Elamur Research⌕
☰
home / papers / 2026 / DPN-LE: Dual Personality Neuron Localization and Editing for Large Language Models

DPN-LE: Dual Personality Neuron Localization and Editing for Large Language Models

Paper · 2026 · Lifan Zheng, Xue Yang, Jiawei Chen et al.

Summary

Neuron-editing approaches to steering the personality of large language models, such as NPTI, alter tens of thousands of MLP neurons per trait and harm general abilities. The authors first measure this cost on LLaMA-3-8B-Instruct using GSM8K, HotpotQA and TriviaQA. They also run a PCA analysis of activations. Two findings stand out: individual neurons serve both personality and general knowledge, and high and low poles of a Big Five trait occupy clearly separated regions of activation space. Building on this, they introduce DPN-LE, a training-free technique. It forms layer-wise steering vectors from mean MLP activation differences between high-trait and low-trait samples. It then keeps only neurons that pass two tests: a Cohen's d effect-size threshold and a top-quantile activation-magnitude threshold. Sparse additive interventions on these neurons, in a uniform variant and a layer-weighted variant (DPN-LE w), control personality at inference time. Personality scores stay close to NPTI while edits drop by roughly 97%, which helps preserve reasoning. This makes the method useful for role-play and survey-simulation applications that require a capable model.

Key points

  • Preliminary analysis with NPTI (γ = 1.4) cut GSM8K accuracy from a baseline of 75.36 by 5.16 to 66.03 points. The average losses were 16.00 points for high-trait editing and 40.79 points for low-trait editing.
  • DPN-LE modifies about 660–750 neurons per trait direction (~0.5%), compared with about 21–22k for NPTI. This is an average reduction of 96.7%. Each trait needs only 1,000 contrastive sample pairs.
  • On LLaMA-3-8B-Instruct the average trait scores were:
    • DPN-LE: 9.11, with fluency of 9.08
    • DPN-LE w: 9.10
    • NPTI: 9.43
    • Simple Prompt: 9.35
    • P2: 8.69
    • PAS: 6.93
  • Ablations identify γ in the range 0.8–1.0 and the Q995 quantile (0.5% of neurons per layer) as the best trade-off. DPN-LE w holds fluency better at strong γ. Generalization is confirmed on Qwen2.5-7B-Instruct.
  • Limitations:
    • DPN-LE w still causes notable GSM8K drops for Extraversion-low (−17.89%) and Neuroticism-high (−11.37%).
    • Alignment on IPIP-NEO-300 is weaker than PAS and NPTI.
    • Only single traits are edited.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksComputer Vision and Pattern Recognition · 3,486 works

🧭 Topics 1Research subjects and languages the work is about.Language models & pretraining · 3,894 works

Cite

Cite as Lifan Zheng, Xue Yang, Jiawei Chen, Chenyan WU, Jingyuan Zhang, Fanheng Kong, Xinyi Zeng, Xiang Chen and Yu Tian. 2026. DPN-LE: Dual Personality Neuron Localization and Editing for Large Language Models. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.1528

@inproceedings{zheng2026dpn,
  title = {{DPN-LE: Dual Personality Neuron Localization and Editing for Large Language Models}},
  author = {Lifan Zheng, Xue Yang, Jiawei Chen, Chenyan WU, Jingyuan Zhang, Fanheng Kong, Xinyi Zeng, Xiang Chen and Yu Tian},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1528},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1528}
}

