Elamur Research⌕
☰
home / papers / 2026 / Are Emotion and Rhetoric Neurons in LLM? Neuron Recognition and Adaptive Masking for Emotion-Rhetoric Prediction Steering

Are Emotion and Rhetoric Neurons in LLM? Neuron Recognition and Adaptive Masking for Emotion-Rhetoric Prediction Steering

Paper · 2026 · Li Zheng, Xin Zhang, Shuyi He et al.

Summary

The paper asks whether large language models contain individual neurons tied to specific emotions and rhetorical devices, and whether these neurons can be reliably tested and used for control. The authors study FFN neurons in Llama-3.1-8B-Instruct for six emotions (happiness, sadness, anger, fear, surprise, disgust) and four devices (metaphor, hyperbole, humor, sarcasm). They select candidate neurons by combining activation-frequency counts, probability normalization and entropy-based selectivity filtering.

Standard verification, which forces neurons to zero or replaces them with mean values, sometimes leaves accuracy unchanged or even raises it. To fix this, the authors propose adaptive masking. It picks neurons dynamically, weakens their activations instead of removing them, and tunes the intervention through feedback. This produces a consistent accuracy drop, which makes causal claims about neuron function more trustworthy.

Injecting neuron activations can push predictions on non-target inputs toward a chosen class. Adding rhetoric neurons also improves emotion recognition. The work offers a route to fine-grained, interpretable steering of affective and rhetorical behaviour in LLMs.

Key points

  • Model: Llama-3.1-8B-Instruct, which has 32 Transformer blocks. The analysis targets neurons in the FFN layers.
  • Evaluation uses five datasets: DailyDialogue, HYPO, TroFi, IAC-v2 and ColBERT.
  • Emotion and rhetoric neurons are concentrated and most strongly activated in the upper layers. Masking the top layers hurts performance more than masking middle or bottom layers.
  • Rhetoric-neuron signals make emotional features easier to separate. Combining emotion and rhetoric neurons beat both emotion neurons alone and external optimization (prompting or fine-tuning) on DailyDialogue.
  • Steering success is measured as the share of samples classified into the target class after injection. Coverage rose for all emotion categories and also for rhetoric. A stated limitation is the reliance on static functional vectors rather than context-aware adjustment.

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

🎓Wuhan UniversityWuhan · China47 works · 1,246 citations🎓RMIT UniversityMelbourne · Australia19 works · 704 citations

 Fields, topics and concepts

📚 Shelves 6The 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 worksPsychology · 1,649 worksExperimental and Cognitive Psychology · 930 works

🧭 Topics 3Research subjects and languages the work is about.Language models & pretraining · 3,894 worksSentiment & emotion · 1,035 worksInterpretability · 149 works

Cite

Cite as Li Zheng, Xin Zhang, Shuyi He, Fei Li, Chong Teng, Jiang-Ming Yang, Donghong Ji and Zhuang Li. 2026. Are Emotion and Rhetoric Neurons in LLM? Neuron Recognition and Adaptive Masking for Emotion-Rhetoric Prediction Steering. 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.193

@inproceedings{zheng2026are,
  title = {{Are Emotion and Rhetoric Neurons in LLM? Neuron Recognition and Adaptive Masking for Emotion-Rhetoric Prediction Steering}},
  author = {Li Zheng, Xin Zhang, Shuyi He, Fei Li, Chong Teng, Jiang-Ming Yang, Donghong Ji and Zhuang Li},
  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.193},
  url = {https://doi.org/10.18653/v1/2026.acl-long.193}
}

