Yuran Bian

Your Prompts Are Not Safe: Output-Free Membership Inference via Prompt Vectors in Vision-Language Tuning

Prompt tuning enables Vision-Language Models (VLMs) to efficiently adapt to new tasks through learnable prompt vectors. This naturally raises a question: do these prompts leak private information about their training data? While Membership Inference Attacks (MIAs) can quantify this risk, current met