WAVEFORM ATLAS — RELAXATION AND STATISTICAL METHODS
3 November 2026

This public subset contains our C5 fixtures, trajectories and numerical checks, plus the statistical repair and its archived synthetic qualification. These are our calculations, not data supplied by Marija Vucelja or her coauthors. Literature is credited in sources.json.

Run in a Python 3.12 environment:
python -m pip install -r statistical-repair/requirements.txt
python run_c5_validation.py
python run_statistical_validation.py

The wrappers use temporary directories and do not overwrite archived evidence. C5 reproduction includes the original implementation and a separately implemented check. Other portfolio benchmarks (C1–C4, B and hardware programmes) are outside this subset.

The website algebraic evaluator is mpemba.js; its parity record is c5-browser-validation.json. Model temperatures and times are dimensionless. All C5 trajectories are nonstationary and ineligible for stationary surrogate inference. Hardware holds and instrument-transfer requirements remain in force.

The original unsuccessful hand-picked C5 model is reconstructed from a source comment; the original failure log was unavailable. Selection and finite-grid limitations remain documented. A passing finite check is not a theorem, physical experiment or readiness score.

The reader-facing teaching examples are in story.js, with explicit illustrative inputs in worked-examples.json. They are not instrument measurements. Run node check_story.js to check their algebra and probability coordinates. edition-validation.json also records source-preservation and browser-model parity. The moving introduction uses the unchanged mpemba.js evaluator and C5 fixtures.
