arXiv 2602.158922026-07-15cs.CVcs.AI

视觉语言模型中的自我中心偏差

Egocentric Bias in Vision-Language Models

Cognitive Science Program, University of California, Berkeley（加州大学伯克利分校认知科学项目）；Department of Electrical and Computer Engineering, University of California San Diego（加州大学圣地亚哥分校电气与计算机工程系）；School of Computer Science, Georgia Institute of Technology & Emory University（佐治亚理工学院计算机科学学院及埃默里大学）；Department of Computer Science, Johns Hopkins University（约翰霍普金斯大学计算机科学系）；Department of Cognitive Science, University of California San Diego（加州大学圣地亚哥分校认知科学系）；Equal Advising Department of Computer Science & Wilmer Eye Institute, Johns Hopkins University（约翰霍普金斯大学计算机科学系及威尔默眼科研究所）；Language Technologies Institute, Carnegie Mellon University（卡内基梅隆大学语言技术研究所）；Robotics Institute, Carnegie Mellon University（卡内基梅隆大学机器人研究所）；Weinberg Institute for Cognitive Science, University of Michigan（密歇根大学韦恩伯格认知科学研究所）

Maijunxian Wang, Yijiang Li, Bingyang Wang, Tianwei Zhao, Ran Ji, Qingying Gao, Emmy Liu, Hokin Deng, Dezhi Luo

备注：: Accepted at CogSci 2026 (Best Undergraduate Student Paper)

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出FlipSet基准，揭示视觉语言模型在第二级视觉视角推理中存在系统性自我中心偏差，表明模型在整合社会认知与空间操作方面存在根本性不足。

摘要：

视觉视角的推理——推断从他人视角世界如何呈现——是社会认知的基础。我们引入FlipSet，一个用于评估视觉语言模型中第二级视觉视角推理（L2 VPT）的诊断基准。该任务要求模拟从另一个代理视角对二维字符字符串进行180度旋转，隔离空间变换与三维场景复杂性。评估103个VLMs发现系统性的自我中心偏差：绝大多数表现低于随机水平，其中约四分之三的错误重现了摄像机视角。控制实验揭示了组合缺陷——模型在孤立情况下实现高理论-心准确率和高于随机水平的内部旋转，但当需要整合时却彻底失败。这种分离表明当前VLMs缺乏将社会意识与空间操作结合的机制，暗示了基于模型的空间推理的根本限制。FlipSet为多模态系统诊断视角推理能力提供了认知基础的测试平台。

英文摘要：

Visual perspective taking--inferring how the world appears from another's viewpoint--is foundational to social cognition. We introduce FlipSet, a diagnostic benchmark for Level-2 visual perspective taking (L2 VPT) in vision-language models. The task requires simulating 180-degree rotations of 2D character strings from another agent's perspective, isolating spatial transformation from 3D scene complexity. Evaluating 103 VLMs reveals systematic egocentric bias: the vast majority perform below chance, with roughly three-quarters of errors reproducing the camera viewpoint. Control experiments expose a compositional deficit--models achieve high theory-of-mind accuracy and above-chance mental rotation in isolation, yet fail catastrophically when integration is required. This dissociation indicates that current VLMs lack the mechanisms needed to bind social awareness to spatial operations, suggesting fundamental limitations in model-based spatial reasoning. FlipSet provides a cognitively grounded testbed for diagnosing perspective-taking capabilities in multimodal systems.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。