aiml_fake_news-benchmark

What problem does it solve?

This Skill automates the benchmarking of fake-news detectors by generating and evaluating domain-diverse articles to assess robustness against adversarial content.

Core Features & Use Cases

• Automated multi-domain benchmarking across six misinformation domains (public_health, election_interference, financial_manipulation, military_disinfo, science_denial, fabricated_event).
• YAML frontmatter driven metadata for easy discovery and integration, plus prompt variants to support few-shot and zeroshot setups for evaluation.
• Use Case: Research teams can run repeatable detector evaluations, compare models, and quantify false positive/negative rates in controlled settings.

Quick Start

Run the benchmark on the target fake-news classifier using the provided dataset and prompts to obtain a domain-coverage report.