Test Case Generation from Large Language Mode - Systematic Mapping Studies (2023 - 2025) - Code Results

This spreadsheet is the primary coding artifact of a Systematic Mapping Study analyzing 128 primary studies (2023–2025) on the use of Large Language Models (LLMs) in software testing. Each row corresponds to one study and records the prompt engineering technique(s) used (coded PE1–PE11), the fine-tuning strategy if any (coded F1–F7), the specific LLM(s) evaluated (coded M1–M11), and the testing lifecycle stage(s)/activity(ies) addressed. Stages and activities follow the taxonomy proposed in the associated paper. There are four lifecycle stages: T1 (Requirements Management), T2 (Artifact Generation), T3 (Verification and Prioritization), and T4 (Report and Regression). These are decomposed into thirteen activities (A1–A13, e.g., requirements extraction, element localization, test/input/oracle generation, runtime verification, prioritization, repair, refactoring, validation, quality assessment, log analysis). Cells in the "Stages/Activities" column list only the activities the LLM performs through reasoning, generation, or decision-making within each study's pipeline. This excludes steps done by deterministic auxiliary tools (parsers, static analyzers, symbolic execution engines, coverage tools) or by manual human evaluation, even when those steps are part of the same proposed technique.