Teaching machine-learned models to collaborate using language-based backpropagation

Abstract

Techniques for utilizing a plurality of machine-learned models to perform a task are presented herein. The system can process, using a first machine-learned model from the plurality of machine-learned models, the task to determine a first subtask. Additionally, the system can process, using a second machine-learned model from the plurality of machine-learned models, the first subtask to generate a first output. Moreover, the system can process, using the first machine-learned model, the first output to generate a first response associated with a task. Furthermore, the system can receive, from a user device, user feedback associated with the first response. Subsequently, the system can process, using the first machine-learned model, the user feedback to generate a first feedback for the second machine-learned model.

Description

TEACHING MACHINE-LEARNED MODELS TO COLLABORATE USINGLANGUAGE-BASED B ACKPROPAGATIONFIELD

[0001] The present disclosure relates generally to machine learning. More particularly, the present disclosure relates to systems and methods for teaching machine-learned models to collaborate using language-based backpropagation.BACKGROUND

[0002] Machine-learned models, such as Large language models (LLMs), have demonstrated state of the art performance across a wide range of tasks, such as writing, summarization, translation, coding, and much more. As such, these models are being deployed as central parts of a user’s workflow, such as a chatbot, coding assistant, or an email writing assistant.

[0003] In conventional systems, LLMs can be pre trained in a self or unsupervised manner on a large corpus of text and then fine-tuned using techniques such as reinforcement learning. Fine-tuning an LLM involves adjusting a pre-trained model on a smaller, task-specific dataset to improve its performance on a particular application. The LLM is fine-tuned on a smaller, more specific dataset related to the task it needs to perform (e.g., sentiment analysis, question answering, or summarization). The dataset for fine-tuning is usually labeled, with examples demonstrating the correct outputs for given inputs. Fine-tuning allows a general- purpose LLM to be adapted for specific tasks, making it more effective and efficient in those applications.SUMMARY

[0004] Aspects and advantages of embodiments of the present 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] One example aspect of the present disclosure is directed to an example system for utilizing a plurality of machine-learned models to perform a task. The system can include one or more processors. Additionally, the system can include one or more non-transitory computer-readable media that collectively store instructions that, when executed by the one or more processors, cause the computing system to perform operations. The operations can include processing, using a first machine-learned model from the plurality of machine-learnedmodels, the task to determine a first subtask. Additionally, the instructions can include processing, using a second machine-learned model from the plurality of machine-learned models, the first subtask to generate a first output. Moreover, the instructions can include processing, using the first machine-learned model, the first output to generate a first response associated with the task. Furthermore, the instructions can include receiving, from a user device, user feedback associated with the first response. Subsequently, the instructions can include processing, using the first machine-learned model, the user feedback to generate a first feedback for the second machine-learned model.

[0006] In some instances, the system can transmit the first feedback to the second machine-learned model. Additionally, the system can process, using the second-machine- leamed model, the first feedback to generate a second output. Moreover, the system can process, using the first machine-learned model, the second output to generate a second response associated with the task. Furthermore, the system can present the second response to the user device, the second response being modified based on the user feedback to the first response.

[0007] In some instances, the system can update a parameter of the second machine- learned model based on the first feedback.

[0008] In some instances, the system can process, using the first machine-learned model, the user feedback to determine that the user feedback is associated with the first subtask. Additionally, the system can transmit the first feedback to the second machine-learned model based on the determination that the user feedback is associated with the first subtask.

[0009] In some instances, the first feedback is a score.

[0010] In some instances, the first feedback is a prompt.

[0011] In some instances, the system can transmit the first subtask to the second machine-learned model.

[0012] In some instances, the plurality of machine-learned models can include a third machine-learned model. Additionally, the system can process, using the first machine-learned model, the task to determine a second subtask. Moreover, the system can process, using the third machine-learned model, the second subtask to generate a second output, wherein the first response is based on the first output and the second output. Furthermore, the system can process, using the first machine-learned model, the user feedback to generate a second feedback for the third machine-learned model.

[0013] In some instances, the first machine-learned model and the second machine- learned model can be stored in the one or more non-transitory computer-readable media of the computing system.

[0014] In some instances, the first machine-learned model can be stored in the user device and the second machine-learned model is stored in the computing system.

[0015] In some instances, the first machine-learned model can be stored in the computing system and the second machine-learned model is stored in another system.

[0016] In some instances, the system can obtain (e.g., receive), from the user device, the task.

[0017] In some instances, the system can cause a presentation of the first response on a graphical user interface of the user device.

[0018] In some instances, the user feedback is a user input on the graphical user interface.

[0019] In some instances, the first response is an audio presentation, and the user feedback is audio data.

[0020] Another example aspect of the present disclosure is directed to an example computer-implemented method for utilizing a plurality of machine-learned models to perform a task. The method can include processing, using a first machine-learned model from the plurality of machine-learned models, the task to determine a first subtask. Additionally, the method can include processing, using a second machine-learned model from the plurality of machine-learned models, the first subtask to generate a first output. Moreover, the method can include processing, using the first machine-learned model, the first output to generate a first response associated with a task. Furthermore, the method can include receiving, from a user device, user feedback associated with the first response. Subsequently, the instructions can include processing, using the first machine-learned model, the user feedback to generate a first feedback for the second machine-learned model.

