Elamur Research⌕
☰
home / papers / 2026 / Probing Social Identity Bias in Chinese LLMs with Gendered Pronouns and Social Groups

Probing Social Identity Bias in Chinese LLMs with Gendered Pronouns and Social Groups

Paper · 2026 · Geng Liu, Li Feng, Junjie Mu et al.

Summary

The paper asks whether Chinese large language models show social identity bias, meaning they treat an ingroup more favourably and an outgroup more harshly, as earlier work by Hu et al. (2025) found for English models. The authors adapt that method to Mandarin. They use eight sentence-completion starters framed with 我们 ("we") or with one of two third-person plurals: the default 他们 or the feminine-marked 她们. The starters are tested in generic form and combined with 240 social groups relevant in China. Ten base and instruction-tuned models produced nearly 300k completions, which were scored for sentiment and toxicity. The models generally write more positively about "we" than about "they". Base models show especially strong negativity toward outgroups. Instruction tuning narrows sentiment gaps, but toxicity differences tend to remain. Several models also produce more toxic text for 她们 than for 他们. This shows that Mandarin gender marking can expose bias that English-only tests would miss. The work offers a language-aware evaluation framework for Chinese LLMs.

Key points

  • Scale: 297,600 generated responses from ten models, whose developers are DeepSeek, Tencent (Hunyuan), Baidu (ERNIE), 01.AI's Yi, Zhipu AI's GLM, Baichuan and Alibaba's Qwen. Both open checkpoints and API systems were included.
  • Reducing refusals: instruction-tuned models often declined bare starters. To counter this, each prompt began with one neutral sentence from a pool of 2,000 drawn from the ChineseWebText corpus (quality score ≥ 0.9). The same pool was used for every framing condition.
  • Measurement: sentiment labels come from majority voting across three Chinese classifiers, including the Aliyun and Baidu NLP sentiment services. Toxicity is scored as a second, separate dimension.
  • Most affected categories: bias was strongest for "Gender", "Level of Education" and "Nationality".
  • Real-world check: an exploratory analysis of 4,079 Chinese user–assistant exchanges from WildChat found similar ingroup–outgroup patterns. The authors caution that these dialogues come mostly from English-centric models and contain few 她们 cases.

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

🎓University of Science and Technology of ChinaHefei · China108 works · 112,698 citations🎓Politecnico di MilanoMilan · Italy5 works · 258 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksSocial Sciences · 1,772 worksSafety Research · 543 works

🧭 Topics 2Research subjects and languages the work is about.Language models & pretraining · 3,320 worksInterpretability · 130 works

Cite

Cite as Geng Liu, Li Feng, Junjie Mu, Mengxiao Zhu and Francesco Pierri. 2026. Probing Social Identity Bias in Chinese LLMs with Gendered Pronouns and Social Groups. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.1597

@inproceedings{liu2026probing,
  title = {{Probing Social Identity Bias in Chinese LLMs with Gendered Pronouns and Social Groups}},
  author = {Geng Liu, Li Feng, Junjie Mu, Mengxiao Zhu and Francesco Pierri},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1597},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1597}
}

