{
  "model": "Hsheet",
  "lattice": "square",
  "n_sites": 4,
  "boundary": "O",
  "params": {
    "N_up": 2,
    "N_dn": 2,
    "R": 1,
    "basis": "none"
  },
  "instance_id": "Hsheet/square_4_O_R1_none",
  "rows": [
    {
      "energy": -1.439,
      "sigma": null,
      "energy_variance": null,
      "dof": 4,
      "einf": null,
      "v_score": null,
      "method": "FCI",
      "ansatz": null,
      "method_detail": "full CI; cc-pVXZ in PySCF, Appendix D/E; correlation energy from X = 3, 4; extrapolated: 1/M → 0 (complete basis)",
      "method_as_published": "FCI [full CI; cc-pVXZ in PySCF, Appendix D/E; correlation energy from X = 3, 4; extrapolated: complete basis set]",
      "family": "other",
      "bound_type": "extrapolated",
      "bound_type_reason": "extrapolated to the complete basis set or to zero truncation error; not a bound (RULES.md 13.4; assigned during source reading)",
      "reference": "J. Lin, G. Goldshlager & L. Lin, Explicitly antisymmetrized neural network layers for variational Monte Carlo simulation, J. Comput. Phys. 474, 111765 (2023), arXiv:2112.03491, doi:10.1016/j.jcp.2022.111765",
      "peer_reviewed": true,
      "source": "track3-2026-11-12",
      "provenance": "primary",
      "verified": {
        "checked_on": "2026-11-12",
        "method": "source text of arXiv:2112.03491 read locally (pypdf layout text; the generator finds every value again on its line), no LLM transcription of numbers; single reading",
        "reported_as": "-1.4390 (Table 2, a = 1.0, column FCI; total energy)",
        "note": "J. Comput. Phys. 474, 111765 Table 2 (body not in the journal accessibility text); arXiv v1 Table II, text line 1010 | Hartree, total energy of the square H4 of side a = 1.0 bohr (Fig. 9: \"The side length is a\"), stored per atom (x 1/4); neutral, N_up = N_dn = 2; the paper estimates the extrapolation error at a few 1e-4 Ha | version read: arXiv v1: the journal table body is missing from the Safari accessibility text",
        "secondary_of": null
      }
    },
    {
      "energy": -1.424518,
      "sigma": 0.000007,
      "energy_variance": null,
      "dof": 4,
      "einf": null,
      "v_score": null,
      "method": "VMC",
      "ansatz": "FermiNet-FA",
      "method_detail": "factorized antisymmetric layer of rank 1 (FA-1) with backflow-based Jastrow, VMCNet code, KFAC",
      "method_as_published": "FermiNet-FA [VMC, factorized antisymmetric layer of rank 1 (FA-1) with backflow-based Jastrow, VMCNet code, KFAC]",
      "family": "other NQS",
      "bound_type": "variational",
      "bound_type_reason": "the energy of a stated wave function (Hartree-Fock, DMRG at a stated bond dimension, MRCI, VMC, a neural-network state); an upper bound in the instance's basis (RULES.md 13.4; assigned during source reading)",
      "reference": "J. Lin, G. Goldshlager & L. Lin, Explicitly antisymmetrized neural network layers for variational Monte Carlo simulation, J. Comput. Phys. 474, 111765 (2023), arXiv:2112.03491, doi:10.1016/j.jcp.2022.111765",
      "peer_reviewed": true,
      "source": "track3-2026-11-12",
      "provenance": "primary",
      "verified": {
        "checked_on": "2026-11-12",
        "method": "source text of arXiv:2112.03491 read locally (pypdf layout text; the generator finds every value again on its line), no LLM transcription of numbers; single reading",
        "reported_as": "-1.424518(7) (Table 2, a = 1.0, column FA-1; total energy)",
