Elamur Research⌕
☰
home / papers / 2026 / Preconditioned Test-Time Adaptation for Out-of-Distribution Debiasing in Narrative Generation

Preconditioned Test-Time Adaptation for Out-of-Distribution Debiasing in Narrative Generation

Paper · 2026 · Hanwen Shen, Ting Ying, Jiajie Lu et al.

Summary

Debiased LLMs usually handle familiar or mildly biased prompts well, but they often fail on unfamiliar, strongly bias-laden prompts. The authors treat this as a distribution-shift problem in long-form narrative generation. They used Qwen3-4B-based OOD detectors, with WritingPrompts as the in-distribution reference and RealToxicityPrompts as the candidate set. These detectors show that toxic prompts lie far outside the reference distribution, and static models perform worse on them.

The proposed framework, CAP-TTA, recasts debiasing as episodic test-time adaptation. Text is generated in segments while a bias-risk score is tracked. Only when that score exceeds a threshold does the system take a few update steps on a LoRA adapter, which is reset for each prompt. These steps are guided by a diagonal preconditioner computed offline. Safe update data come from WikiText-2 and from a typed SafeBank built on HolisticBias.

CAP-TTA matches the bias scores of the best self-correction baselines while raising fluency by about 12%. It also runs faster than AdamW or SGD updates and limits catastrophic forgetting. This points toward selective, adaptive debiasing that preserves narrative quality.

Key points

  • OOD detection (RTP vs. WritingPrompts, Qwen3-4B): AUROC is 99.22% for kNN (k=10) and 98.81% for Mahalanobis, both indicating Far-OOD. LLR gives 70.74%, suggesting Near-OOD.
  • Compared systems include Qwen3-4B, DeepSeek-R1-8B, DeepSeek-R1-8B-Debiased, Mistral-7B-Instruct and Qwen3-4B-Sherlock (self-correction), scored with the BB Bias score on ID and OOD sets. Base-model bias appears to rise from roughly 0.29 (ID) to 0.45 (OOD) for Qwen3-4B, though the extracted table layout is ambiguous.
  • In the fluency-oriented setting, a DiD analysis shows a marginally significant bias reduction over the base model (p=0.07).
  • Human evaluation used binary bias judgments and was approved by the Stevens IRB (Protocol 2026-002).
  • Code is released at github.com/hshen13/debias_tta.

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 MacauMacao · China33 works · 2,223 citations🎓Stevens Institute of TechnologyHoboken · United States10 works · 95 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 1Research subjects and languages the work is about.Bias, fairness & safety · 517 works

Cite

Cite as Hanwen Shen, Ting Ying, Jiajie Lu and Shanshan Wang. 2026. Preconditioned Test-Time Adaptation for Out-of-Distribution Debiasing in Narrative Generation. 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.919

@inproceedings{shen2026preconditioned,
  title = {{Preconditioned Test-Time Adaptation for Out-of-Distribution Debiasing in Narrative Generation}},
  author = {Hanwen Shen, Ting Ying, Jiajie Lu and Shanshan Wang},
  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.919},
  url = {https://doi.org/10.18653/v1/2026.acl-long.919}
}

