@inproceedings{bergmanis-etal-2026-tildeopen,
title = "{T}ilde{O}pen {LLM}: Leveraging Curriculum Learning to Achieve Equitable Language Representation",
author = "Bergmanis, Toms and
Pretkalni{\c{n}}{\v{s}}, Ingus J{\={a}}nis and
Kronis, Martins and
Nicmanis, Davis and
Jelinska, Je{\c{l}}izaveta and
Rozis, Roberts and
V{\={i}}ksna, Rinalds and
Pinnis, Marcis",
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-1.834/",
doi = "10.63317/2emdserc8sfq",
pages = "10634--10652",
abstract = "Large language models often underperform in many European languages due to the dominance of English and a few high-resource languages in training data. This paper presents TildeOpen LLM, a 30-billion-parameter open-weight foundational model trained on 34 European languages to promote linguistic equity and improve performance for low-resource languages. To address the data imbalance, we combine dataset upsampling with a curriculum-based training schedule that alternates between uniform and natural language distributions. The resulting model performs favorably compared to other multilingual LLMs despite being trained with significantly fewer computing resources. Evaluation across multiple multilingual benchmarks shows that TildeOpen surpasses existing open-weight models in text generation and comprehension, particularly for Baltic, Finno-Ugric, and Slavic languages. Human evaluations confirm an up to tenfold reduction in linguistic errors relative to leading baselines. The model and associated resources are fully open-weight and publicly available at huggingface.co/TildeAI/TildeOpen-30b. These outcomes demonstrate that careful data curation and balanced training strategies can substantially enhance multilingual model quality without increasing model size or training volume."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="bergmanis-etal-2026-tildeopen">
<titleInfo>
<title>TildeOpen LLM: Leveraging Curriculum Learning to Achieve Equitable Language Representation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Toms</namePart>
<namePart type="family">Bergmanis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ingus</namePart>
<namePart type="given">Jānis</namePart>
<namePart type="family">Pretkalniņš</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Martins</namePart>
<namePart type="family">Kronis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Davis</namePart>
<namePart type="family">Nicmanis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jeļizaveta</namePart>
<namePart type="family">Jelinska</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Roberts</namePart>
<namePart type="family">Rozis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rinalds</namePart>
<namePart type="family">Vīksna</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marcis</namePart>
<namePart type="family">Pinnis</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>Large language models often underperform in many European languages due to the dominance of English and a few high-resource languages in training data. This paper presents TildeOpen LLM, a 30-billion-parameter open-weight foundational model trained on 34 European languages to promote linguistic equity and improve performance for low-resource languages. To address the data imbalance, we combine dataset upsampling with a curriculum-based training schedule that alternates between uniform and natural language distributions. The resulting model performs favorably compared to other multilingual LLMs despite being trained with significantly fewer computing resources. Evaluation across multiple multilingual benchmarks shows that TildeOpen surpasses existing open-weight models in text generation and comprehension, particularly for Baltic, Finno-Ugric, and Slavic languages. Human evaluations confirm an up to tenfold reduction in linguistic errors relative to leading baselines. The model and associated resources are fully open-weight and publicly available at huggingface.co/TildeAI/TildeOpen-30b. These outcomes demonstrate that careful data curation and balanced training strategies can substantially enhance multilingual model quality without increasing model size or training volume.</abstract>
<identifier type="citekey">bergmanis-etal-2026-tildeopen</identifier>
<identifier type="doi">10.63317/2emdserc8sfq</identifier>
<location>
<url>https://aclanthology.org/2026.lrec-1.834/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>10634</start>
<end>10652</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T TildeOpen LLM: Leveraging Curriculum Learning to Achieve Equitable Language Representation
%A Bergmanis, Toms
%A Pretkalniņš, Ingus Jānis
%A Kronis, Martins
%A Nicmanis, Davis
%A Jelinska, Jeļizaveta
%A Rozis, Roberts
%A Vīksna, Rinalds
%A Pinnis, Marcis
%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 bergmanis-etal-2026-tildeopen
%X Large language models often underperform in many European languages due to the dominance of English and a few high-resource languages in training data. This paper presents TildeOpen LLM, a 30-billion-parameter open-weight foundational model trained on 34 European languages to promote linguistic equity and improve performance for low-resource languages. To address the data imbalance, we combine dataset upsampling with a curriculum-based training schedule that alternates between uniform and natural language distributions. The resulting model performs favorably compared to other multilingual LLMs despite being trained with significantly fewer computing resources. Evaluation across multiple multilingual benchmarks shows that TildeOpen surpasses existing open-weight models in text generation and comprehension, particularly for Baltic, Finno-Ugric, and Slavic languages. Human evaluations confirm an up to tenfold reduction in linguistic errors relative to leading baselines. The model and associated resources are fully open-weight and publicly available at huggingface.co/TildeAI/TildeOpen-30b. These outcomes demonstrate that careful data curation and balanced training strategies can substantially enhance multilingual model quality without increasing model size or training volume.
%R 10.63317/2emdserc8sfq
%U https://aclanthology.org/2026.lrec-1.834/
%U https://doi.org/10.63317/2emdserc8sfq
%P 10634-10652
Markdown (Informal)
[TildeOpen LLM: Leveraging Curriculum Learning to Achieve Equitable Language Representation](https://aclanthology.org/2026.lrec-1.834/) (Bergmanis et al., LREC 2026)
ACL
- Toms Bergmanis, Ingus Jānis Pretkalniņš, Martins Kronis, Davis Nicmanis, Jeļizaveta Jelinska, Roberts Rozis, Rinalds Vīksna, and Marcis Pinnis. 2026. TildeOpen LLM: Leveraging Curriculum Learning to Achieve Equitable Language Representation. In Proceedings of the Fifteenth Language Resources and Evaluation Conference, pages 10634–10652, Palma de Mallorca, Spain. ELRA Language Resource Association.