Conceptexploiters
Exploiter
An auxiliary agent mechanism in self-play training that is specifically trained to find the weaknesses of the main agent. Serving as a sparring partner, it forces the main agent to patch vulnerabilities and improve robustness. It was introduced by DeepMind in the AlphaStar project.
Timeline (last 90 days)
Oct 29
For small-scale projects, fully reproducing AlphaStar-level training is unrealistic, but "periodically training an exploiter targeting the current policy and adding it to the league" is a simplified version that can be implemented on single-machine resources.
Unverified50%