References

31 references, 11 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Marilynn B. Brewer. 1999. The psychology of prejudice: Ingroup love or outgroup hate? Journal of Social Issues, 55(3):429–444. find
  2. Jianghao Chen, Pu Jian, Tengxiao Xi, Dongyi Yi, Qianlong Du, Chenglin Ding, Guibo Zhu, Chengqing Zong, Jinqiao Wang, and Jiajun Zhang. 2023. Chinesewebtext: Large-scale high-quality chinese web text extracted with effective evaluation model. Preprint, arXiv:2311.01149. — [arXiv:2311.01149](https://arxiv.org/abs/2311.01149) find
  3. Zina Chkirbene, Ridha Hamila, Ala Gouissem, and Unal Devrim. 2024. Large language models (llm) in industry: A survey of applications, challenges, and trends. In 2024 IEEE 21st International Conference on Smart Communities: Improving Quality of Life using AI, Robotics and IoT (HONET), pages 229–234. find
  4. Marta Costa-jussà, Pierre Andrews, Eric Smith, Prangthip Hansanti, Christophe Ropers, Elahe Kalbassi, Cynthia Gao, Daniel Licht, and Carleigh Wood. 2023. Multilingual holistic bias: Extending descriptors and patterns to unveil demographic biases in languages at scale. In Proceedings of the 2023 Conf… find
  5. Ameet Deshpande, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, and Karthik Narasimhan. 2023. Toxicity in chatgpt: Analyzing persona-assigned language models. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 1236–1270, Singapore. Association for Computational Ling…
  6. Leon Fröhling, Gianluca Demartini, and Dennis Assenmacher. 2025. Personas with attitudes: Controlling LLMs for diverse data annotation. In Proceedings of the The 9th Workshop on Online Abuse and Harms (WOAH), pages 468–481, Vienna, Austria. Association for Computational Linguistics. find
  7. Isabel O. Gallegos, Ryan A. Rossi, Joe Barrow, Md Mehrab Tanjim, Sungchul Kim, Franck Dernoncourt, Tong Yu, Ruiyi Zhang, and Nesreen K. Ahmed. 2024. Bias and fairness in large language models: A survey. Computational Linguistics, 50(3):1097– 1179.
  8. Deep Ganguli, Liane Lovitt, Jackson Kernion, Amanda Askell, Yuntao Bai, Saurav Kadavath, Ben Mann, Ethan Perez, Nicholas Schiefer, Kamal Ndousse, and 1 others. 2022. Red teaming language models to reduce harms: Methods, scaling behaviors, and lessons learned. arXiv preprint arXiv:2209.07858.
  9. Miles Hewstone, Mark Rubin, and Hazel Willis. 2002. Intergroup bias. Annual Review of Psychology, 53:575–604. find
  10. Tiancheng Hu, Yara Kyrychenko, Steve Rathje, Nigel Collier, Sander van der Linden, and Jon Roozenbeek. 2025. Generative language models exhibit social identity biases. Nature Computational Science, 5(1):65–75. find
  11. C-T James Huang, Y-H Audrey Li, and Yafei Li. 2009. The syntax of Chinese. Cambridge University Press. find
  12. Leilei Jiang, Guixiang Zhu, Jianshan Sun, Jie Cao, and Jia Wu. 2025. Exploring the occupational biases and stereotypes of chinese large language models. Scientific Reports, 15(1):18777. find
  13. Hannah Rose Kirk, Bertie Vidgen, Paul Röttger, and Scott A Hale. 2024. The benefits, risks and bounds of personalizing the alignment of large language models to individuals. Nature Machine Intelligence, 6(4):383–392. find
  14. Keita Kurita, Nidhi Vyas, Ayush Pareek, Alan W Black, and Yulia Tsvetkov. 2019. Measuring bias in contextualized word representations. In Proceedings of the First Workshop on Gender Bias in Natural Language Processing, pages 166–172, Florence, Italy. Association for Computational Linguistics.
  15. Michael Lepori. 2020. Unequal representations: Analyzing intersectional biases in word embeddings using representational similarity analysis. In Proceedings of the 28th International Conference on Computational Linguistics, pages 1720–1728, Barcelona, Spain (Online). International Committee on Compu… find
