AI RESEARCH

SWAP: Towards Copyright Auditing of Soft Prompts via Sequential Watermarking

arXiv CS.AI

ArXi:2511.04711v2 Announce Type: replace-cross Large-scale vision-language models, especially CLIP, have nstrated remarkable performance across diverse downstream tasks. Soft prompts, as carefully crafted modules that efficiently adapt vision-language models to specific tasks, necessitate effective