---
title: "Polite but Misaligned — paper by Sun; and Guo (2026)"
description: "The paper investigates whether large language models judge politeness the way people do, and whether strong scores on… 38 references."
authors: ["Rong Wang; Kun Sun;", "Yadong Guo"]
year: 2026
venue: "EMNLP2026"
source: "https://arxiv.org/abs/2609.29001"
doi: "10.48550/arxiv.2609.29001"
arxiv: "2609.29001"
licence: "CC BY-NC-ND 4.0 (arXiv)"
licence_url: "http://creativecommons.org/licenses/by-nc-nd/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/wang-2026-polite-but-misaligned-evaluating-llm/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/wang-2026-polite-but-misaligned-evaluating-llm/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/39/) / Polite but Misaligned: Evaluating LLM Politeness Judgments Against Human Pragmatic Norms

# Polite but Misaligned: Evaluating LLM Politeness Judgments Against Human Pragmatic Norms

**Authors:** Rong Wang; Kun Sun; and Yadong Guo · **Year:** 2026 · **Venue:** EMNLP2026 · **Licence:** [CC BY-NC-ND 4.0 (arXiv)](http://creativecommons.org/licenses/by-nc-nd/4.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2609.29001) · [PDF](https://arxiv.org/pdf/2609.29001) · [DOI 10.48550/arxiv.2609.29001](https://doi.org/10.48550/arxiv.2609.29001) · [arXiv 2609.29001](https://arxiv.org/abs/2609.29001) · [OpenAlex](https://openalex.org/W7214300154)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
The paper investigates whether large language models judge politeness the way people do, and whether strong scores on aggregate metrics hide systematic bias. The authors used two English online datasets. The first is 3,000 requests drawn from the Stanford Politeness Corpus, which has continuous crowd ratings. The second is 3,000 class-balanced items from the Hugging Face frfede/politeness-corpus, labelled Impolite, Neutral or Polite. Seven LLMs were prompted zero-shot with six templates that varied role framing, access to pragmatic cues, and output format. The authors then measured correlation, error and agreement with the human labels and between models. They also defined new metrics that capture the direction of disagreement, and five linguistically trained annotators audited 318 high-disagreement cases. The models agreed with each other more than with humans. They aligned with humans mainly when explicit politeness markers were present. They also systematically drifted toward Neutral and under-predicted Impolite, and this bias remained when expert consensus replaced crowd labels. The authors argue that polite-sounding output does not show human-like pragmatic judgment, and that evaluation should examine biased disagreement patterns against several human references.

## Key points
- **Models evaluated:** gemini-2.5-flash, gpt-4.1, claude-3.5-sonnet, grok-3 and deepseek-chat, all accessed through OpenRouter with default sampling. The open-weight llama-3-8b-instruct and llama-3.1-8b-instruct were added for the categorical dataset only. No fine-tuning or output correction was applied.
- **Datasets:**
  - The Stanford Politeness Corpus contains 10,957 requests from Wikipedia Talk and Stack Exchange, each rated by five annotators. A score-balanced 3,000 subset was used.
  - The frfede corpus contains 16,428 items, with 5,476 per class. A 1,000-per-class sample was used.
- **Metrics for the continuous task:** Pearson r, MAE, and a "Close Rate," defined as the share of items where the model and human scores differ by less than 0.5.
- **Metrics for the categorical task:** accuracy, macro-F1, Cohen's κ, and Fleiss' κ across models. The authors also introduced directional diagnostics:
  - neutral-compression difference and ratio;
  - impoliteness-suppression difference;
  - per-class rate of shifts from Polite or Impolite to Neutral.
- **Main findings:**
  - Inter-model agreement exceeded model–human agreement.
  - Rapport-building strategies were over-represented among misaligned cases.
  - Expert-framed and function-aware prompts changed calibration but did not close the gap.
  - Expert agreement in the 318-case audit was only fair-to-moderate.
  - Prompts, metrics, model outputs and scripts are released on GitHub (rong4ivy/llm-politeness).

## Interesting ideas
- Accuracy alone can hide a classifier that defaults to the middle class; counting how often each extreme class collapses into the default label exposes this bias directly.
- Reporting model-to-model agreement alongside model-to-human agreement reveals that several systems can share a bias, so high consensus among them does not prove they are right.
- Treat crowd labels and expert consensus as two separate references, and audit only a deliberately chosen set of high-disagreement items, without using that set to estimate overall error rates.
- Models treat each extra politeness marker as making a text more polite, whereas humans see a limit beyond which piling on markers turns into over-politeness or insincerity.
- Build prompt templates by varying controlled factors one at a time: role framing, which cues the model may use, and output format. This isolates how sensitive results are to each prompt choice.

