Large Language Models Assistance in core Model-Driven Engineering activities

Dataset for the paper: "Large Language Models Assistance in core Model-Driven Engineering activities", accepted at MODELS 2026. The artifact documents the review protocol, database search results, screening decisions, included papers, extracted variables for analysis, and summary tables used to classify the current state of research on LLM-supported MDE. Purpose of the artifact The goal of this artifact is to make the literature review transparent, inspectable, and reusable. In particular, it allows readers and reviewers to: - inspect the systematic literature review protocol;- verify the search scope, inclusion criteria, and exclusion criteria;- trace the screening process from raw database results to the final set of included papers;- inspect the classification of papers according to MDE tasks and LLM-related dimensions;- reuse the extracted data for further analysis or comparison with future studies. Repository content. ├── LLM_MDE_Literature_Review Protocol and Analysis-final.xlsx└── README.md Workbook structure The Excel workbook is organized as follows. Sheet Purpose Protocol Defines the review objective, research question, scope, search string, inclusion/exclusion criteria, and screening summary. Query results (Scopus) and ECIC Raw Scopus search results with screening labels and metadata. Query results (IEEE) and ECIC Raw IEEE Xplore search results with screening labels and metadata. Query results (ACM) and ECIC Raw ACM Digital Library search results with screening labels and metadata. Set of paper included - 2nd stage Papers retained after the first screening stage and assessed during full-text screening. Final Analysis Final included papers and detailed extracted variables used for the analysis. RQ1 - modeling support Classification of included papers according to the MDE/modeling support dimensions RQ1 - LLM support Classification of included papers according to LLM support, roles, prompting, architecture, evaluation, and technology readiness RQ1- analysis Aggregated tables summarizing the distribution of tasks, LLM roles, and human roles. RQ1- heat Matrix-style summary tables used to analyze relationships between categories and supported activities Screening process The workbook records a two-stage screening process. Stage 1: database search and title/abstract screening Step Result Scopus results 1,246 IEEE Xplore results 1,187 ACM Digital Library results 998 Total raw records 3,431 Duplicates removed 259 Records excluded after title/abstract/keyword screening 2,988 Records remaining after Stage 1 184 Stage 2: full-text screening Step Result Papers assessed in full text 184 Papers remaining after full-text selection 96 Final number of included papers 96