# Exploring Design Choices for Building Language-Specific LLMs

arXiv: https://arxiv.org/abs/2406.14670 · HTML: https://mlsanity.com/paper/2406.14670/

## Key facts

- Cited by 22 papers in the map (data through 14 Oct 2026).
- Builds on 30 of its 46 references that are in the map.
- First posted on arXiv 20 Jun 2024.

## Builds on (most cited first, 30 of 30)

- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Neural Machine Translation of Rare Words with Subword Units](https://mlsanity.com/paper/1508.07909/) (2015)
- [SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processing](https://mlsanity.com/paper/1808.06226/) (2018)
- [BLOOM: A 176B-Parameter Open-Access Multilingual Language Model](https://mlsanity.com/paper/2211.05100/) (2022)
- [Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone](https://mlsanity.com/paper/2404.14219/) (2024)
- [No Language Left Behind: Scaling Human-Centered Machine Translation](https://mlsanity.com/paper/2207.04672/) (2022)
- [XNLI: Evaluating Cross-lingual Sentence Representations](https://mlsanity.com/paper/1809.05053/) (2018)
- [The FLORES-101 Evaluation Benchmark for Low-Resource and Multilingual Machine Translation](https://mlsanity.com/paper/2106.03193/) (2021)
- [XL-Sum: Large-Scale Multilingual Abstractive Summarization for 44 Languages](https://mlsanity.com/paper/2106.13822/) (2021)
- [XCOPA: A Multilingual Dataset for Causal Commonsense Reasoning](https://mlsanity.com/paper/2005.00333/) (2020)
- [How Good is Your Tokenizer? On the Monolingual Performance of Multilingual Language Models](https://mlsanity.com/paper/2012.15613/) (2020)
- [Efficient and Effective Text Encoding for Chinese LLaMA and Alpaca](https://mlsanity.com/paper/2304.08177/) (2023)
- [Aya Model: An Instruction Finetuned Open-Access Multilingual Language Model](https://mlsanity.com/paper/2402.07827/) (2024)
- [Do All Languages Cost the Same? Tokenization in the Era of Commercial Language Models](https://mlsanity.com/paper/2305.13707/) (2023)
- [The BigScience ROOTS Corpus: A 1.6TB Composite Multilingual Dataset](https://mlsanity.com/paper/2303.03915/) (2023)
- [Multilingual LAMA: Investigating Knowledge in Multilingual Pretrained Language Models](https://mlsanity.com/paper/2102.00894/) (2021)
- [Towards Leaving No Indic Language Behind: Building Monolingual Corpora, Benchmark and Models for Indic Languages](https://mlsanity.com/paper/2212.05409/) (2022)
- [LLaMA Beyond English: An Empirical Study on Language Capability Transfer](https://mlsanity.com/paper/2401.01055/) (2024)
- [WECHSEL: Effective initialization of subword embeddings for cross-lingual transfer of monolingual language models](https://mlsanity.com/paper/2112.06598/) (2021)
- [As Good as New. How to Successfully Recycle English GPT-2 to Make Models for Other Languages](https://mlsanity.com/paper/2012.05628/) (2020)
- [How to Adapt Your Pretrained Multilingual Model to 1600 Languages](https://mlsanity.com/paper/2106.02124/) (2021)
- [Expanding Pretrained Models to Thousands More Languages via Lexicon-based Adaptation](https://mlsanity.com/paper/2203.09435/) (2022)
- [Fast Vocabulary Transfer for Language Model Compression](https://mlsanity.com/paper/2402.09977/) (2024)
- [Efficient Domain Adaptation of Language Models via Adaptive Tokenization](https://mlsanity.com/paper/2109.07460/) (2021)
- [Efficiently Adapting Pretrained Language Models To New Languages](https://mlsanity.com/paper/2311.05741/) (2023)
- [GreenPLM: Cross-Lingual Transfer of Monolingual Pre-Trained Language Models at Almost No Cost](https://mlsanity.com/paper/2211.06993/) (2022)
- [Cedille: A large autoregressive French language model](https://mlsanity.com/paper/2202.03371/) (2022)
- [Embedding structure matters: Comparing methods to adapt multilingual vocabularies to new languages](https://mlsanity.com/paper/2309.04679/) (2023)
- [RakutenAI-7B: Extending Large Language Models for Japanese](https://mlsanity.com/paper/2403.15484/) (2024)
- [Impact of Tokenization on LLaMa Russian Adaptation](https://mlsanity.com/paper/2312.02598/) (2023)

