Detection of meaning drift in a document

Abstract

A computing device for generating, managing, or editing a document includes one or more memories to store instructions and one or more processors to execute the instructions to perform operations, the operations including: obtaining a prompt indicating an intended meaning of a document, receiving a plurality of inputs editing a first version of the document to produce a second version of the document, determining whether specified editing criteria associated with editing the document are satisfied, in response to the specified editing criteria being satisfied, implementing one or more machine-learned models to determine whether specified drift criteria associated with the intended meaning of the document is satisfied, based on the prompt, the edits to the first version of the document, and the second version of the document, and providing an output based on whether the specified drift criteria associated with the intended meaning of the document is satisfied.

Description

DETECTION OF MEANING DRIFT IN A DOCUMENTFIELD

[0001] This disclosure relates generally to machine learning processes and machine- learned devices and systems. More particularly, the disclosure relates to implementing one or more machine-learned models to monitor or detect a meaning drift for content in a document which is caused by edits to the document.BACKGROUND

[0002] A computer can receive input(s). The computer can execute instructions to process the input(s) to generate output(s) using a parameterized model. The computer can obtain feedback on its performance in generating the outputs with the model. The computer can generate feedback by evaluating its performance. The computer can receive feedback from an external source. The computer can update parameters of the model based on the feedback to improve its performance. In this manner, the computer can iteratively “learn” to generate the desired outputs. The resulting model is often referred to as a machine-learned model.

[0003] In the creation and refinement process of a document, the content may go through many modifications, from one or multiple users and agents, that may cause the generated content to deviate significantly from the original intent.SUMMARY

[0004] Aspects and advantages of embodiments of the disclosure will be set forth in part in the following description, or can be learned from the description, or can be learned through practice of the embodiments.

[0005] Example aspects of the disclosure provide an example computing device that includes one or more processors and one or more example non-transitory computer-readable media storing instructions that are executable by the one or more processors to cause the computing device to perform example operations. In some implementations, the example operations can include obtaining a prompt indicating an intended meaning of a document, receiving a plurality of inputs editing a first version of the document to produce a second version of the document, determining whether one or more specified editing criteria associated withediting the document are satisfied, in response to the one or more specified editing criteria being satisfied, implementing one or more machine-learned models to determine whether specified drift criteria associated with the intended meaning of the document is satisfied, based on the prompt, the edits to the first version of the document, and the second version of the document, and providing an output based on whether the specified drift criteria associated with the intended meaning of the document is satisfied.

[0006] In some implementations, the one or more specified editing criteria includes a number of edits to the first version of the document exceeding a threshold number of edits.

[0007] In some implementations, the one or more specified editing criteria includes a threshold duration of time elapsing since a last edit to the first version of the document.

[0008] In some implementations, the document is associated with a computing program and includes code, and the one or more specified editing criteria includes edited code in the document being capable of being compiled.

[0009] In some implementations, the prompt is received from a user to generate, via the one or more machine-learned models, the document.

[0010] In some implementations, the prompt is retrieved by the one or more machine- learned models by determining a prompt which would generate the document.

[0011] In some implementations, the plurality of inputs editing the first version of the document are provided by a plurality of users, and the one or more machine-learned models are configured to determine whether the drift criteria associated with the intended meaning of the document is satisfied, based on the prompt, the edits to the first version of the document, the second version of the document, and information associated with the plurality of users.

[0012] In some implementations, the information associated with the plurality of users includes a defined role for respective users among the plurality of users editing the document.

[0013] In some implementations, the specified drift criteria associated with the intended meaning of the document is satisfied when a global meaning of the second version of the document deviates from the intended meaning of the document by more than a threshold drift level, and the specified drift criteria associated with the intended meaning of the document is not satisfied when the global meaning of the second version of the document deviates from the intended meaning of the document by less than the threshold drift level.

[0014] In some implementations, when the specified drift criteria associated with the intended meaning of the document is satisfied, the output includes automatically reverting or revising the second version of the document.

[0015] In some implementations, when the specified drift criteria associated with the intended meaning of the document is satisfied, the output includes providing, for presentation to a user via a user interface, at least one of a notification or a recommendation to revert or revise the second version of the document.

[0016] In some implementations, the document includes at least one of an image, a video, or an audio recording, the plurality of inputs editing the first version of the document include editing characteristics to change a quality of the at least one of the image, the video, or the audio recording to produce the second version of the document, and the specified drift criteria associated with the intended meaning of the document is satisfied when a resulting quality of the second version of the document is less than a threshold quality level indicated by the prompt.

[0017] In some implementations, the document includes a computing program comprising code, the plurality of inputs editing the first version of the document include editing the code to produce the second version of the document, and the specified drift criteria associated with the intended meaning of the document is satisfied when a latency associated with the second version of the document is greater than a threshold latency level indicated by the prompt.

[0018] In some implementations, the plurality of inputs editing the first version of the document include a first input editing a first portion of the document and a second input editing a second portion of the document, and the specified drift criteria associated with the intended meaning of the document is satisfied when the first input editing the first portion of the document conflicts with the second input editing the second portion of the document.

[0019] Example aspects of the disclosure provide an example computer-implemented method. In some implementations, the example computer-implemented method can include obtaining, by a computing system comprising one or more processors, a prompt indicating an intended meaning of a document; receiving, by the computing system, a plurality of inputs editing a first version of the document to produce a second version of the document; determining, by the computing system, whether one or more specified editing criteria associated with editing the document are satisfied; in response to the one or more specified editing criteria being satisfied, determining, via one or more machine-learned models, whether specified driftcriteria associated with the intended meaning of the document is satisfied, based on the prompt, the edits to the first version of the document, and the second version of the document; and providing, by the computing system, an output based on whether the specified drift criteria associated with the intended meaning of the document is satisfied.

[0020] In some implementations, the one or more specified editing criteria includes: a number of edits to the first version of the document exceeding a threshold number of edits, or a threshold duration of time elapsing since a last edit to the first version of the document.

[0021] In some implementations, the plurality of inputs editing the first version of the document are provided by a plurality of users, and the one or more machine-learned models determine whether the drift criteria associated with the intended meaning of the document is satisfied, based on the prompt, the edits to the first version of the document, the second version of the document, and information associated with the plurality of users.

[0022] In some implementations, the computer-implemented method includes determining, via the one or more machine-learned models, the specified drift criteria associated with the intended meaning of the document is satisfied when a global meaning of the second version of the document deviates from the intended meaning of the document by more than a threshold drift level.

[0023] In some implementations, the computer-implemented method includes determining, via the one or more machine-learned models, the specified drift criteria associated with the intended meaning of the document is satisfied when a resulting quality of the second version of the document is less than a threshold quality level indicated by the prompt.

[0024] The computer-implemented method may execute any of the operations of the computing device as described herein.

[0025] Example aspects of the disclosure provide one or more example non-transitory computer-readable media storing instructions that are executable by one or more processors to cause a computing system to perform example operations. In some implementations, the example operations can include obtaining a prompt indicating an intended meaning of a document; receiving a plurality of inputs editing a first version of the document to produce a second version of the document; determining whether one or more specified editing criteria associated with editing the document are satisfied; in response to the one or more specified editing criteria being satisfied, implementing one or more machine-learned models to determinewhether specified drift criteria associated with the intended meaning of the document is satisfied, based on the prompt, the edits to the first version of the document, and the second version of the document; and providing an output based on whether the specified drift criteria associated with the intended meaning of the document is satisfied.

[0026] The non-transitory computer-readable medium may store additional instructions to execute other aspects and operations of the computing device and computer-implemented method as described herein.

[0027] Other example aspects of the disclosure are directed to other systems, methods, apparatuses, tangible non-transitory computer-readable media, and devices for performing functions described herein. These and other features, aspects, and advantages of various implementations will become better understood with reference to the following description and appended claims. The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate implementations of the disclosure and, together with the description, help explain the related principles.BRIEF DESCRIPTION OF THE DRAWINGS

[0028] FIG. 1A is an example system according to one or more example embodiments of the disclosure;

[0029] FIG. IB is an example block diagram of a computing system, according to one or more example embodiments of the disclosure;

[0030] FIG. 2 illustrates a flow diagram of an example, non-limiting computer- implemented method, according to one or more example embodiments of the disclosure;

[0031] FIG. 3 illustrates a flow diagram of an example, non-limiting computer- implemented method, according to one or more example embodiments of the disclosure;

[0032] FIG. 4 illustrates a flow diagram of an example, non-limiting computer- implemented method, according to one or more example embodiments of the disclosure;

[0033] FIG. 5 illustrates a flow diagram of an example, non-limiting computer- implemented method, according to one or more example embodiments of the disclosure;

[0034] FIG. 6 illustrates a block diagram of a meaning drift detection application, according to one or more example embodiments of the disclosure;

[0035] FIG. 7 is a flow chart diagram illustrating an example method for training a machine-learned model according to example implementations of aspects of the disclosure;

[0036] FIG. 8 is a block diagram of an example processing flow for using machine- learned model(s) to process input(s) to generate output(s) according to example implementations of aspects of the disclosure;

[0037] FIG. 9 is a block diagram of an example sequence processing model according to example implementations of aspects of the disclosure;

[0038] FIG. 10 is a block diagram of an example technique for populating an example input sequence for processing by a sequence processing model according to example implementations of aspects of the disclosure;

[0039] FIG. 11 is a block diagram of an example model development platform according to example implementations of aspects of the disclosure;

[0040] FIG. 12 is a block diagram of an example training workflow for training a machine-learned model according to example implementations of aspects of the disclosure;

[0041] FIG. 13 is a block diagram of an inference system for operating one or more machine-learned model(s) to perform inference according to example implementations of aspects of the disclosure;

[0042] FIG. 14 is a block diagram of an example networked computing system according to example implementations of aspects of the disclosure;

[0043] FIG. 15 is a block diagram of an example computing device according to example implementations of aspects of the disclosure; and

[0044] FIG. 16 is a block diagram of an example computing device according to example implementations of aspects of the disclosure.DETAILED DESCRIPTION

[0045] Reference now will be made to embodiments of the disclosure, one or more examples of which are illustrated in the drawings, wherein like reference characters denote like elements. Each example is provided by way of explanation of the disclosure and is not intended to limit the disclosure. In fact, it will be apparent to those skilled in the art that various modifications and variations can be made to disclosure without departing from the scope or spirit of the disclosure. For instance, features illustrated or described as part of one embodiment canbe used with another embodiment to yield a still further embodiment. Thus, it is intended that the disclosure covers such modifications and variations as come within the scope of the appended claims and their equivalents.

[0046] Agents powered by machine-learned models (e.g., large language models (LLMs), large multimodal models, etc.) are implemented in existing editing contexts. For example, the LLMs may be configured to write documents from scratch given a prompt or to edit content using provided instructions. An editing agent may be configured to update the original text of a document based on comments provided by the user. The content may include text, code, or other types of content that can be edited.

[0047] Content editing with machine-learned models (e.g., large language models (LLMs), large multimodal models, etc.) is becoming increasingly popular. For example, a user may request a machine-learned model to directly write blocks of texts or generate images using a prompt, as well as edit existing content for refining the content. Refining the document or imagery (e.g., via editing sessions by one user, multiple users, by an agent, etc.) may occur over a long duration of time (e.g., over many hours). Machine-learned models (e.g., large language models (LLMs), large multimodal models, etc.) can be configured to consider the context and the provided instructions (or history of instructions) from the user to generate a prediction (e.g., predicted content) based on the inputs. However, when working with an extensive amount of content (e.g., several pages of text, several minutes or hours of audio or video, multiple building blocks of a large application in the context of software architecture, etc.), the user and / or machine-learned models may not be able to monitor or keep track of all the changes and edits to the content, or monitor or keep track of content which has previously been modified. Thus, after many edits to the content, the global context may no longer make sense due to a meaning drift from the original content. That is, the generated content can deviate significantly from the original intent of the user, particularly over longer editing sessions where there are many modelbased editing sessions and interventions.

[0048] According to examples of the disclosure, methods and computing systems described herein are implemented to monitor the meaning drift and utilize a global meaning consistency mechanism to track and update separate parts of the content that are affected due to changes (edits) introduced in another part of the content.

[0049] In some implementations, one or more machine-learned models may be configured to consider as inputs: (1) original or prior versions of a document, (2) the updated version of the document, and (3) the difference between the document and the updated document. The one or more machine-learned models may be configured to provide as an output a determination as to whether the updated document remains consistent with the original intent or intended meaning of the document. That is, the one or more machine-learned models may be configured to determine whether a global meaning of the updated document has drifted beyond a threshold drift level compared to the global meaning of the original document or a prior version of the document.

[0050] For example, the output may include an updated global meaning and / or differences between the original intent or meaning of the document and the updated global meaning. For example, the output may include one or more indications which notify a user of an inconsistency with respect to a changed part of the document (e.g., inconsistencies between the updated version of the document and the original or prior version of the document, inconsistencies between different portions of the updated version of the document, etc.), which notify a user that the updated document remains consistent with the original intent or intended meaning of the document, which provide suggestions or recommendations for revising the changed part of the document so that the updated document remains consistent with the original intent or intended meaning of the document, which automatically makes a revision to the changed part of the document so that the updated document remains consistent with the original intent or intended meaning of the document, etc.

[0051] In some implementations, the one or more machine-learned models may be configured to be implemented in response to a determination that the user has edited the document. In some implementations, the one or more machine-learned models may be configured to be implemented in response to a determination that the user has edited the document and after a particular number of changes (edits) have been made. In some implementations, the one or more machine-learned models may be configured to be implemented in response to a determination that the user has edited the document and after a pause (e.g., a threshold duration of time) since a last edit (change) has occurred. In some implementations, the one or more machine-learned models may be configured to be implemented in response to a determination that the user has edited the document and in the case of the edits including changesto code in a computing program that the code is capable of being compiled. Therefore, computing resources (e.g., processor cycles) may be conserved by selectively implementing the one or more machine-learned models according to particular criteria being satisfied (e.g., a certain number of edits, a certain delay after a last edit, a condition that code in a document which has been edited be capable of being compiled, etc.).

[0052] In some implementations, the one or more machine-learned models may be configured to be implemented with respect to an entirety of the document to determine the cohesiveness and consistency of the document. In some implementations, the one or more machine-learned models may be configured to be implemented with respect to only a portion of the document, thereby conserving computing resources such as processor cycles. For example, the portion of the document may correspond to only those portions of the document which have been edited. For example, the portion of the document may correspond to one or more particular sections of the document (e g., a background section, a summary section, a conclusion section, etc.) which have been edited, and thus the one or more machine-learned models may be configured to be implemented on a section-by-section basis.

[0053] In some implementations, the one or more machine-learned models may be configured to be implemented in response to particular types of edits being made. For example, the one or more machine-learned models may be configured to identify a type of edit being made to the document. For example, in response to a first type of edit being made to the document, the one or more machine-learned models may be implemented to determine whether the updated document remains consistent with the original intent or intended meaning of the document. For example, the first type of edit may be a substantive edit which the one or more machine-learned models determines (with some confidence threshold level being satisfied) is likely to have affected the content, structure, and meaning of the document. Such substantive edits may include reorganizing sections or paragraphs, the addition of new information, the removal of information, a modification to the tone or style of the document, etc. For example, in response to a second type of edit being made to the document, the one or more machine-learned models may not be implemented to determine whether the updated document remains consistent with the original intent or intended meaning of the document. For example, the second type of edit may be a non- substantive edit which the one or more machine-learned models determines (with some confidence threshold level being satisfied) is not likely to have affected the content, structure,and meaning of the document. Such non-substantive edits may include correcting typographical or grammatical errors (e.g., spelling, punctuation, etc ), changes in formatting (e.g., changing font size), etc. Therefore, computing resources (e.g., processor cycles) may be conserved by selectively implementing the one or more machine-learned models based on the types of edits to the document.

[0054] As an example implementation, local edits made to a document over time via a computing system can cause unintended changes to the global meaning of a document. For example, a document may be directed to a story with multiple characters where the user edits a particular local fragment that describes a character’s appearance, style and tone of voice. In some implementations, the computing system may be configured to store each edit (e.g., in a local memory, at a server computing system, in a database, etc.). The edits may be stacked over each other, together with a final variant. According to the examples of the disclosure, the computing system may include one or more machine-learned models configured to analyze or scan the entire document for consistency. In some implementations, in response to identifying outdated fragments (e.g., referring to the old appearance, using the old voice tone) the computing system (e.g., the one or more machine-learned models) may be configured to automatically update the outdated fragments. In some implementations, in response to identifying outdated fragments (e.g., referring to the old appearance, using the old voice tone) the computing system (e.g., the one or more machine-learned models) may be configured to provide an output (e.g., via comments, via proposed edits) for presentation to the user to make changes to the document to improve the consistency of the overall document. Thus, although edits to a document may be local, certain changes may change the tone or meaning of the entire document, and as described according to this example implementation, the user can keep track of those changes at a global level.

[0055] As another example implementation, a user may want to intentionally and explicitly make changes that affect multiple parts of a document. In some implementations, the computing system may be configured to provide a user interface having various control knobs (e.g., user interface elements) configured to enable a user to edit a document at various levels (e.g., a high or top level, an intermediate level, a low-level, etc.). For example, the computing system may be configured to provide a user interface having various control knobs (e.g., user interface elements) configured to enable a user to edit a hierarchical storyline of a video atvarious levels. For example, the user may be able to edit it at a top-level (e.g., a character ventures through the woods and makes new friends), intermediate level (e.g., the character no longer uses a boat to travel from A to B, but a bike through the woods) or at a low-level (e.g., the character accidentally slips).

