FEAT is a linear-complexity foundation model designed specifically for extremely large-scale structured (tabular) data.
March 18, 2026
Original Paper
FEAT: A Linear-Complexity Foundation Model for Extremely Large Structured Data
arXiv · 2603.16513
The Takeaway
It solves the O(N^2) scaling limit of attention and the representation collapse of linear models in the structured data domain. For industries like finance or healthcare with massive cross-sample dependencies, FEAT offers 40x faster inference and linear scaling.
From the abstract
Structured data is foundational to healthcare, finance, e-commerce, and scientific data management. Large structured-data models (LDMs) extend the foundation model paradigm to unify heterogeneous datasets for tasks such as classification, regression, and decision support. However, existing LDMs face major limitations. First, most rely on sample-wise self-attention, whose O(N^2) complexity limits the sample count. Second, linear sequence models often degrade representations due to hidden-state co