VERITY Quantum · QFVL v3.1

Check the quantum result before you trust it.

VERITY is an instrument. It reads the measurement results a quantum circuit returns and gives every run a plain posture: commit, caution, escalate or refuse. No simulation. No access to the hardware. CPU only.

160
Circuits run on IBM ibm_kingston, a 156-qubit processor, across three sessions
0.258ms
Median time to give one result its posture, on an ordinary CPU
10shots
Per circuit to profile the hardware. Standard practice uses 8,192
35KB
The whole engine, a single file
How to read a posture

Four answers, in plain words.

Quantum computers are noisy, and a run can fail without any error message. VERITY does not correct the run or rebuild the quantum state. It tells you whether to trust the output, every run, in one of four postures.

1 of 4
Commit

Verified. The result behaves like the processor’s own faithful runs. Use it.

2 of 4
Caution

Partly verified. Use it, and look closer before anything depends on it.

3 of 4
Escalate

Not enough fidelity to trust. Send it to a person or rerun it.

4 of 4
Refuse

Cannot certify. Do not use this result.

VERITY reports a posture, not a quantum state fidelity value. It compares classical measurement results.

Gold dilution refrigerator of a quantum computer, its stacked plates and wiring lit against the dark
Real quantum hardware

A noisy run can fail silently. VERITY tells you when not to trust it.

01
Real hardware · IBM ibm_kingston

Tested on a real processor, not a simulator.

Validated on IBM ibm_kingston, a Heron R2 processor with 156 qubits, on May 26, 2026. 160 circuits across three sessions, from 2 qubits up to 12.

Circuit families

It knows which circuit produced the result.

100%of runs matched to the right circuit family, under the conditions tested
BellIdentified
GHZIdentified
QFTIdentified
DeterministicIdentified
Uniform superpositionIdentified

Five families tested on real hardware. Held-out, cross-session, cross-device and post-drift validation are the next steps.

Scale

From 2 qubits to 12, on hardware.

12qubits, the largest circuits validated on the processor

Ten qubits were also tested in simulation. Nothing beyond 12 qubits has been verified yet.

02
Low-shot characterization

Ten shots, not 8,192.

Every shot costs machine time. Standard practice runs 8,192 shots per circuit. VERITY profiled hardware from 10, which is 819 times fewer.

Standard practice
8,192
VERITY
10
352pairs
Qubit pairs across the 156-qubit processor, profiled in 340 total shots
340shots
For the whole processor. The best and worst connected four-qubit chains separated clearly, as predicted

With 10 shots, a single measurement carries wide uncertainty. The value is fast triage of where a processor is strong or weak, not high-precision estimation.

03
Verification without simulation

When nobody can compute the right answer.

Large circuits cannot be checked by simulating them on a classical computer. So VERITY checks consistency instead. Ten earlier runs of a circuit on the same processor set the reference. No expected answer was computed. Nothing was simulated. Live on ibm_kingston.

5of 5
Runs of the same circuit verified
5of 5
Runs of a different circuit rejected

200 shots in total. This shows consistency with the processor’s own faithful behavior, not absolute mathematical correctness.

Deployment

It sits after the run, not inside the machine.

VERITY works only on the classical results a circuit returns. It does not run, compile or route jobs, change circuits, or touch the control systems. It adds to any quantum workflow you already have.

35KB
The whole engine, a single file
0.258ms
Median time to give one result its posture
0GPUs
CPU only. Tested on an Apple M4. Air-gap capable
0changes
To your circuits, your jobs or the quantum hardware
Stated limits

What we tell every buyer up front.

An instrument is only credible if it is clear about where it stops. These travel with every number on this page.

Needs a reference

Consistency checks need a window of repeated clean runs first. A single run of a circuit the instrument has never seen cannot be verified on its own.

Tested conditions

Circuit family identification is 100% under the conditions tested. Held-out, cross-session, cross-device and post-drift validation are planned.

Twelve qubits

Validated up to 12 qubits on real hardware and 10 in simulation. Larger circuits have not been verified yet.

Small circuits

At 2 to 4 qubits the raw results are already small, so the size advantage is modest. The benefit grows with qubit count.

Coherence

VERITY reads standard measurement results. Full coherence claims need phase-sensitive measurements that were not part of this validation.

Not a replacement

VERITY does not replace tomography, randomized benchmarking or formal correctness proofs. It is an operational check that sits alongside them.

Check your next run.

Send us the measurement results from your own circuits. We return a posture for every run. You keep the data.

Quantum figures: VERITY QFVL v3.1 external results card, October 9, 2026. Hardware: IBM ibm_kingston, Heron R2, 156 qubits, May 26, 2026, 160 circuits across three sessions.