TruthStance: An Annotated Dataset of Conversations on Truth Social

Dataset Contents This release and each of its prior versions (v1, v2, v3) include the following files: - ground_truth.xlsx (v3) Human annotations for the argument mining and stance detection tasks. This file contains two sheets (one per task), each with 750 labeled examples (1,500 total annotations). Refer to the "Ground Truth label" column for the final label. Each row also includes the post/comment ID/URL and its text content. - new_truths-all.csv (v2) A collection of 523,360 unique comments scraped from Truth Social conversation threads (2022–2025). - new_authors-all.csv (v1) Metadata for 90,563 unique authors appearing in the scraped posts and comments. Code used for data collection and preprocessing is available here: Truth Social Conversational Data If using this dataset for research, please cite our paper: Ameen, Fathima, et al. "TruthStance: An Annotated Dataset of Conversations on Truth Social." arXiv e-prints (2026): arXiv-2602.