arXiv subjects

Zekun He

Publications and source records attributed to Zekun He.

4 recordsLinked to original sources

arXiv2025

Classical reservoir approach for efficient molecular ground state preparation

Ground state preparation is a central application of quantum algorithms for electronic structure. We introduce the classical reservoir approach, a low cost variational ansatz tailored to near-term hardware, requiring only nearest-neighbor interactions on a machine with square-lattice connectivity. Unlike traditional methods built from the classically efficient Hartree Fock theory, our ansatz operates in localized molecular orbitals to study previously unexplored regions of the variational parameter space. Numerical benchmarks demonstrate chemical accuracy across diverse systems and bond lengths; notably, significantly reduced circuit depths are attainable when relaxed error thresholds (e.g., tens of E_h) are permissible. We benchmark the method on hydrogen chains, N_2, O_2, CO, BeH_2, and H_2O, the latter corresponding to an effective 24 qubit calculation.

quant-ph↗

arXiv2025

Efficient Fermi-Hubbard model ground-state preparation by coupling to a classical reservoir in the instantaneous-response limit

Preparing the ground state of the Fermi-Hubbard model is challenging, in part due to the exponentially large Hilbert space, which complicates efficiently finding a path from an initial state to the ground state using the variational principle. In this work, we propose an approach for ground state preparation of interacting models by involving a classical reservoir, simplified to the instantaneous-response limit, which can be described using a Hamiltonian formalism. The resulting time evolution operator consist of spin-adapted nearest-neighbor hopping and on-site interaction terms similar to those in the Hubbard model, without expanding the Hilbert space. We can engineer the coupling to rapidly drive the system from an initial product state to its interacting ground state by numerically minimizing the final state energy. This ansatz closely resembles the Hamiltonian variational ansatz, offering a fresh perspective on it.

quant-ph↗

arXiv2024

Hunyuan-Large: An Open-Source MoE Model with 52 Billion Activated Parameters by Tencent

In this paper, we introduce Hunyuan-Large, which is currently the largest open-source Transformer-based mixture of experts model, with a total of 389 billion parameters and 52 billion activation parameters, capable of handling up to 256K tokens. We conduct a thorough evaluation of Hunyuan-Large's superior performance across various benchmarks including language understanding and generation, logical reasoning, mathematical problem-solving, coding, long-context, and aggregated tasks, where it outperforms LLama3.1-70B and exhibits comparable performance when compared to the significantly larger LLama3.1-405B model. Key practice of Hunyuan-Large include large-scale synthetic data that is orders larger than in previous literature, a mixed expert routing strategy, a key-value cache compression technique, and an expert-specific learning rate strategy. Additionally, we also investigate the scaling laws and learning rate schedule of mixture of experts models, providing valuable insights and guidances for future model development and optimization. The code and checkpoints of Hunyuan-Large are released to facilitate future innovations and applications. Codes: https://github.com/Tencent/Hunyuan-Large Models: https://huggingface.co/tencent/Tencent-Hunyuan-Large

cs.CL↗

arXiv2024

Hidden local adiabatic ramp in the modulated time evolution and the quantum approximate optimization algorithm

Adiabatic state preparation provides an analytical solution for generating the ground state of a target Hamiltonian, starting from an easily prepared ground state of the initial Hamiltonian. While effective for time-dependent Hamiltonians with an energy gap to the first coupled excited state, the process becomes exceedingly slow as the gap narrows. Rather than strictly following the adiabatic theorem, a more robust approach allows controlled diabatic excitations during the evolution and numerically optimizes the path to eliminate these excitations by the end. In this work, this is achieved via modulated time evolution, using a time dependent oscillating field lambda(t) to modulate the Hamiltonian, in conjunction with a transverse field B(t) whose optimized shape closely resembles a local adiabatic ramp. Beyond modulated time evolution, the quantum approximate optimization algorithm (QAOA), which also employs a transverse field defined as beta(t) over gamma(t), exhibits a shape similar to the local adiabatic ramp. This resemblance offers a more intuitive and physically motivated way to understand the QAOA algorithm through the lens of time evolution.

quant-ph↗