In different NLP tasks, detecting harmful content is crucial for online environments, especially with the growing influence of social media. However, previous research has two main issues: 1) a lack of data in low-resource settings, and 2) inconsistent definitions and criteria for judging harmful content, requiring classification models to be robust to spurious features and diverse. We propose Toxicraft, a novel framework for synthesizing datasets of harmful information to address these weaknesses. With only a small amount of seed data, our framework can generate a wide variety of synthetic, yet remarkably realistic, examples of toxic information. Experimentation across various datasets showcases a notable enhancement in detection model robustness and adaptability, surpassing or close to the gold labels.
@article{arxiv.2409.14740,
title = {ToxiCraft: A Novel Framework for Synthetic Generation of Harmful Information},
author = {Zheng Hui and Zhaoxiao Guo and Hang Zhao and Juanyong Duan and Congrui Huang},
journal= {arXiv preprint arXiv:2409.14740},
year = {2025}
}