---
title: "Identity-Robust Language Model Generation via Content Integrity"
description: "Building on this, they propose Identity-Robust Generation (IRG), a training-free, plug-and-play pipeline with three stages… 50 references."
authors: ["Miao Zhang", "Kelly Chen", "Mehrab Tanjim", "Rumi Chunara"]
year: 2026
venue: "Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)"
source: "https://aclanthology.org/2026.acl-long.884/"
doi: "10.18653/v1/2026.acl-long.884"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/zhang-2026-identity-robust-language-model-generation/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/zhang-2026-identity-robust-language-model-generation/)

[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/36/) / Identity-Robust Language Model Generation via Content Integrity Preservation

# Identity-Robust Language Model Generation via Content Integrity Preservation

**Authors:** [Miao Zhang](https://md.research.elamur.ai/free/en/authors/miao-zhang/), Kelly Chen, Mehrab Tanjim and Rumi Chunara · **Year:** 2026 · **Venue:** [Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2026](https://md.research.elamur.ai/free/en/venues/acl/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.acl-long.884/) · [PDF](https://aclanthology.org/2026.acl-long.884.pdf) · [DOI 10.18653/v1/2026.acl-long.884](https://doi.org/10.18653/v1/2026.acl-long.884) · [OpenAlex](https://openalex.org/W7166888904)

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

## Summary
The paper examines a quieter form of LLM bias: when a prompt mentions who the user is (age, religion, employment status, etc.), the quality of answers to objective questions can drop, even though the identity has no bearing on the correct answer. Using Llama3.3-70B-Instruct, the authors show that accuracy on TruthfulQA shifts across identities. Probes trained on attention-head activations, however, show that the model's internal factual knowledge stays stable. The distortion therefore happens during generation, not in what the model knows.

Building on this, they propose Identity-Robust Generation (IRG), a training-free, plug-and-play pipeline with three stages:
1. **Detection:** find identity mentions and test whether each one is needed to answer the query.
2. **Neutral generation:** strip non-essential mentions and answer the neutralized query.
3. **Optional personalization:** restyle the answer for the user, then check that its meaning still matches the neutral version.

Across four benchmarks and 18 identities, IRG reduces identity-dependent bias by an average of 66.3% compared with vanilla prompting. It also beats prompt-based defenses without lowering answer quality. Because it works only on inputs and outputs, the method is practical for closed or deployed models.

## Key points
- The evaluation covers 18 identities in six categories: education, religion, race, career, age and gender. It measures disparities in factuality, utility, ambiguity resolution and safety; the named datasets include TruthfulQA, MMLU-Pro and AmbigQA.
- In the motivating analysis, the reported probe accuracy is averaged over the top 10 attention heads of Llama3.3-70B-Instruct. Internal truth representations stay consistent across identities while output accuracy varies.
- IRG cuts identity-induced bias by 66.3% on average relative to vanilla prompting and outperforms existing prompt-based debiasing. It requires no fine-tuning, activation steering or annotated data.
- The method stays effective when multiple identities are combined in one prompt (e.g., "an unemployed man"), a setting where bias tends to grow.
- Personalization was tested by adapting readability for high school vs. PhD users, measured with Flesch–Kincaid Grade Level differences. Plain style instructions increased bias, while IRG achieved strong stylistic adaptation with consistently low bias on the PB measure.

*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

- [New York University](https://md.research.elamur.ai/free/en/universities/new-york-university/) — University, New York, United States · 219 works · 103,487 citations

📚 Shelves (6): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works · [🤝 Social Sciences](https://md.research.elamur.ai/free/en/shelves/social-sciences/) · 1,954 works · [🤝 Safety Research](https://md.research.elamur.ai/free/en/shelves/social-sciences/safety-research/) · 626 works

🧭 Topics (1): [🧠 Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · 3,894 works

## Cite

Cite as: Miao Zhang, Kelly Chen, Mehrab Tanjim and Rumi Chunara. 2026. Identity-Robust Language Model Generation via Content Integrity Preservation. 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.

```bibtex
@inproceedings{zhang2026identity,
  title = {{Identity-Robust Language Model Generation via Content Integrity Preservation}},
  author = {Miao Zhang, Kelly Chen, Mehrab Tanjim and Rumi Chunara},
  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.884},
  url = {https://doi.org/10.18653/v1/2026.acl-long.884}
}
```

