A municipal rain-barrel rebate instructor wants a written exam item that is “open the tool, run with RAG and without, and score the two packs.” What should a candidate recognize about that request?
Select an answer to reveal the explanation.
Short Explanation
Lab time is practice; the written exam asks you to summarize the idea. HO-4.1.2 compare-with-and-without-RAG work is not examinable. A fair item might still ask which architecture grounds cases in the current rebate handbook.
Full Explanation
Hands-on objective HO-4.1.2 covers comparing generation with and without RAG in a lab, but that compare-and-score exercise is not examinable on the written CT-GenAI exam. Fair written items stay at summarize depth: candidates explain RAG’s purpose, retrieve-then-generate flow, and how grounding in a current municipal rebate handbook improves case relevance without opening a tool or scoring dual packs. Treating hands-on RAG compare labs as the only examinable form of Chapter 4 content confuses practice-lab objectives with written K2 coverage and would exclude legitimate architecture questions. Requiring every written item to launch a training job before answering mixes fine-tuning operations into a RAG awareness item and exceeds examinable depth. Describing RAG solely as pasting the whole repository into the prompt restates the anti-pattern the syllabus rejects and is not a valid exam constraint. Exam caveat: HO labels mark non-examinable labs; the written paper still may ask which architecture grounds cases in the current rain-barrel rebate handbook. Operational check: if a proposed item tells candidates to open a tool, run dual packs, and score them, rewrite it as a summarize question about retrieve-then-generate grounding instead.