zhibinlan/UME-R1-2B
Image-Text-to-Text
•
2B
•
Updated
•
277
•
5
UME-R1 is a framework designed to endow multimodal embedding models with the flexibility to switch between discriminative and generative embeddings