Qingsong Health Group (02661) has officially launched its evidence-based medicine intelligent agent product, "Zheng Yuanfang." The product is designed with evidence-based medicine methodology as its core logic, targeting real clinical scenarios to provide doctors with explainable, verifiable, and traceable medical decision support. This marks an advancement for AI from being a "medical Q&A tool" to an "evidence-based decision-making assistant." Unlike current medical AI products based on general large models enhanced with medical data, Zheng Yuanfang does not prioritize generative capabilities. Instead, it incorporates the evidence-based medicine system at its foundation, making evidence prioritization and traceability its native design principles. In terms of capabilities, Zheng Yuanfang not only offers literature and data support but also simulates expert research processes through a deep research engine. It can perform cross-language literature retrieval, authority assessment, and conflict verification, forming a complete reasoning path with confidence explanations. Additionally, its "reverse inquiry" feature proactively prompts users to supplement relevant test results or medical history when information is insufficient, helping to build a complete evidence loop and serving as a collaborative intelligent assistant for doctors. When using Zheng Yuanfang, each response not only provides conclusions but also clearly labels the clinical guidelines, literature sources, and evidence levels on which they are based. This mechanism avoids the issue of AI hallucinations that may seem plausible but are unverifiable, offering a higher safety margin for clinical decision-making. In authoritative benchmark tests, Zheng Yuanfang demonstrated outstanding professional accuracy. In the CMB 2023 Chinese Practicing Physician Qualification Exam benchmark test, it achieved a perfect score with 100% accuracy, becoming the first AI system in China to attain full marks in a national-level medical examination. Several mainstream general large models did not reach this level in similar tests. In more challenging oncology senior and associate senior exams, Zheng Yuanfang achieved state-of-the-art (SOTA) results in complex clinical reasoning scenarios, significantly outperforming multiple domestic and international counterparts, including OpenEvidence, showcasing its leading capabilities in real clinical decision support. In terms of data and knowledge systems, Zheng Yuanfang has built a knowledge base covering over 50 million authoritative medical data entries in Chinese and English. It systematically integrates international medical literature with core resources from the Chinese medical system, including large-scale Chinese medical journals, guideline data, and other internationally authoritative medical literature, as well as a self-constructed structured medical knowledge graph. On this foundation, Zheng Yuanfang achieves dual alignment between Chinese medical guidelines and international evidence-based systems, effectively addressing the issue of international medical AI products being unsuitable for Chinese clinical environments. In its initial phase, the product collaborated with over 3,000 professional doctors in model training, making it more aligned with real clinical scenarios. As a comprehensive health technology platform covering health management, medical services, and health security, Qingsong Health Group has long accumulated data and technical capabilities while actively cooperating with medical institutions and authoritative media to embed intelligent capabilities into real medical scenarios, promoting industry upgrades and the construction of a full-lifecycle health ecosystem. The launch of Zheng Yuanfang signifies a critical leap for medical AI from "being able to answer questions" to "being able to support decisions." By systematizing and toolizing evidence-based medicine methodology, Zheng Yuanfang is expected to become essential infrastructure connecting medical knowledge with clinical practice, providing more reliable technical support for clinical decision-making.
Comments