Political Compass Evaluations
Definition
Standardized benchmark batteries mapping language model responses against multidimensional political scales.
Key Characteristics
- Automated Likert-scale scoring of propositional statements.
- Robustness testing against adversarial and flipped prompts.
- Clustering of distinct model families by normative consensus.
Applications
- Public model leaderboards
- Fairness auditing
Mentions in Source
- “Evaluates multi-turn model completions across economic and social axes.” — ai-political-compass