Elamur Research⌕
☰
home / papers / 2026 / BILLY: Steering Large Language Models via Merging Persona Vectors for Creative Generation

BILLY: Steering Large Language Models via Merging Persona Vectors for Creative Generation

Paper · 2026 · Tsung-Min Pai, Jui-I Wang, Li-Chun Lu et al.

Summary

Multi-LLM setups, where several role-playing models debate an idea, tend to produce more creative output than a single model. They are slow and expensive, however, because many rounds of text must be generated and read. The authors propose BILLY, a training-free alternative that packs several viewpoints into one model. They build on the persona-vector pipeline of Chen et al. (2025) and a custom persona dataset. For each persona (e.g., Creative Professional, Environmentalist), they obtain an activation-space direction from responses generated under opposing system prompts. These responses are scored by an LLM judge and filtered by a threshold. The resulting vectors are combined offline into one composite vector, which is added to the model's activations during inference. On four creativity-oriented benchmarks, BILLY matches or beats single-model prompting and multi-LLM baselines while sharply cutting inference time and token cost. Projection analyses suggest steering controls the model's internal state better than multi-role prompting. This matters because it offers a cheap, interpretable way to obtain multi-perspective creativity.

Key points

  • The pipeline has three stages: contrastive extraction of persona vectors, offline fusion into one steering vector, and injection at inference. It requires no fine-tuning and no communication between models.
  • In projection analysis, multi-role prompting (CRE+ENV) shifted activations positively only along the creative axis and negatively along the environmentalist one. BILLY (CRE+ENV) was positive on both axes.
  • Single-persona vectors were selective, each mainly moving its own axis. The analysis used 10 neutral and 10 evaluation questions, with per-layer plots including Layer 20.
  • Blending effects are non-linear. CRE+ENV projects less on the creative axis than CRE alone, yet it scores higher on Originality. The environmentalist vector appears to contribute content, and the creative vector reframes it.
  • Limitation: vectors are merged by simple averaging. Learned or task-specific weighting is left for future work.

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

🎓Massachusetts Institute of TechnologyCambridge · United States211 works · 69,000 citations🎓National Taiwan UniversityTaipei · Taiwan29 works · 2,116 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 Tsung-Min Pai, Jui-I Wang, Li-Chun Lu, Shao-Hua Sun, Hung-yi Lee and Kai-Wei Chang. 2026. BILLY: Steering Large Language Models via Merging Persona Vectors for Creative Generation. Proceedings of the 19th Conference of the European Chapter 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.eacl-long.369

@inproceedings{pai2026billy,
  title = {{BILLY: Steering Large Language Models via Merging Persona Vectors for Creative Generation}},
  author = {Tsung-Min Pai, Jui-I Wang, Li-Chun Lu, Shao-Hua Sun, Hung-yi Lee and Kai-Wei Chang},
  booktitle = {Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.eacl-long.369},
  url = {https://doi.org/10.18653/v1/2026.eacl-long.369}
}

