Learning Transferable Negative Prompts for Out-of-Distribution Detection

TL;DR

NegPrompt introduces negative prompts learned solely from ID data to improve out-of-distribution detection, effectively handling open-vocabulary scenarios and surpassing existing methods in various benchmarks.

Contribution

The paper proposes NegPrompt, a novel method that learns transferable negative prompts from ID data for enhanced OOD detection without external outlier data.

Findings

Outperforms state-of-the-art prompt-learning-based OOD detection methods.

Effective in both closed- and open-vocabulary classification scenarios.

Maintains high detection accuracy with novel class labels.

Abstract

Existing prompt learning methods have shown certain capabilities in Out-of-Distribution (OOD) detection, but the lack of OOD images in the target dataset in their training can lead to mismatches between OOD images and In-Distribution (ID) categories, resulting in a high false positive rate. To address this issue, we introduce a novel OOD detection method, named 'NegPrompt', to learn a set of negative prompts, each representing a negative connotation of a given class label, for delineating the boundaries between ID and OOD images. It learns such negative prompts with ID data only, without any reliance on external outlier data. Further, current methods assume the availability of samples of all ID classes, rendering them ineffective in open-vocabulary learning scenarios where the inference stage can contain novel ID classes not present during training. In contrast, our learned negative prompts are transferable to novel class labels. Experiments on various ImageNet benchmarks show that NegPrompt surpasses state-of-the-art prompt-learning-based OOD detection methods and maintains a consistent lead in hard OOD detection in closed- and open-vocabulary classification scenarios. Code is available at https://github.com/mala-lab/negprompt.