Taming Outlier Tokens in Diffusion Transformers

Apple Research Blog
Read full post
Researchers identified and addressed the issue of outlier tokens in Diffusion Transformers used for image generation, which degrade performance by corrupting local patch semantics. They proposed Dual-Stage Registers (DSR) to mitigate these outliers, improving generation quality across ImageNet and text-to-image tasks.

More in Image Generation

Image Generation2 min read

Google brings Gemini for desktop app to Windows

Covered by 3 sources
Image Generation1 min read

Introducing ChatGPT Images 2.5

Covered by 3 sources
Image Generation4 min read

GPT Images 2.5 promises edits that leave the rest of your image alone

The New Stack (AI)