ExaGPT introduces a novel approach for detecting LLM-generated text that provides clear explanations humans can understand and verify.

• Uses example-based comparison to identify text similarity patterns between suspect text and known samples
• Provides interpretable results by highlighting specific text spans that influenced the detection decision
• Achieves detection accuracy comparable to existing methods while offering transparency
• Helps protect academic integrity by enabling educators to explain detection results to students

This research addresses critical concerns in education by providing a tool that not only detects AI-generated content but also explains its reasoning—reducing false accusations and helping educators maintain fair assessment environments.

ExaGPT: Example-Based Machine-Generated Text Detection for Human Interpretability