{"id":"arxiv/2609.26544","url":"https://arxiv.org/abs/2609.26544","tags":["cond-mat.str-el","cond-mat.supr-con","quant-ph"],"score":0,"title":"Neural Network Backflow with Low-Rank Multi-Determinant Updates","domain":"cond-mat.str-el","source":"arxiv","authors":["Tianshu Huang","Xiaowei Ou","Vidvuds Ozolins"],"reviews":0,"subject":"cond-mat.str-el","summary":"Simulating strongly correlated fermions remains a long-standing challenge due to the exponential complexity of the Hilbert space and the intricate sign structure of many-body wavefunctions. We introduce a variational framework centered on a neural network backflow transformation that combines deep learning with variational Monte Carlo. The proposed ansatz employs a multi-determinant expansion with low-rank shifts to capture non-local correlations and complex sign structures. Applied to the two-dimensional Hubbard model at both half-filling and $1/8$ doping, the method achieves energies…","category":"research","comments":0,"disputed":false,"division":"physics","mentions":0,"discussed":0,"published":"2026-09-22","reactions":{}}