arXiv cs.CL· Mullosharaf K. Arabov, Svetlana S. Khaybullina·· 13 小时前AI 评分20
面向低资源黏着语巴什基尔语的大语言模型参数高效适配：语料、数据集与可复现基准
Parameter-Efficient Adaptation of Large Language Models to a Low-Resource Agglutinative Language: A Corpus, Datasets, and a Reproducible Benchmark for the Bashkir Language
AI 导读
研究发布巴什基尔语开放资源套件与适配方法基准，含71,567篇文档、46.9M tokens的Bashkir Web Corpus及三个标注数据集。在8个架构（82M-8B）上对比全量微调、QLoRA与上下文学习，YandexGPT-5-Lite-8B结合持续预训练在二分类上达macro-F1 0.748，高于XLM-R的0.653和mBERT的0.650。
正文 · 原文
Abstract:Bashkir is a low-resource agglutinative language written in Cyrillic and remains poorly represented in modern LLMs. We present an open resource suite and a systematic benchmark of adaptation methods: the Bashkir Web Corpus (71,567 documents, 46.9M tokens, 16 sources) and three labelled datasets for binary, multiclass and multilabel text classification; the benchmark covers the first two. Experiments use eight architectures (82M-8B: DistilGPT2, GPT-2 base/medium, Phi-3-mini, Mistral-7B, YandexGPT-5-Lite-8B, mBERT, XLM-R) and three adaptation groups (full fine-tuning, QLoRA, in-context learning), plus continual pretraining (CPT). (1) QLoRA on 3.8-8B models outperforms encoder-only fine-tuning on binary classification: YandexGPT-5-Lite-8B with CPT reaches macro-F1 0.748 vs 0.653 (XLM-R) and 0.650 (mBERT); pairwise differences are significant after Holm correction (p<0.001). (2) YandexGPT-5-Lite-8B significantly outperforms Mistral-7B and Phi-3-mini on multiclass classification, BPB and native-speaker ratings (mean 3.28 vs 2.52 and 2.22; n=100; alpha=0.69-0.77; p<0.05); on binary classification its QLoRA advantage is not significant (p>=0.084). (3) CPT effects depend on model and task: perplexity improves for Phi-3-mini (+0.58) and YandexGPT-5-Lite-8B (+0.33) but worsens for Mistral-7B (-1.15, p<0.001); on binary classification CPT helps Mistral-7B most (+4.0 macro-F1). (4) Tokeniser fertility on Bashkir varies 2.67x (2.92-7.80) and tracks adaptation quality, while perplexity across tokenisers is not comparable. Cyrillic tokenisers with Russian pretraining are preferable; LLM judges reproduce native ratings for grammar and coherence (rho=0.33-0.46) but not language correctness (rho=0.16, p=0.11); native evaluation remains necessary. All artefacts are open-licensed.
| Comments: | Preprint. 29 pages, 1 figure, 17 tables |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.31397 [cs.CL] |
| (or arXiv:2610.31397v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.31397 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Mullosharaf Arabov Am [view email]
[v1]
Thu, 22 Oct 2026 14:59:55 UTC (1,117 KB)
来源：arXiv cs.CL · arxiv.org