview the opening of the answer, then unlock the full walkthrough.

I would define a precedence rule and make the agent follow it consistently. For factual or time-sensitive questions, fresh retrieved evidence should outrank memory unless there is a very specific reason not to trust the retrieval result.

I would also inspect how memory...

Upgrade to view official answer

Related Questions