[0056] The control knobs (e.g., user interface elements) provided to the user enables the user to make changes at any level. Further, the one or more machine-learned models may be configured to implement a plurality of edits automatically to follow the user inputs changing the document at various levels or to propose changes in accordance with the user inputs. In some implementations, the one or more machine-learned models may be configured to enable the user to bulk accept all or a plurality of edits. In some implementations, the one or more machine- learned models may be configured to enable the user to accept edits one by one. After the changes are complete, the one or more machine-learned models may be configured to analyze or scan the entire document for consistency.

[0057] As another example implementation, a document to be edited may include a computing program. The computing system may include one or more machine-learned models configured to maintain an overall intent or goal with respect to the computing program in response to edits made to the computing program (e.g., changes to code) which can cause performance of the computing program to worsen. For example, a prompt may indicate the intent or goal associated with the document (e.g., “This software enables a latency-efficient approach to content management”). For example, the computing system (e.g., the one or more machine-learned models) may be configured to determine whether a particular edit (or a plurality of edits) may make the overall system less efficient and not match or satisfy the overall prompt anymore. The computing system may receive inputs from a user editing a particular function (e.g., when editing an article, the user introduces numerous search function calls). The computing system (e.g., the one or more machine-learned models) may determine the edits make the overall system less efficient and cause the overall prompt to not be satisfied any longer. In response to this determination, the computing system (e.g., the one or more machine-learned models) may be configured to provide an output indicating the particular edits are worsening the performance of the computing program. For example, the output may include a comment or notification such as: “It looks like you are introducing multiple slow calls to the search function. The content management system is no longer latency efficient.” The computing system (e.g., theone or more machine-learned models) may be configured to provide or automatically implement a solution that can satisfy the overall intent of the prompt. For example, the computing system (e.g., the one or more machine-learned models) may be configured to provide, for presentation to the user, alternative implementations to edit the code (e.g., in a more efficient manner) without affecting overall latency negatively or making a more minimal change.

[0058] In some implementations, the computing system (e.g., the one or more machine- learned models) may be configured to determine whether an overall intent or goal with respect to the computing program is maintained in response to edits made to the computing program, if particular criteria are satisfied. For example, the particular criteria may include that the code in the computing program can be compiled. Therefore, computing resources (e.g., processor cycles) may be conserved by having the particular criteria be satisfied before implementing the one or more machine-learned models.

[0059] As another example implementation, a document may be edited by a plurality of users which can cause meaning drift with respect to the document when edits by the plurality of users conflict with one another. For example, when different users simultaneously (or disjointedly) edit the document conflicting changes may be made to completely different parts of the content. In some implementations, the computing system (e.g., the one or more machine- learned models) may be configured to determine whether an overall intent or goal with respect to the document is maintained in response to the edits made to the document by the plurality of users. In some implementations, the one or more machine-learned models may be configured to be implemented in response to a determination that the plurality of users have edited the document and after a particular number of changes (edits) have been made by the plurality of users. In some implementations, the one or more machine-learned models may be configured to be implemented in response to a determination that the plurality of users have edited the document and after a pause (e.g., a predetermined duration of time) since a last edit (change) has occurred by any one particular user and before a next edit. In some implementations, the one or more machine-learned models may be configured to perform batch processing with respect to the changes made by the plurality of users, rather than be implemented with respect to respective changes by individual users. In some implementations, the one or more machine-learned models may be configured to be implemented in response to a determination that the plurality of users have edited the document and in the case of the edits including changes to code in a computingprogram that the code is capable of being compiled. Therefore, computing resources (e.g., processor cycles) may be conserved by selectively implementing the one or more machine- learned models according to particular criteria being satisfied (e.g., a certain number of edits, a certain delay after a last edit, a condition that code in a document which has been edited be capable of being compiled, etc.).

[0060] In some implementations, the computing system (e.g., the one or more machine- learned models) may be configured to freeze or suspend a capability for users to edit or change the document to allow for sufficient time for the computing system (e.g., the one or more machine-learned models) to determine whether an overall intent or goal with respect to the document is maintained in response to the edits made to the document by the plurality of users.

[0061] In some implementations, the computing system may be configured to provide, for presentation to the user, a user interface having control knobs (e.g., user interface elements) which are configured to enable the plurality of users to revert or resolve inconsistencies in the document (e.g., through a chat interface), whereby the inputs to the user interface may be provided to the one or more machine-learned models to reevaluate whether an overall intent or goal with respect to the document is maintained in response to the edits made to the document by the plurality of users.

[0062] One or more technical benefits of the disclosure include ensuring that the meaning (e.g., the goal or intent) with respect to content in a document remains consistent in response to edits to the document by one or more users. Therefore, the accuracy and quality of a document can be maintained. Further, as described herein, in some implementations computing resources can be conserved or efficiently utilized by selectively implementing one or more machine- learned models to determine whether an overall (global) intent or goal with respect to the document is maintained in response to the edits made to the document by the one or more users. Further, as described herein, in some implementations computing resources can be conserved or efficiently utilized by detecting changes to a computing program (e.g., changes to code in the computing program) which result in the edited computing program not satisfying the goal or intent of the computing program (e.g., negatively affect the performance of the computing program).

[0063] Therefore, aspects of the disclosure provide technical effects, benefits, and / or improvements in computing technology and the technology of generating accurate documentswhich satisfy specifications or specified criteria, via one or more computing devices (e.g., a user computing device, a server computing system, and combinations thereof) by implementing one or more machine-learned models to detect a meaning drift caused by edits to the document, as described herein.

[0064] Referring now to the drawings, FIG. 1 A is an example system according to one or more example embodiments of the disclosure. FIG. 1A illustrates an example of a system 1000 which includes a computing device 100, an external computing device 200, a server computing system 300, and external content 500, which may be in communication with one another over a network 400. For example, the computing device 100 and the external computing device 200 can include any of a personal computer, a smartphone, a tablet computer, a laptop, a global positioning service device, a smartwatch, and the like. The network 400 may include any type of communications network including a wired or wireless network, or a combination thereof. The network 400 may include a local area network (LAN), wireless local area network (WLAN), wide area network (WAN), personal area network (PAN), virtual private network (VPN), or the like. For example, wireless communication between elements of the example embodiments may be performed via a wireless LAN, Wi-Fi, Bluetooth, ZigBee, Wi-Fi direct (WFD), ultra wideband (UWB), infrared data association (IrDA), Bluetooth low energy (BLE), near field communication (NFC), a radio frequency (RF) signal, and the like. For example, wired communication between elements of the example embodiments may be performed via a pair cable, a coaxial cable, an optical fiber cable, an Ethernet cable, and the like. Communication over the network 400 can use a wide variety of communication protocols (e.g., TCP / IP, HTTP, SMTP, FTP), encodings or formats (e.g., HTML, XML), and / or protection schemes (e.g., VPN, secure HTTP, SSL).

[0065] As will be explained in more detail below, in some implementations the computing device 100 and / or server computing system 300 may form part of an application system which can provide a tool for users to generate, manage, or edit a document (e.g., a text document, via one or more machine-learned models. A document can include any recorded information or material, and may include text, computer programs, images, audio, video, or any other medium that can convey information.

[0066] In some example embodiments, the server computing system 300 may obtain data from one or more of a document edits data store 350 and a machine-learned model data store370, to implement various operations and aspects of the application system as disclosed herein. The document edits data store 350 and machine-learned model data store 370 may be integrally provided with the server computing system 300 (e.g., as part of the one or more memory devices 320 of the server computing system 300) or may be separately (e.g., remotely) provided. Further, document edits data store 350 and machine-learned model data store 370 can be combined as a single data store (database) or may include a plurality of respective data stores. Data stored in one data store (e.g., the document edits data store 350) may overlap with some data stored in another data store. In some implementations, one data store (e.g., the machine- learned model data store 370) may reference data that is stored in another data store (e.g., the document edits data store 350).

[0067] In some implementations, the document edits data store 350 can store a record (history) of edits to a document over time, prior versions of a document, updated versions of a document, differences between an updated version of a document and a prior version of a document, etc. In some implementations, the information stored in the document edits data store 350 can be associated with and / or stored according to a particular user or a plurality of users. In some implementations, the information stored in the document edits data store 350 can be associated with and / or stored according to a particular type of document. In some implementations, the information stored in the document edits data store 350 can be associated with and / or stored according to a particular application that is associated with the document (e.g., a particular software application that is executed based on code from the document, a particular application that displays imagery, video, or outputs audio, that is associated with the document, etc.).

[0068] Machine-learned model data store 370 can store machine-learned models which can be retrieved and implemented by the server computing system 300 for generating distilled or fine-tuned machine-learned models (e.g., distilled or fine-tuned generative machine-learned models) that, in some implementations, can also be provided to the computing device 100. Machine-learned model data store 370 can also store distilled or fine-tuned machine-learned models (e.g., distilled or fine-tuned generative machine-learned models) which can be retrieved and implemented by the computing device 100. In some implementations, the computing device 100 can retrieve and implement machine-learned models which are large parameter models that have not been fine-tuned or distilled. The machine-learned models (including large parametermodels and distilled or fine-tuned models) stored at the machine-learned model data store 370 can include generative machine-learned models respectively associated with different types of applications or types of documents (e.g., pertaining to different genres or subjects and / or to different types of media) that may be implemented across a variety of domains (e.g., artistic, healthcare, gaming, education, financial, customer engagement such as chatbots and virtual assistants, engineering / science, entertainment, etc.). The machine-learned models may include large language models (e.g., the Bidirectional Encoder Representations from Transformers (BERT) large language model) and general, multimodal models (e.g., Gemini). The machine- learned models may include generative artificial intelligence (Al) models (e.g., Bard) which may implement generative adversarial networks (GANs), transformers, variational autoencoders (VAEs), neural radiance fields (NeRFs), and the like.

[0069] External content 500 can be any form of external content including news articles, webpages, video files, audio files, written descriptions, ratings, game content, social media content, photographs, commercial offers, transportation method, weather conditions, sensor data obtained by various sensors, or other suitable external content. The computing device 100, external computing device 200, and server computing system 300 can access external content 500 over network 400. External content 500 can be searched by computing device 100, external computing device 200, and server computing system 300 according to known searching methods and search results can be ranked according to relevance, popularity, or other suitable attributes, including location-specific filtering or promotion.

[0070] FIG. IB is an example block diagram of a computing system, according to one or more example embodiments of the disclosure. Referring now to FIG. IB, example block diagrams of a system 1100 including a computing device 100 and server computing system 300 according to one or more example embodiments of the disclosure will now be described. Although computing device 100 is represented in FIG. IB, features of the computing device 100 described herein are also applicable to the external computing device 200.

[0071] The computing device 100 may include one or more processors 110, one or more memory devices 120, an application system 130, a position determination device 140, an input device 150, a display device 160, an output device 170, and a capture device 180. The server computing system 300 may include one or more processors 310, one or more memory devices 320, and an application system 330.

[0072] For example, the one or more processors 110, 310 can be any suitable processing device that can be included in a computing device 100 or server computing system 300. For example, the one or more processors 110, 310 may include one or more of a processor, processor cores, a controller and an arithmetic logic unit, a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), an image processor, a microcomputer, a field programmable array, a programmable logic unit, an application-specific integrated circuit (ASIC), a microprocessor, a microcontroller, etc., and combinations thereof, including any other device capable of responding to and executing instructions in a defined manner. The one or more processors 110, 310 can be a single processor or a plurality of processors that are operatively connected, for example in parallel.

[0073] The one or more memory devices 120, 320 can include one or more non- transitoiy computer-readable storage mediums, including a Read Only Memory (ROM), Programmable Read Only Memory (PROM), Erasable Programmable Read Only Memory (EPROM), and flash memory, a USB drive, a volatile memory device including a Random Access Memory (RAM), a hard disk, floppy disks, a Blu-ray disk, or optical media such as CD ROM discs and DVDs, and combinations thereof. However, examples of the one or more memory devices 120, 320 are not limited to the above description, and the one or more memory devices 120, 320 may be realized by other various devices and structures as would be understood by those skilled in the art.

[0074] For example, the one or more memory devices 120 can also include data 122 and instructions 124 that can be retrieved, manipulated, created, or stored by the one or more processors 110. In some example embodiments, such data can be accessed and used as input to implement meaning drift detection application 132, and to execute the instructions to perform operations including: obtaining a prompt indicating an intended meaning of a document; receiving a plurality of inputs editing a first version of the document to produce a second version of the document; determining whether one or more specified editing criteria associated with editing the document are satisfied; in response to the one or more specified editing criteria being satisfied, implementing one or more machine-learned models to determine whether specified drift criteria associated with the intended meaning of the document is satisfied, based on the prompt, the edits to the first version of the document, and the second version of the document;and providing an output based on whether the specified drift criteria associated with the intended meaning of the document is satisfied, as described according to examples of the disclosure.

[0075] For example, the one or more memory devices 320 can also include data 322 and instructions 324 that can be retrieved, manipulated, created, or stored by the one or more processors 310. In some example embodiments, such data can be accessed and used as input to implement meaning drift detection application 332, and to execute the instructions to perform operations including: obtaining a prompt indicating an intended meaning of a document; receiving a plurality of inputs editing a first version of the document to produce a second version of the document; determining whether one or more specified editing criteria associated with editing the document are satisfied; in response to the one or more specified editing criteria being satisfied, implementing one or more machine-learned models to determine whether specified drift criteria associated with the intended meaning of the document is satisfied, based on the prompt, the edits to the first version of the document, and the second version of the document; and providing an output based on whether the specified drift criteria associated with the intended meaning of the document is satisfied, as described according to examples of the disclosure.

[0076] In some example embodiments, the computing device 100 includes an application system 130. For example, the application system 130 may include the meaning drift detection application 132. The application system 130 can include various other applications including document applications, text messaging applications, email applications, dictation applications, virtual keyboard applications, browser applications, map applications, social media applications, navigation applications, etc.

[0077] According to examples of the disclosure, the meaning drift detection application 132 may be executed by the computing device 100 to provide a user of the computing device 100 a way to generate, manage, and edit documents. In some implementations, the meaning drift detection application 132 may be part of another application (e.g., a document application, text messaging application, email application, dictation application, virtual keyboard application, browser application, map application, social media application, navigation application, etc.) or may be a standalone application. The meaning drift detection application 132 may be configured to be dynamically interactive according to various user inputs. Example implementations of the meaning drift detection application 132 are described herein, however the disclosure is notlimited to these examples as various modifications may be made to the embodiments described herein.

[0078] In some examples, one or more aspects of the meaning drift detection application 132 may be implemented by the meaning drift detection application 332 of the server computing system 300 which may be remotely located, to generate, manage, or edit a document, via one or more machine-learned models, in response to receiving an input from a user via the computing device 100. In some examples, one or more aspects of the meaning drift detection application 332 may be implemented by the meaning drift detection application 132 of the computing device 100, to generate, manage, or edit a document, via one or more machine-learned models, in response to receiving an input from a user via the computing device 100.

[0079] In some example embodiments, the computing device 100 includes a position determination device 140. Position determination device 140 can determine a current geographic location of the computing device 100 and communicate the geographic location to server computing system 300 over network 400. The position determination device 140 can be any device or circuitry for analyzing the position of the computing device 100. For example, the position determination device 140 can determine actual or relative position by using a satellite navigation positioning system (e.g. a GPS system, a Galileo positioning system, the GLObal Navigation satellite system (GLONASS), the BeiDou Satellite Navigation and Positioning system), an inertial navigation system, a dead reckoning system, based on an IP address, by using triangulation and / or proximity to cellular towers or WiFi hotspots, and / or other suitable techniques for determining a position of the computing device 100. For example, in some implementations the meaning drift detection application 132 may be configured to generate or provide suggestions or recommendations with respect to content in a document based on a location of the computing device 100 as determined via the position determination device 140.

[0080] The computing device 100 may include an input device 150 configured to receive an input from a user and may include, for example, one or more of a keyboard (e.g., a physical keyboard, virtual keyboard, etc ), a mouse, a joystick, a button, a switch, an electronic pen or stylus, a gesture recognition sensor (e.g., to recognize gestures of a user including movements of a body part), an input sound device or speech recognition sensor (e.g., a microphone to receive a voice input such as a voice command or a voice query), a track ball, a remote controller, a portable (e.g., a cellular or smart) phone, a tablet PC, a pedal or footswitch, a virtual -realitydevice, and so on. The input device 150 may also be embodied by a touch-sensitive display having a touchscreen capability, for example. For example, the input device 150 may be configured to receive an input from a user associated with the input device 150 for providing a prompt which can identify an intent or goal with respect to a document (e.g., “this computer program enables a latency-efficient approach to content management,” “this video should have smooth transitions between scenes and should have high quality sound without background noise or distortion,” “this picture should have minimal artifacts and a high dynamic range with accurate and vibrant colors,” etc.). For example, the input device 150 may be configured to receive an input from a user associated with the input device 150 for providing edits to the document, for communicating with other users, for accepting or declining suggestions or recommendations provided by the computing system with respect to edits to the document, etc.

[0081] The computing device 100 may include a display device 160 which displays information viewable by the user (e.g., a user interface screen). For example, the display device 160 may be a non-touch sensitive display or a touch-sensitive display. The display device 160 may include a liquid crystal display (LCD), a light emitting diode (LED) display, an organic light emitting diode (OLED) display, active matrix organic light emitting diode (AMOLED), flexible display, 3D display, a plasma display panel (PDP), a cathode ray tube (CRT) display, and the like, for example. However, the disclosure is not limited to these example displays and may include other types of displays. The display device 160 can be used by the application system 130 provided at the computing device 100 to display information to a user relating to an input (e.g., information relating to the document, information relating to generating, managing, and editing a document, etc., user interface screens having user interface elements which are selectable by the user for generating, managing, or editing the document, etc.).

