Unlocking Complex Visual Generation via Closed-Loop Verified Reasoning

Benchmark Model Rank Results
image-generation-on-wiseCLVR (4B)–Overall: 0.74Cultural: 0.73Time: 0.72Space: 0.83Biology: 0.73…
image-generation-on-wiseCLVR (9B)–Overall: 0.76Cultural: 0.76Time: 0.73Space: 0.83Biology: 0.76…
text-to-image-generation-on-genevalCLVR (9B)–Overall: 0.88
text-to-image-generation-on-genevalFLUX.2 4B + CLVR + PPRL + DSWM–Overall: 0.87