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.
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...
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: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...
Recent advances in diffusion models have shown impressive performance in controllable image generation and dense prediction tasks. However, existing approaches typically treat diffusion-based controllable generation and dense prediction as separate tasks, overlooking the potential benefits of jointly modeling the heterogeneous distributions.
Pixel diffusion models generate RGB images directly but tend to miss fine‑scale natural‑image statistics. The authors introduce an adversarial post‑training step that adds an adversarial loss to the model’s output at non‑high‑noise timesteps, without changing the architecture or sampling procedure. This approach improves distribution fidelity, coverage, prompt alignment, and perceptual quality across two pixel backbones, and restores missing high‑frequency spectral power while avoiding memorization or mode dropping.
The paper investigates why Diffusion Transformers perform well when predicting clean data but poorly when predicting noise or velocity. It introduces the concept of *residual-stream burden*, explaining that noisy targets force the residual stream to carry noise-dependent variation through the network, limiting representation flexibility. Experiments show that spectral concentration of clean targets reduces this burden, and the authors propose Spatially Indexed Hyper-Connections (SiHC) to expand residual bandwidth, achieving state‑of‑the‑art FID on ImageNet.
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. The approach combines Token-Choice routing, fine-grained routed experts, a shared expert, and bias-based load balancing, while also employing Pixel-Space Hierarchical Flow Matching to adjust spatial resolution during denoising. PixelMoE outperforms dense counterparts at comparable active parameter counts, achieving strong FID scores on class-conditional ImageNet-1K.
arXiv:2607. 19332v1 Announce Type: new Abstract: Generative models have undergone many generations of evolution, from VAEs/GANs to diffusion/flow matching.
arXiv:2610.28086v1 Announce Type: cross Abstract: Real-time video generation, interactive world models, and on-device image synthesis are almost a reality. Diffusion Transformers (DiT) achieve excell...
arXiv:2410.13925v2 Announce Type: replace Abstract: \textit{Nature is infinitely resolution-free}. In the context of this reality, existing diffusion models, such as Diffusion Transformers, often fac...
Generative models have undergone many generations of evolution, from VAEs/GANs to diffusion/flow matching. Along the way, the underlying techniques have become more complicated and various beliefs about what drives strong empirical performance have taken hold.
arXiv:2607. 27372v1 Announce Type: new Abstract: The deep learning revolution, kicked off by AlexNet, taught us that end-to-end training beats decomposing a problem into hand-designed stages.