HalleluBERT: Let Every Token That Has Meaning Bear Its Weight

Raphael Schmitt

doi:10.63317/3qdqexx4e9i2

HalleluBERT: Let Every Token That Has Meaning Bear Its Weight

Correct Metadata for

Use this form to create a GitHub issue with structured data describing the correction. You will need a GitHub account. Once you create that issue, the correction will be reviewed by a staff member.

⚠️ Mobile Users: Submitting this form to create a new issue will only work with github.com, not the GitHub Mobile app.

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>. You may use ... for bold, ... for italic, ... for underline, <sc>...</sc> for small-caps, <tt>...<tt> for typewriter text, <url>...</url> for URLs, <a href=...> for hyperlinks, and <par/> for paragraph breaks.

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 external publication .)

Authors concatenated from the text boxes above:

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

Abstract

Transformer-based models have advanced NLP, yet Hebrew still lacks a RoBERTa encoder that is trained at scale and released in both base and large variants. We present HalleluBERT, a RoBERTa-based encoder family trained from scratch on 49.1 GB of deduplicated Hebrew web text and Wikipedia using a Hebrew-specific byte-level BPE vocabulary. On native Hebrew benchmarks for named entity recognition (BMC, NEMO) and sentiment classification (SMCD), HalleluBERT outperforms monolingual and multilingual baselines, and yields the highest unweighted mean score across the three benchmarks. We release model weights and tokenizer under the MIT license to support reproducible Hebrew NLP research.

Anthology ID:: 2026.lrec-1.236
Volume:: Proceedings of the Fifteenth Language Resources and Evaluation Conference
Month:: May
Year:: 2026
Address:: Palma de Mallorca, Spain
Editors:: Stelios Piperidis, Núria Bel, Henk van den Heuvel, Nancy Ide, Simon Krek, Antonio Toral
Venue:: LREC
SIG:
Publisher:: ELRA Language Resource Association
Note:
Pages:: 3022–3030
Language:
External URL:: https://lrec.elra.info/lrec2026-main-236
DOI:: 10.63317/3qdqexx4e9i2
Bibkey:
Cite (ACL):: Raphael Schmitt. 2026. HalleluBERT: Let Every Token That Has Meaning Bear Its Weight. In Proceedings of the Fifteenth Language Resources and Evaluation Conference, pages 3022–3030, Palma de Mallorca, Spain. ELRA Language Resource Association.
Cite (Informal):: HalleluBERT: Let Every Token That Has Meaning Bear Its Weight (Schmitt, LREC 2026)
Copy Citation:

External Cite Search Fix data

Export citation

BibTeX
MODS XML
Endnote
Preformatted

@inproceedings{schmitt-2026-hallelubert,
    title = "{H}allelu{BERT}: Let Every Token That Has Meaning Bear Its Weight",
    author = "Schmitt, Raphael",
    editor = "Piperidis, Stelios  and
      Bel, N{\'u}ria  and
      van den Heuvel, Henk  and
      Ide, Nancy  and
      Krek, Simon  and
      Toral, Antonio",
    booktitle = "Proceedings of the Fifteenth Language Resources and Evaluation Conference",
    month = may,
    year = "2026",
    address = "Palma de Mallorca, Spain",
    publisher = "ELRA Language Resource Association",
    url = "https://aclanthology.org/2026.lrec-1.236/",
    doi = "10.63317/3qdqexx4e9i2",
    pages = "3022--3030",
    abstract = "Transformer-based models have advanced NLP, yet Hebrew still lacks a RoBERTa encoder that is trained at scale and released in both base and large variants. We present HalleluBERT, a RoBERTa-based encoder family trained from scratch on 49.1{~}GB of deduplicated Hebrew web text and Wikipedia using a Hebrew-specific byte-level BPE vocabulary. On native Hebrew benchmarks for named entity recognition (BMC, NEMO) and sentiment classification (SMCD), HalleluBERT outperforms monolingual and multilingual baselines, and yields the highest unweighted mean score across the three benchmarks. We release model weights and tokenizer under the MIT license to support reproducible Hebrew NLP research."
}

