CLAIMar 21

SozKZ: Training Efficient Small Language Models for Kazakh from Scratch

arXiv:2603.2085414.41 citations
AI Analysis

This addresses the need for efficient language technology for Kazakh speakers, offering a viable solution for low-resource languages with incremental improvements in model design.

The paper tackled the problem of underserved Kazakh language in multilingual models by training small, dedicated language models from scratch with a language-appropriate tokenizer, achieving competitive performance such as 30.3% accuracy on Kazakh cultural QA and 25.5% on topic classification, surpassing larger multilingual models.

Kazakh, a Turkic language spoken by over 22 million people, remains underserved by existing multilingual language models, which allocate minimal capacity to low-resource languages and employ tokenizers ill-suited to agglutinative morphology. We present SozKZ, a family of Llama-architecture language models (50M-600M parameters) trained entirely from scratch on 9 billion tokens of Kazakh text with a dedicated 50K BPE tokenizer. We evaluate all models on three Kazakh benchmarks -- multiple-choice cultural QA, reading comprehension (Belebele), and topic classification (SIB-200) -- alongside five multilingual baselines ranging from 500M to 3B parameters. Our 600M model achieves 30.3% accuracy on Kazakh cultural QA, approaching the 32.0% of Llama-3.2-1B (2x larger), and 25.5% on SIB-200 topic classification, surpassing all evaluated multilingual models up to 2B parameters. We observe consistent scaling from 50M to 600M, with MC QA accuracy rising from 22.8% to 30.3%, suggesting that further scaling remains beneficial. These results demonstrate that small, dedicated models trained from scratch with a language-appropriate tokenizer offer a viable path for low-resource language technology, achieving competitive performance at a fraction of the computational cost. All models and the tokenizer are released under open licenses.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes