Elamur Research⌕
☰
home / papers / 2026 / Know Your Place: Diagnosing Implicit Social Adaptation Failures in Chinese Large Language Models

Know Your Place: Diagnosing Implicit Social Adaptation Failures in Chinese Large Language Models

Paper · 2026 · Yu Tian, Jie Xing, Ziming Li et al.

Summary

The paper asks whether Chinese LLMs adapt their conversational style to social signals that are only implied in a user's message, rather than stated in an instruction. The authors call the observed shortfall Social Agnosia. They use the term to describe behavior, not internal cognition: models overlook hints about the interlocutor's power, emotional intensity and expertise. To measure this, they build C-ISA, a diagnostic benchmark based on Communication Accommodation Theory. Three roughly independent binary dimensions (Social Power, Affective Arousal and Epistemic Status) define eight social archetypes. Test items are derived from neutral seed queries through a two-model generation-and-checking pipeline. Several LLMs are compared with and without explicit social instructions. When given explicit cues, models shift strategy appropriately. Without them, their answers become uniform and tone-deaf, especially toward high-power speakers, where they over-explain and hedge. The same gap appears in English-centric models, which suggests the cause lies in alignment practices rather than in the language. The findings matter for dialogue agents deployed in high-context cultures.

Key points

  • C-ISA contains 4,000 test instances covering a 2 × 2 × 2 matrix (P, A, K) of eight archetypes. They are built from 500 neutral seed queries using a Dual-Model Generation–Verification Loop. The dataset is released at github.com/ty373/C-ISA.
  • In high-power settings, models tend toward a default the authors label Defensive Submission. It shows up as longer replies, more negations and avoidance of simply doing the task. Explicit conditioning moves models toward short, execution-first answers.
  • Breaking the failure down by dimension, the gap is smallest for Epistemic Status and largest for Power and Affective Arousal. The authors read this as a sign that training data under-covers asymmetric, high-pressure interactions.
  • In the S4 "Arrogant Boss" scenario, Distinct-2 rises from implicit to explicit conditions:
    • Qwen2.5-72B: 0.436 → 0.707
    • Llama-3.1-70B-Instruct: 0.534 → 0.596
    • Grok-4: 0.436 → 0.489
  • Scoring relies on LLM-as-a-Judge, so the authors draw conclusions only from relative contrasts under the same judge, not from absolute scores.

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

🎓Inner Mongolia UniversityHohhot · China3 works · 1 citation

 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 1Research subjects and languages the work is about.Language models & pretraining · 3,320 works

Cite

Cite as Yu Tian, Jie Xing, Ziming Li, Jiang Li, Zehua Duo, Tian Lan, Xu Liu, Guanglai Gao and Xiangdong Su. 2026. Know Your Place: Diagnosing Implicit Social Adaptation Failures in Chinese Large Language Models. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.1148

@inproceedings{tian2026know,
  title = {{Know Your Place: Diagnosing Implicit Social Adaptation Failures in Chinese Large Language Models}},
  author = {Yu Tian, Jie Xing, Ziming Li, Jiang Li, Zehua Duo, Tian Lan, Xu Liu, Guanglai Gao and Xiangdong Su},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.1148},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1148}
}

