arXiv 2408.005502024-08-01cs.CVcs.AIcs.CL

缓解大型视觉语言模型中的多语言幻觉

Mitigating Multilingual Hallucination in Large Vision-Language Models

School of Computer Science & Technology, Huazhong University of Science and Technology（华中科技大学计算机科学与技术学院）；School of Computing Science and Technology, Fudan University（复旦大学计算机科学技术学院）；Wangxuan Institute of Computer Technology, Peking University（北京大学王选计算机研究所）；School of Computing Science and Technology, Zhejiang Gongshang University（浙江工商大学计算机科学与技术学院）；Department of Computer Science and Engineering, The Chinese University of Hong Kong（香港中文大学计算机科学与工程学系）

Xiaoye Qu, Mingyang Song, Wei Wei, Jianfeng Dong, Yu Cheng

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文首次针对大型视觉语言模型的多语言幻觉问题，提出两阶段多语言幻觉去除框架MHR，通过跨语言对齐与直接偏好优化，在13种语言上平均提升准确率19.0%。

摘要：

尽管大型视觉语言模型（LVLMs）在广泛任务中展现出卓越能力，但它们仍存在幻觉问题，即模型在给定图像-查询输入时会生成看似合理但错误的答案。当以非英语语言查询图像时，这种幻觉现象更为严重，而现有缓解LVLMs幻觉的方法仅考虑英语场景。本文首次尝试缓解LVLMs中这一重要的多语言幻觉问题。通过深入的实验分析，我们发现LVLMs中的多语言幻觉是一个系统性问题，可能源于多语言能力的不足或多模态能力的欠缺。为此，我们提出了一个两阶段的多语言幻觉去除（MHR）框架，旨在同时提升高资源语言和低资源语言的抗幻觉能力。我们不依赖繁琐的多语言资源人工标注，而是充分利用LVLM的固有能力，提出了一种新颖的跨语言对齐方法：为每个图像-查询输入生成多个响应，然后为每种语言识别幻觉感知数据对。这些数据对最终用于直接偏好优化，以促使LVLMs偏好非幻觉响应。实验结果表明，我们的MHR显著减少了LVLMs的幻觉生成。值得注意的是，在我们扩展的多语言POPE基准上，我们的框架在13种不同语言上的准确率平均提升了19.0%。我们的代码和模型权重可在https://github.com/ssmisya/MHR获取。

英文摘要：

While Large Vision-Language Models (LVLMs) have exhibited remarkable capabilities across a wide range of tasks, they suffer from hallucination problems, where models generate plausible yet incorrect answers given the input image-query pair. This hallucination phenomenon is even more severe when querying the image in non-English languages, while existing methods for mitigating hallucinations in LVLMs only consider the English scenarios. In this paper, we make the first attempt to mitigate this important multilingual hallucination in LVLMs. With thorough experiment analysis, we found that multilingual hallucination in LVLMs is a systemic problem that could arise from deficiencies in multilingual capabilities or inadequate multimodal abilities. To this end, we propose a two-stage Multilingual Hallucination Removal (MHR) framework for LVLMs, aiming to improve resistance to hallucination for both high-resource and low-resource languages. Instead of relying on the intricate manual annotations of multilingual resources, we fully leverage the inherent capabilities of the LVLM and propose a novel cross-lingual alignment method, which generates multiple responses for each image-query input and then identifies the hallucination-aware pairs for each language. These data pairs are finally used for direct preference optimization to prompt the LVLMs to favor non-hallucinating responses. Experimental results show that our MHR achieves a substantial reduction in hallucination generation for LVLMs. Notably, on our extended multilingual POPE benchmark, our framework delivers an average increase of 19.0% in accuracy across 13 different languages. Our code and model weights are available at https://github.com/ssmisya/MHR

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。