Abstract: We introduce a novel generalization of the Clauser-Horne-Shimony-Holt (CHSH) game to a multiplayer setting, i.e., Hypercube game, where all $m$ players are required to assign values to vertices on corresponding facets of an $m$-dimensional hypercube. The players win if and only if their answers satisfy both parity and consistency conditions. We completely characterize the maximum winning probabilities (game value) under classical, quantum and no-signalling strategies, respectively. In contrast to the original CHSH game designed to demonstrate the superiority of quantumness, we find that the quantum advantages in the Hypercube game significantly decrease as the number of players increase. Notably, the quantum value decays exponentially fast to the classical value as $m$ increases, while the no-signalling value always remains to be one.
Introduction
In quantum foundations, nonlocal games serve as a primary benchmark for demonstrating the non-classical nature of quantum mechanics. The most prominent example is the bipartite Clauser-Horne-Shimony-Holt (CHSH) game, where two players, Alice and Bob, can achieve a higher winning probability by sharing entanglement than is possible with any classical strategy. While the bipartite case is well-established, understanding how quantum advantages scale in multipartite (multiplayer) scenarios remains a significant challenge. This work explores this frontier by introducing the Hypercube game ($HC_m$), a scalable $m$-player generalization of the CHSH game based on the geometry of an $m$-dimensional hypercube.
Methods: The $m$-player Hypercube Game
The $HC_m$ game is defined by the geometric constraints of an $m$-dimensional hypercube. Each of the $m$ players is assigned a specific "facet" of the cube based on a bit ($0$ or $1$) provided by a referee. The players' task is to assign values ($+1$ or $-1$) to all vertices on their assigned facet.
To win, the players' assignments must satisfy two primary conditions:
- Parity: The product of assignments on a player's facet must meet a specific requirement (typically $1$, with a variation for the first player).
- Consistency: If two players' facets share a vertex, they must assign the same value to that vertex.
The study derives the maximum winning probabilities ($\omega$) for three strategy types: Classical (shared randomness), Quantum (shared entanglement, specifically GHZ states), and No-Signalling (the theoretical limit of correlations allowed by relativity).
Key Results and Insights
The analysis reveals a surprising trend: while quantum players always maintain an edge over classical ones, this advantage vanishes exponentially fast as the number of participants increases. The gap between the quantum and classical values narrows rapidly as $m$ increases.
The paper provides exact values for the maximum winning probability:
- Classical Value ($\omega_c$): $\frac{1}{2} + \frac{1}{2^m}$. As $m \to \infty$, this approaches $0.5$ (random guessing).
- Quantum Value ($\omega_q$): Approximates to $\omega_q \approx \frac{1}{2} + \frac{1}{2^m} + O\left(\frac{m-1}{4^m}\right)$.
- No-Signalling Value ($\omega_{ns}$): Always 1.0.
This suggests that as the hypercube's dimension increases, the density of consistency checks becomes so high that quantum entanglement cannot effectively overcome the constraints. In contrast, the persistent gap between quantum and no-signalling values highlights that quantum entanglement is a significantly "weaker" form of non-locality than what is theoretically permitted by the laws of relativity alone.
Discussion and Future Work
The exponential decay of quantum advantage observed here is a specific property of the Hypercube game's structure, differing from other multipartite games like the Mermin-GHZ game. This "reality check" on quantum scaling has implications for large-scale quantum networks. Future research directions include investigating the uniqueness of the optimal quantum strategy (rigidity), analyzing parallel repetition of the game, and applying these insights to enhance Device-Independent Quantum Key Distribution (DIQKD) protocols in multi-user environments.
Summary updated