References

58 references, 16 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Safinah Ali, Ayat Abodayeh, Zahra Dhuliawala, Cynthia Breazeal, and Hae Won Park. 2025. Towards inclusive co-creative child-robot interaction: Can social robots support neurodivergent children’s creativity? In Proceedings of the 2025 ACM/IEEE International Conference on Human-Robot Interaction. find
  2. Yuanpu Cao, Tianrong Zhang, Bochuan Cao, Ziyi Yin, Lu Lin, Fenglong Ma, and Jinghui Chen. 2025. Personalized steering of large language models: versatile steering vectors through bi-directional preference optimization. In Proceedings of the 38th International Conference on Neural Information Process… find
  3. Mert Cemri, Melissa Z. Pan, Shuyi Yang, Lakshya A. Agrawal, Bhavya Chopra, Rishabh Tiwari, Kurt Keutzer, Aditya Parameswaran, Dan Klein, Kannan Ramchandran, Matei Zaharia, Joseph E. Gonzalez, and Ion Stoica. 2025. Why do multi-agent llm systems fail? find
  4. Tuhin Chakrabarty, Philippe Laban, Divyansh Agarwal, Smaranda Muresan, and Chien-Sheng Wu. 2023. Art or artifice? large language models and the false promise of creativity. arXiv:2309.14556. — [arXiv:2309.14556](https://arxiv.org/abs/2309.14556) find
  5. Chi-Min Chan, Weize Chen, Yusheng Su, Jianxuan Yu, Wei Xue, Shanghang Zhang, Jie Fu, and Zhiyuan Liu. 2024. Chateval: Towards better LLM-based evaluators through multi-agent debate. In The Twelfth International Conference on Learning Representations.
  6. Jiangjie Chen, Xintao Wang, Rui Xu, Siyu Yuan, Yikai Zhang, Wei Shi, Jian Xie, Shuang Li, Ruihan Yang, Tinghui Zhu, Aili Chen, Nianqi Li, Lida Chen, Caiyu Hu, Siye Wu, Scott Ren, Ziquan Fu, and Yanghua Xiao. 2024a. From persona to personalization: A survey on role-playing language agents.
  7. Jing Chen, Xinyu Zhu, Cheng Yang, Chufan Shi, Yadong Xi, Yuxiang Zhang, Junjie Wang, Jiashu Pu, Tian Feng, Yujiu Yang, and Rongsheng Zhang. 2024b. HoLLMwood: Unleashing the creativity of large language models in screenwriting via role playing. In Findings of the Association for Computational Linguis… find
  8. Runjin Chen, Andy Arditi, Henry Sleight, Owain Evans, and Jack Lindsey. 2025. Persona vectors: Monitoring and controlling character traits in language models. find
  9. Weize Chen, Yusheng Su, Jingwei Zuo, Cheng Yang, Chenfei Yuan, Chi-Min Chan, Heyang Yu, Yaxi Lu, Yi-Hsin Hung, Chen Qian, Yujia Qin, Xin Cong, Ruobing Xie, Zhiyuan Liu, Maosong Sun, and Jie Zhou. 2024c. Agentverse: Facilitating multi-agent collaboration and exploring emergent behaviors. In Internati…
  10. Bonnie Cramond, Juanita Matthews-Morgan, Deborah Bandalos, and Li Zuo. 2005. A report on the 40year follow-up of the torrance tests of creative thinking: Alive and well in the new millennium. Gifted Child Quarterly - GIFTED CHILD QUART. find
  11. Niall Creech, Natalia Criado Pacheco, and Simon Miles. 2021. Resource allocation in dynamic multiagent systems. find
  12. M. Diehl and W. Stroebe. 1987. Productivity loss in brainstorming groups: Toward the solution of a riddle. Journal of Personality and Social Psychology. find
  13. Shiying Ding, Xinyi Chen, Yan Fang, Wenrui Liu, Yiwu Qiu, and Chunlei Chai. 2023. Designgpt: Multi-agent collaboration in design. find
  14. Yilun Du, Shuang Li, Antonio Torralba, Joshua B. Tenenbaum, and Igor Mordatch. 2023. Improving factuality and reasoning in language models through multiagent debate. arXiv:2305.14325.
  15. Gregory J Feist. 1998. A meta-analysis of personality in scientific and artistic creativity. Personality and social psychology review. find
  16. Giorgio Franceschelli and Mirco Musolesi. 2023. On the creativity of large language models. arXiv:2304.00008. — [arXiv:2304.00008](https://arxiv.org/abs/2304.00008) find
  17. Carlos Gómez-Rodríguez and Paul Williams. 2023. A confederacy of models: a comprehensive evaluation of LLMs on creative writing. In Findings of the Association for Computational Linguistics: EMNLP. find
  18. Taicheng Guo, Xiuying Chen, Yaqi Wang, Ruidi Chang, Shichao Pei, Nitesh V Chawla, Olaf Wiest, and Xiangliang Zhang. 2024. Large language model based multi-agents: A survey of progress and challenges. In IJCAI.
  19. Shanshan Han, Qifan Zhang, Yuhang Yao, Weizhao Jin, and Zhaozhuo Xu. 2025. Llm multi-agent systems: Challenges and open problems. find