[0021] Another example aspect of the present disclosure is directed to one or more non- transitory computer-readable media. The one or more non-transitory computer-readable media can store instructions that are executable by a computing system to perform example operations. The operations can include processing, using a first machine-learned model from the plurality of machine-learned models, the task to determine a first subtask. Additionally, the instructions can include processing, using a second machine-learned model from the plurality of machine-learned models, the first subtask to generate a first output. Moreover, the instructions can include processing, using the first machine-learned model, the first output to generate a first response associated with the task. Furthermore, the instructions can include receiving, from a user device, user feedback associated with the first response. Subsequently,the instructions can include processing, using the first machine-learned model, the user feedback to generate a first feedback for the second machine-learned model.

[0022] Other aspects of the present disclosure are directed to various systems, apparatuses, non-transitory computer-readable media, user interfaces, and electronic devices.

[0023] These and other features, aspects, and advantages of various embodiments of the present disclosure 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 example embodiments of the present disclosure and, together with the description, serve to explain the related principles.BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Detailed discussion of embodiments directed to one of ordinary skill in the art is set forth in the specification, which makes reference to the appended figures, in which:

[0025] Figure 1 depicts a block diagram of an example system for generating structured data entry to be stored in a knowledge graph according to example embodiments of the present disclosure.

[0026] Figure 2 depicts a block diagram of an example system for processing user input using a machine-learned model to generate an insight according to example embodiments of the present disclosure.

[0027] Figure 3 depicts a flow chart diagram of an example method for utilizing a plurality of machine-learned models to generate a response to a user query according to example embodiments of the present disclosure.

[0028] Figure 4 depicts a flow chart diagram of an example method for providing feedback to one or more machine-learned model based on a determination of a subtask according to example embodiments of the present disclosure

[0029] Figure 5A depicts a block diagram of an example computing system that performs insight summary generation according to example embodiments of the present disclosure.

[0030] Figure 5B depicts a block diagram of an example computing device that performs insight summary generation according to example embodiments of the present disclosure.

[0031] Figure 5C depicts a block diagram of an example computing device that performs insight summary generation according to example embodiments of the present disclosure.

[0032] Figure 6 depicts a flowchart of a method for training one or more machine- learned models according to aspects of the present disclosure.

[0033] Figure 7 is a block diagram of an example processing flow for using machine- learned model(s) to process input(s) to generate output(s).

[0034] Figure 8 is a block diagram of an example implementation of an example machine-learned model configured to process sequences of information.

[0035] Figure 9 is a block diagram of an example technique for populating an example input sequence.

[0036] Figure 10 is a block diagram of an example model development platform that can facilitate creation, adaptation, and refinement of example machine-learned models.

[0037] Figure 11 is a block diagram of an example training flow for training a machine- learned development model.

[0038] Figure 12 is a block diagram of an inference system for operating one or more machine-learned model(s) to perform inference.

[0039] Reference numerals that are repeated across plural figures are intended to identify the same features in various implementations.DETAILED DESCRIPTION

[0040] Generally, the present disclosure is directed to techniques for training multiple collaborating large language model (LLM) agents that can involve back propagating errors in responses to the collaborating LLM agents. In some instances, the system can utilize a plurality of available LLMs with different specializations. For example, the plurality of LLMs that can perform a wide range of tasks, such as writing, summarization, translation, coding, and be deployed as central parts of a user’s workflow (e.g., a chatbot, coding assistant, an email writing assistant). Additionally, the plurality of available LLMs can act on behalf of a user, serve as a subject matter expert, or offer a particular service. In order to accomplish a task, the system can utilize the plurality of available LLMs to work together. For example, a first LLM associated with a first user can work together with a second LLM associated with a second user to collaboratively organize an event that takes into account both the first user and second user’s preferences. Alternatively, an orchestrator LLM might need to coordinate with other specialized LLMs to access knowledge or perform actions (e.g., booking different parts of a trip).

[0041] In conventional systems, LLMs can be aligned using techniques such as reinforcement learning. The alignment step can be important in order to make the LLM act inline with human preferences. Current alignment techniques assume that a single model is being aligned. Given the significant potential of multi-LLM collaboration, the system described herein trains multiple models to work together by applying alignment techniques when multiple models contribute to a single output.

[0042] According to some embodiments, the system teaches LLMs to work together more effectively by using alignment techniques, such as supervised fine-tuning (SFT) and / or reinforcement learning based on human feedback (RLHF). In conventional systems, alignment techniques such as SFT or RLHF may only work for the case of a single LLM because the system does not take into account which LLM in an ensemble contributed which part of the answer. Existing techniques are unable to handle the situation in which multiple models work together to generate a single output (e.g., one LLM querying another and then editing the answer), which can be the case when a cascade or ensemble of LLM models are used.

[0043] In some instances, the system can utilize backpropagation techniques to train different types of neural network. The backpropagation techniques can include computing a loss relative to a target and an adjustment is applied to all nodes in the network based on the partial derivative of the loss with respect to the weight in question.

[0044] Systems and methods of the present disclosure can provide a variety of technical effects and benefits, such as improved accuracy of machine-learned outputs; reduced computational cost (e.g., electricity cost, processor usage) of machine-learned language generation; and reduced cost (e.g., computational cost, labor cost) of generating an output.

[0045] For example, systems and methods according to example aspects of the present disclosure can provide improved accuracy of machine-learned outputs by enabling machine- learned models to be trained to work together. Using techniques described herein, the system can train the plurality of different models in an efficient manner, without having to train each model individually, which may also lead to worse performance in practice. As an example, some alternative methods may be configured to provide machine-generated outputs to a user without a mechanism to evaluate the outputs’ accuracy or to filter out inaccurate outputs. Advantageously, systems and methods according to the present disclosure can reduce a rate of factual errors in outputs generated by a machine-learned generative language model itself, and in outputs provided to the user by a computing system comprising the machine-learned generative language model, compared to alternative methods with fewer or less effective error prevention mechanisms.