References

36 references, 16 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Robert B Arundale. 2020. Communicating & relating: Constituting face in everyday interacting. Oxford University Press. find
  2. Molly E Ireland, Richard B Slatcher, Paul W Eastwick, Lauren E Scissors, Eli J Finkel, and James W Pennebaker. 2011. Language style matching predicts relationship initiation and stability. Psychological science, 22(1):39–44. find
  3. Terra Blevins, Susanne Schmalwieser, and Benjamin Roth. 2025. Do language models accommodate their users? a study of linguistic convergence. arXiv preprint arXiv:2508.03276. find
  4. Fatima Jahara, Mark Dredze, and Sharon Levy. 2025. Evaluating implicit biases in llm reasoning through logic grid puzzles. arXiv preprint arXiv:2511.06160. find
  5. Penelope Brown. 1987. Politeness: Some universals in language usage, volume 4. Cambridge university press. find
  6. Robert Kirk, Ishita Mediratta, Christoforos Nalmpantis, Jelena Luketina, Eric Hambro, Edward Grefenstette, and Roberta Raileanu. 2023. Understanding the effects of rlhf on llm generalisation and diversity. arXiv preprint arXiv:2310.06452.
  7. Shreyas Chaudhari, Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan, Ameet Deshpande, and Bruno Castro da Silva. 2025. Rlhf deciphered: A critical analysis of reinforcement learning from human feedback for llms. ACM Computing Surveys, 58(2):1–37. find
  8. Katherine M Collins, Albert Q Jiang, Simon Frieder, Lionel Wong, Miri Zilka, Umang Bhatt, Thomas Lukasiewicz, Yuhuai Wu, Joshua B Tenenbaum, William Hart, and 1 others. 2024. Evaluating language models for mathematics through interactions. Proceedings of the National Academy of Sciences, 121(24):e23… find
  9. Andrea Ferrario, Alberto Termine, and Alessandro Fac-chini. 2024. Addressing social misattributions of large language models: An hcxai-based approach. arXiv preprint arXiv:2403.17873. — [arXiv:2403.17873](https://arxiv.org/abs/2403.17873) find
  10. Howard Giles, Justine Coupland, and Nikolas Coupland. 1991. Contexts of accommodation: Developments in applied sociolinguistics. Cambridge University Press. find
  11. Team GLM, Aohan Zeng, Bin Xu, Bowen Wang, Chenhui Zhang, Da Yin, Dan Zhang, Diego Rojas, Guanyu Feng, Hanlin Zhao, and 1 others. 2024. Chatglm: A family of large language models from glm-130b to glm-4 all tools. arXiv preprint arXiv:2406.12793.
  12. Yueguo Gu. 1990. Politeness phenomena in modern chinese. Journal of pragmatics, 14(2):237–257. find
  13. Taicheng Guo, Xiuying Chen, Yaqi Wang, Ruidi Chang, Shichao Pei, Nitesh V Chawla, Olaf Wiest, and Xiangliang Zhang. 2024. Large language model based multi-agents: A survey of progress and challenges. arXiv preprint arXiv:2402.01680.
  14. Edward T Hall. 1976. Beyond culture. Anchor. find
  15. John Heritage. 2012. Epistemics in action: Action formation and territories of knowledge. Research on language & social interaction, 45(1):1–29. find
  16. Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, and 1 others. 2024. Gpt-4o system card. arXiv preprint arXiv:2410.21276.
  17. Youngeun Koo, Jiwoo Lee, Dojun Park, Seohyun Park, and Sungeun Lee. 2025. Evaluating large language models on understanding korean indirect speech acts. arXiv preprint arXiv:2502.10995. — [arXiv:2502.10995](https://arxiv.org/abs/2502.10995) find
  18. Cheng Li, Mengzhuo Chen, Jindong Wang, Sunayana Sitaram, and Xing Xie. 2024. Culturellm: Incorporating cultural differences into large language models. Advances in Neural Information Processing Systems, 37:84799–84838.
  19. Aixin Liu, Aoxue Mei, Bangcai Lin, Bing Xue, Bingxuan Wang, Bingzheng Xu, Bochao Wu, Bowei Zhang, Chaofan Lin, Chen Dong, and 1 others. 2025. Deepseek-v3. 2: Pushing the frontier of open large language models. arXiv preprint arXiv:2512.02556.
