Single layer tiny Co4 outpaces GPT-2 and GPT-BERT

Noor Ul Zain; Mohsin Raza Naseem; Ahsan Adeel

Single layer tiny Co4 outpaces GPT-2 and GPT-BERT

Noor Ul Zain, Mohsin Raza Naseem, Ahsan Adeel

Correct Metadata for

Important: The Anthology treat PDFs as authoritative. Please use this form only to correct data that is out of line with the PDF. See our corrections guidelines if you need to change the PDF.

Title Adjust the title. Retain tags such as <fixed-case>.

Authors Adjust author names and order to match the PDF.

Abstract Correct abstract if needed. Retain XML formatting tags such as <tex-math>.

Verification against PDF Ensure that the new title/authors match the snapshot below. (If there is no snapshot or it is too small, consult the PDF.)

Authors concatenated from the text boxes above:

ALL author names match the snapshot above—including middle initials, hyphens, and accents.

Abstract

We show that a tiny Co⁴ machine (CITATION) with a single layer, two heads, and 8M parameters, operating at O(N) computational cost (where N is the number of input tokens), in just 2 epochs outpaces GPT-2 (124M, 12 layers, O(N²)) and GPT-BERT (30M, 12 layers, O(N²)), both trained for 10 epochs. Co⁴ achieves orders-of-magnitude greater training efficiency on 10M tokens, demonstrating sample-efficient pretraining. On the BabyLM challenge evaluation pipeline, Co⁴ performs comparably or better across complex benchmarks, showing strong zero-shot and fine-tuning performance on SuperGLUE tasks. Specifically, Co⁴ outperforms GPT-2 in 5 out of 7 zero-shot metrics and 6 out of 7 fine-tuning tasks, and GPT-BERT in 4 out of 7 metrics in both cases. These results strongly suggest a need to rethink prevailing deep learning paradigms and associated scaling laws.

Anthology ID:: 2025.babylm-main.24
Volume:: Proceedings of the First BabyLM Workshop
Month:: November
Year:: 2025
Address:: Suzhou, China
Editors:: Lucas Charpentier, Leshem Choshen, Ryan Cotterell, Mustafa Omer Gul, Michael Y. Hu, Jing Liu, Jaap Jumelet, Tal Linzen, Aaron Mueller, Candace Ross, Raj Sanjay Shah, Alex Warstadt, Ethan Gotlieb Wilcox, Adina Williams
Venue:: BabyLM
SIG:
Publisher:: Association for Computational Linguistics
Note:
Pages:: 313–322
Language:
URL:: https://aclanthology.org/2025.babylm-main.24/
DOI:
Bibkey:
Cite (ACL):: Noor Ul Zain, Mohsin Raza Naseem, and Ahsan Adeel. 2025. Single layer tiny Co4 outpaces GPT-2 and GPT-BERT. In Proceedings of the First BabyLM Workshop, pages 313–322, Suzhou, China. Association for Computational Linguistics.
Cite (Informal):: Single layer tiny Co4 outpaces GPT-2 and GPT-BERT (Zain et al., BabyLM 2025)
Copy Citation:
PDF:: https://aclanthology.org/2025.babylm-main.24.pdf

PDF Cite Search Fix data

Export citation

BibTeX
MODS XML
Endnote
Preformatted

@inproceedings{zain-etal-2025-single,
    title = "Single layer tiny Co4 outpaces {GPT}-2 and {GPT}-{BERT}",
    author = "Zain, Noor Ul  and
      Naseem, Mohsin Raza  and
      Adeel, Ahsan",
    editor = "Charpentier, Lucas  and
      Choshen, Leshem  and
      Cotterell, Ryan  and
      Gul, Mustafa Omer  and
      Hu, Michael Y.  and
      Liu, Jing  and
      Jumelet, Jaap  and
      Linzen, Tal  and
      Mueller, Aaron  and
      Ross, Candace  and
      Shah, Raj Sanjay  and
      Warstadt, Alex  and
      Wilcox, Ethan Gotlieb  and
      Williams, Adina",
    booktitle = "Proceedings of the First BabyLM Workshop",
    month = nov,
    year = "2025",
    address = "Suzhou, China",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.babylm-main.24/",
    pages = "313--322",
    ISBN = "TODO",
    abstract = "We show that a tiny $Co^{4}$ machine (CITATION) with a single layer, two heads, and 8M parameters, operating at $O(N)$ computational cost (where \textit{N} is the number of input tokens), in just 2 epochs outpaces GPT-2 (124M, 12 layers, $O(N^2)$) and GPT-BERT (30M, 12 layers, $O(N^2)$), both trained for 10 epochs. $Co^{4}$ achieves orders-of-magnitude greater training efficiency on 10M tokens, demonstrating sample-efficient pretraining. On the BabyLM challenge evaluation pipeline, $Co^{4}$ performs comparably or better across complex benchmarks, showing strong zero-shot and fine-tuning performance on SuperGLUE tasks. Specifically, $Co^{4}$ outperforms GPT-2 in 5 out of 7 zero-shot metrics and 6 out of 7 fine-tuning tasks, and GPT-BERT in 4 out of 7 metrics in both cases. These results strongly suggest a need to rethink prevailing deep learning paradigms and associated scaling laws."
}

