ResearchBenchmarksText-to-Image Generation on MSCOCO (CLIP-T)Text-to-Image Generation on MSCOCO (CLIP-T)Follow31.8CLIP-TUni-X30.65630.95331.2531.547Sep 29, 2025Evaluation resultsMethodMethodLinksCLIP-TUni-X# Params.=3B / 4.5B, T...# Params.=3B / 4.5B, Trained on more image-text data=true2025.0931.8EMU3# Params.=8B# Params.=8B2025.0931.3Liquid# Params.=7B# Params.=7B2025.0930.7