GRAID: Synthetic Data Generation with Geometric Constraints and Multi-Agentic Reflection for Harmful Content Detection
Rad, Melissa Kazemi, Purpura, Alberto, Kumar, Himanshu, Chen, Emily, Sorower, Mohammad Shahed
–arXiv.org Artificial Intelligence
We address the problem of data scarcity in harmful text classification for guardrailing applications and introduce GRAID (Geometric and Reflective AI-Driven Data Augmentation), a novel pipeline that leverages Large Language Models (LLMs) for dataset augmentation. GRAID consists of two stages: (i) generation of geometrically controlled examples using a constrained LLM, and (ii) augmentation through a multi-agentic reflective process that promotes stylistic diversity and uncovers edge cases. This combination enables both reliable coverage of the input space and nuanced exploration of harmful content. Using two benchmark data sets, we demonstrate that augmenting a harmful text classification dataset with GRAID leads to significant improvements in downstream guardrail model performance.
arXiv.org Artificial Intelligence
Aug-26-2025
- Country:
- North America > United States (0.46)
- Genre:
- Research Report
- New Finding (0.67)
- Experimental Study (0.46)
- Research Report
- Industry:
- Law Enforcement & Public Safety (1.00)
- Information Technology > Security & Privacy (1.00)
- Government (1.00)
- Law (0.94)
- Education (0.93)
- Health & Medicine > Therapeutic Area (0.68)
- Media > News (0.67)
- Technology: