Elamur Research⌕
☰
home / papers / 2023 / CValues: Measuring the Values of Chinese Large Language Models from Safety to Responsibility

CValues: Measuring the Values of Chinese Large Language Models from Safety to Responsibility

Paper · 2023 · Guohai Xu, Jiayi Liu, Ming Yan et al.

Summary

Existing Chinese LLM benchmarks such as C-EVAL, M3KE and GAOKAO-Bench mostly test knowledge and reasoning. They largely ignore whether models behave in line with human values. CValues addresses this gap with a two-tier framework for Chinese. The basic tier, safety, requires that answers be free of harmful content. The higher tier, responsibility, asks whether a model gives constructive guidance and humane care while considering its wider social impact.

The prompts were gathered with experts in the loop. Crowdworkers tried to break an early ChatPLUG chatbot, and the questions that succeeded became safety prompts. Domain specialists, recruited through the "100 Bottles of Poison for AI" initiative, wrote deliberately tricky responsibility questions. Models were then judged two ways: by trained annotators, and automatically through two-option multiple-choice items.

Most Chinese LLMs proved fairly safe but clearly weaker on responsibility. The two evaluation modes also diverged. Multiple-choice tests mainly capture whether a model recognizes bad behavior, while human review captures what it actually generates. The authors therefore recommend using both.

Key points

  • Human-evaluation set: 2,100 prompts in total.
    • 1,300 safety prompts from crowdworkers, covering 10 scenarios (an extension of the Sun et al. 2023 taxonomy).
    • 800 responsibility prompts from 8 experts (100 each), in domains such as law, psychology and environmental science.
  • Automatic-evaluation set: 4,312 multiple-choice prompts. They were built by generating LLM responses, having humans label them safe or unsafe, rewriting them with ChatGPT, and combining them into templated two-option questions.
  • Human labeling: each response was rated by three specialized annotators, and the final label was decided by majority vote.
  • Ziya-LLaMA case: the model scored high on multiple-choice accuracy but low in human evaluation. This illustrates that recognizing harmful content does not guarantee avoiding it.
  • Released resources (ModelScope and GitHub):
    • The benchmark and code.
    • CValues-Comparison: 145k prompts with paired positive and negative responses.
    • 100PoisonMpts: expert questions and answers, described as the first Chinese AI-governance dataset.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🏢Alibaba GroupHangzhou · China64 works · 14,113 citations🎓Beijing Jiaotong UniversityBeijing · China6 works · 1,570 citations

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 9,086 worksArtificial Intelligence · 8,889 worksArts and Humanities · 9,296 worksLanguage and Linguistics · 9,206 worksSocial Sciences · 744 worksSafety Research · 221 works

🧭 Topics 2Research subjects and languages the work is about.Language models & pretraining · 1,343 worksBias, fairness & safety · 209 works

Cite

Cite as Guohai Xu, Jiayi Liu, Ming Yan, Haotian Xu, Jinghui Si, Zhuoran Zhou, Peng Yi, Xing Gao, Jitao Sang, Rong Zhang, Ji Zhang, Chao Peng, Fei Huang and Jingren Zhou. 2023. CValues: Measuring the Values of Chinese Large Language Models from Safety to Responsibility. arXiv:2307.09705.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2307.09705 · arXiv 2307.09705

@misc{xu2023cvalues,
  title = {{CValues: Measuring the Values of Chinese Large Language Models from Safety to Responsibility}},
  author = {Guohai Xu, Jiayi Liu, Ming Yan, Haotian Xu, Jinghui Si, Zhuoran Zhou, Peng Yi, Xing Gao, Jitao Sang, Rong Zhang, Ji Zhang, Chao Peng, Fei Huang and Jingren Zhou},
  year = {2023},
  doi = {10.48550/arxiv.2307.09705},
  eprint = {2307.09705},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2307.09705}
}

