CoChat
Sign Up
Cited by 148Speech and dialogue systemsTopic ModelingPersona Design and Applications

The DOI matches its Crossref record; 105 of the 127 references it cites were found in a registry or an index.

Two Tales of Persona in LLMs: A Survey of Role-Playing and Personalization

Open access
Tseng, Yu-Min, Huang, Yu-Chao, Hsiao, Teng-Yun, Chen, Wei-Lin, Huang, Chao-Wei, Meng, Yu, Chen, Yun-Nung·2024·Findings of the Association for Computational Linguistics: EMNLP 2024
DOI 10.18653/v1/2024.findings-emnlp.969
Open this paper in CoChat — free accountOpen PDF

Abstract

The concept of persona, originally adopted in dialogue literature, has re-surged as a promising framework for tailoring large language models (LLMs) to specific context (e.g., personalized search, LLM-as-a-judge).However, the growing research on leveraging persona in LLMs is relatively disorganized and lacks a systematic taxonomy.To close the gap, we present a comprehensive survey to categorize the current state of the field.We identify two lines of research, namely (1) LLM Role-Playing, where personas are assigned to LLMs, and (2) LLM Personalization, where LLMs take care of user personas.Additionally, we introduce existing methods for LLM personality evaluation.To the best of our knowledge, we present the first survey for role-playing and personalization in LLMs under the unified view of persona.We continuously maintain a paper collection to foster future endeavors.1 * Equal contribution. 1 https://github.com/MiuLab/PersonaLLM-Survey

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.18653/v1/2024.findings-emnlp.969

Cites

0 citation errors · 22 not checkable

Statistics

Figures

Figure 1 In Role-Playing , LLMs act according to assigned personas ( i.e. , roles) under a defined environment. For example, given role names with descriptions , LLMs role-play in a social simulation game. For Personalization , LLMs consider user personas to generate tailored responses to the same question. Dashed rectangles are prompts and solid rectangles are LLMs' responses.
Figure 1 In Role-Playing , LLMs act according to assigned personas ( i.e. , roles) under a defined environment. For example, given role names with descriptions , LLMs role-play in a social simulation game. For Personalization , LLMs consider user personas to generate tailored responses to the same question. Dashed rectangles are prompts and solid rectangles are LLMs' responses.CC-BY · From the PDF
Figure 2 The taxonomy of LLM role-playing and LLM personalization.
Figure 2 The taxonomy of LLM role-playing and LLM personalization.CC-BY · From the PDF
Figure 3 An illustration of four LLM role-playing environments: Software Development (§2.1.1), Game (§2.1.2), Medical Application (§2.1.3), and LLM as Evaluators (§2.1.4). For each environment, we provide a simple scenario with a task description ( red-bordered ) and relevant personas ( i.e. , roles; blue-bordered ). The dashed rectangle represents an example LLM role-playing prompt template. In addition to the above environments, past research also proposes general frameworks applicable to different environments (§5.1).
Figure 3 An illustration of four LLM role-playing environments: Software Development (§2.1.1), Game (§2.1.2), Medical Application (§2.1.3), and LLM as Evaluators (§2.1.4). For each environment, we provide a simple scenario with a task description ( red-bordered ) and relevant personas ( i.e. , roles; blue-bordered ). The dashed rectangle represents an example LLM role-playing prompt template. In addition to the above environments, past research also proposes general frameworks applicable to different environments (§5.1).CC-BY · From the PDF
Figure 4 An illustration of five types of personalized LLMs: Recommendation (§3.1), Search (§3.2), Education (§3.3), Healthcare (§3.4), and Dialogue (§3.5). On the left side, dashed rectangles are prompts, and solid rectangles are the responses of LLMs. On the right side, we depict multi-turn interactions between LLMs and users.
Figure 4 An illustration of five types of personalized LLMs: Recommendation (§3.1), Search (§3.2), Education (§3.3), Healthcare (§3.4), and Dialogue (§3.5). On the left side, dashed rectangles are prompts, and solid rectangles are the responses of LLMs. On the right side, we depict multi-turn interactions between LLMs and users.CC-BY · From the PDF

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

The publisher hasn’t deposited this paper’s reference list.

