Answer: Few-shot Prompting

## Explanation **Few-shot Prompting** is the correct answer because: 1. **Definition**: Few-shot prompting involves providing the model with a small number of examples (typically 2-10) before asking it to perform a task on new input. 2. **Key characteristic**: The developer provides "two example inputs and outputs before the main query" - this is the defining feature of few-shot prompting. 3. **Contrast with other options**: - **Zero-shot Prompting**: No examples are provided; the model must understand the task from the prompt alone. - **Negative Prompting**: Typically refers to specifying what NOT to include in generated content, not providing examples. - **Guided Prompting**: Not a standard term in prompt engineering; may refer to structured prompting but doesn't specifically mean providing examples. 4. **Application**: In text classification tasks like sentiment analysis (positive/negative reviews), few-shot prompting helps the model understand the desired format and criteria by showing concrete examples. **Why this approach works**: By providing examples, the model learns the pattern and can apply similar reasoning to new, unseen text reviews, improving accuracy compared to zero-shot approaches.

Author: Jin H