Research

Paper

AI LLM March 12, 2026

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

Authors

Junjie Wu, Xuan Kan, Zihao He, Shunwen Tan, Bo Pan, Kaitai Zhang

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.

Metadata

arXiv ID: 2603.11665
Provider: ARXIV
Primary Category: cs.CL
Published: 2026-03-12
Fetched: 2026-03-14 05:03

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.11665v1</id>\n    <title>Multi-Task Reinforcement Learning for Enhanced Multimodal LLM-as-a-Judge</title>\n    <updated>2026-03-12T08:32:38Z</updated>\n    <link href='https://arxiv.org/abs/2603.11665v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.11665v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>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.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CL'/>\n    <published>2026-03-12T08:32:38Z</published>\n    <arxiv:primary_category term='cs.CL'/>\n    <author>\n      <name>Junjie Wu</name>\n    </author>\n    <author>\n      <name>Xuan Kan</name>\n    </author>\n    <author>\n      <name>Zihao He</name>\n    </author>\n    <author>\n      <name>Shunwen Tan</name>\n    </author>\n    <author>\n      <name>Bo Pan</name>\n    </author>\n    <author>\n      <name>Kaitai Zhang</name>\n    </author>\n  </entry>"
}