LoD: Loss-difference OOD Detection by Intentionally Label-Noisifying Unlabeled Wild Data

TL;DR

This paper introduces LoD, a novel framework for OOD detection that uses label-noisifying of unlabeled data to improve model safety without threshold tuning, supported by theoretical and experimental validation.

Contribution

LoD is the first to intentionally add label noise to unlabeled data to enhance OOD detection and eliminate threshold dependence, with a solid theoretical basis and extensive experiments.

Findings

LoD outperforms existing OOD detection methods.

LoD eliminates the need for threshold tuning in OOD filtering.

Theoretical analysis supports LoD's effectiveness.

Abstract

Using unlabeled wild data containing both in-distribution (ID) and out-of-distribution (OOD) data to improve the safety and reliability of models has recently received increasing attention. Existing methods either design customized losses for labeled ID and unlabeled wild data then perform joint optimization, or first filter out OOD data from the latter then learn an OOD detector. While achieving varying degrees of success, two potential issues remain: (i) Labeled ID data typically dominates the learning of models, inevitably making models tend to fit OOD data as IDs; (ii) The selection of thresholds for identifying OOD data in unlabeled wild data usually faces dilemma due to the unavailability of pure OOD samples. To address these issues, we propose a novel loss-difference OOD detection framework (LoD) by \textit{intentionally label-noisifying} unlabeled wild data. Such operations not only enable labeled ID data and OOD data in unlabeled wild data to jointly dominate the models' learning but also ensure the distinguishability of the losses between ID and OOD samples in unlabeled wild data, allowing the classic clustering technique (e.g., K-means) to filter these OOD samples without requiring thresholds any longer. We also provide theoretical foundation for LoD's viability, and extensive experiments verify its superiority.