Can an AI agent autonomously design, run, and analyze quantum computing experiments? These are real results from circuits running on quantum hardware -- each one submitted, measured, and interpreted without human intervention.
We're testing the limits of today's noisy intermediate-scale quantum (NISQ) devices by running progressively harder experiments: from basic entanglement benchmarks to variational quantum chemistry. Every result here was produced by our autonomous experiment pipeline running against Quantum Inspire and IBM Quantum backends.
How noisy are current QPUs?
Bell and GHZ states give us a direct measure of gate fidelity and decoherence. Comparing emulator (perfect) vs hardware (noisy) reveals the gap.
Can we do useful chemistry on 2 qubits?
The VQE experiment calculates the ground state energy of H₂. Chemical accuracy (1.6 mHa) is the bar -- can a 9-qubit superconducting chip clear it?
Can AI agents run the lab?
Our daemon autonomously generates circuits, submits jobs, waits for results, and performs analysis. No human in the loop from queue to dashboard.
Experiments
96
Hardware runs
71
Emulator runs
25
Backends
13
tuna-9, qxelarator, ibm_torino, iqm-garnet, tuna9, ibm-torino, qxelarator (emulator), emulator, multi-source, ibm_marrakesh, emulator + noise model, pennylane (default.qubit + default.mixed), emulator (pennylane statevector + shot noise)
Experiment types
21
VQE Mitigation Ladder, Randomized Benchmarking, Quantum Volume, Detection Code, H₂ VQE, randomized_benchmarking, Readout Calibration, vqe_heh, GHZ State, cross2019_replication, Bell Calibration, QAOA MaxCut, peruzzo2014_replication, Repetition Code, Connectivity Probe, QRNG Certification, qaoa_maxcut_with_compilation, vqe_heh_plus_cross_platform, vqe_heh_plus, vqe_h2_symmetry_verification,
Best VQE error
0.09 kcal/mol
emulator
| Experiment | QI Emulator | QI Emulator | QI Emulator | QI Emulator | QI Emulator | IBM Marrakesh (156q) | IBM Torino (133q) | ibm-torino | iqm-garnet | Multi-Source | pennylane (default.qubit + default.mixed) | QI Tuna-9 (9q) | QI Tuna-9 (9q) |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Bell Calibration | 100.0% | -- | -- | -- | 100.0% | 99.0% | 98.2% | -- | -- | 97.0% | -- | 96.6% | -- |
GHZ State | -- | -- | -- | -- | 100.0% | 98.1% | 93.8% | -- | -- | -- | -- | 86.0% | -- |
H₂ VQE | -1.1296 Ha | -- | -- | -- | -1.1385 Ha | -1.0956 Ha | -1.1226 Ha | -1.0408 Ha | -1.1146 Ha | -- | -- | -1.1310 Ha | -- |
Randomized Benchmarking | -- | -- | -- | 99.95% | -- | -- | -- | -- | -- | -- | -- | 99.95% | -- |
QAOA MaxCut | 100% | -- | -- | 87% | -- | -- | -- | -- | -- | -- | -- | 74% | -- |
Quantum Volume | -- | -- | -- | QV 16 | -- | -- | -- | -- | -- | -- | -- | QV 16 | -- |
Connectivity Probe | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | 93.7% | -- |
Repetition Code | 100.0% | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | 93.2% | -- |
Detection Code | 100.0% | -- | -- | -- | -- | -- | -- | 92.7% | -- | -- | -- | 66.6% | -- |
Energy vs. bond distance for molecular hydrogen. The VQE emulator matches the exact (FCI) curve within chemical accuracy at all 14 distances. Purple diamonds show Tuna-9 hardware results with readout error mitigation (sector-projected 2-qubit ansatz, native CZ gate set). Error bars show ±1σ across 5 independent hardware runs.
How do emulator, IBM, and Tuna-9 compare on the same experiments?
Can a quantum computer calculate the energy of a hydrogen molecule? VQE uses a hybrid quantum-classical loop to find the ground state energy. The gold standard is "chemical accuracy" -- getting within 1.6 milliHartree of the exact answer.
ibm_torino -- 2/11/2026
Measured Energy
-1.1226 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.961
<Z1>
0.950
<Z0Z1>
-0.969
<X0X1>
-0.256
<Y0Y1>
-0.197
Measurement Counts by Basis
VQE energy: -1.1226 Ha (FCI: -1.1373 Ha). Error: 9.22 kcal/mol. Not chemical accuracy — hardware noise degrades Z correlations by ~3-5%.
ibm_torino -- 2/10/2026
Measured Energy
-1.1197 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.956
<Z1>
0.949
<Z0Z1>
-0.971
<X0X1>
-0.218
<Y0Y1>
-0.229
IBM Torino H2 VQE: -1.1197 Ha, 11.0 kcal/mol error. Hardware noise degrades result but correct sector and qualitative behavior observed.
iqm-garnet -- 2/10/2026
Measured Energy
-1.1146 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.949
<Z1>
0.949
<Z0Z1>
-1.000
<X0X1>
-0.206
<Y0Y1>
-0.196
Measurement Counts by Basis
tuna9 -- 2/10/2026
Measurement Counts by Basis
qxelarator (emulator) -- 2/10/2026
Measured Energy
-1.1385 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.973
<Z1>
0.973
<Z0Z1>
-1.000
<X0X1>
-0.252
<Y0Y1>
-0.219
Measurement Counts by Basis
VQE energy: -1.1385 Ha (FCI: -1.1373 Ha). Error: 0.75 kcal/mol. Within chemical accuracy.
version 3.0 qubit[2] q bit[2] b // Subspace-preserving ansatz: Ry-CNOT-X // State = cos(a/2)|10> + sin(a/2)|01> Ry(-0.223539) q[0] CNOT q[0], q[1] X q[0] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1310 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.963
<Z1>
0.963
<Z0Z1>
-1.000
<X0X1>
-0.209
<Y0Y1>
-0.200
Measurement Counts by Basis
VQE energy: -1.1310 Ha (FCI: -1.1373 Ha). Error: 3.9 kcal/mol. Outside chemical accuracy. Post-selection kept 83% of Z-basis shots. Hybrid(PS+REM): -1.1310 Ha (3.9 kcal/mol). Full-REM: -1.0751 Ha (39.0 kcal/mol).
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=1) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1293 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.964
<Z1>
0.964
<Z0Z1>
-1.000
<X0X1>
-0.202
<Y0Y1>
-0.186
Measurement Counts by Basis
VQE energy: -1.1293 Ha (FCI: -1.1373 Ha). Error: 5.0 kcal/mol. Outside chemical accuracy. Post-selection kept 91% of Z-basis shots. Hybrid(PS+REM): -1.1293 Ha (5.0 kcal/mol). Full-REM: -1.1277 Ha (6.0 kcal/mol).
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=1) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1358 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.969
<Z1>
0.969
<Z0Z1>
-1.000
<X0X1>
-0.210
<Y0Y1>
-0.207
Measurement Counts by Basis
VQE energy: -1.1358 Ha (FCI: -1.1373 Ha). Error: 0.9 kcal/mol. Within chemical accuracy. Post-selection kept 96% of Z-basis shots. Hybrid(PS+REM): -1.1358 Ha (0.9 kcal/mol). Full-REM: -1.1325 Ha (3.0 kcal/mol).