References

34 references, 21 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. BelleGroup. 2023. Belle. LianjiaTech/BELLE. https://github.com/ ↗
  2. Ondřej Bojar, Christian Buck, Christian Federmann, Barry Haddow, Philipp Koehn, Johannes Leveling, Christof Monz, Pavel Pecina, Matt Post, Herve Saint-Amand, et al. 2014. Findings of the 2014 workshop on statistical machine translation. In Proceedings of the ninth workshop on statistical machine tr…
  3. Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter,…
  4. Aakanksha Chowdhery, Sharan Narang, Jacob Devlin, Maarten Bosma, Gaurav Mishra, Adam Roberts, Paul Barham, Hyung Won Chung, Charles Sutton, Sebastian Gehrmann, Parker Schuh, Kensen Shi, Sasha Tsvyashchenko, Joshua Maynez, Abhishek Rao, Parker Barnes, Yi Tay, Noam Shazeer, Vinodkumar Prabhakaran, Emi…
  5. Paul F Christiano, Jan Leike, Tom Brown, Miljan Martic, Shane Legg, and Dario Amodei. 2017. Deep reinforcement learning from human preferences. Advances in neural information processing systems, 30. find
  6. Yiming Cui, Ziqing Yang, and Xin Yao. 2023. Efficient and effective text encoding for chinese llama and alpaca. CoRR, abs/2304.08177.
  7. Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2021. Measuring massive multitask language understanding. In 9th International Conference on Learning Representations, ICLR 2021, Virtual Event, Austria, May 3-7, 2021. OpenReview.net.
  8. Karl Moritz Hermann, Tomas Kocisky, Edward Grefenstette, Lasse Espeholt, Will Kay, Mustafa Suleyman, and Phil Blunsom. 2015. Teaching machines to read and comprehend. Advances in neural information processing systems, 28.
  9. Yuzhen Huang, Yuzhuo Bai, Zhihao Zhu, Junlei Zhang, Jinghan Zhang, Tangjun Su, Junteng Liu, Chuancheng Lv, Yikai Zhang, Jiayi Lei, Yao Fu, Maosong Sun, and Junxian He. 2023. C-eval: A multi-level multi-discipline chinese evaluation suite for foundation models. CoRR, abs/2305.08322.
  10. Percy Liang, Rishi Bommasani, Tony Lee, Dimitris Tsipras, Dilara Soylu, Michihiro Yasunaga, Yian Zhang, Deepak Narayanan, Yuhuai Wu, Ananya Kumar, Benjamin Newman, Binhang Yuan, Bobby Yan, Ce Zhang, Christian Cosgrove, Christopher D. Manning, Christopher Ré, Diana Acosta-Navas, Drew A. Hudson, Eric… find
  11. Chuang Liu, Renren Jin, Yuqi Ren, Linhao Yu, Tianyu Dong, Xiaohan Peng, Shuting Zhang, Jianxiang Peng, Peiyi Zhang, Qingqing Lyu, Xiaowen Su, Qun Liu, and Deyi Xiong. 2023. M3KE: A massive multi-level multi-subject knowledge evaluation benchmark for chinese large language models. CoRR, abs/2305.1026… find
  12. ModelScope. 2021. PLUG: Pre-training for Language Understanding and Generation. https://modelscope.cn/models/damo/nlp_plug_text-generation_27B/summary. ↗
  13. OpenAI. 2022. Chatgpt: Optimizing language models for dialogue. OpenAI Blog. find
  14. OpenAI. 2023. GPT-4 technical report. abs/2303.08774.
  15. CoRR, OpenLMLab. 2023. Moss. https://github.com/OpenLMLab/MOSS. ↗
  16. Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. 2022. Training language models to follow instructions with human feedback. Advances in Neural Information Processing Systems, 35:27730–27744.
  17. Pranav Rajpurkar, Jian Zhang, Konstantin Lopyrev, and Percy Liang. 2016. Squad: 100,000+ questions for machine comprehension of text. arXiv preprint arXiv:1606.05250.
  18. Aarohi Srivastava, Abhinav Rastogi, Abhishek Rao, Abu Awal Md Shoeb, Abubakar Abid, Adam Fisch, Adam R. Brown, Adam Santoro, Aditya Gupta, Adrià Garriga-Alonso, Agnieszka Kluska, Aitor Lewkowycz, Akshat Agarwal, Alethea Power, Alex Ray, Alex Warstadt, Alexander W. Kocurek, Ali Safaya, Ali Tazarv, Al…
