Analyzing and Mitigating Object Hallucination in Large Vision-Language Models

TL;DR

This paper introduces LURE, a post-hoc algorithm that significantly reduces object hallucination in large vision-language models by analyzing key factors like co-occurrence and uncertainty, improving description accuracy.

Contribution

The paper presents LURE, a novel, model-agnostic method for rectifying object hallucination in LVLMs through statistical analysis and reconstruction techniques.

Findings

LURE achieves 23% improvement in hallucination metrics.

LURE outperforms previous approaches in GPT and human evaluations.

LURE is seamlessly integrable with any LVLM.

Abstract

Large vision-language models (LVLMs) have shown remarkable abilities in understanding visual information with human languages. However, LVLMs still suffer from object hallucination, which is the problem of generating descriptions that include objects that do not actually exist in the images. This can negatively impact many vision-language tasks, such as visual summarization and reasoning. To address this issue, we propose a simple yet powerful algorithm, LVLM Hallucination Revisor (LURE), to post-hoc rectify object hallucination in LVLMs by reconstructing less hallucinatory descriptions. LURE is grounded in a rigorous statistical analysis of the key factors underlying object hallucination, including co-occurrence (the frequent appearance of certain objects alongside others in images), uncertainty (objects with higher uncertainty during LVLM decoding), and object position (hallucination often appears in the later part of the generated text). LURE can also be seamlessly integrated with any LVLMs. We evaluate LURE on six open-source LVLMs, achieving a 23% improvement in general object hallucination evaluation metrics over the previous best approach. In both GPT and human evaluations, LURE consistently ranks at the top. Our data and code are available at https://github.com/YiyangZhou/LURE.