@inproceedings{deshpande-etal-2023-honey,
title = "Honey, {I} Shrunk the Language: Language Model Behavior at Reduced Scale.",
author = "Deshpande, Vijeta and
Pechi, Dan and
Thatte, Shree and
Lialin, Vladislav and
Rumshisky, Anna",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2023",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.findings-acl.326",
doi = "10.18653/v1/2023.findings-acl.326",
pages = "5298--5314",
abstract = "In recent years, language models have drastically grown in size, and the abilities of these models have been shown to improve with scale. The majority of recent scaling laws studies focused on high-compute high-parameter count settings, leaving the question of when these abilities begin to emerge largely unanswered. In this paper, we investigate whether the effects of pre-training can be observed when the problem size is reduced, modeling a smaller, reduced-vocabulary language. We show the benefits of pre-training with masked language modeling (MLM) objective in models as small as 1.25M parameters, and establish a strong correlation between pre-training perplexity and downstream performance (GLUE benchmark). We examine downscaling effects, extending scaling laws to models as small as {\textasciitilde}1M parameters. At this scale, we observe a break of the power law for compute-optimal models and show that the MLM loss does not scale smoothly with compute-cost (FLOPs) below $2.2 \times 10^{15}$ FLOPs. We also find that adding layers does not always benefit downstream performance.Our filtered pre-training data, reduced English vocabulary, and code are available at \url{https://github.com/text-machine-lab/mini_bert}$github.com/text-machine-lab/mini\_bert$",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="deshpande-etal-2023-honey">
<titleInfo>
<title>Honey, I Shrunk the Language: Language Model Behavior at Reduced Scale.</title>
</titleInfo>
<name type="personal">
<namePart type="given">Vijeta</namePart>
<namePart type="family">Deshpande</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dan</namePart>
<namePart type="family">Pechi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shree</namePart>
<namePart type="family">Thatte</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vladislav</namePart>
<namePart type="family">Lialin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anna</namePart>
<namePart type="family">Rumshisky</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2023</title>
</titleInfo>
<name type="personal">
<namePart type="given">Anna</namePart>
<namePart type="family">Rogers</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jordan</namePart>
<namePart type="family">Boyd-Graber</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Naoaki</namePart>
<namePart type="family">Okazaki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Toronto, Canada</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>In recent years, language models have drastically grown in size, and the abilities of these models have been shown to improve with scale. The majority of recent scaling laws studies focused on high-compute high-parameter count settings, leaving the question of when these abilities begin to emerge largely unanswered. In this paper, we investigate whether the effects of pre-training can be observed when the problem size is reduced, modeling a smaller, reduced-vocabulary language. We show the benefits of pre-training with masked language modeling (MLM) objective in models as small as 1.25M parameters, and establish a strong correlation between pre-training perplexity and downstream performance (GLUE benchmark). We examine downscaling effects, extending scaling laws to models as small as ~1M parameters. At this scale, we observe a break of the power law for compute-optimal models and show that the MLM loss does not scale smoothly with compute-cost (FLOPs) below 2.2 \times 10¹5 FLOPs. We also find that adding layers does not always benefit downstream performance.Our filtered pre-training data, reduced English vocabulary, and code are available at https://github.com/text-machine-lab/mini_bertgithub.com/text-machine-lab/mini_bert</abstract>
<identifier type="citekey">deshpande-etal-2023-honey</identifier>
<identifier type="doi">10.18653/v1/2023.findings-acl.326</identifier>
<location>
<url>https://aclanthology.org/2023.findings-acl.326</url>
</location>
<part>
<date>2023-07</date>
<extent unit="page">
<start>5298</start>
<end>5314</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Honey, I Shrunk the Language: Language Model Behavior at Reduced Scale.
%A Deshpande, Vijeta
%A Pechi, Dan
%A Thatte, Shree
%A Lialin, Vladislav
%A Rumshisky, Anna
%Y Rogers, Anna
%Y Boyd-Graber, Jordan
%Y Okazaki, Naoaki
%S Findings of the Association for Computational Linguistics: ACL 2023
%D 2023
%8 July
%I Association for Computational Linguistics
%C Toronto, Canada
%F deshpande-etal-2023-honey
%X In recent years, language models have drastically grown in size, and the abilities of these models have been shown to improve with scale. The majority of recent scaling laws studies focused on high-compute high-parameter count settings, leaving the question of when these abilities begin to emerge largely unanswered. In this paper, we investigate whether the effects of pre-training can be observed when the problem size is reduced, modeling a smaller, reduced-vocabulary language. We show the benefits of pre-training with masked language modeling (MLM) objective in models as small as 1.25M parameters, and establish a strong correlation between pre-training perplexity and downstream performance (GLUE benchmark). We examine downscaling effects, extending scaling laws to models as small as ~1M parameters. At this scale, we observe a break of the power law for compute-optimal models and show that the MLM loss does not scale smoothly with compute-cost (FLOPs) below 2.2 \times 10¹5 FLOPs. We also find that adding layers does not always benefit downstream performance.Our filtered pre-training data, reduced English vocabulary, and code are available at https://github.com/text-machine-lab/mini_bertgithub.com/text-machine-lab/mini_bert
%R 10.18653/v1/2023.findings-acl.326
%U https://aclanthology.org/2023.findings-acl.326
%U https://doi.org/10.18653/v1/2023.findings-acl.326
%P 5298-5314
Markdown (Informal)
[Honey, I Shrunk the Language: Language Model Behavior at Reduced Scale.](https://aclanthology.org/2023.findings-acl.326) (Deshpande et al., Findings 2023)
ACL