References

64 references, 24 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Afra Feyza Akyürek, Sejin Paik, Muhammed Kocyigit, Seda Akbiyik, Serife Leman Runyun, and Derry Wijaya. 2022. On measuring social biases in prompt-based multi-task learning. In Findings of the Association for Computational Linguistics: NAACL 2022, pages 551–564, Seattle, United States. Association f… find
  2. Cem Anil, Esin Durmus, Nina Panickssery, Mrinank Sharma, Joe Benton, Sandipan Kundu, Joshua Batson, Meg Tong, Jesse Mu, Daniel Ford, Francesco Mosconi, Rajashree Agrawal, Rylan Schaeffer, Naomi Bashkansky, Samuel Svenningsen, Mike Lambert, Ansh Radhakrishnan, Carson Denison, Evan Hubinger, and 15 ot… find
  3. Howard S. Becker. 1967. Whose side are we on? Social Problems, 14(3):239–247. find
  4. Su Lin Blodgett, Solon Barocas, Hal Daumé III, and Hanna Wallach. 2020. Language (technology) is power: A critical survey of “bias” in NLP. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 5454– 5476, Online. Association for Computational Linguistics.
  5. Tolga Bolukbasi, Kai-Wei Chang, James Y. Zou, Venkatesh Saligrama, and Adam T. Kalai. 2016. Man is to computer programmer as woman is to homemaker? debiasing word embeddings. In Advances in Neural Information Processing Systems, volume 29.
  6. Asli Celikyilmaz, Elizabeth Clark, and Jianfeng Gao. 2020. Evaluation of text generation: A survey. arXiv preprint arXiv:2006.14799. find
  7. Yupeng Chang, Yi Chang, and Yuan Wu. 2026. BA-loRA: Bias-alleviating low-rank adaptation to mitigate catastrophic inheritance in large language models. In The Fourteenth International Conference on Learning Representations. find
  8. Sumanth Dathathri, Andrea Madotto, Janice Lan, Jane Hung, Eric Frank, Piero Molino, Jason Yosinski, and Kawin Ethayarajh and Dan Jurafsky. 2022. The authenticity gap in human evaluation. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 6056–6070, Abu D… find
  9. Angela Fan, Mike Lewis, and Yann Dauphin. 2018. Hierarchical neural story generation. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 889–898.
  10. Joseph L. Fleiss. 1971. Measuring nominal scale agreement among many raters. Psychological Bulletin, 76(5):378–382. find
  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. Computational Linguistics, 50(3):1097– 1179.
  12. Yossi Gandelsman, Yu Sun, Xinlei Chen, and Alexei A. Efros. 2022. Test-time training with masked autoencoders. Advances in Neural Information Processing Systems, 35:29374–29385. find
  13. Zhitong Gao, Shipeng Yan, and Xuming He. 2023. ATTA: Anomaly-aware test-time adaptation for out-of-distribution detection in segmentation. Preprint, arXiv:2309.05994. — [arXiv:2309.05994](https://arxiv.org/abs/2309.05994) find
  14. Iker García-Ferrero, David Montero, and Román Orús. 2025. Refusal steering: Fine-grained control over LLM refusal behaviour for sensitive topics. Preprint, arXiv:2512.16602. find
  15. Samuel Gehman, Suchin Gururangan, Maarten Sap, Yejin Choi, and Noah A. Smith. 2020. RealToxicityPrompts: Evaluating neural toxic degeneration in language models. In Findings of the Association for Computational Linguistics: EMNLP 2020, pages 3356–3369, Online. Association for Computational Linguisti…
  16. Gracjan Góral, Marysia Winkels, and Steven Basart. 2025. Depth-wise activation steering for honest language models. Preprint, arXiv:2512.07667. — [arXiv:2512.07667](https://arxiv.org/abs/2512.07667) find
  17. Kevin A. Hallgren. 2012. Computing inter-rater reliability for observational data: An overview and tutorial. Tutorials in Quantitative Methods for Psychology, 8(1):23–34. find
  18. Rujun Han, Yanfei Chen, Zoey CuiZhu, Lesly Miculicich, Guan Sun, Yuanjun Bi, Weiming Wen, Hui Wan, Chunfeng Wen, Solène Maître, George Lee, Vishy Tirumalashetty, Emily Xue, Zizhao Zhang, Salem Haykal, Burak Gokturk, Tomas Pfister, and Chen-Yu Lee. 2025. Deep researcher with test-time diffusion. arXi… — [arXiv:2507.16075](https://arxiv.org/abs/2507.16075) find
  19. Sandra Harding. 1992. Rethinking standpoint epistemology: What is “strong objectivity?”. The Centennial Review, 36(3):437–470. find
  20. Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2021. LoRA: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685.
  21. Evan Hubinger, Carson Denison, Jesse Mu, Mike Lambert, Meg Tong, Monte MacDiarmid, Tamera Lanham, Daniel M. Ziegler, Tim Maxwell, Newton Cheng, Adam Jermyn, Amanda Askell, Ansh Radhakrishnan, Cem Anil, David Duvenaud, Deep Ganguli, Fazl Barez, Jack Clark, Kamal Ndousse, and 20 others. 2024. Sleeper… find
  22. Wonje Jeung, Dongjae Jeon, Ashkan Yousefpour, and Jonghyun Choi. 2024. Large language models still exhibit bias in long text. arXiv preprint arXiv:2410.17519. find
