Xiaomi AI launches the powerful MiLM-6B model: with 6.4 billion parameters, ranking 10th in the C-Eval list

According to news from this website on August 11, Xiaomi has not officially announced that it will directly join the AI ​​large model track. However, Xiaomi’s AI large model MiLM-6B has quietly appeared in C-Eval and CMMLU large model reviews. list, as of now, Xiaomi’s large model ranks 10th on the C-Eval overall list and 1st in the same parameter magnitude.

According to the information obtained by this site based on the GitHub project page, MiLM-6B is a huge-scale pre-trained language model developed by Xiaomi, with a parameter scale of 6.4 billion

According to the data provided by C-Eval, the MiLM-6B model has the highest performance in STEM (Science, Technology, Engineering) and mathematics education), showing a high accuracy rate in all 20 subjects, especially in multiple items such as metrology, physics, chemistry and biology

It is reported that the MiLM-6B model shows relatively good accuracy in most liberal arts subjects, but in subjects involving "abstract thinking" such as There is still room for improvement in Law, Mathematics, Programming, Probability Theory and Discrete Mathematics

In addition to "Education and Geography", the model achieved high accuracy in 8 other social science test subjects Rate. Among the 11 subjects of humanities, MiLM-6B performed well in the field of "History and Law"

MiLM-6B, a large AI model, was exposed for the first time and has a high degree of completion. In the future, Xiaomi's virtual assistant services, such as "Xiao Ai Classmates", may adopt this model

Articles on this site contain external jump links, which are designed to save selection time and provide more information, but the results for reference only

The above is the detailed content of Xiaomi AI launches the powerful MiLM-6B model: with 6.4 billion parameters, ranking 10th in the C-Eval list. For more information, please follow other related articles on the PHP Chinese website!