[0082] The computing device 100 may include an output device 170 to provide an output to the user and may include, for example, one or more of an audio device (e.g., one or more speakers), a haptic device to provide haptic feedback to a user (e.g., a vibration device), a light source (e.g., one or more light sources such as LEDs which provide visual feedback to a user), a thermal feedback system, and the like.

[0083] The computing device 100 may include a capture device 180 that is capable of capturing media content, according to various examples of the disclosure. For example, the capture device 180 can include an image capturer 182 (e.g., a camera) which is configured tocapture images (e.g., photos, video, and the like). For example, the capture device 180 can include a sound capturer 184 (e.g., a microphone) which is configured to capture sound or audio (e.g., an audio recording). The media content captured by the capture device 180 may be transmitted to one or more of the server computing system 300, document edits data store 350 and machine-learned model data store 370, for example, via network 400. For example, in some implementations, content which is captured by the capture device 180 may be provided as an input to one or more machine-learned models to generate, manage, or edit a document (e.g., an image, a video, an audio recording, etc.), for example.

[0084] In accordance with example embodiments of the disclosure, the server computing system 300 can include one or more processors 310 and one or more memory devices 320 as described herein. The server computing system 300 may also include an application system 330 which is similar to the application system 130 described herein.

[0085] For example, the application system 330 may include the meaning drift detection application 332 which performs functions similar to those discussed herein with respect to meaning drift detection application 132. In some implementations, one or more machine-learned models (e.g., generative machine-learned models, large language models, etc.) associated with the application system 330 may be configured to generate, manage, or edit a document, via one or more machine-learned models.

[0086] For example, one or more machine-learned models (e.g., generative machine- learned models, large language models, etc.) associated with the application system 330 may be configured to perform a first action (e.g., generate or provide recommendations when the one or machine-learned models determines edits to a document change the overall meaning of the document by more than a threshold level), while the computing device 100 may be configured to perform a second action (e.g., receive inputs from a user accepting or declining the suggestion or recommendation with respect to maintaining the intent or goal with respect to the document). For example, one or more machine-learned models (e.g., generative machine-learned models, large language models, etc.) associated with the application system 130 may be configured to perform a first action (e.g., receive edits to a document), while the server computing system 300 may be configured to perform a second action (e.g., implementing one or machine-learned models to determine whether edits to the document changes the overall meaning of the document by more than the threshold level).

[0087] Examples of the disclosure are directed to computer implemented methods for generating, managing, and editing a document via one or more machine-learned models and for providing a user interface for generating, managing, and editing the document by implementing one or more machine-learned models with respect to the edited document.

[0088] The flow diagram of FIG. 2 illustrates a method 2000 for generating, managing, and editing a document by implementing one or more machine-learned models. Although shown in a particular sequence or order, unless otherwise specified, the order of the processes can be modified. Thus, the illustrated embodiments should be understood only as examples, and the illustrated processes can be performed in a different order, and some processes can be performed in parallel. Additionally, one or more processes can be omitted in various embodiments. Thus, not all processes are required in every embodiment. Other process flows are possible.

[0089] Referring to FIG. 2, at operation 2100 the method 2000 includes a computing device receiving a prompt from a user which is associated with an intended meaning with respect to generating, managing, or editing a document, according to examples of the disclosure. As described herein, the computing device may be embodied as computing device 100, server computing system 300, or combinations thereof. For example, the document can include any recorded information or material, and may include text, computer programs, images, audio, video, or any other medium that can convey information. For example, the document may be associated with any kind of application (e.g., a document application, text messaging application, email application, media (audio, visual) application, browser application, map application, social media application, navigation application, computer programming application, etc.). For example, the input providing the prompt may be provided by the user via the input device 150. For example, the input providing the prompt may be provided by the user via a voice input, the selection of a user interface element which suggests possible prompts, via a text entry, etc. For example, the prompt may indicate an intent or goal with respect to the document (e.g., “this computer program enables a latency -efficient approach to content management,” “this video should have smooth transitions between scenes and should have high quality sound without background noise or distortion,” “this picture should have minimal artifacts and a high dynamic range with accurate and vibrant colors,” etc.). The intent or goal may be a measurable, objective, feature that defines a specification of the document (e.g., a particular quality or attribute of the document, performance characteristics associated with code in a computing program, etc.). Theintent or goal indicated in the prompt may indicate a level of consistency and cohesiveness that can also be defined or measured (e.g., based on training data).

[0090] At operation 2200 the method 2000 includes the computing device receiving one or more inputs editing the document. For example, one or more inputs may be received editing a first version of the document to produce a second version of the document. For example, the first version of the document may be the original document, or may be a previously edited version of the document. The second version of the document may correspond to an updated version of the document after the one or more edits have been received via the one or more inputs. The edits (changes) to the document may include additions, deletions, and / or modifications to the content of the document (e.g., adding or rearranging content in a paper, removing noise from an audio recording, rearranging parts of the audio recording, applying special effects, graphic elements, or filters to an image, adding transitions to a video, stabilizing footage of a video, adjusting colors and tones in a video or image, combining footage from different videos, creating a collage of images, etc.).

[0091] At operation 2300 the method 2000 includes the computing device implementing one or more machine-learned models to determine whether the intended meaning has deviated based on the prompt from operation 2100 and the edits received at operation 2200.

[0092] As described herein, in some implementations, operation 2300 may be selectively performed.

[0093] FIG. 3 illustrates a flow diagram of an example, non-limiting computer- implemented method, according to one or more example embodiments of the disclosure. The flow diagram of FIG. 3 illustrates a method 3000 for selectively implementing the one or more machine-learned models at operation 2300. Although shown in a particular sequence or order, unless otherwise specified, the order of the processes can be modified. Thus, the illustrated embodiments should be understood only as examples, and the illustrated processes can be performed in a different order, and some processes can be performed in parallel. Additionally, one or more processes can be omitted in various embodiments. Thus, not all processes are required in every embodiment. Other process flows are possible.

[0094] At operation 3100 the method 3000 includes receiving edits to the document, for example, as described with respect to operation 2200 of FIG. 2. At operation 3200 the method 3000 includes the computing device (e.g., meaning drift detection application 6100 from FIG. 6)determining whether the number of edits made at operation 3100 has exceeded a threshold level. For example, the threshold level may correspond to a specified number of edits (e.g., five, ten, etc.).

[0095] If the number of edits made at operation 3100 has exceeded the threshold level, the method may continue to operation 3300 and the one or more machine-learned models may be implemented to determine whether a global (overall) meaning of the second version of the document has drifted (changed or deviated) beyond a threshold drift level compared to the global (overall) meaning of the first version of the document which is ascertained based on the prompt provided by the user (e.g., at operation 2100 of FIG. 2). Operation 3300 may be similar to operation 2300 of FIG. 2.

[0096] If the number of edits made at operation 3100 has not exceeded the threshold level, the method may return to operation 3200 to monitor for additional edits to the document. Therefore, computing resources (e.g., processor cycles) may be conserved by selectively implementing the one or more machine-learned models according to particular criteria being satisfied (e.g., the certain number of edits) so that the one or more machine-learned models are not implemented on a repeated basis needlessly.

[0097] FIG. 4 illustrates a flow diagram of an example, non-limiting computer- implemented method, according to one or more example embodiments of the disclosure. The flow diagram of FIG. 4 illustrates a method 4000 for selectively implementing the one or more machine-learned models at operation 2300. Although shown in a particular sequence or order, unless otherwise specified, the order of the processes can be modified. Thus, the illustrated embodiments should be understood only as examples, and the illustrated processes can be performed in a different order, and some processes can be performed in parallel. Additionally, one or more processes can be omitted in various embodiments. Thus, not all processes are required in every embodiment. Other process flows are possible.

[0098] At operation 4100 the method 4000 includes receiving edits to the document, for example, as described with respect to operation 2200 of FIG. 2. At operation 4200 the method 4000 includes the computing device (e.g., meaning drift detection application 6100 from FIG. 6) determining whether the time since the last edit made at operation 4100 has exceeded a threshold duration of time. For example, the threshold duration of time may correspond to a specified time (e.g., five seconds, ten seconds, one minute, etc ).

[0099] If the time since the last edit has exceeded the threshold duration of time, the method may continue to operation 4300 and the one or more machine-learned models may be implemented to determine whether a global (overall) meaning of the second version of the document has drifted (changed or deviated) beyond a threshold drift level compared to the global (overall) meaning of the first version of the document which is ascertained based on the prompt provided by the user (e.g., at operation 2100 of FIG. 2). Operation 4300 may be similar to operation 2300 of FIG. 2.

[0100] If the time since the last edit has not exceeded the threshold duration of time, the method may return to operation 4200 to monitor for a further edit or to wait until the threshold duration of time is exceeded. Therefore, computing resources (e.g., processor cycles) may be conserved by selectively implementing the one or more machine-learned models according to particular criteria being satisfied (e.g., a pause of a certain length occurring since a last edit) so that the one or more machine-learned models are not implemented on a repeated basis needlessly.

[0101] FIGS. 3 and 4 are example methods for selectively performing operation 2300 of FIG. 2. However, the disclosure is not limited to the example of FIGS. 3 and 4 and other conditions or criteria may be utilized by the computing device to determine whether to perform operation 2300 after an edit to a document by one or more users. For example, the computing device (e.g., meaning drift detection application 6100 from FIG. 6) may be configured to determine whether a threshold duration of time has passed since the last time the one or more machine-learned models were implemented at operation 2300, and if the threshold duration of time has passed, the one or more machine-learned models may be implemented again to determine whether a global (overall) meaning of the second version of the document has drifted (changed or deviated) beyond a threshold drift level compared to the global (overall) meaning of the first version of the document which is ascertained based on the prompt provided by the user (e.g., at operation 2100 of FIG. 2). For example, the threshold duration of time may correspond to a specified time (e.g., five seconds, ten seconds, one minute, etc.). Therefore, computing resources (e.g., processor cycles) may be conserved by selectively implementing the one or more machine-learned models according to particular criteria being satisfied (e.g., a certain length of time occurring since the last implementation of the one or more machine-learned models) so that the one or more machine-learned models are not implemented on a repeated basis needlessly.

[0102] For example, when the edits to a document include changing code for a computing program, the one or more machine-learned models may be configured to be implemented only if the code is capable of being compiled. Therefore, computing resources (e.g., processor cycles) may be conserved by selectively implementing the one or more machine- learned models according to particular criteria being satisfied (e.g., a condition that code in a document which has been edited be capable of being compiled).

[0103] The one or more machine-learned models may be configured to determine whether the intended (global) meaning has drifted (changed, deviated), based on the prompt from operation 2100 and the edits received at operation 2200. The one or more machine-learned models may include a zero-shot model where the foundation model is given a prompt such as “Your task is to keep track of the global meaning of a document and to detect conflicts due to updates to a document,” where the updates are provided as one of the inputs to the model. In some implementations, the zero-shot model may also be fine-tuned, or parameter-efficient finetuned, based on training data. For example, the one or more machine-learned models are provided with the first version of the document (e.g., the original document), the detailed edits (which may be stored, for example, in the document edits data store 350), and the second version of the document (e.g., the final or updated document) having the edits applied. The one or more machine-learned models may be configured to output an observation and an updated global meaning, In some implementations, the one or more machine-learned models may be configured to provide feedback to the individual edits and reject some of them with a motivation or rationale (e.g., an output of “accepting this edit results in a complete global meaning change - are you sure?”). In some implementations, the one or more machine-learned models may be configured to recognize that some edits are in conflict with each other and automatically resolve the conflict or provide feedback to the user (e.g., “it looks like the edit in this part of the document is contrary to the edit in this other part”) and provide a resolution mechanism.

[0104] In some implementations, the one or more machine-learned models may be configured to analyze or examine a first difference between the second version of the document and the first version of the document. In some implementations, the computing device (e.g., the meaning drift detection application 6100 which includes the one or more machine-learned models) may determine whether the first difference exceeds a first threshold difference level. If the first difference exceeds the first threshold difference level, the computing device (e.g., themeaning drift detection application 6100 which includes the one or more machine-learned models) may determine the intended (global) meaning of the document has deviated. If the first difference does not exceed the first threshold difference level, the computing device (e.g., the meaning drift detection application 6100 which includes the one or more machine-learned models) may determine the intended (global) meaning of the document has not deviated. At operation 2400 the computing device can provide an output according to whether the intended meaning of the document has deviated as determined at operation 2300.

[0105] In some implementations, the one or more machine-learned models may be configured to first assess the global (overall) meaning of the second version of the document and then analyze or examine a second difference between the global meaning associated with the second version of the document and the global (intended) meaning associated with the first version of the document. In some implementations, the computing device (e g., the meaning drift detection application 6100 which includes the one or more machine-learned models) may determine whether the second difference exceeds a second threshold difference level. If the second difference exceeds the second threshold difference level, the computing device (e.g., the meaning drift detection application 6100 which includes the one or more machine-learned models) may determine the intended (global) meaning of the document has deviated. If the second difference does not exceed the second threshold difference level, the computing device (e.g., the meaning drift detection application 6100 which includes the one or more machine- learned models) may determine the intended (global) meaning of the document has not deviated. At operation 2400 the computing device can provide an output according to whether the intended meaning of the document has deviated as determined at operation 2300.

[0106] In some implementations, the computing device (e.g., the meaning drift detection application 6100 which includes the one or more machine-learned models) may determine whether drift criteria associated with the intended meaning for the document is satisfied (e.g., based on the prompt, a first version of the document, the edits to the first version of the document, and the second version of the document). For example, if the drift criteria is satisfied, the computing device (e.g., the meaning drift detection application 6100 which includes the one or more machine-learned models) may determine the intended (global) meaning of the document has deviated. For example, if the drift criteria is not satisfied, the computing device (e.g., the meaning drift detection application 6100 which includes the one or more machine-learnedmodels) may determine the intended (global) meaning of the document has not deviated. For example, the computing device (e.g., the meaning drift detection application 6100 which includes the one or more machine-learned models) may determine the intended (global) meaning of the document has deviated if the second version of the document does not satisfy certain specifications (e.g., performance requirements such as a latency specification for a computing program that is edited). For example, the computing device (e.g., the meaning drift detection application 6100 which includes the one or more machine-learned models) may determine the intended (global) meaning of the document has deviated if the second version of the document includes a threshold amount of content which conflicts or is contrary to specifications provided in the prompt (e.g., content in a second part of a document conflicts with edits made to content in a first part of the document, such as an edited description of an appearance of a character is inconsistent with a description of the appearance of the character in another part of the document). For example, the computing device (e.g., the meaning drift detection application 6100 which includes the one or more machine-learned models) may determine specified drift criteria associated with the intended meaning of the document is satisfied (the intended or global meaning of the document has deviated) if the latency associated with the updated computing program (second version of the document) is greater than a threshold latency level which can be indicated in the prompt. For example, after editing a document which includes at least one of an image, a video, or an audio recording, the computing device (e.g., the meaning drift detection application 6100 which includes the one or more machine-learned models) may determine specified drift criteria associated with the intended meaning of the document is satisfied (the intended or global meaning of the document has deviated) if a resulting quality of the second version of the document (e.g., the updated image, update video, updated audio recording) is less than a threshold quality level which can be indicated in the prompt.

[0107] FIG. 5 illustrates a flow diagram of an example, non-limiting computer- implemented method, according to one or more example embodiments of the disclosure. The flow diagram of FIG. 5 illustrates a method 5000 for providing an output based on whether the intended (global) meaning of the document has deviated as determined at operation 2300 of FIG. 2. Although shown in a particular sequence or order, unless otherwise specified, the order of the processes can be modified. Thus, the illustrated embodiments should be understood only as examples, and the illustrated processes can be performed in a different order, and some processescan be performed in parallel. Additionally, one or more processes can be omitted in various embodiments. Thus, not all processes are required in every embodiment. Other process flows are possible.

[0108] At operation 5100 the method 5000 includes determining whether the intended (global) meaning of the document has deviated as determined at operation 2300 of FIG. 2. When the computing device (e.g., the meaning drift detection application 6100 which includes the one or more machine-learned models) determines the intended (global) meaning has drifted (changed, deviated), based on the prompt from operation 2100 and the edits received at operation 2200, the method 5000 can continue to operation 5300 and / or operation 5400. For example, at operation 5300, the meaning drift detection application 6100 may be configured to automatically revert or revise the second version of the document so that the intended (global) meaning of the document is maintained with respect to the first version of the document and / or with respect to the intent provided in the prompt.

[0109] For example, at operation 5400, the meaning drift detection application 6100 may be configured to provide, for presentation to the user (e.g., via the output device 170), a notification (e.g., feedback) and / or a recommendation to revert or revise the second version of the document. In some implementations, the recommendation may include a resolution mechanism (e.g., selectable user interface elements that, when selected, can reverse, add, or modify edited content and / or can edit other portions of content so that the overall meaning of the document is maintained). For example, the output may include a comment or notification such as: “It looks like you are introducing multiple slow calls to the search function. The content management system is no longer latency efficient.” The meaning drift detection application 6100 (e.g., the one or more machine-learned models) may be configured to provide or automatically implement a solution that can satisfy the overall intent of the prompt. For example, the meaning drift detection application 6100 (e.g., the one or more machine-learned models) may be configured to provide, for presentation to the user, alternative implementations to edit code in a document (e.g., to be executed in a more efficient manner) without affecting overall latency negatively or making a more minimal change.