References

48 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Issa Annamoradnejad and Gohar Zoghi. 2024. Colbert: Using bert sentence embedding in parallel neural networks for computational humor. Expert Systems with Applications, 249:123685. find
  2. Vinamra Benara, Chandan Singh, John X Morris, Richard J Antonello, Ion Stoica, Alexander G Huth, and Jianfeng Gao. 2024. Crafting interpretable embeddings for language neuroscience by asking llms questions. Advances in neural information processing systems, 37:124137. find
  3. Julia Birke and Anoop Sarkar. 2006. A clustering approach for nearly unsupervised recognition of nonliteral language. In 11th Conference of the European chapter of the association for computational linguistics, pages 329–336. find
  4. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901.
  5. Carlos Busso, Murtaza Bulut, Chi-Chun Lee, Abe Kazemzadeh, Emily Mower, Samuel Kim, Jeannette N Chang, Sungbok Lee, and Shrikanth S Narayanan. 2008. Iemocap: Interactive emotional dyadic motion capture database. Language resources and evaluation, 42(4):335–359. find
  6. Santiago Castro, Devamanyu Hazarika, Verónica Pérez-Rosas, Roger Zimmermann, Rada Mihalcea, and Soujanya Poria. 2019. Towards multimodal sarcasm detection (an _obviously_ perfect paper). In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4619–4629.
  7. Alessio Cocchieri, Luca Ragazzi, Paolo Italiani, Giuseppe Tagliavini, and Gianluca Moro. 2025. “what do you call a dog that is incontrovertibly true? dogma”: Testing llm generalization through humor. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1… find
  8. Yihe Deng, Weitong Zhang, Zixiang Chen, and Quanquan Gu. 2023. Rephrase and respond: Let large language models ask better questions for themselves. arXiv preprint arXiv:2311.04205.
  9. Dario Di Palma, Alessandro De Bellis, Giovanni Serve-dio, Vito Walter Anelli, Fedelucio Narducci, and Tommaso Di Noia. 2025. Llamas have feelings too: Unveiling sentiment and emotion representations in llama models through probing. arXiv preprint arXiv:2505.16491.
  10. Tao Feng, Lizhen Qu, Zhuang Li, Haolan Zhan, Yuncheng Hua, and Reza Haf. 2024. IMO: Greedy layer-wise sparse representation learning for out-of-distribution text classification with pre-trained models. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume… find
  11. Susan R Fussell and Mallie M Moss. 2014. Figurative language in emotional communication. In Social and cognitive approaches to interpersonal communication, pages 113–141. Psychology Press. find
  12. Palaash Goel, Dushyant Singh Chauhan, and Md Shad Akhtar. 2025. Target-augmented shared fusion-based multimodal sarcasm explanation generation. In Findings of the Association for Computational Linguistics: NAACL 2025, pages 8480–8493. find
  13. Nabil Hossain, John Krumm, Tanvir Sajed, and Henry Kautz. 2020. Stimulating creativity with funlines: A case study of humor generation in headlines. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics: System Demonstrations, pages 256–262. find
  14. Zhiting Hu, Zichao Yang, Xiaodan Liang, Ruslan Salakhutdinov, and Eric P Xing. 2017. Toward controlled generation of text. In International conference on machine learning, pages 1587–1596. PMLR.
  15. Yuanxiang Huangfu, Peifeng Li, Yaxin Fan, and Qiaoming Zhu. 2025. Non-emotion-centric empathetic dialogue generation. In Proceedings of the 31st International Conference on Computational Linguistics, pages 989–999. find
  16. Hiruni Maleesa Jayasinghe, Kok Wai Wong, and Anupiya Nugaliyadde. 2025. A systematic review of interpretability and explainability for speech emotion features in automatic speech emotion recognition. Pattern recognition, page 112122. find
  17. Minseo Kim, Taemin Kim, Thu Hoang Anh Vo, Yu-gyeong Jung, and Uichin Lee. 2025. Exploring modular prompt design for emotion and mental health recognition. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems, pages 1–18. find
  18. David Konstan. 2007. Rhetoric and emotion. A companion to Greek rhetoric, 411:25. find
  19. Jaewook Lee, Woojin Lee, Oh-Woog Kwon, and Harksoo Kim. 2025. Do large language models have “emotion neurons”? investigating the existence and role. In Findings of the Association for Computational Linguistics: ACL 2025, pages 15617–15639. find
  20. 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 Natural Language Processing (Volume 1: Long Papers), pages 986–995.
  21. Zhipeng Li, Binglin Wu, Yingyi Zhang, Xianneng Li, Kai Li, and Weizhi Chen. 2025. Cusmer: Multimodal intent recognition in customer service via data augment and llm merge. In Companion Proceedings of the ACM on Web Conference 2025, pages 3058– 3062. find
  22. Jian Liao, Yu Feng, Yujin Zheng, Jun Zhao, Suge Wang, and Jianxing Zheng. 2025. My words imply your opinion: Reader agent-based propagation enhancement for personalized implicit emotion analysis. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Lo… find
  23. Shereen Oraby, Vrindavan Harrison, Lena Reed, Ernesto Hernandez, Ellen Riloff, and Marilyn Walker. 2016. Creating and characterizing a diverse corpus of sarcasm in dialogue. In Proceedings of the 17th annual meeting of the special interest group on discourse and dialogue, pages 31–41.
  24. Jiao Ou, Junda Lu, Che Liu, Yihong Tang, Fuzheng Zhang, Di Zhang, and Kun Gai. 2024. Dialogbench: Evaluating llms as human-like dialogue systems. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume… find
  25. Alec Radford, Rafal Jozefowicz, and Ilya Sutskever. 2018. Learning to generate reviews and discovering sentiment.
