Project TAME (Traditional Chinese LLM)
Participated in building a Traditional Chinese experts LLM that outperforms GPT-4 in Traditional Chinese capabilities, focusing on manufacturing datasets.
LLMPre-TrainingDataset CurationBenchmarkingSovereign AI
Project TAME: Traditional Chinese Experts LLM
I was an active participant in Project TAME, an initiative dedicated to building a highly capable Large Language Model (LLM) specialized for Traditional Chinese.
Contributions
- SOTA Traditional Chinese: Helped develop an LLM that outperforms GPT-4 in Traditional Chinese language capabilities.
- Data & Benchmarking: Responsible for collecting extensive manufacturing training data and establishing robust benchmarking datasets.
- Advanced Training: Gained and applied deep insights into state-of-the-art LLM training techniques throughout the project lifecycle.
- 🤗 Model Weights: Llama-3-Taiwan-70B-Instruct