Skip to content

Metriq Submission Readiness

Date checked: 2026-05-06

This note records the readiness gate for submitting scpn-quantum-control results to Metriq.

Current Metriq Workflow

The current Metriq platform describes benchmark contribution through the metriq-gym command-line workflow. The public FAQ states that new benchmark results should be added with the upload command from the metriq-gym CLI.

The Metriq-Gym documentation describes benchmark execution as:

mgym job dispatch <benchmark-config.json> --provider <provider> --device <device>
mgym job poll <job-id>

and documents a suite of benchmark definitions such as mirror circuits, EPLG, BSEQ, and WIT.

Readiness Decision

No Metriq submission was made in this pass.

Reason:

  • the committed SCPN benchmark artefacts are valid for the project papers, but they are not Metriq-Gym result artefacts;
  • Metriq expects benchmark outputs generated through its own schemas and upload path;
  • submitting project-specific Rust/VQE/DLA/FIM benchmark tables as Metriq results would blur benchmark definitions and violate the repository's bounded-claim discipline.

Local Installation State

metriq-gym was installed into an isolated user-level virtual environment:

/home/anulum/.venvs/scpn-metriq

The command-line entry point is:

/home/anulum/.venvs/scpn-metriq/bin/mgym

Validation:

/home/anulum/.venvs/scpn-metriq/bin/python -m pip check
/home/anulum/.venvs/scpn-metriq/bin/mgym --help
./.venv-linux/bin/python -m pip check

Both the isolated Metriq environment and the project environment reported no broken Python requirements after the final repair.

The installation is intentionally not inside the repository and not inside .venv-linux. A first attempt to install metriq-gym directly into .venv-linux introduced incompatible dependency pins around Qiskit, NumPy, and Click. Those conflicts were removed from .venv-linux, and the project environment was rechecked before proceeding.

Valid Submission Gate

A future Metriq submission is valid only after one of these paths is completed:

  1. Run an existing Metriq-Gym benchmark on a supported local simulator or QPU backend and upload the resulting Metriq artefact through the documented mgym path.
  2. Propose an SCPN/Kuramoto-XY benchmark definition upstream to Metriq-Gym, wait for review/acceptance, then run and upload results using that accepted schema.

The first path is lower risk and should be preferred for an initial submission.

Suggested Initial Benchmark

A no-QPU readiness run was completed with a local simulator first and is recorded in docs/publication/metriq_local_smoke_2026-05-06.md.

The executed smoke workflow was:

mgym job dispatch /tmp/scpn_metriq_bv_local_smoke_2026-05-06.json \
  --provider local \
  --device aer_simulator
mgym job poll b96914ac-2e2f-461e-84a8-b61d81300fb2

The run used the standard Metriq-Gym Bernstein--Vazirani schema, reported local simulator execution, and did not upload a result.

Only after local schema, environment, and upload behaviour are understood should any hardware benchmark be considered. Hardware execution must remain explicitly approved and budget-gated.

Claim Boundary

This note does not claim:

  • any Metriq result submission;
  • any accepted Metriq benchmark;
  • any Metriq score;
  • any QPU execution;
  • any performance comparison beyond the existing project-specific benchmark artefacts.

It records the submission gate required to keep Metriq results scientifically and procedurally valid. The local smoke result confirms CLI readiness only; it is not a submitted or endorsed project benchmark.