Compositional Text-to-Image Generation Via Region-aware Bimodal Direct Preference Optimization

Benchmark Model Rank Results
text-to-image-generation-on-t2i-compbenchSDXL-BiDPO#4Color: 0.7935Texture: 0.7136Spatial: 0.2341Shape: 0.6047…
text-to-image-generation-on-dpg-benchSDXL-BiDPO#74DPG-Bench Score: 78.84
text-to-image-generation-on-genevalSDXL-BiDPO#148Overall: 0.62