Onur Yararbas

Major: Computer Science
Mentors: Rimika Jaiswal, Professor Murphy Niu, Professor Leon Balents

Benchmarking ADAPT-VQE Against VQE with Kinetic Terms

Variational quantum eigensolvers (VQE) often suffer from energy plateaus that trap classical optimizers in suboptimal regions of the cost landscape. We investigate the integration of a kinetic rotation term, parametrized Pauli-Y gates inserted at each circuit depth, designed to reshape the landscape and mitigate local minima. Our work benchmarks this approach against ADAPT-VQE, a state-of-the-art algorithm that adaptively constructs the variational ansatz from a predefined operator pool. Using IBM’s Qiskit, we simulate the 1D transverse-field Ising model and evaluate performance via the energy ratio E(VQE)/E(true ground state) and fidelity via overlap with the exact ground state. Our results indicate that ADAPT-VQE requires a highly expressive operator pool to reach high fidelity. However, under these conditions it constructs significantly deeper circuits with more parameters than traditional VQE and VQE with a kinetic term. Our results suggest that kinetic terms effectively reduce plateau-induced stagnation and offer a scalable path to reliable ground-state preparation, especially for shallow or fixed-depth circuits.