Fetching the paper…

Synth$^2$: Boosting Visual-Language Models with Synthetic Captions and Image Embeddings · Around