Problem
Research questions and friction points this paper is trying to address.
Chinese LLM-generated text detection
out-of-domain inputs
adversarial samples
reliability
model ensemble
EnsemJudge: Enhancing Reliability in Chinese LLM-Generated Text Detection through Diverse Model Ensembles
🤖 AI Summary
This work addresses the limited robustness of existing methods for detecting machine-generated Chinese text, particularly in out-of-domain and adversarial settings, an area predominantly studied in English contexts. To bridge this gap, we propose the first Chinese-specific ensemble learning framework that integrates a multi-model voting mechanism with Chinese-pretrained language models. Trained and evaluated on the NLPCC2025 Shared Task dataset, our approach significantly enhances detection stability and accuracy under distribution shifts and adversarial conditions. It achieves state-of-the-art performance by outperforming all baselines and securing first place in NLPCC2025 Shared Task 1, thereby demonstrating its effectiveness and superiority in real-world Chinese text detection scenarios.
📝 Abstract
Large Language Models (LLMs) are widely applied across various domains due to their powerful text generation capabilities. While LLM-generated texts often resemble human-written ones, their misuse can lead to significant societal risks. Detecting such texts is an essential technique for mitigating LLM misuse, and many detection methods have shown promising results across different datasets. However, real-world scenarios often involve out-of-domain inputs or adversarial samples, which can affect the performance of detection methods to varying degrees. Furthermore, most existing research has focused on English texts, with limited work addressing Chinese text detection. In this study, we propose EnsemJudge, a robust framework for detecting Chinese LLM-generated text by incorporating tailored strategies and ensemble voting mechanisms. We trained and evaluated our system on a carefully constructed Chinese dataset provided by NLPCC2025 Shared Task 1. Our approach outperformed all baseline methods and achieved first place in the task, demonstrating its effectiveness and reliability in Chinese LLM-generated text detection. Our code is available at https://github.com/johnsonwangzs/MGT-Mini.
Innovation
Methods, ideas, or system contributions that make the work stand out.
model ensemble
Chinese LLM-generated text detection
robustness
adversarial samples
out-of-domain detection
🔎 Similar Papers
No similar papers found.
