Advanced Toxic Comment Classification Using Multi-‎Architecture Generative AI Techniques

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Generative AI improves classification accuracy and F1-score for toxic comments in the Jigsaw dataset, suggesting new detection methods.

Achieving a toxic F1-score of 98.10% demonstrates the effectiveness of generative AI in classification tasks.
The approach improves accuracy to 96.22% by fine-tuning GPT-2 with synthetic data and addressing class imbalance.
A multi-architecture generative AI framework enhances traditional models, highlighting a new path for detection systems.
The study indicates potential advancements in mitigating toxic behavior online through data augmentation techniques.

Sushma et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark