"DualFL: Bridging the Granularity Gap for Reliable Method and Statement Fault Localization"

"This dataset contains the complete pre-processed benchmarks and the associated data generation pipeline for the research on \"Dual-Granularity for Fault Localization\". It provides the necessary data to reproduce the experiments on a dual-granularity graph neural network model that processes program information at both method-level and statement-level.The dataset is organized into two main parts:data\/: The fully processed graph embeddings for the complete Defects4J 1.2 suite (including Chart, Lang, Time, Math, Mockito, and Closure projects). These .pkl files are ready for direct use in model training and evaluation.Data_work_flow\/: A comprehensive three-stage data processing pipeline scripts (covering Linux and Windows environments) that transform raw Defects4J artifacts into enriched graph representations using GraphCodeBERT and Node2Vec."