  20. Eric Mayor. 2025. Markers of synchrony in large language model conversational agreements and disagreements. find
  21. Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, and 1 others. 2022. Training language models to follow instructions with human feedback. Advances in neural information processing systems, 35:27730–27744.
  22. Joon Sung Park, Joseph O’Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S Bernstein. 2023. Generative agents: Interactive simulacra of human behavior. In Proceedings of the 36th annual acm symposium on user interface software and technology, pages 1–22. find
  23. Murray Shanahan, Kyle McDonell, and Laria Reynolds. 2023. Role play with large language models. Nature, 623(7987):493–498.
  24. Mrinank Sharma, Meg Tong, Tomasz Korbak, David Duvenaud, Amanda Askell, Samuel R Bowman, Newton Cheng, Esin Durmus, Zac Hatfield-Dodds, Scott R Johnston, and 1 others. 2023. Towards understanding sycophancy in language models. arXiv preprint arXiv:2310.13548.
  25. Zhivar Sourati, Alireza S Ziabari, and Morteza Dehghani. 2025. The homogenizing effect of large language models on human expression and thought. arXiv preprint arXiv:2508.01491. — [arXiv:2508.01491](https://arxiv.org/abs/2508.01491) find
  26. Kimi Team, Yifan Bai, Yiping Bao, Guanduo Chen, Jiahao Chen, Ningxin Chen, Ruijue Chen, Yanru Chen, Yuankun Chen, Yutian Chen, and 1 others. 2025. Kimi k2: Open agentic intelligence. arXiv preprint arXiv:2507.20534.
  27. Quan Tu, Shilong Fan, Zihang Tian, Tianhao Shen, Shuo Shang, Xin Gao, and Rui Yan. 2024. Charactereval: A chinese benchmark for role-playing conversational agent evaluation. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1183…
  28. Neeraj Varshney, Wenlin Yao, Hongming Zhang, Jianshu Chen, and Dong Yu. 2023. A stitch in time saves nine: Detecting and mitigating hallucinations of llms by validating low-confidence generation. arXiv preprint arXiv:2307.03987.
  29. Chaoyi Wang and Yuan Sun. 2025. Mseed: Human preference dataset for multidimensional safety enhancement and evaluation of large language models. DATA INTELLIGENCE, 7(4):1243–1269. find
  30. Noah Wang, Zy Peng, Haoran Que, Jiaheng Liu, Wangchunshu Zhou, Yuhan Wu, Hongcheng Guo, Ruitong Gan, Zehao Ni, Jian Yang, and 1 others. 2024. Rolellm: Benchmarking, eliciting, and enhancing role-playing abilities of large language models. In Findings of the Association for Computational Linguistics:…
  31. Alexander Wei, Nika Haghtalab, and Jacob Steinhardt. 2023. Jailbroken: How does llm safety training fail? Advances in Neural Information Processing Systems, 36:80079–80110.
  32. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, and 1 others. 2025. Qwen3 technical report. arXiv preprint arXiv:2505.09388.
  33. Longfei Yun, Chenyang An, Zilong Wang, Letian Peng, and Jingbo Shang. 2025. The price of format: Diversity collapse in llms. arXiv preprint arXiv:2505.18949. find
  34. Xiangkai Zeng, Cheng Yang, Cunchao Tu, Zhiyuan Liu, and Maosong Sun. 2018. Chinese liwc lexicon expansion via hierarchical classification of word embeddings with sememe attention. In Proceedings of the AAAI conference on artificial intelligence, volume 32. find
  35. Jinfeng Zhou, Yuxuan Chen, Yihan Shi, Xuanming Zhang, Leqi Lei, Yi Feng, Zexuan Xiong, Miao Yan, Xunzhi Wang, Yaru Cao, and 1 others. 2025. Socialeval: Evaluating social intelligence of large language models. arXiv preprint arXiv:2506.00900. find
  36. Xuhui Zhou, Hao Zhu, Leena Mathur, Ruohong Zhang, Haofei Yu, Zhengyang Qi, Louis-Philippe Morency, Yonatan Bisk, Daniel Fried, Graham Neubig, and 1 others. 2023. Sotopia: Interactive evaluation for social intelligence in language agents. arXiv preprint arXiv:2310.11667.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tian-2026-know-your-place-diagnosing-implicit/