Adapting Large Language Models to Low-Resource Tibetan: A Two-Stage Continual and Supervised Fine-Tuning Study

TL;DR

This paper presents a two-stage fine-tuning approach for adapting large language models to Tibetan, significantly improving translation quality and understanding of adaptation dynamics in low-resource language settings.

Contribution

It introduces a novel two-stage adaptation process combining Continual Pretraining and Supervised Fine-Tuning for Tibetan, with detailed analysis of model layer adaptations.

Findings

Perplexity decreased from 2.98 to 1.54

Translation BLEU score improved from 0.046 to 0.261

Layer-wise analysis reveals adaptation mainly in embedding and output layers

Abstract

Adapting large language models (LLMs) to low-resource languages remains a major challenge due to data scarcity and cross-lingual drift. This work presents a two-stage adaptation of Qwen2.5-3B to Tibetan, a morphologically rich and underrepresented language. We employ Continual Pretraining (CPT) to establish Tibetan linguistic grounding, followed by Supervised Fine-Tuning (SFT) for task and translation specialization. Empirical evaluations demonstrate a consistent decrease in perplexity (from 2.98 $\rightarrow$ 1.54) and substantial improvements in Chinese$\rightarrow$Tibetan translation quality (BLEU: 0.046 $\rightarrow$ 0.261; chrF: 2.2 $\rightarrow$ 6.6). Layer-wise analysis across 435 layers in Qwen3-4B reveals that adaptation primarily concentrates on embedding and output heads, with mid--late MLP projections encoding domain-specific transformations. Our findings suggest that CPT constructs a Tibetan semantic manifold while SFT sharpens task alignment with minimal representational disruption. This study provides the first quantitative exploration of Tibetan adaptation dynamics for LLMs, and offers an open, reproducible framework for extending multilingual foundation models to low-resource settings.