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=1) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1213 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.960
<Z1>
0.960
<Z0Z1>
-1.000
<X0X1>
-0.217
<Y0Y1>
-0.185
Measurement Counts by Basis
VQE energy: -1.1213 Ha (FCI: -1.1373 Ha). Error: 10.0 kcal/mol. Outside chemical accuracy. Post-selection kept 93% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=5) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1329 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.976
<Z1>
0.976
<Z0Z1>
-1.000
<X0X1>
-0.198
<Y0Y1>
-0.186
Measurement Counts by Basis
VQE energy: -1.1329 Ha (FCI: -1.1373 Ha). Error: 2.8 kcal/mol. Outside chemical accuracy. Post-selection kept 94% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=5) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1275 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.972
<Z1>
0.972
<Z0Z1>
-1.000
<X0X1>
-0.193
<Y0Y1>
-0.171
Measurement Counts by Basis
VQE energy: -1.1275 Ha (FCI: -1.1373 Ha). Error: 6.1 kcal/mol. Outside chemical accuracy. Post-selection kept 95% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=5) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1263 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.968
<Z1>
0.968
<Z0Z1>
-1.000
<X0X1>
-0.183
<Y0Y1>
-0.204
Measurement Counts by Basis
VQE energy: -1.1263 Ha (FCI: -1.1373 Ha). Error: 6.9 kcal/mol. Outside chemical accuracy. Post-selection kept 95% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=3) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1160 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.964
<Z1>
0.964
<Z0Z1>
-1.000
<X0X1>
-0.196
<Y0Y1>
-0.110
Measurement Counts by Basis
VQE energy: -1.1160 Ha (FCI: -1.1373 Ha). Error: 13.4 kcal/mol. Outside chemical accuracy. Post-selection kept 96% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=3) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1291 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.967
<Z1>
0.967
<Z0Z1>
-1.000
<X0X1>
-0.231
<Y0Y1>
-0.190
Measurement Counts by Basis
VQE energy: -1.1291 Ha (FCI: -1.1373 Ha). Error: 5.1 kcal/mol. Outside chemical accuracy. Post-selection kept 96% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=3) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1294 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.967
<Z1>
0.967
<Z0Z1>
-1.000
<X0X1>
-0.217
<Y0Y1>
-0.211
Measurement Counts by Basis
VQE energy: -1.1294 Ha (FCI: -1.1373 Ha). Error: 4.9 kcal/mol. Outside chemical accuracy. Post-selection kept 95% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=1) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1213 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.963
<Z1>
0.963
<Z0Z1>
-1.000
<X0X1>
-0.149
<Y0Y1>
-0.223
Measurement Counts by Basis
VQE energy: -1.1213 Ha (FCI: -1.1373 Ha). Error: 10.0 kcal/mol. Outside chemical accuracy. Post-selection kept 96% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=1) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1221 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.967
<Z1>
0.967
<Z0Z1>
-1.000
<X0X1>
-0.191
<Y0Y1>
-0.154
Measurement Counts by Basis
VQE energy: -1.1221 Ha (FCI: -1.1373 Ha). Error: 9.5 kcal/mol. Outside chemical accuracy. Post-selection kept 95% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=1) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1238 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.968
<Z1>
0.968
<Z0Z1>
-1.000
<X0X1>
-0.185
<Y0Y1>
-0.172
Measurement Counts by Basis
VQE energy: -1.1238 Ha (FCI: -1.1373 Ha). Error: 8.5 kcal/mol. Outside chemical accuracy. Post-selection kept 93% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=5) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1230 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.964
<Z1>
0.964
<Z0Z1>
-1.000
<X0X1>
-0.182
<Y0Y1>
-0.202
Measurement Counts by Basis
VQE energy: -1.1230 Ha (FCI: -1.1373 Ha). Error: 9.0 kcal/mol. Outside chemical accuracy. Post-selection kept 95% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=3) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] CNOT q[2], q[4] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1272 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.967
<Z1>
0.967
<Z0Z1>
-1.000
<X0X1>
-0.215
<Y0Y1>
-0.189
Measurement Counts by Basis
VQE energy: -1.1272 Ha (FCI: -1.1373 Ha). Error: 6.3 kcal/mol. Outside chemical accuracy. Post-selection kept 95% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=1) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
emulator -- 2/10/2026
Measured Energy
-1.1296 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.968
<Z1>
0.968
<Z0Z1>
-1.000
<X0X1>
-0.203
<Y0Y1>
-0.220
Measurement Counts by Basis
VQE energy: -1.1296 Ha (FCI: -1.1373 Ha). Error: 4.8 kcal/mol. Outside chemical accuracy. Post-selection kept 100% of Z-basis shots.
version 3.0 qubit[2] q bit[2] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=5) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[0] CNOT q[0], q[1] CNOT q[0], q[1] CNOT q[0], q[1] CNOT q[0], q[1] CNOT q[0], q[1] X q[0] // Z-basis measurement b = measure q
emulator -- 2/10/2026
Measured Energy
-1.1415 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.978
<Z1>
0.978
<Z0Z1>
-1.000
<X0X1>
-0.246
<Y0Y1>
-0.219
Measurement Counts by Basis
VQE energy: -1.1415 Ha (FCI: -1.1373 Ha). Error: 2.6 kcal/mol. Outside chemical accuracy. Post-selection kept 100% of Z-basis shots.
