AnesBench introduces the first comprehensive benchmark for evaluating large language models' reasoning capabilities in the specialized field of anesthesiology.

• Assesses reasoning across three levels: factual knowledge, clinical reasoning, and advanced problem-solving
• Provides cross-lingual evaluation materials for broader applicability
• Identifies key factors influencing LLM performance in specialized medical domains
• Establishes a foundation for improving AI safety in critical healthcare applications

This research addresses the critical gap between general medical AI capabilities and specialized clinical needs, providing a structured framework to evaluate and improve LLM safety before deployment in high-stakes medical environments.

AnesBench: Multi-Dimensional Evaluation of LLM Reasoning in Anesthesiology