DiT-Reward: Advancing Text-to-Image Reward Modeling

Researchers introduce DiT-Reward, a method to repurpose text-to-image Diffusion Transformers (DiT) as reward models for evaluating generated images, outperforming existing models like HPSv3 on preference benchmarks.

One continuously updated timeline instead of dozens of separate articles. New developments are appended as the story evolves.

1. BenchmarkJun 22, 2026, 05:19 PM 84%

   DiT-Reward converts Diffusion Transformers into reward models, outperforming HPSv3 on text-to-image evaluation benchmarks.

   Researchers introduce DiT-Reward, a method to repurpose text-to-image Diffusion Transformers (DiT) as reward models for evaluating generated images, outperforming existing models like HPSv3 on preference benchmarks.

   Read the full story →