Beyond Unimodal Shortcuts: MLLMs as Cross-Modal Reasoners for Grounded Named Entity Recognition

TL;DR

This paper investigates how Multimodal Large Language Models can be used for grounded named entity recognition, identifies modality bias issues, and proposes a novel reasoning framework to improve cross-modal verification and performance.

Contribution

It introduces Modality-aware Consistency Reasoning (MCR), a new framework that reduces modality bias in MLLMs for GMNER through structured reasoning and optimization techniques.

Findings

MCR significantly reduces modality bias in MLLMs.

The proposed method outperforms existing baselines on GMNER tasks.

Experiments demonstrate improved cross-modal reasoning and grounding accuracy.

Abstract

Grounded Multimodal Named Entity Recognition (GMNER) aims to extract text-based entities, assign them semantic categories, and ground them to corresponding visual regions. In this work, we explore the potential of Multimodal Large Language Models (MLLMs) to perform GMNER in an end-to-end manner, moving beyond their typical role as auxiliary tools within cascaded pipelines. Crucially, our investigation reveals a fundamental challenge: MLLMs exhibit $\textbf{modality bias}$, including visual bias and textual bias, which stems from their tendency to take unimodal shortcuts rather than rigorous cross-modal verification. To address this, we propose Modality-aware Consistency Reasoning ($\textbf{MCR}$), which enforces structured cross-modal reasoning through Multi-style Reasoning Schema Injection (MRSI) and Constraint-guided Verifiable Optimization (CVO). MRSI transforms abstract constraints into executable reasoning chains, while CVO empowers the model to dynamically align its reasoning trajectories with Group Relative Policy Optimization (GRPO). Experiments on GMNER and visual grounding tasks demonstrate that MCR effectively mitigates modality bias and achieves superior performance compared to existing baselines.