[0046] As another example, systems and methods according to example aspects of the present disclosure may in some instances reduce a computational cost of generating machine-learned outputs compared to some alternative methods with a similar accuracy. For example, in some instances, the factual accuracy of a machine-learned language output can be increased by receiving user feedback and modifying one or more models in the plurality of models based on the user feedback. For example, user feedback can indicate that a subtask associated with the answer is inaccurate, and the system can update a parameter of the model that generated the output associated with that subtask based on the user feedback. As a result, the system can provide accurate answers with smaller sized models (e.g., number of parameters). This in turn can reduce the computational cost (e.g., electricity cost, processor usage, memory usage, hardware cost) of training the machine-learned model and a computational cost of generating outputs with the machine-learned model after training. For example, a large increase in model complexity (e.g., doubling of parameter count) may only lead to a small marginal increase in accuracy. Additionally, the increase in accuracy may in some instances have a log-linear relationship with model complexity, meaning that increased complexity will lead to diminishing returns in accuracy as model complexity increases. Advantageously, systems and methods according to some aspects of the present disclosure can provide substantially improved accuracy compared to alternative methods, without increasing the complexity of the machine-learned language model. In this manner, for instance, systems and methods according to some aspects of the present disclosure can provide accurate machine-learned output at reduced computational cost (e.g., model training costs, inference costs) compared to alternative methods having a similar accuracy.

[0047] A technical effect of example implementations of the present disclosure is increased energy efficiency in performing operations using machine-learned models, thereby improving the functioning of computers implementing such models. For instance, example implementations can provide for more energy-efficient training operations or model updates by targeting a specific model associated with the subtask in a plurality of models to fine-tune (e.g., update one or more parameters) based on user feedback. Therefore, by focusing only on the model associated with the subtask that was incorrect, the system does not need to update all of the different models that were utilized in generating the full response. In some scenarios, increased energy efficiency can provide for less energy to be used to perform a given number of inference or training tasks (e.g., less energy expended to maintain the model in memory, less energy expended to perform calculations within the model, such as computing gradients, backpropagating a loss). In some scenarios, increased energy efficiency can provide for more inference or training tasks to be completed for a given energy budget (e.g., a larger quantity of training iterations). In some scenarios, greater expressivity afforded by systems and methodsof the present disclosure can provide for a given level of functionality to be obtained in fewer training iterations, thereby expending a smaller energy budget. In some scenarios, greater expressivity afforded by systems and methods of the present disclosure can provide for an extended level of functionality to be obtained in a given number of training iterations, thereby more efficiently using a given energy budget.

[0048] With reference now to the Figures, example embodiments of the present disclosure will be discussed in further detail.Example Systems

[0049] Figure 1 depicts a block diagram of an example system 100 for generating a response 170 for a user query 115 according to example embodiments of the present disclosure. In some instances, the system 100 can be, comprise, be comprised by, or share one or more properties with a computing device or system described below with respect to Figures 5A-5C (e.g., server computing system 730, training computing system 750, computing device 10, computing device 50).

[0050] According to some embodiments, the system 100 can receive a user query 115 from a user device 110 and feed the user query to a first machine-learned model 120 to generate a response 170. The first machine-learned model 120 can be a model in a plurality of models 160. The plurality of models can include a first machine-learned model 120, a second machine- learned model 130, and third machine-learned model 140, a fourth machine-learned model 150, and so on.

[0051] The first machine-learned model 120 can process the user query and determine that it has a first subtask 122 (e.g., making a dinner reservation) and a second subtask 124 (e.g., making a movie reservation). Based on this determination, the first-machine-leamed model 120 can send a first request associated with the first subtask 122 to the second machine-learned model 130 and a second request associated with the second subtask 124 to the third machine- learned model. 140. For example, the first subtask 122 can be associated with making a dinner reservation, and the second subtask 124 can be associated with making a movie reservation. In this example, the first machine-learned model 120 can request the second machine-learned model 130 that specializes in dinner reservations to select a restaurant, determine an optimal dinner time, and book a reservation. Additionally, in this example, after the selection of the restaurant and dinner time, the second machine-learned model 130 can submit a request to a fourth machine-learned model 150 to book the dinner reservation for the selected time. Continuing with this example, the first machine-learned model 120 can request the thirdmachine-learned model 140 that specializes in movies to select a movie to watch. The second machine-learned model 130 can generate a first output and the third machine-learned model 140 can generate a second output that is transmitted to the first machine-learned model 120. The first machine-learned model can generate a response 170 to the user query based on the first output and the second output.

[0052] In some instances, the user device can provide user feedback to the response 170. For example, the user feedback can include modifying the second output (e.g., changing the movie or movie time). In this example, the first machine-learned model 120 can determine that the user feedback is associated with the second subtask, and as a result, send the user feedback to the third machine-learned model 140.Example Model Arrangements

[0053] Figure 2 depicts a flow diagram of a system 200 according to example embodiments of the present disclosure. The system 200 can include a user device 110 (e.g., smartphone) and a plurality of machine-learned models (e.g., a first machine-learned model 120, and a second machine-learned model 130)

[0054] The plurality of machine-learned models can communicate with each other and work together to generate a response to the user query efficiently. By orchestrating a plurality of machine-learned models effectively, the system 200 can generate improved responses than conventional systems, while minimizing operational overhead.

[0055] At 210, the user device 110 can transmit a user query to the first machine- learned model 120. In a first example, the user query can be to make a dinner reservation and a movie reservation.

