This research introduces an explainable transformer architecture that predicts soil liquefaction risk by integrating multiple data streams, offering a breakthrough for construction safety in seismic regions.

• Combines three data streams: spectral seismic encoding, soil stratigraphy tokenization, and site-specific features
• Processes data from 165 case histories across 11 major earthquakes
• Employs Fast Fourier Transform for seismic waveform encoding
• Provides interpretable results through SHapley Additive exPlanations (SHAP)

This innovation enables more accurate prediction of earthquake-induced soil liquefaction, a critical advancement for engineering safer structures in seismic zones. The model's interpretability allows engineers to understand key risk factors, potentially saving lives and infrastructure through improved construction planning.

Original Paper: Evaluating and Explaining Earthquake-Induced Liquefaction Potential through Multi-Modal Transformers