01▲Latency-Optimal Adaptive Split Inference for Privacy-Preserving Cloud-Edge-End Collaboration arxiv.org The surprising bit is they make FHE less awful by splitting inference across device, edge, and cloud, and optimizing the cut.arxivcloud-computingdistributed-inferenceedge-computingfheiotoptimizationprivacy0 pts/metaleak/5 days ago/1 comment