Are foundation models for computer vision good conformal predictors?

TL;DR

This paper evaluates the uncertainty quantification capabilities of vision foundation models under Conformal Prediction, revealing their suitability, the impact of calibration, and the promise of APS in real-world scenarios.

Contribution

It provides a comprehensive empirical analysis of vision foundation models with conformal prediction, highlighting the effectiveness of Vision Transformers and APS for reliable uncertainty estimation.

Findings

Foundation models are suitable for conformalization, especially Vision Transformers.

Calibrating confidence predictions can degrade conformal set efficiency.

APS is promising for maintaining coverage guarantees in vision models.

Abstract

Recent advances in self-supervision and contrastive learning have brought the performance of foundation models to unprecedented levels in a variety of tasks. Fueled by this progress, these models are becoming the prevailing approach for a wide array of real-world vision problems, including risk-sensitive and high-stakes applications. However, ensuring safe deployment in these scenarios requires a more comprehensive understanding of their uncertainty modeling capabilities, which has received little attention. In this work, we delve into the behaviour of vision and vision-language foundation models under Conformal Prediction (CP), a statistical framework that provides theoretical guarantees of marginal coverage of the true class. Across extensive experiments including popular vision classification benchmarks, well-known foundation vision models, and three CP methods, our findings reveal that foundation models are well-suited for conformalization procedures, particularly those integrating Vision Transformers. We also show that calibrating the confidence predictions of these models, a popular strategy to improve their uncertainty quantification, actually leads to efficiency degradation of the conformal set on adaptive CP methods. Furthermore, few-shot adaptation of Vision-Language Models (VLMs) to downstream tasks, whose popularity is surging, enhances conformal scores compared to zero-shot predictions. Last, our empirical study exposes APS as particularly promising in the context of vision foundation models, as it does not violate the marginal coverage guarantees across multiple challenging, yet realistic scenarios.