Elamur Research⌕
☰
home / papers / 2026 / Beyond Static Persona Consistency: Dynamic Persona Coherence in LLM Role-Playing

Beyond Static Persona Consistency: Dynamic Persona Coherence in LLM Role-Playing

Paper · 2026 · Yirui QI, Xiaoming Zhang, Ruilin Zeng et al.

Summary

LLM role-playing agents usually treat a character's persona as one fixed attribute. In long, pressured conversations they therefore either repeat themselves mechanically or abandon the character. This paper proposes Dynamic Persona Coherence, which separates a stable identity layer from an adaptive layer whose psychological state should change in a lawful way as the dialogue history grows. The authors implement this with the L/M/S Psychological State Model. L holds time-invariant traits, while M (accumulated meaning and stress) and S (short-term affect) are updated each turn by identity-conditioned deltas clipped to a 0–10 range. A correction loop then scores responses with the Persona Consistency Critic (PCC), stores cases in the Persona Case Repository (PCR) and repairs drifting outputs through the Persona Drift Suppressor (PDS). The system was tested with GPT-4o, Claude-3.5-Sonnet and DeepSeek-V3.2 across five personas, with main experiments in Chinese. It raises average PCC from about 0.73 to 0.92 and removes catastrophic drift. The analysis shows that baselines mainly fail to track cumulative emotional state rather than forgetting traits, which is a useful diagnosis for long-session character agents.

Key points

  • Average PCC improves by 26.8% (0.73 → 0.92). Gains per configuration range from +16% to +84%. The largest jump is for Claude-3.5 with an extroverted persona, where PCC rises from 0.50 to 0.91 and drift falls from 9.5% to 0%.
  • The S-score rises from 0.57 to 0.98 and the L-score reaches 1.0. The M-score dips slightly, from 0.89 to 0.86.
  • In the ablation, cumulative S/M tracking accounts for about 85% of the gain and PDS correction for about 15%.
  • PDS repaired 55 of 61 severe violations (90.2%) across 2,185 generation tasks in 15 configurations.
  • GPT-4o is the main judge, and results hold when Claude-3.5-Sonnet or DeepSeek-V3 serve as judges instead. Preliminary English and Japanese tests reach PCC 0.93–0.95.

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

🎓Beihang UniversityBeijing · China65 works · 8,452 citations

 Fields, topics and concepts

📚 Shelves 4The 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 works

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

Cite

Cite as Yirui QI, Xiaoming Zhang, Ruilin Zeng, Mengyao Liu, Ziyi Zhou, Dezhuang Miao, Bingyu Yan and Zhenyu Guan. 2026. Beyond Static Persona Consistency: Dynamic Persona Coherence in LLM Role-Playing. 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.1336

@inproceedings{qi2026beyond,
  title = {{Beyond Static Persona Consistency: Dynamic Persona Coherence in LLM Role-Playing}},
  author = {Yirui QI, Xiaoming Zhang, Ruilin Zeng, Mengyao Liu, Ziyi Zhou, Dezhuang Miao, Bingyu Yan and Zhenyu Guan},
  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.1336},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1336}
}

