Think in Strokes, Not Pixels: Process-Driven Image Generation via Interleaved Reasoning
Open paper
Benchmark
Model
Rank
Results
image-generation-on-wise
BAGEL (7B) (process-driven SFT)
–
Overall: 0.76
text-to-image-generation-on-geneval
Process-driven BAGEL-7B
–
Overall: 0.83
Rank counts only results with a code link.