Code and derived data for reproducing the observable-selected magic calculations.
This repository contains reproducibility materials only. Manuscript text, LaTeX source, submission files, manuscript PDFs, and generated figure files are kept separately. All paths below are relative to the repository root.
src/: simulation, validation, and plotting scripts.data/: generated CSV and JSON data tables used by the figures and checks.results/: validation files and intermediate result summaries. Generated figure files are written here locally and are excluded from version control.raw_data_package/README.md: description of the raw inputs and how they are regenerated.
Create a Python environment and install the listed packages:
python -m pip install -r requirements.txtThe Munich Quantum Toolkit (MQT) Bench validation uses optional public benchmark
circuits through mqt.bench and qiskit. The other scripts run from generated
circuits and fixed random seeds.
Run a small validation pass:
python src/reproduce_all.py --quick --skip-mqtRun the MQT Bench quick validation as well when mqt.bench and qiskit are
available:
python src/reproduce_all.py --quickRegenerate the non-MQT data products and figures:
python src/reproduce_all.py --skip-mqtRegenerate the full data products, including MQT Bench:
python src/reproduce_all.pyRegenerate figure files locally from the included data tables:
python src/build_mechanism_figure.py
python src/r41_observable_dark_gap.py
python src/plot_r41.py
python src/r41_proxy_counterexamples.py
python src/r41_stress_workloads.py --plot-only
python src/r41_2d_stress_workloads.py --plot-only
python src/r41_estimation_workload_benchmark.py --plot-onlyThe full reproduction command also regenerates the observable-blind validations, the analytic same-support positive-weight separation, and the random same-support coefficient validation. It also runs the one- and two-dimensional local-window calculations after quick validations.
The figure-generation scripts write local outputs to results/.