Attention Alignment Between Humans and Vision-Language Models
人类与视觉语言模型之间的注意力对齐
Isaac R. Christian, Udith Haputhanthrige, Hanna Hornfeld, Declan Campbell, Samuel Nastase, Taylor Webb, Michael Graziano
机构
*
Princeton Neuroscience Institute, Princeton University(普林斯顿大学普林斯顿神经科学研究所)
;
Department of Psychology, Princeton University(普林斯顿大学心理学系)
;
Department of Computer Science, Princeton University(普林斯顿大学计算机科学系)
;
Department of Psychology and Center for Computational Language Sciences, University of Southern California(南加州大学心理学系与计算语言科学中心)
;
Department of Psychology, Université de Montréal(蒙特利尔大学心理学系)
专题命中
其他LLM
:language model(title,abstract)
AI总结
本研究比较了六种视觉语言模型的空间注意力图与人类注视热图,发现解码器架构(LSTM vs Transformer)主导对齐程度,LSTM解码器对齐度更高但空间分散且任务区分度低,而Transformer解码器注意力更集中且任务区分度强。
机构
*
Mondo Robotics
;
The Hong Kong University of Science and Technology (Guangzhou)(香港科技大学(广州))
;
The Hong Kong University of Science and Technology(香港科技大学)
;
Artificial General Intelligence Institute, University of Science and Technology of China(中国科学技术大学通用人工智能研究院)
Learning social norms enhances compatibility in dynamic human-AI coordination
学习社会规范可增强动态人机协作中的兼容性
Yi Yang, Siyuan Liu, Xin Gao, Huamu Sun, Chao Liu, Qing Zhou, Bingbing Nie
机构
*
School of Vehicle and Mobility, Tsinghua University, Beijing, China(清华大学车辆与移动系统学院)
;
State Key Laboratory of Intelligent Green Vehicle and Mobility, Tsinghua University, Beijing 100084, China(清华大学智能绿色车辆与移动系统国家重点实验室)
;
State Key Laboratory of Cognitive Neuroscience and Learning & IDG/McGovern Institute for Brain Research, Beijing Normal University, Beijing, China(北京师范大学认知神经科学与学习国家重点实验室)
;
Beijing Key Laboratory of Safe AI and Superalignment, Beijing, China(北京安全人工智能与超对齐关键实验室)
;
Beijing Institute of AI Safety and Governance, Beijing, China(北京人工智能安全与治理研究院)
专题命中
其他LLM
:LLM(abstract);large language model(abstract);language model(abstract);分类 cs.AI
CommentsPublished in /Handbook of Democracy in the Era of Artificial Intelligence/ edited by Evangelos Pournaras, Srijoni Majumdar, Carina Ines Hausladen, and Dirk Helbing. 2026