@article{xue-etal-2022-byt5,
title = "{B}y{T}5: Towards a Token-Free Future with Pre-trained Byte-to-Byte Models",
author = "Xue, Linting and
Barua, Aditya and
Constant, Noah and
Al-Rfou, Rami and
Narang, Sharan and
Kale, Mihir and
Roberts, Adam and
Raffel, Colin",
editor = "Roark, Brian and
Nenkova, Ani",
journal = "Transactions of the Association for Computational Linguistics",
volume = "10",
year = "2022",
address = "Cambridge, MA",
publisher = "MIT Press",
url = "https://aclanthology.org/2022.tacl-1.17/",
doi = "10.1162/tacl_a_00461",
pages = "291--306",
abstract = "Most widely used pre-trained language models operate on sequences of tokens corresponding to word or subword units. By comparison, token-free models that operate directly on raw text (bytes or characters) have many benefits: They can process text in any language out of the box, they are more robust to noise, and they minimize technical debt by removing complex and error-prone text preprocessing pipelines. Because byte or character sequences are longer than token sequences, past work on token-free models has often introduced new model architectures designed to amortize the cost of operating directly on raw text. In this paper, we show that a standard Transformer architecture can be used with minimal modifications to process byte sequences. We characterize the trade-offs in terms of parameter count, training FLOPs, and inference speed, and show that byte-level models are competitive with their token-level counterparts. We also demonstrate that byte-level models are significantly more robust to noise and perform better on tasks that are sensitive to spelling and pronunciation. As part of our contribution, we release a new set of pre-trained byte-level Transformer models based on the T5 architecture, as well as all code and data used in our experiments.1"
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="xue-etal-2022-byt5">
<titleInfo>
<title>ByT5: Towards a Token-Free Future with Pre-trained Byte-to-Byte Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Linting</namePart>
<namePart type="family">Xue</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aditya</namePart>
<namePart type="family">Barua</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Noah</namePart>
<namePart type="family">Constant</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rami</namePart>
<namePart type="family">Al-Rfou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sharan</namePart>
<namePart type="family">Narang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mihir</namePart>
<namePart type="family">Kale</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Adam</namePart>
<namePart type="family">Roberts</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Colin</namePart>
<namePart type="family">Raffel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2022</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<genre authority="bibutilsgt">journal article</genre>
<relatedItem type="host">
<titleInfo>
<title>Transactions of the Association for Computational Linguistics</title>
</titleInfo>
<originInfo>
<issuance>continuing</issuance>
<publisher>MIT Press</publisher>
<place>
<placeTerm type="text">Cambridge, MA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">periodical</genre>
<genre authority="bibutilsgt">academic journal</genre>
</relatedItem>
<abstract>Most widely used pre-trained language models operate on sequences of tokens corresponding to word or subword units. By comparison, token-free models that operate directly on raw text (bytes or characters) have many benefits: They can process text in any language out of the box, they are more robust to noise, and they minimize technical debt by removing complex and error-prone text preprocessing pipelines. Because byte or character sequences are longer than token sequences, past work on token-free models has often introduced new model architectures designed to amortize the cost of operating directly on raw text. In this paper, we show that a standard Transformer architecture can be used with minimal modifications to process byte sequences. We characterize the trade-offs in terms of parameter count, training FLOPs, and inference speed, and show that byte-level models are competitive with their token-level counterparts. We also demonstrate that byte-level models are significantly more robust to noise and perform better on tasks that are sensitive to spelling and pronunciation. As part of our contribution, we release a new set of pre-trained byte-level Transformer models based on the T5 architecture, as well as all code and data used in our experiments.1</abstract>
<identifier type="citekey">xue-etal-2022-byt5</identifier>
<identifier type="doi">10.1162/tacl_a_00461</identifier>
<location>
<url>https://aclanthology.org/2022.tacl-1.17/</url>
</location>
<part>
<date>2022</date>
<detail type="volume"><number>10</number></detail>
<extent unit="page">
<start>291</start>
<end>306</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Journal Article
%T ByT5: Towards a Token-Free Future with Pre-trained Byte-to-Byte Models
%A Xue, Linting
%A Barua, Aditya
%A Constant, Noah
%A Al-Rfou, Rami
%A Narang, Sharan
%A Kale, Mihir
%A Roberts, Adam
%A Raffel, Colin
%J Transactions of the Association for Computational Linguistics
%D 2022
%V 10
%I MIT Press
%C Cambridge, MA
%F xue-etal-2022-byt5
%X Most widely used pre-trained language models operate on sequences of tokens corresponding to word or subword units. By comparison, token-free models that operate directly on raw text (bytes or characters) have many benefits: They can process text in any language out of the box, they are more robust to noise, and they minimize technical debt by removing complex and error-prone text preprocessing pipelines. Because byte or character sequences are longer than token sequences, past work on token-free models has often introduced new model architectures designed to amortize the cost of operating directly on raw text. In this paper, we show that a standard Transformer architecture can be used with minimal modifications to process byte sequences. We characterize the trade-offs in terms of parameter count, training FLOPs, and inference speed, and show that byte-level models are competitive with their token-level counterparts. We also demonstrate that byte-level models are significantly more robust to noise and perform better on tasks that are sensitive to spelling and pronunciation. As part of our contribution, we release a new set of pre-trained byte-level Transformer models based on the T5 architecture, as well as all code and data used in our experiments.1
%R 10.1162/tacl_a_00461
%U https://aclanthology.org/2022.tacl-1.17/
%U https://doi.org/10.1162/tacl_a_00461
%P 291-306
Markdown (Informal)
[ByT5: Towards a Token-Free Future with Pre-trained Byte-to-Byte Models](https://aclanthology.org/2022.tacl-1.17/) (Xue et al., TACL 2022)
ACL