What’s Out There
← Back to swiping
DemographicsFreeCompany

Nemotron-Personas-Korea

NVIDIA

Synthetic Korean personas whose ages, jobs, and regions follow official national statistics.

Persona records synthesized to match real Korean distributions of age, sex, marital status, education, occupation, and region, using sources such as KOSIS, the Supreme Court, the National Health Insurance Service, and the Korea Rural Economic Institute. One row is one fictional person described in Korean. It is aimed at seeding culturally grounded synthetic data.

Access

Cost
FreeNo charge.
License
Open licenseCC BY 4.0
Sign-up
No sign-up
How you get it
API and Bulk downloadParquet

What's in it

One row is
One row per synthetic persona
Covers
South Korea
Years
Statistical sources as of the 2020s
Updated
One-time release
Size
1 to 10 million personas
namesexagemarital statuseducationoccupationregionpersona text

Quality

3 of 5

Novel in tying generated personas to real marginal distributions, but the narrative text is LLM-written and joint distributions are modeled.

  • Personas are fictional
  • Only margins follow official statistics
  • Text quality depends on the generating model

Connects to

  • PersonaHub: Persona set with no demographic grounding, for contrast
  • FinePersonas: Larger LLM-extracted persona collection

Questions it could answer

  • Do synthetic occupation shares match official labor statistics by region?
  • How varied is the persona text within the same demographic cell?

Similar data sets

  • PersonaHub

    Short persona descriptions that steer an LLM into generating varied synthetic data.

    Language & textTencent AI Lab (proj-persona)Free

  • Cosmopedia

    Synthetic textbooks, blog posts, and stories written by Mixtral to mimic what web pretraining text teaches.

    Language & textHugging Face (SmolLM team)Free

  • TinyStories

    Millions of short stories in a three-year-old's vocabulary, used to show tiny models can write coherent text.

    Language & textRonen Eldan and Yuanzhi Li (Microsoft Research)Free

  • WildChat-1M

    A million real ChatGPT conversations with state, country, and hashed-IP fields attached.

    Language & textAllen Institute for AI and University of WashingtonFree

  • Llama-Nemotron Post-Training Dataset

    Millions of math, code, and reasoning examples used to post-train NVIDIA's Llama-Nemotron models.

    Language & textNVIDIAFree

  • Wittgenstein Centre Human Capital Data Explorer

    Population projections to 2100 for 200 countries that track not just how many people, but how educated.

    DemographicsWittgenstein Centre for Demography and Global Human CapitalFree