Temporal Alignment Guidance: On-Manifold Sampling in Diffusion Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Park, Youngrok, Jung, Hojung, Bae, Sangmin, Yun, Se-Young
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917007477178368
author Park, Youngrok
Jung, Hojung
Bae, Sangmin
Yun, Se-Young
author_facet Park, Youngrok
Jung, Hojung
Bae, Sangmin
Yun, Se-Young
contents Diffusion models have achieved remarkable success as generative models. However, even a well-trained model can accumulate errors throughout the generation process. These errors become particularly problematic when arbitrary guidance is applied to steer samples toward desired properties, which often breaks sample fidelity. In this paper, we propose a general solution to address the off-manifold phenomenon observed in diffusion models. Our approach leverages a time predictor to estimate deviations from the desired data manifold at each timestep, identifying that a larger time gap is associated with reduced generation quality. We then design a novel guidance mechanism, `Temporal Alignment Guidance' (TAG), attracting the samples back to the desired manifold at every timestep during generation. Through extensive experiments, we demonstrate that TAG consistently produces samples closely aligned with the desired manifold at each timestep, leading to significant improvements in generation quality across various downstream tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2510_11057
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Temporal Alignment Guidance: On-Manifold Sampling in Diffusion Models
Park, Youngrok
Jung, Hojung
Bae, Sangmin
Yun, Se-Young
Machine Learning
Artificial Intelligence
Diffusion models have achieved remarkable success as generative models. However, even a well-trained model can accumulate errors throughout the generation process. These errors become particularly problematic when arbitrary guidance is applied to steer samples toward desired properties, which often breaks sample fidelity. In this paper, we propose a general solution to address the off-manifold phenomenon observed in diffusion models. Our approach leverages a time predictor to estimate deviations from the desired data manifold at each timestep, identifying that a larger time gap is associated with reduced generation quality. We then design a novel guidance mechanism, `Temporal Alignment Guidance' (TAG), attracting the samples back to the desired manifold at every timestep during generation. Through extensive experiments, we demonstrate that TAG consistently produces samples closely aligned with the desired manifold at each timestep, leading to significant improvements in generation quality across various downstream tasks.
title Temporal Alignment Guidance: On-Manifold Sampling in Diffusion Models
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2510.11057