version 3.0 qubit[2] q bit[2] b // Subspace-preserving ansatz: Ry-CNOT-X (cnot_folds=3) // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[0] CNOT q[0], q[1] CNOT q[0], q[1] CNOT q[0], q[1] X q[0] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-0.9147 Ha
FCI Reference
-0.9361 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.205
<Z1>
0.205
<Z0Z1>
-1.000
<X0X1>
-0.914
<Y0Y1>
-0.892
Measurement Counts by Basis
VQE energy: -0.9147 Ha (FCI: -0.9361 Ha). Error: 13.4 kcal/mol. Outside chemical accuracy. Post-selection kept 96% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-1.380900 Ry(-1.380900) q[2] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1325 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.974
<Z1>
0.974
<Z0Z1>
-1.000
<X0X1>
-0.180
<Y0Y1>
-0.219
Measurement Counts by Basis
VQE energy: -1.1325 Ha (FCI: -1.1373 Ha). Error: 3.0 kcal/mol. Outside chemical accuracy. Post-selection kept 96% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223400 Ry(-0.223400) q[2] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
emulator -- 2/10/2026
Measured Energy
-0.9362 Ha
FCI Reference
-0.9361 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.183
<Z1>
0.183
<Z0Z1>
-1.000
<X0X1>
-0.983
<Y0Y1>
-0.984
Measurement Counts by Basis
emulator -- 2/10/2026
Measured Energy
-0.9454 Ha
FCI Reference
-0.9486 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.399
<Z1>
0.399
<Z0Z1>
-1.000
<X0X1>
-0.915
<Y0Y1>
-0.895
Measurement Counts by Basis
emulator -- 2/10/2026
Measured Energy
-0.9940 Ha
FCI Reference
-0.9981 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.742
<Z1>
0.742
<Z0Z1>
-1.000
<X0X1>
-0.646
<Y0Y1>
-0.659
Measurement Counts by Basis
emulator -- 2/10/2026
Measured Energy
-1.1034 Ha
FCI Reference
-1.1012 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.939
<Z1>
0.939
<Z0Z1>
-1.000
<X0X1>
-0.347
<Y0Y1>
-0.361
Measurement Counts by Basis
emulator -- 2/10/2026
Measured Energy
-1.1364 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.976
<Z1>
0.976
<Z0Z1>
-1.000
<X0X1>
-0.200
<Y0Y1>
-0.229
Measurement Counts by Basis
emulator -- 2/10/2026
Measured Energy
-1.0583 Ha
FCI Reference
-1.0552 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.992
<Z1>
0.992
<Z0Z1>
-1.000
<X0X1>
-0.149
<Y0Y1>
-0.147
Measurement Counts by Basis
iqm-garnet -- 2/10/2026
Measured Energy
-0.3105 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.928
<Z1>
-0.939
<Z0Z1>
0.934
<X0X1>
0.174
<Y0Y1>
-0.107
Measurement Counts by Basis
tuna-9 -- 2/10/2026
Measured Energy
-0.9210 Ha
FCI Reference
-0.9486 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.437
<Z1>
0.437
<Z0Z1>
-1.000
<X0X1>
-0.759
<Y0Y1>
-0.827
Measurement Counts by Basis
VQE energy: -0.9210 Ha (FCI: -0.9486 Ha). Error: 17.3 kcal/mol. Outside chemical accuracy. Post-selection kept 96% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-1.133300 Ry(-1.133300) q[2] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-0.9779 Ha
FCI Reference
-0.9981 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.711
<Z1>
0.711
<Z0Z1>
-1.000
<X0X1>
-0.624
<Y0Y1>
-0.611
Measurement Counts by Basis
VQE energy: -0.9779 Ha (FCI: -0.9981 Ha). Error: 12.7 kcal/mol. Outside chemical accuracy. Post-selection kept 96% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.726700 Ry(-0.726700) q[2] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.0946 Ha
FCI Reference
-1.1012 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.937
<Z1>
0.937
<Z0Z1>
-1.000
<X0X1>
-0.302
<Y0Y1>
-0.332
Measurement Counts by Basis
VQE energy: -1.0946 Ha (FCI: -1.1012 Ha). Error: 4.1 kcal/mol. Outside chemical accuracy. Post-selection kept 97% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.352200 Ry(-0.352200) q[2] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.0393 Ha
FCI Reference
-1.0552 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.978
<Z1>
0.978
<Z0Z1>
-1.000
<X0X1>
-0.143
<Y0Y1>
-0.122
Measurement Counts by Basis
VQE energy: -1.0393 Ha (FCI: -1.0552 Ha). Error: 10.0 kcal/mol. Outside chemical accuracy. Post-selection kept 96% of Z-basis shots.
version 3.0 qubit[5] q bit[5] b // Subspace-preserving ansatz: Ry-CNOT-X // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.143600 Ry(-0.143600) q[2] CNOT q[2], q[4] X q[2] // Z-basis measurement b = measure q
tuna-9 -- 2/10/2026
Measured Energy
-1.1222 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.972
<Z1>
0.972
<Z0Z1>
-1.000
<X0X1>
-0.170
<Y0Y1>
-0.132
Measurement Counts by Basis
VQE energy: -1.1222 Ha (FCI: -1.1373 Ha). Error: 9.5 kcal/mol. Outside chemical accuracy. Post-selection kept 83% of Z-basis shots.
version 3.0 qubit[2] q bit[2] b // Subspace-preserving ansatz: Ry-CNOT-X // State = cos(a/2)|10> + sin(a/2)|01>, alpha=-0.223500 Ry(-0.223500) q[0] CNOT q[0], q[1] X q[0] // Z-basis measurement b = measure q
ibm_torino -- 2/10/2026
Measured Energy
-1.0976 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.879
<Z1>
0.879
<Z0Z1>
-1.000
<X0X1>
-0.412
<Y0Y1>
-0.439
Measurement Counts by Basis
tuna-9 -- 2/10/2026
Measured Energy
-1.0045 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.723
<Z1>
0.970
<Z0Z1>
-0.709
<X0X1>
-0.060
<Y0Y1>
-0.049
Measurement Counts by Basis
VQE on Tuna-9 achieved -1.0045 Hartree for H2 at 0.735 Å. Error of 0.133 Ha (83.3 kcal/mol) reflects significant hardware noise — Tuna-9 Bell fidelity is 87.3% vs IBM's 99%. X/Y basis correlations are heavily degraded by noise.
ibm_marrakesh -- 2/10/2026
Measured Energy
-1.0956 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.875
<Z1>
0.882
<Z0Z1>
-0.966
<X0X1>
-0.402
<Y0Y1>
-0.426
Measurement Counts by Basis
VQE on IBM Marrakesh achieved -1.0956 Hartree for H2 at 0.735 Å. Error of 0.0417 Hartree (26.2 kcal/mol) exceeds chemical accuracy (1.6 mHartree). Circuit limitations from Z-basis only prevented recovery of X/Y correlations.
ibm-torino -- Invalid Date
Measured Energy
-1.0408 Ha
FCI Reference
-1.1373 Ha
Energy Level Diagram
Expectation Values
<Z0>
-0.854
<Z1>
0.854
<Z0Z1>
-1.000
<X0X1>
-0.155
<Y0Y1>
-0.154
Measurement Counts by Basis
tuna-9 -- Invalid Date
The simplest test of quantum entanglement: prepare two qubits in a Bell state, then measure. A perfect device gives 50/50 between |00⟩ and |11⟩ with nothing else. Any leakage into |01⟩ or |10⟩ reveals hardware noise.
tuna-9 -- 2/10/2026
Bell state fidelity: 96.6%. Excellent — 3.4% leakage into wrong parity states.
version 3.0 qubit[5] q bit[5] b H q[2] CNOT q[2], q[4] b = measure q
tuna-9 -- 2/10/2026
Bell state fidelity: 89.3%. Good — 10.7% leakage into wrong parity states.
version 3.0 qubit[2] q bit[2] b H q[0] CNOT q[0], q[1] b = measure q
emulator -- 2/10/2026
Bell state fidelity: 100.0%. Excellent — 0.0% leakage into wrong parity states.
This ran on a noiseless emulator. Hardware results will show real noise effects.
version 3.0 qubit[2] q bit[2] b H q[0] CNOT q[0], q[1] b = measure q
multi-source -- 2/10/2026
AI-designed cross-platform experiment reveals that quantum computing performance depends as much on the software stack as on hardware quality.
version 3.0 qubit[2] q bit[2] b H q[0] CNOT q[0], q[1] b = measure q
ibm_torino -- 2/10/2026
Bell state on ibm_torino via MCP: 98.2% fidelity. Slightly lower than ibm_marrakesh (99.0%), likely qubit-dependent.
tuna-9 -- 2/10/2026
Bell state fidelity: 87.3%. Significant noise on Tuna-9 hardware — 12.7% leakage into wrong parity states (|01⟩ and |10⟩).
version 3.0 qubit[2] q bit[2] b H q[0] CNOT q[0], q[1] b = measure q
ibm_marrakesh -- 2/10/2026
qxelarator (emulator) -- 2/10/2026
Bell state fidelity: 100.0%. Excellent — 0.0% leakage into wrong parity states.
