QEncode Suite v3 uses a fixed execution pipeline, certified-only eligibility, and transparent rank-based scoring so every result is reproducible and directly comparable.
1. PySCF CASCI — computes the active-space FCI reference energy using the exact CASCI Hamiltonian. This is the VQE target: all gaps are measured against E_CASCI, not full-system FCI.
2. Classical comparison — HF, MP2, CCSD, and CCSD(T) energies are computed for each molecule to establish the classical baseline. An entry is marked Beats Classical when its VQE gap is smaller than the CCSD(T) correlation energy.
3. Qubit Hamiltonian — PennyLane qchem.molecular_hamiltonian with Jordan-Wigner, Bravyi-Kitaev, or Parity mapping (via OpenFermion bridge for Parity).
4. Z2 symmetry tapering — reduces qubit count via qchem.taper(). The tapered HF state is computed with qchem.taper_hf().
5. VQE — COBYLA optimizer, 500 iterations, 5 random multi-starts. Ansatz: UCCSD (tapered excitations) or Hardware-Efficient (HEA). Circuit metrics (depth, 2Q gates, T-gate estimate) are recorded post-VQE.
Suite v3 match + trust_level = certified (gap < 0.01 Ha). Validated entries appear in the Research tab only.
Sort by lowest |E_VQE − E_CASCI| (Hartrees). Lower is better. Chemical accuracy threshold: 1.6 × 10⁻³ Ha.
Sort by fewest two-qubit gates (tie-break: circuit depth). Requires transpiled circuit metrics.
Rank-normalised combined score: 0.5 × (gap_rank / (N−1)) + 0.5 × (cost_rank / (N−1)) ∈ [0, 1]. Lower is better. Requires circuit metrics; N = certified entries with metrics.
All v3 entries are stored as signed JSON artifacts in releases/v3/db/ in the public GitHub repository.
Leaderboard CSVs are generated by scripts/export_leaderboard_v3.py directly from those entry files — no manual editing.
Each entry includes PySCF energies, VQE trajectory, circuit metrics, FCIDUMP integrals (base64), and a Ed25519 signature for tamper-detection.
