Nature Human BehaviourMultimodal large language models can make context-sensitive hate speech evaluations aligned with human judgementThomas Davidson2025-12-15Full textDOI https://doi.org/10.1038/s41562-025-02360-w