        "note": "J. Comput. Phys. 474, 111765 Table 2 (body not in the journal accessibility text); arXiv v1 Table II, text line 1010 | Hartree, total energy of the square H4 of side a = 1.0 bohr (Fig. 9: \"The side length is a\"), stored per atom (x 1/4); neutral, N_up = N_dn = 2 | version read: arXiv v1: the journal table body is missing from the Safari accessibility text",
        "secondary_of": null
      }
    },
    {
      "energy": -1.438799,
      "sigma": 0.000004,
      "energy_variance": null,
      "dof": 4,
      "einf": null,
      "v_score": null,
      "method": "VMC",
      "ansatz": "FermiNet-GA",
      "method_detail": "generic antisymmetric layer (GA), VMCNet code, KFAC",
      "method_as_published": "FermiNet-GA [VMC, generic antisymmetric layer (GA), VMCNet code, KFAC]",
      "family": "other NQS",
      "bound_type": "variational",
      "bound_type_reason": "the energy of a stated wave function (Hartree-Fock, DMRG at a stated bond dimension, MRCI, VMC, a neural-network state); an upper bound in the instance's basis (RULES.md 13.4; assigned during source reading)",
      "reference": "J. Lin, G. Goldshlager & L. Lin, Explicitly antisymmetrized neural network layers for variational Monte Carlo simulation, J. Comput. Phys. 474, 111765 (2023), arXiv:2112.03491, doi:10.1016/j.jcp.2022.111765",
      "peer_reviewed": true,
      "source": "track3-2026-11-12",
      "provenance": "primary",
      "verified": {
        "checked_on": "2026-11-12",
        "method": "source text of arXiv:2112.03491 read locally (pypdf layout text; the generator finds every value again on its line), no LLM transcription of numbers; single reading",
        "reported_as": "-1.438799(4) (Table 2, a = 1.0, column GA; total energy)",
        "note": "J. Comput. Phys. 474, 111765 Table 2 (body not in the journal accessibility text); arXiv v1 Table II, text line 1014 | Hartree, total energy of the square H4 of side a = 1.0 bohr (Fig. 9: \"The side length is a\"), stored per atom (x 1/4); neutral, N_up = N_dn = 2 | version read: arXiv v1: the journal table body is missing from the Safari accessibility text",
        "secondary_of": null
      }
    },
    {
      "energy": -1.424429,
      "sigma": 0.000007,
      "energy_variance": null,
      "dof": 4,
      "einf": null,
      "v_score": null,
      "method": "VMC",
      "ansatz": "FermiNet",
      "method_detail": "1 determinant, FermiNet repository (JAX branch)",
      "method_as_published": "FermiNet [VMC, 1 determinant, FermiNet repository (JAX branch)]",
      "family": "other NQS",
      "bound_type": "variational",
      "bound_type_reason": "the energy of a stated wave function (Hartree-Fock, DMRG at a stated bond dimension, MRCI, VMC, a neural-network state); an upper bound in the instance's basis (RULES.md 13.4; assigned during source reading)",
      "reference": "J. Lin, G. Goldshlager & L. Lin, Explicitly antisymmetrized neural network layers for variational Monte Carlo simulation, J. Comput. Phys. 474, 111765 (2023), arXiv:2112.03491, doi:10.1016/j.jcp.2022.111765",
      "peer_reviewed": true,
      "source": "track3-2026-11-12",
      "provenance": "primary",
      "verified": {
        "checked_on": "2026-11-12",
        "method": "source text of arXiv:2112.03491 read locally (pypdf layout text; the generator finds every value again on its line), no LLM transcription of numbers; single reading",
        "reported_as": "-1.424429(7) (Table 5, FermiNet repository, 1 det; total energy)",
        "note": "J. Comput. Phys. 474, 111765 Table 5 (body not in the journal accessibility text); arXiv v1 Table V, text line 1412 | Hartree, total energy, stored per atom (x 1/4). Table 5 does not state a; its VMCNet entries repeat the a = 1.0 values of Table 2 (-1.424531, -1.438804), so a = 1.0 bohr | version read: arXiv v1: the journal table body is missing from the Safari accessibility text | reader confidence medium",
