Exploring Causes and Mitigation of Hallucinations in Large Vision Language Models

TL;DR

This paper investigates hallucinations in Large Vision-Language Models, analyzing their patterns and developing a detection and mitigation pipeline to improve caption accuracy during inference.

Contribution

It introduces an automated hallucination detection pipeline and a decoding strategy to reduce hallucinations in image captioning tasks.

Findings

Not all tokens are influenced by image input during captioning.

Image dependency signals can detect hallucinations effectively.

The proposed decoding strategy reduces hallucination rates at inference.

Abstract

Large Vision-Language Models (LVLMs) integrate image encoders with Large Language Models (LLMs) to process multi-modal inputs and perform complex visual tasks. However, they often generate hallucinations by describing non-existent objects or attributes, compromising their reliability. This study analyzes hallucination patterns in image captioning, showing that not all tokens in the generation process are influenced by image input and that image dependency can serve as a useful signal for hallucination detection. To address this, we develop an automated pipeline to identify hallucinated objects and train a token-level classifier using hidden representations from parallel inference passes-with and without image input. Leveraging this classifier, we introduce a decoding strategy that effectively controls hallucination rates in image captioning at inference time.