Research works printed but not linked: 17

  1. Tom DeMarco and Tim Lister. 2013. Peopleware: pro- ductive projects and teams. Addison-Wesley.
  2. Ameet Deshpande, EunJeong Hwang, Vishvak Mura- hari, Joon Sung Park, Diyi Yang, Ashish Sabharwal, Karthik Narasimhan, and Ashwin Kalyan, editors. 2024. Proceedings of the 1st Workshop on Person- alization of Generative AI Systems (PERSONALIZE 2024). Association for Computational Linguistics, St. Julians, Malta.
  3. Mihail Eric, Rahul Goel, Shachi Paul, Abhishek Sethi, Sanchit Agarwal, Shuyang Gao, Adarsh Kumar, Anuj Goyal, Peter Ku, and Dilek Hakkani-Tur. 2020. Mul- tiWOZ 2.1: A consolidated multi-domain dialogue dataset with state corrections and state tracking base- lines. In Proceedings of the Twelfth Language Re- sources and Evaluation Conference, pages 422-428, Marseille, France. European Language Resources Association. 19
  4. Jun Hu, Wenwen Xia, Xiaolu Zhang, Chilin Fu, We- ichang Wu, Zhaoxin Huan, Ang Li, Zuoli Tang, and Jun Zhou. 2024. Enhancing sequential recommenda- tion via llm-based semantic embedding learning. In Companion Proceedings of the ACM on Web Confer- ence 2024, pages 103-111.
  5. Daniil Huryn, William M. Hutsell, and Jinho D. Choi. 2022. Automatic generation of large-scale multi-turn dialogues from Reddit. In Proceedings of the 29th International Conference on Computational Linguis- tics, pages 3360-3373, Gyeongju, Republic of Korea. International Committee on Computational Linguis- tics.
  6. Guohao Li, Hasan Hammoud, Hani Itani, Dmitrii Khizbullin, and Bernard Ghanem. 2024a. Camel: Communicative agents for" mind" exploration of large language model society. Advances in Neural Information Processing Systems, 36. 2, 5, 8
  7. Lei Li, Yongfeng Zhang, and Li Chen. 2023a. Person- alized prompt learning for explainable recommen- dation. ACM Transactions on Information Systems, 41(4):1-26. 5, 18, 19
  8. Yanran Li, Hui Su, Xiaoyu Shen, Wenjie Li, Ziqiang Cao, and Shuzi Niu. 2017. DailyDialog: A manually labelled multi-turn dialogue dataset. In Proceedings of the Eighth International Joint Conference on Nat- ural Language Processing (Volume 1: Long Papers), pages 986-995, Taipei, Taiwan. Asian Federation of Natural Language Processing. 19
  9. Zekun Li, Baolin Peng, Pengcheng He, Michel Galley, Jianfeng Gao, and Xifeng Yan. 2024c. Guiding large language models via directional stimulus prompting. Advances in Neural Information Processing Systems, 36. 2, 7
  10. Jianmo Ni, Jiacheng Li, and Julian McAuley. 2019. Jus- tifying recommendations using distantly-labeled re- views and fine-grained aspects. In Proceedings of the 2019 conference on empirical methods in natural language processing and the 9th international joint conference on natural language processing (EMNLP- IJCNLP), pages 188-197. 18, 19
  11. M Swift and J Allen. 2010. Towards a personal health management assistant. Journal of biomedical infor- matics, 43(5):S13-S16.
  12. Ning Wu, Ming Gong, Linjun Shou, Shining Liang, and Daxin Jiang. 2023b. Large language models are diverse role-players for summarization evaluation. In CCF International Conference on Natural Language Processing and Chinese Computing, pages 695-707. Springer. 2, 4
  13. Xuhai Xu, Bingsheng Yao, Yuanzhe Dong, Saadia Gabriel, Hong Yu, James Hendler, Marzyeh Ghas- semi, Anind K Dey, and Dakuo Wang. 2024. Mental- llm: Leveraging large language models for mental health prediction via online text data. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 8(1):1-32.
  14. Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik R Narasimhan, and Yuan Cao. 2022b. React: Synergizing reasoning and acting in language models. In The Eleventh International Conference on Learning Representations. 3
  15. Saizheng Zhang, Emily Dinan, Jack Urbanek, Arthur Szlam, Douwe Kiela, and Jason Weston. 2018a. Per- sonalizing dialogue agents: I have a dog, do you have pets too? In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Vol- ume 1: Long Papers), pages 2204-2213, Melbourne, Australia. Association for Computational Linguistics. 2, 20 · DOI · Paper page
  16. Wanjun Zhong, Lianghong Guo, Qiqi Gao, He Ye, and Yanlin Wang. 2024. Memorybank: Enhancing large language models with long-term memory. Proceed- ings of the AAAI Conference on Artificial Intelligence, 38(17):19724-19731. · DOI · Paper page
  17. USER:] That is all I need to know. Thanks, goodbye. [SYSTEM:] Thank you so much for Cambridge TownInfo center. Have a great day! Table 5: An example of ToD modeling from the MultiWOZ dataset (Zang et al., 2020).

Cited by

Sorted by citations

Related

See all

Metadata

Authors
Tseng, Y.-M., Huang, Y.-C., Hsiao, T.-Y., Chen, W.-L., Huang, C.-W., Meng, Y., & Chen, Y.-N.
Affiliations
National Taiwan University (TW)
ORCID
Venue
Findings of the Association for Computational Linguistics: EMNLP 2024
Publisher
Association for Computational Linguistics
Published
2024
Pages
16612-16631
DOI
10.18653/v1/2024.findings-emnlp.969
OpenAlex
W4404781205
Cited by
87
References
0
Citations by year
2026: 47 · 2025: 35 · 2024: 5
FWCI
25.12
Type
conference paper
Language
en
Open access
Licence
cc-by
Locations
  • Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) (publishedVersion, cc-by)
  • arXiv (Cornell University) (submittedVersion, cc-by-sa)
  • arXiv (Cornell University)
Funders
  • National Taiwan University (113L900902, 113L900903)
  • National Science and Technology Council
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Speech and dialogue systems
  • Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
  • Physical Sciences › Computer Science › Human-Computer Interaction › Persona Design and Applications
Keywords
personas, large language models, LLM personalization, personality assessment, personalized search, LLM-as-a-judge

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review