One Sample is Enough to Make Conformal Prediction Robust

TL;DR

This paper introduces a novel robust conformal prediction method that achieves robustness with only a single model forward pass, significantly reducing computational costs while maintaining smaller prediction sets.

Contribution

It proposes RCP1, a single-sample robust conformal prediction method that certifies the conformal procedure itself, applicable to classification and regression tasks.

Findings

RCP1 achieves robustness with one forward pass.

RCP1 produces smaller average prediction sets.

The method is task-agnostic and extends to conformal risk control.

Abstract

For any black-box model, conformal prediction (CP) returns prediction sets guaranteed to include the true label with high adjustable probability. Robust CP (RCP) extends the guarantee to the worst case noise up to a pre-defined magnitude. For RCP, a well-established approach is to use randomized smoothing since it is applicable to any black-box model and provides smaller sets compared to deterministic methods. However, smoothing-based robustness requires many model forward passes per each input which is computationally expensive. We show that conformal prediction attains some robustness even with a single forward pass on a randomly perturbed input. Using any binary certificate we propose a single sample robust CP (RCP1). Our approach returns robust sets with smaller average set size compared to SOTA methods which use many (e.g. 100) passes per input. Our key insight is to certify the conformal procedure itself rather than individual conformity scores. Our approach is agnostic to the task (classification and regression). We further extend our approach to smoothing-based robust conformal risk control.