        "secondary_of": null
      }
    },
    {
      "energy": -1.438796,
      "sigma": 0.000005,
      "energy_variance": null,
      "dof": 4,
      "einf": null,
      "v_score": null,
      "method": "VMC",
      "ansatz": "FermiNet",
      "method_detail": "16 determinants, VMCNet code, KFAC",
      "method_as_published": "FermiNet [VMC, 16 determinants, VMCNet code, KFAC]",
      "family": "other NQS",
      "bound_type": "variational",
      "bound_type_reason": "the energy of a stated wave function (Hartree-Fock, DMRG at a stated bond dimension, MRCI, VMC, a neural-network state); an upper bound in the instance's basis (RULES.md 13.4; assigned during source reading)",
      "reference": "J. Lin, G. Goldshlager & L. Lin, Explicitly antisymmetrized neural network layers for variational Monte Carlo simulation, J. Comput. Phys. 474, 111765 (2023), arXiv:2112.03491, doi:10.1016/j.jcp.2022.111765",
      "peer_reviewed": true,
      "source": "track3-2026-11-12",
      "provenance": "primary",
      "verified": {
        "checked_on": "2026-11-12",
        "method": "source text of arXiv:2112.03491 read locally (pypdf layout text; the generator finds every value again on its line), no LLM transcription of numbers; single reading",
        "reported_as": "-1.438796(5) (Table 2, a = 1.0, column FermiNet 16 dets; total energy)",
        "note": "J. Comput. Phys. 474, 111765 Table 2 (body not in the journal accessibility text); arXiv v1 Table II, text line 1014 | Hartree, total energy of the square H4 of side a = 1.0 bohr (Fig. 9: \"The side length is a\"), stored per atom (x 1/4); neutral, N_up = N_dn = 2 | version read: arXiv v1: the journal table body is missing from the Safari accessibility text",
        "secondary_of": null
      }
    },
    {
      "energy": -1.438796,
      "sigma": 0.000005,
      "energy_variance": null,
      "dof": 4,
      "einf": null,
      "v_score": null,
      "method": "VMC",
      "ansatz": "FermiNet",
      "method_detail": "2 determinants, FermiNet repository (JAX branch)",
      "method_as_published": "FermiNet [VMC, 2 determinants, FermiNet repository (JAX branch)]",
      "family": "other NQS",
      "bound_type": "variational",
      "bound_type_reason": "the energy of a stated wave function (Hartree-Fock, DMRG at a stated bond dimension, MRCI, VMC, a neural-network state); an upper bound in the instance's basis (RULES.md 13.4; assigned during source reading)",
      "reference": "J. Lin, G. Goldshlager & L. Lin, Explicitly antisymmetrized neural network layers for variational Monte Carlo simulation, J. Comput. Phys. 474, 111765 (2023), arXiv:2112.03491, doi:10.1016/j.jcp.2022.111765",
      "peer_reviewed": true,
      "source": "track3-2026-11-12",
      "provenance": "primary",
      "verified": {
        "checked_on": "2026-11-12",
        "method": "source text of arXiv:2112.03491 read locally (pypdf layout text; the generator finds every value again on its line), no LLM transcription of numbers; single reading",
        "reported_as": "-1.438796(5) (Table 5, FermiNet repository, 2 det; total energy)",
        "note": "J. Comput. Phys. 474, 111765 Table 5 (body not in the journal accessibility text); arXiv v1 Table V, text line 1014 | Hartree, total energy, stored per atom (x 1/4). Table 5 does not state a; its VMCNet entries repeat the a = 1.0 values of Table 2 (-1.424531, -1.438804), so a = 1.0 bohr. Same digits as Table 2's VMCNet 16-determinant entry, a different run | version read: arXiv v1: the journal table body is missing from the Safari accessibility text | reader confidence medium",
