dorsal/arxiv
View SchemaSpectral Generative Flow Models: A Physics-Inspired Replacement for Vectorized Large Language Models
| Authors | Andrew Kiruluta |
|---|---|
| Categories | |
| ArXiv ID | 2601.08893vv1 |
| URL | https://arxiv.org/abs/2601.08893 |
| License | http://creativecommons.org/licenses/by/4.0/ |
Abstract
We introduce Spectral Generative Flow Models (SGFMs), a physics-inspired alternative to transformer-based large language models. Instead of representing text or video as sequences of discrete tokens processed by attention, SGFMs treat generation as the evolution of a continuous field governed by constrained stochastic dynamics in a multiscale wavelet basis. This formulation replaces global attention with local operators, spectral projections, and Navier--Stokes-like transport, yielding a generative mechanism grounded in continuity, geometry, and physical structure. Our framework provides three key innovations: (i) a field-theoretic ontology in which text and video are unified as trajectories of a stochastic partial differential equation; (ii) a wavelet-domain representation that induces sparsity, scale separation, and computational efficiency; and (iii) a constrained stochastic flow that enforces stability, coherence, and uncertainty propagation. Together, these components define a generative architecture that departs fundamentally from autoregressive modeling and diffusion-based approaches. SGFMs offer a principled path toward long-range coherence, multimodal generality, and physically structured inductive bias in next-generation generative models.
{
"annotation_id": "08dded9f-8d4d-4197-9188-2a0afb0b86c6",
"date_created": "2026-02-17T05:53:20.168000Z",
"date_modified": "2026-02-17T05:53:20.168000Z",
"file_hash": "80f4e670338c43a3a5a7ba144f46dd85edfcaae9de0234dedf861d57f3fb9759",
"private": false,
"record": {
"abstract": "We introduce Spectral Generative Flow Models (SGFMs), a physics-inspired alternative to transformer-based large language models. Instead of representing text or video as sequences of discrete tokens processed by attention, SGFMs treat generation as the evolution of a continuous field governed by constrained stochastic dynamics in a multiscale wavelet basis. This formulation replaces global attention with local operators, spectral projections, and Navier--Stokes-like transport, yielding a generative mechanism grounded in continuity, geometry, and physical structure.\n Our framework provides three key innovations: (i) a field-theoretic ontology in which text and video are unified as trajectories of a stochastic partial differential equation; (ii) a wavelet-domain representation that induces sparsity, scale separation, and computational efficiency; and (iii) a constrained stochastic flow that enforces stability, coherence, and uncertainty propagation. Together, these components define a generative architecture that departs fundamentally from autoregressive modeling and diffusion-based approaches. SGFMs offer a principled path toward long-range coherence, multimodal generality, and physically structured inductive bias in next-generation generative models.",
"arxiv_id": "2601.08893",
"authors": [
"Andrew Kiruluta"
],
"categories": [
"cs.LG",
"cs.CL"
],
"license": "http://creativecommons.org/licenses/by/4.0/",
"title": "Spectral Generative Flow Models: A Physics-Inspired Replacement for Vectorized Large Language Models",
"url": "https://arxiv.org/abs/2601.08893",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "b677460d-8d2d-4ec9-9de9-cea0de823040",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}