mlsanityData to 14 Oct 2026

Interpreting CFD Surrogates through Sparse Autoencoders

arXiv 2507.16069 ↗First posted 21 Jul 2025cs.CE cs.LGcited by 3

On the map: Mechanistic interpretability → Sparse autoencoder training and evaluation

Abstract (arXiv)

Learning-based surrogate models have become a practical alternative to high-fidelity CFD solvers, but their latent representations remain opaque and hinder adoption in safety-critical or regulation-bound settings. This work introduces a posthoc interpretability framework for graph-based surrogate models used in computational fluid dynamics (CFD) by leveraging sparse autoencoders (SAEs). By obtaining an overcomplete basis in the node embedding space of a pretrained surrogate, the method extracts a dictionary of interpretable latent features. The approach enables the identification of monosemantic concepts aligned with physical phenomena such as vorticity or flow structures, offering a model-agnostic pathway to enhance explainability and trustworthiness in CFD applications.

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Placed in Sparse autoencoder training and evaluation, one of 13 directions in Mechanistic interpretability. Builds on 12 of its 22 references that are in the map. First posted on arXiv 21 Jul 2025.

Builds on

12 of 22 references.

Closest ancestors · 4

Same direction, newest first

  1. 2024Enhancing Neural Network Interpretability with Feature-Aligned Sparse Autoencoders2 Nov
  2. 2024Decoding Dark Matter: Specialized Sparse Autoencoders for Interpreting Rare Concepts in Foundation Models1 Nov
  3. 2024Efficient Dictionary Learning with Switch Sparse Autoencoders10 Oct
  4. 2024Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 29 Aug

Other references · 8

  1. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,420
  2. 2020Learning Mesh-Based Simulation with Graph Networkscited by 868
  3. 2024Scaling and evaluating sparse autoencoderscited by 553
  4. 2024Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoderscited by 300
  5. 2013k-Sparse Autoencoderscited by 215
  6. 2024Improving Dictionary Learning with Gated Sparse Autoencoderscited by 179
  7. 2025Universal Sparse Autoencoders: Interpretable Cross-Model Concept Alignmentcited by 62
  8. 2025Interpretable and Testable Vision Features via Sparse Autoencoderscited by 25

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026Sparse Autoencoders as a Steering Basis for Phase Synchronization in Graph-Based CFD Surrogatescited by 1
  2. 2026Courant: a State-Adaptive Perceiver-Based Neural Surrogate with Local Support and Interpretable Field Decompositioncited by 0
  3. 2026PhysSAE: Mechanistic Interpretability of PINNs with Sparse Autoencoderscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026FluidFlow: a flow-matching generative model for fluid dynamics surrogates on unstructured meshes
  2. 2025A Mixture of Experts Gating Network for Enhanced Surrogate Modeling in External Aerodynamics
  3. 2026Inpainting physics: self-supervised learning for context-driven fluid simulation
  4. 2023Uncertainty-aware Surrogate Models for Airfoil Flow Simulations with Denoising Diffusion Probabilistic Models
  5. 2025FoilDiff: A Hybrid Transformer Backbone for Diffusion-based Modelling of 2D Airfoil Flow Fields
  6. 2025Learning Distributions of Complex Fluid Simulations with Diffusion Graph Networks
  7. 2025Guiding diffusion models to reconstruct flow fields from sparse data
  8. 2026Multi-fidelity graph-based neural networks architectures to learn Navier-Stokes solutions on non-parametrized 2D domains