A new neural network, "Menon," has been presented at Novosibirsk State University (NSU). The development can compete with ChatGPT and "GigaChat," according to the university's press service.
"Menon" is based on the Chinese architecture "Kwen" with adaptation for the Russian language and culture. A database of over 700,000 Russian-language tasks was used to train the model.
The problem book consists of more than 20 tasks aimed at testing various abilities of Russian-language neural networks, such as common sense, logical reasoning, extracting information from the text useful for answering a question, and so on. "Monsters" of artificial intelligence, such as GPT-4o, GigaChat, giant open neural networks with tens of billions of parameters, compete in solving these problems, and our neural network, only one and a half billion parameters in size, tens of times smaller than others, looks quite good there.
In the future, the developers plan to continue improving the model in applied and scientific fields. Among the applied tasks is the creation of an assistant for students and applicants. In the scientific direction, the team will focus on improving the learning mechanism and increasing the accuracy of the model when working with various types of data.
Earlier, www1.ru reported that Sber presented the Kandinsky Video 4.0 neural network.
Read materials on the topic:
Sber expands access to the GigaChat chatbot in Telegram for all users
From concept to video in a few minutes: Sber presented the beta version of Kandinsky Video 1.1
