LLMTSDiff Replication Package- Benchmarks for Test Smell Diffusion (LLM, SBST, and Human-Written Tests)

Description This archive contains the complete replication package for the study: On the Diffusion of Test Smells in LLM-Generated Unit Tests The artifact includes three benchmarks: Benchmark-1: LLM-generated unit tests from https://dl.acm.org/doi/pdf/10.1145/3691620.3695330 Benchmark-2: LLM-generated unit tests (alternative configuration) from https://arxiv.org/pdf/2409.17561 Human-Written Benchmark: Test suites collected from CATLM, Defects4J, and SF110 Each benchmark is provided in a minimal JSONL/CSV schema including: stable test identifiers dataset metadata generator metadata SHA1 hashes of test and production files ingestion traceability information The JSONL format contains full source code.The CSV format contains a flat metadata representation. Due to its size, the Human-Written benchmark is archived exclusively in this Zenodo release. All scripts used for dataset construction and analysis are available in the GitHub replication repository. Structure Benchmark-1/Benchmark-2/Human_Written/ Reproducibility The GitHub (https://anonymous.4open.science/r/LLMTSDiff-B341) repository contains: all preprocessing scripts test smell analysis pipeline correlation and non-linear analysis scripts The DOI of this archive ensures long-term reproducibility and versioned artifact citation.