Evaluation and Analysis of Hallucination in Large Vision-Language Models

Junyang Wang, Yiyang Zhou, Guohai Xu, Pengcheng Shi, Chenlin Zhao +7 more
8/29/2023
cs.LGcs.AIcs.CL

Official code: junyangwang0410/HaELM (19 stars, last commit 2023-09-26).

Abstract

Large Vision-Language Models (LVLMs) have recently achieved remarkable success. However, LVLMs are still plagued by the hallucination problem, which limits the practicality in many scenarios. Hallucination refers to the information of LVLMs' responses that does not exist in the visual input, which poses potential risks of substantial consequences. There has been limited work studying hallucination evaluation in LVLMs. In this paper, we propose Hallucination Evaluation based on Large Language Models (HaELM), an LLM-based hallucination evaluation framework. HaELM achieves an approximate 95% performance comparable to ChatGPT and has additional advantages including low cost, reproducibility, privacy preservation and local deployment. Leveraging the HaELM, we evaluate the hallucination in current LVLMs. Furthermore, we analyze the factors contributing to hallucination in LVLMs and offer helpful suggestions to mitigate the hallucination problem. Our training data and human annotation hallucination data will be made public soon.

View on arXivView PDF

Code Implementations(1)

junyangwang0410/HaELMOfficial100%

An automatic MLLM hallucination detection framework

190PythonSep 26, 20233 years ago

Similar Papers

A Survey on Hallucination in Large Vision-Language Models

94% match1 repo962024

Hal-Eval: A Universal and Fine-grained Hallucination Evaluation Framework for Large Vision Language Models

94% match1 repo112024

A Comprehensive Analysis for Visual Object Hallucination in Large Vision-Language Models

92% match2025

Reference-free Hallucination Detection for Large Vision-Language Models

92% match1 repo42024

Analyzing and Mitigating Object Hallucination in Large Vision-Language Models

92% match1 repo1592026

Hallucination of Multimodal Large Language Models: A Survey

91% match1 repo1.0k2024

Cite this paper

@article{wang2023evaluation,
  title  = {Evaluation and Analysis of Hallucination in Large Vision-Language Models},
  author = {Junyang Wang and Yiyang Zhou and Guohai Xu and Pengcheng Shi and Chenlin Zhao and Haiyang Xu and Qinghao Ye and Ming Yan and Ji Zhang and Jihua Zhu and Jitao Sang and Haoyu Tang},
  year   = {2023},
  eprint = {2308.15126},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2308.15126v3}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2308.15126.svg)](https://codeofpaper.com/paper/2308.15126)

Discussion