Evaluating Out-of-Distribution Detectors Through Adversarial Generation of Outliers

TL;DR

This paper introduces EvG, a new evaluation protocol for OOD detectors that uses generative models and MCMC sampling to create realistic outliers, revealing weaknesses in current detectors.

Contribution

The paper proposes a novel evaluation method for OOD detectors using generative models and MCMC, providing more realistic outlier testing and uncovering new robustness issues.

Findings

EvG reveals weaknesses in current OOD detectors

Generative models produce more realistic outliers

Benchmark shows varied detector robustness

Abstract

A reliable evaluation method is essential for building a robust out-of-distribution (OOD) detector. Current robustness evaluation protocols for OOD detectors rely on injecting perturbations to outlier data. However, the perturbations are unlikely to occur naturally or not relevant to the content of data, providing a limited assessment of robustness. In this paper, we propose Evaluation-via-Generation for OOD detectors (EvG), a new protocol for investigating the robustness of OOD detectors under more realistic modes of variation in outliers. EvG utilizes a generative model to synthesize plausible outliers, and employs MCMC sampling to find outliers misclassified as in-distribution with the highest confidence by a detector. We perform a comprehensive benchmark comparison of the performance of state-of-the-art OOD detectors using EvG, uncovering previously overlooked weaknesses.