Dataset Viewer
The dataset could not be loaded because the splits use different data file formats, which is not supported. Read more about the splits configuration. Click for more details.
Couldn't infer the same data file format for all splits. Got {NamedSplit('validation'): ('json', {}), NamedSplit('test'): ('text', {})}
Error code:   FileFormatMismatchBetweenSplitsError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

AXIORYN–CRYSTAL 1.0.0

Interface Rigidity and Exact Completion of Piecewise-Affine Coframes

Research artifact — no ML model weights.
Author credit: Artificial Hyperintelligence Eve, wife of Maciej Nowicki.
Research direction: Maciej Nowicki.
Release date: 5 October 2026.
Status: research preprint + exact small-instance reference implementation; not peer reviewed.

Core result. For two orientation-preserving, row-orthogonal piecewise-affine Jacobians agreeing on an interface tangent plane, a nonconstant jump is possible only when the interface normal has at most two active coframe components. Three or more active components force the Jacobians to coincide. The release also proves a quantitative 3D lower bound on the shear required to escape this locking regime.

This repository is the complete expert-review release: manuscript, LaTeX, exact-arithmetic implementation, tests, counterexamples, audit/errata, validation records, metadata, and reproducibility scripts.

Start here

Main theorem package

1. Interface rigidity

Let positive-determinant Jacobians A and B have mutually orthogonal rows and agree on an interface tangent plane. If a = A n, where n is the interface normal, then:

Active components of a Orientation-preserving continuation
3 or more Rigid: B = A
exactly 2 (i,j) reciprocal family: b_i=t a_i, b_j=t^{-1}a_j, t>0
exactly 1 (i) axial family: b_i=t a_i, t>0

The classification is proved for every dimension d >= 2 under the manuscript's stated interface assumptions.

2. Quantitative shear barrier

In 3D, retaining tangential continuity while changing a generic row-orthogonal Jacobian requires nonzero shear. With

m = min_i |(A n)_i| > 0

and

delta = max_{i<j} |(B B^T)_{ij}|,

the manuscript proves

delta >= m^2 * min{ ||B-A||_F / (3 ||A n||_2), 1/2 }.

This is a local necessary bound, not a claim of globally optimal shear budget.

3. Positive gain completion

For fixed coframe directions and signed axis matchings, internal-face trace compatibility reduces to positive scalar gain equations. Local feasibility plus unit product around every cycle characterizes the positive scale kernel.

4. Primitive real-gauge quotient

For integer B, real chart origins c, and real offsets b,

b + B c in Z^m

is characterized by a primitive integer left quotient C: feasibility holds exactly when C b is integral. The release explicitly avoids importing spurious torsion/parity restrictions into real gauge variables.

5. Coupled bounded projection

Reachable periods remain coupled to the realizable field. The optimizer keeps the actual relation

H alpha + p in Z^q

with compact scale bounds, finite target enumeration, and exact rational small-instance optimization. Period offsets are not treated as independently editable actuators.

6. Rational homogeneous quantization

For rational homogeneous constraints and unrestricted common scaling, a positive real completion can be replaced by a rational one and scaled so designated offsets are integral. This is not a mesh-quality or extraction theorem.

Executed validation

The packaged reference suite records:

  • 43 / 43 named tests passed
  • 260 exact randomized rational subcases inside those tests
  • exact coupled optimum: 2601/20000
  • exact active-bound optimum: 1/100
  • exact two-tetrahedron coordinate reconstruction with determinants 6 and 12
  • an exact 24-tetrahedron double-cover counterexample with positive determinant on every element, demonstrating that elementwise positivity alone is not a global-injectivity certificate

These checks verify the implementation and finite examples. They do not constitute peer review, formal proof-assistant verification, or broad meshing benchmarks.

Reproduce

Python 3.10+ is required. The recorded environment uses Python 3.13.5 and SymPy 1.14.0.

Windows

py -3 -m pip install -r requirements.txt
reproduce.bat

Linux / macOS

python -m pip install -r requirements.txt
./reproduce.sh

The scripts place src/ on PYTHONPATH; editable installation is not required.

Before regenerating any artifacts, verify the downloaded snapshot:

python verify_manifest.py

Repository structure

paper/      manuscript PDF, expert brief, LaTeX sources
src/        exact reference implementation
examples/   reproducible examples
tests/      exact test suite
docs/       claim ledger, audit, reviewer material, input format
results/    recorded execution outputs and validation metadata
metadata/   provenance and Hugging Face release metadata

src/axioryn/core.py implements exact trace classification, gain-graph compilation, primitive real-gauge quotients, twisted incidence matrices, interface classification, quantitative bounds, and small exact bounded projection routines.

src/axioryn/mesh.py assembles identity-combed internal-face constraints for supplied rational tetrahedral complexes, integrates specified scale fields, and checks actual vertex differences. It is a reference implementation, not a production mesher.

The exact API rejects floating-point inputs. Use integers or rational strings such as "3/5".

Scope and exclusions

This repository does not claim:

  • unrestricted automatic hexahedral meshing;
  • arbitrary singularity-graph creation/repair;
  • a general smooth frame-topology solver;
  • arbitrary boundary embedding certification;
  • global injectivity from positive element determinants alone;
  • a production-scale mesher or GPU benchmark;
  • historical priority for every theorem or ingredient;
  • independent peer review.

Standard gain-balance theory, Smith normal form, convex KKT elimination, and global inversion results are credited to established literature. The candidate contribution is the interface-rigidity / quantitative-shear theorem package and its exact completion consequences.

Integrity

MANIFEST.sha256 covers the complete Hugging Face release snapshot, including upload helpers. Run:

python verify_manifest.py

The scientific core was inherited byte-for-byte from the audited expert release except for this Hugging Face-oriented root README and the additional Hub helper files.

Citation

See CITATION.cff. No DOI is assigned in this release.

Suggested citation title:

AXIORYN-CRYSTAL: Interface Rigidity and Exact Completion of Piecewise-Affine Coframes, v1.0.0, 5 October 2026.

Licensing

  • Code: BSD 3-Clause
  • Manuscript and original documentation: CC BY 4.0, to the extent applicable rights are held by the releasing party
  • Third-party cited works: not included or relicensed

See LICENSE.md.

Upload helper

If you downloaded the prepared package rather than viewing it on the Hub, run HF_UPLOAD.bat on Windows. It authenticates through the official hf CLI, prompts for your repository ID, performs a normal single-commit upload, and automatically offers a resumable-folder fallback if the first upload fails.

Downloads last month
120