arXiv 2410.231142025-07-17cs.CVcs.AIcs.CLcs.LG

面向大型视觉-语言模型的统一三元组级幻觉评估

Unified Triplet-Level Hallucination Evaluation for Large Vision-Language Models

The Hong Kong University of Science and Technology（香港科学与技术大学）

Junjie Wu, Tsz Ting Chung, Kai Chen, Dit-Yan Yeung

备注：: Accepted by TMLR 2025. Project Page: https://kaichen1998.github.io/projects/tri-he/

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出统一三元组级幻觉评估框架和Tri-HE基准，同时衡量大型视觉-语言模型的对象幻觉与关系幻觉，发现关系幻觉更严重，并设计免训练方法有效缓解幻觉。

摘要：

尽管大型视觉-语言模型（LVLMs）在视觉-语言推理中表现出色，但它们可能生成给定图像中并不存在的幻觉内容。现有的大多数LVLM幻觉基准仅限于评估与对象相关的幻觉。然而，两个对象之间关系上的潜在幻觉，即关系幻觉，仍缺乏研究。为弥补这一不足，我们设计了一个统一框架，以同时衡量LVLMs中的对象幻觉和关系幻觉。该框架的核心思想是通过从LVLMs的响应中提取的（对象，关系，对象）三元组来评估幻觉，使其易于推广到不同的视觉-语言任务。基于该框架，我们进一步提出了Tri-HE，一个新颖的三元组级幻觉评估基准，可用于同时研究对象幻觉和关系幻觉。通过在Tri-HE上的全面评估，我们观察到在现有LVLMs中，关系幻觉问题甚至比对象幻觉更为严重，这突显了一个此前被忽视的、影响LVLMs可靠性的问题。此外，基于我们的发现，我们设计了一种简单的免训练方法，可有效缓解LVLMs的幻觉。我们的数据集和用于复现实验的代码已公开，地址为https://github.com/wujunjie1998/Tri-HE。

英文摘要：

Despite the outstanding performance in vision-language reasoning, Large Vision-Language Models (LVLMs) might generate hallucinated contents that do not exist in the given image. Most existing LVLM hallucination benchmarks are constrained to evaluate the object-related hallucinations. However, the potential hallucination on the relations between two objects, i.e., relation hallucination, still lacks investigation. To remedy that, we design a unified framework to measure the object and relation hallucination in LVLMs simultaneously. The core idea of our framework is to evaluate hallucinations via (object, relation, object) triplets extracted from LVLMs' responses, making it easily generalizable to different vision-language tasks. Based on our framework, we further introduce Tri-HE, a novel Triplet-level Hallucination Evaluation benchmark which can be used to study both object and relation hallucination at the same time. With comprehensive evaluations on Tri-HE, we observe that the relation hallucination issue is even more serious than object hallucination among existing LVLMs, highlighting a previously neglected problem towards reliable LVLMs. Moreover, based on our findings, we design a simple training-free approach that effectively mitigates hallucinations for LVLMs. Our dataset and code for the reproduction of our experiments are available publicly at https://github.com/wujunjie1998/Tri-HE.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。