Seonggil Lee

Rough-Torsional Adaptive Sparse Multi-Axis Scaling (RT-ASMAS) Transcending MoE and Dense Scaling Laws via UROA, STCT, and RQIG

The paradigm of scaling Large Language Models relies heavily on Mixture of Experts (MoE) to decouple parameter count from active computational cost. However, standard token-choice MoE routing suffers from load imbalance, while pure dense scaling faces catastrophic diminishing ret...

Structured State Space Duality (SSD) Super-Elevated: Rough Topological Operators, Complex Torsion, and Information Geometry

Standard Transformer Self-Attention suffers from quadratic temporal and space complexity O(n^2) [5]. The Structured State Space Duality (SSD) framework resolves this bottleneck by proving an exact mathematical equivalence between 1-semiseparable matrix transformations and selecti...