What is AI Arena?
AI Arena is an Ethereum native game where players all over the world can design, train and battle NFT characters powered by artificial neural networks!
To learn more about the game, see
We are in the process of migrating to a platform fighting game, where the objective is to knock your opponent off of a platform. The switch will not happen immediately, so researchers will still be creating models for the traditional fighter version of the game. But once the switch happens, the state space will expand dramatically (given the additional mobility a platform fighting game provides), and the policies are expected to be much more diverse. In the traditional fighting game, we found that models were quickly converging on the same set of local optima. We expect the search space for viable policies to be much larger with a platform fighting game, and thus, expect a more diverse set of models.
What Role do Researchers Play?
AI Arena is structured as a global Meta Competition.
Researchers compete against each other to create the best learning algorithm in the Research Competition. Researchers train their models in a python environment that ArenaX Labs has created for this competition. Eventually, researchers will have an opportunity to sell their models into the Gaming Competition.
To learn more about the Meta Competition, see
How do Researchers Participate?
To get started, install our python package and head on over to our colab notebook to see some starter code on training your model in python.
pip install [package coming soon]
To learn more about the starter code, check out
To dig into the code for the environment, visit our repo: