QMBLthe quantum many-body leaderboard

All instances / Hubbard

Hubbard rectangular 4x8, U = 8, n = 0.875

Hubbard/rectangular-4x8_32_P_14_8

Record

-0.76690(1) E/site

HFPS (hidden-fermion Pfaffian state)

Chen et al. (2025) · lowest eligible strict variational bound (rules §6)

The Hamiltonian

H = −t ∑⟨ij⟩,σ (cc + h.c.) + U ∑i ni↑ni↓

t = 1, U = 8, N = N = 14 on 32 sites.

lattice
rectangular-4x8
sites
32
boundary
periodic
Nf
14
U
8
energies
3

Stored as a total energy in the convention of VarBench; quoted here as E/site, which is what the papers report (how the conversion works).

Every published energy

Grouped by what the number is. Ranking happens only inside the first group.

Strict variational upper bounds (3)

The record is the lowest eligible energy in this group, and only this group.

E/siteσVar(E)V-score methodsourceyear
-0.76690(1)record 1.0e-5 5.22e-1 2.7e-3 HFPS (hidden-fermion Pfaffian state)
peer reviewed
Chen et al. (2025) 2025
-0.7641(3) 3.0e-4 n/a n/a RHB (hierarchical backflow K = 2 + nonlocal FNN factor, Ndet = 5, Nneuron = 100)
preprint
Zhou, Zhou & Liu, Locality-Induc n/a
-0.76331(6) 6.3e-5 2.10e-2 1.1e-4 VMC Hidden Fermion Determinant State Ansatz (N_hidden = 16. Single hidden layer fully connected net with alpha = 14). Soft mean-field constraint for lambda = 8 stripe order. Moreno et al. (2022) 2022

How these numbers were read

Rows added after the VarBench import record where the number came from.

-0.76690(1) — HFPS (hidden-fermion Pfaffian state)

Checked 2026-09-15 · arXiv PDF text extracted locally with pypdf in layout mode, no LLM transcription

Reported as. -0.76690(1) per site, sigma^2/N_site = 0.016320

Table IV of arXiv:2507.10705, "Raw data of Fig. 7 in the L x 4 lattice with 1/8 doping and U = 8", row 8 x 4; Sec. III states Fig. 7 shows "L x 4 PBC lattices". 32 sites, 28 electrons, U = 8, periodic: this instance. The same table's 16 x 4 value -0.76413(3) is the HFPS row on rectangular-4x16_64_P_28_8, and its 4 x 4 value -0.74177(5) sits 3e-5 above the ED energy -0.7418 quoted by arXiv:2606.00924. New record, 0.47% below the VarBench HFDS row -0.7633125 and below the RHB -0.7641(3) of arXiv:2606.00924 on the same lattice. Variance: sigma^2/N_site stored as a total (x 32). Its V-score, 2.7e-3, is ~25x the VarBench HFDS row's 1.1e-4 at a higher energy; the HFDS variance is the likelier outlier, since on 4x16 the same VarBench ansatz has ten times the per-site variance and this paper puts HFDS 3e-4 above exact already at 4x4. Read from the arXiv version; the PNAS version was not compared (pnas.org refuses this sandbox).

-0.7641(3) — RHB (hierarchical backflow K = 2 + nonlocal FNN factor...

Checked 2026-09-15 · arXiv PDF text extracted locally with pypdf in layout mode, no LLM transcription

Reported as. -0.7641 per site, table-wide sampling error ~0.0003

Table IV of arXiv:2606.00924, "Energies per site for the Hubbard model at nh = 0.125, U = 8 for different systems under periodic boundary conditions. The energy sampling errors are around 0.0003", row 4 x 8, column RHB (Ndet = 5, Nneuron = 100, the setting adopted for all large-scale runs after Table III). Instance identity: the same row's HFDS column, -0.7633, is the VarBench HFDS row -0.7633125. sigma is the caption's table-wide "around 0.0003". Layout mode shifts this row one column left in sweep-tables.tsv; read here from the printed table. Internal check: the same ansatz gives -0.7410 on 4 x 4 against ED -0.7418 (Table III), above exact as it must be.

Coverage

Newest published number here: 2025. 2 rows added from the 2026-09 literature sweep, on top of the VarBench snapshot of 2024-10-22.

This instance as data

JSON · source file on GitHub · open an issue about this instance