[0110] When the computing device (e.g., the meaning drift detection application 6100 which includes the one or more machine-learned models) determines the intended (global) meaning has not drifted (changed, deviated), based on the prompt from operation 2100 and theedits received at operation 2200, the method 5000 can continue to operation 5200. For example, at operation 5200, the edits to the document can be accepted. In some implementations, the meaning drift detection application 6100 may accept the edits without notifying the user such that implementation of the one or more machine-learned models is transparent to (hidden from) the user. In some implementations, the meaning drift detection application 6100 may accept the edits and notify the user that the edits were accepted so that the user is assured the edits have not changed the overall meaning of the document and that the document remains consistent and cohesive according to the intent provided in the prompt.

[0111] FIG. 6 illustrates a block diagram of a meaning drift detection application, according to one or more example embodiments of the disclosure.

[0112] Referring to FIG. 6, the meaning drift detection application 6100 (which may correspond to meaning drift detection application 132 and / or meaning drift detection application 332) may include a conditioning parameters generator 3110, one or more sequence processing models 3120, one or more large language models 3130, and one or more generative machine- learned models 3140. The meaning drift detection application 6100 may receive inputs 6200 from a user as discussed above with respect to operation 2100 and operation 2200 of FIG. 2, operation 3100 of FIG. 3, and operation 4100 of FIG. 4. Conditioning parameters generator 6110 may be configured to generate conditioning parameters based at least in part on the input, wherein the conditioning parameters provide values for one or more conditions associated with content to be generated which relates at least in part to the input 6200, the first version of the document 6300 (e.g., an original document or a prior version of the document), and the second version of the document 6400 (e.g., the edited or updated version of the document). Other inputs to the meaning drift detection application 6100 may include information stored in the document edits data store 350, the machine-learned model data store 370, etc.

[0113] To generate the conditioning parameters, the conditioning parameters generator 6110 may be configured to retrieve values for the one or more conditions associated with the input. For example, to generate the conditioning parameters, the conditioning parameters generator 6110 may be configured to extract the values for the one or more conditions from the input. The input may include information indicative of the user's intent or requirements. In some implementations, the conditioning parameters generator 6110 (or the one or more sequence processing models 6120 or the one or more large language models 6130) may be configured toextract information from the input 6200 to identify values for the one or more conditions, and the conditioning parameters generator 6110 may be configured to generate the conditioning parameters based on the extracted values. For example, the input itself may identify an intent, purpose, or meaning for a document to be edited or generated (e.g., “this software enables a latency-efficient approach to content management”) or an attribute or feature associated with editing the document (e.g., “make the main character funny and speak with an Irish accent”) that can be used to generate the conditioning parameters for generating, managing, and editing the document.

[0114] To generate the conditioning parameters, the conditioning parameters generator 6110 may be configured to infer the values for the one or more conditions from the input. The input may include information indicative of the user's intent or requirements. In some implementations, the conditioning parameters generator 6110 (or the one or more sequence processing models 6120 or the one or more large language models 6130) may be configured to infer information from the input 6200 to identify values for the one or more conditions, and the conditioning parameters generator 6110 may be configured to generate the conditioning parameters based on the inferred values. For example, the input may include a reference to a quality or objective with respect to the document (“latency-efficient,” “funny,” etc.), and the conditioning parameters generator 6110 (or the one or more sequence processing models 6120 or the one or more large language models 6130) may be configured to infer a value based on the input. For example, an input requesting the meaning drift detection application 6100 to ensure an overall global meaning of the document (e.g., a computing program) is “latency-efficient” may infer a value based on the context of the document. For example, a computing program associated with a real-time environment may be associated with a latency on the order of microseconds to milliseconds, while a computing program associated with a networking environment may be associated with a latency on the order of ten milliseconds to a couple hundred milliseconds. For example, the meaning drift detection application 6100 may be configured to ascertain an inferred value based on information via external content 500 which may provide information about acceptable or standard values for a particular context.

[0115] In some implementations, the conditioning parameters generator 6110 may be configured to infer the values for the one or more conditions from the input by providing the input to one or more sequence processing models 6120, wherein the one or more sequenceprocessing models 6120 are configured to output the values for the one or more conditions in response to or based on the query. The one or more sequence processing models 6120 may include one or more machine-learned models which are configured to process and analyze sequential data and to handle data that occurs in a specific order or sequence, including time series data, natural language text, or any other data with a temporal or sequential structure.

[0116] The one or more sequence processing models 6120 may receive an input including text and tokenize the input by breaking down the sequence of text into small units (tokens) to provide a structured representation of the input sequence. The one or more sequence processing models 6120 may represent the tokens as vectors in a continuous vector space by mapping each token to a high-dimensional vector, where the relationships between tokens (words) are reflected in the geometric relationships between their corresponding vector. For example, the one or more sequence processing models 6120 may receive an input including the text “the main character wears a blue t-shirt” and tokenize the input by breaking down the sequence of text into small units (tokens) (e.g., “main character,” “wears,” “blue,” and “t-shirt”), thereby providing a structured representation of the input sequence. In a word embedding, semantically similar words are closer together in the vector space. For example, the vectors for "t-shirt" and "shirt" might be close to each other because of their semantic relationship, while the vectors for “blue” and “white” may be far apart compared to the vectors for “blue” and “cobalt”.

[0117] The one or more large language models 6130 can be, or otherwise include, a model that has been trained on a large corpus of language training data in a manner that provides the one or more large language models 6130 with the capability to perform multiple language tasks. For example, the one or more large language models 6130 can be trained to perform summarization tasks, conversational tasks, simplification tasks, oppositional viewpoint tasks, etc. In particular, the one or more large language models 6130 can be trained to process a variety of outputs to generate a language output. For example, the one or more large language models 6130 can process an embedding generated by a machine-learned embedding generation model, portions of source content (e.g., document chunk(s)) identified using an embedding generation model, language outputs generated using the one or more large language models 6130 or some other model, etc.

[0118] The one or more generative machine-learned models 6140 may include a deep neural network or a generative adversarial network (GAN), variational autoencoders, stablediffusion machine-learned models, visual transformers, neural radiance fields (NeRFs), etc., to generate content (e.g., an image, a video, an audio recording, a text document, etc.) with values for conditions associated with one or more features. For example, the computing device may include a database (e.g., machine-learned model data store 370) which is configured to store a plurality of generative machine-learned models respectively associated with a plurality of different types of content (e.g., different genres or subjects, different kinds of content including imagery, videos, audio, and text, different formats of content including outlines, reports, spreadsheets, different styles or purposes including to persuade, inform, entertain, explain, etc.). In some implementations, the computing device may be configured to retrieve, from among the one or more generative machine-learned models 6140, a generative machine-learned model associated with a particular type of content relating to the input (e.g., a generative machine- learned model associated with audio content, video content, imagery content, text content, etc.).

[0119] In some implementations, the one or more generative machine-learned models 6140 may be trained on a large dataset of content (e.g., a large corpus of language training data) with corresponding information about the conditions associated with the content. During training, the one or more generative machine-learned models 6140 learn relationships between elements in an output (e.g., content) and conditions that influence them. This may involve the computing device adjusting each generative machine-learned model’s internal parameters to generate realistic or accurate content (e.g., grammatically correct content, coherent content, etc.) based on the training data. The one or more generative machine-learned models 6140 may be trained on one or more training datasets including reference imagery, reference audio, reference videos, reference text documents, etc. The one or more training datasets may include values for the one or more conditions.

[0120] In some implementations, the meaning drift detection application 6100 (e.g., the one or more large language models and / or the one or more generative machine-learned models 6140) are configured to generate an output which includes a user interface output 6500. For example, if the computing device (e.g., the meaning drift detection application 6100 which includes the one or more machine-learned models) determines the intended (global) meaning has drifted (changed, deviated), based on the prompt from operation 2100 provided from input 6200 and the edits received at operation 2200 from input 6200 (or stored in the document edits data store 350), the meaning drift detection application 6100 may be configured to provide, forpresentation to the user, a notification (e.g., feedback) and / or a recommendation to revert or revise the second version of the document, through the user interface.

[0121] In some implementations, the meaning drift detection application 6100 (e.g., the one or more large language models and / or the one or more generative machine-learned models 6140) are configured to generate an output which includes a drift corrected version of the document 6600. For example, if the computing device (e.g., the meaning drift detection application 6100 which includes the one or more machine-learned models) determines the intended (global) meaning has drifted (changed, deviated), based on the prompt from operation 2100 provided from input 6200 and the edits received at operation 2200 from input 6200 (or stored in the document edits data store 350), the meaning drift detection application 6100 may be configured to automatically revert or revise the second version of the document to generate the drift corrected version of the document 6600, so that the intended (global) meaning of the document is maintained with respect to the first version of the document and / or with respect to the intent provided in the prompt.

[0122] In some implementations, the automatic corrections or revisions, or the recommendations, which are generated by the meaning drift detection application 6100, may be based on the conditioning parameters (and corresponding values for the one or more conditions) to make decisions for generating content. For example, the meaning drift detection application 6100 may recommend or automatically change a term or feature to be consistent with the intended global meaning of the document (e.g., changing an English accent of a character to an Irish accent to be consistent with an intended meaning as provided in the prompt).

[0123] In some implementations, after a drift correction version of the document 6600 is generated and / or after a recommendation or feedback is provided via user interface output 6500, the user can provide feedback or a further input relating to the output, and such feedback or further input can be used to refine the one or more machine-learned models. In some implementations, the one or more machine-learned models may be updated or refined based on feedback from the user relating to whether the meaning drift detection application 6100 has correctly identified whether or not the intended (global) meaning of the second version of the document has deviated based on the prompt from operation 2100 and the edits received at operation 2200.

[0124] An example operation of the meaning drift detection application 6100 is described herein. For example, the meaning drift detection application 6100 may receive an input (prompt) which includes information associated an intent or purpose (meaning) associated with the document. In some implementations, the document may have been created (generated) by the one or more machine-learned models (e.g., the one or more generative machine-learned models 6140) with a particular high-level prompt, which can be retrieved by the meaning drift detection application 6100. If the document was not initially generated via a prompt, then the meaning drift detection application 6100 may be configured to determine a prompt which would have made the one or more machine-learned models generate the document.

[0125] In some implementations, the meaning drift detection application 6100 may be configured to provide an output to confirm with the user whether the user agrees with global meaning as ascertained by the meaning drift detection application 6100.

[0126] Subsequently, the user may make edits or revisions to the document (e.g., to improve the form of the document, to add context, to satisfy user preferences, to provide expert knowledge, etc.). The meaning drift detection application 6100 may be configured to implement one or more machine-learned models as described herein (e.g., with respect to operation 2300) at a particular endpoint. As described herein, the endpoint may be defined based on various criteria (e.g., after a certain number of edits, after a pause since the last edit of a particular duration of time, etc.). The meaning drift detection application 6100 may be configured to batch together the edits that have been made up until the endpoint and then implement the one or more machine-learned models to check for consistency of the entire document, together with the previously determined or identified prompt.

[0127] In some implementations, the one or more machine-learned models of the meaning drift detection application 6100 may be configured to be implemented with respect to only a portion of the document, thereby conserving computing resources such as processor cycles. For example, the portion of the document may correspond to only those portions of the document which have been edited. For example, the portion of the document may correspond to one or more particular sections of the document (e.g., a background section, a summary section, a conclusion section, etc.) which have been edited, and thus the one or more machine-learned models may be configured to be implemented on a section-by-section basis.

[0128] In some implementations, the one or more machine-learned models of the meaning drift detection application 6100 may be configured to be implemented in response to particular types of edits being made. For example, the one or more machine-learned models may be configured to identify a type of edit being made to the document. For example, in response to a first type of edit being made to the document, the one or more machine-learned models may be implemented to determine whether the updated document remains consistent with the original intent or intended meaning of the document. For example, the first type of edit may be a substantive edit which the one or more machine-learned models determines (with some confidence threshold level being satisfied) is likely to have affected the content, structure, and meaning of the document. Such substantive edits may include reorganizing sections or paragraphs, the addition of new information, the removal of information, a modification to the tone or style of the document, etc. For example, in response to a second type of edit being made to the document, the one or more machine-learned models may not be implemented to determine whether the updated document remains consistent with the original intent or intended meaning of the document. For example, the second type of edit may be a non-substantive edit which the one or more machine-learned models determines (with some confidence threshold level being satisfied) is not likely to have affected the content, structure, and meaning of the document. Such non-substantive edits may include correcting typographical or grammatical errors (e.g., spelling, punctuation, etc.), changes in formatting (e.g., changing font size), etc. Therefore, computing resources (e.g., processor cycles) may be conserved by selectively implementing the one or more machine-learned models based on the types of edits to the document.

[0129] Edits to the document may be made manually by the user or edits may be made by implementation of one or more machine-learned models according to an input (prompt) provided by the user (e.g., “change this fragment such that the past tense is used” or “in this fragment, the main character wears a blue t-shirt”). The prompts for editing the document may be stored together with the provided edits (regardless of the mechanism) and batched together up until the specified endpoint when the one or more machine-learned models are implemented to check for consistency of the entire document.

[0130] The one or more machine-learned models of the meaning drift detection application 6100 may be configured to process the edits to the document (e.g., manual or raw edits, prompts for editing the document, etc.).

[0131] In some implementations, a plurality of users may be performing a combination of edits to the document. For example, when different users simultaneously (or disjointedly) edit the document conflicting changes may be made to completely different parts of the content. In some implementations, the one or more machine-learned models of the meaning drift detection application 6100 may be configured to determine whether an overall intent or goal with respect to the document is maintained in response to the edits made to the document by the plurality of users. In some implementations, the one or more machine-learned models may be configured to be implemented (for determining whether a deviation in the meaning of the document has occurred) in response to a determination that the plurality of users have edited the document and after a particular number of changes (edits) have been made by the plurality of users. In some implementations, the one or more machine-learned models may be configured to be implemented in response to a determination that the plurality of users have edited the document and after a pause (e.g., a predetermined duration of time) since a last edit (change) has occurred by any one particular user and before a next edit. In some implementations, the one or more machine- learned models may be configured to perform batch processing with respect to the changes made by the plurality of users, rather than be implemented with respect to respective changes by individual users. In some implementations, the one or more machine-learned models may be configured to be implemented in response to a determination that the plurality of users have edited the document and in the case of the edits including changes to code in a computing program that the code is capable of being compiled. Therefore, computing resources (e.g., processor cycles) may be conserved by selectively implementing the one or more machine- learned models according to particular criteria being satisfied (e.g., a certain number of edits, a certain delay after a last edit, a condition that code in a document which has been edited be capable of being compiled, etc.).

[0132] In some implementations, the computing device (e.g., the meaning drift detection application 6100 having the one or more machine-learned models) may be configured to freeze or suspend a capability for users to edit or change the document to allow for sufficient time for the meaning drift detection application 6100 (e.g., the one or more machine-learned models) to determine whether an overall intent or goal with respect to the document is maintained in response to the edits made to the document by the plurality of users.

[0133] In some implementations, the computing device (e.g., the meaning drift detection application 6100 having the one or more machine-learned models) may be configured to provide, for presentation to the users, a user interface having control knobs (e.g., user interface elements) which are configured to enable the plurality of users to revert or resolve inconsistencies in the document (e.g., through a chat interface whereby the users can reach a consensus on the changes). The inputs to the user interface may be provided to the one or more machine-learned models to reevaluate whether an overall intent or goal with respect to the document is maintained in response to the edits made to the document by the plurality of users.

[0134] In some instances, the meaning drift detection application 6100 may be configured to determine whether an overall intent or goal with respect to the document is maintained based on the edits to the document, the prompt to the document, and information associated with the plurality of users. For example, the information associated with the plurality of users can include a defined role for respective users among the plurality of users editing the document. For example, a user may be associated with a certain role in a content management system that is used to edit the document (e.g., the user may be an editor that is responsible for revising the style of the document, the user may be an expert with respect to a particular topic of the document, etc.). For example, the information associated with a user can indicate an intent or goal of the document that can be used to determine whether a meaning drift has occurred.

[0135] In some implementations, the meaning drift detection application 6100 may be configured to determine whether an overall intent or goal with respect to a particular section of a document is maintained. Thus, the one or more machine-learned models of the meaning drift detection application 6100 may be implemented on a secti on-by-section basis.

[0136] In some implementations, the user may provide an input (e.g., updated prompt) altering the intended (global) meaning of the document. Based on the updated meaning of the document, the meaning drift detection application 6100 may be configured to determine whether an overall intent or goal with respect to the document is maintained based on the edits to the document.

[0137] FIG. 7 depicts a flowchart of a method 700 for training one or more machine- learned models according to aspects of the disclosure. For instance, an example machine-learned model can include one or more of a LLM, a generative machine-learned model, etc. For example, the one or more machine-learned models may be configured to implement theoperations of FIGS. 2 through 5, of the meaning drift detection application 6100, etc., as described herein.

[0138] FIG. 7 is a flow chart diagram illustrating an example method for training a machine-learned model according to example implementations of aspects of the disclosure. One or more portion(s) of example method 700 can be implemented by a computing system that includes one or more computing devices such as, for example, computing systems described with reference to the other drawings. Each respective portion of example method 700 can be performed by any (or any combination) of one or more computing devices. Moreover, one or more portion(s) of example method 700 can be implemented on the hardware components of the device(s) described herein, for example, to train one or more systems or models. FIG. 7 depicts elements performed in a particular order for purposes of illustration and discussion. Those of ordinary skill in the art, using the disclosures provided herein, will understand that the elements of any of the methods discussed herein can be adapted, rearranged, expanded, omitted, combined, or modified in various ways without deviating from the scope of the disclosure. FIG. 7 is described with reference to elements / terms described with respect to other systems and drawings for exemplary illustrated purposes and is not meant to be limiting. One or more portions of example method 700 can be performed additionally, or alternatively, by other systems.

