Fetching the paper…

TokLIP: Marry Visual Tokens to CLIP for Multimodal Comprehension and Generation · Around