TwinnyAI Personas Dataset
19,100+ downloads of structured persona data.
Downloads
19.1K+
Hosts
HF + Kaggle
Access
Open
Metric profile
01 / Signal
19.1K+
Community downloads
Hugging Face + Kaggle
02 / Signal
Structured
Schema consistency
fine-tune ready fields
03 / Signal
Open
Access barrier
no gating, no request form
Overview
A curated, structured dataset of AI personas designed for training and evaluating persona-consistent language models — the data foundation underneath the TWINNY.AI digital twin work.
Published openly on both Hugging Face and Kaggle, where it has passed 19,100 downloads.
What makes it work
Structured, not scraped
Each persona carries consistent fields — traits, tone, domain and behavioural cues — so it is usable for supervised fine-tuning without cleanup.
Evaluation-ready
The schema supports persona-consistency benchmarking as well as generation, making it useful on both sides of the training loop.
Community adoption
Sustained download volume across two platforms indicates real downstream usage rather than a one-off release.
Architecture
- Consistent persona schema across all records
- Published as Hugging Face dataset + Kaggle mirror
- Designed for SFT and persona-consistency evaluation
Stack
- Hugging Face
- Kaggle
- LLM Data
Structured AI personas dataset published on Hugging Face & Kaggle (19,100+ downloads).
Interested?