While recent advancements in commercial large language models (LM) have shown
promising results in medical tasks, their closed-source nature poses
significant privacy and security concerns, hindering their widespread use in
the medical field. Despite efforts to create open-source models, their limited
parameters often result in insufficient multi-step reasoning capabilities
required for solving complex medical problems. To address this, we introduce
Meerkat-7B, a novel medical AI system with 7 billion parameters. Meerkat-7B was
trained using our new synthetic dataset consisting of high-quality
chain-of-thought reasoning paths sourced from 18 medical textbooks, along with
diverse instruction-following datasets. Our system achieved remarkable accuracy
across seven medical benchmarks, surpassing GPT-3.5 by 13.1%, as well as
outperforming the previous best 7B models such as MediTron-7B and BioMistral-7B
by 13.4% and 9.8%, respectively. Notably, it surpassed the passing threshold of
the United States Medical Licensing Examination (USMLE) for the first time for
a 7B-parameter model. Additionally, our system offered more detailed free-form
responses to clinical queries compared to existing 7B and 13B models,
approaching the performance level of GPT-3.5. This significantly narrows the
performance gap with large LMs, showcasing its effectiveness in addressing
complex medical challenges.

Meerkat-7B 是一个具有 70 亿参数的新型医疗人工智能系统，通过使用来自 18 本医学教科书的高品质推理路径和多样的指令跟随数据集进行训练，成功地在七个医学基准测试中取得了出色的准确性，在超过 GPT-3.5 13.1％的同时，还超过了先前的 7B 模型（如 MediTron-7B 和 BioMistral-7B）分别达到了 13.4％和 9.8％的性能，并且首次超过了 United States Medical Licensing Examination（USMLE）的 7B 参数模型通过阈值。此外，相比现有的 7B 和 13B 模型，我们的系统对临床问题提供了更详细的自由形式回答，接近于 GPT-3.5 的性能水平，这大大缩小了与大型语言模型之间的性能差距，展示了它在应对复杂的医学挑战方面的有效性。