# MSF: Efficient Diffusion Model Via Multi-Scale Latent Factorize

arXiv: https://arxiv.org/abs/2501.13349 · HTML: https://mlsanity.com/paper/2501.13349/

## 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 19 of its 58 references that are in the map.
- First posted on arXiv 23 Jan 2025.

## Placed in

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

## Builds on (most cited first, 19 of 19)

- [Denoising Diffusion Implicit Models](https://mlsanity.com/paper/2010.02502/) (2020)
- [Score-Based Generative Modeling through Stochastic Differential Equations](https://mlsanity.com/paper/2011.13456/) (2020)
- [Classifier-Free Diffusion Guidance](https://mlsanity.com/paper/2207.12598/) (2022)
- [SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis](https://mlsanity.com/paper/2307.01952/) (2023)
- [Auto-Encoding Variational Bayes](https://mlsanity.com/paper/1312.6114/) (2013)
- [Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets](https://mlsanity.com/paper/2311.15127/) (2023)
- [Conditional Generative Adversarial Nets](https://mlsanity.com/paper/1411.1784/) (2014)
- [Imagen Video: High Definition Video Generation with Diffusion Models](https://mlsanity.com/paper/2210.02303/) (2022)
- [DPM-Solver++: Fast Solver for Guided Sampling of Diffusion Probabilistic Models](https://mlsanity.com/paper/2211.01095/) (2022)
- [PixArt-α: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis](https://mlsanity.com/paper/2310.00426/) (2023)
- [SiT: Exploring Flow and Diffusion-based Generative Models with Scalable Interpolant Transformers](https://mlsanity.com/paper/2401.08740/) (2024)
- [PixArt-Sigma: Weak-to-Strong Training of Diffusion Transformer for 4K Text-to-Image Generation](https://mlsanity.com/paper/2403.04692/) (2024)
- [Emu: Enhancing Image Generation Models Using Photogenic Needles in a Haystack](https://mlsanity.com/paper/2309.15807/) (2023)
- [PIXART-δ: Fast and Controllable Image Generation with Latent Consistency Models](https://mlsanity.com/paper/2401.05252/) (2024)
- [Playground v3: Improving Text-to-Image Alignment with Deep-Fusion Large Language Models](https://mlsanity.com/paper/2409.10695/) (2024)
- [FiT: Flexible Vision Transformer for Diffusion Model](https://mlsanity.com/paper/2402.12376/) (2024)
- [Pyramid Diffusion for Fine 3D Large Scene Generation](https://mlsanity.com/paper/2311.12085/) (2023)
- [FiTv2: Scalable and Improved Flexible Vision Transformer for Diffusion Model](https://mlsanity.com/paper/2410.13925/) (2024)
- [Diffusion Models Need Visual Priors for Image Generation](https://mlsanity.com/paper/2410.08531/) (2024)

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

- [Adaptive Domain Shift in Diffusion Models for Cross-Modality Image Translation](https://mlsanity.com/paper/2601.18623/) (2026)
- [Accelerating Diffusion Decoders via Multi-Scale Sampling and One-Step Distillation](https://mlsanity.com/paper/2603.19570/) (2026)

## Abstract (arXiv)

While diffusion-based generative models have made significant strides in visual content creation, conventional approaches face computational challenges, especially for high-resolution images, as they denoise the entire image from noisy inputs. This contrasts with signal processing techniques, such as Fourier and wavelet analyses, which often employ hierarchical decompositions. Inspired by such principles, particularly the idea of signal separation, we introduce a diffusion framework leveraging multi-scale latent factorization. Our framework uniquely decomposes the denoising target, typically latent features from a pretrained Variational Autoencoder, into a low-frequency base signal capturing core structural information and a high-frequency residual signal that contributes finer, high-frequency details like textures. This decomposition into base and residual components directly informs our two-stage image generation process, which first produces the low-resolution base, followed by the generation of the high-resolution residual. Our proposed architecture facilitates reduced sampling steps during the residual learning stage, owing to the inherent ease of modeling residual information, which confers advantages over conventional full-resolution generation techniques. This specific approach of decomposing the signal into a base and a residual, conceptually akin to how wavelet analysis can separate different frequency bands, yields a more streamlined and intuitive design distinct from generic hierarchical models. Our method, name (Multi-Scale Factorization), demonstrates its effectiveness by achieving FID scores of 2.08 (256×256) and 2.47 (512×512) on class-conditional ImageNet benchmarks, outperforming the DiT baseline (2.27 and 3.04 respectively) while also delivering a 4× speed-up with the same number of sampling steps.