C-ReD: A Comprehensive Chinese Benchmark for AI-Generated Text Detection Derived from Real-World Prompts

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.

📄 Paper 💻 Dataset

🔥 Overview

C-ReD is a large-scale Chinese benchmark dataset for AI-generated text detection, designed to evaluate model robustness across:

• Multiple domains (News, Q&A, Film Review, Academic Writing, Composition)
• Multiple LLMs (ChatGPT, Qwen, DeepSeek, Claude, etc.)
• Multiple real-world prompt types
• Both in-domain and out-of-domain evaluation settings

Unlike prior datasets, C-ReD emphasizes:

• 🧠 Real-world prompt simulation
• 🌐 Strong generator diversity (including Chinese LLMs)
• 📚 Multi-domain realistic text distribution
• 🔄 Cross-domain & cross-model generalization

---

📊 Dataset Statistics

• Total samples: 128,610
  • Human-written: 12,997
  • AI-generated: 115,613
• Domains: 5 core domains
• Generators: 9 state-of-the-art LLMs
• Unified schema across all samples

Example Schema

{
  "id": 1,
  "text": "...",
  "label": 0,
  "domain": "news",
  "generator": "GPT-4o",
  "prompt": "...",
  "length": 512
}

---

🧪 Test Script & Code Sources

Our evaluation and baseline implementations are partially built upon the following open-source repositories:

• ImBD
• Lastde Detector
• DNA-DetectLLM
• LAPD

We sincerely thank the authors for releasing their code.

---

📜 Citation

If you find our work useful, please cite:

@inproceedings{qing2026CReD,
  title={{C}-{R}e{D}: A Comprehensive Chinese Benchmark for {AI}-Generated Text Detection Derived from Real-World Prompts},
  author={Qing, Chenxi and Wu, Junxi and Liu, Zheng and Qiu, Yixiang and Yu, Hongyao and Chen, Bin and Wu, Hao and Xia, Shu-Tao},
  booktitle={Findings of the Association for Computational Linguistics: ACL 2026},
  pages={42703--42733},
  year={2026}
}