mlsanityData to 7 Oct 2026

Separable Neural Architectures as Physical World Models: from Mathematical Theory to Applications

arXiv 2606.14934 ↗First posted 12 Jun 2026cs.LG cs.AIcited by 0

On the map: World models → Latent predictive world models

Abstract (arXiv)

This work introduces the Separable Neural Architecture (SNA), a function representational class combining neural approximation with tensor decomposition. The SNA decouples localized coordinate functions (atoms) from global interactions governed by a sparse, low-rank interaction object. This architecture possesses a compact and smooth inductive bias well-suited for solving partial differential equations (PDEs). When viewed as a Galerkin trial space under the variational SNA (VSNA) framework, the formulation satisfies classical variational guarantees under Lax-Milgram: well-posedness, quasi-optimality, convergence, and stability. In high-dimensional spatiotemporal--parametric PDEs, the VSNA mitigates the curse of dimensionality by scaling algebraically rather than exponentially. Exploiting an entirely factorized, tensor-native alternating least squares (ALS) optimization framework reduces this cost to linear in dimension. The VSNA is validated across elliptic, hyperbolic, and parabolic systems, demonstrating close alignment with predicted algebraic and spectral scaling rates. We showcase the SNA as a "solve once, query anywhere" physical world model via two engineering case studies: a 7D parametric manufacturing simulation and an experimental thermal-to-property inversion pipeline for Inconel 718. The VSNA executes a 1,000,000-query Monte Carlo sweep in 102s on a standard laptop CPU, yielding a 150,000x speedup over a full-grid finite element baseline hosted on an NVIDIA A100 GPU. It further enables real-time generative inverse-mode reconstructions under 100ms. These results demonstrate that the SNA serves as a compact mathematical substrate for continuous parameter manifolds to enable real-time inversion, optimization loops, and rapid uncertainty propagation.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Latent predictive world models, one of 11 directions in World models. Builds on 14 of its 59 references that are in the map. First posted on arXiv 12 Jun 2026.

Builds on

14 of 59 references.

References · 14

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  2. 2016XGBoost: A Scalable Tree Boosting Systemcited by 5,120
  3. 2020Fourier Neural Operator for Parametric Partial Differential Equationscited by 2,806
  4. 2019DeepONet: Learning nonlinear operators for identifying differential equations based on the universal approximation theorem of operatorscited by 1,965
  5. 2018World Modelscited by 1,444
  6. 2022Scientific Machine Learning through Physics-Informed Neural Networks: Where we are and What's nextcited by 603
  7. 2021A physics-informed variational DeepONet for predicting the crack path in brittle materialscited by 98
  8. 2022An Introduction to Kernel and Operator Learning Methods for Homogenization by Self-consistent Clustering Analysiscited by 6
  9. 2025A Kernel-based Resource-efficient Neural Surrogate for Multi-fidelity Prediction of Aerodynamic Fieldcited by 3
  10. 2025A Unified Generative-Predictive Framework for Deterministic Inverse Designcited by 3
  11. 2025Explainable Hierarchical Deep Learning Neural Networks (Ex-HiDeNN)cited by 2
  12. 2026Leviathan: Decoupling Input and Output Representations in Language Modelscited by 2
2 more
  1. 2026Agile Reinforcement Learning through Separable Neural Architecture and Applicationscited by 2
  2. 2024Convolutional Hierarchical Deep Learning Neural Networks-Tensor Decomposition (C-HiDeNN-TD): a scalable surrogate modeling approach for large-scale physical systemscited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Merging Memory and Space: A State Space Neural Operator
  2. 2025Towards a Foundation Model for Partial Differential Equations Across Physics Domains
  3. 2026EvoPINN: Agentic Discovery of Executable Algorithms for Physics-Informed Neural Networks
  4. 2026Learning Physical Operators using Neural Operators
  5. 2026Geometry-Preserving Neural Architectures on Manifolds with Boundary
  6. 2026When Attention Beats Fourier: Multi-Scale Transformers for PDE Solving on Irregular Domains
  7. 2025Wave-PDE Nets: Trainable Wave-Equation Layers as an Alternative to Attention
  8. 2026Learning to Discretize: Diffusion-Based Adaptive Mesh with Spectral Guidance