High-Quality Thermal Gibbs Sampling with Quantum Annealing Hardware

Quantum Annealing (QA) was originally intended for accelerating the solution of combinatorial optimization tasks that have natural encodings as Ising models. However, recent experiments on QA hardware platforms demonstrated that, in operating regime corresponding to weak interactions, behaves like a noisy Gibbs sampler at hardware-specific effective temperature. This work builds those insights and identifies class small hardware-native models are robust noise effects proposes procedure executing these maximize sampling performance. Experimental results indicate proposed protocol high-quality samples from Furthermore, we show this temperature can be adjusted by modulating annealing time energy scale. The provides an approach using model presenting potential new opportunities applications machine learning physics simulation.