Saved in:
Bibliographic Details
Main Authors: Ma, Feipeng, Xue, Hongwei, Wang, Guangting, Zhou, Yizhou, Rao, Fengyun, Yan, Shilin, Zhang, Yueyi, Wu, Siying, Shou, Mike Zheng, Sun, Xiaoyan
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2405.19333
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929364490256384
author Ma, Feipeng
Xue, Hongwei
Wang, Guangting
Zhou, Yizhou
Rao, Fengyun
Yan, Shilin
Zhang, Yueyi
Wu, Siying
Shou, Mike Zheng
Sun, Xiaoyan
author_facet Ma, Feipeng
Xue, Hongwei
Wang, Guangting
Zhou, Yizhou
Rao, Fengyun
Yan, Shilin
Zhang, Yueyi
Wu, Siying
Shou, Mike Zheng
Sun, Xiaoyan
contents Most multi-modal tasks can be formulated into problems of either generation or embedding. Existing models usually tackle these two types of problems by decoupling language modules into a text decoder for generation, and a text encoder for embedding. To explore the minimalism of multi-modal paradigms, we attempt to achieve only one model per modality in this work. We propose a Multi-Modal Generative Embedding Model (MM-GEM), whereby the generative and embedding objectives are encapsulated in one Large Language Model. We also propose a PoolAggregator to boost efficiency and enable the ability of fine-grained embedding and generation. A surprising finding is that these two objectives do not significantly conflict with each other. For example, MM-GEM instantiated from ViT-Large and TinyLlama shows competitive performance on benchmarks for multimodal embedding models such as cross-modal retrieval and zero-shot classification, while has good ability of image captioning. Additionally, MM-GEM can seamlessly execute region-level image caption generation and retrieval tasks. Besides, the advanced text model in MM-GEM brings over 5% improvement in Recall@1 for long text and image retrieval.
format Preprint
id arxiv_https___arxiv_org_abs_2405_19333
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Multi-Modal Generative Embedding Model
Ma, Feipeng
Xue, Hongwei
Wang, Guangting
Zhou, Yizhou
Rao, Fengyun
Yan, Shilin
Zhang, Yueyi
Wu, Siying
Shou, Mike Zheng
Sun, Xiaoyan
Computer Vision and Pattern Recognition
Most multi-modal tasks can be formulated into problems of either generation or embedding. Existing models usually tackle these two types of problems by decoupling language modules into a text decoder for generation, and a text encoder for embedding. To explore the minimalism of multi-modal paradigms, we attempt to achieve only one model per modality in this work. We propose a Multi-Modal Generative Embedding Model (MM-GEM), whereby the generative and embedding objectives are encapsulated in one Large Language Model. We also propose a PoolAggregator to boost efficiency and enable the ability of fine-grained embedding and generation. A surprising finding is that these two objectives do not significantly conflict with each other. For example, MM-GEM instantiated from ViT-Large and TinyLlama shows competitive performance on benchmarks for multimodal embedding models such as cross-modal retrieval and zero-shot classification, while has good ability of image captioning. Additionally, MM-GEM can seamlessly execute region-level image caption generation and retrieval tasks. Besides, the advanced text model in MM-GEM brings over 5% improvement in Recall@1 for long text and image retrieval.
title Multi-Modal Generative Embedding Model
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2405.19333