Multi-Task Reinforcement Learning for Enhanced Multimodal LLM-as-a-Judge

TL;DR

This paper introduces MT-RL-Judge, a multi-task reinforcement learning framework that improves multimodal large language models' ability to evaluate across diverse visual tasks, aligning better with human judgment.

Contribution

The paper presents a novel multi-task reinforcement learning approach that enhances the generalization and consistency of MLLM-based judges across multiple visual evaluation tasks.

Findings

MT-RL-Judge outperforms baselines in judgment consistency.

It shows higher correlation with human preferences.

The approach generalizes well to out-of-distribution tasks.

Abstract

Multimodal Large Language Models (MLLMs) have been widely adopted as MLLM-as-a-Judges due to their strong alignment with human judgment across various visual tasks. However, most existing judge models are optimized for single-task scenarios and struggle to generalize to diverse contexts, which is a critical requirement for reliable evaluation. To address this limitation, we propose Multi-Task Reinforcement Learning for MLLM-as-a-Judge (MT-RL-Judge), a framework that jointly optimizes the judge model across multiple tasks, leveraging the generalization capabilities of RL. Experimental results against several strong baselines demonstrate that MT-RL-Judge outperforms strong baselines in both judgment consistency and correlation with human preferences. Furthermore, our approach exhibits robust generalization on out-of-distribution tasks, further validating its effectiveness.