@inproceedings{panahi-etal-2026-synthllm,
title = "{S}ynth{LLM}: An {LLM}-based Scalable Synthetic Data Generation Pipeline for Low-Resource Languages",
author = "Panahi, Solmaz and
Nedumpozhimana, Vasudevan and
Kelleher, John",
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-1.844/",
doi = "10.63317/36i5afj23ivf",
pages = "10776--10791",
abstract = "Large Language Models (LLMs) have enabled scalable synthetic data generation, yet their effective adaptation to low-resource languages remains underexplored. We introduce an LLM-based generate and annotate paradigm to create synthetic datasets for low-resource NLP classification tasks. The framework employs a smaller model for text generation and a stronger model for automatic annotation. Using Farsi Natural Language Inference (NLI) as a case study, we construct a large-scale synthetic dataset of 100,000 labeled instances. We provide a systematic empirical analysis of annotation quality, label-distribution effects, and training regimes. We compare GPT-4o-mini, Aya-23-35B, and DeBERTa as annotators and examine how annotation variability propagates to downstream performance. Our results show that a warm-up phase with synthetic data consistently outperforms data mixing and reversed ordering. Notably, open-source annotation (Aya-23-35B) achieves comparable downstream performance to the proprietary model (GPT-4o-mini), with significant cost implications for deploying pipelines in low-resource settings. The dataset and code are publicly available at \url{https://huggingface.co/datasets/Solmazp/text2entail}."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="panahi-etal-2026-synthllm">
<titleInfo>
<title>SynthLLM: An LLM-based Scalable Synthetic Data Generation Pipeline for Low-Resource Languages</title>
</titleInfo>
<name type="personal">
<namePart type="given">Solmaz</namePart>
<namePart type="family">Panahi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vasudevan</namePart>
<namePart type="family">Nedumpozhimana</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">John</namePart>
<namePart type="family">Kelleher</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 (LLMs) have enabled scalable synthetic data generation, yet their effective adaptation to low-resource languages remains underexplored. We introduce an LLM-based generate and annotate paradigm to create synthetic datasets for low-resource NLP classification tasks. The framework employs a smaller model for text generation and a stronger model for automatic annotation. Using Farsi Natural Language Inference (NLI) as a case study, we construct a large-scale synthetic dataset of 100,000 labeled instances. We provide a systematic empirical analysis of annotation quality, label-distribution effects, and training regimes. We compare GPT-4o-mini, Aya-23-35B, and DeBERTa as annotators and examine how annotation variability propagates to downstream performance. Our results show that a warm-up phase with synthetic data consistently outperforms data mixing and reversed ordering. Notably, open-source annotation (Aya-23-35B) achieves comparable downstream performance to the proprietary model (GPT-4o-mini), with significant cost implications for deploying pipelines in low-resource settings. The dataset and code are publicly available at https://huggingface.co/datasets/Solmazp/text2entail.</abstract>
<identifier type="citekey">panahi-etal-2026-synthllm</identifier>
<identifier type="doi">10.63317/36i5afj23ivf</identifier>
<location>
<url>https://aclanthology.org/2026.lrec-1.844/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>10776</start>
<end>10791</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T SynthLLM: An LLM-based Scalable Synthetic Data Generation Pipeline for Low-Resource Languages
%A Panahi, Solmaz
%A Nedumpozhimana, Vasudevan
%A Kelleher, John
%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 panahi-etal-2026-synthllm
%X Large Language Models (LLMs) have enabled scalable synthetic data generation, yet their effective adaptation to low-resource languages remains underexplored. We introduce an LLM-based generate and annotate paradigm to create synthetic datasets for low-resource NLP classification tasks. The framework employs a smaller model for text generation and a stronger model for automatic annotation. Using Farsi Natural Language Inference (NLI) as a case study, we construct a large-scale synthetic dataset of 100,000 labeled instances. We provide a systematic empirical analysis of annotation quality, label-distribution effects, and training regimes. We compare GPT-4o-mini, Aya-23-35B, and DeBERTa as annotators and examine how annotation variability propagates to downstream performance. Our results show that a warm-up phase with synthetic data consistently outperforms data mixing and reversed ordering. Notably, open-source annotation (Aya-23-35B) achieves comparable downstream performance to the proprietary model (GPT-4o-mini), with significant cost implications for deploying pipelines in low-resource settings. The dataset and code are publicly available at https://huggingface.co/datasets/Solmazp/text2entail.
%R 10.63317/36i5afj23ivf
%U https://aclanthology.org/2026.lrec-1.844/
%U https://doi.org/10.63317/36i5afj23ivf
%P 10776-10791
Markdown (Informal)
[SynthLLM: An LLM-based Scalable Synthetic Data Generation Pipeline for Low-Resource Languages](https://aclanthology.org/2026.lrec-1.844/) (Panahi et al., LREC 2026)
ACL