Request access to X-POSE
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
Important: Please ensure you are logged into Hugging Face using your official institutional or university email address. Requests from personal email accounts (e.g., Gmail, Yahoo) will be declined.
By requesting access to this dataset, you agree to the following terms:
- Research Use Only: This dataset is strictly for non-commercial, scientific, or academic research purposes.
- No Redistribution: You agree not to share, distribute, or re-publish this dataset (or any portion of it) with third parties.
- Full Liability: You assume full responsibility for your use of the data, including ensuring compliance with copyright laws and protecting the rights of third parties. The dataset providers hold no liability for infringements arising from misuse.
Log in or Sign Up to review the conditions and access this dataset content.
tags: - multimodal - multimodal misinformation - fact-checking size_categories: - 1K<n<10K
X-POSE: X Posts with Evidence
X-POSE is a multimodal fact-checking dataset designed to support research on evidence-based misinformation detection. It contains 5,704 image-text pairs from X (2017–2025), balanced between factually correct and misleading posts, with crowdsourced annotations from Community Notes. This repository includes raw web articles for evidence retrieval and 22,226 evidence excerpts extracted through a VLM-assisted pipeline.
The dataset was introduced in Evidence Triangulation for Multimodal Fact-Checking in the Wild, accepted at ECCV 2026.
Citation: Papadopoulos, S. I., Chrysidis, Z., Koutlis, C., Papadopoulos, S., & Petrantonakis, P. C. (2026, September). Evidence Triangulation for Multimodal Fact-Checking in the Wild. In European Conference on Computer Vision (pp. 209-227). Cham: Springer Nature Switzerland.
- Downloads last month
- 25