arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2402.01345cs.CVcs.AIcs.CLcs.LG
收藏

Skip \ :一种减少大视觉语言模型幻觉的简单方法

Skip \n: A Simple Method to Reduce Hallucination in Large Vision-Language Models

National University of Singapore(新加坡国立大学)

Zongbo Han, Zechen Bai, Haiyang Mei, Qianli Xu, Changqing Zhang, Mike Zheng Shou

AI总结:

该研究发现大视觉语言模型中与段落分隔符\ \ 相关的语义偏移偏差是诱发多模态幻觉的关键因素,提出跳过\ 输出的简单方法,可有效缓解模型幻觉问题。

摘要:

大视觉语言模型(LVLMs)的最新进展展现出了借助人类语言理解视觉信息的出色能力。尽管取得了这些进展,LVLMs仍面临多模态幻觉的挑战,比如生成视觉信息中不存在的物体的文本描述。然而,多模态幻觉的潜在根本原因仍未得到充分探究。本文提出了一种新的视角,认为LVLMs的固有偏差可能是幻觉产生的一个关键因素。具体而言,我们系统地识别出一种与段落分隔符(\ \ )相关的语义偏移偏差:训练数据中“\ \ ”前后的内容往往呈现出显著的语义变化。这种模式会让模型推断“\ \ ”之后的内容应当与之前幻觉描述较少的内容存在明显差异,从而增加“\ \ ”之后出现幻觉描述的概率。我们已在多个公开可用的LVLMs上验证了这一假设。此外,我们发现,在生成的描述中刻意插入“\ \ ”会诱发更多幻觉。我们提出了一种简单的方法,通过跳过“\ ”的输出来有效缓解LVLMs的幻觉问题。

英文摘要:

Recent advancements in large vision-language models (LVLMs) have demonstrated impressive capability in visual information understanding with human language. Despite these advances, LVLMs still face challenges with multimodal hallucination, such as generating text descriptions of objects that are not present in the visual information. However, the underlying fundamental reasons of multimodal hallucinations remain poorly explored. In this paper, we propose a new perspective, suggesting that the inherent biases in LVLMs might be a key factor in hallucinations. Specifically, we systematically identify a semantic shift bias related to paragraph breaks (\n\n), where the content before and after '\n\n' in the training data frequently exhibit significant semantic changes. This pattern leads the model to infer that the contents following '\n\n' should be obviously different from the preceding contents with less hallucinatory descriptions, thereby increasing the probability of hallucinatory descriptions subsequent to the '\n\n'. We have validated this hypothesis on multiple publicly available LVLMs. Besides, we find that deliberately inserting '\n\n' at the generated description can induce more hallucinations. A simple method is proposed to effectively mitigate the hallucination of LVLMs by skipping the output of '\n'.

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。

↑