Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wang, Wenqi, Tan, Reuben, Zhu, Pengyue, Yang, Jianwei, Yang, Zhengyuan, Wang, Lijuan, Kolobov, Andrey, Gao, Jianfeng, Gong, Boqing
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2505.05456
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866917080449679360
author Wang, Wenqi
Tan, Reuben
Zhu, Pengyue
Yang, Jianwei
Yang, Zhengyuan
Wang, Lijuan
Kolobov, Andrey
Gao, Jianfeng
Gong, Boqing
author_facet Wang, Wenqi
Tan, Reuben
Zhu, Pengyue
Yang, Jianwei
Yang, Zhengyuan
Wang, Lijuan
Kolobov, Andrey
Gao, Jianfeng
Gong, Boqing
contents Spatial intelligence (SI) represents a cognitive ability encompassing the visualization, manipulation, and reasoning about spatial relationships, underpinning disciplines from neuroscience to robotics. We introduce SITE, a benchmark dataset towards SI Thorough Evaluation in a standardized format of multi-choice visual question-answering, designed to assess large vision-language models' spatial intelligence across diverse visual modalities (single-image, multi-image, and video) and SI factors (figural to environmental scales, spatial visualization and orientation, intrinsic and extrinsic, static and dynamic). Our approach to curating the benchmark combines a bottom-up survey about 31 existing datasets and a top-down strategy drawing upon three classification systems in cognitive science, which prompt us to design two novel types of tasks about view-taking and dynamic scenes. Extensive experiments reveal that leading models fall behind human experts especially in spatial orientation, a fundamental SI factor. Moreover, we demonstrate a positive correlation between a model's spatial reasoning proficiency and its performance on an embodied AI task.
format Preprint
id arxiv_https___arxiv_org_abs_2505_05456
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SITE: towards Spatial Intelligence Thorough Evaluation
Wang, Wenqi
Tan, Reuben
Zhu, Pengyue
Yang, Jianwei
Yang, Zhengyuan
Wang, Lijuan
Kolobov, Andrey
Gao, Jianfeng
Gong, Boqing
Computer Vision and Pattern Recognition
Spatial intelligence (SI) represents a cognitive ability encompassing the visualization, manipulation, and reasoning about spatial relationships, underpinning disciplines from neuroscience to robotics. We introduce SITE, a benchmark dataset towards SI Thorough Evaluation in a standardized format of multi-choice visual question-answering, designed to assess large vision-language models' spatial intelligence across diverse visual modalities (single-image, multi-image, and video) and SI factors (figural to environmental scales, spatial visualization and orientation, intrinsic and extrinsic, static and dynamic). Our approach to curating the benchmark combines a bottom-up survey about 31 existing datasets and a top-down strategy drawing upon three classification systems in cognitive science, which prompt us to design two novel types of tasks about view-taking and dynamic scenes. Extensive experiments reveal that leading models fall behind human experts especially in spatial orientation, a fundamental SI factor. Moreover, we demonstrate a positive correlation between a model's spatial reasoning proficiency and its performance on an embodied AI task.
title SITE: towards Spatial Intelligence Thorough Evaluation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2505.05456