visual encoding

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…