  23. James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A. Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, Demis Hassabis, Claudia Clopath, Dharshan Kumaran, and Raia Hadsell. 2017. Overcoming catastrophic forgetting in neural networks… find
  24. Kimin Lee, Kibok Lee, Honglak Lee, and Jinwoo Shin. 2018. A simple unified framework for detecting out-of-distribution samples and adversarial attacks. In Advances in Neural Information Processing Systems. find
  25. 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, and 3… find
  26. Alisa Liu, Maarten Sap, Ximing Lu, Swabha Swayamdipta, Chandra Bhagavatula, Noah A. Smith, and Yejin Choi. 2021. DExperts: Decoding-time controlled text generation with experts and anti-experts. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th I…
  27. Shih-Yang Liu, Chien-Yi Wang, Hongxu Yin, Pavlo Molchanov, Yu-Chiang Frank Wang, Kwang-Ting Cheng, and Min-Hung Chen. 2024. DoRA: weight-decomposed low-rank adaptation. In Proceedings of the 41st International Conference on Machine Learning, ICML’24. JMLR.org. find
  28. James Martens and Roger Grosse. 2015. Optimizing neural networks with kronecker-factored approximate curvature. In International Conference on Machine Learning, pages 2408–2417. PMLR. find
  29. Nicholas Meade, Elinor Poole-Dayan, and Siva Reddy. 2022. An empirical survey of the effectiveness of debiasing techniques for pre-trained language models. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1878–1898, Dublin, Ire…
  30. Shira Mitchell, Eric Potash, Solon Barocas, Alexander D’Amour, and Kristian Lum. 2021. Algorithmic fairness: Choices, assumptions, and definitions. Annual Review of Statistics and Its Application, 8:141–163. find
  31. 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…
  32. 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…
  33. Shuaicheng Niu, Jiaxiang Wu, Yifan Zhang, Zhiquan Wen, Yaofo Chen, Peilin Zhao, and Mingkui Tan. 2023. Towards stable test-time adaptation in dynamic wild world. In International Conference on Learning Representations. find
  34. Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul F. Christiano, Jan Leike, and Ryan Lowe. 2022. Tr… — Training language models to follow instructions with human feedback
  35. German I. Parisi, Ronald Kemker, Jose L. Part, Christopher Kanan, and Stefan Wermter. 2019. Continual lifelong learning with neural networks: A review. Neural Networks, 113:54–71.
  36. Xiangyu Qi, Yi Zeng, Tinghao Xie, Pin-Yu Chen, Ruoxi Jia, Prateek Mittal, and Peter Henderson. 2023. Fine-tuning aligned language models compromises safety, even when users do not intend to! Preprint, arXiv:2310.03693.
  37. Joaquin Quiñonero-Candela, Masashi Sugiyama, Anton Schwaighofer, and Neil D. Lawrence, editors. 2008. Dataset Shift in Machine Learning. The MIT Press. find
  38. Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. 2023. Direct preference optimization: Your language model is secretly a reward model. In Advances in Neural Information Processing Systems, volume 36, pages 53728–53741. Curran Associates, Inc.
  39. Shauli Ravfogel, Yanai Elazar, Hila Gonen, Michael Twiton, and Yoav Goldberg. 2020. Null it out: Guarding protected attributes by iterative nullspace projection. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 7237–7256.
  40. Jie Ren, Peter J. Liu, Emily Fertig, Jasper Snoek, Ryan Poplin, Mark A. DePristo, Joshua V. Dillon, and Balaji Lakshminarayanan. 2019. Likelihood ratios for out-of-distribution detection. In Advances in Neural Information Processing Systems. find
  41. Maarten Sap, Dallas Card, Saadia Gabriel, Yejin Choi, and Noah A. Smith. 2019. The risk of racial bias in hate speech detection. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 1668–1678, Florence, Italy. Association for Computational Linguistics.
  42. Timo Schick, Sahana Udupa, and Hinrich Schütze. 2021. Self-diagnosis and self-debiasing: A proposal for reducing corpus-based bias in NLP. Transactions of the Association for Computational Linguistics, 9:1408– 1424.
  43. Hendrik Schuff, Lindsey Vanderlyn, Heike Adel, and Ngoc Thang Vu. 2023. How to do human evaluation: A brief introduction to user studies in NLP. Natural Language Engineering, 29(5):1199–1222. find
  44. Emily Sheng, Kai-Wei Chang, Premkumar Natarajan, and Nanyun Peng. 2019. The woman worked as a babysitter: On biases in language generation. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Process…
  45. Charlie Snell, Jaehoon Lee, Kelvin Xu, and Aviral Kumar. 2024. Scaling LLM test-time compute optimally can be more effective than scaling model parameters. arXiv preprint arXiv:2408.03314.
