Colored Noise Diffusion Sampling

Hugging Face Daily Papers Papers
diffusion-models image-synthesis sampling-method frequency-dependent spectral-bias stochastic-solver training-free

Summary

Introduces Colored Noise Sampling (CNS), a training-free stochastic solver for diffusion models that dynamically allocates energy based on frequency-dependent schedules, improving image quality metrics like FID significantly on ImageNet-256.

Diffusion models achieve state-of-the-art image synthesis, with their generative trajectories fundamentally exhibiting a spectral bias, resolving low-frequency global structures early and high-frequency fine details later. Conventional stochastic differential equation (SDE) solvers fail to account for this dynamic, naively injecting uniform white noise throughout the entire process and misusing the finite energy budget. In this work, we establish a mathematical framework that reconsiders SDE inference as a targeted, frequency-decoupled energy transfer. Leveraging this framework, we introduce Colored Noise Sampling (CNS), a novel, training-free stochastic solver. Rather than injecting uniform white noise, CNS utilizes a dynamic, timestep- and frequency-dependent schedule that more efficiently allocates injected energy toward structurally unresolved frequency bands. By actively exploiting the model's inherent spectral bias, CNS systematically steers the generated distribution toward the true data manifold. Extensive experiments demonstrate that CNS significantly outperforms standard ODE and SDE baselines as a strictly plug-and-play, inference-time sampler substitution across diverse architectures (SiT, JiT, FLUX). Compared to standard sampling on ImageNet-256, CNS achieves substantial unguided FID reductions, improving from 8.26 to 6.27 on SiT-XL/2, 32.39 to 26.69 on JiT-B/16, and 11.88 to 8.31 on JiT-H/16, while yielding consistent relative FID improvements with Classifier-Free Guidance. Project page is available at https://hadardavidson.github.io/CNS/.
Original Article
View Cached Full Text

Cached at: 05/29/26, 07:00 AM

Paper page - Colored Noise Diffusion Sampling

Source: https://huggingface.co/papers/2605.30332

Abstract

Diffusion models exhibit spectral bias in image synthesis, and a new sampling method called Colored Noise Sampling addresses this by dynamically allocating energy based on frequency-dependent schedules, leading to improved image quality metrics.

Diffusion modelsachieve state-of-the-art image synthesis, with their generative trajectories fundamentally exhibiting aspectral bias, resolving low-frequency global structures early and high-frequency fine details later. Conventionalstochastic differential equation(SDE) solvers fail to account for this dynamic, naively injecting uniformwhite noisethroughout the entire process and misusing the finite energy budget. In this work, we establish a mathematical framework that reconsiders SDE inference as a targeted, frequency-decoupledenergy transfer. Leveraging this framework, we introduceColored Noise Sampling(CNS), a novel, training-free stochastic solver. Rather than injecting uniformwhite noise, CNS utilizes a dynamic, timestep- andfrequency-dependent schedulethat more efficiently allocates injected energy toward structurally unresolved frequency bands. By actively exploiting the model’s inherentspectral bias, CNS systematically steers the generated distribution toward the true data manifold. Extensive experiments demonstrate that CNS significantly outperforms standardODEand SDE baselines as a strictly plug-and-play, inference-time sampler substitution across diverse architectures (SiT, JiT, FLUX). Compared to standard sampling on ImageNet-256, CNS achieves substantial unguidedFIDreductions, improving from 8.26 to 6.27 on SiT-XL/2, 32.39 to 26.69 on JiT-B/16, and 11.88 to 8.31 on JiT-H/16, while yielding consistent relativeFIDimprovements withClassifier-Free Guidance. Project page is available at https://hadardavidson.github.io/CNS/.

View arXiv pageView PDFProject pageGitHub0Add to collection

Get this paper in your agent:

hf papers read 2605\.30332

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.30332 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.30332 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.30332 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Class-frequency Guided Noise Schedule for Diffusion Models

arXiv cs.LG

This paper proposes a class-frequency guided noise schedule for diffusion models that assigns larger-scale noises to low-frequency classes to improve generation quality on imbalanced datasets, demonstrating substantial improvements over baselines.

Constrained Code Generation with Discrete Diffusion

arXiv cs.CL

This paper introduces Constrained Diffusion for Code (CDC), a training-free neurosymbolic inference framework that integrates constraint satisfaction directly into the reverse denoising process of discrete diffusion models for code generation. CDC consistently improves constraint satisfaction in functional correctness, security, and syntax across benchmarks, outperforming existing diffusion and autoregressive baselines.

Cyclic Denoising Reveals Ultrastable Memories in Diffusion Models

arXiv cs.LG

Cyclic denoising is introduced as a novel extraction attack that reveals ultrastable memorized training images in diffusion models by repeatedly noising and denoising samples. The technique requires no gradients or weight inspection and has implications for privacy auditing.