Releases a million-scale human preference dataset (29M pairs) specifically for text-to-image editing tasks.
arXiv · March 17, 2026 · 2603.14916
The Takeaway
Large-scale preference data for image *editing* (vs. generation) has been a major bottleneck. This release allows for the training of robust reward models to align editing tools with human notions of quality and instruction adherence.
From the abstract
Recent text-guided image editing (TIE) models have achieved remarkable progress, while many edited images still suffer from issues such as artifacts, unexpected editings, unaesthetic contents. Although some benchmarks and methods have been proposed for evaluating edited images, scalable evaluation models are still lacking, which limits the development of human feedback reward models for image editing. To address the challenges, we first introduce \textbf{EditHF-1M}, a million-scale image editing