[0139] At 702, example method 700 can include obtaining a training instance. A set of training data can include a plurality of training instances divided between multiple datasets (e.g., a training dataset, a validation dataset, or testing dataset). A training instance can be labeled or unlabeled. Although referred to in example method 700 as a “training” instance, it is to be understood that runtime inferences can form training instances when a model is trained using an evaluation of the model’s performance on that runtime instance (e.g., online training / leaming). Example data types for the training instance and various tasks associated therewith are described throughout the disclosure.

[0140] At 704, example method 700 can include processing, using one or more machine- learned models, the training instance to generate an output. The output can be directly obtained from the one or more machine-learned models or can be a downstream result of a chain of processing operations that includes an output of the one or more machine-learned models.

[0141] At 706, example method 700 can include receiving an evaluation signal associated with the output. The evaluation signal can be obtained using a loss function. Various determinations of loss can be used, such as mean squared error, likelihood loss, cross entropy loss, hinge loss, contrastive loss, or various other loss functions. The evaluation signal can be computed using known ground-truth labels (e.g., supervised learning), predicted or estimated labels (e.g., semi- or self-supervised learning), or without labels (e.g., unsupervised learning). The evaluation signal can be a reward (e.g., for reinforcement learning). The reward can be computed using a machine-learned reward model configured to generate rewards based on output(s) received. The reward can be computed using feedback data describing human feedback on the output(s).

[0142] At 708, example method 700 can include updating the machine-learned model using the evaluation signal. For example, values for parameters of the machine-learned model(s) can be learned, in some embodiments, using various training or learning techniques, such as, for example, backwards propagation. For example, the evaluation signal can be backpropagated from the output (or another source of the evaluation signal) through the machine-learned model(s) to update one or more parameters of the model(s) (e.g., based on a gradient of the evaluation signal with respect to the parameter value(s)). For example, system(s) containing one or more machine-learned models can be trained in an end-to-end manner. Gradient descent techniques can be used to iteratively update the parameters over a number of training iterations. In some implementations, performing backwards propagation of errors can include performing truncated backpropagation through time. Example method 700 can include implementing a number of generalization techniques (e.g., weight decays, dropouts, etc.) to improve the generalization capability of the models being trained.

[0143] In some implementations, example method 700 can be implemented for training a machine-learned model from an initialized state to a fully trained state (e.g., when the model exhibits a desired performance profile, such as based on accuracy, precision, recall, etc.).

[0144] In some implementations, example method 700 can be implemented for particular stages of a training procedure. For instance, in some implementations, example method 700 can be implemented for pre-training a machine-learned model. Pre-training can include, for instance, large-scale training over potentially noisy data to achieve a broad base of performance levels across a variety of tasks / data types. In some implementations, example method 700 can beimplemented for fine-tuning a machine-learned model. Fine-tuning can include, for instance, smaller-scale training on higher-quality (e.g., labeled, curated, etc.) data. Fine-tuning can affect all or a portion of the parameters of a machine-learned model. For example, various portions of the machine-learned model can be “frozen” for certain training stages. For example, parameters associated with an embedding space can be “frozen” during fine-tuning (e.g., to retain information learned from a broader domain(s) than present in the fine-tuning dataset(s)). An example fine-tuning approach includes reinforcement learning. Reinforcement learning can be based on user feedback on model performance during use.Example Machine-Learned Models

[0145] FIG. 8 is a block diagram of an example processing flow for using machine- learned model(s) 1 to process input(s) 2 to generate output(s) 3.

[0146] Machine-learned model(s) 1 can be or include one or multiple machine-learned models or model components. Example machine-learned models can include neural networks (e.g., deep neural networks). Example machine-learned models can include non-linear models or linear models. Example machine-learned models can use other architectures in lieu of or in addition to neural networks. Example machine-learned models can include decision tree based models, support vector machines, hidden Markov models, Bayesian networks, linear regression models, k-means clustering models, etc.

[0147] Example neural networks can include feed-forward neural networks, recurrent neural networks (RNNs), including long short-term memory (LSTM) based recurrent neural networks, convolutional neural networks (CNNs), diffusion models, generative-adversarial networks, or other forms of neural networks. Example neural networks can be deep neural networks. Some example machine-learned models can leverage an attention mechanism such as self-attention. For example, some example machine-learned models can include multi-headed self-attention models.

[0148] Machine-learned model(s) 1 can include a single or multiple instances of the same model configured to operate on data from input(s) 2. Machine-learned model(s) 1 can include an ensemble of different models that can cooperatively interact to process data from input(s) 2. For example, machine-learned model(s) 1 can employ a mixture-of-experts structure. See, e.g., Zhou et al., Mixture-of-Experts with Expert Choice Routing, ARXIV:2202.09368V2 (Oct. 14, 2022).

[0149] Input(s) 2 can generally include or otherwise represent various types of data. Input(s) 2 can include one type or many different types of data. Output(s) 3 can be data of the same type(s) or of different types of data as compared to input(s) 2. Output(s) 3 can include one type or many different types of data.

[0150] Example data types for input(s) 2 or output(s) 3 include natural language text data, software code data (e.g., source code, object code, machine code, or any other form of computer-readable instructions or programming languages), machine code data (e.g., binary code, assembly code, or other forms of machine-readable instructions that can be executed directly by a computer's central processing unit), assembly code data (e.g., low-level programming languages that use symbolic representations of machine code instructions to program a processing unit), genetic data or other chemical or biochemical data, image data, audio data, audiovisual data, haptic data, biometric data, medical data, financial data, statistical data, geographical data, astronomical data, historical data, sensor data generally (e.g., digital or analog values, such as voltage or other absolute or relative level measurement values from a real or artificial input, such as from an audio sensor, light sensor, displacement sensor, etc.), and the like. Data can be raw or processed and can be in any format or schema.

[0151] In multimodal inputs 2 or outputs 3, example combinations of data types include image data and audio data, image data and natural language data, natural language data and software code data, image data and biometric data, sensor data and medical data, etc. It is to be understood that any combination of data types in an input 2 or an output 3 can be present.

[0152] An example input 2 can include one or multiple data types, such as the example data types noted above. An example output 3 can include one or multiple data types, such as the example data types noted above. The data type(s) of input 2 can be the same as or different from the data type(s) of output 3. It is to be understood that the example data types noted above are provided for illustrative purposes only. Data types contemplated within the scope of the disclosure are not limited to those examples noted above.Example Machine-Learned Sequence Processing Models

[0153] FIG. 9 is a block diagram of an example implementation of an example machine- learned model configured to process sequences of information. For instance, an example implementation of machine-learned model(s) 1 can include machine-learned sequence processing model(s) 4. An example system can pass input(s) 2 to sequence processing model(s)4. Sequence processing model(s) 4 can include one or more machine-learned components. Sequence processing model(s) 4 can process the data from input(s) 2 to obtain an input sequence5. Input sequence 5 can include one or more input elements 5-1, 5-2, . . . , 5-A , etc. obtained from input(s) 2. Sequence processing model 4 can process input sequence 5 using prediction layer(s) 6 to generate an output sequence 7. Output sequence 7 can include one or more output elements 7-1, 7-2, . . . , 7-N, etc. generated based on input sequence 5. The system can generate output(s) 3 based on output sequence 7.

[0154] Sequence processing model(s) 4 can include one or multiple machine-learned model components configured to ingest, generate, or otherwise reason over sequences of information. For example, some example sequence processing models in the text domain are referred to as “Large Language Models,” or LLMs. See, e.g., PaLM 2 Technical Report, GOOGLE, https: / / ai.google / static / documents / palm2techreport.pdf (n.d.). Other example sequence processing models can operate in other domains, such as image domains, see, e.g., Dosovitskiy et al., An Image is Worth 16x16 Words: Transformer s for Image Recognition at Scale, ARXIV:2010.11929v2 (Jun. 3, 2021), audio domains, see, e.g., Agostinelli et a\., MusicLM: Generating Music From Text, ARXIV:2301.11325V1 (Jan. 26, 2023), biochemical domains, see, e.g., Jumper et al., Highly accurate protein structure prediction with AlphaFold, 596 Nature 583 (Aug. 26, 2021), by way of example. Sequence processing model(s) 4 can process one or multiple types of data simultaneously. Sequence processing model(s) 4 can include relatively large models (e.g., more parameters, computationally expensive, etc.), relatively small models (e.g., fewer parameters, computationally lightweight, etc ), or both.

[0155] In general, sequence processing model(s) 4 can obtain input sequence 5 using data from input(s) 2. For instance, input sequence 5 can include a representation of data from input(s) 2 in a format understood by sequence processing model(s) 4. One or more machine-learned components of sequence processing model(s) 4 can ingest the data from input(s) 2, parse the data into pieces compatible with the processing architectures of sequence processing model(s) 4 (e.g., via “tokenization”), and project the pieces into an input space associated with prediction layer(s) 6 (e.g., via “embedding”).

[0156] Sequence processing model(s) 4 can ingest the data from input(s) 2 and parse the data into a sequence of elements to obtain input sequence 5. For example, a portion of input datafrom input(s) 2 can be broken down into pieces that collectively represent the content of the portion of the input data. The pieces can provide the elements of the sequence.

[0157] Elements 5-1, 5-2, . . . , 5-M can represent, in some cases, building blocks for capturing or expressing meaningful information in a particular data domain. For instance, the elements can describe “atomic units” across one or more domains. For example, for textual input source(s), the elements can correspond to groups of one or more words or sub-word components, such as sets of one or more characters.

[0158] For example, elements 5-1, 5-2, . . . , 5-M can represent tokens obtained using a tokenizer. For instance, a tokenizer can process a given portion of an input source and output a series of tokens (e.g., corresponding to input elements 5-1, 5-2, . . . , 5-M) that represent the portion of the input source. Various approaches to tokenization can be used. For instance, textual input source(s) can be tokenized using a byte-pair encoding (BPE) technique. See, e.g., Kudo et al., SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processing, PROCEEDINGS OF THE 2018 CONFERENCE ON EMPIRICAL METHODS IN NATU AL LA GUAGE PROCESSING (System Demonstrations), pages 66-71 (October 31-November 4, 2018), https: / / aclanthology.org / D18-2012.pdf. Image-based input source(s) can be tokenized by extracting and serializing patches from an image.

[0159] In general, arbitrary data types can be serialized and processed into input sequence 5. It is to be understood that element(s) 5-1, 5-2, . . . , 5-M depicted in FIG. 9 can be the tokens or can be the embedded representations thereof.

[0160] Prediction layer(s) 6 can predict one or more output elements 7-1, 7-2, . . . , 7-N based on the input elements. Prediction layer(s) 6 can include one or more machine-learned model architectures, such as one or more layers of learned parameters that manipulate and transform the input(s) to extract higher-order meaning from, and relationships between, input element(s) 5-1, 5-2, . . . , 5-M. In this manner, for instance, example prediction layer(s) 6 can predict new output element(s) in view of the context provided by input sequence 5.

[0161] Prediction layer(s) 6 can evaluate associations between portions of input sequence5 and a particular output element. These associations can inform a prediction of the likelihood that a particular output follows the input context. For example, consider the textual snippet, “The carpenter’s toolbox was small and heavy. It was full of .” Example prediction layer(s)6 can identify that “It” refers back to “toolbox” by determining a relationship between therespective embeddings. Example prediction layer(s) 6 can also link “It” to the attributes of the toolbox, such as “small” and “heavy.” Based on these associations, prediction layer(s) 6 can, for instance, assign a higher probability to the word “nails” than to the word “sawdust.”

[0162] A transformer is an example architecture that can be used in prediction layer(s) 4. See, e.g., Vaswani et al., Attention Is All You Need, ARXIV: 1706.03762v7 (Aug. 2, 2023). A transformer is an example of a machine-learned model architecture that uses an attention mechanism to compute associations between items within a context window. The context window can include a sequence that contains input sequence 5 and potentially one or more output element(s) 7-1, 7-2, . . . , 7-N. A transformer block can include one or more attention layer(s) and one or more post-attention layer(s) (e.g., feedforward layer(s), such as a multi-layer perceptron).

[0163] Prediction layer(s) 6 can include other machine-learned model architectures in addition to or in lieu of transformer-based architectures. For example, recurrent neural networks (RNNs) and long short-term memory (LSTM) models can also be used, as well as convolutional neural networks (CNNs). In general, prediction layer(s) 6 can leverage various kinds of artificial neural networks that can understand or generate sequences of information.

[0164] Output sequence 7 can include or otherwise represent the same or different data types as input sequence 5. For instance, input sequence 5 can represent textual data, and output sequence 7 can represent textual data. Input sequence 5 can represent image, audio, or audiovisual data, and output sequence 7 can represent textual data (e g., describing the image, audio, or audiovisual data). It is to be understood that prediction layer(s) 6, and any other interstitial model components of sequence processing model(s) 4, can be configured to receive a variety of data types in input sequence(s) 5 and output a variety of data types in output sequence(s) 7.

[0165] Output sequence 7 can have various relationships to input sequence 5. Output sequence 7 can be a continuation of input sequence 5. Output sequence 7 can be complementary to input sequence 5. Output sequence 7 can translate, transform, augment, or otherwise modify input sequence 5. Output sequence 7 can answer, evaluate, confirm, or otherwise respond to input sequence 5. Output sequence 7 can implement (or describe instructions for implementing) an instruction provided via input sequence 5.

[0166] Output sequence 7 can be generated autoregressively. For instance, for some applications, an output of one or more prediction layer(s) 6 can be passed through one or more output layers (e.g., softmax layer) to obtain a probability distribution over an output vocabulary (e.g., a textual or symbolic vocabulary) conditioned on a set of input elements in a context window. In this manner, for instance, output sequence 7 can be autoregressively generated by sampling a likely next output element, adding that element to the context window, and regenerating the probability distribution based on the updated context window, and sampling a likely next output element, and so forth.

[0167] Output sequence 7 can also be generated non-autoregressively. For instance, multiple output elements of output sequence 7 can be predicted together without explicit sequential conditioning on each other. See, e.g., Saharia et al., Non- Autoregressive Machine Translation with Latent Alignments, ARXIV:2004.07437V3 (NOV. 16, 2020).

[0168] Output sequence 7 can include one or multiple portions or elements. In an example content generation configuration, output sequence 7 can include multiple elements corresponding to multiple portions of a generated output sequence (e.g., a textual sentence, values of a discretized waveform, computer code, etc.). In an example classification configuration, output sequence 7 can include a single element associated with a classification output. For instance, an output “vocabulary” can include a set of classes into which an input sequence is to be classified. For instance, a vision transformer block can pass latent state information to a multilayer perceptron that outputs a likely class value associated with an input image.

[0169] FIG. 10 is a block diagram of an example technique for populating an example input sequence 8. Input sequence 8 can include various functional elements that form part of the model infrastructure, such as an element 8-0 obtained from a task indicator 9 that signals to any model(s) that process input sequence 8 that a particular task is being performed (e.g., to help adapt a performance of the model(s) to that particular task). Input sequence 8 can include various data elements from different data modalities. For instance, an input modality 10-1 can include one modality of data. A data-to-sequence model 11-1 can process data from input modality 10-1 to project the data into a format compatible with input sequence 8 (e.g., one or more vectors dimensioned according to the dimensions of input sequence 8) to obtain elements 8-1, 8-2, 8-3. Another input modality 10-2 can include a different modality of data. A data-to-sequence model 11-2 can project data from input modality 10-2 into a format compatible with input sequence 8 to obtain elements 8-4, 8-5, 8-6. Another input modality 10-3 can include yet another different modality of data. A data-to-sequence model 11-3 can project data from input modality 10-3 into a format compatible with input sequence 8 to obtain elements 8-7, 8-8, 8-9.

[0170] Input sequence 8 can be the same as or different from input sequence 5. Input sequence 8 can be a multimodal input sequence that contains elements that represent data from different modalities using a common dimensional representation. For instance, an embedding space can have P dimensions. Input sequence 8 can be configured to contain a plurality of elements that have P dimensions. In this manner, for instance, example implementations can facilitate information extraction and reasoning across diverse data modalities by projecting data into elements in the same embedding space for comparison, combination, or other computations therebetween.

[0171] For example, elements 8-0, . . . , 8-9 can indicate particular locations within a multidimensional embedding space. Some elements can map to a set of discrete locations in the embedding space. For instance, elements that correspond to discrete members of a predetermined vocabulary of tokens can map to discrete locations in the embedding space that are associated with those tokens. Other elements can be continuously distributed across the embedding space. For instance, some data types can be broken down into continuously defined portions (e.g., image patches) that can be described using continuously distributed locations within the embedding space.

[0172] In some implementations, the expressive power of the embedding space may not be limited to meanings associated with any particular set of tokens or other building blocks. For example, a continuous embedding space can encode a spectrum of high-order information. An individual piece of information (e.g., a token) can map to a particular point in that space: for instance, a token for the word “dog” can be projected to an embedded value that points to a particular location in the embedding space associated with canine-related information. Similarly, an image patch of an image of a dog on grass can also be projected into the embedding space. In some implementations, the projection of the image of the dog can be similar to the projection of the word “dog” while also having similarity to a projection of the word “grass,” while potentially being different from both. In some implementations, the projection of the image patch may not exactly align with any single projection of a single word. In some implementations, theprojection of the image patch can align with a combination of the projections of the words “dog” and “grass.” In this manner, for instance, a high-order embedding space can encode information that can be independent of data modalities in which the information is expressed.

