DATA.md

Data map

All paths below are relative to publication/. Original file hashes appear in provenance/source-inventory.json; data is retained as supplied, not regenerated silently for publication. The protocols and source code define the observables.

Location	Contents	Use and limits
pilot/config.json	Original configuration and master seed	Defines the source-off pilot; preserve the accompanying pre-execution receipt
data/pilot_run/model_inputs.json	Retained sampled inputs	Reconstruct the delivered pilot models
data/pilot_run/per_unit.json, per_unit_summary.csv	Unit-level measures	Respect pairing and independent model-unit counts
data/pilot_run/results.json	Pilot summaries and intervals	Original primary estimand, not the later factorial contrast
data/pilot_run/trajectories_n4.npz, trajectories_n6.npz	Retained pilot trajectory arrays	Inspect keys/shapes before analysis; do not assume raw quantum states are included
data/reviewer_replay/	Reimplementation, factorial and neutral-control arrays and receipts	Same-project, model-assisted review evidence, not independent-team replication
data/factorial_analysis/	Cell summaries and finite-family intervals	Recomputed from retained factorial arrays by analyze_factorial.py
data/review_additions/	Sign tests, transport checks and initial-floor diagnostics	Secondary analyses; distinguish them from the predeclared primary
figures/	Supplied PNG and SVG figures	Prefer SVG for scaling; cite the associated work and version

Read compressed NumPy arrays without pickle execution:

from pathlib import Path
import numpy as np
path = Path('publication/data/reviewer_replay/factorial_n6.npz')
with np.load(path, allow_pickle=False) as archive:
    for key in archive.files:
        print(key, archive[key].shape, archive[key].dtype)
    print(archive['columns'])

Column-labelled factorial matrices contain per-unit observables, not all time-resolved quantum states. Use the accompanying analysis code for the exact initial/post-window conventions and the injection/dynamics pairing. Do not treat multiple arms or time points as independent experimental units.