2024 ICML ICML 2024

Seesaw: Compensating for Nonlinear Reduction with Linear Computations for Private Inference