Hamilton-Zero: A Neural Tensor-Network Foundation Model for Ground States of Arbitrary Quadratic Qubit Hamiltonians
Researchers have developed a neural tensor-network foundation model that can efficiently compute ground states of arbitrary quadratic qubit Hamiltonians. The model, called Hamilton-Zero, uses manifold variational optimisation to find the ground state, and is trained on a dataset of hundreds of thousands of different Hamiltonian systems. This allows for the computation of ground states for system sizes up to 8100 qubits, which is beyond the reach of classical simulation method
Researchers have developed a neural tensor-network foundation model that can efficiently compute ground states of arbitrary quadratic qubit Hamiltonians. The model, called Hamilton-Zero, uses manifold variational optimisation to find the ground state, and is trained on a dataset of hundreds of thousands of different Hamiltonian systems. This allows for the computation of ground states for system sizes up to 8100 qubits, which is beyond the reach of classical simulation methods.
---
Why it matters: This matters because it brings us closer to achieving quantum advantage in computing ground states of complex systems, a long-standing challenge in quantum computing.
Source: https://arxiv.org/abs/2608.11911
This article was originally published at: https://arxiv.org/abs/2608.11911