References

27 references, 12 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, and et al. 2023. Gpt-4 technical report. Computing Research Repository, arXiv:2303.08774. Version 6.
  2. Anthropic. 2024. The Claude 3 model family: Opus, sonnet, haiku. Claude-3 Model Card. find
  3. William Fleeson. 2001. Toward a structure-and process-integrated view of personality: Traits as density distributions of states. Journal of personality and social psychology, 80(6):1011. find
  4. Ivar Frisch and Mario Giulianelli. 2024. LLM agents in interaction: Measuring personality consistency and linguistic alignment in interacting populations of large language models. In Proceedings of the 1st Workshop on Personalization of Generative AI Systems (PERSONALIZE 2024), pages 102–111, St. Ju… find
  5. Cheng Li, Ziang Leng, Chenxi Yan, Junyi Shen, Hao Wang, Weishi Mi, Yaying Fei, Xiaoyang Feng, Song Yan, HaoSheng Wang, and et al. 2023. Chatharuhi: Reviving anime character in reality via large language model. Computing Research Repository, arXiv:2308.09597. Version 1.
  6. Kenneth Li, Tianle Liu, Naomi Bashkansky, David Bau, Fernanda Viégas, Hanspeter Pfister, and Martin Wattenberg. 2024. Measuring and controlling instruction (in)stability in language model dialogs. Computing Research Repository, arXiv:2402.10962. Version 4. — [arXiv:2402.10962](https://arxiv.org/abs/2402.10962) find
  7. Aixin Liu, Bei Feng, Bing Xue, Bingxuan Wang, Bochao Wu, Chengda Lu, Chenggang Zhao, Chengqi Deng, Chenyu Zhang, Chong Ruan, and et al. 2024. Deepseek-v3 technical report. Computing Research Repository, arXiv:2412.19437. Version 2. find
  8. Keming Lu, Bowen Yu, Chang Zhou, and Jingren Zhou. 2024. Large language models are superpositions of all characters: Attaining arbitrary role-play via self-alignment. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 7828–7840,…
  9. Ameet Deshpande, Karthik R Narasimhan, and Vishvak Murahari. 2025. PersonaGym: Evaluating persona agents and LLMs. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 6999–7022, Suzhou, China. Association for Computational Linguistics. find
  10. Klaus R Scherer. 2001. Appraisal considered as a process of multilevel sequential checking. Oxford University Press. find
  11. Murray Shanahan, Kyle McDonell, and Laria Reynolds. 2023. Role play with large language models. Nature, 623(7987):493–498.
  12. Yunfan Shao, Linyang Li, Junqi Dai, and Xipeng Qiu. 2023. Character-LLM: A trainable agent for role-playing. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 13153–13187, Singapore. Association for Computational Linguistics.
  13. Jisu Shin, Juhyun Oh, Eunsu Kim, Hoyun Song, and Alice Oh. 2025. Spotting out-of-character behavior: Atomic-level evaluation of persona fidelity in open-ended generation. In Findings of the Association for Computational Linguistics: ACL 2025, pages 26312–26332, Vienna, Austria. Association for Compu… find
  14. Craig A Smith and Richard S Lazarus. 1990. Emotion and adaptation. The Guilford Press. find
  15. Ala N Tak and Jonathan Gratch. 2023. Is gpt a computational model of emotion? In 2023 11th International Conference on Affective Computing and Intelligent Interaction (ACII), pages 1–8. IEEE. find
  16. Walter Mischel and Yuichi Shoda. 1995. A cognitive-affective system theory of personality: reconceptualizing situations, dispositions, dynamics, and invariance in personality structure. Psychological review, 102(2):246. find
  17. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, and et al. 2023. Llama 2: Open foundation and fine-tuned chat models. Computing Research Repository, arXiv:2307.09288. Version 2.
  18. Humza Naveed, Asad Ullah Khan, Shi Qiu, Muhammad Saqib, Saeed Anwar, Muhammad Usman, Naveed Akhtar, Nick Barnes, and Ajmal Mian. 2025. A comprehensive overview of large language models. ACM Transactions on Intelligent Systems and Technology, 16(5):1–72.
  19. Yu-Min Tseng, Yu-Chao Huang, Teng-Yun Hsiao, Wei-Lin Chen, Chao-Wei Huang, Yu Meng, and Yun-Nung Chen. 2024. Two tales of persona in LLMs: A survey of role-playing and personalization. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 16612–16631, Miami, Florida, USA. A…
  20. Cindy Nguyen, D Carrion, and MK Badawy. 2024. Comparative performance of claude and gpt models in basic radiological imaging tasks. MedRxiv, pages 2024–11. find
  21. Joon Sung Park, Joseph O’Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, Michael S Bernstein, and et al. 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
  22. Vinay Samuel, Henry Peng Zou, Yue Zhou, Shreyas Chaudhari, Ashwin Kalyan, Tanmay Rajpurohit, 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…
  23. Noah Wang, Z.y. Peng, Haoran Que, Jiaheng Liu, Wangchunshu Zhou, Yuhan Wu, Hongcheng Guo, Ruitong Gan, Zehao Ni, Jian Yang, Man Zhang, Zhaoxiang Zhang, Wanli Ouyang, Ke Xu, Wenhao Huang, Jie Fu, and Junran Peng. 2024. RoleLLM: Benchmarking, eliciting, and enhancing role-playing abilities of large la…
  24. Saizheng Zhang, Emily Dinan, Jack Urbanek, Arthur Szlam, Douwe Kiela, and Jason Weston. 2018. Personalizing dialogue agents: I have a dog, do you have pets too? In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2204–2213, Melbou…
  25. Wayne Xin Zhao, Kun Zhou, Junyi Li, Tianyi Tang, Xiaolei Wang, Yupeng Hou, Yingqian Min, Beichen Zhang, Junjie Zhang, Zican Dong, and et al. 2023. A survey of large language models. Computing Research Repository, arXiv:2303.18223. Version 16.
  26. Jinfeng Zhou, Zhuang Chen, Dazhen Wan, Bosi Wen, Yi Song, Jifan Yu, Yongkang Huang, Pei Ke, Guanqun Bi, Libiao Peng, JiaMing Yang, Xiyao Xiao, Sahand Sabour, Xiaohan Zhang, Wenjing Hou, Yijia Zhang, Yuxiao Dong, Hongning Wang, Jie Tang, and Minlie Huang. 2024. CharacterGLM: Customizing social charac… find
  27. Jinfeng Zhou, Yongkang Huang, Bosi Wen, Guanqun Bi, Yuxuan Chen, Pei Ke, Zhuang Chen, Xiyao Xiao, Libiao Peng, and Kuntian Tang. 2025. Characterbench: Benchmarking character customization of large language models. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 261… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/qi-2026-beyond-static-persona-consistency-dynamic/