# Limitations
The Fusion Transformer-XL model’s reliance on a single dataset introduces potential biases, limiting its generalizability across diverse language styles and contexts. Additionally, the model’s performance is influenced by the quality and quantity of labeled data, highlighting the need for larger, more diverse datasets. This study also focuses solely on linguistic features, excluding multimedia content and user metadata that could improve detection accuracy.

# Validity
The study's results are largely shaped by the dataset characteristics and linguistic focus, which may affect the model's applicability in varied settings. Expanding the dataset and incorporating additional data types would likely enhance the model's robustness and adaptability to a broader range of cyberbullying scenarios.