Spectral Generative Flow Models: A Physics-Inspired Replacement for Vectorized Large Language Models

TL;DR

Spectral Generative Flow Models (SGFMs) offer a physics-inspired, continuous field approach to generative modeling, replacing traditional attention mechanisms with spectral and stochastic dynamics for improved coherence and efficiency.

Contribution

SGFMs introduce a novel physics-inspired framework that unifies text and video generation through stochastic PDEs in the wavelet domain, departing from autoregressive and diffusion models.

Findings

Provides a unified field-theoretic ontology for text and video

Achieves computational efficiency via wavelet sparsity and scale separation

Enforces stability and coherence through constrained stochastic flows

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.