NextFlow: Unified Sequential Modeling Activates Multimodal Understanding and Generation

Benchmark Model Rank Results
text-to-image-generation-on-prism-benchNextFlow-RL (7B)#1Accuracy: 78.8
image-editing-on-gedit-bench-enNextFlow RL#6Overall: 7.87
image-generation-on-wiseNextFlow-RL (7B)#23Overall: 0.62
image-editing-on-imgedit-dataNextFlow–Overall: 4.44Add: 4.34Adjust: 4.68Extract: 4.23Replace: 4.47…
image-editing-on-imgedit-dataNextFlow-RL–Overall: 4.49Add: 4.33Adjust: 4.68Extract: 4.22Replace: 4.50…
text-to-image-generation-on-wiseNextFlow–Overall: 0.59Cultural: 0.62Time: 0.60Space: 0.70Biology: 0.54…
text-to-image-generation-on-wiseNextFlow-RL–Overall: 0.62Cultural: 0.63Time: 0.63Space: 0.77Biology: 0.58…