Xing Xie
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
Specify Privacy Yourself: Assessing Inference-Time Personalized Privacy Preservation Ability of Large Vision-Language Models
According to the abstract, Xingqi Wang, Xiaoyuan Yi, Xing Xie and Jia Jia present SPY-Bench, with 32,700 samples, to test whether vision-language models respect privacy limits that users specify in words. Across 21 models, most performed poorly; the best-performing o4-mini reached about 24% compliance accuracy.