OpenAI o1模型在诊断测试中与急诊医生表现相当

OpenAI’s o1 model is showing early signs of clinical potential after outperforming physicians in a controlled study using real emergency room cases. In a blinded evaluation based on anonymized patient records, the system produced more accurate diagnoses than two attending doctors, highlighting how advanced reasoning models are beginning to move beyond benchmarks and into real-world medical scenarios. The study is notable because it moves away from simplified exam-style questions and instead tests performance against messy, incomplete, and time-sensitive clinical data—the kind of information doctors actually work with in emergency settings. That shift makes the results more meaningful, even as researchers caution that such systems are not ready to replace human clinicians. …