@inproceedings{patil-etal-2025-regional,
title = "Regional-{T}iny{S}tories: A Small Language Model Framework for Evaluating Language Learning, Tokenizers, and Datasets",
author = "Patil, Nirvan and
Inamdar, Malhar Abhay and
Gosai, Agnivo and
Pathak, Guruprasad and
Joshi, Anish and
Joshirao, Anish and
Dandekar, Raj and
Dandekar, Rajat and
Panat, Sreedath",
editor = "Inui, Kentaro and
Sakti, Sakriani and
Wang, Haofen and
Wong, Derek F. and
Bhattacharyya, Pushpak and
Banerjee, Biplab and
Ekbal, Asif and
Chakraborty, Tanmoy and
Singh, Dhirendra Pratap",
booktitle = "Proceedings of the 14th International Joint Conference on Natural Language Processing and the 4th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics",
month = dec,
year = "2025",
address = "Mumbai, India",
publisher = "The Asian Federation of Natural Language Processing and The Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-ijcnlp.142/",
pages = "2318--2367",
ISBN = "979-8-89176-303-6",
abstract = "Small, resource-efficient language models are pivotal for extending high-quality text generation to low-resource and regional languages{---}the true frontier of linguistic equity in AI. Yet research largely prioritises massive English-centric systems, leaving regional-centric (low-resource) language modelling underexplored, particularly how tokenizer design, dataset diversity, and linguistic structure shape the inference of Small Language Models (SLMs) under realistic computational and data constraints. We present Regional-TinyStories, a lightweight framework that treats SLMs as cost-effective stand-ins for LLMs, enabling rapid, variable-wise inference-based analysis. Extending TinyStories to Hindi, Marathi, and Bangla, we release datasets of 2M synthetic and translated stories per language and train over 20 SLMs spanning 5{--}157M parameters. Using this framework, we (i) uncover contrasts between form-oriented (grammar, fluency) and content-oriented (context, completeness, creativity) metrics; (ii) chart language-specific learning dynamics; (iii) rank tokenizers, showing Indic-specific Sarvam-1 outperforming SUTRA and generic Tiktoken (GPT-2) across all metrics; and (iv) demonstrate that dataset semantic quality (translation vs. synthetic) strongly governs downstream generation. Validation through an LLM-as-Judge ensemble (GPT-4o, LLaMA-3.3-70B) and a 100+ participant human study confirms these trends while exposing systematic score inflation in automated evaluations. Regional-TinyStories offers a reproducible path to benchmark tokenizers, datasets, and SLM designs for scalable, context-faithful generation in low-resource settings."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="patil-etal-2025-regional">
<titleInfo>
<title>Regional-TinyStories: A Small Language Model Framework for Evaluating Language Learning, Tokenizers, and Datasets</title>
</titleInfo>
<name type="personal">
<namePart type="given">Nirvan</namePart>
<namePart type="family">Patil</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Malhar</namePart>
<namePart type="given">Abhay</namePart>
<namePart type="family">Inamdar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Agnivo</namePart>
<namePart type="family">Gosai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guruprasad</namePart>
<namePart type="family">Pathak</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anish</namePart>
<namePart type="family">Joshi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anish</namePart>
<namePart type="family">Joshirao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Raj</namePart>
<namePart type="family">Dandekar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rajat</namePart>
<namePart type="family">Dandekar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sreedath</namePart>
<namePart type="family">Panat</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-12</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 14th International Joint Conference on Natural Language Processing and the 4th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics</title>
</titleInfo>
<name type="personal">
<namePart type="given">Kentaro</namePart>
<namePart type="family">Inui</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sakriani</namePart>
<namePart type="family">Sakti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haofen</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Derek</namePart>
<namePart type="given">F</namePart>
<namePart type="family">Wong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pushpak</namePart>
<namePart type="family">Bhattacharyya</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Biplab</namePart>
<namePart type="family">Banerjee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Asif</namePart>
<namePart type="family">Ekbal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tanmoy</namePart>
<namePart type="family">Chakraborty</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dhirendra</namePart>
<namePart type="given">Pratap</namePart>
<namePart type="family">Singh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>The Asian Federation of Natural Language Processing and The Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Mumbai, India</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-303-6</identifier>