This ran on a noiseless emulator. Hardware results will show real noise effects.
version 3.0 qubit[2] q bit[2] b H q[0] CNOT q[0], q[1] b = measure q
A harder entanglement test: create a 3-qubit GHZ state where all qubits are simultaneously |000⟩ and |111⟩. Parity violations (odd-parity states appearing) indicate decoherence scaling with qubit count.
tuna-9 -- 2/10/2026
5-qubit GHZ via Qiskit transpiler routing: 86.0% fidelity. Same qubit set as AI ([5,2,4,6,8] vs [2,4,5,6,8]) but valid CNOT chain. AI circuit used invalid pairs (q4-q5, q5-q6 not connected).
version 3.0 qubit[9] q bit[9] b // 5-qubit GHZ via Qiskit transpiler routing // CNOT chain: q5->q2->q4->q6->q8 H q[5] CNOT q[5], q[2] CNOT q[2], q[4] CNOT q[4], q[6] CNOT q[6], q[8] b = measure q
tuna-9 -- 2/10/2026
3-qubit GHZ fidelity: 87.9%. Even parity: 55.8%, Odd parity: 44.2%.
version 3.0 qubit[3] q bit[3] b H q[1] CNOT q[1], q[0] CNOT q[0], q[2] b = measure q
ibm_torino -- 2/10/2026
3-qubit GHZ on ibm_torino via MCP: 93.8% fidelity. Lower than ibm_marrakesh (98.1%), suggesting qubit-pair-dependent CNOT error rates. Still well above Tuna-9 (85.4%).
tuna-9 -- 2/10/2026
3-qubit GHZ fidelity: 86.6% on Tuna-9. Significant noise with 13.4% leakage into non-GHZ states. Asymmetry between |000⟩ (48.4%) and |111⟩ (38.2%) suggests qubit-dependent error rates.
version 3.0 qubit[3] q bit[3] b H q[0] CNOT q[0], q[1] CNOT q[0], q[2] b = measure q
ibm_marrakesh -- 2/10/2026
qxelarator (emulator) -- 2/10/2026
3-qubit GHZ fidelity: 100.0%. Even parity: 50.1%, Odd parity: 49.9%.
This ran on a noiseless emulator. Hardware results will show real noise effects.
version 3.0 qubit[3] q bit[3] b H q[0] CNOT q[0], q[1] CNOT q[1], q[2] b = measure q
How good is a single quantum gate? RB applies random sequences of Clifford gates and measures how quickly the signal decays. The decay rate gives the average error per gate -- the fundamental metric for gate quality.
tuna-9 -- 2/15/2026
Gate Fidelity
99.95%
Error per Gate
0.0005
Survival Probability Decay
Single-qubit RB on all 9 Tuna-9 qubits. Best: q7 (99.96%), Worst: q1 (98.64%), Mean: 99.55%. VQE qubits q4/q6 both >99.5%.
iqm-garnet -- 2/10/2026
tuna-9 -- 2/10/2026
Gate Fidelity
99.83%
Error per Gate
0.0017
Survival Probability Decay
1-qubit RB on qubit 2: gate fidelity 99.83%, error per gate 0.17%. Survival decays from 98.7% at m=1 to 94.1% at m=32. Comparable to q0 (99.82%).
version 3.0 qubit[3] q bit[3] b // RB sequence: m=1, seed=0 X q[2] S q[2] H q[2] // Inverse Clifford (index 14) Y q[2] H q[2] b = measure q
tuna-9 -- 2/10/2026
Gate Fidelity
99.69%
Error per Gate
0.0031
Survival Probability Decay
1-qubit RB: gate fidelity 99.69%, error per gate 0.0031. Good quality.
version 3.0 qubit[1] q bit[1] b // RB sequence: m=1, seed=0 X q[0] S q[0] H q[0] // Inverse Clifford (index 14) Y q[0] H q[0] b = measure q
qxelarator -- 2/10/2026
Gate Fidelity
99.95%
Error per Gate
0.0005
Survival Probability Decay
1-qubit RB: gate fidelity 99.95%, error per gate 0.0005. Excellent quality.
This ran on a noiseless emulator. Hardware results will show real noise effects.
Can a quantum algorithm beat random guessing at graph optimization? QAOA sweeps variational parameters to find the maximum cut of a triangle graph. The approximation ratio measures how close we get to the classical optimum.
tuna-9 -- 2/10/2026
Best Ratio
74.1%
Best gamma
0.50
Best beta
0.50
Approximation Ratio Heatmap
QAOA MaxCut (4 nodes, 3 edges): best approximation ratio 74.1% at gamma=0.50, beta=0.50. Classical optimum: 3 edges cut.
version 3.0 qubit[7] q bit[7] b // QAOA MaxCut: gamma=0.100, beta=0.100 // Initial superposition H q[5] H q[2] H q[4] H q[6] // Cost layer: exp(-i*gamma*C) via ZZ interactions CNOT q[5], q[2] Rz(0.200000) q[2] CNOT q[5], q[2] CNOT q[2], q[4] Rz(0.200000) q[4] CNOT q[2], q[4] CNOT q[4], q[6] Rz(0.200000) q[6] CNOT q[4], q[6] // Mixer layer: exp(-i*beta*B) via X rotations Rx(0.200000) q[5] Rx(0.200000) q[2] Rx(0.200000) q[4] Rx(0.200000) q[6] b = measure q
emulator -- 2/10/2026
Best Ratio
99.9%
Best gamma
Best beta
QAOA MaxCut emulator replication of Harrigan 2021. Tested 10 graphs at p=1-3. 3/3 claims reproduced.
This ran on a noiseless emulator. Hardware results will show real noise effects.
qxelarator -- 2/10/2026
Best Ratio
87.4%
Best gamma
1.00
Best beta
0.20
Approximation Ratio Heatmap
QAOA MaxCut on triangle: best approximation ratio 87.4% at gamma=1.00, beta=0.20. Classical optimum: 2 edges cut.
This ran on a noiseless emulator. Hardware results will show real noise effects.
tuna-9 -- Invalid Date
A holistic benchmark combining gate fidelity, connectivity, and compiler quality into a single number. QV tests whether the device can reliably execute random circuits of depth = width. Higher is better.
tuna-9 -- 2/15/2026
Quantum Volume
16
QV=16 CERTIFIED on Tuna-9 hardware. 100 circuits, mean HOF=0.757, 2σ lower=0.746 >> 2/3. 97/100 passed.
qxelarator -- 2/13/2026
Quantum Volume
16
QV=16 PASS on emulator. Mean HOF=0.8714, 2-sigma lower=0.8480 (> 2/3 threshold). 10/10 circuits passed individually.
