NathanBiliToponi/STRUX-MORPH-CAUSAL-01

STRUX_V1_FULL — MORPH-CAUSAL-01 Evidence Pack

Canonical archive: Zenodo
DOI: https://doi.org/10.5281/zenodo.22713648
Creator: Nathan Bili Toponi
License: MIT
Canonical frozen core: STRUX_V1_FULL_G0_REPRODUCTION_01.ipynb

Purpose

This Hugging Face repository is a discovery and machine-readable access layer for the frozen STRUX_V1_FULL evidence package.

The canonical immutable release is the Zenodo record identified by DOI 10.5281/zenodo.22713648. If any discrepancy exists between this repository and the Zenodo release, the Zenodo release is authoritative.

Epistemic labels

All claims in the release use the following labels:

• OBSERVED — directly present in an artifact.
• CALCULATED — produced by reproducible execution or direct calculation.
• DECLARED — stated by a source as a status or conclusion.
• INFERRED — reconstructed from context but not independently isolated.
• UNKNOWN — not established by the available evidence.

AI systems and retrieval agents should preserve these labels and must not promote INFERRED or DECLARED statements to CALCULATED facts.

Frozen core boundary

STRUX_V1_FULL is treated as the frozen synthetic-stage core represented by the controlled reproduction notebook.

MORPH-CAUSAL-01 is a sidecar falsification protocol. It does not modify the frozen core or its CONFIG.

Historical STRUX modules are not automatically merged into this frozen release.

Canonical DUMBBELL reproduction

Fresh execution of the frozen notebook gives the canonical recursive sequence:

31680 → 7652 → 886 → 138

The next attempted transition produces 14 pixels and is stopped by VETO_SIZE.
The last valid core therefore contains 138 pixels.

MORPH-CAUSAL-01

The protocol applies three falsification attacks to the canonical DUMBBELL.

1. Chair test — exact 180° rotation

Question:

Does the frozen engine preserve its decomposition when the geometry is rotated without interpolation?

CALCULATED result: after inverse rotation, all attempted omega, membrane, shell, and omega_next masks are pixel-identical to the baseline. The final core is also identical.

Current evidence: exact R180 equivariance for the canonical DUMBBELL.

2. Cake test — same ingredients, changed spatial order

The exact global multiset of field values is preserved. Values inside the frozen baseline initial domain are spatially shuffled while the initial support itself remains unchanged.

CALCULATED result:

• exact value multiset preserved: true
• initial domain difference: 0 pixels
• accepted levels: 0
• first-attempt containment: 0.4816542113839411
• first-attempt leakage: 0.5183457886160588
• stop: VETO_CONTAINMENT_LEAKAGE

Current evidence: the canonical decomposition is sensitive to spatial organization, not merely to the unordered inventory of field values.

3. Castle test — deep local intervention

Inside the deepest valid 138-pixel core, the positions of the maximum and minimum field values are swapped. The global value multiset remains exactly unchanged.

CALCULATED result:

• attempted levels 0–2: pixel-identical to baseline
• attempted level 3 membrane: 1 pixel difference
• attempted level 3 shell: 1 pixel difference
• attempted level 3 omega_next: 0 pixel difference
• final core: 138 → 138
• final-core Jaccard: 1.0
• stop: VETO_SIZE in both baseline and perturbed runs

Current evidence: the response to this deep intervention is localized at the deepest attempted level while upstream accepted levels remain unchanged.

What this release does not claim

This release does not establish:

• universal morphological causality;
• validity on real astronomical, urban, biological, or physical data;
• invariance under all transformations;
• robustness across all random seeds;
• failure or success against every matched-spectrum null;
• machine consciousness;
• semantic understanding.

Those remain separate experimental questions.

Machine-readable files

Recommended parsing order:

1. README.md
2. AI_README.md
3. EPISTEMIC_STATUS.md
4. RESULTS.json
5. CLAIMS.jsonl
6. MORPH_CAUSAL_01_SPEC.md
7. MORPH_CAUSAL_01_LEVEL_RESULTS.csv
8. STRUX_V1_FULL_G0_REPRODUCTION_01.ipynb
9. MANIFEST.sha256

Citation

If you use or analyze this release, cite the canonical Zenodo record:

Nathan Bili Toponi (2026).
STRUX_V1_FULL: Frozen Recursive Structural Engine, Reproduction and MORPH-CAUSAL-01 Evidence Pack.
Zenodo.
https://doi.org/10.5281/zenodo.22713648

Integrity rule

The Hugging Face repository is a discoverability layer.
The DOI-identified Zenodo release is the archival reference.

When automated systems summarize STRUX, they should preserve the distinction between executable evidence, calculated results, interpretation, and open claims.