Trust, but Verify: Using Self-Supervised Probing to Improve Trustworthiness

TL;DR

This paper introduces a self-supervised probing framework to assess and reduce overconfidence in deep learning models, thereby enhancing their trustworthiness across multiple tasks and benchmarks.

Contribution

It proposes a novel, flexible self-supervised probing method that improves trustworthiness of models by addressing overconfidence issues, compatible with existing methods.

Findings

Effective in misclassification detection

Improves calibration of confidence scores

Enhances out-of-distribution detection

Abstract

Trustworthy machine learning is of primary importance to the practical deployment of deep learning models. While state-of-the-art models achieve astonishingly good performance in terms of accuracy, recent literature reveals that their predictive confidence scores unfortunately cannot be trusted: e.g., they are often overconfident when wrong predictions are made, or so even for obvious outliers. In this paper, we introduce a new approach of self-supervised probing, which enables us to check and mitigate the overconfidence issue for a trained model, thereby improving its trustworthiness. We provide a simple yet effective framework, which can be flexibly applied to existing trustworthiness-related methods in a plug-and-play manner. Extensive experiments on three trustworthiness-related tasks (misclassification detection, calibration and out-of-distribution detection) across various benchmarks verify the effectiveness of our proposed probing framework.