## For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

- **I1** (learn): Add shift-to-default rates to the evaluation of LLM teacher tiers and silver labels (e.g., speech-act or sentiment-like labels), so a teacher that hedges toward "neutral/other" gets caught before its labels are trained on.
- **I2** (learn): When several witness parsers or LLM teachers agree on silver data, still compare against a small human gold set, because shared training biases can make the witnesses agree while all being wrong.
- **I3** (frames): For prag's speech-act, implicature and irony labels, keep a small expert-audited set of hard cases as a diagnostic, kept separate from the main benchmark score.
- **I4** (frames): prag could model politeness cues as saturating rather than additive, and treat an excess of mitigators or flattery as a trigger for an irony or strategic-exaggeration check.
- **I5** (learn): Run teacher prompts as a small factorial design so that label changes caused by phrasing can be measured and filtered out of the training data.

*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 Tübingen](https://md.research.elamur.ai/free/en/universities/university-of-tubingen/) — University, Tübingen, Germany · 75 works · 41,081 citations
- [Tongji University](https://md.research.elamur.ai/free/en/universities/tongji-university/) — University, Shanghai, China · 19 works · 2,162 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

🧭 Topics (2): [🧠 Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · 3,320 works · [English](https://md.research.elamur.ai/free/en/topics/english/) · 2,263 works

