Supplementary Material for "Adapting Foundation Models for Generating Institutional Documents in Legal and Public Security Contexts: A Systematic Review"

Supplementary Material for the Systematic Literature Review "Adapting Foundation Models for Legal Document Generation in Regulated Institutional Contexts: A Systematic Review" Overview This repository contains the supplementary material for the systematic literature review (SLR) titled "Adapting Foundation Models for Generating Institutional Documents in Legal and Public Security Contexts: A Systematic Review", authored by [double-blind]. The review was conducted following the PRISMA 2020 guidelines (Page et al., 2021) and the systematic review protocol proposed by Kitchenham and Charters (2007), covering studies published between January 2021 and December 2025. The search was performed across four indexed databases (Scopus, Web of Science, IEEE Xplore, and ACM Digital Library) and complemented by backward snowballing on the references of the included studies. The purpose of this deposit is to ensure the transparency, reproducibility, and auditability of the selection, extraction, and quality assessment processes, in accordance with the recommendations of Crawford (2025) and the PRISMA 2020 statement. Repository Contents This deposit includes four files that document the full methodological process of the review: 1. MappingSelectionProcess.pdf This document provides a textual description of the study selection process, from the initial database search to the final corpus of included studies. It details each step of the PRISMA 2020 workflow (identification, deduplication, screening, eligibility) and explains how the exclusion criteria (EC1–EC5) were applied at each stage. The document cross-references the two accompanying spreadsheets (MappingSLR.xlsx and MappingSnowballing.xlsx), which contain the detailed records of each selection decision. 2. MappingSLR.xlsx This spreadsheet contains the detailed record of the study selection process based on the automated search in the four indexed databases. It is organized into five sheets, each corresponding to one stage of the PRISMA workflow: P1 – Initial Extraction from Databases: complete list of the 200 records retrieved from the four databases (Scopus, Web of Science, IEEE Xplore, ACM Digital Library), with bibliographic metadata. P2 – Duplicates Removed: list of the 51 duplicate records identified and removed (20 via automatic detection by Rayyan and 31 via manual inspection). P3 – Removed During Screening: list of the 112 records excluded during the title and abstract screening, with the specific exclusion criterion (EC1–EC4) applied to each. P4 – Removed During Full-Text Screening: list of the 14 studies excluded after full-text reading, with the specific exclusion criterion (EC1, EC3, EC4, EC5) applied to each. P5 – Final Screening: list of the 23 studies included in the final corpus via database search. 3. MappingSnowballing.xlsx This spreadsheet contains the detailed record of the backward snowballing process conducted on the references of the 23 studies included via database search, following the recommendations of Wohlin (2014). It is organized into three sheets: P1 – Initial Extraction: list of the 14 candidate studies identified through reference analysis. P2 – Removed During Full-Text Screening: list of the 5 studies excluded after full-text reading, with the specific exclusion criterion applied to each. P3 – Final Screening: list of the 9 studies included in the final corpus via snowballing. The combined result of MappingSLR.xlsx (23 studies) and MappingSnowballing.xlsx (9 studies) yields the final corpus of 32 primary studies analyzed in the review. 4. DataExtraction&QualityAssessment.xlsx This spreadsheet contains the systematic data extraction form and the quality assessment of the 32 included primary studies. It is organized into two sheets: Data Extraction: structured extraction of technical, methodological, and contextual information from each study, organized into six analytical dimensions: (i) study identification, (ii) model characteristics, (iii) data engineering, (iv) ethical aspects, (v) evaluation, and (vi) synthesis. The fields were defined a priori based on the four research questions (RQ1–RQ4) of the review and on the quality criteria (QA1–QA7). Quality Assessment: disaggregated scoring of each study against the seven quality criteria (QA1–QA7) defined in Section 2.5 of the manuscript, using a trichotomous scale (0; 0.5; 1), with a maximum score of 7 points per study. Intended Use This material is intended for: Verification and auditability of the selection, extraction, and quality assessment decisions reported in the manuscript Replication of the systematic review workflow by other researchers Reuse of the extracted data for secondary analyses, meta-research, or updates of the review Reference in systematic mapping studies or scoping reviews on adjacent topics Users are encouraged to consult the main manuscript for the methodological context, the research questions, and the synthesis of findings. How to Cite If you use this material in your research, please cite both the main article and this supplementary dataset. Main article Double Bind (2026). Adapting Foundation Models for Generating Institutional Documents in Legal and Public Security Contexts: A Systematic Review. Inteligencia Artificial, [volume], [number], [pages]. DOI: [to be assigned upon publication]. This supplementary dataset Double Bind (2026). Supplementary material for the systematic literature review "Adapting Foundation Models for Generating Institutional Documents in Legal and Public Security Contexts" [Data set]. Zenodo. DOI: 10.5281/zenodo.19476784. BibTeX @article{2026adapting, author = double-blind, title = {Adapting Foundation Models for Legal Document Generation in Regulated Institutional Contexts: A Systematic Review}, journal = {Inteligencia Artificial}, year = {2026}, volume = {}, number = {}, pages = {}, doi = {10.5281/zenodo.19476784} } @dataset{2026supplementary, author = double-blind, title = {Supplementary material for the systematic literature review "Adapting Foundation Models for Legal Document Generation in Regulated Institutional Contexts: A Systematic Review"}, year = {2026}, publisher = {Zenodo}, doi = {10.5281/zenodo.19476784} } License This supplementary material is released under the Creative Commons Attribution 4.0 International License (CC BY 4.0). You are free to share and adapt the material for any purpose, including commercially, provided that proper attribution is given to the original authors. The full terms of the license are available at https://creativecommons.org/licenses/by/4.0/. Authors and Contact double-blind References Cited in This README Kitchenham, B., & Charters, S. (2007). Guidelines for performing Systematic Literature Reviews in Software Engineering. Keele University and Durham University. Technical Report EBSE-2007-01. Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D., et al. (2021). The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ, 372, n71. https://doi.org/10.1136/bmj.n71 Wohlin, C. (2014). Guidelines for snowballing in systematic literature studies and a replication in software engineering. In Proceedings of the 18th International Conference on Evaluation and Assessment in Software Engineering (EASE '14). https://doi.org/10.1145/2601248.2601268