Saved in:
Bibliographic Details
Main Authors: Xiong, Tianyi, Ge, Yi, Li, Ming, Zhang, Zuolong, Kulkarni, Pranav, Wang, Kaishen, He, Qi, Zhu, Zeying, Liu, Chenxi, Chen, Ruibo, Zheng, Tong, Chen, Yanshuo, Wang, Xiyao, Zhang, Renrui, Chen, Wenhu, Huang, Heng
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2511.21662
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912964127227904
author Xiong, Tianyi
Ge, Yi
Li, Ming
Zhang, Zuolong
Kulkarni, Pranav
Wang, Kaishen
He, Qi
Zhu, Zeying
Liu, Chenxi
Chen, Ruibo
Zheng, Tong
Chen, Yanshuo
Wang, Xiyao
Zhang, Renrui
Chen, Wenhu
Huang, Heng
author_facet Xiong, Tianyi
Ge, Yi
Li, Ming
Zhang, Zuolong
Kulkarni, Pranav
Wang, Kaishen
He, Qi
Zhu, Zeying
Liu, Chenxi
Chen, Ruibo
Zheng, Tong
Chen, Yanshuo
Wang, Xiyao
Zhang, Renrui
Chen, Wenhu
Huang, Heng
contents Large multimodal models (LMMs) are increasingly adopted as judges in multimodal evaluation systems due to their strong instruction following and consistency with human preferences. However, their ability to follow diverse, fine-grained evaluation criteria remains underexplored. We develop Multi-Crit, a benchmark for evaluating multimodal judges on their capacity to follow pluralistic criteria and produce reliable criterion-level judgments. Covering both open-ended generation and verifiable reasoning tasks, Multi-Crit is built through a rigorous data curation pipeline that gathers challenging response pairs with multi-criterion human annotations. It further introduces three novel metrics for systematically assessing pluralistic adherence, criterion-switching flexibility, and the ability to recognize criterion-level preference conflicts. Comprehensive analysis of 25 LMMs reveals that 1) proprietary models still struggle to maintain consistent adherence to pluralistic criteria--especially in open-ended evaluation; 2) open-source models lag further behind in flexibly following diverse criteria; and 3) critic fine-tuning with holistic judgment signals enhances visual grounding but fails to generalize to pluralistic criterion-level judgment. Additional analyses on reasoning fine-tuning, test-time scaling, and boundary consistency between open-source and proprietary models further probe the limits of current multimodal judges. As a pioneering study, Multi-Crit lays the foundation for building reliable and steerable multimodal AI evaluation.
format Preprint
id arxiv_https___arxiv_org_abs_2511_21662
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Multi-Crit: Benchmarking Multimodal Judges on Pluralistic Criteria-Following
Xiong, Tianyi
Ge, Yi
Li, Ming
Zhang, Zuolong
Kulkarni, Pranav
Wang, Kaishen
He, Qi
Zhu, Zeying
Liu, Chenxi
Chen, Ruibo
Zheng, Tong
Chen, Yanshuo
Wang, Xiyao
Zhang, Renrui
Chen, Wenhu
Huang, Heng
Computer Vision and Pattern Recognition
Large multimodal models (LMMs) are increasingly adopted as judges in multimodal evaluation systems due to their strong instruction following and consistency with human preferences. However, their ability to follow diverse, fine-grained evaluation criteria remains underexplored. We develop Multi-Crit, a benchmark for evaluating multimodal judges on their capacity to follow pluralistic criteria and produce reliable criterion-level judgments. Covering both open-ended generation and verifiable reasoning tasks, Multi-Crit is built through a rigorous data curation pipeline that gathers challenging response pairs with multi-criterion human annotations. It further introduces three novel metrics for systematically assessing pluralistic adherence, criterion-switching flexibility, and the ability to recognize criterion-level preference conflicts. Comprehensive analysis of 25 LMMs reveals that 1) proprietary models still struggle to maintain consistent adherence to pluralistic criteria--especially in open-ended evaluation; 2) open-source models lag further behind in flexibly following diverse criteria; and 3) critic fine-tuning with holistic judgment signals enhances visual grounding but fails to generalize to pluralistic criterion-level judgment. Additional analyses on reasoning fine-tuning, test-time scaling, and boundary consistency between open-source and proprietary models further probe the limits of current multimodal judges. As a pioneering study, Multi-Crit lays the foundation for building reliable and steerable multimodal AI evaluation.
title Multi-Crit: Benchmarking Multimodal Judges on Pluralistic Criteria-Following
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.21662