  16. Charles N Li and Sandra A Thompson. 1989. Mandarin Chinese: A functional reference grammar. Univ of California Press. find
  17. Yanyang Li, Jianqiao Zhao, Duo Zheng, Zi-Yuan Hu, Zhi Chen, Xiaohui Su, Yongfeng Huang, Shijia Huang, Dahua Lin, Michael Lyu, and Liwei Wang. 2023. CLEVA: Chinese language models EVAluation platform. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing: System De…
  18. Geng Liu, Carlo Alberto Bono, and Francesco Pierri. 2025a. Comparing diversity, negativity, and stereotypes in chinese-language ai technologies: an investigation of baidu, ernie and qwen. PeerJ Computer Science, 11:e2694. find
  19. Geng Liu, Li Feng, Carlo Alberto Bono, Songbo Yang, Mengxiao Zhu, and Francesco Pierri. 2025b. Evaluating prompt-driven chinese large language models: The influence of persona assignment on stereotypes and safeguards. arXiv preprint arXiv:2506.04975. — [arXiv:2506.04975](https://arxiv.org/abs/2506.04975) find
  20. Chandler May, Alex Wang, Shikha Bordia, Samuel R. Bowman, and Rachel Rudinger. 2019. On measuring social biases in sentence encoders. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and…
  21. Bonan Min, Hayley Ross, Elior Sulem, Amir Pouran Ben Veyseh, Thien Huu Nguyen, Oscar Sainz, Eneko Agirre, Ilana Heintz, and Dan Roth. 2023. Recent advances in natural language processing via large pre-trained language models: A survey. ACM Computing Surveys, 56(2):1–40.
  22. Moin Nadeem, Anna Bethke, and Siva Reddy. 2021. StereoSet: Measuring stereotypical bias in pretrained language models. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long…
  23. Nikita Nangia, Clara Vania, Rasika Bhalerao, and Samuel R. Bowman. 2020. CrowS-pairs: A challenge dataset for measuring social biases in masked language models. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1953–1967, Online. Association for…
  24. Ali Omrani, Alireza Salkhordeh Ziabari, Charles Yu, Preni Golazizian, Brendan Kennedy, Mohammad Atari, Heng Ji, and Morteza Dehghani. 2023. Social-group-agnostic bias mitigation via the stereotype content model. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguisti… find
  25. Alicia Parrish, Angelica Chen, Nikita Nangia, Vishakh Padmakumar, Jason Phang, Jana Thompson, Phu Mon Htut, and Samuel Bowman. 2022. BBQ: A hand-built bias benchmark for question answering. In Findings of the Association for Computational Linguistics: ACL 2022, pages 2086–2105, Dublin, Ireland. Asso…
  26. Mohaimenul Azam Khan Raiaan, Md. Saddam Hossain Mukta, Kaniz Fatema, Nur Mohammad Fahad, Sadman Sakib, Most Marufatul Jannat Mim, Jubaer Ahmad, Mohammed Eunus Ali, and Sami Azam. 2024. A review on large language models: Architectures, applications, taxonomies, open issues and challenges. IEEE Access… find
  27. Mubashar Raza, Zarmina Jahangir, Muhammad Bilal Riaz, Muhammad Jasim Saeed, and Muhammad Awais Sattar. 2025. Industrial applications of large language models. Scientific Reports, 15(1):13755. find
  28. Henri Tajfel and John C Turner. 2004. The social identity theory of intergroup behavior. In Political psychology, pages 276–293. Psychology Press. find
  29. Yuemei Xu, Ling Hu, Jiayi Zhao, Zihan Qiu, Kexin Xu, Yuqi Ye, and Hanwen Gu. 2025. A survey on multilingual large language models: Corpora, alignment, and bias. Frontiers of Computer Science, 19(11):1911362. find
  30. Ge Zhang, Yizhi Li, Yaoyao Wu, Linyuan Zhang, Chenghua Lin, Jiayi Geng, Shi Wang, and Jie Fu. 2023. Corgi-pm: A chinese corpus for gender bias probing and mitigation. arXiv preprint. find
  31. Wenting Zhao, Xiang Ren, Jack Hessel, Claire Cardie, Yejin Choi, and Yuntian Deng. 2024. Wildchat: 1m chatgpt interaction logs in the wild. arXiv preprint arXiv:2405.01470.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/liu-2026-probing-social-identity-bias-chinese/