Uses SMT solvers to formally verify the physical consistency of tree-based ML models across their entire input domain.
March 19, 2026
Original Paper
Formal verification of tree-based machine learning models for lateral spreading
arXiv · 2603.16983
The Takeaway
It challenges the status quo of 'black-box' hazard prediction by proving whether a model obeys physical laws (like monotonicity or safety thresholds). This allows practitioners to iteratively apply constraints until a model is provably physically consistent, which is critical for safety-critical civil engineering applications.
From the abstract
Machine learning models for geotechnical hazard prediction can achieve high accuracy while learning physically inconsistent relationships from sparse or biased training data. Current remedies (post-hoc explainability, such as SHAP and LIME, and training-time constraints) either diagnose individual predictions approximately or restrict model capacity without providing exhaustive guarantees. This paper encodes trained tree ensembles as logical formulas in a Satisfiability Modulo Theories (SMT) sol