Efficient Training of Robust Traditional Chinese LLaMA-1B on a Single Consumer GPU: Continual Pre-training, SFT, and DPO
This work addresses a practical reliability gap for deploying Traditional Chinese SLMs in cost-effective, on-device applications, offering an incremental improvement through a reproducible, adapter-only pipeline.
The paper tackled the problem of token-level instability in Traditional Chinese (TC) Small Language Models (SLMs), where models unpredictably emit non-TC characters or code-switch into other languages, by developing PureTC-1B, a three-stage stabilization pipeline using LoRA adapters. The result was a 51.3% relative reduction in non-TC output tokens on a benchmark and up to a 77.2% relative reduction in incorrect-language tokens on a Named Entity Translation task compared to baseline models.
Small Language Models (SLMs) enable cost-effective, on-device and latency-sensitive AI applications, yet their deployment in Traditional Chinese (TC) remains hindered by token-level instability - models unpredictably emit non-TC characters or code-switch into other languages. We address this practical reliability gap by creating PureTC-1B, a three-stage stabilization pipeline for Llama-3.2-1B-Instruct (an open-weight, instruction-tuned model released by Meta) using parameter-efficient LoRA adapters. Our method combines Continual Pre-Training (CPT) on TC-centric corpora, Supervised Fine-Tuning (SFT) with instruction data, and Direct Preference Optimization (DPO) using TC-adherence preferences to improve monolingual robustness without full-model retraining. On a benchmark designed to simulate real-world usage, PureTC-1B achieves a 51.3% relative reduction (micro-average) in non-TC output tokens versus the base model. On a Named Entity Translation (NET) task, PureTC-1B further reduces incorrect-language tokens by 77.2% relative to Llama-3B and 57.2% relative to Qwen-1.5B, indicating that robust TC adherence is attainable even at the 1B scale. The pipeline is reproducible, adapter-only, and hardware-friendly, offering practitioners a practical recipe to enhance language stability for TC and potentially other non-English languages.