[0173] Task indicator 9 can include a model or model component configured to identify a task being performed and inject, into input sequence 8, an input value represented by element 8-0 that signals which task is being performed. For instance, the input value can be provided as a data type associated with an input modality and projected along with that input modality (e.g., the input value can be a textual task label that is embedded along with other textual data in the input; the input value can be a pixel-based representation of a task that is embedded along with other image data in the input; etc.). The input value can be provided as a data type that differs from or is at least independent from other input(s). For instance, the input value represented by element 8-0 can be a learned within a continuous embedding space.

[0174] Input modalities 10-1, 10-2, and 10-3 can be associated with various different data types (e.g., as described above with respect to input(s) 2 and output(s) 3).

[0175] Data-to-sequence models 11-1, 11-2, and 11-3 can be the same or different from each other. Data-to-sequence models 11-1, 11-2, and 11-3 can be adapted to each respective input modality 10-1, 10-2, and 10-3. For example, a textual data-to-sequence model can subdivide a portion of input text and project the subdivisions into element(s) in input sequence 8 (e.g., elements 8-1, 8-2, 8-3, etc.). An image data-to-sequence model can subdivide an input image and project the subdivisions into element(s) in input sequence 8 (e.g., elements 8-4, 8-5, 8-6, etc.). An arbitrary datatype data-to-sequence model can subdivide an input of that arbitrary datatype and project the subdivisions into element(s) in input sequence 8 (e.g., elements 8-7, 8-8, 8-9, etc.).

[0176] Data-to-sequence models 11-1, 11-2, and 11-3 can form part of machine-learned sequence processing model(s) 4. Data-to-sequence models 11-1, 11-2, and 11-3 can be jointly trained with or trained independently from machine-learned sequence processing model(s) 4. Data-to-sequence models 11-1, 11-2, and 11-3 can be trained end-to-end with machine-learned sequence processing model(s) 4.Example Machine-Learned Model Development Platform

[0177] FIG. 11 is a block diagram of an example model development platform 12 that can facilitate creation, adaptation, and refinement of example machine-learned models (e.g.,machine-learned model(s) 1, sequence processing model(s) 4, etc.). Model development platform 12 can provide a number of different toolkits that developer systems can employ in the development of new or adapted machine-learned models.

[0178] Model development platform 12 can provide one or more model libraries 13 containing building blocks for new models. Model libraries 13 can include one or more pretrained foundational models 13-1, which can provide a backbone of processing power across various tasks. Model libraries 13 can include one or more pre-trained expert models 13-2, which can be focused on performance in particular domains of expertise. Model libraries 13 can include various model primitives 13-3, which can provide low-level architectures or components (optionally pre-trained), which can be assembled in various arrangements as desired.

[0179] Model development platform 12 can receive selections of various model components 14. Model development platform 12 can pass selected model components 14 to a workbench 15 that combines selected model components 14 into a development model 16.

[0180] Workbench 15 can facilitate further refinement and adaptation of development model 16 by leveraging a number of different toolkits integrated with model development platform 12. For example, workbench 15 can facilitate alignment of the development model 16 with a desired performance profile on various tasks using a model alignment toolkit 17.

[0181] Model alignment toolkit 17 can provide a number of tools for causing development model 16 to generate outputs aligned with desired behavioral characteristics. Alignment can include increasing an accuracy, precision, recall, etc. of model outputs. Alignment can include enforcing output styles, schema, or other preferential characteristics of model outputs. Alignment can be general or domain-specific. For instance, a pre-trained foundational model 13-1 can begin with an initial level of performance across multiple domains. Alignment of the pre-trained foundational model 13-1 can include improving a performance in a particular domain of information or tasks (e.g., even at the expense of performance in another domain of information or tasks).

[0182] Model alignment toolkit 17 can integrate one or more dataset(s) 17-1 for aligning development model 16. Curated dataset(s) 17-1 can include labeled or unlabeled training data. Dataset(s) 17-1 can be obtained from public domain datasets. Dataset(s) 17-1 can be obtained from private datasets associated with one or more developer system(s) for the alignment of bespoke machine-learned model(s) customized for private use-cases.

[0183] Pre-training pipelines 17-2 can include a machine-learned model training workflow configured to update development model 16 over large-scale, potentially noisy datasets. For example, pre-training can leverage unsupervised learning techniques (e.g., denoising, etc.) to process large numbers of training instances to update model parameters from an initialized state and achieve a desired baseline performance. Pre-training pipelines 17-2 can leverage unlabeled datasets in dataset(s) 17-1 to perform pre-training. Workbench 15 can implement a pre-training pipeline 17-2 to pre-train development model 16.

[0184] Fine-tuning pipelines 17-3 can include a machine-learned model training workflow configured to refine the model parameters of development model 16 with higher- quality data. Fine-tuning pipelines 17-3 can update development model 16 by conducting supervised training with labeled dataset(s) in dataset(s) 17-1. Fine-tuning pipelines 17-3 can update development model 16 by conducting reinforcement learning using reward signals from user feedback signals. Workbench 15 can implement a fine-tuning pipeline 17-3 to fine-tune development model 16.

[0185] Prompt libraries 17-4 can include sets of inputs configured to induce behavior aligned with desired performance criteria. Prompt libraries 17-4 can include few-shot prompts (e.g., inputs providing examples of desired model outputs for prepending to a desired runtime query), chain-of-thought prompts (e.g., inputs providing step-by-step reasoning within the exemplars to facilitate thorough reasoning by the model), and the like.

[0186] Example prompts can be retrieved from an available repository of prompt libraries 17-4. Example prompts can be contributed by one or more developer systems using workbench 15.

[0187] In some implementations, pre-trained or fine-tuned models can achieve satisfactory performance without exemplars in the inputs. For instance, zero-shot prompts can include inputs that lack exemplars. Zero-shot prompts can be within a domain within a training dataset or outside of the training domain(s).

[0188] Prompt libraries 17-4 can include one or more prompt engineering tools. Prompt engineering tools can provide workflows for retrieving or learning optimized prompt values. Prompt engineering tools can facilitate directly learning prompt values (e.g., input element values) based one or more training iterations. Workbench 15 can implement prompt engineering tools in development model 16.

[0189] Prompt libraries 17-4 can include pipelines for prompt generation. For example, inputs can be generated using development model 16 itself or other machine-learned models. In this manner, for instance, a first model can process information about a task and output an input for a second model to process in order to perform a step of the task. The second model can be the same as or different from the first model. Workbench 15 can implement prompt generation pipelines in development model 16.

[0190] Prompt libraries 17-4 can include pipelines for context injection. For instance, a performance of development model 16 on a particular task can improve if provided with additional context for performing the task. Prompt libraries 17-4 can include software components configured to identify desired context, retrieve the context from an external source (e.g., a database, a sensor, etc.), and add the context to the input prompt. Workbench 15 can implement context injection pipelines in development model 16.

[0191] Although various training examples described herein with respect to model development platform 12 refer to “pre-training” and “fine-tuning,” it is to be understood that model alignment toolkit 17 can generally support a wide variety of training techniques adapted for training a wide variety of machine-learned models. Example training techniques can correspond to the example training method 700 described above.

[0192] Model development platform 12 can include a model plugin toolkit 18. Model plugin toolkit 18 can include a variety of tools configured for augmenting the functionality of a machine-learned model by integrating the machine-learned model with other systems, devices, and software components. For instance, a machine-learned model can use tools to increase performance quality where appropriate. For instance, deterministic tasks can be offloaded to dedicated tools in lieu of probabilistically performing the task with an increased risk of error. For instance, instead of autoregressively predicting the solution to a system of equations, a machine-learned model can recognize a tool to call for obtaining the solution and pass the system of equations to the appropriate tool. The tool can be a traditional system of equations solver that can operate deterministically to resolve the system of equations. The output of the tool can be returned in response to the original query. In this manner, tool use can allow some example models to focus on the strengths of machine-learned models — e.g., understanding an intent in an unstructured request for a task — while augmenting the performance of the model by offloadingcertain tasks to a more focused tool for rote application of deterministic algorithms to a well- defined problem.

[0193] Model plugin toolkit 18 can include validation tools 18-1. Validation tools 18-1 can include tools that can parse and confirm output(s) of a machine-learned model. Validation tools 18-1 can include engineered heuristics that establish certain thresholds applied to model outputs. For example, validation tools 18-1 can ground the outputs of machine-learned models to structured data sources (e.g., to mitigate “hallucinations”).

[0194] Model plugin toolkit 18 can include tooling packages 18-2 for implementing one or more tools that can include scripts or other executable code that can be executed alongside development model 16. Tooling packages 18-2 can include one or more inputs configured to cause machine-learned model(s) to implement the tools (e.g., few-shot prompts that induce a model to output tool calls in the proper syntax, etc ). Tooling packages 18-2 can include, for instance, fine-tuning training data for training a model to use a tool.

[0195] Model plugin toolkit 18 can include interfaces for calling external application programming interfaces (APIs) 18-3. For instance, in addition to or in lieu of implementing tool calls or tool code directly with development model 16, development model 16 can be aligned to output instruction that initiate API calls to send or obtain data via external systems.

[0196] Model plugin toolkit 18 can integrate with prompt libraries 17-4 to build a catalog of available tools for use with development model 16. For instance, a model can receive, in an input, a catalog of available tools, and the model can generate an output that selects a tool from the available tools and initiates a tool call for using the tool.

[0197] Model development platform 12 can include a computational optimization toolkit 19 for optimizing a computational performance of development model 16. For instance, tools for model compression 19-1 can allow development model 16 to be reduced in size while maintaining a desired level of performance. For instance, model compression 19-1 can include quantization workflows, weight pruning and sparsification techniques, etc. Tools for hardware acceleration 19-2 can facilitate the configuration of the model storage and execution formats to operate optimally on different hardware resources. For instance, hardware acceleration 19-2 can include tools for optimally sharding models for distributed processing over multiple processing units for increased bandwidth, lower unified memory requirements, etc. Tools for distillation 19- 3 can provide for the training of lighter- weight models based on the knowledge encoded indevelopment model 16. For instance, development model 16 can be a highly performant, large machine-learned model optimized using model development platform 12. To obtain a lightweight model for running in resource-constrained environments, a smaller model can be a “student model” that learns to imitate development model 16 as a “teacher model.” In this manner, for instance, the investment in learning the parameters and configurations of development model 16 can be efficiently transferred to a smaller model for more efficient inference.

[0198] Workbench 15 can implement one, multiple, or none of the toolkits implemented in model development platform 12. Workbench 15 can output an output model 20 based on development model 16. Output model 20 can be a deployment version of development model 16. Output model 20 can be a development or training checkpoint of development model 16. Output model 20 can be a distilled, compressed, or otherwise optimized version of development model 16.

[0199] FIG. 12 is a block diagram of an example training flow for training a machine- learned development model 16. One or more portion(s) of the example training flow can be implemented by a computing system that includes one or more computing devices such as, for example, computing systems described with reference to the other drawings. Each respective portion of the example training flow can be performed by any (or any combination) of one or more computing devices. Moreover, one or more portion(s) of the example training flow can be implemented on the hardware components of the device(s) described herein, for example, to train one or more systems or models. FIG. 12 depicts elements performed in a particular order for purposes of illustration and discussion. Those of ordinary skill in the art, using the disclosures provided herein, will understand that the elements of any of the methods discussed herein can be adapted, rearranged, expanded, omitted, combined, or modified in various ways without deviating from the scope of the disclosure. FIG. 12 is described with reference to elements / terms described with respect to other systems and drawings for exemplary illustrated purposes and is not meant to be limiting. One or more portions of the example training flow can be performed additionally, or alternatively, by other systems.

[0200] Initially, development model 16 can persist in an initial state as an initialized model 21. Development model 16 can be initialized with weight values. Initial weight valuescan be random or based on an initialization schema. Initial weight values can be based on prior pre-training for the same or for a different model.

[0201] Initialized model 21 can undergo pre-training in a pre-training stage 22. Pretraining stage 22 can be implemented using one or more pre-training pipelines 17-2 over data from dataset(s) 17-1. Pre-training can be omitted, for example, if initialized model 21 is already pre-trained (e.g., development model 16 contains, is, or is based on a pre-trained foundational model or an expert model).

[0202] Pre-trained model 23 can then be a new version of development model 16, which can persist as development model 16 or as a new development model. Pre-trained model 23 can be the initial state if development model 16 was already pre-trained. Pre-trained model 23 can undergo fine-tuning in a fine-tuning stage 24. Fine-tuning stage 24 can be implemented using one or more fine-tuning pipelines 17-3 over data from dataset(s) 17-1. Fine-tuning can be omitted, for example, if a pre-trained model as satisfactory performance, if the model was already fine-tuned, or if other tuning approaches are preferred.

[0203] Fine-tuned model 29 can then be a new version of development model 16, which can persist as development model 16 or as a new development model. Fine-tuned model 29 can be the initial state if development model 16 was already fine-tuned. Fine-tuned model 29 can undergo refinement with user feedback 26. For instance, refinement with user feedback 26 can include reinforcement learning, optionally based on human feedback from human users of finetuned model 25. As reinforcement learning can be a form of fine-tuning, it is to be understood that fine-tuning stage 24 can subsume the stage for refining with user feedback 26. Refinement with user feedback 26 can produce a refined model 27. Refined model 27 can be output to downstream system(s) 28 for deployment or further development.

[0204] In some implementations, computational optimization operations can be applied before, during, or after each stage. For instance, initialized model 21 can undergo computational optimization 29-1 (e.g., using computational optimization toolkit 19) before pre-training stage 22. Pre-trained model 23 can undergo computational optimization 29-2 (e.g., using computational optimization toolkit 19) before fine-tuning stage 24. Fine-tuned model 25 can undergo computational optimization 29-3 (e.g., using computational optimization toolkit 19) before refinement with user feedback 26. Refined model 27 can undergo computational optimization 29-4 (e.g., using computational optimization toolkit 19) before output todownstream system(s) 28. Computational optimization(s) 29-1, . . . , 29-4 can all be the same, all be different, or include at least some different optimization techniques.Example Machine-Learned Model Inference System

[0205] FIG. 13 is a block diagram of an inference system for operating one or more machine-learned model(s) 1 to perform inference (e.g., for training, for deployment, etc.). A model host 31 can receive machine-learned model(s) 1. Model host 31 can host one or more model instance(s) 31-1, which can be one or multiple instances of one or multiple models. Model host 31 can host model instance(s) 31-1 using available compute resources 31-2 associated with model host 31.

[0206] Model host 31 can perform inference on behalf of one or more client(s) 32. Client(s) 32 can transmit an input request 33 to model host 31. Using input request 33, model host 31 can obtain input(s) 2 for input to machine-learned model(s) 1. Machine-learned model(s) 1 can process input(s) 2 to generate output(s) 3. Using output(s) 3, model host 31 can return an output payload 34 for responding to input request 33 from client(s) 32. Output payload 34 can include or be based on output(s) 3.

[0207] Model host 31 can leverage various other resources and tools to augment the inference task. For instance, model host 31 can communicate with tool interfaces 35 to facilitate tool use by model instance(s) 31-1. Tool interfaces 35 can include local or remote APIs. Tool interfaces 35 can include integrated scripts or other software functionality. Model host 31 can engage online learning interface(s) 36 to facilitate ongoing improvements to machine-learned model(s) 1. For instance, online learning interface(s) 36 can be used within reinforcement learning loops to retrieve user feedback on inferences served by model host 31. Model host 31 can access runtime data source(s) 37 for augmenting input(s) 2 with additional contextual information. For instance, runtime data source(s) 37 can include a knowledge graph 37-1 that facilitates structured information retrieval for information associated with input request(s) 33 (e.g., a search engine service). Runtime data source(s) 37 can include public or private, external or local database(s) 37-2 that can store information associated with input request(s) 33 for augmenting input(s) 2. Runtime data source(s) 37 can include account data 37-3 which can be retrieved in association with a user account corresponding to a client 32 for customizing the behavior of model host 31 accordingly.

[0208] Model host 31 can be implemented by one or multiple computing devices or systems. Client(s) 2 can be implemented by one or multiple computing devices or systems, which can include computing devices or systems shared with model host 31.

[0209] For example, model host 31 can operate on a server system that provides a machine-learning service to client device(s) that operate client(s) 32 (e.g., over a local or wide- area network). Client device(s) can be end-user devices used by individuals. Client device(s) can be server systems that operate client(s) 32 to provide various functionality as a service to downstream end-user devices.

[0210] In some implementations, model host 31 can operate on a same device or system as client(s) 32. Model host 31 can be a machine-learning service that runs on-device to provide machine-learning functionality to one or multiple applications operating on a client device, which can include an application implementing client(s) 32. Model host 31 can be a part of a same application as client(s) 32. For instance, model host 31 can be a subroutine or method implemented by one part of an application, and client(s) 32 can be another subroutine or method that engages model host 31 to perform inference functions within the application. It is to be understood that model host 31 and client(s) 32 can have various different configurations.

[0211] Model instance(s) 31-1 can include one or more machine-learned models that are available for performing inference. Model instance(s) 31-1 can include weights or other model components that are stored on in persistent storage, temporarily cached, or loaded into highspeed memory. Model instance(s) 31-1 can include multiple instance(s) of the same model (e.g., for parallel execution of more requests on the same model). Model instance(s) 31-1 can include instance(s) of different model(s). Model instance(s) 31-1 can include cached intermediate states of active or inactive model(s) used to accelerate inference of those models. For instance, an inference session with a particular model may generate significant amounts of computational results that can be re-used for future inference runs (e.g., using a KV cache for transformer-based models). These computational results can be saved in association with that inference session so that session can be executed more efficiently when resumed.