        "secondary_of": null
      }
    },
    {
      "energy": -1.438804,
      "sigma": 0.000005,
      "energy_variance": null,
      "dof": 4,
      "einf": null,
      "v_score": null,
      "method": "VMC",
      "ansatz": "FermiNet",
      "method_detail": "2 determinants, VMCNet code, KFAC",
      "method_as_published": "FermiNet [VMC, 2 determinants, VMCNet code, KFAC]",
      "family": "other NQS",
      "bound_type": "variational",
      "bound_type_reason": "the energy of a stated wave function (Hartree-Fock, DMRG at a stated bond dimension, MRCI, VMC, a neural-network state); an upper bound in the instance's basis (RULES.md 13.4; assigned during source reading)",
      "reference": "J. Lin, G. Goldshlager & L. Lin, Explicitly antisymmetrized neural network layers for variational Monte Carlo simulation, J. Comput. Phys. 474, 111765 (2023), arXiv:2112.03491, doi:10.1016/j.jcp.2022.111765",
      "peer_reviewed": true,
      "source": "track3-2026-11-12",
      "provenance": "primary",
      "verified": {
        "checked_on": "2026-11-12",
        "method": "source text of arXiv:2112.03491 read locally (pypdf layout text; the generator finds every value again on its line), no LLM transcription of numbers; single reading",
        "reported_as": "-1.438804(5) (Table 2, a = 1.0, column FermiNet 2 dets; total energy)",
        "note": "J. Comput. Phys. 474, 111765 Table 2 (body not in the journal accessibility text); arXiv v1 Table II, text line 1014 | Hartree, total energy of the square H4 of side a = 1.0 bohr (Fig. 9: \"The side length is a\"), stored per atom (x 1/4); neutral, N_up = N_dn = 2 | version read: arXiv v1: the journal table body is missing from the Safari accessibility text",
        "secondary_of": null
      }
    },
    {
      "energy": -1.438805,
      "sigma": 0.000004,
      "energy_variance": null,
      "dof": 4,
      "einf": null,
      "v_score": null,
      "method": "VMC",
      "ansatz": "FermiNet",
      "method_detail": "one full determinant (orbitals depend on all electrons, no spin block structure), VMCNet code, KFAC",
      "method_as_published": "FermiNet [VMC, one full determinant (orbitals depend on all electrons, no spin block structure), VMCNet code, KFAC]",
      "family": "other NQS",
      "bound_type": "variational",
      "bound_type_reason": "the energy of a stated wave function (Hartree-Fock, DMRG at a stated bond dimension, MRCI, VMC, a neural-network state); an upper bound in the instance's basis (RULES.md 13.4; assigned during source reading)",
      "reference": "J. Lin, G. Goldshlager & L. Lin, Explicitly antisymmetrized neural network layers for variational Monte Carlo simulation, J. Comput. Phys. 474, 111765 (2023), arXiv:2112.03491, doi:10.1016/j.jcp.2022.111765",
      "peer_reviewed": true,
      "source": "track3-2026-11-12",
      "provenance": "primary",
      "verified": {
        "checked_on": "2026-11-12",
        "method": "source text of arXiv:2112.03491 read locally (pypdf layout text; the generator finds every value again on its line), no LLM transcription of numbers; single reading",
        "reported_as": "-1.438805(4) (Table 2, a = 1.0, column FermiNet 1 full det; total energy)",
        "note": "J. Comput. Phys. 474, 111765 Table 2 (body not in the journal accessibility text); arXiv v1 Table II, text line 1014 | Hartree, total energy of the square H4 of side a = 1.0 bohr (Fig. 9: \"The side length is a\"), stored per atom (x 1/4); neutral, N_up = N_dn = 2 | version read: arXiv v1: the journal table body is missing from the Safari accessibility text",
