Beemo introduces the first benchmark for detecting hybrid authorship - content initially generated by AI but later edited by humans.

• Addresses a critical gap in MGT (Machine-Generated Text) detection by focusing on multi-author scenarios
• Provides datasets from security, linguistics, creative writing, and educational domains
• Creates a new evaluation framework more aligned with real-world usage patterns
• Demonstrates that current detection tools struggle with hybrid content

This research is crucial for security applications as hybrid texts represent a more sophisticated challenge for content authentication systems, potentially enabling more convincing misrepresentations or academic dishonesty that evade current detection methods.

Beemo: Benchmark of Expert-edited Machine-generated Outputs