Entanglement-Enhanced Learning

Entanglement is a powerful resource for quantum learning, yet rigorously characterizing its advantage can be challenging.We present theoretical results showing that entanglement yields an exponential reduction in sample complexity for specific learning tasks in both discrete-variable and continuous-variable settings [1,2]. We further analyze the robustness of these entanglement-assisted protocols under practical imperfections and identify operating regimes in which the advantage provably persists. Finally, we implement the protocols on superconducting qubits [3] and photonic modes [4], empirically demonstrating a provable quantum advantage of entanglement-enhanced learning.