Artifacts of the study: Evaluating LLM-Generated Unit Tests with Mutation Testing: ChatGPT vs DeepSeek

This repository provides the experimental artifacts for the paper "Evaluating LLM-Generated Unit Tests with Mutation Testing: ChatGPT vs DeepSeek", accepted at the Technical Track of the Brazilian Symposium on Software Quality (SBQS 2025). The artifacts include: Target Java classes from the Defects4J dataset. Prompts used to generate test suites with ChatGPT (GPT-4o) and DeepSeek V3. LLM-generated test suites (five runs per class/model). Base test classes used to generate fixed sets of mutants in PIT for reproducibility. Full PIT mutation testing reports with Mutation Score and Mutation Coverage results. Error logs from compilation and runtime failures. Cyclomatic complexity metrics for quantitative analysis. Consolidated spreadsheets with experimental results. The repository ensures transparency and reproducibility of the experiment, supporting future research on the use of Large Language Models (LLMs) in software testing.