petergpt

A benchmark tool that tests whether AI models can detect and challenge nonsensical prompts instead of confidently answering invalid questions. It evaluates models across multiple domains using 100 carefully crafted nonsense questions.