TabPFN Extensions for Interpretable Geotechnical Modelling

TL;DR

This paper evaluates the application of TabPFN, a tabular foundation model, for interpretable geotechnical site characterization, demonstrating its effectiveness in classification and imputation tasks with uncertainty analysis.

Contribution

It presents a comprehensive evaluation workflow for TabPFN in geotechnical modeling, emphasizing interpretability and uncertainty quantification without retraining.

Findings

Embeddings show label-consistent clay/sand grouping.

Iterative imputation reduces RMSE across targets, with TabPFN performing best on four.

SHAP attributions align with known geotechnical correlations.

Abstract

Geotechnical site characterisation relies on sparse, heterogeneous borehole data, where uncertainty quantification and interpretability matter as much as predictive accuracy. We evaluate TabPFN~\citep{Hollmann2025}, a tabular foundation model, and its \texttt{tabpfn-extensions} library on two geotechnical tasks: (1) soil-type classification from N-value and shear-wave velocity data as a controlled illustrative case, and (2) iterative imputation of five mechanical parameters ($s_\mathrm{u}$, $E_{\mathrm{u}}$, ${\sigma'}_\mathrm{p}$, $C_\mathrm{c}$, $C_\mathrm{v}$) in BM/AirportSoilProperties/2/2025. Without retraining, we apply cosine-similarity analysis to TabPFN embeddings, visualise predictive distributions, and compute SHAP attributions. On the regression benchmark we compare TabPFN with mean imputation, linear regression, random forests, XGBoost, and HBM; introduce a proxy decomposition of predictive uncertainty across context-perturbation classes; and propagate marginal $C_\mathrm{c}$ and ${\sigma'}_\mathrm{p}$ distributions through a one-dimensional consolidation model to obtain the reliability index $\beta$ and serviceability exceedance probability $P_\mathrm{f}$. Embeddings exhibit label-consistent Clay/Sand grouping; iterative imputation reduces RMSE for all five targets, with TabPFN lowest on four; SHAP attributions are consistent with the Skempton compression-index correlation and the inverse preconsolidation-pressure-water-content dependence; the within-posterior component is largest in the proxy decomposition. We position the contribution as a worked evaluation workflow that may complement established methods for data-scarce geotechnics, not as algorithmic innovation.