[0056] At 220, the first machine-learned model 120 can determine that the user query is associated with a first subtask, and the second machine-learned model can process the first subtask. Based on this determination, the first machine-learned model can transmit a first request associated with the first subtask to the second machine-learned model 130. Continuing with the first example, the first machine-learned model 120, which can be an orchestrator model, can determine that the user query is associated with a first subtask (e.g., selecting a restaurant and making a reservation). Based on this determination, the orchestrator model can select a second machine-learned model 130 to select the restaurant based on user preferences and book the reservation. In this first example, the second machine-learned model 130 can be specialized for selecting restaurants based on user preferences and booking the restaurant. Insome instances, the second-machine-learned model can have access to a user’s calendar and a restaurant booking system to be able to make the dinner reservation.

[0057] In a first scenario, the second machine-learned model 130 can process the first subtask and determine that the user query is associated with a second subtask, and that the second subtask can be processed by a fourth machine-learned model 150. Continuing with the first example, the second machine-learned model 130 can be a model that specializes in selecting a restaurant based on reviews, rankings, user preferences, past user experiences, user feedback, and so on. Additionally, after selecting a restaurant, the second machine-learned model 130 can instruct the fourth machine-learned model 150 to book the restaurant. In this example, the fourth machine-learned model 150 may be stored in the user device 110 and have access to a user’s calendar and be authorized by a restaurant booking system to make the dinner reservation.

[0058] At 230, the second machine-learned model 130 can process the first request and generate a first output associated with the first subtask. The second machine-learned model 130 can send the first output to the first machine-learned model 120. Continuing with the first example, the first output can be a confirmation of a dinner reservation at a first time at a first restaurant.

[0059] At 240, the first machine-learned model 120 can process the first output to generate a first response. In some instances, the first machine-learned 120 can process a plurality of different outputs from a plurality of different models to generate the first response. Continuing with the first example, the first response can be a confirmation of a dinner reservation at a first time at a first restaurant.

[0060] Additionally, the first response can also include a confirmation of a movie reservation at a second time of a first movie. For example, the first machine-learned model 120, which can be an orchestrator model, can determine that the user query is associated with a third subtask associated with making a movie reservation. Moreover, the orchestrator model can determine that the third machine-learned model 140 is specialized in making a movie reservation. Thus, the first machine-learned model 120 can send instructions to the third machine-learned model 140 to perform an action associated with the third subtask (e.g., booking a movie reservation).

[0061] At 250, the user device 110 can provide user feedback to the first machine- learned model 120. The user feedback can be natural language data that is received from a user. For example, the user feedback can be that the dinner reservation is too early.

[0062] At 260, the first machine-learned model 120 can determine that the user feedback is associated with the first subtask. Subsequently, the first machine-learned model 120 can send the user feedback associated with the first subtask to the second machine-learned model 130.

[0063] In some instances, at 270, the second machine-learned model 130 can process the user feedback to generate a second output. Continuing with the first example, the second machine-learned model 130 can update the dinner reservation to a second time based on the user feedback. The second output can be sent to the first machine-learned model 120. Additionally, or alternatively, at 290, the system 200 can update a parameter of the second machine-learned model 130 based on the user feedback.

[0064] Continuing with the first scenario described above, the second machine-learned model 130 can process the user feedback and determine that the user feedback is associated with a second subtask that was processed by the fourth machine-learned model 150. Thus the second machine-learned model 130 can send the user feedback to the fourth machine-learned model 150. Additionally, in some instances, the fourth machine-learned model can provide model feedback to the second machine-learned model. The model feedback can include what the fourth machine-learned model would have needed to get the second subtask solved correctly (e.g. additional constraint, different prompt). For example, the mechanism can be based on the second machine-learned model 130 querying the fourth machine-learned model 150 with "here is what I got from you, here is what the user expected, here is the original query I sent you."

[0065] At 280, the first machine-learned model can generate a second response to the user query by processing the second output.Example Methods

[0066] Figure 3 depicts a flow chart diagram of an example method for utilizing a plurality of machine-learned models to generate a response to a user query according to example embodiments of the present disclosure. Although Figure 3 depicts steps performed in a particular order for purposes of illustration and discussion, the methods of the present disclosure are not limited to the particularly illustrated order or arrangement. The various steps of method 300 can be omitted, rearranged, combined, and / or adapted in various ways without deviating from the scope of the present disclosure.

[0067] In some instances, the system can obtain (e.g., receive), from the user device, the task.

[0068] At 302, the system can process, using a first machine-learned model from a plurality of machine-learned models, the task to determine a first subtask.

[0069] In some instances, the system can transmit the first subtask to the second machine-learned model.

[0070] At 304, the system can process, using a second machine-learned model from the plurality of machine-learned models, the first subtask to generate a first output.

[0071] In some instances, the plurality of machine-learned models can include a third machine-learned model. Additionally, the system can process, using the first machine-learned model, the task to determine a second subtask. Moreover, the system can process, using the third machine-learned model, the second subtask to generate a second output, wherein the first response is based on the first output and the second output. Furthermore, the system can process, using the first machine-learned model, the user feedback to generate a second feedback for the third machine-learned model.

[0072] At 306, the system can process, using the first machine-learned model, the first output to generate a first response associated with a task.

[0073] In some instances, the system can cause a presentation of the first response on a graphical user interface of the user device.

[0074] At 308, the system can receive, from a user device, user feedback associated with the first response.

[0075] In some instances, the user feedback is a user input on the graphical user interface.

[0076] In some instances, the system can process, using the first machine-learned model, the user feedback to determine that the user feedback is associated with the first subtask. Additionally, the system can transmit the first feedback to the second machine-learned model based on the determination that the user feedback is associated with the first subtask.

