arXiv 2406.008062024-06-02cs.LG

通过大语言模型设想离群点暴露以进行分布外检测

Envisioning Outlier Exposure by Large Language Models for Out-of-Distribution Detection

Hong Kong Baptist University（香港浸会大学）；Hefei University of Technology（合肥工业大学）；University of Nottingham（诺丁汉大学）；University of California, Santa Cruz（加州大学圣克鲁兹分校）；The University of Sydney（悉尼大学）；Sydney AI Centre（悉尼人工智能中心）

Chentao Cao, Zhun Zhong, Zhanke Zhou, Yang Liu, Tongliang Liu, Bo Han

备注：: ICML 2024

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出利用大语言模型（LLM）的专家知识和推理能力设想离群点暴露（EOE），通过视觉相似性生成离群标签和基于惩罚的评分函数，实现无需实际OOD数据的零样本分布外检测，在多任务中达到SOTA。

摘要：

在开放世界场景中部署机器学习模型时，检测分布外（OOD）样本至关重要。随着CLIP等视觉-语言模型的出现，无需在分布内（ID）数据上进行训练的零样本OOD检测成为可能。现有方法仅使用闭集标签构建基于文本的分类器。然而，这在很大程度上限制了CLIP从大型开放标签空间中识别样本的固有能力。在本文中，我们提出利用大语言模型（LLM）的专家知识和推理能力来设想潜在的离群点暴露（Envisioning potential Outlier Exposure，简称EOE），以解决这一限制，且无需访问任何实际的OOD数据。由于更好地适应开放世界场景，EOE可以推广到不同的任务，包括远、近和细粒度的OOD检测。在技术上，我们设计了（1）基于视觉相似性的LLM提示，以生成专门用于OOD检测的潜在离群类别标签，以及（2）基于潜在离群点惩罚的新评分函数，以有效区分困难OOD样本。在实验上，EOE在不同的OOD任务中实现了最先进的性能，并且可以有效地扩展到ImageNet-1K数据集。代码公开于：https://github.com/tmlr-group/EOE。

英文摘要：

Detecting out-of-distribution (OOD) samples is essential when deploying machine learning models in open-world scenarios. Zero-shot OOD detection, requiring no training on in-distribution (ID) data, has been possible with the advent of vision-language models like CLIP. Existing methods build a text-based classifier with only closed-set labels. However, this largely restricts the inherent capability of CLIP to recognize samples from large and open label space. In this paper, we propose to tackle this constraint by leveraging the expert knowledge and reasoning capability of large language models (LLM) to Envision potential Outlier Exposure, termed EOE, without access to any actual OOD data. Owing to better adaptation to open-world scenarios, EOE can be generalized to different tasks, including far, near, and fine-grained OOD detection. Technically, we design (1) LLM prompts based on visual similarity to generate potential outlier class labels specialized for OOD detection, as well as (2) a new score function based on potential outlier penalty to distinguish hard OOD samples effectively. Empirically, EOE achieves state-of-the-art performance across different OOD tasks and can be effectively scaled to the ImageNet-1K dataset. The code is publicly available at: https://github.com/tmlr-group/EOE.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。