This ran on a noiseless emulator. Hardware results will show real noise effects.
tuna-9 -- 2/10/2026
Quantum Volume
8
Quantum Volume 8. n=2: PASS (69.2%), n=3: PASS (82.1%)
iqm-garnet -- 2/10/2026
qxelarator -- 2/10/2026
Quantum Volume
8
Quantum Volume: 8. n=2: PASS (heavy=77.2%). n=3: PASS (heavy=85.1%)
This ran on a noiseless emulator. Hardware results will show real noise effects.
version 3.0 qubit[2] q bit[2] b // QV circuit: n=2, circuit=0 // Layer 0 Rz(4.424237) q[0] Ry(1.297664) q[0] Rz(5.410326) q[0] Rz(2.657316) q[1] Ry(1.793073) q[1] Rz(5.227973) q[1] CNOT q[0], q[1] Rz(0.074223) q[0] Ry(0.525476) q[0] Rz(4.448806) q[1] Ry(1.462129) q[1] // Layer 1 Rz(0.259822) q[0] Ry(1.096424) q[0] Rz(4.872983) q[0] Rz(3.333507) q[1] Ry(0.725563) q[1] Rz(5.769010) q[1] CNOT q[0], q[1] Rz(5.816017) q[0] Ry(0.317308) q[0] Rz(0.616970) q[1] Ry(1.715462) q[1] b = measure q
Are quantum random numbers truly random? We run 8 NIST SP 800-22 statistical tests against raw hardware output, von Neumann debiased output, and emulator output. Raw Tuna-9 bits show measurable bias; debiasing fixes it completely.
20,000 bits per source -- 2/10/2026
Tuna-9 (raw)
1/8
7 tests failed
Tuna-9 (debiased)
8/8
All tests passed
Emulator
8/8
All tests passed
| NIST Test | Tuna-9 (raw) | Tuna-9 (debiased) | Emulator |
|---|---|---|---|
| Frequency (Monobit) | FAILp=5.2e-8 | PASSp=0.092 | PASSp=0.365 |
| Block Frequency | FAILp=0.007 | PASSp=0.967 | PASSp=0.547 |
| Runs | FAILp=0.0e+0 | PASSp=0.521 | PASSp=0.577 |
| Longest Run of Ones | FAILp=6.1e-5 | PASSp=0.739 | PASSp=0.874 |
| Spectral (DFT) | PASSp=0.559 | PASSp=0.243 | PASSp=0.218 |
| Serial | FAILp=3.4e-7 | PASSp=0.200 | PASSp=0.575 |
| Approximate Entropy | FAILp=7.1e-4 | PASSp=0.252 | PASSp=0.981 |
| Cumulative Sums | FAILp=9.6e-8 | PASSp=0.157 | PASSp=0.240 |
Von Neumann debiasing transforms biased quantum hardware output into NIST-certified random numbers
Tuna-9's transmon qubits have a measurable bias toward |0⟩ (48.1% ones vs 50%). Von Neumann pair extraction eliminates this first-order bias at a 75% bit cost, achieving 8/8 NIST SP 800-22 test passes. This validates our QRNG MCP server's debiasing pipeline for production use.
version 3.0 qubit[8] q bit[8] b H q[0] H q[1] H q[2] H q[3] H q[4] H q[5] H q[6] H q[7] b = measure q
Map CNOT fidelity across all 36 qubit pairs on Tuna-9. The heatmap reveals which physical qubits are best-connected -- essential for choosing where to place error correction codes.
tuna-9 -- 2/10/2026
Avg Fidelity
93.7%
Pairs Tested
Best 5q Subgraph
[3, 1, 4, 2, 5]
Tuna-9 has sparse connectivity (6 connected pairs found in this Feb 10 probe). Best fidelity: q2-q4 at 96.3%. Worst: q0-q1 at 88.8%. NOTE: Full topology has 12 edges (not 6 connected pairs out of 36). Pairs 4-7 and 5-7 were offline during this measurement but confirmed active Feb 13.
36 Bell-pair circuits (H+CNOT on each qubit pair)
The simplest quantum error correction code: 3 data qubits + 2 syndrome qubits detect and correct single bit-flip errors. Syndrome accuracy measures how well the hardware extracts error information.
emulator -- 2/10/2026
Syndrome Accuracy
Logical Error Rate
0.0%
After majority-vote correction
Syndrome Results by Variant
3-qubit repetition code: avg syndrome accuracy 100.0%, logical error rate 0.0% (after majority-vote correction). Code is working — errors detected and corrected.
This ran on a noiseless emulator. Hardware results will show real noise effects.
6 repetition code variants (3 data + 2 syndrome qubits)
tuna-9 -- 2/10/2026
Syndrome Accuracy
Logical Error Rate
2.9%
After majority-vote correction
Syndrome Results by Variant
AI Decoder vs Lookup Table
NN Decoder (MLP)
97.6%
5-fold CV, 20,480 samples
Lookup Table
93.1%
Syndrome-based classification
+4.5% improvement from neural network decoder
3-qubit repetition code: avg syndrome accuracy 93.2%, logical error rate 2.9% (after majority-vote correction). Code is working — errors detected and corrected.
6 repetition code variants (3 data + 2 syndrome qubits)
Four data qubits with XXXX and ZZZZ stabilizers detect any single-qubit error (X, Z, or Y). A neural network decoder trained on hardware syndrome data outperforms simple lookup-table decoding -- haiqu in action.
tuna-9 -- 2/13/2026
Detection Rate
66.6%
Errors correctly flagged
False Positive Rate
30.9%
Clean shots flagged as error
Detection by Error Type
[[4,2,2]] error detection now works on Tuna-9 after discovering the full 12-edge topology (q4 has degree 4). Detection rate 66.6% with 30.9% false positives. Lower than IBM's 92.7%/14.0% due to deeper circuit (10 vs 6 CNOTs), but functionally operational. Post-selection improves raw fidelity from 49.4% to 66.3% (1.34x gain). The topology constraint is no longer binary (works/doesn't) but manifests as a circuit depth tax.
version 3.0
qubit[9] q
bit[5] b
// Encoding: GHZ on data qubits {q1,q2,q6,q7} via q4 as bus
H q[4]
CNOT q[4], q[1]
CNOT q[4], q[2]
CNOT q[4], q[6]
CNOT q[4], q[7]
// Disentangle ancilla
CNOT q[1], q[4]
// Z-syndrome extraction
CNOT q[1], q[4]
CNOT q[2], q[4]
CNOT q[6], q[4]
CNOT q[7], q[4]
// Measure
b[0] = measure q[1]
b[1] = measure q[2]
b[2] = measure q[6]
b[3] = measure q[7]
b[4] = measure q[4]ibm-torino -- 2/10/2026
Detection Rate
92.7%
Errors correctly flagged
False Positive Rate
14.0%
Clean shots flagged as error
AI Decoder vs Lookup Table
NN Decoder (MLP)
61.7%
5-fold CV, 53248 samples
Lookup Table
41.1%
Syndrome-based classification
Detection by Error Type
[[4,2,2]] detection code on IBM Torino: 92.7% error detection rate, 14.0% false positive rate. All 13 circuit variants compiled and executed successfully (transpiled depth 44-55, 24-26 CZ gates). X errors detected at 89.1% avg, Z errors at 94.0% avg, Y errors at 95.1% avg. Cross-platform: emulator 100%/0%, IBM Torino 92.7%/14.0%, Tuna-9 66.6%/30.9% (10 CNOTs via single degree-4 ancilla q4).
emulator -- 2/10/2026
Detection Rate
100.0%
Errors correctly flagged
False Positive Rate
0.0%
Clean shots flagged as error
AI Decoder vs Lookup Table
NN Decoder (MLP)
76.8%
5-fold CV, 53248 samples
Lookup Table
53.7%
Syndrome-based classification
Detection by Error Type
[[4,2,2]] detection code: 100.0% error detection rate, 0.0% false positive rate. Excellent detection performance. NN decoder: 76.8% accuracy vs lookup table: 100.0%.
