Back to Home
Research.
My academic publications, pre-prints, and explorations into the mathematics and architecture of emerging AI models.
Published2026
SBLA-Cortical Unifying 158-Bit Ternary Synapses O1 Dendritic Fast-Weight Memory and Homeostatic Spiking Predictive Coding for Ultra-Efficient Foundation Models
Sumit Shrivastava
Abstract
"This paper introduces a novel approach to ultra-efficient foundation models by unifying 158-bit ternary synapses with dendritic fast-weight memory. By leveraging homeostatic spiking predictive coding, the proposed SBLA-Cortical architecture significantly reduces computational overhead while maintaining robust learning capabilities."
Document Preview