Debiased Negative Mining Improves Out-of-distribution Detection with Pre-trained Vision-Language Models

2026-05-22

AI Summary

Current out-of-distribution (OOD) detection methods based on vision-language models suffer from performance limitations due to false negatives introduced during negative sample mining. This work proposes a debiased negative sampling mechanism that, for the first time, formulates the correction of negative label distribution bias as a tractable Monte Carlo sampling process. By leveraging labeled in-distribution data and unlabeled wild corpora to approximate the true negative label distribution, the approach effectively mitigates the false negative problem. Extensive experiments demonstrate that the proposed method consistently outperforms existing techniques across diverse OOD detection settings, establishing new state-of-the-art performance.

Abstract

Aiming at identifying unexpected inputs from unknown classes, out-of-distribution (OOD) detection has emerged as a pivotal approach to enhancing the reliability of machine learning models. This paper focuses on the burgeoning paradigm of post-hoc OOD detection with pre-trained vision-language models (VLMs), where a popular pipeline is to detect OOD inputs by examining their affinities between ID labels and negative labels, i.e., those semantically different from ID labels. Due to the unavailability of target OOD labels, existing works predominantly rely on heuristic rules to mine negative labels from unlabeled wild corpus data. Despite the empirical success, we argue that the power of VLM-based OOD detection has yet to be fully unleashed since the notorious false negative problem is far from addressed in the literature. With this motivation, we are interested in addressing the challenge of mining true negative labels for OOD scoring. To this end, we develop a theoretical framework for correcting the sampling bias of negatives labels by indirectly approximating the distribution of negative labels. Perhaps surprisingly, we show that the debiased negative mining can be naturally converted into Monte-Carlo sampling based on ID labels and the unlabeled wild corpus data. Extensive experiments empirically manifest that our method establishes a new state-of-the-art in a variety of OOD detection setups. Code is publicly available at \href{https://github.com/60pen9/Debiased-Negative-Mining-Improves-OOD-Detection-with-Pre-trained-VLMs}{\textcolor{red}{here}}.

Problem

• out-of-distribution detection
• vision-language models
• negative mining
• false negative
• sampling bias

Innovation

• Debiased Negative Mining
• Out-of-Distribution Detection
• Vision-Language Models
• Monte-Carlo Sampling
• False Negative Problem