- [← previous Identifying the Periodicity of Information in Natural Language](https://md.research.elamur.ai/free/en/papers/ou-2026-identifying-periodicity-information-natural-language/)
- [next → Ideology Prediction of German Political Texts](https://md.research.elamur.ai/free/en/papers/schneider-2026-ideology-prediction-german-political-texts/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/36/)

## References

50 references, 15 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Sandhini Agarwal, Lama Ahmad, Jason Ai, Sam Altman, Andy Applebaum, Edwin Arbus, Rahul K Arora, Yu Bai, Bowen Baker, Haiming Bao, and 1 others. 2025. [gpt-oss-120b & gpt-oss-20b model card](https://md.research.elamur.ai/free/en/papers/agarwal-2025-gpt-oss-120b-gpt-oss/). arXiv preprint arXiv:2508.10925.
2. Haozhe An, Connor Baumler, Abhilasha Sancheti, and Rachel Rudinger. 2025. On the mutual influence of gender and occupation in LLM representations. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1663–1680, Vienna, Austria. Ass… [find](https://scholar.google.com/scholar?q=Haozhe+An%2C+Connor+Baumler%2C+Abhilasha+Sancheti%2C+and+Rachel+Rudinger%2E+2025%2E+On+the+mutual+influence+of+gender+and+occupation+in+LLM+representations%2E+In+Proceedings+of+the+63rd+Annual+Meeting+of+the+Asso)
3. Amos Azaria and Tom Mitchell. 2023. [The internal state of an LLM knows when it’s lying](https://md.research.elamur.ai/free/en/papers/azaria-2023-internal-state-llm-knows-when/). In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 967–976, Singapore. Association for Computational Linguistics.
4. Pragyan Banerjee, Abhinav Java, Surgan Jandial, Simra Shahid, Shaz Furniturewala, Balaji Krishnamurthy, and Sumit Bhatia. 2024. All should be equal in the eyes of lms: Counterfactually aware fair text generation. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 1767… [find](https://scholar.google.com/scholar?q=Pragyan+Banerjee%2C+Abhinav+Java%2C+Surgan+Jandial%2C+Simra+Shahid%2C+Shaz+Furniturewala%2C+Balaji+Krishnamurthy%2C+and+Sumit+Bhatia%2E+2024%2E+All+should+be+equal+in+the+eyes+of+lms%3A+Counterfactually+aware+fair+text)
5. Tilman Beck, Hendrik Schuff, Anne Lauscher, and Iryna Gurevych. 2024. [Sensitivity, performance, robustness: Deconstructing the effect of sociodemographic prompting](https://md.research.elamur.ai/free/en/papers/beck-2024-sensitivity-performance-robustness-deconstructing-effect/). In Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers),…
6. Camilla Casula, Sebastiano Vecellio Salto, Elisa Leonardelli, and Sara Tonelli. 2025. Job unfair: An investigation of gender and occupational bias in free-form text completions by LLMs. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 22770–22788, Suzh… [find](https://scholar.google.com/scholar?q=Camilla+Casula%2C+Sebastiano+Vecellio+Salto%2C+Elisa+Leonardelli%2C+and+Sara+Tonelli%2E+2025%2E+Job+unfair%3A+An+investigation+of+gender+and+occupational+bias+in+free%2Dform+text+completions+by+LLMs%2E+In+Proceedings)
7. Ruoxi Cheng, Haoxuan Ma, Shuirong Cao, Jiaqi Li, Ai-hua Pei, Zhiqiang Wang, Pengliang Ji, Haoyu Wang, and Jiaqi Huo. 2024. Reinforcement learning from multi-role debates as feedback for bias mitigation in llms. arXiv preprint arXiv:2404.10160. — [arXiv:2404.10160](https://arxiv.org/abs/2404.10160) [find](https://scholar.google.com/scholar?q=Ruoxi+Cheng%2C+Haoxuan+Ma%2C+Shuirong+Cao%2C+Jiaqi+Li%2C+Ai%2Dhua+Pei%2C+Zhiqiang+Wang%2C+Pengliang+Ji%2C+Haoyu+Wang%2C+and+Jiaqi+Huo%2E+2024%2E+Reinforcement+learning+from+multi%2Drole+debates+as+feedback+for+bias+mitigatio)