[0212] Compute resource(s) 31-2 can include one or more processors (central processing units, graphical processing units, tensor processing units, machine-learning accelerators, etc.) connected to one or more memory devices. Compute resource(s) 31-2 can include a dynamic pool of available resources shared with other processes. Compute resource(s) 31-2 can includememory devices large enough to fit an entire model instance in a single memory instance. Compute resource(s) 31-2 can also shard model instance(s) across multiple memory devices (e.g., using data parallelization or tensor parallelization, etc.). This can be done to increase parallelization or to execute a large model using multiple memory devices which individually might not be able to fit the entire model into memory.

[0213] Input request 33 can include data for input(s) 2. Model host 31 can process input request 33 to obtain input(s) 2. Input(s) 2 can be obtained directly from input request 33 or can be retrieved using input request 33. Input request 33 can be submitted to model host 31 via an API.

[0214] Model host 31 can perform inference over batches of input requests 33 in parallel. For instance, a model instance 31-1 can be configured with an input structure that has a batch dimension. Separate input(s) 2 can be distributed across the batch dimension (e.g., rows of an array). The separate input(s) 2 can include completely different contexts. The separate input(s)2 can be multiple inference steps of the same task. The separate input(s) 2 can be staggered in an input structure, such that any given inference cycle can be operating on different portions of the respective input(s) 2. In this manner, for instance, model host 31 can perform inference on the batch in parallel, such that output(s) 3 can also contain the batch dimension and return the inference results for the batched input(s) 2 in parallel. In this manner, for instance, batches of input request(s) 33 can be processed in parallel for higher throughput of output payload(s) 34.

[0215] Output payload 34 can include or be based on output(s) 3 from machine-learned model(s) 1. Model host 31 can process output(s) 3 to obtain output payload 34. This can include chaining multiple rounds of inference (e.g., iteratively, recursively, across the same model(s) or different model(s)) to arrive at a final output for a task to be returned in output payload 34. Output payload 34 can be transmitted to client(s) 32 via an API.

[0216] Online learning interface(s) 36 can facilitate reinforcement learning of machine- learned model(s) 1. Online learning interface(s) 36 can facilitate reinforcement learning with human feedback (RLHF). Online learning interface(s) 36 can facilitate federated learning of machine-learned model(s) 1.

[0217] Model host 31 can execute machine-learned model(s) 1 to perform inference for various tasks using various types of data. For example, various different input(s) 2 and output(s)3 can be used for various different tasks. In some implementations, input(s) 2 can be orotherwise represent image data. Machine-learned model(s) 1 can process the image data to generate an output. As an example, machine-learned model(s) 1 can process the image data to generate an image recognition output (e.g., a recognition of the image data, a latent embedding of the image data, an encoded representation of the image data, a hash of the image data, etc.). As another example, machine-learned model(s) 1 can process the image data to generate an image segmentation output. As another example, machine-learned model(s) 1 can process the image data to generate an image classification output. As another example, machine-learned model(s) 1 can process the image data to generate an image data modification output (e.g., an alteration of the image data, etc.). As another example, machine-learned model(s) 1 can process the image data to generate an encoded image data output (e.g., an encoded and / or compressed representation of the image data, etc.). As another example, machine-learned model(s) 1 can process the image data to generate an upscaled image data output. As another example, machine-learned model(s) 1 can process the image data to generate a prediction output.

[0218] In some implementations, the task is a computer vision task. In some cases, input(s) 2 includes pixel data for one or more images and the task is an image processing task. For example, the image processing task can be image classification, where the output is a set of scores, each score corresponding to a different object class and representing the likelihood that the one or more images depict an object belonging to the object class. The image processing task may be object detection, where the image processing output identifies one or more regions in the one or more images and, for each region, a likelihood that region depicts an object of interest. As another example, the image processing task can be image segmentation, where the image processing output defines, for each pixel in the one or more images, a respective likelihood for each category in a predetermined set of categories. For example, the set of categories can be foreground and background. As another example, the set of categories can be object classes. As another example, the image processing task can be depth estimation, where the image processing output defines, for each pixel in the one or more images, a respective depth value. As another example, the image processing task can be motion estimation, where the network input includes multiple images, and the image processing output defines, for each pixel of one of the input images, a motion of the scene depicted at the pixel between the images in the network input.

[0219] In some implementations, input(s) 2 can be or otherwise represent natural language data. Machine-learned model(s) 1 can process the natural language data to generate anoutput. As an example, machine-learned model(s) 1 can process the natural language data to generate a language encoding output. As another example, machine-learned model(s) 1 can process the natural language data to generate a latent text embedding output. As another example, machine-learned model(s) 1 can process the natural language data to generate a translation output. As another example, machine-learned model(s) 1 can process the natural language data to generate a classification output. As another example, machine-learned model(s) 1 can process the natural language data to generate a textual segmentation output. As another example, machine-learned model(s) 1 can process the natural language data to generate a semantic intent output. As another example, machine-learned model(s) 1 can process the natural language data to generate an upscaled text or natural language output (e.g., text or natural language data that is higher quality than the input text or natural language, etc.). As another example, machine-learned model(s) 1 can process the natural language data to generate a prediction output (e.g., one or more predicted next portions of natural language content).

[0220] In some implementations, input(s) 2 can be or otherwise represent speech data (e.g., data describing spoken natural language, such as audio data, textual data, etc.). Machine- learned model(s) 1 can process the speech data to generate an output. As an example, machine- learned model(s) 1 can process the speech data to generate a speech recognition output. As another example, machine-learned model(s) 1 can process the speech data to generate a speech translation output. As another example, machine-learned model(s) 1 can process the speech data to generate a latent embedding output. As another example, machine-learned model(s) 1 can process the speech data to generate an encoded speech output (e.g., an encoded and / or compressed representation of the speech data, etc.). As another example, machine-learned model(s) 1 can process the speech data to generate an upscaled speech output (e.g., speech data that is higher quality than the input speech data, etc.). As another example, machine-learned model(s) 1 can process the speech data to generate a textual representation output (e.g., a textual representation of the input speech data, etc.). As another example, machine-learned model(s) 1 can process the speech data to generate a prediction output.

[0221] In some implementations, input(s) 2 can be or otherwise represent latent encoding data (e.g., a latent space representation of an input, etc.). Machine-learned model(s) 1 can process the latent encoding data to generate an output. As an example, machine-learned model(s) 1 can process the latent encoding data to generate a recognition output. As anotherexample, machine-learned model(s) 1 can process the latent encoding data to generate a reconstruction output. As another example, machine-learned model(s) 1 can process the latent encoding data to generate a search output. As another example, machine-learned model(s) 1 can process the latent encoding data to generate a reclustering output. As another example, machine- learned model(s) 1 can process the latent encoding data to generate a prediction output.

[0222] In some implementations, input(s) 2 can be or otherwise represent statistical data. Statistical data can be, represent, or otherwise include data computed and / or calculated from some other data source. Machine-learned model(s) 1 can process the statistical data to generate an output. As an example, machine-learned model(s) 1 can process the statistical data to generate a recognition output. As another example, machine-learned model(s) 1 can process the statistical data to generate a prediction output. As another example, machine-learned model(s) 1 can process the statistical data to generate a classification output. As another example, machine- learned model(s) 1 can process the statistical data to generate a segmentation output. As another example, machine-learned model(s) 1 can process the statistical data to generate a visualization output. As another example, machine-learned model(s) 1 can process the statistical data to generate a diagnostic output.

[0223] In some implementations, input(s) 2 can be or otherwise represent sensor data. Machine-learned model(s) 1 can process the sensor data to generate an output. As an example, machine-learned model(s) 1 can process the sensor data to generate a recognition output. As another example, machine-learned model(s) 1 can process the sensor data to generate a prediction output. As another example, machine-learned model(s) 1 can process the sensor data to generate a classification output. As another example, machine-learned model(s) 1 can process the sensor data to generate a segmentation output. As another example, machine-learned model(s) 1 can process the sensor data to generate a visualization output. As another example, machine-learned model(s) 1 can process the sensor data to generate a diagnostic output. As another example, machine-learned model(s) 1 can process the sensor data to generate a detection output.

[0224] In some implementations, machine-learned model(s) 1 can be configured to perform a task that includes encoding input data for reliable and / or efficient transmission or storage (and / or corresponding decoding). For example, the task may be an audio compression task. The input may include audio data and the output may comprise compressed audio data. Inanother example, the input includes visual data (e.g., one or more images or videos), the output comprises compressed visual data, and the task is a visual data compression task. In another example, the task may comprise generating an embedding for input data (e.g., input audio or visual data). In some cases, the input includes audio data representing a spoken utterance and the task is a speech recognition task. The output may comprise a text output which is mapped to the spoken utterance. In some cases, the task comprises encrypting or decrypting input data. In some cases, the task comprises a microprocessor performance task, such as branch prediction or memory address translation.

[0225] In some implementations, the task is a generative task, and machine-learned model(s) 1 can be configured to output content generated in view of input(s) 2. For instance, input(s) 2 can be or otherwise represent data of one or more modalities that encodes context for generating additional content.

[0226] In some implementations, the task can be a text completion task. Machine- learned model(s) 1 can be configured to process input(s) 2 that represent textual data and to generate output(s) 3 that represent additional textual data that completes a textual sequence that includes input(s) 2. For instance, machine-learned model(s) 1 can be configured to generate output(s) 3 to complete a sentence, paragraph, or portion of text that follows from a portion of text represented by input(s) 2.

[0227] In some implementations, the task can be an instruction following task. Machine- learned model(s) 1 can be configured to process input(s) 2 that represent instructions to perform a function and to generate output(s) 3 that advance a goal of satisfying the instruction function (e.g., at least a step of a multi-step procedure to perform the function). Output(s) 3 can represent data of the same or of a different modality as input(s) 2. For instance, input(s) 2 can represent textual data (e.g., natural language instructions for a task to be performed) and machine-learned model(s) 1 can process input(s) 2 to generate output(s) 3 that represent textual data responsive to the instructions (e.g., natural language responses, programming language responses, machine language responses, etc.). Input(s) 2 can represent image data (e g., image-based instructions for a task to be performed, optionally accompanied by textual instructions) and machine-learned model(s) 1 can process input(s) 2 to generate output(s) 3 that represent textual data responsive to the instructions (e.g., natural language responses, programming language responses, machine language responses, etc.). One or more output(s) 3 can be iteratively or recursively generated tosequentially process and accomplish steps toward accomplishing the requested functionality. For instance, an initial output can be executed by an external system or be processed by machine- learned model(s) 1 to complete an initial step of performing a function. Multiple steps can be performed, with a final output being obtained that is responsive to the initial instructions.

[0228] In some implementations, the task can be a question answering task. Machine- learned model(s) 1 can be configured to process input(s) 2 that represent a question to answer and to generate output(s) 3 that advance a goal of returning an answer to the question (e.g., at least a step of a multi-step procedure to perform the function). Output(s) 3 can represent data of the same or of a different modality as input(s) 2. For instance, input(s) 2 can represent textual data (e.g., natural language instructions for a task to be performed) and machine-learned model(s) 1 can process input(s) 2 to generate output(s) 3 that represent textual data responsive to the question (e.g., natural language responses, programming language responses, machine language responses, etc.). Input(s) 2 can represent image data (e g., image-based instructions for a task to be performed, optionally accompanied by textual instructions) and machine-learned model(s) 1 can process input(s) 2 to generate output(s) 3 that represent textual data responsive to the question (e g., natural language responses, programming language responses, machine language responses, etc.). One or more output(s) 3 can be iteratively or recursively generated to sequentially process and accomplish steps toward answering the question. For instance, an initial output can be executed by an external system or be processed by machine-learned model(s) 1 to complete an initial step of obtaining an answer to the question (e.g., querying a database, performing a computation, executing a script, etc.). Multiple steps can be performed, with a final output being obtained that is responsive to the question.

[0229] In some implementations, the task can be an image generation task. Machine- learned model(s) 1 can be configured to process input(s) 2 that represent context regarding a desired portion of image content. The context can include text data, image data, audio data, etc. Machine-learned model(s) 1 can be configured to generate output(s) 3 that represent image data that depicts imagery related to the context. For instance, machine-learned model(s) 1 can be configured to generate pixel data of an image. Values for channel(s) associated with the pixels in the pixel data can be selected based on the context (e.g., based on a probability determined based on the context).

[0230] In some implementations, the task can be an audio generation task. Machine- learned model(s) 1 can be configured to process input(s) 2 that represent context regarding a desired portion of audio content. The context can include text data, image data, audio data, etc. Machine-learned model(s) 1 can be configured to generate output(s) 3 that represent audio data related to the context. For instance, machine-learned model(s) 1 can be configured to generate waveform data in the form of an image (e.g., a spectrogram). Values for channel(s) associated with pixels of the image can be selected based on the context. Machine-learned model(s) 1 can be configured to generate waveform data in the form of a sequence of discrete samples of a continuous waveform. Values of the sequence can be selected based on the context (e.g., based on a probability determined based on the context).

[0231] In some implementations, the task can be a data generation task. Machine-learned model(s) 1 can be configured to process input(s) 2 that represent context regarding a desired portion of data (e.g., data from various data domains, such as sensor data, image data, multimodal data, statistical data, etc.). The desired data can be, for instance, synthetic data for training other machine-learned models. The context can include arbitrary data type(s). Machine-learned model(s) 1 can be configured to generate output(s) 3 that represent data that aligns with the desired data. For instance, machine-learned model(s) 1 can be configured to generate data values for populating a dataset. Values for the data object(s) can be selected based on the context (e.g., based on a probability determined based on the context).Example Computing Systems and Devices

[0232] FIG. 14 is a block diagram of an example networked computing system that can perform aspects of example implementations of the disclosure. The system can include a number of computing devices and systems that are communicatively coupled over a network 49. An example computing device 50 is described to provide an example of a computing device that can perform any aspect of the disclosure (e.g., implementing model host 31, client(s) 32, or both). An example server computing system 60 is described as an example of a server computing system that can perform any aspect of the disclosure (e.g., implementing model host 31, client(s) 32, or both). Computing device 50 and server computing system(s) 60 can cooperatively interact (e.g., over network 49) to perform any aspect of the disclosure (e.g., implementing model host 31, client(s) 32, or both). Model development platform system 70 is an example system that can host or serve model development platform(s) 12 for development ofmachine-learned models. Third-party system(s) 80 are example system(s) with which any of computing device 50, server computing system(s) 60, or model development platform system(s) 70 can interact in the performance of various aspects of the disclosure (e.g., engaging third-party tools, accessing third-party databases or other resources, etc.).

[0233] Network 49 can be any type of communications network, such as a local area network (e.g., intranet), wide area network (e.g., Internet), or some combination thereof and can include any number of wired or wireless links. In general, communication over network 49 can be carried via any type of wired or wireless connection, using a wide variety of communication protocols (e.g., TCP / IP, HTTP, SMTP, FTP), encodings or formats (e.g., HTML, XML), or protection schemes (e.g., VPN, secure HTTP, SSL). Network 49 can also be implemented via a system bus. For instance, one or more devices or systems of FIG. 14 can be co-located with, contained by, or otherwise integrated into one or more other devices or systems.

[0234] Computing device 50 can be any type of computing device, such as, for example, a personal computing device (e.g., laptop or desktop), a mobile computing device (e.g., smartphone or tablet), a gaming console or controller, a wearable computing device, an embedded computing device, a server computing device, a virtual machine operating on a host device, or any other type of computing device. Computing device 50 can be a client computing device. Computing device 50 can be an end-user computing device. Computing device 50 can be a computing device of a service provided that provides a service to an end user (who may use another computing device to interact with computing device 50).

[0235] Computing device 50 can include one or more processors 51 and a memory 52. Processor(s) 51 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. Memory 52 can include one or more non-transitory computer-readable storage media, such as HBM, RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. Memory 52 can store data 53 and instructions 54 which can be executed by processor(s) 51 to cause computing device 50 to perform operations. The operations can implement any one or multiple features described herein. The operations can implement example methods and techniques described herein.

[0236] Computing device 50 can also include one or more input components that receive user input. For example, a user input component can be a touch-sensitive component (e.g., atouch -sensitive display screen or a touch pad) that is sensitive to the touch of a user input object (e.g., a finger or a stylus). The touch-sensitive component can serve to implement a virtual keyboard. Other example user input components include a microphone, camera, LIDAR, a physical keyboard or other buttons, or other means by which a user can provide user input.

[0237] Computing device 50 can store or include one or more machine-learned models 55. Machine-learned models 55 can include one or more machine-learned model(s) 1, such as a sequence processing model 4. Machine-learned models 55 can include one or multiple model instance(s) 31-1. Machine-learned model(s) 55 can be received from server computing system(s) 60, model development platform system 70, third party system(s) 80 (e.g., an application distribution platform), or developed locally on computing device 50. Machine-learned model(s) 55 can be loaded into memory 52 and used or otherwise implemented by processor(s) 51. Computing device 50 can implement multiple parallel instances of machine-learned model(s) 55.

[0238] Server computing system(s) 60 can include one or more processors 61 and a memory 62. Processor(s) 61 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. Memory 62 can include one or more non-transitory computer-readable storage media, such as HBM, RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. Memory 62 can store data 63 and instructions 64 which can be executed by processor(s) 61 to cause server computing system(s) 60 to perform operations. The operations can implement any one or multiple features described herein. The operations can implement example methods and techniques described herein.