[0077] At 310, the system can process, using the first machine-learned model, the user feedback to generate the first feedback for the second machine-learned model.

[0078] In some instances, the system can transmit the first feedback to the second machine-learned model. Additionally, the system can process, using the second-machine- leamed model, the first feedback to generate a second output. Moreover, the system can process, using the first machine-learned model, the second output to generate a second response associated with the task. Furthermore, the system can present the second response to the user device, the second response being modified based on the user feedback to the first response.

[0079] In some instances, the system can update a parameter of the second machine- learned model based on the first feedback.

[0080] In some instances, the first feedback is a score.

[0081] In some instances, the first feedback is a prompt.

[0082] In some instances, the first machine-learned model and the second machine- learned model can be stored in the one or more non-transitory computer-readable media of the computing system.

[0083] In some instances, the first machine-learned model can be stored in the user device and the second machine-learned model is stored in the computing system.

[0084] In some instances, the first machine-learned model can be stored in the computing system and the second machine-learned model is stored in another system.

[0085] In some instances, the first response is an audio presentation, and the user feedback is audio data.

[0086] Figure 4 depicts a flow chart diagram of an example method for providing feedback to one or more machine-learned models based on a determination of a subtask according to example embodiments of the present disclosure. Although Figure 3 depicts steps performed in a particular order for purposes of illustration and discussion, the methods of the present disclosure are not limited to the particularly illustrated order or arrangement. The various steps of method 400 can be omitted, rearranged, combined, and / or adapted in various ways without deviating from the scope of the present disclosure.

[0087] At 402, the system can receive, from a user device, user feedback associated with the first response.

[0088] At 404, the system can process, using the first machine-learned model, the user feedback to determine that the user feedback is associated with the first subtask.

[0089] At 406, the system can transmit the user feedback to the second machine-learned model.

[0090] At 408, the system can process, using the second-machine-leamed model, the user feedback to generate a second output.

[0091] At 410, the system can update a parameter of the second machine-learned model based on the first feedback.Example Devices and Systems

[0092] Figure 5 A depicts a block diagram of an example computing system according to example embodiments of the present disclosure. The system 700 includes a user computingdevice 702, a server computing system 730, and a training computing system 750 that are communicatively coupled over a network 780.

[0093] The user computing device 702 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, or any other type of computing device.

[0094] The user computing device 702 includes one or more processors 712 and a memory 714. The one or more processors 712 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. The memory 714 can include one or more non-transitory computer-readable storage media, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory 714 can store data 716 and instructions 718 which are executed by the processor 712 to cause the user computing device 702 to perform operations.

[0095] In some implementations, the user computing device 702 can store or include one or more machine-learned models 720, such as machine-learned generation models 108 or machine-learned evaluation models 112. For example, the machine-learned models 720 can be or can otherwise include various machine-learned models such as neural networks (e.g., deep neural networks) or other types of machine-learned models, including non-linear models and / or linear models. Neural networks can include feed-forward neural networks, recurrent neural networks (e.g., long short-term memory recurrent neural networks), convolutional neural networks or other forms of 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 (e.g., transformer models). Example machine-learned models 720 are discussed with reference to Figures 1-3.

[0096] In some implementations, the one or more machine-learned models 720 can be received from the server computing system 730 over network 780, stored in the user computing device memory 714, and then used or otherwise implemented by the one or more processors 712. In some implementations, the user computing device 702 can implement multiple parallel instances of a single machine-learned model 720 (e.g., to perform parallel insight summary generation or evaluation across multiple instances of machine-learned generation model 108 or machine-learned evaluation model 112).

[0097] Additionally or alternatively, one or more machine-learned models 740 (e.g., machine-learned generation models 108, machine-learned evaluation models 112, etc.) can beincluded in or otherwise stored and implemented by the server computing system 730 that communicates with the user computing device 702 according to a client-server relationship. For example, the machine-learned models 740 can be implemented by the server computing system 730 as a portion of a web service (e.g., an advertising analytics service, etc.). Thus, one or more models 720 can be stored and implemented at the user computing device 702 and / or one or more models 740 can be stored and implemented at the server computing system 730.

[0098] The user computing device 702 can also include one or more user input components 722 that receives user input. For example, the user input component 722 can be a touch-sensitive component (e.g., a touch-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, a traditional keyboard, or other means by which a user can provide user input.

[0099] The server computing system 730 includes one or more processors 732 and a memory 734. The one or more processors 732 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. The memory 734 can include one or more non-transitory computer-readable storage media, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory 734 can store data 736 and instructions 738 which are executed by the processor 732 to cause the server computing system 730 to perform operations.

[0100] In some implementations, the server computing system 730 includes or is otherwise implemented by one or more server computing devices. In instances in which the server computing system 730 includes plural server computing devices, such server computing devices can operate according to sequential computing architectures, parallel computing architectures, or some combination thereof.

[0101] As described above, the server computing system 730 can store or otherwise include one or more machine-learned models 740 (e.g., machine-learned generation models 108, machine-learned evaluation models 112, etc.). For example, the models 740 can be or can otherwise include various machine-learned models. Example machine-learned models include neural networks or other multi-layer non-linear models. Example neural networks include feed forward neural networks, deep neural networks, recurrent neural networks, and convolutional 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-atention models (e.g., transformer models). Example models 740 are discussed with reference to Figures 1-3.

[0102] The user computing device 702 and / or the server computing system 730 can train the models 720 and / or 740 via interaction with the training computing system 750 that is communicatively coupled over the network 780. The training computing system 750 can be separate from the server computing system 730 or can be a portion of the server computing system 730.

