Advancing Multimodal Judge Models through a Capability-Oriented Benchmark and MCTS-Driven Data Generation

TL;DR

This paper introduces M-JudgeBench, a comprehensive benchmark for evaluating multimodal large language models as judges, and proposes Judge-MCTS, a data generation framework to improve judge model reliability and capabilities.

Contribution

It presents a new ten-dimensional, capability-oriented benchmark and a MCTS-driven data construction method to enhance the evaluation and training of MLLM-based judge models.

Findings

M-JudgeBench reveals weaknesses in existing judge models.

Judge-MCTS improves judge model performance and reliability.

M-JudgeBench and Judge-MCTS outperform previous benchmarks and methods.

Abstract

Using Multimodal Large Language Models (MLLMs) as judges to achieve precise and consistent evaluations has gradually become an emerging paradigm across various domains. Evaluating the capability and reliability of MLLM-as-a-judge systems is therefore essential for ensuring trustworthy assessment. Existing judge benchmarks categorize samples by task types but fail to capture the fundamental judgment capabilities required for reliable evaluation. In this work, we introduce M-JudgeBench, a ten-dimensional capability-oriented benchmark designed to comprehensively assess the judgment abilities of MLLMs. Our benchmark decomposes evaluation into pairwise Chain-of-Thought (CoT) comparison, length bias avoidance, and process error detection tasks, jointly covering ten fine-grained subtasks. This design enables diagnosis of model reliability across reasoning styles, response lengths, and cross-model variations. Systematic evaluation uncovers the systematic weaknesses in existing MLLM-as-a-judge systems. To address this issue, we further propose Judge-MCTS, a data construction framework generating pairwise reasoning trajectories with various correctness and length. Using Judge-MCTS, we construct an MCTS-augmented dataset and train M-Judger, a series of strong judge models. Extensive experiments demonstrate the superiority of M-Judger on existing judge benchmarks as well as M-JudgeBench. Overall, our work establishes a more principled foundation for evaluating MLLM-as-a-judge through M-JudgeBench and Judge-MCTS framework, paving the way for future research on judge model evaluation and capability-driven judge training.