Scaling Synthetic Data Creation with 1,000,000,000 Personas

Tao Ge and colleagues propose a persona-driven data synthesis methodology that uses various perspectives within a large language model to create diverse synthetic data. They introduce Persona Hub, a collection of 1 billion personas automatically curated from web data, which they describe as about 13% of the world's population. The authors say these personas can tap nearly every perspective encapsulated within the LLM, and they demonstrate use cases including math and logic problems, user prompts, knowledge-rich texts, game NPCs and tools. They claim the approach is versatile, scalable, flexible and easy to use.