[0103] The training computing system 750 includes one or more processors 752 and a memory 754. The one or more processors 752 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. The memory 754 can include one or more non-transitory computer-readable storage media, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory 754 can store data 756 and instructions 758 which are executed by the processor 752 to cause the training computing system 750 to perform operations. In some implementations, the training computing system 750 includes or is otherwise implemented by one or more server computing devices.

[0104] The training computing system 750 can include a model trainer 760 that trains the machine-learned models 720 and / or 740 stored at the user computing device 702 and / or the server computing system 730 using various training or learning techniques, such as, for example, backwards propagation of errors. For example, a loss function can be back propagated through the model(s) to update one or more parameters of the model(s) (e.g., based on a gradient of the loss function). Various loss functions can be used such as mean squared error, likelihood loss, cross entropy loss, hinge loss, and / or various other loss functions. Gradient descent techniques can be used to iteratively update the parameters over a number of training iterations. The training can implement supervised learning, unsupervised learning, reinforcement learning, and other training techniques.

[0105] In some implementations, performing backwards propagation of errors can include performing truncated backpropagation through time. The model trainer 760 can perform a number of generalization techniques (e.g., weight decays, dropouts, etc.) to improve the generalization capability of the models being trained.

[0106] In some implementations, the model(s) 720 can be pre-trained before domainspecific alignment. For instance, a model 720 can be pre trained over a general corpus of training data and fine-tuned on a more targeted corpus of training data. A model 720 can bealigned using prompts that are designed to elicit domain-specific outputs. Prompts can be designed to include learned prompt values (e.g., soft prompts). The trained model(s) 720 may be validated prior to their use using input data other than the training data and may be further updated or refined during their use based on additional feedback / inputs.

[0107] In particular, the model trainer 760 can train the machine-learned models 720 and / or 740 based on a set of training data 762. Training data 762 for the machine-learned generation model can include, for example, input-output pairs comprising structured insight data 106 as inputs, and example outputs 110, 118 as outputs. Training data 762 for the machine- learned evaluation model 112 can include, for example, input-output pairs comprising candidate outputs 110 as inputs, and evaluations 114 (e.g., numerical evaluation scores, etc.) as outputs. In some instances, training data 762 for the machine-learned evaluation model can also include structured insight data 106 as part of the inputs of each input-output pair. In some instances, training data 762 for the machine-learned generation model 108 or machine-learned evaluation model 112 can include, for example, input-output pairs comprising structured insight data 106 as inputs, and feedback inputs 324 as outputs.

[0108] In some implementations, if the user has provided consent, the training examples can be provided by the user computing device 702. Thus, in such implementations, the model 720 provided to the user computing device 702 can be trained by the training computing system 750 on user-specific data received from the user computing device 702. In some instances, this process can be referred to as personalizing the model.

[0109] The model trainer 760 includes computer logic utilized to provide desired functionality. The model trainer 760 can be implemented in hardware, firmware, and / or software controlling a general purpose processor. For example, in some implementations, the model trainer 760 includes program files stored on a storage device, loaded into a memory and executed by one or more processors. In other implementations, the model trainer 760 includes one or more sets of computer-executable instructions that are stored in a tangible computer- readable storage medium such as RAM, hard disk, or optical or magnetic media.

[0110] The network 780 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 the network 780 can be carried via any type of wired and / or wireless connection, using 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).

[0111] Figure 5 A illustrates one example computing system that can be used to implement the present disclosure. Other computing systems can be used as well. For example, in some implementations, the user computing device 702 can include the model trainer 760 and the training dataset 762. In such implementations, the models 720 can be both trained and used locally at the user computing device 702. In some of such implementations, the user computing device 702 can implement the model trainer 760 to personalize the models 720 based on userspecific data.

[0112] Figure 5B depicts a block diagram of an example computing device 10 that performs according to example embodiments of the present disclosure. The computing device 10 can be a user computing device or a server computing device.

[0113] The computing device 10 includes a number of applications (e.g., applications 1 through N). Each application contains its own machine learning library and machine-learned model(s). For example, 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, and other applications.

[0114] As illustrated in Figure 5B, 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, and / 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.

[0115] Figure 5C depicts a block diagram of an example computing device 50 that performs according to example embodiments of the present disclosure. The computing device 50 can be a user computing device or a server computing device.

[0116] The computing device 50 includes a number of applications (e.g., applications 1 through N). Each application is 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, 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).

[0117] The central intelligence layer includes a number of machine-learned models. For example, as illustrated in Figure 7C, 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 someimplementations, 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 the computing device 50.

[0118] 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 the computing device 50. As illustrated in Figure 7C, the central device data layer 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, and / 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).

[0119] Example Training Method

[0120] Figure 6 depicts a flowchart of a method 800 for training one or more machine- learned models according to aspects of the present disclosure. For instance, an example machine-learned model can include a machine-learned model 235.

[0121] One or more portion(s) of example method 800 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 figures. Each respective portion of example method 800 can be performed by any (or any combination) of one or more computing devices. Moreover, one or more portion(s) of example method 800 can be implemented on the hardware components of the device(s) described herein, for example, to train one or more systems or models. Figure 6 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 present disclosure. Figure 6 is described with reference to elements / terms described with respect to other systems and figures for exemplary illustrated purposes and is not meant to be limiting. One or more portions of example method 800 can be performed additionally, or alternatively, by other systems.

[0122] At 802, example method 800 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 800 as a “training” instance, it is to be understood that runtime inferences can form training instances when a model is trainedusing an evaluation of the model’s performance on that runtime instance (e.g., online training / learning). Example data types for the training instance and various tasks associated therewith are described throughout the present disclosure.

