Prompt Engineering & ChatGPT Techniques

Which principle from the "Guidelines for Prompting" notebook emphasises allocating enough reasoning time to the model?

• Write clear and specific instructions

• Use examples in prompts

• Give the model time to "think"

• Keep temperature low

In the recommended tactics, wrapping user text inside triple-backticks is an example of:

• Few-shot prompting

• Asking for step-by-step reasoning

• Providing chain-of-thought

• Using delimiters to separate parts of the input

In the Chat format, which role defines the assistant's overall behaviour or persona?

• User

• Assistant

• Function

• System

The "OrderBot" demo is designed primarily to:

• Translate menu items into Spanish

• Take pizza orders from customers

• Classify customer sentiment

• Summarise delivery policies

Showing the model a handful of input-output pairs to steer its replies is called:

• Zero-shot prompting

• Temperature sampling

• Few-shot prompting

• Chain-of-thought suppression

The "Inferring" notebook demonstrates how to extract ___ from product reviews.

• Shipping addresses

• Sentiments and topical categories

• SKU identifiers

• Word counts only

Which explicit instruction signals the model to perform grammar correction?

• "Rephrase eloquently"

• "Translate to English"

• "Proofread and correct"

• "Infer topics"

Requesting output as JSON or HTML illustrates which tactic?

• Let the model explore freely

• Ask for a structured output

• Raise the temperature parameter

• Give the model negative examples

When refining marketing copy, limiting the number of words addresses which common issue listed in the notebook?

• Wrong tone

• Missing table of dimensions

• Focus on wrong details

• Text is too long

Switching the verb from "summarise" to "extract" in a prompt helps the model to:

• Generate code snippets

• Return only the specific information requested

• Increase creativity

• Translate idioms accurately