回归本源：让去噪生成模型真正去噪
Back to Basics: Let Denoising Generative Models Denoise
提出直接预测干净数据的扩散范式，使大图像块（16/32）的纯像素 Transformer 无需分词器与预训练即可生成高分辨率图像，为简化扩散训练目标与像素级生成架构提供可复用思路。
原文依据 simple, large-patch Transformers on pixels can be strong generative models: using no tokenizer, no pre-training, and no extra loss
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论当今的去噪扩散模型并不在经典意义上进行“去噪”，即它们并不直接预测干净图像，而是由神经网络预测噪声或含噪量。本文指出，预测干净数据与预测含噪量在本质上是不同的。根据流形假设，自然数据应位于低维流形上，而含噪量则不然。基于这一假设，我们主张采用直接预测干净数据的模型，这使得看似容量不足的网络也能在极高维空间中有效运作。我们表明，简单的、采用大图像块的像素级 Transformer 可以成为强生成模型：不使用分词器、不做预训练、也不引入额外损失。我们的方法在概念上不过是“Just image Transformers”，我们称之为 JiT。我们报告了 JiT 在 ImageNet 上使用 16 和 32 的大图像块、在 256 和 512 分辨率下具有竞争力的结果，而在这些情形下预测高维含噪量可能会灾难性地失败。借助我们的网络回归到流形这一本源，我们的研究回归本源，并追求一种基于 Transformer 的、直接在原始自然数据上进行扩散的自包含范式。
英文原文摘要 1239 字符
Today's denoising diffusion models do not "denoise" in the classical sense, i.e., they do not directly predict clean images. Rather, the neural networks predict noise or a noised quantity. In this paper, we suggest that predicting clean data and predicting noised quantities are fundamentally different. According to the manifold assumption, natural data should lie on a low-dimensional manifold, whereas noised quantities do not. With this assumption, we advocate for models that directly predict clean data, which allows apparently under-capacity networks to operate effectively in very high-dimensional spaces. We show that simple, large-patch Transformers on pixels can be strong generative models: using no tokenizer, no pre-training, and no extra loss. Our approach is conceptually nothing more than "Just image Transformers", or JiT, as we call it. We report competitive results using JiT with large patch sizes of 16 and 32 on ImageNet at resolutions of 256 and 512, where predicting high-dimensional noised quantities can fail catastrophically. With our networks mapping back to the basics of the manifold, our research goes back to basics and pursues a self-contained paradigm for Transformer-based diffusion on raw natural data.