[0239] In some implementations, server computing system 60 includes or is otherwise implemented by one or multiple server computing devices. In instances in which server computing system 60 includes multiple server computing devices, such server computing devices can operate according to sequential computing architectures, parallel computing architectures, or some combination thereof.

[0240] Server computing system 60 can store or otherwise include one or more machine- learned models 65. Machine-learned model(s) 65 can be the same as or different from machine- learned model(s) 55. Machine-learned models 65 can include one or more machine-learned model(s) 1, such as a sequence processing model 4. Machine-learned models 65 can include oneor multiple model instance(s) 31 -1. Machine-learned model(s) 65 can be received from computing device 50, model development platform system 70, third party system(s) 80, or developed locally on server computing system(s) 60. Machine-learned model(s) 65 can be loaded into memory 62 and used or otherwise implemented by processor(s) 61. Server computing system(s) 60 can implement multiple parallel instances of machine-learned model(s) 65.

[0241] In an example configuration, machine-learned models 65 can be included in or otherwise stored and implemented by server computing system 60 to establish a client-server relationship with computing device 50 for serving model inferences. For instance, server computing system(s) 60 can implement model host 31 on behalf of client(s) 32 on computing device 50. For instance, machine-learned models 65 can be implemented by server computing system 60 as a portion of a web service (e.g., remote machine-learned model hosting service, such as an online interface for performing machine-learned model operations over a network on server computing system(s) 60). For instance, server computing system(s) 60 can communicate with computing device 50 over a local intranet or internet connection. For instance, computing device 50 can be a workstation or endpoint in communication with server computing system(s) 60, with implementation of machine-learned models 65 being managed by server computing system(s) 60 to remotely perform inference (e.g., for runtime or training operations), with output(s) returned (e.g., cast, streamed, etc.) to computing device 50. Machine-learned models 65 can work cooperatively or interoperatively with machine-learned models 55 on computing device 50 to perform various tasks.

[0242] Model development platform system(s) 70 can include one or more processors 71 and a memory 72. Processor(s) 71 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. Memory 72 can include one or more non-transitory computer-readable storage media, such as HBM, RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. Memory 72 can store data 73 and instructions 74 which can be executed by processor(s) 71 to cause model development platform system(s) 70 to perform operations. The operations can implement any one or multiple features described herein. The operations can implement example methods and techniques described herein. Example operations include the functionalitydescribed herein with respect to model development platform 12. This and other functionality can be implemented by developer tool(s) 75.

[0243] Third-party system(s) 80 can include one or more processors 81 and a memory 82. Processor(s) 81 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. Memory 82 can include one or more non-transitory computer-readable storage media, such as HBM, RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. Memory 82 can store data 83 and instructions 84 which can be executed by processor(s) 81 to cause third-party system(s) 80 to perform operations. The operations can implement any one or multiple features described herein. The operations can implement example methods and techniques described herein. Example operations include the functionality described herein with respect to tools and other external resources called when training or performing inference with machine-learned model(s) 1, 4, 16, 20, 55, 65, etc. (e.g., third-party resource(s) 85).

[0244] FIG. 14 illustrates one example arrangement of computing systems that can be used to implement the disclosure. Other computing system configurations can be used as well. For example, in some implementations, one or both of computing system 50 or server computing system(s) 60 can implement all or a portion of the operations of model development platform system 70. For example, computing system 50 or server computing system(s) 60 can implement developer tool(s) 75 (or extensions thereof) to develop, update / train, or refine machine-learned models 1, 4, 16, 20, 55, 65, etc. using one or more techniques described herein with respect to model alignment toolkit 17. In this manner, for instance, computing system 50 or server computing system(s) 60 can develop, update / train, or refine machine-learned models based on local datasets (e.g., for model personalization / customization, as permitted by user data preference selections).

[0245] FIG. 15 is a block diagram of an example computing device 98 that performs according to example embodiments of the disclosure. Computing device 98 can be a user computing device or a server computing device (e.g., computing device 50, server computing system(s) 60, etc.). Computing device 98 can implement model host 31. For instance, computing device 98 can include a number of applications (e.g., applications 1 through N). Each application can contain its own machine learning library and machine-learned model(s). Forexample, each application can include a machine-learned model. Example applications include a text messaging application, an email application, a dictation application, a virtual keyboard application, a browser application, a social media application, a chat application, a meaning drift detection application, etc. As illustrated in FIG. 15, each application can communicate with a number of other components of the computing device, such as, for example, one or more sensors, a context manager, a device state component, or additional components. In some implementations, each application can communicate with each device component using an API (e.g., a public API). In some implementations, the API used by each application is specific to that application.

[0246] FIG. 16 is a block diagram of an example computing device 99 that performs according to example embodiments of the disclosure. Computing device 99 can be the same as or different from computing device 98. Computing device 99 can be a user computing device or a server computing device (e.g., computing device 50, server computing system(s) 60, etc.). Computing device 98 can implement model host 31. For instance, computing device 99 can include a number of applications (e.g., applications 1 through N). Each application can be in communication with a central intelligence layer. Example applications include a text messaging application, an email application, a dictation application, a virtual keyboard application, a browser application, a social media application, a chat application, a meaning drift detection application, etc. In some implementations, each application can communicate with the central intelligence layer (and model(s) stored therein) using an API (e g., a common API across all applications).

[0247] The central intelligence layer can include a number of machine-learned models. For example, as illustrated in FIG. 16, a respective machine-learned model can be provided for each application and managed by the central intelligence layer. In other implementations, two or more applications can share a single machine-learned model. For example, in some implementations, the central intelligence layer can provide a single model for all of the applications. In some implementations, the central intelligence layer is included within or otherwise implemented by an operating system of computing device 99.

[0248] The central intelligence layer can communicate with a central device data layer. The central device data layer can be a centralized repository of data for computing device 99. As illustrated in FIG. 16, the central device data layer can communicate with a number of othercomponents of the computing device, such as, for example, one or more sensors, a context manager, a device state component, or additional components. In some implementations, the central device data layer can communicate with each device component using an API (e.g., a private API).Additional Disclosure

[0249] The technology discussed herein makes reference to servers, databases, software applications, and other computer-based systems, as well as actions taken and information sent to and from such systems. The inherent flexibility of computer-based systems allows for a great variety of possible configurations, combinations, and divisions of tasks and functionality between and among components. For instance, processes discussed herein can be implemented using a single device or component or multiple devices or components working in combination. Databases and applications can be implemented on a single system or distributed across multiple systems. Distributed components can operate sequentially or in parallel.

[0250] Aspects of the disclosure have been described in terms of illustrative embodiments thereof. Any and all features in the following claims can be combined or rearranged in any way possible, including combinations of claims not explicitly enumerated in combination together, as the example claim dependencies listed herein should not be read as limiting the scope of possible combinations of features disclosed herein. Accordingly, the scope of the disclosure is by way of example rather than by way of limitation, and the subject disclosure does not preclude inclusion of such modifications, variations or additions to the disclosure as would be readily apparent to one of ordinary skill in the art. Moreover, terms are described herein using lists of example elements joined by conjunctions such as “and,” “or,” “but,” etc. It should be understood that such conjunctions are provided for explanatory purposes only. Clauses and other sequences of items joined by a particular conjunction such as “or,” for example, can refer to “and / or,” “at least one of’, “any combination of’ example elements listed therein, etc. Terms such as “based on” should be understood as “based at least in part on.”

[0251] Terms used herein are used to describe the example embodiments and are not intended to limit and / or restrict the disclosure. The singular forms “a,” “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. In this disclosure, terms such as "including", "having", “comprising”, and the like are used tospecify features, numbers, steps, operations, elements, components, or combinations thereof, but do not preclude the presence or addition of one or more of the features, elements, steps, operations, elements, components, or combinations thereof.

[0252] The term "and / or" includes a combination of a plurality of related listed items or any item of the plurality of related listed items. For example, the scope of the expression or phrase "A and / or B" includes the item "A", the item "B", and the combination of items "A and B”.

[0253] In addition, the scope of the expression or phrase "at least one of A or B" is intended to include all of the following: (1) at least one of A, (2) at least one of B, and (3) at least one of A and at least one of B. Likewise, the scope of the expression or phrase "at least one of A, B, or C" is intended to include all of the following: (1) at least one of A, (2) at least one of B, (3) at least one of C, (4) at least one of A and at least one of B, (5) at least one of A and at least one of C, (6) at least one of B and at least one of C, and (7) at least one of A, at least one of B, and at least one of C.

[0254] It will be understood that, although the terms first, second, third, etc., may be used herein to describe various elements, the elements are not limited by these terms. Instead, these terms are used to distinguish one element from another element. For example, without departing from the scope of the disclosure, a first element may be termed as a second element, and a second element may be termed as a first element.

[0255] The term “can” should be understood as referring to a possibility of a feature in various implementations and not as prescribing an ability that is necessarily present in every implementation. For example, the phrase “X can perform Y” should be understood as indicating that, in various implementations, X has the potential to be configured to perform Y, and not as indicating that in every instance X must always be able to perform Y. It should be understood that, in various implementations, X might be unable to perform Y and remain within the scope of the disclosure.

[0256] The term “may” should be understood as referring to a possibility of a feature in various implementations and not as prescribing an ability that is necessarily present in every implementation. For example, the phrase “X may perform Y” should be understood as indicating that, in various implementations, X has the potential to be configured to perform Y, and not as indicating that in every instance X must always be able to perform Y. It should beunderstood that, in various implementations, X might be unable to perform Y and remain within the scope of the disclosure.

[0257] To the extent terms including "module", and "unit," and the like are used herein, these terms may refer to, but are not limited to, a software or hardware component or device, such as a Field Programmable Gate Array (FPGA) or Application Specific Integrated Circuit (ASIC), which performs certain tasks. A module or unit may be configured to reside on an addressable storage medium and configured to execute on one or more processors. Thus, a module or unit may include, by way of example, components, such as software components, object-oriented software components, class components and task components, processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays, and variables. The functionality provided for in the components and modules / units may be combined into fewer components and modules / units or further separated into additional components and modules.

[0258] Aspects of the above-described example embodiments may be recorded in non- transitory computer-readable media including program instructions to implement various operations embodied by a computer. The media may also include, alone or in combination with the program instructions, data files, data structures, and the like. Examples of non-transitory computer-readable media include magnetic media such as hard disks, floppy disks, and magnetic tape; optical media such as CD ROM disks, Blu-Ray disks, and DVDs; magneto-optical media such as optical discs; and other hardware devices that are specially configured to store and perform program instructions, such as semiconductor memory, read-only memory (ROM), random access memory (RAM), flash memory, USB memory, and the like. Examples of program instructions include both machine code, such as produced by a compiler, and files containing higher level code that may be executed by the computer using an interpreter. The program instructions may be executed by one or more processors. The described hardware devices may be configured to act as one or more software modules in order to perform the operations of the above-described embodiments, or vice versa. In addition, a non-transitory computer-readable storage medium may be distributed among computer systems connected through a network and computer-readable codes or program instructions may be stored and executed in a decentralized manner. In addition, the non-transitory computer-readable storagemedia may also be embodied in at least one application specific integrated circuit (ASIC) or Field Programmable Gate Array (FPGA).

[0259] Each block of the flowchart illustrations may represent a unit, module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the blocks may occur out of order. For example, two blocks shown in succession may in fact be executed substantially concurrently (simultaneously) or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved.

[0260] While the disclosure has been described with respect to various example embodiments, each example is provided by way of explanation, not limitation of the disclosure. Those skilled in the art, upon attaining an understanding of the foregoing, can readily produce alterations to, variations of, and equivalents to such embodiments. Accordingly, the disclosure does not preclude inclusion of such modifications, variations and / or additions to the disclosed subject matter as would be readily apparent to one of ordinary skill in the art. For example, features illustrated or described as part of one embodiment can be used with another embodiment to yield a still further embodiment. Thus, it is intended that the disclosure covers such alterations, variations, and equivalents.

Claims

WHAT IS CLAIMED IS:

1. A computing device, comprising: one or more memories configured to store instructions; and one or more processors configured to execute the instructions to perform operations, the operations comprising: obtaining a prompt indicating an intended meaning of a document, receiving a plurality of inputs editing a first version of the document to produce a second version of the document, determining whether one or more specified editing criteria associated with editing the document are satisfied, in response to the one or more specified editing criteria being satisfied, implementing one or more machine-learned models to determine whether specified drift criteria associated with the intended meaning of the document is satisfied, based on the prompt, the edits to the first version of the document, and the second version of the document, and providing an output based on whether the specified drift criteria associated with the intended meaning of the document is satisfied.

2. The computing device of claim 1, wherein the one or more specified editing criteria includes a number of edits to the first version of the document exceeding a threshold number of edits.

3. The computing device of claim 1, wherein the one or more specified editing criteria includes a threshold duration of time elapsing since a last edit to the first version of the document.

4. The computing device of claim 1, wherein the document is associated with a computing program and includes code, and the one or more specified editing criteria includes edited code in the document being capable of being compiled.

5. The computing device of claim 1 , wherein the prompt is received from a user to generate, via the one or more machine-learned models, the document.

6. The computing device of claim 1, wherein the prompt is retrieved by the one or more machine-learned models by determining a prompt which would generate the document.

7. The computing device of claim 1, wherein the plurality of inputs editing the first version of the document are provided by a plurality of users, and the one or more machine-learned models are configured to determine whether the drift criteria associated with the intended meaning of the document is satisfied, based on the prompt, the edits to the first version of the document, the second version of the document, and information associated with the plurality of users.

8. The computing device of claim 7, wherein the information associated with the plurality of users includes a defined role for respective users among the plurality of users editing the document.

9. The computing device of claim 1, wherein the specified drift criteria associated with the intended meaning of the document is satisfied when a global meaning of the second version of the document deviates from the intended meaning of the document by more than a threshold drift level, and the specified drift criteria associated with the intended meaning of the document is not satisfied when the global meaning of the second version of the document deviates from the intended meaning of the document by less than the threshold drift level.

10. The computing device of claim 1, wherein when the specified drift criteria associated with the intended meaning of the document is satisfied, the output includes automatically reverting or revising the second version of the document.11 . The computing device of claim 1, wherein when the specified drift criteria associated with the intended meaning of the document is satisfied, the output includes providing, for presentation to a user via a user interface, at least one of a notification or a recommendation to revert or revise the second version of the document.

12. The computing device of claim 1, wherein the document includes at least one of an image, a video, or an audio recording, the plurality of inputs editing the first version of the document include editing characteristics to change a quality of the at least one of the image, the video, or the audio recording to produce the second version of the document, and the specified drift criteria associated with the intended meaning of the document is satisfied when a resulting quality of the second version of the document is less than a threshold quality level indicated by the prompt.

13. The computing device of claim 1, wherein the document includes a computing program comprising code, the plurality of inputs editing the first version of the document include editing the code to produce the second version of the document, and the specified drift criteria associated with the intended meaning of the document is satisfied when a latency associated with the second version of the document is greater than a threshold latency level indicated by the prompt.

14. The computing device of claim 1, wherein the plurality of inputs editing the first version of the document include a first input editing a first portion of the document and a second input editing a second portion of the document, and the specified drift criteria associated with the intended meaning of the document is satisfied when the first input editing the first portion of the document conflicts with the second input editing the second portion of the document.

15. A computer-implemented method, comprising: obtaining, by a computing system comprising one or more processors, a prompt indicating an intended meaning of a document; receiving, by the computing system, a plurality of inputs editing a first version of the document to produce a second version of the document; determining, by the computing system, whether one or more specified editing criteria associated with editing the document are satisfied; in response to the one or more specified editing criteria being satisfied, determining, via one or more machine-learned models, whether specified drift criteria associated with the intended meaning of the document is satisfied, based on the prompt, the edits to the first version of the document, and the second version of the document; and providing, by the computing system, an output based on whether the specified drift criteria associated with the intended meaning of the document is satisfied.

16. The computer-implemented method of claim 15, wherein the one or more specified editing criteria includes: a number of edits to the first version of the document exceeding a threshold number of edits, or a threshold duration of time elapsing since a last edit to the first version of the document.

17. The computer-implemented method of claim 15, wherein the plurality of inputs editing the first version of the document are provided by a plurality of users, and the one or more machine-learned models determine whether the drift criteria associated with the intended meaning of the document is satisfied, based on the prompt, the edits to the first version of the document, the second version of the document, and information associated with the plurality of users.

18. The computer-implemented method of claim 15, further comprising: determining, via the one or more machine-learned models, the specified drift criteria associated with the intended meaning of the document is satisfied when a global meaning of thesecond version of the document deviates from the intended meaning of the document by more than a threshold drift level.

19. The computer-implemented method of claim 15, further comprising: determining, via the one or more machine-learned models, the specified drift criteria associated with the intended meaning of the document is satisfied when a resulting quality of the second version of the document is less than a threshold quality level indicated by the prompt.

20. A non-transitory computer readable medium storing instructions which, when executed by a processor, cause the processor to perform operations for generating an output document, the operations comprising: obtaining a prompt indicating an intended meaning of a document; receiving a plurality of inputs editing a first version of the document to produce a second version of the document; determining whether one or more specified editing criteria associated with editing the document are satisfied; in response to the one or more specified editing criteria being satisfied, implementing one or more machine-learned models to determine whether specified drift criteria associated with the intended meaning of the document is satisfied, based on the prompt, the edits to the first version of the document, and the second version of the document; and providing an output based on whether the specified drift criteria associated with the intended meaning of the document is satisfied.

Images