Exploring the Limits of Out-of-Distribution Detection

TL;DR

This paper demonstrates that large-scale pre-trained transformers significantly enhance near out-of-distribution detection across various data modalities, achieving new state-of-the-art results and effective few-shot outlier exposure methods.

Contribution

The paper introduces the use of large-scale pre-trained transformers for improved near OOD detection and explores novel few-shot outlier exposure techniques, including text-based outlier class identification.

Findings

Transformers improve OOD detection AUROC from 85% to over 96% on CIFAR benchmarks.

Unsupervised pre-training boosts AUROC from 66% to 77% in genomics OOD tasks.

Few-shot outlier exposure with transformers achieves AUROC over 98% with just 1-10 images per OOD class.

Abstract

Near out-of-distribution detection (OOD) is a major challenge for deep neural networks. We demonstrate that large-scale pre-trained transformers can significantly improve the state-of-the-art (SOTA) on a range of near OOD tasks across different data modalities. For instance, on CIFAR-100 vs CIFAR-10 OOD detection, we improve the AUROC from 85% (current SOTA) to more than 96% using Vision Transformers pre-trained on ImageNet-21k. On a challenging genomics OOD detection benchmark, we improve the AUROC from 66% to 77% using transformers and unsupervised pre-training. To further improve performance, we explore the few-shot outlier exposure setting where a few examples from outlier classes may be available; we show that pre-trained transformers are particularly well-suited for outlier exposure, and that the AUROC of OOD detection on CIFAR-100 vs CIFAR-10 can be improved to 98.7% with just 1 image per OOD class, and 99.46% with 10 images per OOD class. For multi-modal image-text pre-trained transformers such as CLIP, we explore a new way of using just the names of outlier classes as a sole source of information without any accompanying images, and show that this outperforms previous SOTA on standard vision OOD benchmark tasks.