Nvidia's Korea-specific synthetic dataset "Nemotron-Persona-Korea" ranked No. 1 in the dataset institutional sector on the global artificial intelligence (AI) development platform Hugging Face.
Nvidia announced the results on the 28th, saying it was "a case where a Korean-language–specialized dataset gained strong attention from the global community and was recognized for its technical completeness and practicality." The company said it is a noteworthy achievement that shows the competitiveness of Korea's AI ecosystem.
Nemotron-Persona-Korea is a synthetic dataset of 6 million entries that precisely reflects Korea's demographic, geographic, and cultural characteristics. It was built based on trusted public and private data from the Korean Statistical Information Service (KOSIS), the Supreme Court, the National Health Insurance Service, the Korea Rural Economic Institute (KREI), and NAVER Cloud, among others.
Key attributes such as name, gender, age, marital status, education level, occupation, and place of residence follow actual statistical figures. By considering the honorific system, regional job patterns, and other Korean-language and cultural contexts, it increased the realism of the data. It also covers older adults, rural areas, education, and occupational groups that were relatively underrepresented in existing datasets. Nvidia said it "helps developers build sophisticated AI systems that better understand Korean culture."
Nvidia designed the dataset in accordance with the Personal Information Protection Act (PIPA), composing it as fully synthetic data that does not include personal information. Nemotron-Persona-Korea is currently released under an open-source license. Nvidia expressed expectations that it "will contribute to expanding data diversity, mitigating model bias, and improving response quality as a key asset for advancing Korea-style sovereign AI."