Quantifying Epistemic Uncertainty in Diffusion Models

2026-02-09

AI Summary

Existing methods struggle to reliably quantify epistemic uncertainty in diffusion models and often conflate it with aleatoric uncertainty. This work proposes FLARE, the first approach to explicitly disentangle epistemic uncertainty in diffusion models, leveraging Fisher information for scalable uncertainty estimation. FLARE efficiently approximates the Fisher information matrix using stochastic subsets of model parameters, enabling tractable and accurate uncertainty quantification. Theoretical analysis reveals that last-layer Laplace approximation is insufficient for this task, whereas FLARE more faithfully captures epistemic variance. Empirical evaluation on synthetic time-series generation tasks demonstrates that FLARE significantly outperforms existing methods, offering superior reliability and practical utility in both uncertainty estimation and data filtering.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Computer Vision: Diffusion Models for Vision

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

To ensure high quality outputs, it is important to quantify the epistemic uncertainty of diffusion models.Existing methods are often unreliable because they mix epistemic and aleatoric uncertainty. We introduce a method based on Fisher information that explicitly isolates epistemic variance, producing more reliable plausibility scores for generated data. To make this approach scalable, we propose FLARE (Fisher-Laplace Randomized Estimator), which approximates the Fisher information using a uniformly random subset of model parameters. Empirically, FLARE improves uncertainty estimation in synthetic time-series generation tasks, achieving more accurate and reliable filtering than other methods. Theoretically, we bound the convergence rate of our randomized approximation and provide analytic and empirical evidence that last-layer Laplace approximations are insufficient for this task.

Problem

• epistemic uncertainty
• diffusion models
• aleatoric uncertainty
• uncertainty quantification
• Fisher information

Innovation

• epistemic uncertainty
• diffusion models
• Fisher information
• FLARE
• Laplace approximation