Think-Then-Generate: Reasoning-Aware Text-to-Image Diffusion with LLM Encoders
Open paper
Benchmark
Model
Rank
Results
image-editing-on-risebench
Qwen-Image-Edit (Think-Then-Generate)
#4
RISEBench Overall Accuracy: 23.9
image-generation-on-wise
Qwen-Image (Think-Then-Generate)
#4
Overall: 0.79
Rank counts only results with a code link.