Saved in:
Bibliographic Details
Main Authors: Zhou, Runjie, Shao, Youbo, Lu, Haoyu, Xing, Bowei, Bai, Tongtong, Chen, Yujie, Zhao, Jie, Sui, Lin, Yao, Haotian, Zhao, Zijia, Yang, Hao, Wu, Haoning, Zhou, Zaida, Zhu, Jinguo, Huang, Zhiqi, Bao, Yiping, Liu, Yangyang, Charles, Y., Zhou, Xinyu
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.02537
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911417600311296
author Zhou, Runjie
Shao, Youbo
Lu, Haoyu
Xing, Bowei
Bai, Tongtong
Chen, Yujie
Zhao, Jie
Sui, Lin
Yao, Haotian
Zhao, Zijia
Yang, Hao
Wu, Haoning
Zhou, Zaida
Zhu, Jinguo
Huang, Zhiqi
Bao, Yiping
Liu, Yangyang
Charles, Y.
Zhou, Xinyu
author_facet Zhou, Runjie
Shao, Youbo
Lu, Haoyu
Xing, Bowei
Bai, Tongtong
Chen, Yujie
Zhao, Jie
Sui, Lin
Yao, Haotian
Zhao, Zijia
Yang, Hao
Wu, Haoning
Zhou, Zaida
Zhu, Jinguo
Huang, Zhiqi
Bao, Yiping
Liu, Yangyang
Charles, Y.
Zhou, Xinyu
contents We introduce WorldVQA, a benchmark designed to evaluate the atomic visual world knowledge of Multimodal Large Language Models (MLLMs). Unlike current evaluations, which often conflate visual knowledge retrieval with reasoning, WorldVQA decouples these capabilities to strictly measure "what the model memorizes." The benchmark assesses the atomic capability of grounding and naming visual entities across a stratified taxonomy, spanning from common head-class objects to long-tail rarities. We expect WorldVQA to serve as a rigorous test for visual factuality, thereby establishing a standard for assessing the encyclopedic breadth and hallucination rates of current and next-generation frontier models.
format Preprint
id arxiv_https___arxiv_org_abs_2602_02537
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle WorldVQA: Measuring Atomic World Knowledge in Multimodal Large Language Models
Zhou, Runjie
Shao, Youbo
Lu, Haoyu
Xing, Bowei
Bai, Tongtong
Chen, Yujie
Zhao, Jie
Sui, Lin
Yao, Haotian
Zhao, Zijia
Yang, Hao
Wu, Haoning
Zhou, Zaida
Zhu, Jinguo
Huang, Zhiqi
Bao, Yiping
Liu, Yangyang
Charles, Y.
Zhou, Xinyu
Computer Vision and Pattern Recognition
Machine Learning
We introduce WorldVQA, a benchmark designed to evaluate the atomic visual world knowledge of Multimodal Large Language Models (MLLMs). Unlike current evaluations, which often conflate visual knowledge retrieval with reasoning, WorldVQA decouples these capabilities to strictly measure "what the model memorizes." The benchmark assesses the atomic capability of grounding and naming visual entities across a stratified taxonomy, spanning from common head-class objects to long-tail rarities. We expect WorldVQA to serve as a rigorous test for visual factuality, thereby establishing a standard for assessing the encyclopedic breadth and hallucination rates of current and next-generation frontier models.
title WorldVQA: Measuring Atomic World Knowledge in Multimodal Large Language Models
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2602.02537