Introducing ⚔️ AI vs. AI ⚔️ a deep reinforcement learning multi-agents competition system

We’re excited to introduce a new tool we created: ⚔️ AI vs. AI ⚔️, a deep reinforcement learning multi-agents competition system.

This tool, hosted on Spaces, allows us to create multi-agent competitions. It is composed of three elements:

• A Space with a matchmaking algorithm that runs the model fights using a background task.
• A Dataset containing the results.
• A Leaderboard that gets the match history results and displays the models’ ELO.

Then, when a user pushes a trained model to the Hub, it gets evaluated and ranked against others. Thanks to that, we can evaluate your agents against other’s agents in a multi-agent setting.

In addition to being a useful tool for hosting multi-agent competitions, we think this tool can also be a robust evaluation technique in multi-agent settings. By playing against a lot of policies, your agents are evaluated against a wide range of behaviors. This should give you a good idea of the quality of your policy.

Let’s see how it works with our first competition host: SoccerTwos Challenge.

How does AI vs. AI works?

AI vs. AI is an open-source tool developed at Hugging Face to rank the strength of reinforcement learning models in a multi-agent setting.

The idea is to get a relative measure of skill rather than an objective one by making the models play against each other continuously and use the matches results to assess their performance compared to all the other models and consequently get a view of the quality of their policy without requiring classic metrics.

The more agents are submitted for a given task or environment, the more representative the rating becomes.

To generate a rating based on match results in a competitive environment, we decided to base the rankings on the ELO rating system.

Conversely, if a low-rated player wins in an upset against a high-rated player, it will cause a more significant effect on both of their ratings.

Here's how works the algorithm:

Our first AI vs. AI challenge experimentation: SoccerTwos Challenge ⚽

Citation