What is the iterative art of designing inputs to get the best possible results from an LLM?

Get ready for the GARP Risk and AI Exam with flashcards and multiple choice questions. Each question comes with hints and explanations. Prepare for success!

Multiple Choice

What is the iterative art of designing inputs to get the best possible results from an LLM?

Prompt engineering is the practice of crafting and iteratively refining prompts to elicit the best possible outputs from an LLM. It’s about shaping the task in the prompt—the instructions, context, format, and constraints—so the model understands what you want and how you want it delivered. The iterative part matters: you test a prompt, check the results, adjust wording, add or remove examples, reorder information, tweak tone or length, and repeat until the answers align with your goals. This approach also includes deciding what kind of guidance to provide (for example, asking for bullet points, a step-by-step process, or a specific style) and setting boundaries to reduce errors or bias. Within prompt engineering, you’ll encounter techniques like few-shot prompting, where you show a few examples to demonstrate the desired format or reasoning, or chain-of-thought prompting, which prompts the model to explain its reasoning steps. Zero-shot prompting uses no examples at all. Those are useful tools, but the broader practice—iteratively designing and refining prompts to steer performance across tasks—is what “prompt engineering” captures.

Prompt engineering is the practice of crafting and iteratively refining prompts to elicit the best possible outputs from an LLM. It’s about shaping the task in the prompt—the instructions, context, format, and constraints—so the model understands what you want and how you want it delivered. The iterative part matters: you test a prompt, check the results, adjust wording, add or remove examples, reorder information, tweak tone or length, and repeat until the answers align with your goals. This approach also includes deciding what kind of guidance to provide (for example, asking for bullet points, a step-by-step process, or a specific style) and setting boundaries to reduce errors or bias.

Within prompt engineering, you’ll encounter techniques like few-shot prompting, where you show a few examples to demonstrate the desired format or reasoning, or chain-of-thought prompting, which prompts the model to explain its reasoning steps. Zero-shot prompting uses no examples at all. Those are useful tools, but the broader practice—iteratively designing and refining prompts to steer performance across tasks—is what “prompt engineering” captures.

More practice questions