Training through self-play
We train policies through repeated interaction and competition, using feedback to improve decisions over successive rounds of training.
Reinforcement learning · Adaptive agents
We build learning systems that turn interaction into capability. Our work combines reinforcement learning, memory, and self-play to train agents that improve through experience.
From training to competition
We trained and deployed a self-play agent that took first place in a generals.bot competition. The system combines recurrent reinforcement learning, a custom self-play pipeline, and an optimized CPU inference implementation.
We train policies through repeated interaction and competition, using feedback to improve decisions over successive rounds of training.
Our recurrent agent uses memory to make decisions from partial observations. This is a starting point for studying behavior over longer horizons.
We aim to make these learning methods reusable across tasks, with language agents and robotics as longer-term directions.