  20. Jessica He, Stephanie Houde, Gabriel E. Gonzalez, Darío Andrés Silva Moran, Steven I. Ross, Michael Muller, and Justin D. Weisz. 2024. Ai and the future of collaborative work: Group ideation with an llm in a virtual canvas. In Proceedings of the 3rd Annual Meeting of the Symposium on Human-Computer… find
  21. Sirui Hong, Mingchen Zhuge, Jonathan Chen, Xiawu Zheng, Yuheng Cheng, Jinlin Wang, Ceyao Zhang, Zili Wang, Steven Ka Shing Yau, Zijuan Lin, Liyang Zhou, Chenyu Ran, Lingfeng Xiao, Chenglin Wu, and Jürgen Schmidhuber. 2024. MetaGPT: Meta programming for a multi-agent collaborative framework. In The T…
  22. Yihan Hou, Manling Yang, Hao Cui, Lei Wang, Jie Xu, and Wei Zeng. 2024. C2ideas: Supporting creative interior color design ideation with a large language model. In Proceedings of the 2024 CHI conference on human factors in computing systems. find
  23. Yu-Kai Hung, Yun-Chien Huang, Ting-Yu Su, Yen-Ting Lin, Lung-Pan Cheng, Bryan Wang, and Shao-Hua Sun. 2025. Simtube: Generating simulated video comments through multimodal ai and user personas. In International Conference on Intelligent User Interfaces. find
  24. Joel Jang, Seungone Kim, Bill Yuchen Lin, Yizhong Wang, Jack Hessel, Luke Zettlemoyer, Hannaneh Hajishirzi, Yejin Choi, and Prithviraj Ammanabrolu. 2023. Personalized soups: Personalized large language model alignment via post-hoc parameter merging. find
  25. Ole Jorgensen, Dylan Cope, Nandi Schoots, and Murray Shanahan. 2023. Improving activation steering in language models with mean-centring. find
  26. Jan Marco Leimeister. 2010. Collective intelligence. Business & Information Systems Engineering. find
  27. Kenneth Li, Oam Patel, Fernanda Viégas, Hanspeter Pfister, and Martin Wattenberg. 2023. Inference-time intervention: eliciting truthful answers from a language model. In Proceedings of the 37th International Conference on Neural Information Processing Systems.
  28. Ren-Wei Liang, Chin Ting Hsu, Chan-Hung Yu, Saransh Agrawal, Shih-Cheng Huang, Chieh-Yen Lin, Shang-Tse Chen, Kuan-Hao Huang, and Shao-Hua Sun. 2026. Adaptive Helpfulness–Harmlessness Alignment with Preference Vectors . In Conference of the European Chapter of the Association for Computational Lingu… find
  29. Max Liu, Chan-Hung Yu, Wei-Hsu Lee, Cheng-Wei Hung, Yen-Chun Chen, and Shao-Hua Sun. 2025. Synthesizing programmatic reinforcement learning policies with large language model guided search. In International Conference on Learning Representations. find
  30. Yiren Liu, Si Chen, Haocong Cheng, Mengxia Yu, Xiao Ran, Andrew Mo, Yiliu Tang, and Yun Huang. 2024. Coquest: Exploring research question co-creation with an llm-based agent. find
  31. Zijun Liu, Yanzhe Zhang, Peng Li, Yang Liu, and Diyi Yang. 2023. Dynamic llm-agent network: An llm-agent collaboration framework with agent team optimization. arXiv:2310.02170. find
  32. Li-Chun Lu, Shou-Jen Chen, Tsung-Min Pai, Chan-Hung Yu, Hung-yi Lee, and Shao-Hua Sun. 2024. LLM Discussion: Enhancing the creativity of large language models via discussion framework and role-play. In Conference on Language Modeling. find
  33. Li-Chun Lu, Miri Liu, Pin Chun Lu, Yufei Tian, Shao-Hua Sun, and Nanyun Peng. 2026. Rethinking Creativity Evaluation: A Critical Analysis of Existing Creativity Evaluations. In Conference of the European Chapter of the Association for Computational Linguistics.
  34. Bart Nooteboom, Wim Van Haverbeke, Geert Duysters, Victor Gilsing, and Ad van den Oord. 2007. Optimal cognitive distance and absorptive capacity. Research Policy. find
  35. Nina Panickssery, Nick Gabrieli, Julian Schulz, Meg Tong, Evan Hubinger, and Alexander Matt Turner. 2024. Steering llama 2 via contrastive activation addition. find
  36. Max Peeperkorn, Tom Kouwenhoven, Dan Brown, and Anna Jordanous. 2024. Is temperature the creativity parameter of large language models? find
  37. Chenglei Si, Diyi Yang, and Tatsunori Hashimoto. 2025. Can LLMs generate novel research ideas? a large-scale human study with 100+ NLP researchers. In The Thirteenth International Conference on Learning Representations.
  38. Tian Liang, Zhiwei He, Wenxiang Jiao, Xing Wang, Yan Wang, Rui Wang, Yujiu Yang, Shuming Shi, and Zhaopeng Tu. 2024. Encouraging divergent thinking in large language models through multi-agent debate. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing.
  39. Dean Keith Simonton. 2004. Creativity in science: Chance, logic, genius, and zeitgeist. Cambridge University Press. find
