Emu3.5: Native Multimodal Models are World Learners

Benchmark Model Rank Results
text-to-image-generation-on-oneig-enEmu3.5 (34B)#2OneIG Overall Score: 0.564
text-to-image-generation-on-tiif-bench-miniEmu3.5 (34B)#2Accuracy: 89.48
image-editing-on-imgedit-dataEmu3.5#5Overall: 4.41Add: 4.61Adjust: 4.32Extract: 3.96Replace: 4.84…
text-to-image-generation-on-cvtg-2kEmu3.5 (34B)#5Word Accuracy: 0.9123
text-to-image-generation-on-longtext-benchEmu3.5 (34B)#6Accuracy: 97.6
text-to-image-generation-on-oneig-zhEmu3.5 (34B)#9OneIG Overall Score: 0.529
image-editing-on-gedit-bench-enEmu3.5 (34B, AR)#13Overall: 7.59
text-to-image-generation-on-genevalEmu3.5 (34B, AR)#52Overall: 0.86