yrc696/ETLCH-instruct_based_on_llama3.2-1b_taiwan_traditional_chinese

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:1BQuant:BF16Ctx Length:32kPublished:Aug 18, 2025License:afl-3.0Architecture:Transformer0.0K Warm

The yrc696/ETLCH-instruct_based_on_llama3.2-1b_taiwan_traditional_chinese is a 1 billion parameter instruction-tuned language model, based on the Llama 3.2 architecture. Developed by researchers from National Tsing Hua University, National Yang Ming Chiao Tung University, and University of Taipei, this model is specifically enhanced for stable and improved Traditional Chinese language output. It significantly outperforms the base Llama 3.2-1B-Instruct model in Chinese text generation, making it suitable for research and further fine-tuning in Traditional Chinese NLP applications.

Loading preview...

ETLCH-instruct: Enhanced Traditional Chinese LLM

This model, ETLCH-instruct_based_on_llama3.2-1b_taiwan_traditional_chinese, is a 1 billion parameter instruction-tuned language model built upon the Llama 3.2-1B-Instruct architecture. It was developed through continued pre-training and fine-tuning by researchers from National Tsing Hua University, National Yang Ming Chiao Tung University, and University of Taipei.

Key Enhancements:

  • Traditional Chinese Output Stability: The primary focus of this model is to significantly enhance the stability and quality of Traditional Chinese language generation. Statistical analysis indicates a significant improvement (p < .05) compared to the base Llama 3.2-1B-Instruct model.
  • Research and Fine-tuning: It is released for public research to expand knowledge boundaries and serves as a strong base for subsequent fine-tuning efforts, particularly for Traditional Chinese NLP tasks.

Use Cases:

  • Traditional Chinese NLP Research: Ideal for academic and research projects focusing on Traditional Chinese language processing.
  • Further Fine-tuning: Provides a robust foundation for developers and researchers to fine-tune for specific Traditional Chinese applications.

For more technical details, refer to the associated paper: https://arxiv.org/abs/2510.01616.