A French OSCE Dialogue Dataset and Controllable Virtual Patient System for Clinical Training

用于临床培训的法语 OSCE 对话数据集与可控虚拟病人系统

The clinical and communication skills of medical students are commonly assessed through Objective Structured Clinical Examinations (OSCEs), which consist of brief scenario-driven simulations of doctor-patient interactions. 医学生的临床和沟通技能通常通过客观结构化临床考试（OSCE）进行评估，该考试由基于简短情景的医患互动模拟组成。

However, training is often limited by the low availability of human standardized patients, motivating the development of realistic virtual patients (VPs). 然而，由于人类标准化病人的可用性较低，培训往往受到限制，这促使了逼真虚拟病人（VP）的开发。

To address this gap, we introduce a French OSCE dialogue dataset comprising 240 student-patient training interactions. 为了弥补这一差距，我们引入了一个包含 240 次学生与病人培训互动的法语 OSCE 对话数据集。

We build upon it a controllable LLM-based pipeline to generate synthetic OSCE dialogues. The pipeline integrates modular components, such as retrieval-based grounding and a reflection loop, to ensure patient fidelity, coherence, and realism. 在此基础上，我们构建了一个基于大语言模型（LLM）的可控流水线，用于生成合成的 OSCE 对话。该流水线集成了模块化组件，例如基于检索的接地（grounding）和反思循环，以确保病人的保真度、连贯性和真实感。

Additionally, we propose a multi-level evaluation framework assessing patient simulation quality, student performance, and linguistic quality, using an LLM-as-a-Judge approach. 此外，我们提出了一种多级评估框架，利用“大模型作为裁判”（LLM-as-a-Judge）的方法，对病人模拟质量、学生表现和语言质量进行评估。

Experiments suggest that controllability modules generally improve patient fidelity and student evaluation consistency. Finally, we implement an interactive prototype in which students can practice with a VP and receive automatic feedback. 实验表明，可控性模块通常能提高病人的保真度和学生评估的一致性。最后，我们实现了一个交互式原型，学生可以在其中与虚拟病人进行练习并获得自动反馈。