DRM: Diffusion-based Reward Model With Step-wise Guidance

Open paper

Benchmark	Model	Rank	Results
text-to-image-generation-on-geneval	SD3.5-M + DRM Step-GRPO	#117	Overall: 0.78