Lingua-SafetyBench is a new benchmark of 100,440 multilingual harmful image-text pairs that reveals current vision-language models have notable safety weaknesses, especially in non-high-resource languages and with text-dominant inputs.
the content in the image
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Lingua-SafetyBench: A Benchmark for Safety Evaluation of Multilingual Vision-Language Models
Lingua-SafetyBench is a new benchmark of 100,440 multilingual harmful image-text pairs that reveals current vision-language models have notable safety weaknesses, especially in non-high-resource languages and with text-dominant inputs.