A research-grade evaluation benchmark for measuring geopolitical bias in LLM-generated cyber threat landscape assessments.
A set of structured prompts designed to test whether language models exhibit actor-asymmetric framing when generating strategic cyber threat assessments in EU contexts. Each prompt describes a cyber incident in a specific critical infrastructure sector, paired with an attribution condition… See the full description on the dataset page:
https://huggingface.co/datasets/eromang/eu-cyber-llm-benchmark-prompts.
