AI and the World's Hard Problems · Quantum physics & quantum computing · arXiv · 2026-10-21
A General Framework for Constructing Local Hidden-state Models to Determine the Steerability
Not all entangled states exhibit quantum steering, and the main challenge in deciding steerability is verifying the existence of a local hidden-state (LHS) model that reproduces all post-measurement assemblages generated by arbitrary measurements. The authors propose a machine-learning framework that batch-samples measurements and uses gradient-based optimization to construct an optimal LHS model. For two-qubit Werner states it saturates the analytical visibility bounds under three Pauli measurements, arbitrary PVMs and arbitrary POVMs, and for two-qutrit isotropic states it reaches the known bounds under arbitrary PVMs, with results suggesting POVMs can offer an advantage over PVMs in revealing steerability.