Conformal Prediction Sets with Improved Conditional Coverage using Trust Scores

2025-01-17

AI Summary

Standard conformal prediction guarantees only marginal coverage, failing to ensure conditional coverage for critical subgroups—such as high-confidence misclassified samples or sensitive demographic groups. To address this, we propose a novel conformal calibration framework that jointly leverages classifier confidence scores and nonparametric trust scores. This work is the first to incorporate trust scores into conformal prediction, replacing conventional univariate calibration with bivariate quantile calibration over the “confidence + trust” score pair. Our method comprises three components: nonparametric trust estimation, extension of marginal conformal prediction, and a joint calibration mechanism. Extensive evaluation across multiple image datasets demonstrates that our approach improves class-conditional coverage by 12–28%, enhances stability of subgroup and sensitive-group coverage, and reduces coverage deviation by over 40%. These gains significantly improve model fairness and reliability—particularly under data scarcity.

Technology Category

Machine Learning: Calibration & Uncertainty QuantificationComputer Vision: Bias, Fairness & PrivacyReasoning under Uncertainty: Relational Probabilistic Models

Application Category

User Modeling, Personalization and Recommendation: Fairness-aware retrieval and rankingSearch and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingSecurity and Privacy: Data transparency and provenance

Abstract

Standard conformal prediction offers a marginal guarantee on coverage, but for prediction sets to be truly useful, they should ideally ensure coverage conditional on each test point. Unfortunately, it is impossible to achieve exact, distribution-free conditional coverage in finite samples. In this work, we propose an alternative conformal prediction algorithm that targets coverage where it matters most--in instances where a classifier is overconfident in its incorrect predictions. We start by dissecting miscoverage events in marginally-valid conformal prediction, and show that miscoverage rates vary based on the classifier's confidence and its deviation from the Bayes optimal classifier. Motivated by this insight, we develop a variant of conformal prediction that targets coverage conditional on a reduced set of two variables: the classifier's confidence in a prediction and a nonparametric trust score that measures its deviation from the Bayes classifier. Empirical evaluation on multiple image datasets shows that our method generally improves conditional coverage properties compared to standard conformal prediction, including class-conditional coverage, coverage over arbitrary subgroups, and coverage over demographic groups.

Problem

• Limited Data
• Prediction Accuracy
• Overconfident Misclassifications

Innovation

• Confidence-aware Classification
• Trust Score
• Enhanced Prediction Accuracy