XLand-MiniGrid

History

2024: Product Announcement

On November 29, 2024, it became known that Russian scientists from the T-Bank AI Research laboratory and the AIRI Institute, in collaboration with students of MIPT, Skoltech and Innopolis, developed XLand-MiniGrid, the first open virtual environment for research in the field of contextual training with reinforcement of artificial intelligence.

According to TASS, Wednesday is published in the public domain for researchers around the world. Experiments at XLand-MiniGrid have already been conducted by specialists from Google DeepMind, the University of California, Berkeley and the University of Oxford.

𝓲

Freepik

The researcher of the AI Alignment scientific group Vyacheslav Blue emphasized that the work attracted the attention of the world scientific community, since the nascent field of contextual training with reinforcement lacked the necessary tools to evaluate new ideas.

The developed environment contains 100 billion examples of artificial intelligence actions in 30 thousand tasks, which allows you to use ready-made data sets for training instead of creating them from scratch. XLand-MiniGrid is based on JAX technology and is capable of billions of operations per second.

Vladislav Kurenkov, head of the Adaptive Agents scientific group at the AIRI Institute, noted that contextual reinforcement training allows you to create systems that can instantly adapt to new scenarios based on external feedback.

Unlike the closed-door developments of large companies, XLand-MiniGrid provides the ability to change learning conditions during the work, which simplifies the modeling of various tasks and helps create more adaptive models of artificial intelligence. Public access to the environment will allow researchers around the world to accelerate the development and testing of new machine learning algorithms.^[1]

Notes