Lingua-SafetyBenchA Benchmark for Safety Evaluation of Multilingual Vision-Language Models.
DatasetWe provide our multilingual and multimodal dataset for safety evaluation here.Lingua-SafetyBench is a large-scale multilingual multimodal safety benchmark designed to evaluate the safety robustness of Vision-Language Large Models (VLLMs) under joint multilingual and multimodal inputs. Unlike prior benchmarks that mainly focus on either multilingual text-only safety or monolingual multimodal… See the full description on the dataset page:
https://huggingface.co/datasets/leaf-pear/Lingua-SafetyBench.
