Search papers, labs, and topics across Lattice.
Canva Research
2
0
4
Method improvements in diffusion transformers may not be reflected in both ImageNet and text-to-image tasks, revealing a critical gap in current evaluation practices.
Forget slow, memory-intensive layered image generation – MRT achieves 10-100x faster inference and 50-90% less GPU memory consumption while outperforming even Qwen-Image-Layered.