8. A. Dubey, A. Jauhri, A. Pandey, A. Kadian, A. Al-Dahle, A. Letman, R. Ganapathy, and 1 others. 2024. [The llama 3 herd of models](https://md.research.elamur.ai/free/en/papers/grattafiori-2024-llama-3-herd-models/). arXiv preprint arXiv:2407.21783.
9. Shaz Furniturewala, Surgan Jandial, Abhinav Java, Pragyan Banerjee, Simra Shahid, Sumit Bhatia, and Kokil Jaidka. 2024. “thinking” fair and slow: On the efficacy of structured prompts for debiasing language models. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Proces… [find](https://scholar.google.com/scholar?q=Shaz+Furniturewala%2C+Surgan+Jandial%2C+Abhinav+Java%2C+Pragyan+Banerjee%2C+Simra+Shahid%2C+Sumit+Bhatia%2C+and+Kokil+Jaidka%2E+2024%2E+%E2%80%9Cthinking%E2%80%9D+fair+and+slow%3A+On+the+efficacy+of+structured+prompts+for+debiasing+la)
10. Isabel O. Gallegos, Ryan Aponte, Ryan A. Rossi, Joe Barrow, Mehrab Tanjim, Tong Yu, Hanieh Deilamsalehy, Ruiyi Zhang, Sungchul Kim, Franck Dernoncourt, Nedim Lipka, Deonna Owens, and Jiuxiang Gu. 2025. Self-debiasing large language models: Zero-shot recognition and reduction of stereotypes. In Proce… [find](https://scholar.google.com/scholar?q=Isabel+O%2E+Gallegos%2C+Ryan+Aponte%2C+Ryan+A%2E+Rossi%2C+Joe+Barrow%2C+Mehrab+Tanjim%2C+Tong+Yu%2C+Hanieh+Deilamsalehy%2C+Ruiyi+Zhang%2C+Sungchul+Kim%2C+Franck+Dernoncourt%2C+Nedim+Lipka%2C+Deonna+Owens%2C+and+Jiuxiang+Gu%2E+2025)
11. 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](https://md.research.elamur.ai/free/en/papers/gallegos-2023-bias-fairness-large-language-models/). Computational Linguistics, 50(3):1097– 1179.
12. Zorik Gekhman, Eyal Ben David, Hadas Orgad, Eran Ofek, Yonatan Belinkov, Idan Szpektor, Jonathan Herzig, and Roi Reichart. 2025. Inside-out: Hidden factual knowledge in llms. In Conference on Language Modeling (COLM) 2025. [find](https://scholar.google.com/scholar?q=Zorik+Gekhman%2C+Eyal+Ben+David%2C+Hadas+Orgad%2C+Eran+Ofek%2C+Yonatan+Belinkov%2C+Idan+Szpektor%2C+Jonathan+Herzig%2C+and+Roi+Reichart%2E+2025%2E+Inside%2Dout%3A+Hidden+factual+knowledge+in+llms%2E+In+Conference+on+Language)
13. Asma Ghandeharioun, Ann Yuan, Marius Guerard, Emily Reif, Michael Lepori, and Lucas Dixon. 2024. Who’s asking? user personas and the mechanics of latent misalignment. Advances in Neural Information Processing Systems, 37:125967–126003. [find](https://scholar.google.com/scholar?q=Asma+Ghandeharioun%2C+Ann+Yuan%2C+Marius+Guerard%2C+Emily+Reif%2C+Michael+Lepori%2C+and+Lucas+Dixon%2E+2024%2E+Who%E2%80%99s+asking%3F+user+personas+and+the+mechanics+of+latent+misalignment%2E+Advances+in+Neural+Information+Pr)
14. Farsheed Haque, Zhe Fu, Depeng Xu, Shuhan Yuan, and Xi Niu. 2025. Fine-tuning LLMs with cross-attention-based weight decay for bias mitigation. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 15785–15798, Suzhou, China. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Farsheed+Haque%2C+Zhe+Fu%2C+Depeng+Xu%2C+Shuhan+Yuan%2C+and+Xi+Niu%2E+2025%2E+Fine%2Dtuning+LLMs+with+cross%2Dattention%2Dbased+weight+decay+for+bias+mitigation%2E+In+Findings+of+the+Association+for+Computational+Linguis)
