英语轻松读发新版了,欢迎下载、更新

Alibaba’s healthcare AI model on par with senior physicians in medical exams

2025-05-28 11:05:08 英文原文

Alibaba Group Holding’s healthcare-dedicated artificial intelligence (AI) model, powered by its advanced Qwen series, has demonstrated capabilities equivalent to experienced doctors and is now integrated into Quark, the company’s flagship consumer-facing AI assistant app.

The AI model successfully passed China’s medical qualification exams, achieving the “Deputy Chief Physician” standard across 12 common medical disciplines, including general medicine, internal medicine, general surgery, obstetrics and gynaecology, and paediatric medicine, Alibaba said on Tuesday.

China’s medical professional certification system ranks healthcare workers across five levels, with “Deputy Chief Physician” at the fourth-highest level.

The healthcare model is powered by Alibaba’s Qwen 2.5-32B foundation model, supported by extensive, high-quality data and advanced multi-stage training, the company said. Alibaba owns the South China Morning Post.

Benchmark data provided by Quark indicates the healthcare model outperformed DeepSeek’s R1 and V3, and OpenAI’s GPT-4o, in test accuracy. It scored a 74.8 per cent accuracy rate at the “Deputy Chief Physician” level and a 56.4 per cent accuracy rate at the top-tier “Chief Physician” standard.

The Qwen logo is displayed next to a humanoid figure. Photo: Shutterstock Images

The Qwen logo is displayed next to a humanoid figure. Photo: Shutterstock Images

According to Alibaba, the healthcare model has been fully integrated into Quark, which automatically deploys the model when a user asks health-related questions. It has been refined for better accuracy in conjunction with hospitals and medical institutions, which are adopting the model in their own applications, the company said.

关于《Alibaba’s healthcare AI model on par with senior physicians in medical exams》的评论


暂无评论

发表评论

摘要

Alibaba's Qwen series AI model, now integrated into its Quark app, has achieved "Deputy Chief Physician" level across 12 medical disciplines after passing China’s medical qualification exams. The model, based on Qwen 2.5-32B foundation and trained with extensive data, outperforms DeepSeek's R1/V3 and OpenAI's GPT-4 in test accuracy. It scores 74.8% at the "Deputy Chief Physician" level and 56.4% at the "Chief Physician" standard. The AI is being adopted by hospitals and medical institutions for their applications.