Beyond Binary: Towards Fine-Grained LLM-Generated Text Detection via Role Recognition and Involvement Measurement

TL;DR

This work addresses the limitation of binary LLM-generated text detection in real-world scenarios involving human–LLM collaboration. It introduces LLM Role Recognition (LLM-RR) and LLM Involvement Measurement (LLM-IM) and presents the LLMDetect benchmark, consisting of the Hybrid News Detection Corpus (HNDC) and DetectEval, to evaluate detectors under cross-context and multi-intensity variations. Experiments with 10 baselines show fine-tuned PLM-based detectors outperform zero-shot approaches, with DeBERTa excelling in cross-context generalization and Longformer handling varying intensity levels; however, advanced LLMs still struggle to detect content they generate. The findings highlight data-leakage risks for zero-shot detectors and demonstrate that role- and involvement-aware detection offers a more robust approach for maintaining content integrity on social platforms.

Abstract

The rapid development of large language models (LLMs), like ChatGPT, has resulted in the widespread presence of LLM-generated content on social media platforms, raising concerns about misinformation, data biases, and privacy violations, which can undermine trust in online discourse. While detecting LLM-generated content is crucial for mitigating these risks, current methods often focus on binary classification, failing to address the complexities of real-world scenarios like human-LLM collaboration. To move beyond binary classification and address these challenges, we propose a new paradigm for detecting LLM-generated content. This approach introduces two novel tasks: LLM Role Recognition (LLM-RR), a multi-class classification task that identifies specific roles of LLM in content generation, and LLM Influence Measurement (LLM-IM), a regression task that quantifies the extent of LLM involvement in content creation. To support these tasks, we propose LLMDetect, a benchmark designed to evaluate detectors' performance on these new tasks. LLMDetect includes the Hybrid News Detection Corpus (HNDC) for training detectors, as well as DetectEval, a comprehensive evaluation suite that considers five distinct cross-context variations and two multi-intensity variations within the same LLM role. This allows for a thorough assessment of detectors' generalization and robustness across diverse contexts. Our empirical validation of 10 baseline detection methods demonstrates that fine-tuned PLM-based models consistently outperform others on both tasks, while advanced LLMs face challenges in accurately detecting their own generated content. Our experimental results and analysis offer insights for developing more effective detection models for LLM-generated content. This research enhances the understanding of LLM-generated content and establishes a foundation for more nuanced detection methodologies.

Figures (10)

• Figure 1: The detection framework toward fine-grained LLM-generated text detection through role recognition and involvement measurement.
• Figure 2: t-SNE visualization of representations from three non-fine-tuned PLMs on the HNDC test data. DeBERTa shows clearer cluster separation, reflecting stronger discriminative ability.
• Figure 3: Average LLM Involvement Ratio Predictions and Golden Label of Variable-Length Extension Experiments
• Figure 4: Average LLM Involvement Ratio Predictions and Golden Label of Multi-Staged Polish Experiments
• Figure 5: Comparison of Confusion Matrices of LLM-RR Task