Design a Prompt That Elicits a Specifically Formatted Response GENERATIVE-AI-ENGINEER-ASSOCIATE Practice Questions (Page 4)

Questions 16–20

1. 16application · medium

   A team is using a foundation model to convert natural-language queries into SQL queries for a Databricks SQL warehouse. The model sometimes returns SQL with extra commentary like 'Here is the query:' or wraps the SQL in markdown code fences. The team wants the output to be only the raw SQL statement. Which approach is most effective?

   • AIncrease the temperature parameter to 0.0 for more deterministic output.
   • BProvide two examples in the prompt: one showing a natural-language input and the corresponding raw SQL output, and another showing a different input and its raw SQL output.
   • CAdd 'Be concise and to the point.' to the user prompt.
   • DAdd 'Do not include any commentary in your response.' to the system prompt.
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   A prompt asks the model to return a JSON object with a 'status' field. Which addition most effectively constrains the possible values for that field?

   • AAdd 'status' as a boolean.
   • BAdd 'status' as a number.
   • CAdd 'status' with allowed values: 'active', 'inactive', 'pending'.
   • DAdd 'status' as a string.
   Reveal Solution Discussion Select an answer first
4. 18expert · hard

   A company is deploying a model that acts as a customer service assistant. The system prompt contains the formatting rule: 'Always respond in JSON format with fields response and sentiment.' A user asks: 'What is your return policy?' The model responds with a plain-text answer instead of JSON. The team suspects the user prompt is overriding the system prompt. What is the most effective way to ensure the model always responds in JSON, regardless of the user's phrasing?

   • AUse a different model that is better at following system prompts.
   • BRemove the system prompt and rely on the user prompt to specify the format each time.
   • CMove the JSON formatting rule to the user prompt and repeat it for every user message.
   • DAdd a few-shot example in the system prompt showing a user question and the corresponding JSON response.
   Reveal Solution Discussion Select an answer first
5. 19foundation · easy

   After testing a prompt, the model returns a JSON object with an extra field that was not requested. What is the most appropriate next step?

   • ARemove the JSON format instruction entirely.
   • BAccept the output as is because the model is always correct.
   • CSwitch to a different model without changing the prompt.
   • DModify the prompt to explicitly list the allowed fields and state 'do not include any other fields'.
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   Which prompt instruction specifies both a field name and its data type for a structured response?

   • AInclude the user's age as a string.
   • BInclude 'age' as an integer.
   • CInclude the user's age.
   • DInclude the user's age in years.
   Reveal Solution Discussion Select an answer first