NVIDIA and Elith release Nemotron-Personas-Finance-Japan, a million-record synthetic dataset for Japanese financial AI
The open dataset pairs official Japanese statistics with household finance and financial literacy data to enable privacy-compliant agentic AI development in finance.
1 source · cross-referenced
- NVIDIA and Elith released Nemotron-Personas-Finance-Japan, an open synthetic dataset with 9 million personas for Japanese financial AI.
NVIDIA and Elith released Nemotron-Personas-Finance-Japan, an open synthetic dataset designed to support agentic AI development in Japan’s financial sector while complying with privacy regulations. The dataset extends the open-source Nemotron-Personas-Japan collection and is, to the authors’ knowledge, the first open persona dataset at million-record scale that pairs Japan’s official statistics with household finance and financial literacy data.
The dataset contains 9 million personas (1 million records × 9 personas each) in natural Japanese, with 50 fields per record covering persona attributes, contextual demographics, labor statistics, and financial indicators. It totals approximately 1.5 billion tokens, including roughly 750 million persona tokens and 950,000 unique names, spanning 1,600+ occupation categories. Coverage includes demographic, regional, asset distribution, and personality trait axes, with diverse persona types such as occupations, hobbies, financial behaviors, and financial concerns.
To build the dataset, the team used NVIDIA’s NeMo Data Designer, a compound AI system for synthetic data generation that supports Jinja templating, Pydantic validation, structured outputs, automated retries, and multiple generation backends. Generation relied on NeMo SDG-PGMs, a probabilistic graphical model framework under Apache-2.0, and gemma-4-31B-it for Japanese text generation. The dataset was refined with input from domestic financial institutions and industry associations, including Sompo Japan, Tokio Marine Holdings, Bank of Fukuoka, Mizuho Financial Group, Meiji Yasuda Life Insurance, and the Financial Data Utilization Association.
The release is licensed under CC BY 4.0 and is intended for both commercial and non-commercial use. It is designed to integrate with the NeMo User Sim framework for executing simulations drawn from diverse, real-world-consistent distributions.
- Nov 11, 2026 · Hugging Face
Hugging Face proposes token pooling-aware training to improve compressed multi-vector retrieval
Trust79 - Nov 11, 2026 · TechCrunch — AI
Fractile raises $720M Series C to develop AI chips and compiler stack
Trust79 - Nov 11, 2026 · The Verge — AI
Google’s Tabletop AI experiment lets users build custom collaborative apps via waitlist
Trust79