15. Jen-tse Huang, Yuhang Yan, Linqi Liu, Yixin Wan, Wenxuan Wang, Kai-Wei Chang, and Michael R. Lyu. 2025. Where fact ends and fairness begins: Redefining AI bias evaluation through cognitive biases. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 10974–10993, Suzhou, Ch… [find](https://scholar.google.com/scholar?q=Jen%2Dtse+Huang%2C+Yuhang+Yan%2C+Linqi+Liu%2C+Yixin+Wan%2C+Wenxuan+Wang%2C+Kai%2DWei+Chang%2C+and+Michael+R%2E+Lyu%2E+2025%2E+Where+fact+ends+and+fairness+begins%3A+Redefining+AI+bias+evaluation+through+cognitive+biases%2E+In+)
16. Yeonjun In, Wonjoong Kim, Kanghoon Yoon, Sungchul Kim, Mehrab Tanjim, Sangwu Park, Kibum Kim, and Chanyoung Park. 2025. Is safety standard same for everyone? user-specific safety evaluation of large language models. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 6652… [find](https://scholar.google.com/scholar?q=Yeonjun+In%2C+Wonjoong+Kim%2C+Kanghoon+Yoon%2C+Sungchul+Kim%2C+Mehrab+Tanjim%2C+Sangwu+Park%2C+Kibum+Kim%2C+and+Chanyoung+Park%2E+2025%2E+Is+safety+standard+same+for+everyone%3F+user%2Dspecific+safety+evaluation+of+large+l)
17. Jingru Jessica Jia, Zehua Yuan, Junhao Pan, Paul McNamara, and Deming Chen. 2024. Decision-making behavior evaluation framework for llms under uncertain context. Advances in Neural Information Processing Systems, 37:113360–113382. [find](https://scholar.google.com/scholar?q=Jingru+Jessica+Jia%2C+Zehua+Yuan%2C+Junhao+Pan%2C+Paul+McNamara%2C+and+Deming+Chen%2E+2024%2E+Decision%2Dmaking+behavior+evaluation+framework+for+llms+under+uncertain+context%2E+Advances+in+Neural+Information+Process)
18. J Peter Kincaid, Robert P Fishburne Jr, Richard L Rogers, and Brad S Chissom. 1975. Derivation of new readability formulas (automated readability index, fog count and flesch reading ease formula) for navy enlisted personnel. Technical report. [find](https://scholar.google.com/scholar?q=J+Peter+Kincaid%2C+Robert+P+Fishburne+Jr%2C+Richard+L+Rogers%2C+and+Brad+S+Chissom%2E+1975%2E+Derivation+of+new+readability+formulas+%28automated+readability+index%2C+fog+count+and+flesch+reading+ease+formula%29+for+)
19. Tomo Lazovich. 2023. Filter bubbles and affective polarization in user-personalized large language model outputs. In Proceedings on, pages 29–37. PMLR. [find](https://scholar.google.com/scholar?q=Tomo+Lazovich%2E+2023%2E+Filter+bubbles+and+affective+polarization+in+user%2Dpersonalized+large+language+model+outputs%2E+In+Proceedings+on%2C+pages+29%E2%80%9337%2E+PMLR%2E)
20. Jingling Li, Zeyu Tang, Xiaoyu Liu, Peter Spirtes, Kun Zhang, Liu Leqi, and Yang Liu. 2024. Prompting fairness: Integrating causality to debias large language models. arXiv preprint arXiv:2403.08743. — [arXiv:2403.08743](https://arxiv.org/abs/2403.08743) [find](https://scholar.google.com/scholar?q=Jingling+Li%2C+Zeyu+Tang%2C+Xiaoyu+Liu%2C+Peter+Spirtes%2C+Kun+Zhang%2C+Liu+Leqi%2C+and+Yang+Liu%2E+2024%2E+Prompting+fairness%3A+Integrating+causality+to+debias+large+language+models%2E+arXiv+preprint+arXiv%3A2403%2E08743%2E+)
