Beyond Binary: Fine-Grained LLM Content Detection
Recognizing the spectrum of human-AI collaboration in text
This research introduces a novel framework that moves beyond simple binary classification of text as human or AI-generated by recognizing the spectrum of collaboration between humans and LLMs.
- Proposes a role recognition approach that identifies the specific contributions of humans vs. LLMs
- Introduces an involvement measurement methodology to quantify the degree of LLM contribution
- Demonstrates superior performance compared to binary detection methods when evaluating mixed-source content
- Enables more nuanced content moderation that reflects real-world collaboration scenarios
This advancement is crucial for security professionals as it addresses the growing challenge of detecting sophisticated mixed-source content that could spread misinformation or violate platform policies while allowing legitimate collaborative content.