フリー問題

NVIDIA-Certified Associate: Generative AI LLMs のフリー問題 1 / 20 問目

問題文

After several rounds of rewording the instruction, an assistant now formats its answers exactly as the team wanted. A stakeholder asks whether this improvement is now permanent for every application that calls the same hosted model. What is the correct answer?

選択肢

  1. No, and the wording will be discarded by the tokenizer before the model sees it, leaving each request to start from the plain model.
  2. Yes. Once a phrasing has been validated the hosted model retains it as a default for later callers, which is why teams converge on one wording before sharing an endpoint.
  3. Yes, but only within the same session, and the endpoint keeps the most recent successful instruction in a short-lived cache and reuses it for later requests.
  4. No. The improvement lives in the text sent with each request, so any caller that does not send the same instruction will see the earlier behavior.

解答・解説を確認するには

正解と解説の確認、回答の記録には無料登録が必要です。登録すると演習モードでフリー問題に回答し、正誤と解説をその場で確認できます。