## Cite

Cite as: Rong Wang; Kun Sun; and Yadong Guo. 2026. Polite but Misaligned: Evaluating LLM Politeness Judgments Against Human Pragmatic Norms. EMNLP2026.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{sun2026polite,
  title = {{Polite but Misaligned: Evaluating LLM Politeness Judgments Against Human Pragmatic Norms}},
  author = {Rong Wang; Kun Sun; and Yadong Guo},
  booktitle = {EMNLP2026},
  year = {2026},
  doi = {10.48550/arxiv.2609.29001},
  eprint = {2609.29001},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2609.29001}
}
```

- [← previous Polimi at CRF Filling 2026: Prompt-Based Information Extraction from Italian Clinical Notes](https://md.research.elamur.ai/free/en/papers/torri-2026-polimi-crf-filling-2026-prompt/)
- [next → Poller: Are LLMs Suitable for Evaluating Poetry Understanding Task?](https://md.research.elamur.ai/free/en/papers/wang-2026-poller-are-llms-suitable-evaluating/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/39/)

## References

38 references, 17 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Marta Andersson and Dan McIntyre. 2025. Can Chat-GPT recognize impoliteness? An exploratory study of the pragmatic awareness of a large language model. Journal of Pragmatics, 239:16–36. [find](https://scholar.google.com/scholar?q=Marta+Andersson+and+Dan+McIntyre%2E+2025%2E+Can+Chat%2DGPT+recognize+impoliteness%3F+An+exploratory+study+of+the+pragmatic+awareness+of+a+large+language+model%2E+Journal+of+Pragmatics%2C+239%3A16%E2%80%9336%2E)
2. Ljubiša Bojić, Dylan Seychell, and Milan Čabarkapa. 2025. Towards new benchmark for ai alignment & sentiment analysis in socially important issues: A comparative study of human and llms in the context of agi. arXiv preprint arXiv:2501.02531. [↗](https://arxiv.org/abs/2501.02531)
3. Penelope Brown and Stephen C. Levinson. 1987. [Politeness](https://md.research.elamur.ai/free/en/papers/brown-1987-politeness-some-universals-language-usage/): Some Universals in Language Usage. Cambridge University Press, Cambridge.
4. Myra Cheng, Sunny Yu, Cinoo Lee, Pranav Khadpe, Lujain Ibrahim, and Dan Jurafsky. 2026. [ELEPHANT: Measuring and understanding social sycophancy in LLMs](https://md.research.elamur.ai/free/en/papers/cheng-2025-elephant-measuring-understanding-social-sycophancy/). In ICLR. [↗](https://doi.org/10.48550/arxiv.2505.13995)
5. Cristian Danescu-Niculescu-Mizil, Moritz Sudhof, Dan Jurafsky, Jure Leskovec, and Christopher Potts. 2013. [A computational approach to politeness with application to social factors](https://md.research.elamur.ai/free/en/papers/mizil-2013-computational-approach-politeness-application-social/). In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 250–259. Association for Computational Linguistics. [↗](https://doi.org/10.48550/arxiv.1306.6078)
6. Aida Mostafazadeh Davani, Mark Díaz, and Vinodkumar Prabhakaran. 2022. [Dealing with disagreements: Looking beyond the majority vote in subjective annotations](https://md.research.elamur.ai/free/en/papers/davani-2022-dealing-disagreements-looking-beyond-majority/). Transactions of the Association for Computational Linguistics, 10:92–110. [↗](https://doi.org/10.1162/tacl_a_00449)
7. Mauajama Firdaus, Asif Ekbal, and Pushpak Bhattacharyya. 2020. Incorporating politeness across languages in customer care responses: Towards building a multi-lingual empathetic dialogue agent. In Proceedings of the 12th Language Resources and Evaluation Conference, pages 4172–4182. [find](https://scholar.google.com/scholar?q=Mauajama+Firdaus%2C+Asif+Ekbal%2C+and+Pushpak+Bhattacharyya%2E+2020%2E+Incorporating+politeness+across+languages+in+customer+care+responses%3A+Towards+building+a+multi%2Dlingual+empathetic+dialogue+agent%2E+In+Proc)
8. Bruce Fraser. 2010. Pragmatic competence: The case of hedging. In Günther Kaltenböck, Wiltrud Mihatsch, and Stefan Schneider, editors, New Approaches to Hedging, volume 9 of Studies in Pragmatics, pages 15–34. Emerald Group Publishing Limited, Bingley, UK. frfede. 2024. Politeness corpus. Accessed: 2026-08-30. [↗](https://doi.org/10.1163/9789004253247_003)
9. Samuel Gehman, Suchin Gururangan, Maarten Sap, Yejin Choi, and Noah A. Smith. 2020. [RealToxicityPrompts: Evaluating neural toxic degeneration in language models](https://md.research.elamur.ai/free/en/papers/gehman-2020-realtoxicityprompts-evaluating-neural-toxic-degeneration/). In Findings of EMNLP 2020, pages 3356–3369.
10. H. Paul Grice. 1975. Logic and conversation. In Peter Cole and Jerry L. Morgan, editors, Syntax and Semantics, Volume 3: Speech Acts, pages 41–58. Academic Press, New York. [↗](https://doi.org/10.1163/9789004368811_003)
11. Thomas Hartvigsen, Saadia Gabriel, Hamid Palangi, Maarten Sap, Dipankar Ray, and Ece Kamar. 2022. [ToxiGen: A large-scale machine-generated dataset for adversarial and implicit hate speech detection](https://md.research.elamur.ai/free/en/papers/hartvigsen-2022-toxigen-large-scale-machine-generated/). In Proceedings of ACL 2022, pages 3309–3326.
12. Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2021. [Measuring massive multitask language understanding](https://md.research.elamur.ai/free/en/papers/hendrycks-2020-measuring-massive-multitask-language-understanding/). Proceedings of the International Conference on Learning Representations. [↗](https://doi.org/10.48550/arxiv.2009.03300)
13. Jiaming Ji, Tianyi Qiu, Boyuan Chen, Borong Zhang, Hantao Lou, Kaile Wang, Yawen Duan, Zhonghao He, Lukas Vierling, Donghai Hong, Jiayi Zhou, Zhaowei Zhang, Fanzhi Zeng, Juntao Dai, Xuehai Pan, Kwan Yee Ng, Aidan O’Gara, Hua Xu, Brian Tse, and 7 others. 2023. [AI alignment: A comprehensive survey](https://md.research.elamur.ai/free/en/papers/ji-2023-ai-alignment-comprehensive-survey/). arXiv preprint arXiv:2310.19852. [↗](https://arxiv.org/abs/2310.19852)
14. Kenji Kitao. 1990. A study of Japanese and American perceptions of politeness in requests. Doshisha Studies in English, 50:178–210. [↗](https://doi.org/10.14988/pa.2017.0000001647)
15. Gabriel Rongyang Lau, Wei Yan Low, Seow Min Koh, Fiona Fui-Hoon Nah, and Andree Hartanto. 2025. Evaluating AI alignment in LLMs: Output analysis of value priorities across 75 models with human benchmarking. arXiv preprint arXiv:2506.12617. [↗](https://arxiv.org/abs/2506.12617)
16. Geoffrey N. Leech. 1983. Principles of Pragmatics. Longman. [find](https://scholar.google.com/scholar?q=Geoffrey+N%2E+Leech%2E+1983%2E+Principles+of+Pragmatics%2E+Longman%2E)
17. Xuying Li, Zhuo Li, Yuji Kosuga, and Victor Bian. 2025. Optimizing safe and aligned language generation: A multi-objective GRPO approach. arXiv preprint arXiv:2503.21819. [↗](https://arxiv.org/abs/2503.21819)
18. Adam Dahlgren Lindström, Leila Methnani, Lea Krause, Petter Ericson, Íñigo Martínez de Rituerto de Troya, Dimitri Coelho Mollo, and Roel Dobbe. 2025. Helpful, harmless, honest? Sociotechnical limits of AI alignment and safety through reinforcement learning from human feedback. Ethics and Information Technology, 27(2):28. [find](https://scholar.google.com/scholar?q=Adam+Dahlgren+Lindstr%C3%B6m%2C+Leila+Methnani%2C+Lea+Krause%2C+Petter+Ericson%2C+%C3%8D%C3%B1igo+Mart%C3%ADnez+de+Rituerto+de+Troya%2C+Dimitri+Coelho+Mollo%2C+and+Roel+Dobbe%2E+2025%2E+Helpful%2C+harmless%2C+honest%3F+Sociotechnical+limits+o)
19. Bing Liu and Lei Zhang. 2012. A survey of opinion mining and sentiment analysis. In Mining text data, pages 415–463. Springer. [↗](https://doi.org/10.1007/978-1-4614-3223-4_13)
20. Miriam A. Locher and Richard J. Watts. 2005. Politeness theory and relational work. Journal of Politeness Research: Language, Behaviour, Culture, 1(1):9–33. [↗](https://doi.org/10.1515/jplr.2005.1.1.9)
21. Bolei Ma, Yuting Li, Wei Zhou, Ziwei Gong, Yang Janet Liu, Katja Jasinskaja, Annemarie Friedrich, Julia Hirschberg, Frauke Kreuter, and Barbara Plank. 2025. [Pragmatics in the era of large language models: A survey on datasets, evaluation, opportunities and challenges](https://md.research.elamur.ai/free/en/papers/ma-2025-pragmatics-era-large-language-models/). In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 8679–8696, Vienna, Austria. Association for Computational Linguistics.
22. Kshitij Mishra, Priyanshu Priya, and Asif Ekbal. 2023. Help me heal: A reinforced polite and empathetic mental health and legal counseling dialogue system for crime victims. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 14408–14416. [find](https://scholar.google.com/scholar?q=Kshitij+Mishra%2C+Priyanshu+Priya%2C+and+Asif+Ekbal%2E+2023%2E+Help+me+heal%3A+A+reinforced+polite+and+empathetic+mental+health+and+legal+counseling+dialogue+system+for+crime+victims%2E+In+Proceedings+of+the+AAAI)
23. Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L. Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, and 1 others. 2022. [Training language models to follow instructions with human feedback](https://md.research.elamur.ai/free/en/papers/ouyang-2022-training-language-models-follow-instructions/). Advances in Neural Information Processing Systems, 35:27730–27744.
24. Frank Robert Palmer. 2001. Mood and modality. Cambridge University Press. [↗](https://doi.org/10.1017/cbo9781139167178)
25. Barbara Plank. 2022. [The “problem” of human label variation: On ground truth in data, modeling and evaluation](https://md.research.elamur.ai/free/en/papers/plank-2022-problem-human-label-variation-ground/). In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 10671–10682, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics. [↗](https://doi.org/10.18653/v1/2022.emnlp-main.731)
26. Priyanshu Priya, Mauajama Firdaus, and Asif Ekbal. 2024. [Computational politeness in natural language processing: A survey](https://md.research.elamur.ai/free/en/papers/priya-2024-computational-politeness-natural-language-processing/). ACM Computing Surveys, 56(9):1–42. Article 241. [↗](https://doi.org/10.1145/3654660)
