@inproceedings{mireshghallah-etal-2024-smaller,
title = "Smaller Language Models are Better Zero-shot Machine-Generated Text Detectors",
author = "Mireshghallah, Niloofar and
Mattern, Justus and
Gao, Sicun and
Shokri, Reza and
Berg-Kirkpatrick, Taylor",
editor = "Graham, Yvette and
Purver, Matthew",
booktitle = "Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers)",
month = mar,
year = "2024",
address = "St. Julian{'}s, Malta",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.eacl-short.25",
pages = "278--293",
abstract = "As large language models are becoming more embedded in different user-facing services, it is important to be able to distinguish between human-written and machine-generated text to verify the authenticity of news articles, product reviews, etc. Thus, in this paper we set out to explore whether it is possible to use one language model to identify machine-generated text produced by another language model, in a zero-shot way, even if the two have different architectures and are trained on different data. We find that overall, smaller models are better universal machine-generated text detectors: they can more precisely detect text generated from both small and larger models, without the need for any additional training/data. Interestingly, we find that whether or not the detector and generator models were trained on the same data is not critically important to the detection success. For instance the OPT-125M model has an AUC of 0.90 in detecting GPT4 generations, whereas a larger model from the GPT family, GPTJ-6B, has AUC of 0.65.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="mireshghallah-etal-2024-smaller">
<titleInfo>
<title>Smaller Language Models are Better Zero-shot Machine-Generated Text Detectors</title>
</titleInfo>
<name type="personal">
<namePart type="given">Niloofar</namePart>
<namePart type="family">Mireshghallah</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Justus</namePart>
<namePart type="family">Mattern</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sicun</namePart>
<namePart type="family">Gao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Reza</namePart>
<namePart type="family">Shokri</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Taylor</namePart>
<namePart type="family">Berg-Kirkpatrick</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-03</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yvette</namePart>
<namePart type="family">Graham</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Matthew</namePart>
<namePart type="family">Purver</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">St. Julian’s, Malta</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>As large language models are becoming more embedded in different user-facing services, it is important to be able to distinguish between human-written and machine-generated text to verify the authenticity of news articles, product reviews, etc. Thus, in this paper we set out to explore whether it is possible to use one language model to identify machine-generated text produced by another language model, in a zero-shot way, even if the two have different architectures and are trained on different data. We find that overall, smaller models are better universal machine-generated text detectors: they can more precisely detect text generated from both small and larger models, without the need for any additional training/data. Interestingly, we find that whether or not the detector and generator models were trained on the same data is not critically important to the detection success. For instance the OPT-125M model has an AUC of 0.90 in detecting GPT4 generations, whereas a larger model from the GPT family, GPTJ-6B, has AUC of 0.65.</abstract>
<identifier type="citekey">mireshghallah-etal-2024-smaller</identifier>
<location>
<url>https://aclanthology.org/2024.eacl-short.25</url>
</location>
<part>
<date>2024-03</date>
<extent unit="page">
<start>278</start>
<end>293</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Smaller Language Models are Better Zero-shot Machine-Generated Text Detectors
%A Mireshghallah, Niloofar
%A Mattern, Justus
%A Gao, Sicun
%A Shokri, Reza
%A Berg-Kirkpatrick, Taylor
%Y Graham, Yvette
%Y Purver, Matthew
%S Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers)
%D 2024
%8 March
%I Association for Computational Linguistics
%C St. Julian’s, Malta
%F mireshghallah-etal-2024-smaller
%X As large language models are becoming more embedded in different user-facing services, it is important to be able to distinguish between human-written and machine-generated text to verify the authenticity of news articles, product reviews, etc. Thus, in this paper we set out to explore whether it is possible to use one language model to identify machine-generated text produced by another language model, in a zero-shot way, even if the two have different architectures and are trained on different data. We find that overall, smaller models are better universal machine-generated text detectors: they can more precisely detect text generated from both small and larger models, without the need for any additional training/data. Interestingly, we find that whether or not the detector and generator models were trained on the same data is not critically important to the detection success. For instance the OPT-125M model has an AUC of 0.90 in detecting GPT4 generations, whereas a larger model from the GPT family, GPTJ-6B, has AUC of 0.65.
%U https://aclanthology.org/2024.eacl-short.25
%P 278-293
Markdown (Informal)
[Smaller Language Models are Better Zero-shot Machine-Generated Text Detectors](https://aclanthology.org/2024.eacl-short.25) (Mireshghallah et al., EACL 2024)
ACL
- Niloofar Mireshghallah, Justus Mattern, Sicun Gao, Reza Shokri, and Taylor Berg-Kirkpatrick. 2024. Smaller Language Models are Better Zero-shot Machine-Generated Text Detectors. In Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers), pages 278–293, St. Julian’s, Malta. Association for Computational Linguistics.