Russian scientists have unveiled a new open environment, XLand-MiniGrid, designed for developing algorithms in contextual reinforcement learning, a distinct branch of AI research. This was reported by “Vedomosti” citing a statement from “T-Bank”.
The project involved researchers from T-Bank AI Research, the AIRI Institute, and students from three universities — MIPT, Skoltech, and Innopolis University.
The development targets areas such as robotics, autonomous transport systems management, and personalized recommendations.
The virtual environment is accessible worldwide. Researchers from Google DeepMind, the University of California, Berkeley, and the University of Oxford have already conducted experiments with it, according to the project authors.
“Our work immediately attracted the attention of other researchers in the field, and articles are already emerging where authors use the environment to test their methods,” commented Vyacheslav Siny, one of the participants in the project.
A scientific paper describing the project has been accepted at the international AI conference NeurIPS 2024 in Vancouver, Canada.
Earlier in December, a virtual AI-based assistant for psychologists was created in Russia, which doubles productivity, service quality, and revenue.