Multi-Crit: Benchmarking Multimodal Judges on Pluralistic Criteria-Following

TL;DR

Multi-Crit introduces a comprehensive benchmark for evaluating multimodal models' ability to follow diverse evaluation criteria, revealing current limitations and guiding future improvements in multimodal AI judging systems.

Contribution

It develops a novel benchmark with new metrics to systematically assess pluralistic adherence and criterion-level judgment in multimodal models, addressing a key gap in evaluation methods.

Findings

Proprietary models struggle with consistent criterion adherence.

Open-source models lag in flexible criterion following.

Fine-tuning improves visual grounding but not pluralistic judgment.

Abstract

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.