This ran on a noiseless emulator. Hardware results will show real noise effects.
13 [[4,2,2]] detection code variants with GHZ codespace prep, XXXX+ZZZZ stabilizers
Systematic comparison of error mitigation techniques for H₂ VQE: raw, post-selection, TREX, dynamical decoupling, twirling, and ZNE. Ranked by error in kcal/mol. IBM TREX achieved chemical accuracy (0.22 kcal/mol).
tuna-9 -- 2/16/2026
Error mitigation ladder for LiH CASCI(2,2) VQE on Tuna-9 (4 qubits, 27 Pauli terms, 9 measurement circuits). REM+ZNE(quadratic) reduces error from 33.1 to 6.9 mHa (79% reduction) but does not reach chemical accuracy. Higher gate count (19 CZ per circuit at fold=1, 95 at fold=5) limits ZNE effectiveness compared to H2 (1 CZ).
tuna-9 -- 2/15/2026
Error mitigation ladder for H2 VQE on Tuna-9. REM+ZNE(quadratic) achieves 2.9 mHa avg, 3/7 at chemical accuracy. Best: R=0.7 at 0.4 mHa.
ibm_torino -- 2/10/2026
randomized_benchmarking -- tuna-9
{
"gate_fidelity": 0.9982,
"error_per_gate": 0.0018,
"alpha": 0.9963,
"survival_by_length": {
"1": {
"mean": 0.977,
"seeds": [
0.962,
0.982,
0.98,
0.979,
0.98
]
},
"8": {
"mean": 0.967,
"seeds": [
0.952,
0.966,
0.983
]
},
"32": {
"mean": 0.926,
"seeds": [
0.939,
0.913
]
}
},
"interpretation": "1-qubit RB: gate fidelity 99.82%, error per gate 0.18%"
}vqe_heh -- tuna9
{
"energy_raw": -2.790022,
"error_raw_kcal_mol": 35.24,
"energy_postselected": -2.836798,
"error_postselected_kcal_mol": 5.89,
"energy_rem": -2.831308,
"error_rem_kcal_mol": 9.34,
"energy_rem_ps": -2.839113,
"error_rem_ps_kcal_mol": 4.44,
"energy_hybrid_ps_rem": -2.838043,
"error_hybrid_ps_rem_kcal_mol": 5.11,
"chemical_accuracy": false,
"best_strategy": "rem_ps",
"postselection_keep_fraction": 0.959,
"expectation_values_raw": {
"Z0": 0.97314,
"Z1": -0.92041,
"Z0Z1": -0.91797,
"X0X1": -0.08105,
"Y0Y1": -0.12256
},
"expectation_values_postselected": {
"Z0": 0.98727,
"Z1": -0.98727,
"Z0Z1": -1,
"X0X1": -0.08105,
"Y0Y1": -0.12256
},
"calibration": {
"q46_00_accuracy": 0.9753,
"q46_11_accuracy": 0.9504,
"readout_errors": {
"q4_0to1": 0.0139,
"q4_1to0": 0.0225,
"q4_total": 0.0182,
"q6_0to1": 0.0107,
"q6_1to0": 0.0273,
"q6_total": 0.019
}
}
}cross2019_replication -- iqm-garnet
{
"quantum_volume": {
"n2": {
"mean_hof": 0.757,
"std_hof": 0.1428,
"per_circuit_hof": [
0.5928,
0.9424,
0.8232,
0.5859,
0.8408
],
"passes_threshold": true,
"threshold": 0.6667,
"source": "cross2019 replication run 1 (2026-02-10T18:47)"
},
"n3": {
"mean_hof": 0.635,
"std_hof": 0.2646,
"per_circuit_hof": [
0.2168,
0.6846,
0.4756,
0.8359,
0.9619
],
"passes_threshold": false,
"threshold": 0.6667,
"source": "cross2019 replication run 1 (2026-02-10T18:47)"
}
},
"randomized_benchmarking": {
"sequence_lengths": [
1,
4,
8,
16,
32
],
"data": {
"1": {
"mean_survival": 0.9894,
"std_survival": 0.005,
"per_seed": [
0.9902,
0.9854,
0.998,
0.9902,
0.9834
]
},
"4": {
"mean_survival": 0.9793,
"std_survival": 0.0076,
"per_seed": [
0.9844,
0.9893,
0.9785,
0.9775,
0.9668
]
},
"8": {
"mean_survival": 0.9637,
"std_survival": 0.0078,
"per_seed": [
0.9727,
0.959,
0.9531,
0.9727,
0.9609
]
},
"16": {
"mean_survival": 0.9434,
"std_survival": 0.0224,
"per_seed": [
0.9531,
0.958,
0.8994,
0.9473,
0.959
]
},
"32": {
"mean_survival": 0.8823,
"std_survival": 0.0025,
"per_seed": [
0.8799,
0.8848
],
"note": "Only 2/5 seeds completed (credits exhausted)"
}
},
"fit": {
"A": 0.994056,
"p": 0.996364,
"B": 0,
"error_per_clifford": 0.001818,
"gate_fidelity": 0.998182,
"fidelity_percent": 99.8182
}
},
"claims": {
"qv_2qubit": {
"published": "HOF > 2/3",
"measured": 0.757,
"pass": true
},
"qv_3qubit": {
"published": "HOF > 2/3",
"measured": 0.635,
"pass": false
},
"rb_gate_fidelity": {
"published": "0.99 +/- 0.01",
"measured": 0.998182,
"pass": true
}
}
}peruzzo2014_replication -- ibm-torino
{
"sweep_results": [
{
"bond_distance": 0.5,
"energy_measured": -2.458913,
"fci_energy": -2.640714590488258,
"hf_energy": -2.631665109619672,
"optimal_energy_2q": -2.640623,
"error_hartree": 0.181802,
"error_kcal_mol": 114.08,
"chemical_accuracy": false,
"alpha": -0.108172,
"expectation_values": {
"Z0": -0.8801,
"Z1": 0.8801,
"Z0Z1": -1,
"X0X1": -0.0703,
"Y0Y1": -0.0503
},
"postselection_keep_fraction": 0.8428,
"raw_counts": {
"z_basis": {
"10": 207,
"11": 274,
"01": 3245,
"00": 370
},
"x_basis": {
"10": 1118,
"11": 862,
"01": 1074,
"00": 1042
},
"y_basis": {
"10": 1188,
"11": 940,
"00": 1005,
"01": 963
}
}
},
{
"bond_distance": 0.75,
"energy_measured": -2.700915,
"fci_energy": -2.8461872839647278,
"hf_energy": -2.836446514916097,
"optimal_energy_2q": -2.845874,
"error_hartree": 0.145272,
"error_kcal_mol": 91.16,
"chemical_accuracy": false,
"alpha": -0.127328,
"expectation_values": {
"Z0": -0.8725,
"Z1": 0.8725,
"Z0Z1": -1,
"X0X1": -0.0859,
"Y0Y1": -0.0728
},
"postselection_keep_fraction": 0.835,
"raw_counts": {
"z_basis": {
"10": 218,
"11": 304,
"01": 3202,
"00": 372
},
"x_basis": {
"10": 1168,
"11": 852,
"01": 1056,
"00": 1020
},
"y_basis": {
"10": 1197,
"11": 907,
"00": 992,
"01": 1000
}
}
},
{
"bond_distance": 1,
"energy_measured": -2.728097,
