Benchmarking¶

This guide describes the benchmarking framework for evaluating the backend selector.

Overview¶

The benchmarking suite evaluates performance using real-world quantum circuits from various application domains.

Benchmark Circuits¶

GHZ States¶

Test multi-qubit entanglement and connectivity.

# H(q0) → CX(q0,q1) → CX(q1,q2) → ...

Bernstein-Vazirani¶

Test parallel operations and phase kickback.

Mirror Circuits¶

Test device fidelity through circuit inversion. The circuit and its inverse should return to |0⟩.

LUCJ Ansatz¶

Chemistry ansatz for quantum chemistry applications. Reference: arXiv:2511.22476

QAOA¶

Quantum Approximate Optimization Algorithm for combinatorial optimization.

Mock Backends¶

The benchmarks use mock backends simulating real IBM processors:

Backend	Qubits	Topology	Generation
IBM Heron R1	133	Heavy-hex	2023
IBM Heron R2	156	Heavy-hex	2024
IBM Nighthawk	120	Square lattice	2025

Running Benchmarks¶

cd enhanced-quantum-backend-selector
poetry run python benchmarks/run_benchmarks.py

Results are saved to benchmarks/results/ with timestamps.

Custom Output Directory¶

poetry run python benchmarks/run_benchmarks.py --output /path/to/results

Result Format¶

{
  "benchmark_suite": "Enhanced Quantum Backend Selector",
  "timestamp": "2026-01-07T01:05:48.123456",
  "results": [
    {
      "name": "GHZ-5",
      "circuit_metrics": {
        "num_qubits": 5,
        "depth": 6,
        "num_nonlocal_gates": 4
      },
      "selection_time_seconds": 0.543,
      "rankings": [
        {
          "rank": 1,
          "backend": "ibm_nighthawk",
          "score": -0.007466,
          "error_percent": 0.747
        }
      ]
    }
  ]
}

Metrics Collected¶

Circuit Metrics:

num_qubits: Number of qubits
depth: Circuit depth
num_nonlocal_gates: Two-qubit gate count

Performance Metrics:

selection_time_seconds: Time to select backend
score: Backend score (negative error)
error_percent: Expected error percentage

Adding New Benchmarks¶

Add circuit function to benchmarks/circuits/benchmark_circuits.py:

def create_new_circuit(num_qubits: int) -> QuantumCircuit:
    qc = QuantumCircuit(num_qubits, name="NewCircuit")
    # Add gates
    qc.measure_all()
    return qc

Export in benchmarks/circuits/__init__.py
Add to suite in benchmarks/run_benchmarks.py

Validation¶

poetry run pytest tests/test_benchmarks.py -v