CareGuard is a fine-tuned Large Language Model (LLM) designed for cyberbullying and harmful language detection in short social media texts, particularly tweets.

CareGuard is an LLM-based text classification model fine-tuned on labeled Twitter data from the Cyberbully Detection Dataset (Kaggle). It builds on a pretrained multilingual LLM backbone and is optimized to identify cyberbullying, harassment, and abusive language patterns in online communication.

Model type: Transformer-based Large Language Model (LLM) for text classification

Language(s): Multilingual (primarily English)

Finetuned from: FacebookAI/xlm-roberta-base

Here is how to use this model in PyTorch: