LABHDUFBA/social-rag: v1.0
Abstract
v1.0 — Supplementary Materials for Social-RAG manuscript (Zenodo) This is the v1.0 archival release of the social-rag repository, created to accompany the Social-RAG manuscript and to support transparency, auditability, and (where feasible) reproducibility of the reported analyses. What's included Documentation docs/ with repository overview, data governance, and ethics/safety guidance. Appendices Theme-specific methodological reports for Vaccine and Rouanet (appendices/), including Portuguese variants. Prompts System prompts (theme-conditioned) and the LLM-as-judge prompt (prompts/), including Portuguese variants. Evaluation artifacts Criteria definition (Table 2 mapping), test question sets (factual + hermeneutic), and raw model outputs (evaluation/criteria, evaluation/questions, evaluation/model-outputs). LLM-as-judge results with full judge/run logs, consolidated tables, and analysis-ready CSVs including per-judge means (evaluation/judge-results). Figures (fully traceable) Source data (figures/data/), plotting scripts (figures/src/), and exported manuscript figures in PNG/SVG (figures/imgs/). Public samples Small anonymized CSV samples and the sampling script (samples/). Reproducibility All figures referenced in the manuscript are backed by: explicit source data (figures/data/), generation scripts (figures/src/), and exported outputs (figures/imgs/). Minimal reproduction steps and scope limitations are documented in: docs/supplementary-overview.md docs/data-governance.md docs/ethics-and-safety.md Data governance & access This release does not include the full raw message-level corpora used in the study. Only sanitized samples and aggregated/derived evaluation artifacts are shared to reduce privacy and re-identification risks. See docs/data-governance.md and docs/ethics-and-safety.md. Citation Please use the metadata in CITATION.cff when citing this repository (Zenodo record will provide the DOI for this archived release). License See LICENSE.md for licensing terms.
Fields of Study & Subjects
Topic Modeling
Misinformation and Its Impacts
Sentiment Analysis and Opinion Mining
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 10 February 2026
- Language
- EN