Oriel — AI & technology worth noticing, every day.

Nemotron-Personas-Finance-Japan: A Synthetic Dataset Capturing Japan's Demographics and Financial Context

Web page · https://huggingface.co/blog/nvidia/nemotron-personas-finance-japan

First seen by Oriel on .

AI summary

Nemotron-Personas-Finance-Japan is an open, fully synthetic Japanese financial persona dataset combining official demographic, labor, household-finance, and literacy distributions for privacy-conscious AI development and evaluation.

  • The CC BY 4.0 release contains 1 million records with nine personas each, totaling 9 million natural Japanese personas, about 1.5 billion tokens, and roughly 950,000 unique names.
  • Each record has 50 fields covering persona details, demographics, labor context, household finances, digital and financial literacy, with more than 1,600 occupation categories.
  • NeMo Data Designer handled templating, validation, structured outputs, and retries, while NeMo SDG-PGMs grounded distributions statistically and gemma-4-31B-it generated Japanese text.
  • Every persona derives from public statistical distributions rather than real-person records, reducing privacy risk while preserving demographic, regional, occupational, asset, and personality diversity.
  • Teams can generate dialogues, simulate customers with NeMo User Sim, test bias and fairness, evaluate sales interactions, red-team customer-facing systems, and build financial AI agents.

#synthetic data #financial ai #japanese ai #nemo data designer

Generated by gpt-5.6-sol. Check the original for details.

Aliases

SourceExternal IDURL
Hugging Face Bloghttps://huggingface.co/blog/nvidia/nemotron-personas-finance-japanhttps://huggingface.co/blog/nvidia/nemotron-personas-finance-japan Read discussion