Xiaomi’s AI Large Model MiLM Officially Passes the Record

On May 16th, according to the official Weibo account of “Xiaomi Corporation”, Xiaomi‘s large language model MiLM has officially passed the filing for large models. The related model is said to be gradually applied in Xiaomi‘s cars, phones, smart home products, etc., and will be “opened up for experience by more users” in the future.

Xiaomi‘s MiLM large model made its first appearance on the C-Eval and CMMLU large model evaluation list in August last year. At that time, this model ranked 10th on the overall C-Eval list and 1st in terms of parameter level.

According to information provided by the GitHub project page, it is known that MiLM-6B is a large-scale pre-trained language model developed by Xiaomi with a parameter scale of 6.4 billion.

According to the information provided by C-Eval, the MiLM-6B model achieved high accuracy in various subjects such as accountancy, physics, chemistry, and biology among all 20 STEM (science, technology, engineering, and mathematics) subjects.

It is reported that the MiLM-6B model has shown relatively good accuracy in most humanities subjects, but there is still room for improvement in subjects involving ‘abstract thinking’ such as ‘law, mathematics, programming, probability theory, and discrete mathematics’.

In 10 ‘social science’ test subjects, except for ‘education and geography’, the model achieved satisfactory accuracy in the other 8 subjects. In the 11 humanities subjects, MiLM-6B showed good accuracy performance based on ‘history and law’.

SEE ALSO: Xiaomi Appoints Two Female Executives