Define prompt diversity and explain why it matters for robust evaluation.

• A. Prompt diversity is using the same prompt across all tests to maintain consistency.
• B. Prompt diversity means using many different phrasings, structures, and formats; it helps ensure performance is not tied to a single prompt style and improves generalization.
• C. Prompt diversity is about varying the length of a single prompt only.
• D. Prompt diversity focuses on changing model architectures.

Answer: (B)

Prompt diversity means using many different phrasings, structures, and formats when evaluating a model. The goal is to avoid judging performance based on a single prompt style and to test whether the model truly understands the task rather than just matching a particular wording.

This matters for robust evaluation because prompts can strongly influence a model’s output. If you only test with one prompt, you might overestimate the model’s abilities if it happens to respond well to that wording, or miss weaknesses that appear with paraphrased or differently structured prompts. By presenting a variety of prompts, you check for consistent performance across linguistic variations, ensure the results reflect genuine capability and generalization, and reveal where the model is brittle or prompt-sensitive. In practice, including diverse prompts—varying phrasing, sentence structure, and formats such as instructions, questions, or step-by-step prompts—leads to more trustworthy assessments and better guidance for real-world use.