Sota Kota and a team of researchers published a study in Nature describing a system that outperforms human experts in the board game Stratego. The AI utilizes reinforcement learning and test-time compute to navigate the game's imperfect information, which requires players to make tactical decisions without knowing the exact rank or position of their opponent's pieces.

The project was developed by Kota and other researchers, including those at OpenAI. The achievement marks a shift in fields previously dominated by human intuition, as the team's general techniques demonstrate that games with hidden information are increasingly susceptible to breakthroughs in reinforcement learning.