AI and the World's Hard Problems · Quantum physics & quantum computing · arXiv · 2026-10-26
Classical Pre-Training of Large Shallow Parametrized Quantum Circuits through Pauli Propagation and deployment on Quantum Hardware
Scaling parametrized quantum circuits is hard because parameter-shift gradients become too expensive and gradient-free optimizers offer limited convergence guarantees. The authors use Pauli propagation to move observables backward through the circuit, obtaining an explicit loss that can be minimized entirely on classical hardware, with truncation and pruning to keep it tractable. On 64-qubit VQE and binary-image generative benchmarks, the pretrained parameters were deployed on IBM's ibm_berlin processor, using fewer gates than IQP-style ansätze.