## Built on by (most cited first, 22 of 22)

- [MEXA: Multilingual Evaluation of English-Centric LLMs via Cross-Lingual Alignment](https://mlsanity.com/paper/2410.05873/) (2024)
- [EMMA-500: Enhancing Massively Multilingual Adaptation of Large Language Models](https://mlsanity.com/paper/2409.17892/) (2024)
- [TituLLMs: A Family of Bangla LLMs with Comprehensive Benchmarking](https://mlsanity.com/paper/2502.11187/) (2025)
- [Efficient Continual Pre-training of LLMs for Low-resource Languages](https://mlsanity.com/paper/2412.10244/) (2024)
- [Facilitating large language model Russian adaptation with Learned Embedding Propagation](https://mlsanity.com/paper/2412.21140/) (2024)
- [Multilingual Arbitrage: Optimizing Data Pools to Accelerate Multilingual Progress](https://mlsanity.com/paper/2408.14960/) (2024)
- [Adapting Chat Language Models Using Only Target Unlabeled Language Data](https://mlsanity.com/paper/2412.11704/) (2024)
- [Stop Taking Tokenizers for Granted: They Are Core Design Decisions in Large Language Models](https://mlsanity.com/paper/2601.13260/) (2026)
- [From English-Centric to Effective Bilingual: LLMs with Custom Tokenizers for Underrepresented Languages](https://mlsanity.com/paper/2410.18836/) (2024)
- [Why We Build Local Large Language Models: An Observational Analysis from 35 Japanese and Multilingual LLMs](https://mlsanity.com/paper/2412.14471/) (2024)
- [Evaluation of LLMs in Medical Text Summarization: The Role of Vocabulary Adaptation in High OOV Settings](https://mlsanity.com/paper/2505.21242/) (2025)
- [Mitigating Catastrophic Forgetting in Target Language Adaptation of LLMs via Source-Shielded Updates](https://mlsanity.com/paper/2512.04844/) (2025)
- [Prune or Retrain: Optimizing the Vocabulary of Multilingual Models for Estonian](https://mlsanity.com/paper/2501.02631/) (2025)
- [How Does "English (US)" Become the Default? Triangulating Structural Bias Towards American English Across the LLM Pipeline](https://mlsanity.com/paper/2604.04204/) (2026)
- [Teaching Old Tokenizers New Words: Efficient Tokenizer Adaptation for Pre-trained Models](https://mlsanity.com/paper/2512.03989/) (2025)
- [Improving Multilingual Capabilities with Cultural and Local Knowledge in Large Language Models While Enhancing Native Performance](https://mlsanity.com/paper/2504.09753/) (2025)
- [Vocabulary Customization for Efficient Domain-Specific LLM Deployment](https://mlsanity.com/paper/2509.26124/) (2025)
- [MultiHashFormer: Hash-based Generative Language Models](https://mlsanity.com/paper/2606.28057/) (2026)
- [Prompt, Translate, Fine-Tune, Re-Initialize, or Instruction-Tune? Adapting LLMs for In-Context Learning in Low-Resource Languages](https://mlsanity.com/paper/2506.19187/) (2025)
- [Reading the News: Adapting Large Language Models to Swedish Journalism Through Continued Pre-Training](https://mlsanity.com/paper/2608.30609/) (2026)
- [From Zero to Hero: An Open LLM Ecosystem for Armenian](https://mlsanity.com/paper/2609.03350/) (2026)
- [To Each Language Its Tokenizer: Modular Tokenizers for Efficient Multilingual LLMs](https://mlsanity.com/paper/2609.15528/) (2026)

## Abstract (arXiv)

Despite rapid progress in large language models (LLMs), their performance on a vast majority of languages remains unsatisfactory. In this paper, we study building language-specific LLMs by adapting monolingual and multilingual LLMs. We conduct systematic experiments on how design choices (base model selection, vocabulary extension, and continued pretraining) impact the adapted LLM, both in terms of efficiency (how many tokens are needed to encode the same amount of information) and end task performance. We find that (1) the initial performance of LLM does not always correlate with the final performance after the adaptation. Adapting an English-centric models can yield better results than adapting multilingual models despite their worse initial performance on low-resource languages. (2) Efficiency can easily improved with simple vocabulary extension and continued pretraining in most LLMs we study, and (3) The optimal adaptation method (choice of the base model, new vocabulary size, training data, initialization strategy) is highly language-dependent, and the simplest embedding initialization works well across various experimental settings. Together, our work lays foundations on efficiently building language-specific LLMs by adapting existing LLMs.