BluffJAX: Adversarial Imperfect Information Games in JAX

2026-10-05

AI Summary

This study addresses the lack of efficient parallelization and mechanistic diversity in existing benchmarks for imperfect-information games in reinforcement learning. To this end, this work proposes an open-source adversarial imperfect-information game suite built upon the JAX framework. The suite introduces a variety of underexplored game variants and leverages GPU acceleration alongside vectorized environments to enable high-throughput simulations, achieving extreme scalability with hundreds of millions of samples per second on single- or multi-GPU configurations. Experimental results demonstrate that the proposed suite significantly outperforms existing libraries in computational efficiency. Furthermore, it establishes standardized baseline results for both reinforcement learning and game-solving algorithms, providing a robust and scalable platform for future research in imperfect-information settings.

Abstract

We introduce BluffJAX: an open-source suite of adversarial imperfect information games in JAX. We provide canonical implementations of games designed for high simulation throughputs and parallelization on GPU accelerators. Our suite consists of well-studied benchmarks such as Texas Hold'Em Poker and Kuhn Poker, as well as games that have not been previously studied in reinforcement learning research, such as Bluff, Stud Poker, and Kemps. We hope that implementing a variety of game mechanics and difficulties will introduce new challenges and foster novel research directions in game-theoretic methods for RL. We benchmark the throughput performance and memory usage of our environments in single and multi-GPU settings, demonstrating scaling of up to hundreds of millions of samples per second, and motivating the usage of BluffJAX over related GPU and CPU-based libraries. We benchmark reinforcement learning, tree search, and game-solving algorithms in JAX in order to provide users with baseline results and facilitate future comparisons.

Problem

• imperfect information games
• reinforcement learning
• JAX
• GPU acceleration
• benchmark environments

Innovation

• Imperfect Information Games
• JAX
• GPU Parallelization
• Reinforcement Learning
• High Throughput