BriefGPT.xyz
Aug, 2023
Ceci n'est pas une pomme: 多模态嵌入中的对抗幻觉
Ceci n'est pas une pomme: Adversarial Illusions in Multi-Modal Embeddings
HTML
PDF
Eugene Bagdasaryan, Vitaly Shmatikov
TL;DR
多模态编码器、敌对幻觉、嵌入空间、接近度、零样本分类。
Abstract
multi-modal encoders
map images, sounds, texts, videos, etc. into a single
embedding space
, aligning representations across modalities (e.g., associate an image of a dog with a barking sound). We show that multi-
→