  46. Yiyou Sun, Yifei Ming, Xiaojin Zhu, and Yixuan Li. 2022. Out-of-distribution detection with deep nearest neighbors. In Proceedings of the 39th International Conference on Machine Learning, volume 162 of Proceedings of Machine Learning Research, pages 20827–20840. PMLR. find
  47. Yu Sun, Xiaolong Wang, Zhuang Liu, John Miller, Alexei A. Efros, and Moritz Hardt. 2020. Test-time training with self-supervision for generalization under distribution shifts. In International Conference on Machine Learning, pages 9229–9248. PMLR. find
  48. Harini Suresh and John V. Guttag. 2021. A framework for understanding sources of harm throughout the machine learning life cycle. In Proceedings of the 1st ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization, EAAMO ’21, pages 1–9, New York, NY, USA. Association for Comput… find
  49. Thomas Yu Chow Tam, Sonish Sivarajkumar, Sumit Kapoor, Alisa V. Stolyar, Katelyn Polanska, Kar-leigh R. McCarthy, Hunter Osterhoudt, Xizhi Wu, Shyam Visweswaran, Sunyang Fu, Piyush Mathur, Giovanni E. Cacciamani, Cong Sun, Yifan Peng, and Yanshan Wang. 2024. A framework for human evaluation of large… find
  50. Chris van der Lee, Albert Gatt, Emiel van Miltenburg, Sander Wubben, and Emiel Krahmer. 2019. Best practices for the human evaluation of automatically generated text. In Proceedings of the 12th International Conference on Natural Language Generation, pages 355–368, Tokyo, Japan. Association for Comp… find
  51. Dequan Wang, Evan Shelhamer, Shaoteng Liu, Bruno Olshausen, and Trevor Darrell. 2021. Tent: Fully test-time adaptation by entropy minimization. In International Conference on Learning Representations. find
  52. Liyuan Wang, Xingxing Zhang, Hang Su, and Jun Zhu. 2024a. A comprehensive survey of continual learning: Theory, method and application. IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(8):5362–5383. find
  53. Qin Wang, Olga Fink, Luc Van Gool, and Dengxin Dai. 2022. Continual test-time domain adaptation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7201–7211. find
  54. Shanshan Wang, Derek Wong, Jingming Yao, and Lidia Chao. 2024b. What is the best way for ChatGPT to translate poetry? In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 14025– 14043, Bangkok, Thailand. Association for Computation… find
  55. Shanshan Wang, Junchao Wu, Fengying Ye, Derek F. Wong, Jingming Yao, and Lidia S. Chao. 2025. Benchmarking the detection of LLMs-generated Modern Chinese poetry. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 9533–9552. Association for Computational Linguistics. find
  56. Tongtong Wu, Linhao Luo, Yuan-Fang Li, Shirui Pan, Thuy-Trang Vu, and Gholamreza Haffari. 2024. Continual learning for large language models: A survey. Preprint, arXiv:2402.01364.
  57. Xi Xiao, Chenrui Ma, Yunbei Zhang, Chen Liu, Zhux-uanzi Wang, Yanshu Li, Lin Zhao, Guosheng Hu, Tianyang Wang, and Hao Xu. 2026a. Not all directions matter: Towards structured and task-aware low-rank model adaptation. In Proceedings of the 64th Annual Meeting of the Association for Computational Lin…
  58. Xi Xiao, Yunbei Zhang, Lin Zhao, Yiyang Liu, Xiaoying Liao, Zheda Mai, Xingjian Li, Xiao Wang, Hao Xu, Jihun Hamm, Xue Lin, Min Xu, Qifan Wang, Tianyang Wang, and Cheng Han. 2026b. Prompt-based adaptation in large-scale vision models: A survey. Preprint, arXiv:2510.13219. — [arXiv:2510.13219](https://arxiv.org/abs/2510.13219) find
  59. Kevin Yang and Dan Klein. 2021. FUDGE: Controlled text generation with future discriminators. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 3511–3535.
  60. Brian Hu Zhang, Blake Lemoine, and Margaret Mitchell. 2018. Mitigating unwanted biases with adversarial learning. arXiv preprint arXiv:1801.07593. — [arXiv:1801.07593](https://arxiv.org/abs/1801.07593) find
  61. Marvin Zhang, Sergey Levine, and Chelsea Finn. 2022. Memo: Test time robustness via adaptation and augmentation. In Advances in Neural Information Processing Systems, volume 35, pages 38629–38642. find
  62. Qingru Zhang, Minshuo Chen, Alexander Bukharin, Pengcheng He, Yu Cheng, Weizhu Chen, and Tuo Zhao. 2023. Adaptive budget allocation for parameter-efficient fine-tuning. In International Conference on Learning Representations. find
  63. Jieyu Zhao, Tianlu Wang, Mark Yatskar, Vicente Ordonez, and Kai-Wei Chang. 2017. Men also like shopping: Reducing gender bias amplification using corpus-level constraints. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 2979–2989.
  64. Jieyu Zhao, Tianlu Wang, Mark Yatskar, Vicente Ordonez, and Kai-Wei Chang. 2018. Gender bias in coreference resolution: Evaluation and debiasing methods. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies,…

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/shen-2026-preconditioned-test-time-adaptation-out/