Back to results
US 2026/0310671 A1Published application
Learning to page in and page out data chunks in a generative neural network
OwnerFlorian Nils Hartmann, Matthew Sharifiinventors
Life of this patent
Estimating the term…
Loading patent…
Loading patent…
OwnerFlorian Nils Hartmann, Matthew Sharifiinventors
Estimating the term…
Methods, systems, and apparatus for using paging operations in a generative neural network at inference. In one aspect, a user query is received. Data chunks for the user query from respective data sources are retrieved. Descriptive paths for data chunks included in each data source are retrieved. A context block for the user query is initialized by adding the data chunks for the user query and the descriptive paths to the context block. An initial input including the user query and the initialized context block is processed using the generative neural network to generate an initial output that includes data specifying paging operations that each reference one or more of the descriptive paths. The context block is updated by executing the paging operations. An updated input that includes the updated context block is processed using the generative neural network to generate a final response to the user query.
A computer implemented method comprising:
A system comprising one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
One or more non-transitory computer storage media encoded with instructions that, when executed by one or more computers, cause the one or more computers to perform operations comprising:
Dependencies are inferred from the claim text.