Download as File

<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zain-etal-2025-single">
    <titleInfo>
        <title>Single layer tiny Co4 outpaces GPT-2 and GPT-BERT</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Noor</namePart>
        <namePart type="given">Ul</namePart>
        <namePart type="family">Zain</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Mohsin</namePart>
        <namePart type="given">Raza</namePart>
        <namePart type="family">Naseem</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Ahsan</namePart>
        <namePart type="family">Adeel</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2025-11</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the First BabyLM Workshop</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Lucas</namePart>
            <namePart type="family">Charpentier</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Leshem</namePart>
            <namePart type="family">Choshen</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Ryan</namePart>
            <namePart type="family">Cotterell</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Mustafa</namePart>
            <namePart type="given">Omer</namePart>
            <namePart type="family">Gul</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Michael</namePart>
            <namePart type="given">Y</namePart>
            <namePart type="family">Hu</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jing</namePart>
            <namePart type="family">Liu</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jaap</namePart>
            <namePart type="family">Jumelet</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Tal</namePart>
            <namePart type="family">Linzen</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Aaron</namePart>
            <namePart type="family">Mueller</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Candace</namePart>
            <namePart type="family">Ross</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Raj</namePart>
            <namePart type="given">Sanjay</namePart>
            <namePart type="family">Shah</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Alex</namePart>
            <namePart type="family">Warstadt</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Ethan</namePart>
            <namePart type="given">Gotlieb</namePart>
            <namePart type="family">Wilcox</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Adina</namePart>
            <namePart type="family">Williams</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Suzhou, China</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">TODO</identifier>
    </relatedItem>
    <abstract>We show that a tiny Co⁴ machine (CITATION) with a single layer, two heads, and 8M parameters, operating at O(N) computational cost (where N is the number of input tokens), in just 2 epochs outpaces GPT-2 (124M, 12 layers, O(N²)) and GPT-BERT (30M, 12 layers, O(N²)), both trained for 10 epochs. Co⁴ achieves orders-of-magnitude greater training efficiency on 10M tokens, demonstrating sample-efficient pretraining. On the BabyLM challenge evaluation pipeline, Co⁴ performs comparably or better across complex benchmarks, showing strong zero-shot and fine-tuning performance on SuperGLUE tasks. Specifically, Co⁴ outperforms GPT-2 in 5 out of 7 zero-shot metrics and 6 out of 7 fine-tuning tasks, and GPT-BERT in 4 out of 7 metrics in both cases. These results strongly suggest a need to rethink prevailing deep learning paradigms and associated scaling laws.</abstract>
    <identifier type="citekey">zain-etal-2025-single</identifier>
    <location>
        <url>https://aclanthology.org/2025.babylm-main.24/</url>
    </location>
    <part>
        <date>2025-11</date>
        <extent unit="page">
            <start>313</start>
            <end>322</end>
        </extent>
    </part>
</mods>
</modsCollection>

Download as File

%0 Conference Proceedings
%T Single layer tiny Co4 outpaces GPT-2 and GPT-BERT
%A Zain, Noor Ul
%A Naseem, Mohsin Raza
%A Adeel, Ahsan
%Y Charpentier, Lucas
%Y Choshen, Leshem
%Y Cotterell, Ryan
%Y Gul, Mustafa Omer
%Y Hu, Michael Y.
%Y Liu, Jing
%Y Jumelet, Jaap
%Y Linzen, Tal
%Y Mueller, Aaron
%Y Ross, Candace
%Y Shah, Raj Sanjay
%Y Warstadt, Alex
%Y Wilcox, Ethan Gotlieb
%Y Williams, Adina
%S Proceedings of the First BabyLM Workshop
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou, China
%@ TODO
%F zain-etal-2025-single
%X We show that a tiny Co⁴ machine (CITATION) with a single layer, two heads, and 8M parameters, operating at O(N) computational cost (where N is the number of input tokens), in just 2 epochs outpaces GPT-2 (124M, 12 layers, O(N²)) and GPT-BERT (30M, 12 layers, O(N²)), both trained for 10 epochs. Co⁴ achieves orders-of-magnitude greater training efficiency on 10M tokens, demonstrating sample-efficient pretraining. On the BabyLM challenge evaluation pipeline, Co⁴ performs comparably or better across complex benchmarks, showing strong zero-shot and fine-tuning performance on SuperGLUE tasks. Specifically, Co⁴ outperforms GPT-2 in 5 out of 7 zero-shot metrics and 6 out of 7 fine-tuning tasks, and GPT-BERT in 4 out of 7 metrics in both cases. These results strongly suggest a need to rethink prevailing deep learning paradigms and associated scaling laws.
%U https://aclanthology.org/2025.babylm-main.24/
%P 313-322

Download as File

Markdown (Informal)

[Single layer tiny Co4 outpaces GPT-2 and GPT-BERT](https://aclanthology.org/2025.babylm-main.24/) (Zain et al., BabyLM 2025)

Single layer tiny Co4 outpaces GPT-2 and GPT-BERT (Zain et al., BabyLM 2025)

ACL

Noor Ul Zain, Mohsin Raza Naseem, and Ahsan Adeel. 2025. Single layer tiny Co4 outpaces GPT-2 and GPT-BERT. In Proceedings of the First BabyLM Workshop, pages 313–322, Suzhou, China. Association for Computational Linguistics.