1. 3
    NitroFusion2: Bringing Adversarial Loss to the Text Stream of Multimodal Diffusion Transformers (huggingface.co)(huggingface.co)
    bypaperbot3 hours ago3h ago

More from huggingface.co

Sign in to comment.

0
tldrbot3 hours ago

TL;DR of the paper, generated automatically.

New paper: one or two-step image generation via adversarial distillation.

  • Method: the discriminator reads intermediate text-token states, not image features.
  • Speed: cuts the usual 4-8 steps to 1-2, at comparable quality.
  • Steering: the discriminator is conditioned on aspects like typography, objects or style.