PoVisLE: Multimodal Benchmark for Polish Cultural Understanding
August 11, 2026
PoVisLE is a monocultural vision-language benchmark containing 1,117 images and 2,366 annotations designed for the Polish language. It evaluates how well models interpret region-specific symbols and visual cues within a specific linguistic and cultural context.
HOW THIS AFFECTS YOU
●
researcherYou can use this to evaluate the cultural and linguistic localization limits of your multimodal models.