Medical artificial intelligence is accelerating towards more specialized and safer directions. According to the Shanghai Artificial Intelligence Laboratory, the MedBench, a Chinese medical large model evaluation benchmark, has recently officially launched its 5.0 version with a comprehensive upgrade.

This upgrade has gathered top research forces in the field of medical AI in China. The research team from the Shanghai Artificial Intelligence Laboratory collaborated with the Guangzhou Laboratory, the team led by Academician Zhong Nanshan from the First Affiliated Hospital of Guangzhou Medical University, and the team led by Academician Ge Junbo from the Zhongshan Hospital of Fudan University to participate in the construction work. The new version focuses on building a specialized evaluation system and has pioneered a medical atomic skill evaluation paradigm, which can achieve a comprehensive correction of AI model hallucinations, continuously reinforcing the safety defense of medical AI.

As the first native medical vertical evaluation system in China, MedBench is also designated by the Shanghai Municipal Committee of Cyberspace Administration and the Shanghai Health Commission as a pre-evaluation carrier for medical AI application technology, playing a crucial role in promoting the standardization and high-quality development of medical large models.