Versor: A Geometric Sequence Architecture Enhanced Scale Generalization and Interpretability via Conformal Algebra
Trương Minh Huy, Edward Hirst · 2026
Read Versor: A Geometric Sequence Architecture Enhanced Scale Generalization and Interpretability via Conformal Algebra by Trương Minh Huy, Edward Hirst (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-10195.txt.
Description
Sequence models usually rely on linear steps that hide structure, but shape itself can carry the links. A 2026 paper by Truong Minh Huy and Edward Hirst introduces Versor, a sequence design using conformal geometric algebra instead of usual linear steps. States embed in a shape space and evolve via rotor moves, showing 3D matching links by design without added shape coding.
Versor beats rivals, graph networks, and shape rivals on chaotic N-body motion, topological puzzles, and standard mixed-media tests. It uses two hundred times fewer weights than rivals, splits focus clearly into nearness and angle parts, and transfers zero-shot across scales where vision models collapse. Custom kernels run over a hundred times faster, cutting per-step latency to about a thousandth of a second.
Catalog record
| CREATOR | Trương Minh Huy, Edward Hirst |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 98 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-10195