Elucidating the SNR-t Bias of Diffusion Probabilistic Models

Hugging Face Daily Papers Papers

Summary

This paper identifies a Signal-to-Noise Ratio timestep (SNR-t) bias in diffusion probabilistic models during inference, where SNR-timestep alignment from training is disrupted at inference time. The authors propose a differential correction method that decomposes samples into frequency components and corrects each separately, improving generation quality across models like IDDPM, ADM, DDIM, EDM, and FLUX with minimal computational overhead.

Diffusion Probabilistic Models have demonstrated remarkable performance across a wide range of generative tasks. However, we have observed that these models often suffer from a Signal-to-Noise Ratio-timestep (SNR-t) bias. This bias refers to the misalignment between the SNR of the denoising sample and its corresponding timestep during the inference phase. Specifically, during training, the SNR of a sample is strictly coupled with its timestep. However, this correspondence is disrupted during inference, leading to error accumulation and impairing the generation quality. We provide comprehensive empirical evidence and theoretical analysis to substantiate this phenomenon and propose a simple yet effective differential correction method to mitigate the SNR-t bias. Recognizing that diffusion models typically reconstruct low-frequency components before focusing on high-frequency details during the reverse denoising process, we decompose samples into various frequency components and apply differential correction to each component individually. Extensive experiments show that our approach significantly improves the generation quality of various diffusion models (IDDPM, ADM, DDIM, A-DPM, EA-DPM, EDM, PFGM++, and FLUX) on datasets of various resolutions with negligible computational overhead. The code is at https://github.com/AMAP-ML/DCW.
Original Article
View Cached Full Text

Cached at: 04/21/26, 07:21 AM

Paper page - Elucidating the SNR-t Bias of Diffusion Probabilistic Models

Source: https://huggingface.co/papers/2604.16044

Abstract

Diffusion probabilistic models suffer from SNR-timestep bias during inference, which is addressed through a differential correction method that processes frequency components separately, improving generation quality across multiple models with minimal computational cost.

Diffusion Probabilistic Modelshave demonstrated remarkable performance across a wide range of generative tasks. However, we have observed that these models often suffer from a Signal-to-Noise Ratio-timestep (SNR-t) bias. This bias refers to the misalignment between the SNR of thedenoising sampleand its corresponding timestep during theinference phase. Specifically, during training, the SNR of a sample is strictly coupled with its timestep. However, this correspondence is disrupted during inference, leading toerror accumulationand impairing thegeneration quality. We provide comprehensive empirical evidence and theoretical analysis to substantiate this phenomenon and propose a simple yet effectivedifferential correctionmethod to mitigate the SNR-t bias. Recognizing that diffusion models typically reconstruct low-frequency componentsbefore focusing on high-frequency details during thereverse denoising process, we decompose samples into variousfrequency componentsand applydifferential correctionto each component individually. Extensive experiments show that our approach significantly improves thegeneration qualityof various diffusion models (IDDPM, ADM, DDIM, A-DPM, EA-DPM, EDM, PFGM++, and FLUX) on datasets of various resolutions with negligible computational overhead. The code is at https://github.com/AMAP-ML/DCW.

View arXiv pageView PDFGitHub71Add to collection

Get this paper in your agent:

hf papers read 2604\.16044

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/2604.16044 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2604.16044 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2604.16044 in a Space README.md to link it from this page.

Collections including this paper3

Similar Articles

Temporal Difference Learning for Diffusion Models

arXiv cs.LG

This paper introduces a temporal difference (TD) learning objective for diffusion models that enforces cross-time consistency along the denoising trajectory. It reformulates denoising as a reinforcement learning policy evaluation problem, showing significant improvements in sample quality (FID), especially for few-step samplers.

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.