Supplementary dataset for a knowledge-guided LLM-GIS framework for relative landslide risk screening in Hong Kong

This dataset provides the numerical records supporting the study “Knowledge-guided LLM-GIS framework for relative landslide risk screening and risk-informed management in Hong Kong”. It includes the complete 100-step Mode A candidate-formulation sequence, the 100-step Mode B weighting-refinement trajectory, evaluation metrics, ROC data, sensitivity analyses, factor weights, normalisation settings, sample-level screening scores, and regional risk-classification thresholds. These data are provided to support transparency, auditability, and reproducibility of the deterministic GIS calculations and the LLM-assisted rule-generation workflow.