Conffusion: Confidence Intervals for Diffusion Models

TL;DR

Conffusion introduces a method to generate confidence intervals for diffusion model outputs efficiently, providing statistical guarantees and outperforming baseline methods in speed and accuracy.

Contribution

We propose Conffusion, a fine-tuning approach that predicts confidence interval bounds in a single pass, significantly improving speed and bounds quality over traditional sampling methods.

Findings

Conffusion is three orders of magnitude faster than baseline methods.

It provides statistically guaranteed confidence intervals for diffusion model outputs.

Conffusion outperforms baseline methods in bounds accuracy.

Abstract

Diffusion models have become the go-to method for many generative tasks, particularly for image-to-image generation tasks such as super-resolution and inpainting. Current diffusion-based methods do not provide statistical guarantees regarding the generated results, often preventing their use in high-stakes situations. To bridge this gap, we construct a confidence interval around each generated pixel such that the true value of the pixel is guaranteed to fall within the interval with a probability set by the user. Since diffusion models parametrize the data distribution, a straightforward way of constructing such intervals is by drawing multiple samples and calculating their bounds. However, this method has several drawbacks: i) slow sampling speeds ii) suboptimal bounds iii) requires training a diffusion model per task. To mitigate these shortcomings we propose Conffusion, wherein we fine-tune a pre-trained diffusion model to predict interval bounds in a single forward pass. We show that Conffusion outperforms the baseline method while being three orders of magnitude faster.