        "secondary_of": null
      }
    },
    {
      "energy": -1.424531,
      "sigma": 0.000006,
      "energy_variance": null,
      "dof": 4,
      "einf": null,
      "v_score": null,
      "method": "VMC",
      "ansatz": "FermiNet",
      "method_detail": "single determinant, VMCNet code, KFAC",
      "method_as_published": "FermiNet [VMC, single determinant, VMCNet code, KFAC]",
      "family": "other NQS",
      "bound_type": "variational",
      "bound_type_reason": "the energy of a stated wave function (Hartree-Fock, DMRG at a stated bond dimension, MRCI, VMC, a neural-network state); an upper bound in the instance's basis (RULES.md 13.4; assigned during source reading)",
      "reference": "J. Lin, G. Goldshlager & L. Lin, Explicitly antisymmetrized neural network layers for variational Monte Carlo simulation, J. Comput. Phys. 474, 111765 (2023), arXiv:2112.03491, doi:10.1016/j.jcp.2022.111765",
      "peer_reviewed": true,
      "source": "track3-2026-11-12",
      "provenance": "primary",
      "verified": {
        "checked_on": "2026-11-12",
        "method": "source text of arXiv:2112.03491 read locally (pypdf layout text; the generator finds every value again on its line), no LLM transcription of numbers; single reading",
        "reported_as": "-1.424531(6) (Table 2, a = 1.0, column FermiNet 1 det; total energy)",
        "note": "J. Comput. Phys. 474, 111765 Table 2 (body not in the journal accessibility text); arXiv v1 Table II, text line 1010 | Hartree, total energy of the square H4 of side a = 1.0 bohr (Fig. 9: \"The side length is a\"), stored per atom (x 1/4); neutral, N_up = N_dn = 2 | version read: arXiv v1: the journal table body is missing from the Safari accessibility text",
        "secondary_of": null
      }
    },
    {
      "energy": -1.2863,
      "sigma": null,
      "energy_variance": null,
      "dof": 4,
      "einf": null,
      "v_score": null,
      "method": "HF",
      "ansatz": null,
      "method_detail": "restricted, restricted Hartree-Fock; cc-pVXZ in PySCF, Appendix D/E; extrapolated: 1/M → 0 (complete basis)",
      "method_as_published": "RHF [restricted Hartree-Fock; cc-pVXZ in PySCF, Appendix D/E; extrapolated: complete basis set]",
      "family": "mean field",
      "bound_type": "extrapolated",
      "bound_type_reason": "extrapolated to the complete basis set or to zero truncation error; not a bound (RULES.md 13.4; assigned during source reading)",
      "reference": "J. Lin, G. Goldshlager & L. Lin, Explicitly antisymmetrized neural network layers for variational Monte Carlo simulation, J. Comput. Phys. 474, 111765 (2023), arXiv:2112.03491, doi:10.1016/j.jcp.2022.111765",
      "peer_reviewed": true,
      "source": "track3-2026-11-12",
      "provenance": "primary",
      "verified": {
        "checked_on": "2026-11-12",
        "method": "source text of arXiv:2112.03491 read locally (pypdf layout text; the generator finds every value again on its line), no LLM transcription of numbers; single reading",
        "reported_as": "-1.2863 (Table 2, a = 1.0, column RHF; total energy)",
        "note": "J. Comput. Phys. 474, 111765 Table 2 (body not in the journal accessibility text); arXiv v1 Table II, text line 1010 | Hartree, total energy of the square H4 of side a = 1.0 bohr (Fig. 9: \"The side length is a\"), stored per atom (x 1/4); neutral, N_up = N_dn = 2 | version read: arXiv v1: the journal table body is missing from the Safari accessibility text",
        "secondary_of": null
      }
    },
    {
      "energy": -1.3358,
      "sigma": null,
