# A Frozen Pixel-Space Diffusion Model Can Guide Itself with Its Own Samples

arXiv: https://arxiv.org/abs/2607.29122 · HTML: https://mlsanity.com/paper/2607.29122/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models.
- Builds on 23 of its 34 references that are in the map.
- First posted on arXiv 31 Jul 2026.

## Placed in

- Diffusion models / [Diffusion foundations and sampling](https://mlsanity.com/diffusion/diffusion-foundations-and-sampling/)

## Builds on (most cited first, 23 of 23)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://mlsanity.com/paper/1505.04597/) (2015)
- [ImageNet Large Scale Visual Recognition Challenge](https://mlsanity.com/paper/1409.0575/) (2014)
- [Diffusion Models Beat GANs on Image Synthesis](https://mlsanity.com/paper/2105.05233/) (2021)
- [Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow](https://mlsanity.com/paper/2209.03003/) (2022)
- [Improved Techniques for Training GANs](https://mlsanity.com/paper/1606.03498/) (2016)
- [Perceptual Losses for Real-Time Style Transfer and Super-Resolution](https://mlsanity.com/paper/1603.08155/) (2016)
- [Improved Precision and Recall Metric for Assessing Generative Models](https://mlsanity.com/paper/1904.06991/) (2019)
- [SiT: Exploring Flow and Diffusion-based Generative Models with Scalable Interpolant Transformers](https://mlsanity.com/paper/2401.08740/) (2024)
- [Understanding Diffusion Objectives as the ELBO with Simple Data Augmentation](https://mlsanity.com/paper/2303.00848/) (2023)
- [Guiding a Diffusion Model with a Bad Version of Itself](https://mlsanity.com/paper/2406.02507/) (2024)
- [Scalable Adaptive Computation for Iterative Generation](https://mlsanity.com/paper/2212.11972/) (2022)
- [PixelFlow: Pixel-Space Generative Models with Flow](https://mlsanity.com/paper/2504.07963/) (2025)
- [PixelDiT: Pixel Diffusion Transformers for Image Generation](https://mlsanity.com/paper/2511.20645/) (2025)
- [DeCo: Frequency-Decoupled Pixel Diffusion for End-to-End Image Generation](https://mlsanity.com/paper/2511.19365/) (2025)
- [DiP: Taming Diffusion Models in Pixel Space](https://mlsanity.com/paper/2511.18822/) (2025)
- [Latent Forcing: Reordering the Diffusion Trajectory for Pixel-Space Image Generation](https://mlsanity.com/paper/2602.11401/) (2026)
- [There is No VAE: End-to-End Pixel-Space Generative Modeling via Self-Supervised Pre-training](https://mlsanity.com/paper/2510.12586/) (2025)
- [Guiding a Diffusion Transformer with the Internal Dynamics of Itself](https://mlsanity.com/paper/2512.24176/) (2025)
- [PixelU: A U-Shaped Transformer for Efficient End-to-End Pixel Diffusion](https://mlsanity.com/paper/2606.27760/) (2026)
- [Neon: Negative Extrapolation From Self-Training Improves Image Generation](https://mlsanity.com/paper/2510.03597/) (2025)

## Built on by (most cited first, 2 of 2)

- [Persistence Forcing: Exploiting Feature Specialization in Pixel-Space Diffusion](https://mlsanity.com/paper/2609.36014/) (2026)
- [How to Guide Your Language Flow](https://mlsanity.com/paper/2609.19356/) (2026)

## Abstract (arXiv)

Pixel-space diffusion models aim to learn an end-to-end generator directly over raw pixels. This is challenging because a single model must capture both global structure and local texture in the same high-dimensional space. While recent work improves pixel diffusion through alternative prediction targets, training objectives, and architectures, these advances typically require training a new model from scratch. We show there is a cheaper, complementary strategy: a frozen, pretrained pixel diffusion model can guide itself. Our key observation is that intermediate layers of a pretrained pixel diffusion transformer can be decoded into coarse predictions that capture the main low-frequency structure, while the final layers progressively refine local, high-frequency details. We therefore attach a lightweight prediction head to an intermediate layer, keep the backbone frozen, and use the discrepancy between the intermediate and final predictions as a self-guidance direction during sampling. To train this head, we further find that real images are not necessary. Instead, model-generated samples suffice and even outperform real images for training the head, especially in enhancing the high-frequency components that pixel diffusion tends to underfit. Across multiple pixel diffusion models on ImageNet, our Synthetic Self-Guidance (SSG) consistently improves generation while adapter training requires less than 1% of full-model training compute: it reduces FID by over 50% across the evaluated JiT variants without classifier-free guidance (CFG) and further improves strong baselines with CFG, e.g., JiT-H/16 from 1.86 to 1.67 and PixelREPA-H/16 from 1.81 to 1.59. Our code is available at https://github.com/zfu006/SSG.