Automating Detection and Root-Cause Analysis of Flaky Tests in Quantum Software — Supplementary Dataset

This package provides the supplementary dataset and prompt templates accompanying the research study, "Automating Detection and Root-Cause Analysis of Flaky Tests in Quantum Software." The study investigates an automated approach to identifying and performing root-cause analysis of flaky tests in quantum computing software systems using Large Language Models (LLMs). Contents dataset.zip — Curated collection of flaky and non-flaky test cases extracted from open-source quantum computing repositories (Qiskit, Microsoft Quantum, NetKet, TensorFlow Quantum). Organised at two granularity levels: Full/ — Complete source files containing the faulty code. Method/ — Extracted method-level code snippets. Each test case includes issue/PR descriptions, faulty (.bug) and fixed (.fix) source code, and patch diffs (code.diff). prompt.zip — LLM prompt templates used for three prompting strategies (zero-shot, cosine-similarity-matched examples, and cosine-similarity-matched examples with preserved comments), covering three sequential research questions: flakiness classification (RQ3), code-informed reclassification (RQ4), and root-cause identification (RQ5). blueprint.xlsx — Master spreadsheet listing all test cases with flaky/non-flaky labels, root-cause categories, fix-pattern types, and pre-computed cosine-similarity example mappings. Structure Data is organised by organization and repository following the pattern:{Full,Method}/{Flaky,Non_Flaky}/{owner}/{repo}/{issue_or_pr}/ Usage To reproduce the experiments, follow the prompt conversation flow documented in prompt/README.md. Each test case is evaluated in a three-turn LLM conversation (RQ3 → RQ4 → RQ5) within a single session thread. Refer to blueprint.xlsx for ground-truth labels and cosine-similarity mappings. License & Citation If you use this dataset, please cite the accompanying paper.