Game, Set, Quantum: Parameterized Quantum Circuit for Correlated Equilibrium in Bayesian Games

Abstract

Strategic decision-making among many agents under incomplete information is central to economics, security, and multi-agent artificial intelligence (AI). Computing equilibria in such settings is challenging because the joint type-action space grows exponentially with the number of players. In binary-type, binary-action Bayesian games, an explicit representation over type-action profiles requires O(22n) entries, making direct linear-programming (LP) formulations memory intensive at moderate player counts. We propose a hybrid quantum-classical framework for approximating Bayes correlated equilibrium using a parameterized quantum circuit (PQC). The PQC represents the conditional strategy distribution with O(nL) trainable parameters, where n is the number of players and L is the circuit depth; for the largest setting studied here, n = 10 and L = 2, this corresponds to 60 trainable angles. The circuit is trained by gradient-based regret minimization with a negative entropy regularizer and a curriculum schedule over player counts. On a poker-style Bayesian game with two to ten players, the proposed solver achieves lower mean clipped regret than MCCFR across all tested player counts and lower regret than DCFR up to eight players, while DCFR performs best at ten players. These results show that compact PQC parameterizations can provide a viable variational representation for approximate equilibrium computation, while highlighting the roles of ansatz expressivity, optimization strategy, and classical simulation cost.