Network Security Internet Technology Development Database Servers Mobile Phone Android Software Apple Software Computer Software News IT Information

In addition to Weibo, there is also WeChat

Please pay attention

WeChat public account

Shulou

Lesser Fullness of Grain released "Xuanyuan 70B" financial model C-Eval and CMMLU ranked first.

2025-03-29 Update From: SLTechnology News&Howtos shulou NAV: SLTechnology News&Howtos > IT Information >

Share

Shulou(Shulou.com)11/24 Report--

On September 22, du Lesser Fullness of Grain announced that the "Xuanyuan 70B" financial model is open source, and all users are free to download and try it, and announced the results of "Xuanyuan 70B" in the two authoritative language model evaluation benchmarks of C-Eval and CMMLU. Data show that in the C-Eval list, the total score of XuanYuan-70B reached 71.9 points; in the CMMLU list, it topped the list with a high score of 71.05, ranking first among all open source models on the two authoritative lists, and it is also the first major financial model in China that ranks first in the two authoritative lists at the same time.

The C-Eval list is a comprehensive examination evaluation set jointly constructed by Tsinghua University, Shanghai Jiaotong University and the University of Edinburgh. It includes almost all the mainstream models at home and abroad, covering 52 disciplines in four major directions: humanities, social sciences, science and technology, and other majors. There are 13948 multiple choice questions and four difficulty levels: middle school, undergraduate, graduate student and occupation. It is one of the most authoritative lists of large models to judge the potential of the model.

CMMLU dataset is a comprehensive Chinese evaluation benchmark, jointly launched by MBZUAI, Shanghai Jiaotong University and Microsoft Research Asia. It is specially used to evaluate the knowledge and reasoning ability of language models in Chinese context, covering a total of 67 topics, involving natural sciences, social sciences, engineering, humanities and common sense.

The financial industry has rich scenes and a high degree of digitization, which is one of the best scenarios for the landing application of large models. However, finance is also an industry with high compliance requirements, and the business decision-making is complex, so there are still many challenges for the real application of large models in financial business.

Xu Dongliang, CTO of du Xiaoman, said, "compared with the general model, the 'Xuanyuan 70B' financial model has a more professional, accurate, controllable and secure understanding of financial knowledge through the massive financial data training deposited in Lesser Fullness of Grain's business scenario. We open source the large model, hoping to reduce the threshold for the deployment and application of the big model in the financial industry, and promote a breakthrough in service efficiency and service experience in the financial industry."

In May this year, Lesser Fullness of Grain opened up China's first financial model with hundreds of billions of parameters, and hundreds of financial institutions have applied for trial. Compared with the previous version, the ability of the new version of the large model is comprehensively improved: in the incremental pre-training and instruction fine-tuning phase, a large amount of financial data from Lesser Fullness of Grain's business scenarios are added to respond to financial problems more professionally. the ability to understand financial knowledge has been significantly improved; at the same time, in the pre-training phase, the context length of the model "Regulatory 70B" has been expanded to 8k, which can handle longer financial reports, research and analysis.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.

Views: 0

*The comments in the above article only represent the author's personal views and do not represent the views and positions of this website. If you have more insights, please feel free to contribute and share.

Share To

IT Information

Wechat

© 2024 shulou.com SLNews company. All rights reserved.

12
Report