OneCAT: Decoder-Only Auto-Regressive Model for Unified Understanding and Generation

Benchmark Model Rank Results
image-editing-on-imgedit-dataOneCAT-3B–Overall: 3.43Add: 3.65Adjust: 3.70Extract: 2.42Replace: 3.92…
text-to-image-generation-on-genevalOneCAT-1.5B–Overall: 0.85
text-to-image-generation-on-genevalOneCAT-3B–Overall: 0.90
visual-question-answering-on-mm-vetOneCAT-1.5B–GPT-4 score: 42.4
visual-question-answering-on-mm-vetOneCAT-3B–GPT-4 score: 52.2