21. Jingling Li, Zeyu Tang, Xiaoyu Liu, Peter Spirtes, Kun Zhang, Liu Leqi, and Yang Liu. 2025. Prompting fairness: Integrating causality to debias large language models. In Proceedings of the International Conference on Learning Representations (ICLR). [find](https://scholar.google.com/scholar?q=Jingling+Li%2C+Zeyu+Tang%2C+Xiaoyu+Liu%2C+Peter+Spirtes%2C+Kun+Zhang%2C+Liu+Leqi%2C+and+Yang+Liu%2E+2025%2E+Prompting+fairness%3A+Integrating+causality+to+debias+large+language+models%2E+In+Proceedings+of+the+Internation)
22. Kenneth Li, Oam Patel, Fernanda Viégas, Hanspeter Pfister, and Martin Wattenberg. 2023a. [Inference-time intervention: Eliciting truthful answers from a language model](https://md.research.elamur.ai/free/en/papers/li-2023-inference-time-intervention-eliciting-truthful/). Advances in Neural Information Processing Systems, 36:41451–41530.
23. Yingji Li, Mengnan Du, Rui Song, Xin Wang, and Ying Wang. 2023b. A survey on fairness in large language models. arXiv preprint arXiv:2308.10149. — [arXiv:2308.10149](https://arxiv.org/abs/2308.10149) [find](https://scholar.google.com/scholar?q=Yingji+Li%2C+Mengnan+Du%2C+Rui+Song%2C+Xin+Wang%2C+and+Ying+Wang%2E+2023b%2E+A+survey+on+fairness+in+large+language+models%2E+arXiv+preprint+arXiv%3A2308%2E10149%2E+%E2%80%94+%5BarXiv%3A2308%2E10149%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2308%2E10149%29)
24. Stephanie Lin, Jacob Hilton, and Owain Evans. 2022. [Truthfulqa: Measuring how models mimic human falsehoods](https://md.research.elamur.ai/free/en/papers/lin-2022-truthfulqa-measuring-how-models-mimic/). In Proceedings of the 60th annual meeting of the association for computational linguistics (volume 1: long papers), pages 3214–3252.
25. Dianqing Liu, Yi Liu, Guoqing Jin, and Zhendong Mao. 2025. Mitigating biases in language models via bias unlearning. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 4160–4178, Suzhou, China. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Dianqing+Liu%2C+Yi+Liu%2C+Guoqing+Jin%2C+and+Zhendong+Mao%2E+2025%2E+Mitigating+biases+in+language+models+via+bias+unlearning%2E+In+Proceedings+of+the+2025+Conference+on+Empirical+Methods+in+Natural+Language+Proc)
26. Siyang Liu, Trisha Maturi, Bowen Yi, Siqi Shen, and Rada Mihalcea. 2024. The generation gap: Exploring age bias in the value systems of large language models. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 19617–19634, Miami, Florida, USA. Associatio… [find](https://scholar.google.com/scholar?q=Siyang+Liu%2C+Trisha+Maturi%2C+Bowen+Yi%2C+Siqi+Shen%2C+and+Rada+Mihalcea%2E+2024%2E+The+generation+gap%3A+Exploring+age+bias+in+the+value+systems+of+large+language+models%2E+In+Proceedings+of+the+2024+Conference+on+)
27. Marlene Lutz, Indira Sen, Georg Ahnert, Elisa Rogers, and Markus Strohmaier. 2025. The prompt makes the person(a): A systematic evaluation of sociodemographic persona prompting for large language models. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 23212–23237, Suz… [find](https://scholar.google.com/scholar?q=Marlene+Lutz%2C+Indira+Sen%2C+Georg+Ahnert%2C+Elisa+Rogers%2C+and+Markus+Strohmaier%2E+2025%2E+The+prompt+makes+the+person%28a%29%3A+A+systematic+evaluation+of+sociodemographic+persona+prompting+for+large+language+mode)
28. Orfeas Menis Mastromichalakis, Giorgos Filandrianos, Maria Symeonaki, and Giorgos Stamou. 2025. Assumed identities: Quantifying gender bias in machine translation of gender-ambiguous occupational terms. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages… [find](https://scholar.google.com/scholar?q=Orfeas+Menis+Mastromichalakis%2C+Giorgos+Filandrianos%2C+Maria+Symeonaki%2C+and+Giorgos+Stamou%2E+2025%2E+Assumed+identities%3A+Quantifying+gender+bias+in+machine+translation+of+gender%2Dambiguous+occupational+term)
