Decoding and controlling emotion in LLMs through human-aligned representational geometry with enhanced interpretability
Authors:
- Xiuwen WuXWXiuwen Wu
- Hao WangHWHao Wang
- Zhiang YanZYZhiang Yan
- Xiaohan TangXTXiaohan Tang
- Pengfei XuPXPengfei Xu
- Wai-Ting SiokWSWai-Ting Siok
- Ping LiPLPing Li
- Jia-Hong GaoJGJia-Hong Gao
- Bingjiang LyuBLBingjiang Lyu
- Lang QinLQLang Qin
- 14 May 2026
- journal article
- Published by Elsevier in Computers in Human Behavior
- Vol. 183
Abstract
No abstract availableFunding Information
- National Center for Protein Sciences, Peking University
- Fundamental Research Funds for the Central Universities
- Peking University (8430503616)
- Peking University (050575)
- Peking University (050550)
- Peking University (7100604651)
- National University's Basic Research Foundation of China
This publication has 64 references indexed in Scilit:
- Evaluating the alignment of AI with human emotionsAdvanced Design Research, 2024
- Language-specific representation of emotion-concept knowledge causally supports emotion inferenceiScience, 2024
- In praise of empathic AITrends in Cognitive Sciences, 2023
- AI as a Mental Health Therapist for AdolescentsJAMA Pediatrics, 2023
- Can AI language models replace human participants?Trends in Cognitive Sciences, 2023
- A systematic review on affective computing: emotion models, databases, and recent advancesInformation Fusion, 2022
- Semantic Space Theory: A Computational Approach to EmotionTrends in Cognitive Sciences, 2021
- Cultural differences in emotion: differences in emotional arousal level between the East and the WestIntegrative Medicine Research, 2016
- The circumplex model of affect: An integrative approach to affective neuroscience, cognitive development, and psychopathologyDevelopment and Psychopathology, 2005
- Cognition and motivation in emotion.American Psychologist, 1991