QUTCC: Quantile Uncertainty Training and Conformal Calibration for Imaging Inverse Problems

ICLR 2026 Conference Submission13847 Authors

Keywords: inverse problems, uncertainty quantification, conformal calibration

TL;DR: We propose QUTCC, a quantile uncertainty training and calibration technique that enables nonlinear, non-uniform scaling of quantile predictions to enable tighter uncertainty estimates in imaging inverse problems.

Abstract: Deep learning models often hallucinate, producing realistic artifacts that are not truly present in the sample. This can have dire consequences for scientific and medical inverse problems, where accuracy is more important than perceptual quality. Uncertainty quantification techniques, such as conformal prediction, can pinpoint outliers and provide guarantees for image regression tasks, improving reliability. However, existing methods predict fixed quantiles and utilize a linear constant scaling factor to calibrate uncertainty bounds, resulting in larger, less informative bounds. We propose QUTCC, a quantile uncertainty training and calibration technique that enables nonlinear, non-uniform scaling of quantile predictions to enable tighter uncertainty estimates. Using a U-Net architecture with a quantile embedding, QUTCC can predict the full conditional distribution of quantiles for each image. After conformal calibration, QUTCC can predict pixel-wise uncertainty intervals that satisfy coverage guarantees and also estimate a pixel-wise conditional probability density function. We evaluate our method on image denoising, quantitative phase imaging, and compressive MRI reconstruction. Our method successfully pinpoints hallucinations in image estimates and consistently achieves tighter uncertainty intervals than prior methods while maintaining the same statistical coverage.

Primary Area: applications to computer vision, audio, language, and other modalities

Submission Number: 13847