NVIDIA 发布法国和比利时 Nemotron 本地化人物数据集
原标题：Fine-Tuning NVIDIA Nemotron to Know Local Rules, Not Just Local Languages
AI 摘要
NVIDIA 发布 Nemotron-Personas-France 和 Nemotron-Personas-Belgium 两个基于地区人口特征的数据集，分别与法国 AI 初创公司 Pleias 及比利时鲁汶大学合作构建，属于覆盖 12 个国家的 Nemotron 开放数据计划。文章以法国护照办理场景为例，说明用这些数据集对 Nemotron 开放模型做后训练，可让模型掌握当地具体规则而非仅语言，从而给出更准确的回答。Pleias 已将三个 Nemotron-Personas 数据集用于 Lamina 脊柱护理助手，把少量真实病例扩展为多样化训练对话。
正文节选
As AI adoption expands globally, people across languages and cultures are using it in more aspects of their daily lives. General-purpose language models serve as a powerful foundation, but they sometimes lack the local knowledge, cultural nuance, and precision a particular application needs. Post-training can adapt an open model to these needs. With suitable data and evaluation, it can improve accuracy and usefulness while supporting safer interactions with the model. This gives countries and in