GME: Improving Universal Multimodal Retrieval by Multimodal LLMs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Xin, Zhang, Yanzhao, Xie, Wen, Li, Mingxin, Dai, Ziqi, Long, Dingkun, Xie, Pengjun, Zhang, Meishan, Li, Wenjie, Zhang, Min
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915220280049664
author Zhang, Xin
Zhang, Yanzhao
Xie, Wen
Li, Mingxin
Dai, Ziqi
Long, Dingkun
Xie, Pengjun
Zhang, Meishan
Li, Wenjie
Zhang, Min
author_facet Zhang, Xin
Zhang, Yanzhao
Xie, Wen
Li, Mingxin
Dai, Ziqi
Long, Dingkun
Xie, Pengjun
Zhang, Meishan
Li, Wenjie
Zhang, Min
contents Universal Multimodal Retrieval (UMR) aims to enable search across various modalities using a unified model, where queries and candidates can consist of pure text, images, or a combination of both. Previous work has attempted to adopt multimodal large language models (MLLMs) to realize UMR using only text data. However, our preliminary experiments demonstrate that more diverse multimodal training data can further unlock the potential of MLLMs. Despite its effectiveness, the existing multimodal training data is highly imbalanced in terms of modality, which motivates us to develop a training data synthesis pipeline and construct a large-scale, high-quality fused-modal training dataset. Based on the synthetic training data, we develop the General Multimodal Embedder (GME), an MLLM-based dense retriever designed for UMR. Furthermore, we construct a comprehensive UMR Benchmark (UMRB) to evaluate the effectiveness of our approach. Experimental results show that our method achieves state-of-the-art performance among existing UMR methods. Last, we provide in-depth analyses of model scaling and training strategies, and perform ablation studies on both the model and synthetic data.
format Preprint
id arxiv_https___arxiv_org_abs_2412_16855
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle GME: Improving Universal Multimodal Retrieval by Multimodal LLMs
Zhang, Xin
Zhang, Yanzhao
Xie, Wen
Li, Mingxin
Dai, Ziqi
Long, Dingkun
Xie, Pengjun
Zhang, Meishan
Li, Wenjie
Zhang, Min
Computation and Language
Information Retrieval
Universal Multimodal Retrieval (UMR) aims to enable search across various modalities using a unified model, where queries and candidates can consist of pure text, images, or a combination of both. Previous work has attempted to adopt multimodal large language models (MLLMs) to realize UMR using only text data. However, our preliminary experiments demonstrate that more diverse multimodal training data can further unlock the potential of MLLMs. Despite its effectiveness, the existing multimodal training data is highly imbalanced in terms of modality, which motivates us to develop a training data synthesis pipeline and construct a large-scale, high-quality fused-modal training dataset. Based on the synthetic training data, we develop the General Multimodal Embedder (GME), an MLLM-based dense retriever designed for UMR. Furthermore, we construct a comprehensive UMR Benchmark (UMRB) to evaluate the effectiveness of our approach. Experimental results show that our method achieves state-of-the-art performance among existing UMR methods. Last, we provide in-depth analyses of model scaling and training strategies, and perform ablation studies on both the model and synthetic data.
title GME: Improving Universal Multimodal Retrieval by Multimodal LLMs
topic Computation and Language
Information Retrieval
url https://arxiv.org/abs/2412.16855