Generative Uncertainty in Diffusion Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Jazbec, Metod, Wong-Toi, Eliot, Xia, Guoxuan, Zhang, Dan, Nalisnick, Eric, Mandt, Stephan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918056121335808
author Jazbec, Metod
Wong-Toi, Eliot
Xia, Guoxuan
Zhang, Dan
Nalisnick, Eric
Mandt, Stephan
author_facet Jazbec, Metod
Wong-Toi, Eliot
Xia, Guoxuan
Zhang, Dan
Nalisnick, Eric
Mandt, Stephan
contents Diffusion models have recently driven significant breakthroughs in generative modeling. While state-of-the-art models produce high-quality samples on average, individual samples can still be low quality. Detecting such samples without human inspection remains a challenging task. To address this, we propose a Bayesian framework for estimating generative uncertainty of synthetic samples. We outline how to make Bayesian inference practical for large, modern generative models and introduce a new semantic likelihood (evaluated in the latent space of a feature extractor) to address the challenges posed by high-dimensional sample spaces. Through our experiments, we demonstrate that the proposed generative uncertainty effectively identifies poor-quality samples and significantly outperforms existing uncertainty-based methods. Notably, our Bayesian framework can be applied post-hoc to any pretrained diffusion or flow matching model (via the Laplace approximation), and we propose simple yet effective techniques to minimize its computational overhead during sampling.
format Preprint
id arxiv_https___arxiv_org_abs_2502_20946
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Generative Uncertainty in Diffusion Models
Jazbec, Metod
Wong-Toi, Eliot
Xia, Guoxuan
Zhang, Dan
Nalisnick, Eric
Mandt, Stephan
Machine Learning
Artificial Intelligence
Diffusion models have recently driven significant breakthroughs in generative modeling. While state-of-the-art models produce high-quality samples on average, individual samples can still be low quality. Detecting such samples without human inspection remains a challenging task. To address this, we propose a Bayesian framework for estimating generative uncertainty of synthetic samples. We outline how to make Bayesian inference practical for large, modern generative models and introduce a new semantic likelihood (evaluated in the latent space of a feature extractor) to address the challenges posed by high-dimensional sample spaces. Through our experiments, we demonstrate that the proposed generative uncertainty effectively identifies poor-quality samples and significantly outperforms existing uncertainty-based methods. Notably, our Bayesian framework can be applied post-hoc to any pretrained diffusion or flow matching model (via the Laplace approximation), and we propose simple yet effective techniques to minimize its computational overhead during sampling.
title Generative Uncertainty in Diffusion Models
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2502.20946