BrainboxAI/medical-training-il

Medical-Training-IL 是一个双语（希伯来语/英语）医学指令调整语料库，专为以色列居民准备A阶段考试而设计，用于训练小型、设备上的医学模型。该数据集结合了高质量的英语医学问答（USMLE风格、基础科学、基于研究的）和约5000个源自希伯来语维基百科的希伯来语本土医学示例。数据集旨在在Google的medgemma-1.5-4b-it模型基础上进行进一步微调，该模型已经预训练了医学内容，我们进一步针对希伯来语和以色列背景进行了专门化。数据集包括来自MedQA（USMLE）、MedMCQA（过滤后）、PubMedQA、希伯来语维基百科医学文章以及手工制作的身份集的示例，总计约21,432个示例，其中约35%为希伯来语，65%为英语。数据集分为训练集（20,361个示例）和验证集（1,071个示例），并提供了详细的记录格式和预期的使用场景。

Medical-Training-IL is a bilingual (Hebrew / English) medical instruction-tuning corpus curated for training small, on-device medical models for Israeli residents preparing for Stage A exams. It combines high-quality English medical QA (USMLE-style, basic sciences, research-grounded) with ~5,000 Hebrew-native medical examples derived from Hebrew Wikipedia. The dataset is designed for continued fine-tuning on top of Googles medgemma-1.5-4b-it — a model already pre-trained on medical content, which is specialized further for Hebrew and Israeli context. The dataset includes examples from MedQA (USMLE), filtered MedMCQA, PubMedQA, Hebrew Wikipedia medical articles, and a handcrafted identity set, totaling ~21,432 examples (~35% Hebrew / 65% English). It is split into train (20,361 examples) and validation (1,071 examples) sets, with a detailed record format and intended use cases provided.