GRAID: Synthetic data generation with geometric constraints and multi-agentic reflection for harmful content detection

A novel pipeline that leverages Large Language Models (LLMs) for dataset augmentation.


Latest publications