  19. Hao Sun, Guangxuan Xu, Jiawen Deng, Jiale Cheng, Chujie Zheng, Hao Zhou, Nanyun Peng, Xiaoyan Zhu, and Minlie Huang. 2022. On the safety of conversational models: Taxonomy, dataset, and benchmark. In Findings of the Association for Computational Linguistics: ACL 2022, Dublin, Ireland, May 22-27, 202…
  20. Hao Sun, Zhexin Zhang, Jiawen Deng, Jiale Cheng, and Minlie Huang. 2023. Safety assessment of chinese large language models. CoRR, abs/2304.10436. find
  21. Yu Sun, Shuohuan Wang, Shikun Feng, Siyu Ding, Chao Pang, Junyuan Shang, Jiaxiang Liu, Xuyi Chen, Yanbin Zhao, Yuxiang Lu, et al. 2021. Ernie 3.0: Large-scale knowledge enhanced pre-training for language understanding and generation. arXiv preprint arXiv:2107.02137.
  22. THUDM. 2023. Chatglm. THUDM/ChatGLM-6B. https://github.com/ ↗
  23. Junfeng Tian, Hehong Chen, Guohai Xu, Ming Yan, Xing Gao, Jianhai Zhang, Chenliang Li, Jiayi Liu, Wenshen Xu, Haiyang Xu, Qi Qian, Wei Wang, Qinghao Ye, Jiejing Zhang, Ji Zhang, Fei Huang, and Jingren Zhou. 2023. Chatplug: Open-domain generative dialogue system with internet-augmented instruction tu… — [arXiv:2304.07849](https://arxiv.org/abs/2304.07849) find
  24. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurélien Rodriguez, Armand Joulin, Edouard Grave, and Guillaume Lample. 2023. Llama: Open and efficient foundation language models. CoRR, ab…
  25. Alex Wang, Amanpreet Singh, Julian Michael, Felix Hill, Omer Levy, and Samuel R Bowman. 2018. Glue: A multi-task benchmark and analysis platform for natural language understanding. arXiv preprint arXiv:1804.07461.
  26. Yidong Wang, Zhuohao Yu, Zhengran Zeng, Linyi Yang, Cunxiang Wang, Hao Chen, Chaoya Jiang, Rui Xie, Jindong Wang, Xing Xie, Wei Ye, Shikun Zhang, and Yue Zhang. 2023a. Pandalm: An automatic evaluation benchmark for LLM instruction tuning optimization. CoRR, abs/2306.05087.
  27. Yizhong Wang, Yeganeh Kordi, Swaroop Mishra, Alisa Liu, Noah A. Smith, Daniel Khashabi, and Hannaneh Hajishirzi. 2023b. Self-instruct: Aligning language models with self-generated instructions. find
  28. Aohan Zeng, Xiao Liu, Zhengxiao Du, Zihan Wang, Hanyu Lai, Ming Ding, Zhuoyi Yang, Yifan Xu, Wendi Zheng, Xiao Xia, Weng Lam Tam, Zixuan Ma, Yufei Xue, Jidong Zhai, Wenguang Chen, Zhiyuan Liu, Peng Zhang, Yuxiao Dong, and Jie Tang. 2023. GLM-130B: an open bilingual pre-trained model. In The Eleventh…
  29. Hui Zeng. 2023. Measuring massive multitask chinese understanding. CoRR, abs/2304.12986. find
  30. Wei Zeng, Xiaozhe Ren, Teng Su, Hui Wang, Yi Liao, Zhiwei Wang, Xin Jiang, ZhenZhang Yang, Kaisheng Wang, Xiaoda Zhang, Chen Li, Ziyan Gong, Yifan Yao, Xinjing Huang, Jun Wang, Jianfeng Yu, Qi Guo, Yue Yu, Yan Zhang, Jin Wang, Hengtao Tao, Dasen Yan, Zexuan Yi, Fang Peng, Fangqing Jiang, Han Zhang,…
  31. Jiaxing Zhang, Ruyi Gan, Junjie Wang, Yuxiang Zhang, Lin Zhang, Ping Yang, Xinyu Gao, Ziwei Wu, Xiaoqun Dong, Junqing He, Jianheng Zhuo, Qi Yang, Yongfeng Huang, Xiayu Li, Yanghan Wu, Junyu Lu, Xinyu Zhu, Weifeng Chen, Ting Han, Kunhao Pan, Rui Wang, Hao Wang, Xiaojun Wu, Zhongshen Zeng, and Chongpe… — [arXiv:2209.02970](https://arxiv.org/abs/2209.02970) find
  32. Xiaotian Zhang, Chunyang Li, Yi Zong, Zhengyu Ying, Liang He, and Xipeng Qiu. 2023. Evaluating the performance of large language models on GAOKAO benchmark. CoRR, abs/2305.12474.
  33. Wanjun Zhong, Ruixiang Cui, Yiduo Guo, Yaobo Liang, Shuai Lu, Yanlin Wang, Amin Saied, Weizhu Chen, and Nan Duan. 2023a. Agieval: A human-centric benchmark for evaluating foundation models. CoRR, abs/2304.06364.
  34. Wanjun Zhong, Ruixiang Cui, Yiduo Guo, Yaobo Liang, Shuai Lu, Yanlin Wang, Amin Saied, Weizhu Chen, and Nan Duan. 2023b. Agieval: A human-centric benchmark for evaluating foundation models. arXiv preprint arXiv:2304.06364.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/xu-2023-cvalues-measuring-values-chinese/