When the Assistant Does Not Know: Answer, Clarify, Retrieve or Decline

HardLLMs & GenAI

Reported in interviews at OpenAI

Problem statement

How would you help users when a busy chat assistant does not know? It currently answers directly despite stale facts, missing personal information and vague requests. It should communicate uncertainty and help through clarification, retrieval, tools or refusal. Common requests must stay fast, with progress on slower paths and citations for factual answers.

Choose how to handle missing knowledge, calculate direct-answer coverage and describe the decision policy.

Show hint

For each kind of query the assistant cannot answer, where does the missing information actually live?

Relevant roles: ML Engineer, LLM Engineer, Gen AI Engineer and AI Engineer.

Related free problems