Detecting and Mitigating Hallucination in Large Vision Language Models via Fine-Grained AI Feedback

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xiao, Wenyi, Huang, Ziwei, Gan, Leilei, He, Wanggui, Li, Haoyuan, Yu, Zhelun, Shu, Fangxun, Jiang, Hao, Zhu, Linchao
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917883719712768
author Xiao, Wenyi
Huang, Ziwei
Gan, Leilei
He, Wanggui
Li, Haoyuan
Yu, Zhelun
Shu, Fangxun
Jiang, Hao
Zhu, Linchao
author_facet Xiao, Wenyi
Huang, Ziwei
Gan, Leilei
He, Wanggui
Li, Haoyuan
Yu, Zhelun
Shu, Fangxun
Jiang, Hao
Zhu, Linchao
contents The rapidly developing Large Vision Language Models (LVLMs) have shown notable capabilities on a range of multi-modal tasks, but still face the hallucination phenomena where the generated texts do not align with the given contexts, significantly restricting the usages of LVLMs. Most previous work detects and mitigates hallucination at the coarse-grained level or requires expensive annotation (e.g., labeling by proprietary models or human experts). To address these issues, we propose detecting and mitigating hallucinations in LVLMs via fine-grained AI feedback. The basic idea is that we generate a small-size sentence-level hallucination annotation dataset by proprietary models, whereby we train a hallucination detection model which can perform sentence-level hallucination detection, covering primary hallucination types (i.e., object, attribute, and relationship). Then, we propose a detect-then-rewrite pipeline to automatically construct preference dataset for training hallucination mitigating model. Furthermore, we propose differentiating the severity of hallucinations, and introducing a Hallucination Severity-Aware Direct Preference Optimization (HSA-DPO) for mitigating hallucination in LVLMs by incorporating the severity of hallucinations into preference learning. Extensive experiments demonstrate the effectiveness of our method.
format Preprint
id arxiv_https___arxiv_org_abs_2404_14233
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Detecting and Mitigating Hallucination in Large Vision Language Models via Fine-Grained AI Feedback
Xiao, Wenyi
Huang, Ziwei
Gan, Leilei
He, Wanggui
Li, Haoyuan
Yu, Zhelun
Shu, Fangxun
Jiang, Hao
Zhu, Linchao
Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
Machine Learning
The rapidly developing Large Vision Language Models (LVLMs) have shown notable capabilities on a range of multi-modal tasks, but still face the hallucination phenomena where the generated texts do not align with the given contexts, significantly restricting the usages of LVLMs. Most previous work detects and mitigates hallucination at the coarse-grained level or requires expensive annotation (e.g., labeling by proprietary models or human experts). To address these issues, we propose detecting and mitigating hallucinations in LVLMs via fine-grained AI feedback. The basic idea is that we generate a small-size sentence-level hallucination annotation dataset by proprietary models, whereby we train a hallucination detection model which can perform sentence-level hallucination detection, covering primary hallucination types (i.e., object, attribute, and relationship). Then, we propose a detect-then-rewrite pipeline to automatically construct preference dataset for training hallucination mitigating model. Furthermore, we propose differentiating the severity of hallucinations, and introducing a Hallucination Severity-Aware Direct Preference Optimization (HSA-DPO) for mitigating hallucination in LVLMs by incorporating the severity of hallucinations into preference learning. Extensive experiments demonstrate the effectiveness of our method.
title Detecting and Mitigating Hallucination in Large Vision Language Models via Fine-Grained AI Feedback
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
Machine Learning
url https://arxiv.org/abs/2404.14233