A GenAI assistant occasionally invents plausible-sounding but false facts (hallucinations). Which TWO techniques directly reduce hallucination? (Select TWO.)
Choose 2 answers.
A.Retrieval-augmented generation (RAG) — ground answers in retrieved source documents
Correct. Grounding in sources is the primary mitigation.
B.Lower the temperature for more deterministic outputs
Correct. Lower temperature = less inventive output.
C.Increase max output tokens
More tokens means more room to hallucinate, not less.
D.Use a larger font in the UI
UI cosmetics can't affect model factuality.
E.Ask users to ignore wrong answers
That's ignoring the problem, not mitigating it.
Why this answer
RAG grounds generation in real retrieved content — the strongest architectural mitigation. Lower temperature makes outputs more deterministic/conservative, reducing creative invention. Token limits affect length, not truthfulness.
Correct answers: A, B · Last reviewed 2026-10-28
Official AWS references
Practice this interactively
Answer without seeing the solution first in the Generative AI Concepts practice session.
Practice Generative AI Concepts