Lingua-SafetyBench: A Benchmark for Safety Evaluation of Multilingual Vision-Language Models

TL;DR

Lingua-SafetyBench is a new comprehensive benchmark with over 100,000 harmful multilingual image-text pairs designed to evaluate and improve the safety of Vision-Language Large Models across diverse languages and modalities.

Contribution

It introduces a large-scale, multilingual, multimodal safety benchmark with data partitioning to analyze vulnerabilities and disparities in current VLLMs.

Findings

Current VLLMs show significant vulnerabilities under joint multilingual and multimodal inputs.

Risks are higher in Non-High-Resource Languages and non-Latin scripts.

Model scaling improves safety more for High-Resource Languages, increasing disparities.

Abstract

The robust safety of Vision-Language Large Models (VLLMs) against joint multilingual and multimodal threats remains severely underexplored. Current benchmarks typically isolate these dimensions, being either multilingual but text-only, or multimodal but monolingual. While recent red-teaming efforts attempt to bridge this gap by rendering harmful prompts as images, their overreliance on typography-style visuals and lack of semantically grounded image-text pairs fail to capture realistic cross-modal interactions under multilingual and multimodal conditions. To address this, we introduce Lingua-SafetyBench, a comprehensive benchmark of 100,440 harmful image-text pairs spanning 10 languages. Crucially, Lingua-SafetyBench explicitly partitions data into image-dominant and text-dominant subsets to precisely disentangle sources of risk. Extensive evaluations reveal that current VLLMs retain non-negligible vulnerabilities under these joint inputs. Linguistically, requests in Non-High-Resource Languages (Non-HRLs) and non-Latin scripts generally pose greater threats. Furthermore, analyzing modality-language interactions uncovers a striking asymmetry: in High-Resource Languages (HRLs), models are most vulnerable to image-dominant risks, whereas in Non-HRLs, text-dominant risks severely degrade safety performance. Finally, a controlled study on the Qwen series demonstrates that while model scaling and iterative upgrades improve overall safety, they disproportionately benefit HRLs. This exacerbates the safety disparity between HRLs and Non-HRLs under text-dominant risks, highlighting that achieving robust safety requires dedicated language- and modality-aware alignment strategies beyond mere scaling. The code and dataset will be available at https://github.com/zsxr15/Lingua-SafetyBench.Warning: this paper contains examples with unsafe content.