Design a Prompt That Elicits a Specifically Formatted Response GENERATIVE-AI-ENGINEER-ASSOCIATE Practice Questions (Page 3)

Questions 11–15

1. 11foundation · easy

   Which prompt component is most effective for teaching a model a specific output format through examples?

   • AA user prompt that asks 'What is the capital of France?'
   • BA few-shot example with two input-output pairs in the desired format.
   • CA single instruction stating 'Use JSON'.
   • DA system prompt that says 'Be concise'.
   Reveal Solution Discussion Select an answer first
2. 12application · medium

   A team is using a model to extract product information from web pages. The output must be a JSON object with fields `product_name` (string), `price` (number), and `in_stock` (boolean). The model sometimes returns `price` as a string like '$19.99' or `in_stock` as 'yes' instead of `true`. Which prompt addition is most effective in enforcing the correct data types?

   • AAdd 'Use the same data types as in the training data.'
   • BAdd 'The price field must be a number without currency symbols, and the in_stock field must be true or false.'
   • CAdd 'Return the product information as a JSON object.'
   • DAdd 'Return the product information accurately.'
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A company is deploying a chat application where users ask questions about internal documents. The application must always return answers in a specific JSON structure with fields `answer` and `source_documents`. The team has a single prompt template that includes both the formatting rules and the user's question. They notice that when users phrase their questions in unusual ways, the model sometimes forgets the JSON format. What is the best way to make the formatting rules more robust?

   • AUse a different model that is better at following instructions.
   • BRemove the formatting rules from the prompt and rely on post-processing to convert the output to JSON.
   • CMove the JSON formatting rules to the system prompt and keep only the user's question in the user prompt.
   • DRepeat the JSON formatting rules at the end of every user prompt.
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A data engineering team uses a Databricks notebook to extract invoice details from unstructured email text. They need the model to return each invoice's vendor name, invoice number, and total amount so the output can be directly parsed by a downstream Python script using the `json` module. The team has already tried a simple prompt like 'Extract the invoice details from this email' but the model returns inconsistent plain-text responses. Which prompt modification is most likely to produce the required output?

   • AAppend 'Use markdown formatting for the response.' to the user prompt.
   • BAdd 'Be very careful and thorough when extracting the details.' to the user prompt.
   • CAdd 'Return the result as a JSON object with keys vendor_name, invoice_number, and total_amount.' to the user prompt.
   • DChange the user prompt to 'Extract the invoice details as a list.'
   Reveal Solution Discussion Select an answer first
6. 15foundation · easy

   In a prompt-based application, where should the instruction 'Always respond in JSON format' be placed to apply to all user requests?

   • AIn the system prompt.
   • BIn the response header.
   • CIn the model's training data.
   • DIn the user prompt.
   Reveal Solution Discussion Select an answer first