12 papers
p-Spin Glass Network Efficient Single-Batch Continual Learning
Vladimer Khasia
Modern sequence models heavily rely on massive memory footprints and large-batch stochastic optimization, barriers that restrict sample efficiency and continual learning. We introd…
HoloByte: Continuous Hyperspherical Distillation for Tokenizer-Free Modeling
Vladimer Khasia
Sequence modeling universally relies on discrete subword tokenization to circumvent the computational intractability of native byte-level attention. However, thi…
BASIS: Balanced Activation Sketching with Invariant Scalars for "Ghost Backpropagation"
Vladimer Khasia
The activation memory required for exact backpropagation scales linearly with network depth, context length, and feature dimensionality, forming an O(L * BN ) spatial bottleneck (w…
Beyond Attention: True Adaptive World Models via Spherical Kernel Operator
Vladimer Khasia
The pursuit of world model based artificial intelligence has predominantly relied on projecting high-dimensional observations into parameterized latent spaces, wherein transition d…
Hybrid Dual-Path Linear Transformations for Efficient Transformer Architectures
Vladimer Khasia
Standard Transformer architectures rely heavily on dense linear transformations, treating feature projection as a monolithic, full-rank operation. We argue that this formulation is…
HAS-VQ: Hessian-Adaptive Sparse Vector Quantization for High-Fidelity LLM Compression
Vladimer Khasia
Post-training quantization is essential for deploying Large Language Models (LLMs) on resource-constrained devices. However, standard integer quantization (e.g., INT4) fundamentall…
Spectral-Window Hybrid (SWH)
Vladimer Khasia
Scaling sequence modeling to extreme contexts requires balancing computational efficiency with representational expressivity. While Transformers provide precise retrieval via the a…
Dynamic Subspace Composition: Efficient Adaptation via Contractive Basis Expansion
Vladimer Khasia
Mixture of Experts (MoE) models scale capacity but often suffer from representation collapse and gradient instability. We propose Dynamic Subspace Composition (DSC), a framework th…
DeepVekua: Geometric-Spectral Representation Learning for Physics-Informed Fields
Vladimer Khasia
We present DeepVekua, a hybrid architecture that unifies geometric deep learning with spectral analysis to solve partial differential equations (PDEs) in sparse data regimes. By le…
The Adaptive Vekua Cascade: A Differentiable Spectral-Analytic Solver for Physics-Informed Representation
Vladimer Khasia
Coordinate-based neural networks have emerged as a powerful tool for representing continuous physical fields, yet they face two fundamental pathologies: spectral bias, which hinder…
The Vekua Layer: Exact Physical Priors for Implicit Neural Representations via Generalized Analytic Functions
Vladimer Khasia
Implicit Neural Representations (INRs) have emerged as a powerful paradigm for parameterizing physical fields, yet they often suffer from spectral bias and the computational expens…
Primal: A Unified Deterministic Framework for Quasi-Orthogonal Hashing and Manifold Learning
Vladimer Khasia
We present Primal, a deterministic feature mapping framework that harnesses the number-theoretic independence of prime square roots to construct robust, tunable vector representati…