← 研踪 PaperTrailarXiv:2512.03976 · v2
Adapting Large Language Models to Low-Resource Tibetan: A Two-Stage Continual and Supervised Fine-Tuning Study
首次提交 2025-12-04 01:06
版本更新 2026-11-16 12:08
雷达观测 2026-11-17 17:12
摘要
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 1.54) and substantial improvements in ChineseTibetan translation quality (BLEU: 0.046 0.261; chrF: 2.2 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.
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v2 · 2026-11-16 12:08 · 最新观测版本
Adapting Large Language Models to Low-Resource Tibetan: A Two-Stage Continual and Supervised Fine-Tuning Study
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 1.54) and substantial improvements in ChineseTibetan translation quality (BLEU: 0.046 0.261; chrF: 2.2 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.
查看此版本解读 →查看此版本原文 ↗