AIF-C01 Practice Question 26: Chain of Thought Prompting
AWS Certified AI Practitioner (AIF-C01) · Test 3 · question 6 of 10
An analyst asks a foundation model multi-step logistics questions: convert units, apply a discount tier, then compute a delivery window. The model replies immediately with a single confident number, and roughly a third of those numbers are wrong even though every fact needed is in the prompt. Which prompt engineering change is most likely to improve accuracy?
Show answer and explanation
Answer:
This is what chain-of-thought (CoT) prompting is for. Adding an instruction such as "think step-by-step", or naming the steps explicitly, makes the model break a complex problem into intermediate steps and reason through them before answering. AWS recommends it exactly for multi-step analysis, maths problems and complex reasoning, where it improves accuracy and also makes the reasoning visible enough to debug (option C). The trade-off is more output tokens and higher latency, which is why it is not recommended for simple tasks.
- A is wrong because temperature controls randomness, not reasoning depth. Setting it to 0 makes the model reliably give the *same* answer, including reliably giving the same wrong answer when it skipped the intermediate steps. - B is wrong because nothing is being truncated; the model is producing a short, complete, incorrect reply. Raising the token cap permits a longer answer but does not ask for one. - D is wrong because the failure is not missing information; the stem says every needed fact is already in the prompt. RAG fixes knowledge gaps, not arithmetic and sequencing errors.
Reference: Give Amazon Nova time to think (chain-of-thought)