  40. Gionnieve Lim and Simon T Perrault. 2024. Rapid aideation: Generating ideas with the self and in collaboration with large language models. arXiv preprint arXiv:2403.12928. — [arXiv:2403.12928](https://arxiv.org/abs/2403.12928) find
  41. Haoyang Su, Renqi Chen, Shixiang Tang, Zhenfei Yin, Xinzhe Zheng, Jinzhe Li, Biqing Qi, Qi Wu, Hui Li, Wanli Ouyang, Philip Torr, Bowen Zhou, and Nanqing Dong. 2025. Many heads are better than one: Improved scientific idea generation by a LLM-based multi-agent system. In Proceedings of the 63rd Annu…
  42. Yi-Cheng Lin, Kang-Chieh Chen, Zhe-Yan Li, Tzu-Heng Wu, Tzu-Hsuan Wu, Kuan-Yu Chen, Hung yi Lee, and Yun-Nung Chen. 2025. Creativity in llm-based multi-agent systems: A survey. find
  43. Ivan Dale Steiner. 1972. Group process and productivity, volume 1. Academic press New York. find
  44. Qiushi Sun, Zhangyue Yin, Xiang Li, Zhiyong Wu, Xipeng Qiu, and Lingpeng Kong. 2023. Corex: Pushing the boundaries of complex reasoning through multi-model collaboration. arXiv:2310.00280. find
  45. Seungjong Sun, Seo Yeon Baek, and Jang Hyun Kim. 2025. Personality vector: Modulating personality of large language models by model merging. find
  46. Ellis Paul Torrance. 1966. Torrance Tests of Creative Thinking. Norms-Technical Manual. Research Edition. Verbal Tests Forms a and B. Figural Tests Forms a and B. Personnel Press. find
  47. Khanh-Tung Tran, Dung Dao, Minh-Duong Nguyen, Quoc-Viet Pham, Barry O’Sullivan, and Hoang D Nguyen. 2025. Multi-agent collaboration mechanisms: A survey of llms. arXiv preprint arXiv:2501.06322. find
  48. Alexander Matt Turner, Lisa Thiergart, Gavin Leech, David Udell, Juan J. Vazquez, Ulisse Mini, and Monte MacDiarmid. 2024. Steering language models with activation engineering.
  49. Wallach, M. A., and Kogan, N. 1965. Modes of thinking in young children: A study of the creativity-intelligence distinction. American Psychological Association. find
  50. Zonglin Yang, Xinya Du, Junxian Li, Jie Zheng, Soujanya Poria, and Erik Cambria. 2024b. Large language models for automated open-domain scientific hypotheses discovery. In Findings of the Association for Computational Linguistics ACL 2024. find
  51. Jesse Zhang, Karl Pertsch Jiahui Zhang, Ziyi Liu, Xiang Ren, Minsuk Chang, Shao-Hua Sun, and Joseph J. Lim. 2023. Bootstrap your own skills: Learning to solve new tasks with large language model guidance. In Conference on Robot Learning. find
  52. Jinghao Zhang, Yuting Liu, Wenjie Wang, Qiang Liu, Shu Wu, Liang Wang, and Tat-Seng Chua. 2025. Personalized text generation with contrastive activation steering. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Association for Compu…
  53. Andy Zou, Long Phan, Sarah Chen, James Campbell, Phillip Guo, Richard Ren, Alexander Pan, Xuwang Yin, Mantas Mazeika, Ann-Kathrin Dombrowski, Shashwat Goel, Nathaniel Li, Michael J. Byun, Zifan Wang, Alex Mallen, Steven Basart, Sanmi Koyejo, Dawn Song, Matt Fredrikson, and 2 others. 2025. Representa…
  54. Zhenhailong Wang, Shaoguang Mao, Wenshan Wu, Tao Ge, Furu Wei, and Heng Ji. 2024. Unleashing the emergent cognitive synergy in large language models: A task-solving agent through multi-persona self-collaboration. In Proceedings of the 2024 Conference of the North American Chapter of the Association…
  55. Haoyuan Wu, Haisheng Zheng, Zhuolun He, and Bei Yu. 2025a. Divergent thoughts toward one goal: LLM-based multi-agent collaboration system for electronic design automation. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: H… find
  56. Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Beibin Li, Erkang Zhu, Li Jiang, Xiaoyun Zhang, Shaokun Zhang, Jiale Liu, Ahmed Hassan Awadallah, Ryen W White, Doug Burger, and Chi Wang. 2023. Autogen: Enabling next-gen llm applications via multi-agent conversation. In arXiv:2308.08155.
  57. Zhengxuan Wu, Aryaman Arora, Atticus Geiger, Zheng Wang, Jing Huang, Dan Jurafsky, Christopher D Manning, and Christopher Potts. 2025b. Axbench: Steering llms? even simple baselines outperform sparse autoencoders. arXiv preprint arXiv:2501.17148. find
  58. Enneng Yang, Li Shen, Guibing Guo, Xingwei Wang, Xiaochun Cao, Jie Zhang, and Dacheng Tao. 2024a. Model merging in llms, mllms, and beyond: Methods, theories, applications and opportunities. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/pai-2026-billy-steering-large-language-models/