29. Sewon Min, Julian Michael, Hannaneh Hajishirzi, and Luke Zettlemoyer. 2020. AmbigQA: Answering ambiguous open-domain questions. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 5783– 5797, Online. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Sewon+Min%2C+Julian+Michael%2C+Hannaneh+Hajishirzi%2C+and+Luke+Zettlemoyer%2E+2020%2E+AmbigQA%3A+Answering+ambiguous+open%2Ddomain+questions%2E+In+Proceedings+of+the+2020+Conference+on+Empirical+Methods+in+Natural+La)
30. Vera Neplenbroek, Arianna Bisazza, and Raquel Fernández. 2025. [Reading between the prompts: How stereotypes shape LLM’s implicit personalization](https://md.research.elamur.ai/free/en/papers/neplenbroek-2025-reading-between-prompts-how-stereotypes/). In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 20378–20411, Suzhou, China. Association for Computational…
31. Anna Neumann, Elisabeth Kirsten, Muhammad Bilal Zafar, and Jatinder Singh. 2025. Position is power: System prompts as a mechanism of bias in large language models (llms). In Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency, pages 573–598. [find](https://scholar.google.com/scholar?q=Anna+Neumann%2C+Elisabeth+Kirsten%2C+Muhammad+Bilal+Zafar%2C+and+Jatinder+Singh%2E+2025%2E+Position+is+power%3A+System+prompts+as+a+mechanism+of+bias+in+large+language+models+%28llms%29%2E+In+Proceedings+of+the+2025+AC)
32. Hadas Orgad, Michael Toker, Zorik Gekhman, Roi Reichart, Idan Szpektor, Hadas Kotek, and Yonatan Belinkov. 2025. [Llms know more than they show: On the intrinsic representation of llm hallucinations](https://md.research.elamur.ai/free/en/papers/orgad-2024-llms-know-more-than-they/). In International Conference on Learning Representations (ICLR 2025).
33. Ingroj Shrestha and Padmini Srinivasan. 2025. LLM bias detection and mitigation through the lens of desired distributions. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 1464–1480, Suzhou, China. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Ingroj+Shrestha+and+Padmini+Srinivasan%2E+2025%2E+LLM+bias+detection+and+mitigation+through+the+lens+of+desired+distributions%2E+In+Proceedings+of+the+2025+Conference+on+Empirical+Methods+in+Natural+Languag)
34. Lihao Sun, Chengzhi Mao, Valentin Hofmann, and Xuechunzi Bai. 2025. [Aligned but blind: Alignment increases implicit bias by reducing awareness of race](https://md.research.elamur.ai/free/en/papers/sun-2025-aligned-but-blind-alignment-increases/). In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 22167–22184, Vienna, Austr…
35. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, and 1 others. 2023. [Llama: Open and efficient foundation language models](https://md.research.elamur.ai/free/en/papers/touvron-2023-llama-open-efficient-foundation-language/). arXiv preprint arXiv:2302.13971.
36. Takuma Udagawa, Yang Zhao, Hiroshi Kanayama, and Bishwaranjan Bhattacharjee. 2025. Bias analysis and mitigation through protected attribute detection and regard classification. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 16–25, Suzhou, China. Association for Compu… [find](https://scholar.google.com/scholar?q=Takuma+Udagawa%2C+Yang+Zhao%2C+Hiroshi+Kanayama%2C+and+Bishwaranjan+Bhattacharjee%2E+2025%2E+Bias+analysis+and+mitigation+through+protected+attribute+detection+and+regard+classification%2E+In+Findings+of+the+Asso)
