AI RESEARCH

FedSpy-LLM: Towards Scalable and Generalizable Data Reconstruction Attacks from Gradients on LLMs

arXiv CS.LG

ArXi:2604.06297v1 Announce Type: cross Given the growing reliance on private data in