1 paper
Yujia Li, Yiqun Zhang, Zihan Cheng +7
Multimodal harmful meme detection is typically formulated as image--text harmfulness classification. A model may correctly predict harmfulness while misidentifying the attacked tar…