      "energy_variance": null,
      "dof": 4,
      "einf": null,
      "v_score": null,
      "method": "HF",
      "ansatz": null,
      "method_detail": "unrestricted, unrestricted Hartree-Fock (S_z = 0); cc-pVXZ in PySCF, Appendix D/E; extrapolated: 1/M → 0 (complete basis)",
      "method_as_published": "UHF [unrestricted Hartree-Fock (S_z = 0); cc-pVXZ in PySCF, Appendix D/E; extrapolated: complete basis set]",
      "family": "mean field",
      "bound_type": "extrapolated",
      "bound_type_reason": "extrapolated to the complete basis set or to zero truncation error; not a bound (RULES.md 13.4; assigned during source reading)",
      "reference": "J. Lin, G. Goldshlager & L. Lin, Explicitly antisymmetrized neural network layers for variational Monte Carlo simulation, J. Comput. Phys. 474, 111765 (2023), arXiv:2112.03491, doi:10.1016/j.jcp.2022.111765",
      "peer_reviewed": true,
      "source": "track3-2026-11-12",
      "provenance": "primary",
      "verified": {
        "checked_on": "2026-11-12",
        "method": "source text of arXiv:2112.03491 read locally (pypdf layout text; the generator finds every value again on its line), no LLM transcription of numbers; single reading",
        "reported_as": "-1.3358 (Table 2, a = 1.0, column UHF; total energy)",
        "note": "J. Comput. Phys. 474, 111765 Table 2 (body not in the journal accessibility text); arXiv v1 Table II, text line 1010 | Hartree, total energy of the square H4 of side a = 1.0 bohr (Fig. 9: \"The side length is a\"), stored per atom (x 1/4); neutral, N_up = N_dn = 2 | version read: arXiv v1: the journal table body is missing from the Safari accessibility text",
        "secondary_of": null
      }
    }
  ],
  "url": "https://qmbl.org/i/Hsheet/square_4_O_R1_none/",
  "per_site_divisor": 4,
  "per_site_label": "E/N (Ha)",
  "best": {
    "exact": null,
    "variational": {
      "energy": -1.438805,
      "sigma": 0.000004,
      "method": "VMC",
      "ansatz": "FermiNet",
      "method_detail": "one full determinant (orbitals depend on all electrons, no spin block structure), VMCNet code, KFAC",
      "reference": "J. Lin, G. Goldshlager & L. Lin, Explicitly antisymmetrized neural network layers for variational Monte Carlo simulation, J. Comput. Phys. 474, 111765 (2023), arXiv:2112.03491, doi:10.1016/j.jcp.2022.111765",
      "bound_type": "variational",
      "energy_per_site": -0.35970125
    },
    "projected": null,
    "extrapolated": {
      "energy": -1.439,
      "sigma": null,
      "method": "FCI",
      "ansatz": null,
      "method_detail": "full CI; cc-pVXZ in PySCF, Appendix D/E; correlation energy from X = 3, 4; extrapolated: 1/M → 0 (complete basis)",
      "reference": "J. Lin, G. Goldshlager & L. Lin, Explicitly antisymmetrized neural network layers for variational Monte Carlo simulation, J. Comput. Phys. 474, 111765 (2023), arXiv:2112.03491, doi:10.1016/j.jcp.2022.111765",
      "bound_type": "extrapolated",
      "energy_per_site": -0.35975
    },
    "approximate": null,
    "below_exact": []
  },
  "no_best_reason": null,
  "record": {
    "energy": -1.439,
    "sigma": null,
    "method": "FCI",
    "ansatz": null,
    "method_detail": "full CI; cc-pVXZ in PySCF, Appendix D/E; correlation energy from X = 3, 4; extrapolated: 1/M → 0 (complete basis)",
    "reference": "J. Lin, G. Goldshlager & L. Lin, Explicitly antisymmetrized neural network layers for variational Monte Carlo simulation, J. Comput. Phys. 474, 111765 (2023), arXiv:2112.03491, doi:10.1016/j.jcp.2022.111765",
    "bound_type": "extrapolated",
    "energy_per_site": -0.35975
  },
  "no_record_reason": null
}
