Research¶
Paper¶
Triadic Neurosymbolic Engine: Prime Factorization as a Neurosymbolic Bridge: Projecting Continuous Embeddings into Discrete Algebraic Space for Deterministic Verification
J. Arturo Ornelas Brand (2026). Independent Researcher.
Benchmarks¶
| Metric | Result | Notes |
|---|---|---|
| Pairwise verification speed | 28.4x faster than cosine | 50K operations |
| Composition guarantee | 100% verified | 5,671 word pairs |
| Hypernym detection accuracy | 100% TP | Contrastive projection at k=6 |
| Model audit scale | 108,694 discrepancies | 2M semantic chains, 2 models |
Key Contributions¶
- LSH-to-prime projection: Maps continuous embeddings to composite integers via locality sensitive hashing, enabling exact algebraic operations
- Four projection modes: Random, PCA, consensus, and contrastive — covering unsupervised through supervised settings
- Scalable graph building: Inverted prime index reduces pairwise comparison from O(N^2) to O(N x F x B)
- Topological model auditing: Compares how different embedding models structure semantic space using shortest-path discrepancy analysis
Known Limitations¶
- Hash overlap is not semantic containment: Subsumption reflects LSH bucket overlap, not genuine semantic containment
- Lossy projection: R^384 to Z is lossy; similar words may receive identical encodings
- Narrow useful k range: k=6--12 is practical; no principled selection method exists
- Low analogy accuracy: 2--10% for discovery (designed for verification, not discovery)
Citation¶
Paper¶
@article{ornelas2026prime,
author = {Ornelas Brand, J. Arturo},
title = {Triadic Neurosymbolic Engine: Prime Factorization as a
Neurosymbolic Bridge: Projecting Continuous Embeddings
into Discrete Algebraic Space for Deterministic Verification},
year = 2026,
month = mar,
doi = {10.5281/zenodo.19205805},
url = {https://doi.org/10.5281/zenodo.19205805}
}
Software¶
@software{ornelas2026triadic,
author = {Ornelas Brand, J. Arturo},
title = {Prime Factorization as a Neurosymbolic Bridge: Projecting
Continuous Embeddings into Discrete Algebraic Space
for Deterministic Verification (Repository)},
year = 2026,
month = feb,
version = {0.3.0},
doi = {10.5281/zenodo.18748671},
url = {https://doi.org/10.5281/zenodo.18748671}
}
Companion Work¶
This engine is the foundation of the Triadic research program:
| Project | What it does | DOI |
|---|---|---|
| TriadicGPT | 40M-parameter GPT that learns prime signatures end-to-end | |
| reptimeline | Tracks how discrete representations evolve during training |
License¶
Business Source License 1.1 (BUSL-1.1)
| Use case | Allowed |
|---|---|
| Individuals / personal projects | Free |
| Academic / research institutions | Free |
| Non-profit organizations | Free |
| For-profit companies (production use) | Requires participation agreement |
Change Date: 2030-03-21 — auto-converts to AGPL-3.0.
Contact: [email protected]