Tag
Researchers from MIT, Carnegie Mellon, NYU, and Stanford developed an AI that decisively defeats top human Stratego players, a game of hidden information, using efficient training methods far cheaper than prior approaches like DeepMind's. The system, published in Nature, generalizes to other strategic games and could aid real-world decision-making under uncertainty such as business negotiations and cybersecurity.
This paper introduces PokaiTrainer, an AI agent for competitive Pokémon VGC that scales belief-state search to handle simultaneous actions and stochastic outcomes, achieving a high Elo rating on live ladders.
FootsiesGym is an open-source environment for studying non-trivial two-player zero-sum imperfect-information games, built on the minimalist fighting game Footsies. It provides a vectorized simulator for efficient training and benchmarks several reinforcement learning algorithms.
This paper models takeover bidding as auction games with imperfect information and studies optimal due diligence strategies using game-theoretic solvers and reinforcement learning, finding that PPO and PPG are effective for large games, and providing a computable threshold for when additional diligence ceases to be beneficial.
EMAgnet introduces parameter-space exponential moving average regularization for policy gradient self-play in large two-player zero-sum games, achieving lower exploitability compared to uniform regularization targets.
MIT researchers co-authored a paper showing that general-purpose policy gradient algorithms can outperform specialized game-theoretic algorithms in imperfect-information games, challenging long-held assumptions in the field.
This paper presents a self-play reinforcement learning framework for the four-player imperfect-information card game Big 2, comparing policy-gradient and value-based methods and finding that PPO with entropy regularization outperforms others.
The author argues that poker is an underrated benchmark for AI agents because it tests reasoning under uncertainty, adaptation, and risk management, and describes an upcoming AI poker arena where builders can submit bots to compete.