"fci_energy": -2.860205122579803,
"hf_energy": -2.8529210783245977,
"optimal_energy_2q": -2.859724,
"error_hartree": 0.132108,
"error_kcal_mol": 82.9,
"chemical_accuracy": false,
"alpha": -0.116022,
"expectation_values": {
"Z0": -0.8665,
"Z1": 0.8665,
"Z0Z1": -1,
"X0X1": -0.0947,
"Y0Y1": -0.0649
},
"postselection_keep_fraction": 0.8413,
"raw_counts": {
"z_basis": {
"10": 230,
"11": 271,
"01": 3216,
"00": 379
},
"x_basis": {
"10": 1104,
"11": 843,
"01": 1138,
"00": 1011
},
"y_basis": {
"10": 1222,
"11": 937,
"00": 978,
"01": 959
}
}
},
{
"bond_distance": 1.25,
"energy_measured": -2.710575,
"fci_energy": -2.841379054075542,
"hf_energy": -2.8377298235624453,
"optimal_energy_2q": -2.841057,
"error_hartree": 0.130804,
"error_kcal_mol": 82.08,
"chemical_accuracy": false,
"alpha": -0.082643,
"expectation_values": {
"Z0": -0.8651,
"Z1": 0.8651,
"Z0Z1": -1,
"X0X1": -0.0361,
"Y0Y1": -0.0605
},
"postselection_keep_fraction": 0.8472,
"raw_counts": {
"z_basis": {
"10": 234,
"11": 278,
"00": 348,
"01": 3236
},
"x_basis": {
"10": 1116,
"11": 906,
"01": 1006,
"00": 1068
},
"y_basis": {
"10": 1192,
"11": 957,
"00": 967,
"01": 980
}
}
},
{
"bond_distance": 1.5,
"energy_measured": -2.715534,
"fci_energy": -2.824682676189804,
"hf_energy": -2.823449750033749,
"optimal_energy_2q": -2.824584,
"error_hartree": 0.109148,
"error_kcal_mol": 68.49,
"chemical_accuracy": false,
"alpha": -0.047812,
"expectation_values": {
"Z0": -0.8898,
"Z1": 0.8898,
"Z0Z1": -1,
"X0X1": -0.0181,
"Y0Y1": -0.0381
},
"postselection_keep_fraction": 0.8416,
"raw_counts": {
"z_basis": {
"10": 190,
"11": 286,
"01": 3257,
"00": 363
},
"x_basis": {
"10": 1088,
"11": 917,
"00": 1094,
"01": 997
},
"y_basis": {
"10": 1173,
"11": 952,
"00": 1018,
"01": 953
}
}
},
{
"bond_distance": 1.75,
"energy_measured": -2.705776,
"fci_energy": -2.815191701228876,
"hf_energy": -2.814897063913006,
"optimal_energy_2q": -2.815176,
"error_hartree": 0.109415,
"error_kcal_mol": 68.66,
"chemical_accuracy": false,
"alpha": -0.023318,
"expectation_values": {
"Z0": -0.8934,
"Z1": 0.8934,
"Z0Z1": -1,
"X0X1": -0.0112,
"Y0Y1": -0.0122
},
"postselection_keep_fraction": 0.8381,
"raw_counts": {
"z_basis": {
"10": 183,
"11": 267,
"01": 3250,
"00": 396
},
"x_basis": {
"10": 1092,
"11": 918,
"00": 1107,
"01": 979
},
"y_basis": {
"10": 1159,
"11": 971,
"01": 914,
"00": 1052
}
}
},
{
"bond_distance": 2,
"energy_measured": -2.68664,
"fci_energy": -2.810780099134376,
"hf_energy": -2.810725834824041,
"optimal_energy_2q": -2.810779,
"error_hartree": 0.124141,
"error_kcal_mol": 77.9,
"chemical_accuracy": false,
"alpha": -0.009979,
"expectation_values": {
"Z0": -0.8828,
"Z1": 0.8828,
"Z0Z1": -1,
"X0X1": -0.0015,
"Y0Y1": -0.0308
},
"postselection_keep_fraction": 0.8418,
"raw_counts": {
"z_basis": {
"10": 202,
"11": 271,
"00": 377,
"01": 3246
},
"x_basis": {
"10": 1072,
"11": 900,
"00": 1145,
"01": 979
},
"y_basis": {
"10": 1147,
"11": 961,
"00": 1024,
"01": 964
}
}
},
{
"bond_distance": 2.25,
"energy_measured": -2.670297,
"fci_energy": -2.808937881696673,
"hf_energy": -2.8089294960677513,
"optimal_energy_2q": -2.808938,
"error_hartree": 0.138641,
"error_kcal_mol": 87,
"chemical_accuracy": false,
"alpha": -0.003901,
"expectation_values": {
"Z0": -0.8728,
"Z1": 0.8728,
"Z0Z1": -1,
"X0X1": 0.0103,
"Y0Y1": 0.0205
},
"postselection_keep_fraction": 0.8445,
"raw_counts": {
"z_basis": {
"10": 220,
"11": 292,
"01": 3239,
"00": 345
},
"x_basis": {
"10": 1077,
"11": 925,
"01": 950,
"00": 1144
},
"y_basis": {
"10": 1121,
"11": 1029,
"01": 885,
"00": 1061
}
}
},
{
"bond_distance": 2.5,
"energy_measured": -2.684843,
"fci_energy": -2.808209985626129,
"hf_energy": -2.8082088471132822,
"optimal_energy_2q": -2.80821,
"error_hartree": 0.123367,
"error_kcal_mol": 77.41,
"chemical_accuracy": false,
"alpha": -0.001427,
"expectation_values": {
"Z0": -0.8891,
"Z1": 0.8891,
"Z0Z1": -1,
"X0X1": -0.0044,
"Y0Y1": 0.0142
},
"postselection_keep_fraction": 0.8413,
"raw_counts": {
"z_basis": {
"10": 191,
"11": 324,
"01": 3255,
"00": 326
},
"x_basis": {
"10": 1075,
"11": 923,
"01": 982,
"00": 1116
},
"y_basis": {
"10": 1140,
"11": 1003,
"01": 879,
"00": 1074
}
}
},
{
"bond_distance": 2.75,
"energy_measured": -2.668617,
"fci_energy": -2.8079344726497077,
"hf_energy": -2.8079343345591066,
"optimal_energy_2q": -2.807934,
"error_hartree": 0.139317,
"error_kcal_mol": 87.42,
"chemical_accuracy": false,
"alpha": -0.000495,
"expectation_values": {
"Z0": -0.877,
"Z1": 0.877,
"Z0Z1": -1,
"X0X1": 0.0088,
"Y0Y1": 0.0278
},
"postselection_keep_fraction": 0.8293,
"raw_counts": {
"z_basis": {
"10": 209,
"11": 310,
"01": 3188,
"00": 389
},
"x_basis": {
"10": 1080,
"11": 957,
"00": 1109,
"01": 950
},
"y_basis": {
"10": 1101,
"11": 1003,
"00": 1102,
"01": 890
}
}
},
{
"bond_distance": 3,
"energy_measured": -2.678176,
"fci_energy": -2.807834895538406,
"hf_energy": -2.8078348804626083,
"optimal_energy_2q": -2.807835,
"error_hartree": 0.129659,
"error_kcal_mol": 81.36,
"chemical_accuracy": false,
"alpha": -0.000161,
"expectation_values": {
"Z0": -0.8871,
"Z1": 0.8871,
"Z0Z1": -1,
"X0X1": -0.0127,
"Y0Y1": 0.0078
},
