ML Security PapersC-ReD: A Comprehensive Chinese Benchmark for AI-Generated Text Detection Derived from Real-World Prompts
Chenxi Qing 1, Junxi Wu 2,1, Zheng Liu 1, Yixiang Qiu 1, Hongyao Yu 1, Bin Chen 3,4, Hao Wu 1, Shu-Tao Xia 1,4
Published on arXiv
2604.11796
Output Integrity Attack
OWASP ML Top 10 — ML09
Key Finding
Enables reliable in-domain detection and strong generalization to unseen LLMs on Chinese text
C-ReD
Novel technique introduced
Recently, large language models (LLMs) are capable of generating highly fluent textual content. While they offer significant convenience to humans, they also introduce various risks, like phishing and academic dishonesty. Numerous research efforts have been dedicated to developing algorithms for detecting AI-generated text and constructing relevant datasets. However, in the domain of Chinese corpora, challenges remain, including limited model diversity and data homogeneity. To address these issues, we propose C-ReD: a comprehensive Chinese Real-prompt AI-generated Detection benchmark. Experiments demonstrate that C-ReD not only enables reliable in-domain detection but also supports strong generalization to unseen LLMs and external Chinese datasets-addressing critical gaps in model diversity, domain coverage, and prompt realism that have limited prior Chinese detection benchmarks. We release our resources at https://github.com/HeraldofLight/C-ReD.
Key Contributions
- Comprehensive Chinese AI-generated text detection benchmark with real-world prompts
- Coverage of 9 LLMs across multiple domains addressing model diversity gaps
- Demonstrates strong cross-model generalization to unseen LLMs and external datasets
🛡️ Threat Analysis
Benchmark for detecting AI-generated text content — this is output integrity and content provenance verification. The paper addresses authenticating whether text was generated by an LLM vs human-written.