AI-102 Question 480
Select 2You are building a chatbot using Azure OpenAI and notice that the responses generated by the model are often vague and lack sufficient detail. You want to improve the response quality by applying prompt engineering techniques. Which of the following actions should you take to achieve this?
- A
Provide examples of desired responses in the prompt.
- B
Use a shorter prompt to minimize the model's focus on extraneous details.
- C
Specify the role or persona the model should assume in the prompt.
- D
Add temperature and token limit settings directly to the prompt text.
- E
Ask the model to summarize its own response for better clarity.
Show answer and explanation
Correct answers: A, C
Explanation
To improve responses using prompt engineering, you can provide examples of desired outputs to guide the model and specify the role or persona the model should assume to give it context and direction. These techniques help the model generate responses that are more aligned with your expectations. Other options, like adjusting model settings (temperature, token limit) or asking for summaries, fall outside the scope of prompt engineering or do not directly address the issue of vague responses.
- A. Correct.
Providing examples of desired responses helps guide the model on the level of detail and structure expected, leading to more precise and useful outputs.
- B. Incorrect.
Using a shorter prompt may reduce extraneous details but does not inherently improve the quality of vague responses. In fact, overly short prompts can lead to less context for the model to work with.
- C. Correct.
Specifying the role or persona the model should assume adds context and direction, which can help tailor the response to the intended use case.
- D. Incorrect.
Temperature and token limits are not part of the prompt text itself; they are configuration settings applied outside of the prompt.
- E. Incorrect.
Asking the model to summarize its own response does not directly address the problem of vague initial responses and is not a standard prompt engineering technique.