Year: 2026
Type: preprint
Abstract: Neural Quantum State Architectures have been shown to have high accuracy on a variety of molecular and lattice systems, often outperforming traditional methods. This work replaces the all-to-all paradigm of current neural quantum states with a localized recurrent sequential backflow mechanism. Results have empirically exhibited approximately quadratic wall-clock scaling over the system sizes tested, namely the hydrogen chain with atom with length of H 10 to H 400 , suggesting a viable path towards a scalable Neural Quantum State.
Source: ChemRxiv
Author: Dylan Liu-Walter
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7215084074