Taibao-7B: An Intellectual Property Domain Large Language Model

Time: Dec. 2023

The current Taibao model is obtained by performing supervised, efficient fine-tuning using LoRA on the foundation model Qwen-7B for instruction tasks. The 7-billion-parameter Qwen-7B foundation model underwent multiple rounds of pretraining on high-quality datasets comprising over 2.5 trillion tokens across Chinese, English, code, mathematics, and specialized domains. This pretraining has endowed the model with a foundational capability for instruction understanding. Its natural language processing and comprehension abilities rank among the best in domestic open-source large models, significantly surpassing other models of similar scale while closely trailing a series of the most advanced closed-source models.

The model focuses on the field of intelligent intellectual property (IP) services, with research and development centered around key aspects of IP creation, management, protection, and utilization. It aims to provide users with comprehensive, efficient, and effective IP services.

We released the LoRA weigth of Taibao-7B, you need combine it with Qwen-7B

More details please see our github repository.

Cite

@misc{Taibao,
  author = {Qiyao Wang, Yuan Lin},
  title = {Taibao-7B: An Intellectual Property Domain Large Language Model},
  year = {2023},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/QiYao-Wang/Taibao}},
}