LinguaMark: Do Multimodal Models Speak Fairly? A Benchmark-Based Evaluation

TL;DR

LinguaMark introduces a multilingual benchmark for evaluating large multimodal models on fairness, bias, and relevance in visual question answering across 11 languages, revealing performance disparities and generalization capabilities.

Contribution

This work presents LinguaMark, a novel multilingual benchmark with 6,875 image-text pairs for evaluating LMMs on bias, relevance, and faithfulness in VQA tasks across multiple languages.

Findings

Closed-source models outperform open-source models overall.

Qwen2.5 shows strong multilingual generalization.

Models exhibit biases and fairness issues across languages.

Abstract

Large Multimodal Models (LMMs) are typically trained on vast corpora of image-text data but are often limited in linguistic coverage, leading to biased and unfair outputs across languages. While prior work has explored multimodal evaluation, less emphasis has been placed on assessing multilingual capabilities. In this work, we introduce LinguaMark, a benchmark designed to evaluate state-of-the-art LMMs on a multilingual Visual Question Answering (VQA) task. Our dataset comprises 6,875 image-text pairs spanning 11 languages and five social attributes. We evaluate models using three key metrics: Bias, Answer Relevancy, and Faithfulness. Our findings reveal that closed-source models generally achieve the highest overall performance. Both closed-source (GPT-4o and Gemini2.5) and open-source models (Gemma3, Qwen2.5) perform competitively across social attributes, and Qwen2.5 demonstrates strong generalization across multiple languages. We release our benchmark and evaluation code to encourage reproducibility and further research.