Show HN: CRED-1 – Open domain credibility dataset for on-device pre-bunking
CRED-1 is an open domain credibility dataset designed to assess the reliability of online content. It includes credibility scores for 2,672 domains known for mis/disinformation and is intended for on-device deployment. The dataset combines various signals to provide a comprehensive evaluation of domain credibility.
- ▪CRED-1 provides credibility scores ranging from 0.0 to 1.0 for 2,672 domains.
- ▪The dataset is fully reproducible and can be rebuilt using a Python pipeline.
- ▪It is designed for privacy-preserving on-device client-side deployment without server calls.
Opening excerpt (first ~120 words) tap to expand
CRED-1: Open Domain Credibility Dataset CRED-1 is an open, reproducible domain-level credibility dataset combining multiple openly-licensed source lists with computed enrichment signals. It provides credibility scores for 2,672 domains known to publish mis/disinformation, conspiracy theories, or other unreliable content. 🎓 Presented at ACM WebSci 2026 (Braunschweig). Landing page: aloth.github.io/agentic-ai-information-integrity/cred-1. First production integration: Trackless Links for iOS and macOS, with free codes for readers and attendees: gutscheinhub.de/ratgeber/trackless-links-cred-1-acm-websci-2026. Paper: A. Loth, M. Kappes, and M.-O. Pahl, "CRED-1: An Open Multi-Signal Domain Credibility Dataset for Automated Pre-Bunking of Online Misinformation," Preprint, 2026.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at GitHub.