1 paper · 1 filter
Zineng Tang, Long Lian, Seun Eisape +6
Despite the recent success of image-text contrastive models like CLIP and SigLIP, these models often struggle with vision-centric tasks that demand high-fidelity image understandin…