  26. Rohit Rajakumar and Mihai Boicu. 2025. Evaluating lightweight transformer models for rhetorical element classification in student essays. Journal of Student-Scientists’ Research, 7. find
  27. Elvis Saravia, Hsien-Chi Toby Liu, Yen-Hao Huang, Junlin Wu, and Yi-Shin Chen. 2018. Carer: Contextualized affect representations for emotion recognition. In Proceedings of the 2018 conference on empirical methods in natural language processing, pages 3687–3697.
  28. Stephen D Smith and Jennifer Kornelsen. 2011. Emotion-dependent responses in spinal cord neurons: a spinal fmri study. NeuroImage, 58(1):269–274. find
  29. Kevin Stowe, Tuhin Chakrabarty, Nanyun Peng, Smaranda Muresan, and Iryna Gurevych. 2021. Metaphor generation with conceptual mappings. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing… find
  30. Hanchen Su, Wei Luo, Yashar Mehdad, Wei Han, Elaine Liu, Wayne Zhang, Mia Zhao, and Joy Zhang. 2025. Llm-friendly knowledge representation for customer support. In Proceedings of the 31st International Conference on Computational Linguistics: Industry Track, pages 496–504. find
  31. Zhiwei Liu, Kailai Yang, Qianqian Xie, Tianlin Zhang, and Sophia Ananiadou. 2024. Emollms: A series of emotional large language models and annotation tools for comprehensive affective analysis. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 5487– 5496. find
  32. Ala N. Tak, Amin Banayeeanzade, Anahita Bolourani, Mina Kian, Robin Jia, and Jonathan Gratch. 2025. Mechanistic interpretability of emotion inference in large language models. In Findings of the Association for Computational Linguistics: ACL 2025, pages 13090–13120. Association for Computational Lin… find
  33. Michael Mohler, Mary Brunson, Bryan Rink, and Marc Tomlinson. 2016. Introducing the lcc metaphor datasets. In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC’16), pages 4221–4227. find
  34. Enrica Troiano, Carlo Strapparava, Gözde Özbal, and Serra Sinem Tekiroğlu. 2018. A computational exploration of exaggeration. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 3296–3304. find
  35. Yuchi Wang, Junliang Guo, Jianhong Bai, Runyi Yu, Tianyu He, Xu Tan, Xu Sun, and Jiang Bian. 2025. Instructavatar: Text-guided emotion and motion control for avatar generation. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 8132–8140. find
  36. Li Zheng, Hao Fei, Fei Li, Bobo Li, Lizi Liao, Donghong Ji, and Chong Teng. 2024. Reverse multi-choice dialogue commonsense inference with graph-of-thought. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 19688– 19696. find
  37. Zhikun Wu, Thomas Weber, and Florian Müller. 2025. One does not simply meme alone: Evaluating co-creativity between llms and humans in the generation of humor. In Proceedings of the 30th International Conference on Intelligent User Interfaces, pages 1082–1092. find
  38. Li Zheng, Donghong Ji, Fei Li, Hao Fei, Shengqiong Wu, Jingye Li, Bobo Li, and Chong Teng. 2025b. Ec-qed: emotion-cause quadruple extraction in dialogs. IEEE Transactions on Audio, Speech and Language Processing. find
  39. Zhenyu Xu, Hailin Xu, Zhouyang Lu, Yingying Zhao, Rui Zhu, Yujiang Wang, Mingzhi Dong, Yuhu Chang, Qin Lv, Robert P Dick, et al. 2024. Can large language models be good companions? an llm-based eyewear system with conversational common ground. Proceedings of the ACM on Interactive, Mobile, Wearable… find
  40. Senqi Yang, Dongyu Zhang, Jing Ren, Ziqi Xu, Xiuzhen Jenny Zhang, Yiliao Song, Hongfei Lin, and Feng Xia. 2025a. Cultural bias matters: A cross-cultural benchmark dataset and sentiment-enriched model for understanding multimodal metaphors. In Proceedings of the 63rd Annual Meeting of the Association… find
  41. Yang Yang, Xunde Dong, and Yupeng Qiang. 2025b. Mse-adapter: A lightweight plugin endowing llms with the capability to perform multimodal sentiment analysis and emotion recognition. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 25642–25650. find
  42. Li Zheng, Tengyue Song, Yuzhe Ding, Xiaorui Wu, Fei Li, Dongdong Xie, Jinbo Li, Chong Teng, and Donghong Ji. 2025c. Improving emotion and intent understanding in multimodal conversations with progressive interaction. IEEE Transactions on Affective Computing. find
  43. Li Zheng, Sihang Wang, Hao Fei, Zuquan Peng, Fei Li, Jianming Fu, Chong Teng, and Donghong Ji. 2025d. Enhancing hyperbole and metaphor detection with their bidirectional dynamic interaction and emotion knowledge. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguist… find
  44. Shanshan Zhong, Zhongzhan Huang, Shanghua Gao, Wushao Wen, Liang Lin, Marinka Zitnik, and Pan Zhou. 2024. Let’s think outside the box: Exploring leap-of-thought in large language models with creative humor generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogniti… find
  45. Jiecheng Zhang, CL Philip Chen, Shuzhen Li, and Tong Zhang. 2025a. Incongruity-aware tension field network for multi-modal sarcasm detection. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 14499–14508. find
  46. Xinjie Zhang, Tenggan Zhang, Lei Sun, Jinming Zhao, and Qin Jin. 2025b. Exploring interpretability in deep learning for affective computing: a comprehensive review. ACM Transactions on Multimedia Computing, Communications and Applications. find
  47. Yunxiang Zhang and Xiaojun Wan. 2022. Mover: Mask, over-generate and rank for hyperbole generation. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 6018–6030. find
  48. Li Zheng, Hao Fei, Ting Dai, Zuquan Peng, Fei Li, Huisheng Ma, Chong Teng, and Donghong Ji. 2025a. Multi-granular multimodal clue fusion for meme understanding. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 26057– 26065. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zheng-2026-are-emotion-rhetoric-neurons-llm/