37. Anvesh Rao Vijjini, Somnath Basu Roy Chowdhury, and Snigdha Chaturvedi. 2025. Exploring safety-utility trade-offs in personalized language models. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologie… [find](https://scholar.google.com/scholar?q=Anvesh+Rao+Vijjini%2C+Somnath+Basu+Roy+Chowdhury%2C+and+Snigdha+Chaturvedi%2E+2025%2E+Exploring+safety%2Dutility+trade%2Doffs+in+personalized+language+models%2E+In+Proceedings+of+the+2025+Conference+of+the+Nations+)
38. Yixin Wan and Kai-Wei Chang. 2025a. The male CEO and the female assistant: Evaluation and mitigation of gender biases in text-to-image generation of dual subjects. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 9174–9190, Vie… [find](https://scholar.google.com/scholar?q=Yixin+Wan+and+Kai%2DWei+Chang%2E+2025a%2E+The+male+CEO+and+the+female+assistant%3A+Evaluation+and+mitigation+of+gender+biases+in+text%2Dto%2Dimage+generation+of+dual+subjects%2E+In+Proceedings+of+the+63rd+Annual+Me)
39. Yixin Wan and Kai-Wei Chang. 2025b. White men lead, black women help? benchmarking and mitigating language agency social biases in LLMs. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 9082–9108, Vienna, Austria. Association f… [find](https://scholar.google.com/scholar?q=Yixin+Wan+and+Kai%2DWei+Chang%2E+2025b%2E+White+men+lead%2C+black+women+help%3F+benchmarking+and+mitigating+language+agency+social+biases+in+LLMs%2E+In+Proceedings+of+the+63rd+Annual+Meeting+of+the+Association+fo)
40. Keyu Wang, Jin Li, Shu Yang, Zhuoran Zhang, and Di Wang. 2025. When truth is overridden: Uncovering the internal origins of sycophancy in large language models. arXiv preprint arXiv:2508.02087. — [arXiv:2508.02087](https://arxiv.org/abs/2508.02087) [find](https://scholar.google.com/scholar?q=Keyu+Wang%2C+Jin+Li%2C+Shu+Yang%2C+Zhuoran+Zhang%2C+and+Di+Wang%2E+2025%2E+When+truth+is+overridden%3A+Uncovering+the+internal+origins+of+sycophancy+in+large+language+models%2E+arXiv+preprint+arXiv%3A2508%2E02087%2E+%E2%80%94+%5BarX)
41. Yubo Wang, Xueguang Ma, Ge Zhang, Yuansheng Ni, Abhranil Chandra, Shiguang Guo, Weiming Ren, Aaran Arulraj, Xuan He, Ziyan Jiang, and 1 others. 2024a. [Mmlu-pro: A more robust and challenging multi-task language understanding benchmark](https://md.research.elamur.ai/free/en/papers/wang-2024-mmlu-pro-more-robust-challenging/). Advances in Neural Information Processing Systems, 37:95266–9529…
42. Yuxia Wang, Haonan Li, Xudong Han, Preslav Nakov, and Timothy Baldwin. 2024b. [Do-not-answer: Evaluating safeguards in LLMs](https://md.research.elamur.ai/free/en/papers/wang-2024-do-not-answer-evaluating-safeguards/). In Findings of the Association for Computational Linguistics: EACL 2024, pages 896–911, St. Julian’s, Malta. Association for Computational Linguistics.
43. Kangda Wei, Hasnat Md Abdullah, and Ruihong Huang. 2025. Mitigating gender bias via fostering exploratory thinking in LLMs. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 6895–6917, Suzhou, China. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Kangda+Wei%2C+Hasnat+Md+Abdullah%2C+and+Ruihong+Huang%2E+2025%2E+Mitigating+gender+bias+via+fostering+exploratory+thinking+in+LLMs%2E+In+Findings+of+the+Association+for+Computational+Linguistics%3A+EMNLP+2025%2C+pa)