[0123] At 804, example method 800 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.

[0124] At 806, example method 800 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).

[0125] At 808, example method 800 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 back propagated 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 b ackpropagation through time. Example method 800 can include implementing a number of generalization techniques (e.g., weight decays, dropouts, etc.) to improve the generalization capability of the models being trained.

[0126] In some implementations, example method 800 can be implemented fortraining 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.).

[0127] In some implementations, example method 800 can be implemented for particular stages of a training procedure. For instance, in some implementations, example method 800 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 800 can be implemented 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

[0128] Figure 7 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.

[0129] 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.

[0130] 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 multiheaded self-attention models.

[0131] 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 frominput(s) 2. For example, machine-learned model(s) 1 can employ a mixture-of-experts structure.

[0132] 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.

[0133] 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.

[0134] 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.

[0135] 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 present disclosure are not limited to those examples noted above.

[0136] Example Machine-Learned Sequence Processing Models

[0137] Figure 8 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 sequence 5. Input sequence 5 can include one or more input elements 5-1, 5-2, . . . , 5-M, 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.

[0138] 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 (LLMs). Other example sequence processing models can operate in other domains, such as image domains, audio domains, biochemical domains, 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.

[0139] 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).

[0140] 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 data from 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.

[0141] 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.

[0142] 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 outputa 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. Imagebased input source(s) can be tokenized by extracting and serializing patches from an image.

[0143] 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 Figure 8 can be the tokens or can be the embedded representations thereof.

[0144] 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.

[0145] Prediction layer(s) 6 can evaluate associations between portions of input sequence 5 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 the respective 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.”

[0146] A transformer is an example architecture that can be used in prediction layer(s) 4. 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 multilayer perceptron).

[0147] 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.

[0148] 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.

[0149] 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.

[0150] Output sequence 7 can be generated autoregressive. 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 autoregressive 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.

[0151] Output sequence 7 can also be generated non-autoregressive. For instance, multiple output elements of output sequence 7 can be predicted together without explicit sequential conditioning on each other. 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.

[0152] Figure 9 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.

[0153] 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.

[0154] 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.

[0155] 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, the projection 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.

[0156] 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 learned within a continuous embedding space.

[0157] 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).

[0158] 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 data type data-to-sequence model can subdivide an input of that arbitrary data type and project the subdivisions into element(s) in input sequence 8 (e.g., elements 8-7, 8-8, 8-9, etc.).

[0159] 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

[0160] Figure 10 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.

[0161] 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.

[0162] 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.

[0163] 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.

[0164] 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 accuracy, precision, and / or recall 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).

[0165] 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.

[0166] 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.

[0167] 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.

[0168] 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.

[0169] 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.

[0170] 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).

[0171] 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 elementvalues) based on one or more training iterations. Workbench 15 can implement prompt engineering tools in development model 16.

[0172] 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.

[0173] 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.

[0174] 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 800 described above.

[0175] 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 offloading certain tasks to a more focused tool for rote application of deterministic algorithms to a well-defined problem.

[0176] 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”).

[0177] 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.

[0178] 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 instructions that initiate API calls to send or obtain data via external systems.

[0179] 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.

[0180] 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 in development 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 ofdevelopment model 16 can be efficiently transferred to a smaller model for more efficient inference.

[0181] 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.

[0182] Figure 11 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 figures. 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. 13 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 present disclosure. FIG. 13 is described with reference to elements / terms described with respect to other systems and figures 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.

[0183] 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 values can 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.

[0184] 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).

[0185] 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 model23 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 has satisfactory performance, if the model was already fine-tuned, or if other tuning approaches are preferred.

[0186] Fine-tuned model 25 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 25 can be the initial state if development model 16 was already fine-tuned. Fine-tuned model 25 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.

[0187] 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 pretraining 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 to downstream 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

[0188] Figure 12 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.

[0189] 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, modelhost 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.

[0190] 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). The knowledge graph 37-1 can be generated by 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.

[0191] 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.

[0192] 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.

[0193] In some implementations, model host 31 can operate on the 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 the 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 subroutineor 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.

[0194] 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 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.

[0195] 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 include memory 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.

[0196] 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.

[0197] 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 performinference 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.

[0198] 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.

[0199] 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.

[0200] 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 or otherwise 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.

[0201] 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 imageprocessing 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.

[0202] 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 an output. 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).

[0203] 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) 1can 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.

[0204] 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 another example, 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.

[0205] 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.

[0206] 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 canprocess 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.

[0207] 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. In another 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.

[0208] 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.

[0209] 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.

[0210] In some implementations, the task can be an instruction following the 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 thatrepresent 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 to sequentially 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.

[0211] 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.

[0212] 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 canbe 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).

[0213] 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).

[0214] 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).Additional Disclosure

[0215] 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.

[0216] While the present subject matter has been described in detail with respect to various specific example embodiments thereof, 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 subject disclosure does not preclude inclusion of such modifications, variations and / or additions to the present subject matter as would be readily apparent to one of ordinary skill in the art. For instance, 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 present disclosure covers such alterations, variations, and equivalents.

[0217] Further to the descriptions above, a user may be provided with controls allowing the user to make an election as to both if and when systems, programs or features described herein may enable collection of user information (e.g., information about a user's search engine queries or other search engine data, location data, online transaction data, a user's preferences, or other user data), and if the user is sent content or communications from a server. In addition, as will be described further below, certain data may be treated in one or more ways before it is stored or used, so that personally identifiable information is removed. For example, a user's identity may be treated so that no personally identifiable information can be determined for the user, or a user's geographic location may be generalized where location information is obtained (such as to a city, ZIP code, or state level), so that a particular location of a user cannot be determined. Thus, the user may have control over what information is collected about the user, how that information is used, and what information is provided to the user.

