LATIDIA · Investigación

Physics2Weight: Arquitectura de red neuronal directa y generación de peso a partir de leyes físicas utilizando modelos de lenguaje grandes

arXiv: 2611.12954v1Tipo de anuncio: nuevo Resumen: Los avances recientes en los modelos de lenguaje grandes (LLM) han habilitado la programación de peso-espacio (WSP), que genera directamente pesos de red neuronal sin optimización basada en gradientes

La noticia

arXiv:2611.12954v1 Announce Type: new Abstract: Recent advances in large language models (LLMs) have enabled weight-space programming (WSP), which directly generates neural network weights without gradient-based optimization. However, extending WSP to physical problems remains challenging, particularly in achieving both solution accuracy and physical consistency. In this paper, we propose Physics2Weight (PtW), a framework that extends WSP to solving partial differential equations by directly generating neural network weights and editing architectures based on governing equations and initial and boundary conditions. PtW introduces physics-constrained prompting and physics-guided architecture editing, combined with iterative weight refinement from WSP, to improve solution accuracy and physical consistency without gradient-based optimization. Experiments on four PDEs, namely the heat,