Breaking 1/ε Barrier in Quantum Zero-Sum Games: Generalizing Metric Subregularity for Spectraplexes
Abstract
Quantum zero-sum games provide a framework for non-local games, quantum interactive proofs, and quantum machine learning, where players optimize a bilinear payoff over quantum states. In contrast to classical bilinear games over polyhedral domains, for which gradient methods achieve linear last-iterate convergence, comparable guarantees over spectraplexes have remained open. Recent work achieved only an O(1/) average-iterate rate and suggested that semidefinite geometry may preclude classical-style linear rates. We refute this obstruction. We prove that quantum zero-sum games admit algorithms with O((1/)) last-iterate convergence to Nash equilibrium. In particular, matrix variants of Nesterov's iterative smoothing and Optimistic Gradient Descent--Ascent match the asymptotic rate of the classical polyhedral case. The key technical ingredient is a new error-bound theory for semidefinite games, establishing metric subregularity of the relevant monotone operator over spectrahedra despite the absence of polyhedral structure. We also give a geometric characterization of Nash equilibria via slack operators, classifying strategic directions as essential, neutral, or non-essential. Under strict complementarity or nondegeneracy, this reduces to a sharp classical-style dichotomy. Finally, we revisit Optimistic Matrix Multiplicative Weights Update. By extending the Quantal Response Equilibrium framework to spectraplex games, we prove an O(1/) last-iterate guarantee, while showing that any O((1/)) speedup for this method must depend on a natural, dimension-dependent condition number. Experiments support the theoretical picture, with Optimistic Gradient Descent--Ascent outperforming Optimistic Matrix Multiplicative Weights Update in the regimes studied.
Turn this paper into a lesson
ArcXiv compiles a structured reading guide from this paper's metadata: plain-English importance, contributions, prerequisite concepts, which sections to read first, flashcards, and a quiz. Grounded in the abstract, never invented.