Behavioral Coherence: A Method for Sensitive-Domain LLM Evaluation

This paper introduces a method for evaluating LLMs in sensitive domains, which could improve safety assessments and governance of AI systems.

Auto-scored: It provides a real but partial step toward better evaluation of LLMs in sensitive domains, aiding risk assessment.