Generative Uncertainty in Diffusion Models
This addresses the challenge of ensuring sample quality in generative models for users relying on automated outputs, though it is incremental as it builds on existing diffusion models.
The paper tackles the problem of detecting low-quality samples in diffusion models by proposing a Bayesian framework for estimating generative uncertainty, demonstrating that it effectively identifies poor-quality samples and significantly outperforms existing uncertainty-based methods.
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.