We introduce OneCAT, a unified multimodal model that seamlessly integrates understanding, generation, and editing within a novel, pure decoder-only transformer architecture. Our framework uniquely ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results