JUICE: Unified Multimodal Modeling in Pretrained Representation Spaces

Open paper

Benchmark	Model	Rank	Results
text-to-image-generation-on-geneval	JUICE (DINO)	#41	Overall: 0.88
text-to-image-generation-on-geneval	JUICE (SigLIP2)	#53	Overall: 0.87