Bootstrap The Original Latent: Learning a Private Model from a Black-box Model
This addresses privacy concerns for model owners and user needs in machine learning, offering an incremental improvement in black-box adaptation methods.
The paper tackles the problem of training private models while using black-box foundation models for guidance, proposing a new setting and training strategy that achieves efficiency and robustness across three datasets without manual augmentations.
In this paper, considering the balance of data/model privacy of model owners and user needs, we propose a new setting called Back-Propagated Black-Box Adaptation (BPBA) for users to better train their private models via the guidance of the back-propagated results of a Black-box foundation/source model. Our setting can ease the usage of foundation/source models as well as prevent the leakage and misuse of foundation/source models. Moreover, we also propose a new training strategy called Bootstrap The Original Latent (BTOL) to fully utilize the foundation/source models. Our strategy consists of a domain adapter and a freeze-and-thaw strategy. We apply our BTOL under BPBA and Black-box UDA settings on three different datasets. Experiments show that our strategy is efficient and robust in various settings without manual augmentations.