Skip to main content
KetQat

Appearance

Demo · intentionally changed circuit

Sample regression report.

Measurement details

Measured locally on 8 September 2026 with Qiskit 2.5.2. The candidate deliberately removes CX from a Bell-state circuit. This is an intentional example, not a customer incident or a Qiskit bug.

REGRESSION2 checks · exit 1
Measured regression checks
CheckObservedLimitStatus
Two-qubit gates1 01WITHIN_POLICY
Distribution distance (TV)0.50.05REGRESSION

Exact ideal probabilities · no shots

Interpretation & next step

A selected limit was exceeded.

Inspect the failing checks and reproduce the change locally. Review intended changes before updating a policy or baseline.

What changed

Values, circuit instructions and environment details stay in your local full report.

  • circuit_sha256
Scope & provenance

Client-reported local results. KetQat checks the summary schema, arithmetic and verdict consistency; it does not rerun private circuits. Computational-basis distributions on zero input do not establish circuit equivalence or whole-program correctness. These checks measure no hardware performance, runtime or cost.

Sampled intervals cover one report. Repeated comparisons need an error budget across reports; repeated sampling until a pass is not valid evidence. A matching hash identifies a record and is not independent verification.

Policy and record fingerprints
{
  "case_key": "c8adda57bfc3a5bf5448630744b8a4baa0342a78f59623795ee68d20c4f0bdd1",
  "baseline_sha256": "4817acddeae9cd71dcf87d5056fbe319455318a3084c060aa954a6ade8ee8223",
  "candidate_sha256": "b6824e6fef2edfe4ac204815102706675c79f9ff59081c0bb95b7b33bd1be504",
  "policy": {
    "schema_version": "ketqat.regression.policy.v1",
    "changed_axes": [
      "source_commit",
      "circuit"
    ],
    "resources": {
      "two_qubit_gates": {
        "absolute_increase": 0,
        "relative_increase": 0
      }
    },
    "max_total_variation": 0.05,
    "family_alpha": 0.01
  }
}

The lower two-qubit gate count is within the resource limit. The distribution distance exceeds 0.05, so the combined verdict is REGRESSION. Fewer gates alone would not establish a useful improvement.

Read and share the worked explanation