A Dataset and Experimental Evaluation of a Parallel Conflict Detection Solution for Model-Based Diagnosis

This article presents a dataset and experimental evaluation of a parallelized variant of Junker’s QuickXPlain algorithm, designed to efficiently compute minimal conflict sets in constraint-based diagnosis tasks. The dataset includes performance benchmarks, conflict traces, and solution metadata for a wide range of configurable diagnosis problems based on real-world and synthetic CSP instances. Our parallel variant leverages multicore architectures to reduce computation time while preserving the completeness and minimality guarantees of QuickXPlain. All evaluations were conducted using reproducible scripts and parameter configurations, enabling comparison across different algorithmic strategies. The provided dataset can be used to replicate experiments, analyze scalability under varying problem sizes, and serve as a baseline for future improvements in conflict explanation algorithms. The full dataset, codebase, and benchmarking scripts are openly available and documented to promote transparency and reusability in constraint-based diagnostic systems research.