culturally-aware

This paper proposes a training-time explainability framework for multilingual hate speech detection, aligning model reasoning with human rationales to improve classification performance and interpretability, evaluated on English and Hinglish datasets.