Although Xiaomi has not officially announced its participation in the AI large model competition, its AI large model MiLM-6B has already appeared on the C-Eval and CMMLU large model evaluation lists. Currently, Xiaomi’s large model ranks 10th on the C-Eval list and 1st in the same parameter category
According to the information on the GitHub project page, IT House learned that MiLM-6B is a large-scale pre-trained language model developed by Xiaomi with a parameter scale of 6.4 billion
The rewritten sentence is as follows: The image source of MiLM-6B is shown on the GitHub page
According to data provided by C-Eval, the MiLM-6B model has shown high accuracy in all 20 subjects of STEM education, including metrology, physics, chemistry, biology, etc.
Please provide the original words so I can rewrite the sentence for you
According to reports, the MiLM-6B model performs relatively accurately in most liberal arts subjects, but there is still room for improvement in subjects involving abstract thinking, such as law, mathematics, programming, probability theory, and discrete mathematics.
Except for "Education and Geography", the model achieved ideal accuracy in the other 8 social science test subjects. Among the 11 subjects in humanities, MiLM-6B performed well in "History and Law"
Xiaomi's virtual assistant service "Xiao Ai Classmate" is expected to use the highly complete AI large model MiLM-6B
in the futureThe above is the detailed content of Xiaomi first exposed the MiLM-6B AI large model with 6.4 billion parameters, which may be used in Xiaoai classmates. For more information, please follow other related articles on the PHP Chinese website!