Replication Package for "Fine-Grained Requirement Conflict Detection and Resolution: A Dataset and LLM-Based Multi-Label Framework"

This replication package accompanies the study “Fine-Grained Requirement Conflict Detection and Resolution: A Dataset and LLM-Based Multi-Label Framework.” The package provides the materials required to reproduce the proposed methodology and the reported experimental results. It includes: (1) the datasets for fine-grained multi-label requirement conflict detection, including the independently sourced and manually verified test set; (2) implementations of the conflict detection and resolution framework; (3) the hybrid dataset-construction pipeline based on retrieval-augmented large language model generation and expert annotation; (4) prompts, configuration files, generation settings, and training hyperparameters; and (5) experimental scripts, evaluation procedures, and supporting documentation. The framework identifies conflicts between functional requirements at the level of events, agents, operations, inputs, outputs, and constraints, and uses the detected conflict types and extracted semantic tuples to support conflict-resolution recommendation generation. The archived version is the version used to produce the results reported in the accompanying manuscript.