SpeeDiff is a latent diffusion model that jointly trains a VAE and diffusion network from scratch, using a Tweedie Pixel Reconstruction loss to prevent latent collapse. The approach allows scaling a transformer‑based architecture and improves representation alignment. Experiments show the SpeeDiff‑XL model trains 140× faster than Vanilla SiT and 61× faster than REPA, achieving an FID of 1.50 on ImageNet 256×256 and 1.53 on 512×512 without guidance.
By Bingliang Zhang, Wenda Chu, Yizhuo Li, Linjie Yang, Yisong Yue, Katherine L. Bouman, Yang Song, Qiushan Guo
The paper revisits pixel-space diffusion models, showing that a vanilla diffusion transformer can generate directly in pixel space by swapping velocity prediction for clean image prediction. It introduces MFlow, an adaptive parameterization that interpolates between these two prediction modes based on data representation and model capacity, and demonstrates competitive performance on ImageNet, UCF‑101, and text‑to‑image tasks. Additionally, the authors add ConvRefiner, a lightweight convolutional module trained post‑hoc to restore high‑frequency details while preserving FID.
By Soumik Mukhopadhyay, Long Mai, Abhinav Shrivastava, Aniruddha Mahapatra
arXiv:2609.39222v1 Announce Type: new
Abstract: High-compression tokenizers are essential for scaling latent image generative models. However, aggressive compression creates a fundamental tradeoff be...
By Xu Huang, Ye Huang, Zijun Liao, Yuwei Niu, Xiaojie Li, Menghan Zhou, De Wen Soh, Xiaotong Li, Daquan Zhou
arXiv:2610.09450v1 Announce Type: new
Abstract: Pixel-space diffusion models avoid the lossy VAE of latent models, which suggests an advantage on downstream tasks where fine-grained detail matters. W...
By Hanqiu Li Cai (SperidLabs), Chema Garabito (SperidLabs)
arXiv:2609.24919v1 Announce Type: new
Abstract: Recent advances in pixel-space diffusion models have narrowed the image quality gap with latent-space diffusion, but still converge more slowly and lag...
By Yongsheng Yu, Wei Xiong, Yichen Sheng, Shiqiu Liu, Jiebo Luo
Unified image restoration (UIR) aims to recover high-quality (HQ) content from low-quality (LQ) images with different degradations using a single model. Most recent methods adapt large pretrained text-to-image (T2I) latent diffusion models for their strong capacity and generative priors.
arXiv:2608. 01298v1 Announce Type: cross Abstract: Diffusion Transformers (DiTs) have emerged as a core architecture in generative modeling due to their scalability and adaptability to multimodal tasks.
By Junno Yun, Ya\c{s}ar Utku Al\c{c}alar, Mehmet Ak\c{c}akaya
PixelMoE introduces a pixel‑space generative framework that applies LLM‑style Mixture‑of‑Experts (MoE) to diffusion models, enabling sparse scaling of model capacity without diffusion‑specific routing. It combines Token‑Choice routing, fine‑grained routed experts, a shared expert, and bias‑based load balancing, and adds Pixel‑Space Hierarchical Flow Matching to vary spatial resolution during denoising. Across model scales, PixelMoE outperforms dense counterparts and achieves competitive FID scores on class‑conditional ImageNet‑1K, with code and pretrained models publicly released.
By Ruixiao Dong, Jianmin Bao, Bozhi Luan, Zijie Meng, Xin Tao, Pengfei Wan, Kun Gai, Li Li, Houqiang Li
arXiv:2609.37080v1 Announce Type: new
Abstract: Latent Diffusion Models (LDMs) typically adopt a two-stage pipeline: an auto-encoder (AE) is first pre-trained to define a latent space, then a diffusi...
By Zhengqiang Zhang, Lingchen Sun, Rongyuan Wu, Qiaosi Yi, Xiangtao Kong, Chaodong Xiao, Lei Zhang
The paper investigates how visual representations contribute to pixel diffusion transformers and finds that higher representation quality does not always improve generative performance, and that separating architecture into specialized decoders mainly redistributes modeling rather than enhancing generative suitability. Based on these findings, the authors propose Representation Guidance (RG), a method that compacts generation-relevant information into a low-dimensional bottleneck and explicitly uses it to steer decoder generation. RG achieves state‑of‑the‑art results on ImageNet 256×256 with fewer parameters, scales efficiently, and extends to text‑to‑image tasks with strong benchmark performance.
By Zihao Pan, Mengting Chen, Cheng Shi, Taihang Hu, Zhengrong Yue, Zikang Wang, Yuan Wang, Wei Li, Sibei Yang
UltraDiff is a new framework for high‑resolution image generation with Latent Diffusion Models that tackles the trade‑off between computational cost and fidelity caused by latent compression. It jointly learns latent diffusion and then distills a marginal score using a dual‑branch DiT called MixTeacher, which models both f8 and f32 distributions and resolves compressed‑space ambiguity with high‑fidelity priors. Vector‑field distillation in the native f32 space forces the student model to learn the teacher’s conditional expectation, filtering out compression artifacts and achieving f8‑level quality while retaining f32 efficiency—generating a 2K image in 6.2 seconds on a single A100 GPU, over 40× faster than the f8 baseline.
By Jingjing Ren, Haitian Zheng, Haoyu Chen, Connelly Barnes, Zhifei Zhang, Yuqian Zhou, Zhe Lin, Lei Zhu
arXiv:2608. 09133v1 Announce Type: cross Abstract: Image super-resolution (SR) with large generative models has recently achieved remarkable perceptual quality, yet maintaining fidelity to the LR observation remains challenging.
By Yu Shi, Yuyao Zhang, Yu-wing Tai