PulseAugur
EN
LIVE 04:21:36

CrossFlow model generates images directly from latent space

Researchers have introduced CrossFlow, a novel cross-space flow formulation that maps noisy latent inputs directly to pixel-space images. This approach bypasses the need for a separate decoder by optimizing a one-step objective that predicts an image rather than a latent displacement. CrossFlow can function as a standalone latent-to-pixel generator or as a decoder replacement for existing latent diffusion pipelines. In experiments on class-conditional ImageNet-1k at 256x256 resolution, CrossFlow-XL achieved a FID score of 1.62 with a single function evaluation, demonstrating the effectiveness of combining latent efficiency with direct pixel supervision. AI

IMPACT This novel cross-space flow formulation could improve the efficiency and quality of image generation models.

RANK_REASON The cluster contains a research paper detailing a new method for image generation.

Read on arXiv cs.CV →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

CrossFlow model generates images directly from latent space

COVERAGE [2]

  1. arXiv cs.CV TIER_1 English(EN) · Xiyuan Wang, Xiao Zhang, Yang Li, Ruoxi Jiang, Zhao Zhong, Liefeng Bo, Muhan Zhang ·

    CrossFlow: One-Step Generation Across Latent and Pixel Spaces

    arXiv:2606.19970v1 Announce Type: new Abstract: Most diffusion and flow-matching generators define the prior, probability path, and prediction target in the same representation space. Latent diffusion improves efficiency by moving this path into an autoencoder latent space, but t…

  2. arXiv cs.CV TIER_1 English(EN) · Muhan Zhang ·

    CrossFlow: One-Step Generation Across Latent and Pixel Spaces

    Most diffusion and flow-matching generators define the prior, probability path, and prediction target in the same representation space. Latent diffusion improves efficiency by moving this path into an autoencoder latent space, but the final sample is still produced by a separatel…