</relatedItem>
<abstract>Small, resource-efficient language models are pivotal for extending high-quality text generation to low-resource and regional languages—the true frontier of linguistic equity in AI. Yet research largely prioritises massive English-centric systems, leaving regional-centric (low-resource) language modelling underexplored, particularly how tokenizer design, dataset diversity, and linguistic structure shape the inference of Small Language Models (SLMs) under realistic computational and data constraints. We present Regional-TinyStories, a lightweight framework that treats SLMs as cost-effective stand-ins for LLMs, enabling rapid, variable-wise inference-based analysis. Extending TinyStories to Hindi, Marathi, and Bangla, we release datasets of 2M synthetic and translated stories per language and train over 20 SLMs spanning 5–157M parameters. Using this framework, we (i) uncover contrasts between form-oriented (grammar, fluency) and content-oriented (context, completeness, creativity) metrics; (ii) chart language-specific learning dynamics; (iii) rank tokenizers, showing Indic-specific Sarvam-1 outperforming SUTRA and generic Tiktoken (GPT-2) across all metrics; and (iv) demonstrate that dataset semantic quality (translation vs. synthetic) strongly governs downstream generation. Validation through an LLM-as-Judge ensemble (GPT-4o, LLaMA-3.3-70B) and a 100+ participant human study confirms these trends while exposing systematic score inflation in automated evaluations. Regional-TinyStories offers a reproducible path to benchmark tokenizers, datasets, and SLM designs for scalable, context-faithful generation in low-resource settings.</abstract>
<identifier type="citekey">patil-etal-2025-regional</identifier>
<location>
<url>https://aclanthology.org/2025.findings-ijcnlp.142/</url>
</location>
<part>
<date>2025-12</date>
<extent unit="page">
<start>2318</start>
<end>2367</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Regional-TinyStories: A Small Language Model Framework for Evaluating Language Learning, Tokenizers, and Datasets
%A Patil, Nirvan
%A Inamdar, Malhar Abhay
%A Gosai, Agnivo
%A Pathak, Guruprasad
%A Joshi, Anish
%A Joshirao, Anish
%A Dandekar, Raj
%A Dandekar, Rajat
%A Panat, Sreedath
%Y Inui, Kentaro
%Y Sakti, Sakriani
%Y Wang, Haofen
%Y Wong, Derek F.
%Y Bhattacharyya, Pushpak
%Y Banerjee, Biplab
%Y Ekbal, Asif
%Y Chakraborty, Tanmoy
%Y Singh, Dhirendra Pratap
%S Proceedings of the 14th International Joint Conference on Natural Language Processing and the 4th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics
%D 2025
%8 December
%I The Asian Federation of Natural Language Processing and The Association for Computational Linguistics
%C Mumbai, India
%@ 979-8-89176-303-6
%F patil-etal-2025-regional
%X Small, resource-efficient language models are pivotal for extending high-quality text generation to low-resource and regional languages—the true frontier of linguistic equity in AI. Yet research largely prioritises massive English-centric systems, leaving regional-centric (low-resource) language modelling underexplored, particularly how tokenizer design, dataset diversity, and linguistic structure shape the inference of Small Language Models (SLMs) under realistic computational and data constraints. We present Regional-TinyStories, a lightweight framework that treats SLMs as cost-effective stand-ins for LLMs, enabling rapid, variable-wise inference-based analysis. Extending TinyStories to Hindi, Marathi, and Bangla, we release datasets of 2M synthetic and translated stories per language and train over 20 SLMs spanning 5–157M parameters. Using this framework, we (i) uncover contrasts between form-oriented (grammar, fluency) and content-oriented (context, completeness, creativity) metrics; (ii) chart language-specific learning dynamics; (iii) rank tokenizers, showing Indic-specific Sarvam-1 outperforming SUTRA and generic Tiktoken (GPT-2) across all metrics; and (iv) demonstrate that dataset semantic quality (translation vs. synthetic) strongly governs downstream generation. Validation through an LLM-as-Judge ensemble (GPT-4o, LLaMA-3.3-70B) and a 100+ participant human study confirms these trends while exposing systematic score inflation in automated evaluations. Regional-TinyStories offers a reproducible path to benchmark tokenizers, datasets, and SLM designs for scalable, context-faithful generation in low-resource settings.
%U https://aclanthology.org/2025.findings-ijcnlp.142/
%P 2318-2367
Markdown (Informal)
[Regional-TinyStories: A Small Language Model Framework for Evaluating Language Learning, Tokenizers, and Datasets](https://aclanthology.org/2025.findings-ijcnlp.142/) (Patil et al., Findings 2025)
ACL
- Nirvan Patil, Malhar Abhay Inamdar, Agnivo Gosai, Guruprasad Pathak, Anish Joshi, Anish Joshirao, Raj Dandekar, Rajat Dandekar, and Sreedath Panat. 2025. Regional-TinyStories: A Small Language Model Framework for Evaluating Language Learning, Tokenizers, and Datasets. In Proceedings of the 14th International Joint Conference on Natural Language Processing and the 4th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics, pages 2318–2367, Mumbai, India. The Asian Federation of Natural Language Processing and The Association for Computational Linguistics.