The paper introduces Priming Models, a two‑stage framework for training one‑step generative models. First, the model is trained on a reconstruction objective such as pixel‑space denoising at a fixed noise level, then it undergoes feature‑space distributional post‑training. This simple reconstruction stage alone is shown to eliminate the difficulty of distributional training, leading to significant improvements in training compute versus generation quality across tasks like class‑conditional image generation, text‑to‑image generation, and world modeling, with a state‑of‑the‑art FID of 1.19 on ImageNet 256×256 using about one‑tenth the compute of previous methods.
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.
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. 19332v1 Announce Type: new Abstract: Generative models have undergone many generations of evolution, from VAEs/GANs to diffusion/flow matching.
arXiv:2609.37147v1 Announce Type: cross Abstract: Distributional Diffusion Models (DDMs) replace the standard mean-prediction denoiser with a \emph{distributional} denoiser trained via a scoring rule...
arXiv:2506. 14753v3 Announce Type: replace-cross Abstract: Diffusion models are well known for their ability to generate a high-fidelity image for an input prompt through an iterative denoising process.
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...
arXiv:2610.16759v2 Announce Type: replace Abstract: Latent diffusion models generate images as codes of an image tokenizer, whose decoder turns them into the pixel image. Recent tokenizers improve ho...
arXiv:2606. 27978v1 Announce Type: cross Abstract: Pixel-space continuous-token autoregressive (AR) generation directly models images as sequences of raw pixel patches, avoiding discrete tokenization or a separately pretrained tokenizer.
arXiv:2609.35763v3 Announce Type: replace Abstract: Distributional training provides collective supervision for one-step visual generation by matching real and generated features in frozen representa...
The paper introduces continuous adversarial flow models, a continuous-time flow framework trained with an adversarial objective that replaces the fixed mean-squared-error criterion of flow matching. By incorporating a learned discriminator, the method guides training toward a different generalized distribution, yielding samples more closely aligned with the target data distribution. Applied as a post‑training step, it markedly improves ImageNet 256px generation metrics—reducing the guidance‑free FID of latent‑space SiT from 8.26 to 3.63 and of pixel‑space JiT from 7.17 to 3.57—and also enhances guided generation and text‑to‑image benchmarks.
The paper introduces Abra, a family of flow‑matching transformers used to systematically study scaling laws for text‑to‑image diffusion models across three orders of magnitude in compute. It finds that diffusion models scale predictably like language models but need far more data, with compute optimality occurring at roughly 200 image tokens per parameter—ten times the optimal ratio for large language models. The study also shows that diffusion models are robust to overtraining, that more data is preferable to larger models, and that scaling predictability extends to generative quality, optimal CFG settings, representation quality, and training curve shapes.
The paper introduces data‑space iteration, a few‑step generation framework that eliminates the need for a fixed probability‑flow discretization. By refining predictions directly in data space with a shared generator, each iteration is trained to produce the best possible sample within its capacity. Experiments on class‑conditional ImageNet 256×256 show that data‑space iteration outperforms standard discretization baselines and matches or improves upon schedule‑searched variants without schedule‑specific training.