Which statement best describes prompt engineering compared to fine-tuning?

• A. It updates model weights to change behavior.
• B. It designs inputs to elicit desired outputs without changing weights.
• C. It discards the original model and uses a new one.
• D. It trains a separate classifier on top of the model.

Answer: (B)

Prompt engineering is about shaping the inputs you give to a model so it produces the outputs you want without changing the model’s internal parameters. By crafting prompts, adding clear instructions, or including examples within the prompt, you steer the model’s behavior while keeping its weights fixed. This makes tweaking and testing different tasks quick and portable across models that share the same architecture.

Fine-tuning, in contrast, updates the model’s internal parameters using task-specific data to alter its behavior more directly. It requires data, training time, and deployment changes, but can yield more consistent results for a given task because the model itself learns from the new information.

The other descriptions imply changing the model’s structure or training in addition to prompting. Replacing the model or using an extra classifier on top goes beyond prompt design and involves separate training or components.