What is the primary goal of the prompt engineering exercise described in the transcript?	To improve an existing prompt to extract key topics from a scholarly article passage into a JSON array of strings.
What specific data structure must the prompt output?	A JSON array of strings containing all the topics mentioned in the text.
What was the first major technique used to improve a poorly performing prompt?	Being simple and direct, explicitly telling the model exactly what output is required (e.g., "extract key topics... into a JSON array of strings").
What prompt engineering technique was used to improve the connection between the instruction and the input data?	Structuring the prompt using XML tags (e.g., `<text>`) around the content being interpolated.
What is the benefit of using XML tags in prompt engineering?	It creates a clearer connection between the instruction (talking about a passage of text) and the provided text data.
What technique was added to ensure the model only returns the required JSON array and no other commentary?	Providing a series of specific, step-by-step instructions, including a constraint like "Do not provide any other text or commentary."
What advanced prompting techniques were mentioned as potential additions to the prompt?	One-shot or multi-shot prompting.