# Mangosteen: An Open Thai Corpus for Language Model Pretraining

arXiv: https://arxiv.org/abs/2507.14664 · HTML: https://mlsanity.com/paper/2507.14664/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 24 of its 31 references that are in the map.
- First posted on arXiv 19 Jul 2025.

## Builds on (most cited first, 24 of 24)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [Unsupervised Cross-lingual Representation Learning at Scale](https://mlsanity.com/paper/1911.02116/) (2019)
- [QLoRA: Efficient Finetuning of Quantized LLMs](https://mlsanity.com/paper/2305.14314/) (2023)
- [The Pile: An 800GB Dataset of Diverse Text for Language Modeling](https://mlsanity.com/paper/2101.00027/) (2020)
- [mT5: A massively multilingual pre-trained text-to-text transformer](https://mlsanity.com/paper/2010.11934/) (2020)
- [Scaling Language Models: Methods, Analysis & Insights from Training Gopher](https://mlsanity.com/paper/2112.11446/) (2021)
- [The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scale](https://mlsanity.com/paper/2406.17557/) (2024)
- [XNLI: Evaluating Cross-lingual Sentence Representations](https://mlsanity.com/paper/1809.05053/) (2018)
- [Bag of Tricks for Efficient Text Classification](https://mlsanity.com/paper/1607.01759/) (2016)
- [The RefinedWeb Dataset for Falcon LLM: Outperforming Curated Corpora with Web Data, and Web Data Only](https://mlsanity.com/paper/2306.01116/) (2023)
- [CCNet: Extracting High Quality Monolingual Datasets from Web Crawl Data](https://mlsanity.com/paper/1911.00359/) (2019)
- [Learning Word Vectors for 157 Languages](https://mlsanity.com/paper/1802.06893/) (2018)
- [CulturaX: A Cleaned, Enormous, and Multilingual Dataset for Large Language Models in 167 Languages](https://mlsanity.com/paper/2309.09400/) (2023)
- [A Monolingual Approach to Contextualized Word Embeddings for Mid-Resource Languages](https://mlsanity.com/paper/2006.06202/) (2020)
- [IndicLLMSuite: A Blueprint for Creating Pre-training and Fine-Tuning Datasets for Indian Languages](https://mlsanity.com/paper/2403.06350/) (2024)
- [FinGPT: Large Generative Models for a Small Language](https://mlsanity.com/paper/2311.05640/) (2023)
- [SEACrowd: A Multilingual Multimodal Data Hub and Benchmark Suite for Southeast Asian Languages](https://mlsanity.com/paper/2406.10118/) (2024)
- [Latxa: An Open Language Model and Evaluation Suite for Basque](https://mlsanity.com/paper/2403.20266/) (2024)
- [PyThaiNLP: Thai Natural Language Processing in Python](https://mlsanity.com/paper/2312.04649/) (2023)
- [WangchanBERTa: Pretraining transformer-based Thai Language Models](https://mlsanity.com/paper/2101.09635/) (2021)
- [SEA-HELM: Southeast Asian Holistic Evaluation of Language Models](https://mlsanity.com/paper/2502.14301/) (2025)
- [SambaLingo: Teaching Large Language Models New Languages](https://mlsanity.com/paper/2404.05829/) (2024)
- [Translation Errors Significantly Impact Low-Resource Languages in Cross-Lingual Learning](https://mlsanity.com/paper/2402.02080/) (2024)

## Built on by (most cited first, 2 of 2)

- [OpenJAI-v1.0: An Open Thai Large Language Model](https://mlsanity.com/paper/2510.06847/) (2025)
- [Counting and Min-Cost Encoding for Tokenization in Large Language Models](https://mlsanity.com/paper/2610.01127/) (2026)

## Abstract (arXiv)

Pre-training data shapes a language model's quality, but raw web text is noisy and demands careful cleaning. Existing large-scale corpora rely on English-centric or language-agnostic pipelines whose heuristics do not capture Thai script or cultural nuances, leaving risky material such as gambling content untreated. Prior Thai-specific efforts customize pipelines or build new ones, yet seldom release their data or document design choices, hindering reproducibility and raising the question of how to construct a transparent, high-quality Thai corpus. We introduce Mangosteen: a 47 billion-token Thai corpus built through a Thai-adapted Dolma pipeline that includes custom rule-based language ID, revised C4/Gopher quality filters, and Thai-trained content filters, plus curated non-web sources such as Wikipedia, Royal Gazette texts, OCR-extracted books, and CC-licensed YouTube subtitles. Systematic ablations using GPT-2 show the pipeline trims CommonCrawl from 202M to 25M documents while raising SEA-HELM NLG from 3 to 11; an 8B-parameter SEA-LION model continually pre-trained on Mangosteen then surpasses SEA-LION-v3 and Llama-3.1 by about four points on Thai benchmarks. We release the full pipeline code, cleaning manifests, corpus snapshot, and all checkpoints, providing a fully reproducible foundation for future Thai and regional LLM research.