1 paper · 1 filter
Tianyu Zhang, Suyuchen Wang, Lu Li +6
We introduce Visual Caption Restoration (VCR), a novel vision-language task that challenges models to accurately restore partially obscured texts using pixel-level hints within ima…