Abstract: There are proposed a method, device, apparatus, and medium for protecting sensitive data. In a method, to-be-processed data is received from a server device. A processing result of a user for the to-be-processed data is received, the processing result comprising sensitive data of the user for the processing of the to-be-processed data. A gradient for training a server model at the server device is determined based on a comparison between the processing result and a prediction result for the to-be-processed data. The gradient is updated in a change direction associated with the gradient so as to generate an updated gradient to be sent to the server device. Noise is added only in the change direction associated with the gradient. The corresponding overhead of processing noise in a plurality of directions can be reduced, and no excessive noise data interfering with training will be introduced to the updated gradient.

Abstract: A model file management method includes that a terminal device receives a storage address of a target model file package from a server and the terminal device obtains the target model file package based on the storage address of the target model file package, where the target model file package is based on a parameter of a model file package locally stored in the terminal device and a parameter of a model file package managed by the server. In an artificial intelligence (AI) field, an application may implement a specific function by using an AI model file. An application is decoupled from an AI model file such that the terminal device performs centralized management on a general model file.

Abstract: The technology of this application relates to a method for controlling a device using a voice and an electronic device, where the technology is applied to the field of artificial intelligence. The method for controlling a device using a voice includes: obtaining a voice instruction of a user, where the voice instruction is used to indicate a target instruction, obtaining user interface information of a current user interface, where the current user interface is a user interface currently displayed by a client device, and determining the target instruction corresponding to the voice instruction, where the target instruction is obtained by using the voice instruction and the user interface information. This technology described herein helps improve voice recognition efficiency.

Abstract: A model file management method includes that a terminal device receives a storage address of a target model file package from a server and the terminal device obtains the target model file package based on the storage address of the target model file package, where the target model file package is based on a parameter of a model file package locally stored in the terminal device and a parameter of a model file package managed by the server. In an artificial intelligence (AI) field, an application may implement a specific function by using an AI model file. An application is decoupled from an AI model file such that the terminal device performs centralized management on a general model file.