27. Crystal Qian, Aaron T. Parisi, Clémentine Bouleau, Vivian Tsai, Maël Lebreton, and Lucas Dixon. 2025. To mask or to mirror: Human-AI alignment in collective reasoning. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 2398–2423, Suzhou, China. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Crystal+Qian%2C+Aaron+T%2E+Parisi%2C+Cl%C3%A9mentine+Bouleau%2C+Vivian+Tsai%2C+Ma%C3%ABl+Lebreton%2C+and+Lucas+Dixon%2E+2025%2E+To+mask+or+to+mirror%3A+Human%2DAI+alignment+in+collective+reasoning%2E+In+Proceedings+of+the+2025+Confe)
28. Kyra L. Rosen, Margaret Sui, Kimia Heydari, Elizabeth J. Enichen, and Joseph C. Kvedar. 2025. The perils of politeness: how large language models may amplify medical misinformation. NPJ Digital Medicine, 8(1):644. [↗](https://doi.org/10.1038/s41746-025-02135-7)
29. Laura Ruis, Akbir Khan, Stella Biderman, Sara Hooker, Tim Rocktäschel, and Edward Grefenstette. 2023. [The goldilocks of pragmatic understanding: Fine-tuning strategy matters for implicature resolution by LLMs](https://md.research.elamur.ai/free/en/papers/ruis-2022-goldilocks-pragmatic-understanding-fine-tuning/). In Advances in Neural Information Processing Systems (NeurIPS).
30. Valery Shulginov, Hasan Berkcan Şimşek, Sergei Kudriashov, Renata Randautsova, and Sofya A. Shevela. 2025. Evaluating the pragmatic competence of large language models in detecting mitigated and unmitigated types of disagreement. In Computational Linguistics and Intellectual Technologies: Proceedings of the International Conference Dialogue 2025, volume 23, pages 345–360. [find](https://scholar.google.com/scholar?q=Valery+Shulginov%2C+Hasan+Berkcan+S%CC%A7ims%CC%A7ek%2C+Sergei+Kudriashov%2C+Renata+Randautsova%2C+and+Sofya+A%2E+Shevela%2E+2025%2E+Evaluating+the+pragmatic+competence+of+large+language+models+in+detecting+mitigated+and+unm)
31. Dan Sperber and Deirdre Wilson. 1986. Relevance: Communication and cognition. Harvard University Press, Cambridge, MA. [find](https://scholar.google.com/scholar?q=Dan+Sperber+and+Deirdre+Wilson%2E+1986%2E+Relevance%3A+Communication+and+cognition%2E+Harvard+University+Press%2C+Cambridge%2C+MA%2E)
32. Settaluri Sravanthi, Meet Doshi, Pavan Tankala, Rudra Murthy, Raj Dabre, and Pushpak Bhattacharyya. 2024. [PUB: A pragmatics understanding benchmark for assessing LLMs’ pragmatics capabilities](https://md.research.elamur.ai/free/en/papers/sravanthi-2024-pub-pragmatics-understanding-benchmark-assessing/). In Findings of the Association for Computational Linguistics: ACL 2024, pages 12075–12097, Bangkok, Thailand. Association for Computational Linguistics.
33. Aarohi Srivastava, Abhinav Rastogi, Abhishek Rao, Abu Awal Md Shoeb, Abubakar Abid, Adam Fisch, Adam R. Brown, Adam Santoro, Aditya Gupta, Adrià Garriga-Alonso, and 1 others. 2023. [Beyond the imitation game: Quantifying and extrapolating the capabilities of language models](https://md.research.elamur.ai/free/en/papers/srivastava-2022-imitation-game-quantifying-extrapolating-capabilities/). Transactions on Machine Learning Research.
34. Mike Yeomans, Alejandro Kantor, and Dustin Tingley. 2018. The politeness package: Detecting politeness in natural language. The R Journal, 10(2):489–502. [↗](https://doi.org/10.32614/rj-2018-079)
35. Ziqi Yin, Hao Wang, Kaito Horio, Daisuke Kawahara, and Satoshi Sekine. 2024. Should we respect LLMs? A cross-lingual study on the influence of prompt politeness on LLM performance. In Proceedings of the Second Workshop on Social Influence in Conversations (SICon 2024), pages 9–35. [find](https://scholar.google.com/scholar?q=Ziqi+Yin%2C+Hao+Wang%2C+Kaito+Horio%2C+Daisuke+Kawahara%2C+and+Satoshi+Sekine%2E+2024%2E+Should+we+respect+LLMs%3F+A+cross%2Dlingual+study+on+the+influence+of+prompt+politeness+on+LLM+performance%2E+In+Proceedings+of+t)
36. Ivan Zakazov, Mikolaj Boronski, Lorenzo Drudi, and Robert West. 2024. Assessing social alignment: Do personality-prompted large language models behave like humans? arXiv preprint arXiv:2412.16772. [↗](https://arxiv.org/abs/2412.16772)
37. Zhexin Zhang, Leqi Lei, Lindong Wu, Rui Sun, Yongkang Huang, Chong Long, Xiao Liu, Xuanyu Lei, Jie Tang, and Minlie Huang. 2024. [SafetyBench: Evaluating the safety of large language models](https://md.research.elamur.ai/free/en/papers/zhang-2023-safetybench-evaluating-safety-large/). In Proceedings of ACL 2024, pages 15537–15553. [↗](https://doi.org/10.48550/arxiv.2309.07045)
38. Haoran Zhao and Robert D. Hawkins. 2025. [Comparing human and LLM politeness strategies in free production](https://md.research.elamur.ai/free/en/papers/zhao-2025-comparing-human-llm-politeness-strategies/). In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 16188–16216, Suzhou, China. Association for Computational Linguistics.

*Updated 1 November 2026*