OES-32 toy simulation (quantum-error-correction-demo)

Classical Python toy simulation of OES-32-style metrics. No qubits, no QEC code; every output is SYNTHETIC.

METRIC_TAG=SYNTHETIC. Every number this script prints comes from hand-written formulas on generated inputs.

What it is

• A single classical Python script, oes32_v14_improved.py, that runs a Monte Carlo loop over hand-written formulas relating a simulated erasure probability to "coherence", "success rate", and a few other labelled quantities.
• A teaching/exploration toy for the OES-32 naming used across the Spark AI NLP repositories.

What it is NOT

• Not a quantum computer implementation and not a quantum error-correcting code (no qubits, no stabilizers, no surface code, no decoder).
• Not a benchmark. Metric names such as "spooky correlation" and "entanglement entropy" are labels for formulas in the script, not quantum measurements.
• Not hardware, field, medical, or production software.
• Not bit-matched to the normative OES-32 residual (see Relationship to ADR-001).

The repository name predates these scope statements and is kept so existing links keep working.

Quickstart

Requires Python 3.8+. tqdm (progress bar) is the only third-party import and is optional.

git clone https://github.com/sparkainlp-x/quantum-error-correction-demo.git
cd quantum-error-correction-demo
python3 -m pip install -r requirements.txt   # optional (tqdm)
python3 oes32_v14_improved.py --iterations 20000 --erasure 0.15 --seed 42

Other options: --iterations N (default 1,000,000), --erasure P (default 0.20), --seed S (reproducible output), -v (verbose logging).

What the script prints

A report with run metadata and SYNTHETIC values for coherence before/after, success rate, "spooky correlation", "entanglement entropy", and collapse rate. The header and footer carry METRIC_TAG=SYNTHETIC, and the success-rate line carries its own SYNTHETIC note.

About the success rate

SYNTHETIC. Recovery success is drawn with probability max(0.95, 1 − 0.7·p_erasure), so the success rate is ≈95% by construction for p_erasure ≥ ~0.07. Reported figures near 95% are sampling outcomes of that hard-coded 0.95 floor, not a measured error-correction rate, and are not comparable with any QEC benchmark.

Tests

There is no unit-test suite. CI (smoke.yml) runs the script twice with --seed 42 and checks that:

1. the report header, footer, and success-rate line are tagged SYNTHETIC;
2. the removed "investor" / unconditional "STABLE" wording has not come back;
3. the two seeded runs produce identical output (timestamp excluded).

To reproduce locally, run the Quickstart command twice and compare the outputs.

Evidence tags

Item	Tag
Every printed metric	SYNTHETIC (formula output on generated inputs)
Any hardware, qubit, or QEC performance interpretation	Not claimed

Tag definitions: sparkainlp-x/.github.

Relationship to ADR-001

The normative OES-32 residual definition is oes32-residual@b77b612 (ADR-001). This demo is not bit-matched to it and is not a Profile A sidecar.

• oes32-residual: normative residual contract
• oes32_engine: Profile A sidecar (Python)
• oes32-hls: Profile A sidecar (C++ HLS prototype; synthesis UNRUN)
• OES-512 (16 × OES-32 blocks, weighted latch) is a TARGET; source not published (oes512-residual)

Citation

Archived on Zenodo: concept DOI 10.5281/zenodo.22985530 (all versions; resolves to the latest). The latest version, v0.1.3, is 10.5281/zenodo.23241551. The first archive, labelled v0.1.0 on Zenodo, is 10.5281/zenodo.22985531; its source tree is git tag v0.1.2 (the same files, re-tagged on the rewritten history).

Citation metadata is in CITATION.cff (GitHub shows a "Cite this repository" button). This is a toy script: if you mention it, cite a tagged release (or the repository URL and commit SHA) and describe its outputs as SYNTHETIC.

Author: Jean-François Brisson, Spark AI NLP, https://sparkainlpx.xyz. Questions: open an issue in this repository.

License

This software is available under the GNU Affero General Public License v3.0 only (AGPL-3.0-only); see LICENSE.

Organizations that want to use it in proprietary products or services without AGPL obligations can contact the author about a commercial license via https://sparkainlpx.xyz.

Versions published before 2026-09-29 were released under the MIT License and remain available under those terms.