The Quark Health Large Model has performed outstandingly in the deputy chief physician title examination, successfully passing the evaluation of 12 disciplines, becoming the first large model in the country to cross this threshold. This exam covered 12 commonly used disciplines including general medicine, common internal medicine, common surgery, obstetrics and gynecology, pediatric internal medicine, oncology, stomatology, otorhinolaryngology, ophthalmology, dermatology and sexually transmitted diseases, psychiatry, and anesthesiology.
The Quark Health Large Model exceeded the pass mark in all test subjects and reached the pass line for a director physician in four disciplines: general medicine, oncology, dermatology and sexually transmitted diseases, and psychiatry. This achievement marks a key step for the large model from "knowledge memory" to "clinical reasoning," demonstrating its potential in medical auxiliary decision-making.
Unlike previous domestic large models that mostly stayed at the初级title examination stage, the Quark Health Large Model achieved a two-level leap from junior to deputy senior titles this time. The model is based on Qwen, built with massive high-quality data and trained through multiple stages, achieving such significant results.
In practical applications, users will find that when using Quark Search, the model can analyze complex medical problems before searching, dynamically retrieving related books, guidelines, drug instructions, and medical papers, thereby improving the accuracy of handling complex cases.
Specifically, in the junior and intermediate title exams, Quark's scores were about 7 points and 10 points higher than the full-blooded base model, respectively. In the deputy chief physician exam, Quark's highest lead reached 30 points. In long-chain reasoning and diagnostic pathway planning, Quark's performance also improved significantly.
It is worth mentioning that in terms of the correct rate of multiple-choice questions, some base models failed to exceed 60%, while Quark boosted the correct rate to 71% with the "medical long thinking" mechanism; in case analysis questions, the correct rate reached 53%. These achievements would not have been possible without the efforts of the Quark Health Algorithm Engineers. They stated that although machine passing exams does not mean it can replace doctors, it shows the enormous potential of machines in assisting diagnosis and treatment decisions, evidence-based retrieval, and patient communication.
Key Points:
🩺 The Quark Health Large Model successfully passed the deputy chief physician exam, becoming the first large model in the country to cross this threshold.
📚 The model exceeded the pass mark in all 12 disciplines and reached the pass line for a director physician in four disciplines.
🤖 Quark improves the accuracy of medical decision-making through analysis and dynamic retrieval, providing efficient health management support for both doctors and patients.
