Visual Tokenization

A text representation approach that renders text as images and processes them with vision models, bypassing conventional character segmentation and vocabulary design. A representative implementation is the PIXEL model. It is one direction explored as an alternative to traditional tokenizers.