A TabPFN-Based Framework for Slope Stability Analysis Using Geometric Features and Shear Strength Parameters

Achieving efficient and accurate assessment of slope stability at a regional scale remains a significant research challenge. This is largely due to the complexity of geological conditions and the limited availability of field data. To address this issue, this study develops a standard two-dimensional (2D) slope database by integrating field investigations with numerical simulations. Geometric features and shear strength parameters including the thickness of the weathered soil layer ( h ), slope height ( H ), slope angles ( α and β ), and cohesion ( c ) are used to construct the mapping relationship with the slope factor of safety (FoS) using machine learning techniques. Four machine learning models, namely TabPFN, CatBoost, XGBoost, and Random Forest, are trained and evaluated to assess slope stability. The results demonstrate that the TabPFN model outperforms the other models in terms of predictive accuracy and robustness. Validation using data from 82 field investigation sites in Nagasaki City, Japan was conducted. TabPFN predictions of 2D FoS closely align with numerical simulation results. These findings underscore the strong potential of TabPFN for reliable assessment of slope stability with limited sample sizes. The proposed framework offers a promising machine-learning surrogate model for numerical simulation. It enhances both the efficiency and accuracy of slope stability assessment at the regional scale. • A standardized 2D slope database integrates field surveys with numerical modeling. • TabPFN provides accurate FoS predictions using geometric and shear strength inputs. • TabPFN outperforms XGBoost, RF, and CatBoost on small-sample prediction. • Validation at 82 Nagasaki sites confirms the reliability of the TabPFN framework.