References

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

  1. David Bau, Jun-Yan Zhu, Hendrik Strobelt, Agata Lapedriza, Bolei Zhou, and Antonio Torralba. 2020. Understanding the role of individual units in a deep neural network. Proceedings of the National Academy of Sciences, 117(48):30071–30078. find
  2. Xubo Cao and Michal Kosinski. 2024. Large language models know how the personality of public figures is perceived by the general public. Scientific Reports, 14(1):6735. find
  3. Jiawei Chen, Yang Yang, Chao Yu, Yu Tian, Zhi Cao, Xue Yang, Linghao Li, Hang Su, and Zhaoxia Yin. 2025. Red teaming large reasoning models. arXiv preprint arXiv:2512.00412.
  4. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, and 1 others. 2021. Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168.
  5. Damai Dai, Li Dong, Yaru Hao, Zhifang Sui, Baobao Chang, and Furu Wei. 2022. Knowledge neurons in pretrained transformers. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 8493– 8502. find
  6. Jia Deng, Tianyi Tang, Yanbin Yin, Wenhao Yang, Wayne Xin Zhao, and Ji-Rong Wen. 2024. Neuron-based personality trait induction in large language models. arXiv preprint arXiv:2410.12327. — [arXiv:2410.12327](https://arxiv.org/abs/2410.12327) find
  7. John M Digman. 1990. Personality structure: Emergence of the five-factor model. Annual review of psychology, 41(1):417–440. find
  8. Florian Dorner, Tom Sühr, Samira Samadi, and Augustin Kelava. 2023. Do personality tests generalize to large language models? In Socially Responsible Language Modelling Research. find
  9. Lewis R Goldberg and 1 others. 1999. A broad-bandwidth, public domain, personality inventory measuring the lower-level facets of several five-factor models. Personality psychology in Europe, 7(1):7– 28. find
  10. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, and 1 others. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783.
  11. Jen-tse Huang, Wenxuan Wang, Eric John Li, Man Ho Lam, Shujie Ren, Youliang Yuan, Wenxiang Jiao, Zhaopeng Tu, and Michael Lyu. 2023. On the humanity of conversational ai: Evaluating the psychological portrayal of llms. In The Twelfth International Conference on Learning Representations. find
  12. Guangyuan Jiang, Manjie Xu, Song-Chun Zhu, Wenjuan Han, Chi Zhang, and Yixin Zhu. 2023. Evaluating and inducing personality in pre-trained language models. Advances in Neural Information Processing Systems, 36:10622–10643.
  13. Hang Jiang, Xiajie Zhang, Xubo Cao, Cynthia Breazeal, Deb Roy, and Jad Kabbara. 2024. Personallm: Investigating the ability of large language models to express personality traits. In Findings of the association for computational linguistics: NAACL 2024, pages 3605–3627.
  14. Mandar Joshi, Eunsol Choi, Daniel S Weld, and Luke Zettlemoyer. 2017. Triviaqa: A large scale distantly supervised challenge dataset for reading comprehension. arXiv preprint arXiv:1705.03551.
  15. Cheng Li, Ziang Leng, Chenxi Yan, Junyi Shen, Hao Wang, Weishi Mi, Yaying Fei, Xiaoyang Feng, Song Yan, HaoSheng Wang, and 1 others. 2023a. Chatharuhi: Reviving anime character in reality via large language model. arXiv preprint arXiv:2308.09597.
  16. Kenneth Li, Oam Patel, Fernanda Viégas, Hanspeter Pfister, and Martin Wattenberg. 2023b. Inference-time intervention: Eliciting truthful answers from a language model. Advances in Neural Information Processing Systems, 36:41451–41530.
  17. Robert R McCrae and Oliver P John. 1992. An introduction to the five-factor model and its applications. Journal of personality, 60(2):175–215. find
  18. Kevin Meng, David Bau, Alex Andonian, and Yonatan Belinkov. 2022a. Locating and editing factual associations in gpt. Advances in neural information processing systems, 35:17359–17372. find
  19. Kevin Meng, Arnab Sen Sharma, Alex Andonian, Yonatan Belinkov, and David Bau. 2022b. Mass-editing memory in a transformer. arXiv preprint arXiv:2210.07229. find
  20. Jesse Mu and Jacob Andreas. 2020. Compositional explanations of neurons. Advances in Neural Information Processing Systems, 33:17153–17163. find
  21. Keyu Pan and Yawen Zeng. 2023. Do llms possess a personality? making the mbti test an amazing evaluation for large language models. arXiv preprint arXiv:2307.16180.
  22. Joon Sung Park, Joseph O’Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S Bernstein. 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
  23. Nina Rimsky, Nick Gabrieli, Julian Schulz, Meg Tong, Evan Hubinger, and Alexander Turner. 2024. Steering llama 2 via contrastive activation addition. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 15504–15522. find
  24. Gregory Serapio-García, Mustafa Safdari, Clément Crepy, Luning Sun, Stephen Fitz, Marwa Abdulhai, Aleksandra Faust, and Maja Matarić. 2023. Personality traits in large language models. Research Square Preprint.
  25. Yunfan Shao, Linyang Li, Junqi Dai, and Xipeng Qiu. 2023. Character-llm: A trainable agent for role-playing. arXiv preprint arXiv:2310.10158.
  26. Alexander Matt Turner, Lisa Thiergart, Gavin Leech, David Udell, Ulisse Mini, and Monte MacDiarmid. 2024. Activation addition: Steering language models without optimization. arXiv preprint arXiv:2308.10248.
  27. Noah Wang, Zy Peng, Haoran Que, Jiaheng Liu, Wangchunshu Zhou, Yuhan Wu, Hongcheng Guo, Ruitong Gan, Zehao Ni, Jian Yang, and 1 others. 2024. Rolellm: Benchmarking, eliciting, and enhancing role-playing abilities of large language models. In Findings of the Association for Computational Linguistics:…
  28. Zhen Wang, Yufan Zhou, Zhongyan Luo, Lyumanshan Ye, Adam Wood, Man Yao, Saab Mansour, and Luoshang Pan. 2025. Deeppersona: A generative engine for scaling deep synthetic personas. arXiv preprint arXiv:2511.07338. — [arXiv:2511.07338](https://arxiv.org/abs/2511.07338) find
  29. Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William Cohen, Ruslan Salakhutdinov, and Christopher D Manning. 2018. Hotpotqa: A dataset for diverse, explainable multi-hop question answering. In Proceedings of the 2018 conference on empirical methods in natural language processing, pages 2369–…
  30. Minjun Zhu, Yixuan Weng, Linyi Yang, and Yue Zhang. 2024. Personality alignment of large language models. arXiv preprint arXiv:2408.11779.
  31. Andy Zou, Long Phan, Sarah Chen, James Campbell, Phillip Guo, Richard Ren, Alexander Pan, Xuwang Yin, Mantas Mazeika, Ann-Kathrin Dombrowski, and 1 others. 2023. Representation engineering: A top-down approach to ai transparency. arXiv preprint arXiv:2310.01405.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zheng-2026-dpn-le-dual-personality-neuron/