Fetching the paper…

Frozen Transformers in Language Models Are Effective Visual Encoder Layers · Around