Versor: A Geometric Sequence Architecture

Researchers introduce Versor, a novel sequence architecture using Conformal Geometric Algebra that significantly outperforms Transformers with 200x fewer parameters and better interpretability. The architecture achieves superior performance on various tasks including N-body dynamics, topological reasoning, and standard benchmarks while offering linear temporal complexity and 100x speedup improvements.

• →Versor uses Conformal Geometric Algebra instead of traditional linear operations, requiring 200x fewer parameters than Transformers while achieving better performance.
• →The architecture demonstrates superior zero-shot scale generalization (0.993 vs 0.070 MCC compared to ViT) and maintains stable predictions on out-of-distribution tests.
• →Custom Clifford kernels provide over 100x cumulative speedup with 1.05ms per-step latency, outperforming optimized Transformer baselines.
• →Versor offers interpretable attention that decomposes into proximity and orientational components, improving model explainability.
• →The system features linear O(L) temporal complexity for dynamical systems and O(L²) complexity for global relational modeling.