MetaMorph: Multimodal Understanding and Generation via Instruction Tuning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Tong, Shengbang, Fan, David, Zhu, Jiachen, Xiong, Yunyang, Chen, Xinlei, Sinha, Koustuv, Rabbat, Michael, LeCun, Yann, Xie, Saining, Liu, Zhuang
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909433364217856
author Tong, Shengbang
Fan, David
Zhu, Jiachen
Xiong, Yunyang
Chen, Xinlei
Sinha, Koustuv
Rabbat, Michael
LeCun, Yann
Xie, Saining
Liu, Zhuang
author_facet Tong, Shengbang
Fan, David
Zhu, Jiachen
Xiong, Yunyang
Chen, Xinlei
Sinha, Koustuv
Rabbat, Michael
LeCun, Yann
Xie, Saining
Liu, Zhuang
contents In this work, we propose Visual-Predictive Instruction Tuning (VPiT) - a simple and effective extension to visual instruction tuning that enables a pretrained LLM to quickly morph into an unified autoregressive model capable of generating both text and visual tokens. VPiT teaches an LLM to predict discrete text tokens and continuous visual tokens from any input sequence of image and text data curated in an instruction-following format. Our empirical investigation reveals several intriguing properties of VPiT: (1) visual generation ability emerges as a natural byproduct of improved visual understanding, and can be unlocked efficiently with a small amount of generation data; (2) while we find understanding and generation to be mutually beneficial, understanding data contributes to both capabilities more effectively than generation data. Building upon these findings, we train our MetaMorph model and achieve competitive performance on both visual understanding and generation. In visual generation, MetaMorph can leverage the world knowledge and reasoning abilities gained from LLM pretraining, and overcome common failure modes exhibited by other generation models. Our results suggest that LLMs may have strong "prior" vision capabilities that can be efficiently adapted to both visual understanding and generation with a relatively simple instruction tuning process.
format Preprint
id arxiv_https___arxiv_org_abs_2412_14164
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle MetaMorph: Multimodal Understanding and Generation via Instruction Tuning
Tong, Shengbang
Fan, David
Zhu, Jiachen
Xiong, Yunyang
Chen, Xinlei
Sinha, Koustuv
Rabbat, Michael
LeCun, Yann
Xie, Saining
Liu, Zhuang
Computer Vision and Pattern Recognition
In this work, we propose Visual-Predictive Instruction Tuning (VPiT) - a simple and effective extension to visual instruction tuning that enables a pretrained LLM to quickly morph into an unified autoregressive model capable of generating both text and visual tokens. VPiT teaches an LLM to predict discrete text tokens and continuous visual tokens from any input sequence of image and text data curated in an instruction-following format. Our empirical investigation reveals several intriguing properties of VPiT: (1) visual generation ability emerges as a natural byproduct of improved visual understanding, and can be unlocked efficiently with a small amount of generation data; (2) while we find understanding and generation to be mutually beneficial, understanding data contributes to both capabilities more effectively than generation data. Building upon these findings, we train our MetaMorph model and achieve competitive performance on both visual understanding and generation. In visual generation, MetaMorph can leverage the world knowledge and reasoning abilities gained from LLM pretraining, and overcome common failure modes exhibited by other generation models. Our results suggest that LLMs may have strong "prior" vision capabilities that can be efficiently adapted to both visual understanding and generation with a relatively simple instruction tuning process.
title MetaMorph: Multimodal Understanding and Generation via Instruction Tuning
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2412.14164