MiniGPT-5: Interleaved Vision-and-Language Generation via Generative Vokens

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zheng, Kaizhi, He, Xuehai, Wang, Xin Eric
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914189690273792
author Zheng, Kaizhi
He, Xuehai
Wang, Xin Eric
author_facet Zheng, Kaizhi
He, Xuehai
Wang, Xin Eric
contents The effectiveness of Multimodal Large Language Models (MLLMs) demonstrates a profound capability in multimodal understanding. However, the simultaneous generation of images with coherent texts is still underdeveloped. Addressing this, we introduce a novel interleaved vision-and-language generation method, centered around the concept of ``generative vokens". These vokens serve as pivotal elements contributing to coherent image-text outputs. Our method is marked by a unique two-stage training strategy for description-free multimodal generation, which does not necessitate extensive descriptions of images. We integrate classifier-free guidance to enhance the alignment of generated images and texts, ensuring more seamless and contextually relevant multimodal interactions. Our model, MiniGPT-5, exhibits substantial improvement over the baseline models on multimodal generation datasets, including MMDialog and VIST. The human evaluation shows MiniGPT-5 is better than the baseline model on more than 56\% cases for multimodal generation, highlighting its efficacy across diverse benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2310_02239
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle MiniGPT-5: Interleaved Vision-and-Language Generation via Generative Vokens
Zheng, Kaizhi
He, Xuehai
Wang, Xin Eric
Computer Vision and Pattern Recognition
Artificial Intelligence
The effectiveness of Multimodal Large Language Models (MLLMs) demonstrates a profound capability in multimodal understanding. However, the simultaneous generation of images with coherent texts is still underdeveloped. Addressing this, we introduce a novel interleaved vision-and-language generation method, centered around the concept of ``generative vokens". These vokens serve as pivotal elements contributing to coherent image-text outputs. Our method is marked by a unique two-stage training strategy for description-free multimodal generation, which does not necessitate extensive descriptions of images. We integrate classifier-free guidance to enhance the alignment of generated images and texts, ensuring more seamless and contextually relevant multimodal interactions. Our model, MiniGPT-5, exhibits substantial improvement over the baseline models on multimodal generation datasets, including MMDialog and VIST. The human evaluation shows MiniGPT-5 is better than the baseline model on more than 56\% cases for multimodal generation, highlighting its efficacy across diverse benchmarks.
title MiniGPT-5: Interleaved Vision-and-Language Generation via Generative Vokens
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2310.02239