arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2507.04590cs.CVcs.CL
收藏

VLM2Vec-V2:推进面向视频、图像和视觉文档的多模态嵌入

VLM2Vec-V2: Advancing Multimodal Embedding for Videos, Images, and Visual Documents

Salesforce Research(Salesforce研究机构);UC Santa Barbara(加州大学圣巴巴拉分校);University of Waterloo(滑铁卢大学);Tsinghua University(清华大学)

Rui Meng, Ziyan Jiang, Ye Liu, Mingyi Su, Xinyi Yang, Yuepeng Fu, Can Qin, Zeyuan Chen, Ran Xu, Caiming Xiong, Yingbo Zhou, Wenhu Chen, Semih Yavuz

AI总结:

本文提出统一多模态嵌入框架 VLM2Vec-V2 及扩展基准 MMEB-V2,支持文本、图像、视频和视觉文档的统一嵌入学习,在视频和文档检索等新任务上取得强劲性能并超越原有图像基准。

摘要:

多模态嵌入模型在实现语义相似度、信息检索和跨模态聚类等各种下游任务中一直至关重要。然而,现有的多模态嵌入模型(如 VLM2Vec、E5-V、GME)主要聚焦于自然图像,对视频和视觉文档等其他视觉形式的支持有限。这限制了它们在现实场景中的适用性,包括 AI 智能体、多模态搜索与推荐以及检索增强生成(RAG)。为弥合这一差距,我们提出了 VLM2Vec-V2,一个用于学习跨多种视觉形式嵌入的统一框架。首先,我们引入了 MMEB-V2,这是一个综合性基准,在 MMEB 的基础上扩展了五种新的任务类型:视觉文档检索、视频检索、时间定位、视频分类和视频问答——涵盖文本、图像、视频和视觉文档输入。其次,我们训练了 VLM2Vec-V2,这是一个支持文本、图像、视频和视觉文档输入的通用嵌入模型。大量实验表明,VLM2Vec-V2 不仅在新引入的视频和文档检索任务上取得了强劲性能,而且在原始图像基准上相比先前基线也有所提升。通过广泛评估,我们的研究为各种多模态嵌入模型的泛化能力提供了见解,并突出了统一嵌入学习的有效策略,为研究和现实场景中更具可扩展性和适应性的表示学习奠定了基础。

英文摘要:

Multimodal embedding models have been crucial in enabling various downstream tasks such as semantic similarity, information retrieval, and clustering over different modalities. However, existing multimodal embeddings like VLM2Vec, E5-V, GME are predominantly focused on natural images, with limited support for other visual forms such as videos and visual documents. This restricts their applicability in real-world scenarios, including AI agents, multi-modal search and recommendation, and retrieval-augmented generation (RAG). To close this gap, we propose VLM2Vec-V2, a unified framework for learning embeddings across diverse visual forms. First, we introduce MMEB-V2, a comprehensive benchmark that extends MMEB with five new task types: visual document retrieval, video retrieval, temporal grounding, video classification and video question answering - spanning text, image, video, and visual document inputs. Next, we train VLM2Vec-V2, a general-purpose embedding model that supports text, image, video, and visual document inputs. Extensive experiments show that VLM2Vec-V2 achieves strong performance not only on the newly introduced video and document retrieval tasks, but also improves over prior baselines on the original image benchmarks. Through extensive evaluation, our study offers insights into the generalizability of various multimodal embedding models and highlights effective strategies for unified embedding learning, laying the groundwork for more scalable and adaptable representation learning in both research and real-world settings.

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。

↑