Generative Uncertainty in Diffusion Models

2025/02/28 by Jazbec, Metod, Wong-Toi, Eliot, Xia, Guoxuan +3 · 2 citations
#Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Machine Learning (cs.LG)

paper · doi:10.48550/arxiv.2502.20946

Abstract

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.

Cited by

• Control-Augmented Autoregressive Diffusion for Data Assimilation
• Mitigating Diffusion Model Hallucinations with Dynamic Guidance

Related

• Shedding Light on Large Generative Networks: Estimating Epistemic Uncertainty in Diffusion Models
• Generative Modeling with Bayesian Sample Inference
• Diffusion Model Guided Sampling with Pixel-Wise Aleatoric Uncertainty Estimation
• Optimizing Decoding Paths in Masked Diffusion Models by Quantifying Uncertainty
• BayesDiff: Estimating Pixel-wise Uncertainty in Diffusion via Bayesian Inference