@inproceedings{chai-etal-2024-autoregressive,
title = "Autoregressive Pre-Training on Pixels and Texts",
author = "Chai, Yekun and
Liu, Qingyi and
Xiao, Jingwu and
Wang, Shuohuan and
Sun, Yu and
Wu, Hua",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.182",
pages = "3106--3125",
abstract = "The integration of visual and textual information represents a promising direction in the advancement of language models. In this paper, we explore the dual modality of language{---}both visual and textual{---}within an autoregressive framework, pre-trained on both document images and texts. Our method employs a multimodal training strategy, utilizing visual data through next patch prediction with a regression head and/or textual data through next token prediction with a classification head. We focus on understanding the interaction between these two modalities and their combined impact on model performance. Our extensive evaluation across a wide range of benchmarks shows that incorporating both visual and textual data significantly improves the performance of pixel-based language models. Remarkably, we find that a unidirectional pixel-based model trained solely on visual data can achieve comparable results to state-of-the-art bidirectional models on several language understanding tasks. This work uncovers the untapped potential of integrating visual and textual modalities for more effective language modeling. We release our code, data, and model checkpoints at https://github.com/ernie-research/pixelgpt.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chai-etal-2024-autoregressive">
<titleInfo>
<title>Autoregressive Pre-Training on Pixels and Texts</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yekun</namePart>
<namePart type="family">Chai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qingyi</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jingwu</namePart>
<namePart type="family">Xiao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shuohuan</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yu</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hua</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yaser</namePart>
<namePart type="family">Al-Onaizan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohit</namePart>
<namePart type="family">Bansal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yun-Nung</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Miami, Florida, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>The integration of visual and textual information represents a promising direction in the advancement of language models. In this paper, we explore the dual modality of language—both visual and textual—within an autoregressive framework, pre-trained on both document images and texts. Our method employs a multimodal training strategy, utilizing visual data through next patch prediction with a regression head and/or textual data through next token prediction with a classification head. We focus on understanding the interaction between these two modalities and their combined impact on model performance. Our extensive evaluation across a wide range of benchmarks shows that incorporating both visual and textual data significantly improves the performance of pixel-based language models. Remarkably, we find that a unidirectional pixel-based model trained solely on visual data can achieve comparable results to state-of-the-art bidirectional models on several language understanding tasks. This work uncovers the untapped potential of integrating visual and textual modalities for more effective language modeling. We release our code, data, and model checkpoints at https://github.com/ernie-research/pixelgpt.</abstract>
<identifier type="citekey">chai-etal-2024-autoregressive</identifier>
<location>
<url>https://aclanthology.org/2024.emnlp-main.182</url>
</location>
<part>
<date>2024-11</date>
<extent unit="page">
<start>3106</start>
<end>3125</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Autoregressive Pre-Training on Pixels and Texts
%A Chai, Yekun
%A Liu, Qingyi
%A Xiao, Jingwu
%A Wang, Shuohuan
%A Sun, Yu
%A Wu, Hua
%Y Al-Onaizan, Yaser
%Y Bansal, Mohit
%Y Chen, Yun-Nung
%S Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing
%D 2024
%8 November
%I Association for Computational Linguistics
%C Miami, Florida, USA
%F chai-etal-2024-autoregressive
%X The integration of visual and textual information represents a promising direction in the advancement of language models. In this paper, we explore the dual modality of language—both visual and textual—within an autoregressive framework, pre-trained on both document images and texts. Our method employs a multimodal training strategy, utilizing visual data through next patch prediction with a regression head and/or textual data through next token prediction with a classification head. We focus on understanding the interaction between these two modalities and their combined impact on model performance. Our extensive evaluation across a wide range of benchmarks shows that incorporating both visual and textual data significantly improves the performance of pixel-based language models. Remarkably, we find that a unidirectional pixel-based model trained solely on visual data can achieve comparable results to state-of-the-art bidirectional models on several language understanding tasks. This work uncovers the untapped potential of integrating visual and textual modalities for more effective language modeling. We release our code, data, and model checkpoints at https://github.com/ernie-research/pixelgpt.
%U https://aclanthology.org/2024.emnlp-main.182
%P 3106-3125
Markdown (Informal)
[Autoregressive Pre-Training on Pixels and Texts](https://aclanthology.org/2024.emnlp-main.182) (Chai et al., EMNLP 2024)
ACL
- Yekun Chai, Qingyi Liu, Jingwu Xiao, Shuohuan Wang, Yu Sun, and Hua Wu. 2024. Autoregressive Pre-Training on Pixels and Texts. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 3106–3125, Miami, Florida, USA. Association for Computational Linguistics.