@inproceedings{chang-etal-2024-scaling,
title = "Scaling Parameter-Constrained Language Models with Quality Data",
author = "Chang, Ernie and
Paltenghi, Matteo and
Li, Yang and
Lin, Pin-Jie and
Zhao, Changsheng and
Huber, Patrick and
Liu, Zechun and
Rabatin, Rastislav and
Shi, Yangyang and
Chandra, Vikas",
editor = "Dernoncourt, Franck and
Preo{\c{t}}iuc-Pietro, Daniel and
Shimorina, Anastasia",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track",
month = nov,
year = "2024",
address = "Miami, Florida, US",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-industry.8/",
doi = "10.18653/v1/2024.emnlp-industry.8",
pages = "80--97",
abstract = "Scaling laws in language modeling traditionally quantify training loss as a function of dataset size and model parameters, providing compute-optimal estimates but often neglecting the impact of data quality on model generalization.In this paper, we extend the conventional understanding of scaling law by offering a microscopic view of data quality within the original formulation {--} \textit{effective training tokens} {--} which we posit to be a critical determinant of performance for parameter-constrained language models.Specifically, we formulate the proposed term of effective training tokens to be a combination of two readily-computed indicators of text:(i) text diversity and (ii) syntheticity as measured by a teacher model.We pretrained over 200 models of 25M to 1.5B parameters on a diverse set of sampled, synthetic data, and estimated the constants that relate text quality, model size, training tokens, and eight reasoning task accuracy scores.We demonstrated the estimated constants yield +0.83 Pearson correlation with true accuracies, and analyze it in scenarios involving widely-used data techniques such as data sampling and synthesis which aim to improve data quality."
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chang-etal-2024-scaling">
<titleInfo>
<title>Scaling Parameter-Constrained Language Models with Quality Data</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ernie</namePart>
<namePart type="family">Chang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Matteo</namePart>
<namePart type="family">Paltenghi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yang</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pin-Jie</namePart>
<namePart type="family">Lin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Changsheng</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Patrick</namePart>
<namePart type="family">Huber</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zechun</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rastislav</namePart>
<namePart type="family">Rabatin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yangyang</namePart>
<namePart type="family">Shi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vikas</namePart>
<namePart type="family">Chandra</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: Industry Track</title>
</titleInfo>
<name type="personal">
<namePart type="given">Franck</namePart>
<namePart type="family">Dernoncourt</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Daniel</namePart>
<namePart type="family">Preoţiuc-Pietro</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anastasia</namePart>
<namePart type="family">Shimorina</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Miami, Florida, US</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Scaling laws in language modeling traditionally quantify training loss as a function of dataset size and model parameters, providing compute-optimal estimates but often neglecting the impact of data quality on model generalization.In this paper, we extend the conventional understanding of scaling law by offering a microscopic view of data quality within the original formulation – effective training tokens – which we posit to be a critical determinant of performance for parameter-constrained language models.Specifically, we formulate the proposed term of effective training tokens to be a combination of two readily-computed indicators of text:(i) text diversity and (ii) syntheticity as measured by a teacher model.We pretrained over 200 models of 25M to 1.5B parameters on a diverse set of sampled, synthetic data, and estimated the constants that relate text quality, model size, training tokens, and eight reasoning task accuracy scores.We demonstrated the estimated constants yield +0.83 Pearson correlation with true accuracies, and analyze it in scenarios involving widely-used data techniques such as data sampling and synthesis which aim to improve data quality.</abstract>
<identifier type="citekey">chang-etal-2024-scaling</identifier>
<identifier type="doi">10.18653/v1/2024.emnlp-industry.8</identifier>
<location>
<url>https://aclanthology.org/2024.emnlp-industry.8/</url>
</location>
<part>
<date>2024-11</date>
<extent unit="page">
<start>80</start>
<end>97</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Scaling Parameter-Constrained Language Models with Quality Data
%A Chang, Ernie
%A Paltenghi, Matteo
%A Li, Yang
%A Lin, Pin-Jie
%A Zhao, Changsheng
%A Huber, Patrick
%A Liu, Zechun
%A Rabatin, Rastislav
%A Shi, Yangyang
%A Chandra, Vikas
%Y Dernoncourt, Franck
%Y Preoţiuc-Pietro, Daniel
%Y Shimorina, Anastasia
%S Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track
%D 2024
%8 November
%I Association for Computational Linguistics
%C Miami, Florida, US
%F chang-etal-2024-scaling
%X Scaling laws in language modeling traditionally quantify training loss as a function of dataset size and model parameters, providing compute-optimal estimates but often neglecting the impact of data quality on model generalization.In this paper, we extend the conventional understanding of scaling law by offering a microscopic view of data quality within the original formulation – effective training tokens – which we posit to be a critical determinant of performance for parameter-constrained language models.Specifically, we formulate the proposed term of effective training tokens to be a combination of two readily-computed indicators of text:(i) text diversity and (ii) syntheticity as measured by a teacher model.We pretrained over 200 models of 25M to 1.5B parameters on a diverse set of sampled, synthetic data, and estimated the constants that relate text quality, model size, training tokens, and eight reasoning task accuracy scores.We demonstrated the estimated constants yield +0.83 Pearson correlation with true accuracies, and analyze it in scenarios involving widely-used data techniques such as data sampling and synthesis which aim to improve data quality.
%R 10.18653/v1/2024.emnlp-industry.8
%U https://aclanthology.org/2024.emnlp-industry.8/
%U https://doi.org/10.18653/v1/2024.emnlp-industry.8
%P 80-97
Markdown (Informal)
[Scaling Parameter-Constrained Language Models with Quality Data](https://aclanthology.org/2024.emnlp-industry.8/) (Chang et al., EMNLP 2024)
ACL
- Ernie Chang, Matteo Paltenghi, Yang Li, Pin-Jie Lin, Changsheng Zhao, Patrick Huber, Zechun Liu, Rastislav Rabatin, Yangyang Shi, and Vikas Chandra. 2024. Scaling Parameter-Constrained Language Models with Quality Data. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 80–97, Miami, Florida, US. Association for Computational Linguistics.