Recently, the "Doughnut Gynecology Large Model" developed by Yisheng Jiankang (Hangzhou) Life Science Technology Co., Ltd. successfully passed the national senior professional title examination for obstetrics and gynecology, marking the first vertical medical large model trained on DingTalk's enterprise-specific AI platform by a startup company to meet professional clinical examination standards. This breakthrough not only demonstrates the practical capabilities of specialized medical AI but also showcases DingTalk's strong technical support in serving enterprises to build their own large models.
Zhu Hong, Chief Technology Officer of DingTalk, introduced that the Doughnut model is the first professional vertical large model born from DingTalk's AI platform. Within a month of enterprise-level training, the accuracy rate quickly increased to 90.2%, and finally successfully passed the national senior professional title examination for obstetrics and gynecology, achieving a key leap from an AI prototype to a clinical-level application.
DingTalk stated that this achievement not only verifies the platform's ability to support enterprises in building their own large models across various industries, but also provides a practical path for AI development in more professional fields such as healthcare, education, and law. Yisheng Jiankang's success case has become a typical example of how enterprises can use DingTalk to build high-precision industry models, further promoting the large-scale implementation of vertical industry AI solutions.
The breakthrough of the Doughnut model also provides a new example for the standardized and professional application of AI technology in medical scenarios. It is expected to be piloted in areas such as obstetrics and gynecology knowledge Q&A, clinical decision support, and doctor training in the future.