Download as File

<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="schmitt-2026-hallelubert">
    <titleInfo>
        <title>HalleluBERT: Let Every Token That Has Meaning Bear Its Weight</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Raphael</namePart>
        <namePart type="family">Schmitt</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-05</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the Fifteenth Language Resources and Evaluation Conference</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Stelios</namePart>
            <namePart type="family">Piperidis</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Núria</namePart>
            <namePart type="family">Bel</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Henk</namePart>
            <namePart type="family">van den Heuvel</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Nancy</namePart>
            <namePart type="family">Ide</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Simon</namePart>
            <namePart type="family">Krek</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Antonio</namePart>
            <namePart type="family">Toral</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>ELRA Language Resource Association</publisher>
            <place>
                <placeTerm type="text">Palma de Mallorca, Spain</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
    </relatedItem>
    <abstract>Transformer-based models have advanced NLP, yet Hebrew still lacks a RoBERTa encoder that is trained at scale and released in both base and large variants. We present HalleluBERT, a RoBERTa-based encoder family trained from scratch on 49.1 GB of deduplicated Hebrew web text and Wikipedia using a Hebrew-specific byte-level BPE vocabulary. On native Hebrew benchmarks for named entity recognition (BMC, NEMO) and sentiment classification (SMCD), HalleluBERT outperforms monolingual and multilingual baselines, and yields the highest unweighted mean score across the three benchmarks. We release model weights and tokenizer under the MIT license to support reproducible Hebrew NLP research.</abstract>
    <identifier type="citekey">schmitt-2026-hallelubert</identifier>
    <identifier type="doi">10.63317/3qdqexx4e9i2</identifier>
    <location>
        <url>https://aclanthology.org/2026.lrec-1.236/</url>
    </location>
    <part>
        <date>2026-05</date>
        <extent unit="page">
            <start>3022</start>
            <end>3030</end>
        </extent>
    </part>
</mods>
</modsCollection>

Download as File

%0 Conference Proceedings
%T HalleluBERT: Let Every Token That Has Meaning Bear Its Weight
%A Schmitt, Raphael
%Y Piperidis, Stelios
%Y Bel, Núria
%Y van den Heuvel, Henk
%Y Ide, Nancy
%Y Krek, Simon
%Y Toral, Antonio
%S Proceedings of the Fifteenth Language Resources and Evaluation Conference
%D 2026
%8 May
%I ELRA Language Resource Association
%C Palma de Mallorca, Spain
%F schmitt-2026-hallelubert
%X Transformer-based models have advanced NLP, yet Hebrew still lacks a RoBERTa encoder that is trained at scale and released in both base and large variants. We present HalleluBERT, a RoBERTa-based encoder family trained from scratch on 49.1 GB of deduplicated Hebrew web text and Wikipedia using a Hebrew-specific byte-level BPE vocabulary. On native Hebrew benchmarks for named entity recognition (BMC, NEMO) and sentiment classification (SMCD), HalleluBERT outperforms monolingual and multilingual baselines, and yields the highest unweighted mean score across the three benchmarks. We release model weights and tokenizer under the MIT license to support reproducible Hebrew NLP research.
%R 10.63317/3qdqexx4e9i2
%U https://aclanthology.org/2026.lrec-1.236/
%U https://doi.org/10.63317/3qdqexx4e9i2
%P 3022-3030

Download as File

Markdown (Informal)

[HalleluBERT: Let Every Token That Has Meaning Bear Its Weight](https://aclanthology.org/2026.lrec-1.236/) (Schmitt, LREC 2026)

HalleluBERT: Let Every Token That Has Meaning Bear Its Weight (Schmitt, LREC 2026)

ACL

Raphael Schmitt. 2026. HalleluBERT: Let Every Token That Has Meaning Bear Its Weight. In Proceedings of the Fifteenth Language Resources and Evaluation Conference, pages 3022–3030, Palma de Mallorca, Spain. ELRA Language Resource Association.