Stable Localized Conformal Prediction via Transduction

2026-05-02

AI Summary

Existing conformal prediction methods suffer from high variability and poor stability in prediction set sizes under limited calibration data, with localized approaches being particularly affected. This work proposes Stable Conformal Prediction (StCP), which formally defines prediction set stability for the first time and introduces a cross-task conformal prediction framework that requires no additional target labels. StCP leverages transductive transfer learning to integrate labeled source-task data with unlabeled target-task data. Theoretical analysis demonstrates that StCP simultaneously guarantees marginal coverage and stability. Empirical results show that StCP significantly outperforms both standard and localized conformal prediction methods under few-shot calibration settings.

Abstract

Existing evaluations of conformal prediction, such as prediction efficiency and test-conditional coverage, are defined in expectation over the calibration data. In practice, when only one calibration set of limited size is available, prediction sets often exhibit high variability in size, especially for methods with localization. We formalize this concern as set stability, defined as the variance of the conditional expectation of the set size given the calibration data. To improve stability without requiring additional target-task labels, we propose Stable Conformal Prediction (StCP), a transfer learning approach that utilizes labeled source-task data and unlabeled target data. Theoretically, we characterize the marginal coverage and stability of StCP; empirically, it delivers more stable prediction sets than standard conformal prediction methods, especially for those with localization, when calibration data are limited.

Problem

• conformal prediction
• set stability
• localization
• calibration data
• prediction set variability

Innovation

• Stable Conformal Prediction
• Set Stability
• Transductive Transfer Learning
• Localized Conformal Prediction
• Calibration Data Efficiency