The adversarial gym for
systems that learn.

Trained adversaries that co-evolve with your system, finding failure modes in RL policies, multi-agent controllers, and autonomous agents before deployment does.