Facebook, Inc.

Maplebear Inc.

Li Tan

John Samuel Barnett

Alexandre Karpenko

Ling Bao

Georges Edouard Maurice Berenger

Christopher James Connolly

Udeepta Dutta Bordoloi

Hugo Johan Van Heuven

Jiangbo Miao

Alex Restrepo

John Pottebaum

Ronnachai Jaroensri

David Mercurio

Maximilian Machedon

Jian Li

David Harry Garcia

Haixun Wang

Xiao Xiao

Taesik Na

Honglu He

A computer system finetunes a machine-learned language model to generate a personalized response to a user request. The system may generate a user representation for each of a plurality of users by applying a transformer model to a sequence of tokens representing a sequence of activities of the user. The system may train an evaluation model coupled to receive a user representation and a response to a user request and generate an estimated evaluation score indicating a level of personalization of the response to the user. The system may finetune a first machine-learned language model to generate a second machine-learned language model. The finetuned machine-learned language model is configured to provide personalized responses for customer services at an online concierge system.

Personalized machine-learned large language model (LLM)

A system may generate a prompt based in part on a search query from a customer client device. The prompt instructs a machine learned model to provide item predictions. And the model was trained by: converting structured data describing items of an online catalog to annotated text data (unstructured data), generating training examples based in part on the annotated text data, and training the model using the training examples. The system may receive item predictions generated by the prompt being applied to the machine learned model, the item predictions may have corresponding item identifiers. The item predictions are processed to identify a recommended item from the item predictions. The processing includes determining item information for the recommended item using an item identifier associated with the recommended item. The item information is provided to the customer client device.

Machine learned models for search and recommendations

Systems, methods, and non-transitory computer-readable media can provide one or more options to adjust, in a virtual three-dimensional space, an image in an initial perspective. At least a first command, provided based on the one or more options, to adjust the image from the initial perspective into an adjusted perspective can be detected. The image can be adjusted, based on at least the first command, into the adjusted perspective in the virtual three-dimensional space to produce an adjusted image. A second command to modify a viewable area of the adjusted image can be detected. At least a portion of the adjusted image can be automatically fitted, when the second command has ceased, within a specified display area while one or more empty regions can be minimized in the specified display area.

Systems and methods for providing image perspective adjustment and automatic fitting

An online system logs information describing presentation of content items, such as actions performed with presented content items or descriptions of presentation of content items, and uses the logged information to recommend other content items. Because of a large amount of information describing content item presentation, the online system samples the information and stores the resulting samples. To improve accuracy of the sampling, the online system adapts the sampling rate for different types of information describing presentation of content items based on amounts of the different types of information received during a specific time interval.

Determining a rate for sampling information describing presentation of content based on previously received information describing presentation of content

Sentence internationalization methods and systems are disclosed. The method may include: providing a developer interface to define an internationalized sentence syntax for an application on a social networking system, the internationalized sentence syntax for translating a natural language expression of a social graph edge of the social networking system; providing a sentence option on the developer interface to define grammar of the internationalized sentence syntax; generating a token structure including a language token to assist translation of the internationalized sentence syntax into a preferred language indicated by the language token, the token structure customizable via the developer interface to configure translation options; associating a social graph attribute with a first token of the token structure; and storing the sentence option and the token structure with the internationalized sentence syntax to facilitate run-time translation of the internationalized sentence syntax into the natural language expression in the preferred language.

Methods and systems for generation of a translatable sentence syntax in a social networking system

Growing community of inventors

Fremont, CA, United States of America

Li Tan