44. Iain Weissburg, Sathvika Anand, Sharon Levy, and Haewon Jeong. 2025. LLMs are biased teachers: Evaluating LLM bias in personalized education. In Findings of the Association for Computational Linguistics: NAACL 2025, pages 5650–5698, Albuquerque, New Mexico. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Iain+Weissburg%2C+Sathvika+Anand%2C+Sharon+Levy%2C+and+Haewon+Jeong%2E+2025%2E+LLMs+are+biased+teachers%3A+Evaluating+LLM+bias+in+personalized+education%2E+In+Findings+of+the+Association+for+Computational+Linguisti)
45. Kyra Wilson and Aylin Caliskan. 2024. Gender, race, and intersectional bias in resume screening via language model retrieval. In Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society, volume 7, pages 1578–1590. [find](https://scholar.google.com/scholar?q=Kyra+Wilson+and+Aylin+Caliskan%2E+2024%2E+Gender%2C+race%2C+and+intersectional+bias+in+resume+screening+via+language+model+retrieval%2E+In+Proceedings+of+the+AAAI%2FACM+Conference+on+AI%2C+Ethics%2C+and+Society%2C+volu)
46. 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](https://md.research.elamur.ai/free/en/papers/yang-2025-qwen3-technical-report/). arXiv preprint arXiv:2505.09388.
47. Urchade Zaratiana, Gil Pasternak, Oliver Boyd, George Hurn-Maloney, and Ash Lewis. 2025. GLiNER2: Schema-driven multi-task learning for structured information extraction. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 130–140,… [find](https://scholar.google.com/scholar?q=Urchade+Zaratiana%2C+Gil+Pasternak%2C+Oliver+Boyd%2C+George+Hurn%2DMaloney%2C+and+Ash+Lewis%2E+2025%2E+GLiNER2%3A+Schema%2Ddriven+multi%2Dtask+learning+for+structured+information+extraction%2E+In+Proceedings+of+the+2025+Co)
48. Tao Zhang, Ziqian Zeng, YuxiangXiao YuxiangXiao, Huiping Zhuang, Cen Chen, James R. Foulds, and Shimei Pan. 2025. GenderAlign: An alignment dataset for mitigating gender bias in large language models. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume… [find](https://scholar.google.com/scholar?q=Tao+Zhang%2C+Ziqian+Zeng%2C+YuxiangXiao+YuxiangXiao%2C+Huiping+Zhuang%2C+Cen+Chen%2C+James+R%2E+Foulds%2C+and+Shimei+Pan%2E+2025%2E+GenderAlign%3A+An+alignment+dataset+for+mitigating+gender+bias+in+large+language+models%2E)
49. Wenting Zhao, Xiang Huang, Zhibin Shen, Jieyu Wu, Yuqing Li, Qianzhen Qi, Tianxiang Sun, Yonatan Bisk, Denny Yu, Linlin Song, and 1 others. 2024. [Wildchat: 1m chatgpt interaction logs in the wild](https://md.research.elamur.ai/free/en/papers/zhao-2024-wildchat-1m-chatgpt-interaction-logs/). In The Twelfth International Conference on Learning Representations (ICLR).
50. Hanzhang Zhou, Zijian Feng, Zixiao Zhu, Junlang Qian, and Kezhi Mao. 2024. Unibias: Unveiling and mitigating llm bias through internal attention and ffn manipulation. Advances in Neural Information Processing Systems, 37:102173–102196. [find](https://scholar.google.com/scholar?q=Hanzhang+Zhou%2C+Zijian+Feng%2C+Zixiao+Zhu%2C+Junlang+Qian%2C+and+Kezhi+Mao%2E+2024%2E+Unibias%3A+Unveiling+and+mitigating+llm+bias+through+internal+attention+and+ffn+manipulation%2E+Advances+in+Neural+Information+Pr)

*Updated 2 November 2026*