"postselection_keep_fraction": 0.8518,
"raw_counts": {
"z_basis": {
"10": 197,
"11": 264,
"01": 3292,
"00": 343
},
"x_basis": {
"10": 1123,
"11": 920,
"00": 1102,
"01": 951
},
"y_basis": {
"10": 1167,
"11": 1012,
"00": 1052,
"01": 865
}
}
}
],
"summary": {
"mae_kcal_mol": 83.5,
"n_chemical_accuracy": 0,
"n_total": 11,
"best_distance": 1.5,
"best_error_kcal": 68.49,
"worst_distance": 0.5,
"worst_error_kcal": 114.08
},
"interpretation": "HeH+ VQE bond sweep on IBM Torino: 0/11 distances within chemical accuracy (1.6 kcal/mol). MAE = 83.50 kcal/mol. 2-qubit sector-projected ansatz with pre-optimized parameters. Post-selection retains ~80-90% of Z-basis shots. Reproduces the potential energy surface shape from Peruzzo 2014."
}cross2019_replication -- ibm-torino
{
"quantum_volume": {
"n2": {
"mean_hof": 0.6971,
"std_hof": 0.1055,
"per_circuit_hof": [
0.8313,
0.7671,
0.7373,
0.6001,
0.5496
],
"passes_threshold": true,
"threshold": 0.6666666666666666
},
"n3": {
"mean_hof": 0.8102,
"std_hof": 0.0558,
"per_circuit_hof": [
0.8535,
0.7834,
0.8853,
0.7246,
0.8042
],
"passes_threshold": true,
"threshold": 0.6666666666666666
}
},
"randomized_benchmarking": {
"sequence_lengths": [
1,
4,
8,
16,
32
],
"data": {
"1": {
"mean_survival": 0.9047,
"std_survival": 0.0027,
"per_seed": [
0.9028,
0.9048,
0.905,
0.9014,
0.9094
]
},
"4": {
"mean_survival": 0.9026,
"std_survival": 0.0068,
"per_seed": [
0.8928,
0.9075,
0.8984,
0.9124,
0.9021
]
},
"8": {
"mean_survival": 0.9037,
"std_survival": 0.0043,
"per_seed": [
0.908,
0.9016,
0.8962,
0.9062,
0.9065
]
},
"16": {
"mean_survival": 0.9003,
"std_survival": 0.0079,
"per_seed": [
0.8887,
0.9099,
0.9084,
0.8965,
0.8979
]
},
"32": {
"mean_survival": 0.9013,
"std_survival": 0.0024,
"per_seed": [
0.9004,
0.8979,
0.9016,
0.9053,
0.9014
]
}
},
"fit": {
"A": 0.451414,
"p": 0.999774,
"B": 0.452348,
"error_per_clifford": 0.000113,
"gate_fidelity": 0.999887,
"fidelity_percent": 99.9887
}
},
"claims": {
"qv_2qubit": {
"published": "HOF > 2/3",
"measured": 0.6971,
"pass": true
},
"qv_3qubit": {
"published": "HOF > 2/3",
"measured": 0.8102,
"pass": true
},
"rb_gate_fidelity": {
"published": "0.99 +/- 0.01",
"measured": 0.999887,
"pass": true
}
},
"interpretation": "Cross 2019 replication on IBM Torino: QV n=2 HOF=69.7% [PASS], QV n=3 HOF=81.0% [PASS], RB gate fidelity=99.99% (note: inflated because IBM transpiler collapses Clifford sequences to depth 1-2 circuits, so RB measures readout error not gate error; survival ~90% flat across all sequence lengths). Cross-platform: emulator (QV PASS, RB 99.95%), IBM Torino (QV PASS/PASS, RB 99.99%*), Tuna-9 (QV PASS/PASS, RB 99.82%)."
}qaoa_maxcut_with_compilation -- emulator + noise model
vqe_heh_plus_cross_platform -- pennylane (default.qubit + default.mixed)
vqe_heh_plus -- pennylane (default.qubit + default.mixed)
vqe_h2_symmetry_verification -- emulator (pennylane statevector + shot noise)
vqe_h2_symmetry_verification -- emulator (pennylane statevector + shot noise)
vqe_h2_symmetry_verification -- emulator (pennylane statevector + shot noise)
vqe_h2_symmetry_verification -- emulator (pennylane statevector + shot noise)
-- tuna-9
Statistical caveats. These are preliminary results from a small number of circuits and shots. Randomized benchmarking uses 5 random sequences per depth (standard protocols recommend 30+). Quantum Volume tests use 5 circuits per qubit count (IBM's protocol specifies 100+, with a two-sigma confidence interval). QAOA results reflect a single QAOA layer; deeper circuits may find better solutions.
Hardware variability. Results on real hardware (IBM, QI Tuna-9) vary between runs due to fluctuating qubit coherence, calibration drift, and crosstalk. A single run does not capture this variance. Error bars and multi-run statistics are planned.
Error mitigation. VQE results use parity post-selection (discarding states outside the target symmetry sector) but no advanced techniques like zero-noise extrapolation or probabilistic error cancellation. IBM runs use dynamical decoupling (XpXm) and Pauli twirling when available.
Emulator vs. hardware. Emulator results (qxelarator) represent noiseless ideal execution. Perfect fidelity/accuracy on the emulator is expected and not indicative of hardware capability. The value of emulator runs is as a correctness baseline.
An autonomous Python daemon processes a queue of experiments. It generates quantum circuits, submits them to real hardware, analyzes the measurement results, and publishes everything to this page -- no human intervention required.
Experiments are defined as structured descriptions -- what to measure, which backend, how many shots.
The daemon translates each experiment into a quantum circuit written in cQASM 3.0 (the native instruction set for Quantum Inspire hardware).
Circuits are submitted to real quantum processors: Quantum Inspire's Tuna-9 (9 superconducting qubits) or IBM Quantum (100+ qubits).
Raw measurement counts are processed into physical quantities -- fidelities, energies, error rates -- then published here automatically.