The paper investigates why diffusion autoencoders can achieve similar image reconstruction quality while exhibiting varying degrees of disentanglement. By tracking training trajectories—plots of reconstruction quality versus disentanglement—it identifies two learning regimes: a reconstruction regime where fidelity improves before disentanglement, and a disentanglement regime where both metrics improve gradually. The authors propose SteeringDRL, which uses gated residual U‑Nets and a noise curriculum to steer training toward better disentanglement, achieving competitive results on attribute and spatial disentanglement benchmarks and reducing seed sensitivity.
arXiv:2607. 05319v1 Announce Type: cross Abstract: We study why diffusion autoencoders can achieve similar image quality while learning substantially different latent structures.
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.
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.
The paper introduces DiffPT, a diffusion‑based pretraining method for classifier backbones aimed at improving performance when training data is scarce. Experiments on time‑series, natural image, and medical imaging datasets show that DiffPT outperforms many self‑supervised baselines, especially in low‑data regimes for medical and time‑series tasks. While contrastive pretraining remains superior on CIFAR‑10 and Tiny ImageNet, DiffPT offers a simple and effective alternative across diverse modalities.
The paper introduces D3S2, a diffusion‑guided dataset distillation framework tailored for semantic segmentation. It tackles long‑tailed class imbalance, pixel‑wise alignment, and high computational cost by first selecting class‑balanced masks and then synthesizing images with a pretrained diffusion model conditioned on those masks. Guided diffusion sampling further refines the data with segmentation‑consistency and class‑wise feature matching losses, achieving superior performance at a 1% compression rate on ADE20K and COCO‑Stuff.
Latent Diffusion-GAN (LD‑GAN) integrates adversarial learning into diffusion models by using the pretrained VAE as an energy‑based discriminator, avoiding extra parameters and preserving the latent priors. It introduces a structural consistency energy to align encoder and decoder features, improving perceptual quality. Experiments show LD‑GAN enhances sample fidelity, sharpness, and diversity across generation tasks while maintaining efficient training dynamics.
arXiv:2411.05005v2 Announce Type: replace-cross Abstract: Beyond high-fidelity image synthesis, diffusion models have recently exhibited promising results in dense visual perception tasks. However, m...
arXiv:2601.15829v3 Announce Type: replace Abstract: Recent years have witnessed the remarkable success of deep learning in remote sensing image interpretation, driven by the availability of large-sca...
PRISM is a feed‑forward single‑image 3D reconstruction framework that splits the task into a parameter‑free geometric warp and a learned residual correction, eliminating the need for iterative diffusion sampling. It uses a two‑stage training strategy—latent supervised distillation for geometric generalization and perceptual fine‑tuning for appearance quality—to generalize from synthetic data. Experiments on three benchmarks show PRISM matches diffusion‑based methods in quality while cutting inference time to just 36 seconds per scene.
arXiv:2607. 12464v1 Announce Type: cross Abstract: When labeled data are scarce, off-the-shelf diffusion models can augment training sets for few-shot medical image classification, but not all generated samples are equally useful for the downstream task.
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.
arXiv:2610.07002v1 Announce Type: new Abstract: Diffusion models learn semantic representations while generating images. In the Decoupled Diffusion Transformer (DDT), a condition encoder provides fea...