Frozen Transformers in Language Models Are Effective Visual Encoder Layers
This research paper reveals that Large Language Models (LLMs), despite being trained solely on textual data, are surprisingly effective encoders for purely visual tasks. This is achieved by using a…
Continue reading