[0218] In particular, in some implementations, the computing system can remove personal information from the obtained data so that information such as specific user identities or user identifiers is not included in the data. For example, in some implementations, the computing system can respectively assign a unique and non-personally identifying identifier to the search engine data for each of the first plurality of users and then remove any personal information from the search engine data. As examples, the unique and non-personally identifying identifier can include a random number, a result of a hash function applied to a user identifier, or some other obfuscated or encrypted identifier which cannot be reverted to obtain the user identifier. Likewise, the computing system can respectively assign the same unique and non-personally identifying identifier to the location data for each of the first plurality of users and then remove personal information from the location data.

[0219] Thus, following a privacy protection process, the computing system can obtain sets of search engine data that are respectively connectable to sets of location data by way of a shared unique and non-personally identifying identifier. However, these sets of data contain no personal information such as specific user identities. As such, the data is anonymous, no userlevel labels are stored, and, as will be described further below, the computation aggregates all data at the location level.

Claims

WHAT IS CLAIMED IS:

1. A computing system for utilizing a plurality of machine-learned models to perform a task, comprising: one or more processors; and one or more non-transitory computer-readable media that collectively store instructions that, when executed by the one or more processors, cause the computing system to perform operations, the operations comprising: processing, using a first machine-learned model from the plurality of machine- learned models, the task to determine a first subtask; processing, using a second machine-learned model from the plurality of machine-learned models, the first subtask to generate a first output; processing, using the first machine-learned model, the first output to generate a first response associated with the task; receiving, from a user device, user feedback associated with the first response; and processing, using the first machine-learned model, the user feedback to generate a first feedback for the second machine-learned model.

2. The computing system of claim 1, the operations further comprising: transmitting the first feedback to the second machine-learned model; processing, using the second machine-learned model, the first feedback to generate a second output; processing, using the first machine-learned model, the second output to generate a second response associated with the task; and presenting the second response to the user device, the second response being modified based on the user feedback to the first response.

3. The computing system of claim 1, the operations further comprising: updating a parameter of the second machine-learned model based on the first feedback.

4. The computing system of claim 1, the operations further comprising:processing, using the first machine-learned model, the user feedback to determine that the user feedback is associated with the first subtask; and transmitting the first feedback to the second machine-learned model based on the determination that the user feedback is associated with the first subtask.

5. The computing system of claim 1, wherein the first feedback is a score.

6. The computing system of claim 1, wherein the first feedback is a prompt, an instruction, or a textual example modifying the first response.

7. The computing system of claim 1, the operations further comprising: transmitting the first subtask to the second machine-learned model.

8. The computing system of claim 1, wherein the plurality of machine-learned models include a third machine-learned model, the operations further comprising: processing, using the first machine-learned model, the task to determine a second subtask; processing, using the third machine-learned model, the second subtask to generate a second output, wherein the first response is based on the first output and the second output; and processing, using the first machine-learned model, the user feedback to generate a second feedback for the third machine-learned model.

9. The computing system of claim 1, wherein the plurality of machine-learned models include a third machine-learned model, the operations further comprising: processing, using the second machine-learned model, the first subtask to determine a second subtask; processing, using the third machine-learned model, the second subtask to generate a second output, wherein the first response is based on the first output and the second output; and processing, using the second machine-learned model, the first feedback to generate a second feedback for the third machine-learned model.

10. The computing system of claim 1, wherein the first machine-learned model and the second machine-learned model are stored in the one or more non-transitory computer-readable media of the computing system.

11. The computing system of claim 1, wherein the first machine-learned model is stored in the user device and the second machine-learned model is stored in the computing system.

12. The computing system of claim 1, wherein the first machine-learned model is stored in the computing system and the second machine-learned model is stored in another system.

13. The computing system of claim 1, the operations further comprising: obtaining, from the user device, the task.

14. The computing system of claim 1, the operations further comprising: causing a presentation of the first response on a graphical user interface of the user device.

15. The computing system of claim 1, wherein the user feedback is a user input on a graphical user interface.

16. The computing system of claim 1, wherein the first response is an audio presentation, and the user feedback is audio data.

17. A computer-implemented method for utilizing a plurality of machine-learned models to perform a task, comprising: processing, using a first machine-learned model from the plurality of machine-learned models, the task to determine a first subtask; processing, using a second machine-learned model from the plurality of machine- learned models, the first subtask to generate a first output; processing, using the first machine-learned model, the first output to generate a first response associated with the task; receiving, from a user device, user feedback associated with the first response; andprocessing, using the first machine-learned model, the user feedback to generate a first feedback for the second machine-learned model.

18. The computer-implemented method of claim 17, the method further comprising: transmitting the first feedback to the second machine-learned model; processing, using the second machine-learned model, the first feedback to generate a second output; processing, using the first machine-learned model, the second output to generate a second response associated with the task; and presenting the second response to the user device, the second response being modified based on the user feedback to the first response.

19. The computer-implemented method of claim 17, further comprising: processing, using the first machine-learned model, the user feedback to determine that the user feedback is associated with the first subtask; and transmitting the first feedback to the second machine-learned model based on the determination that the user feedback is associated with the first subtask.

20. One or more non-transitory computer-readable media storing instructions that are executable by a computing system to perform operations, the operations comprising: processing, using a first machine-learned model from a plurality of machine-learned models, a task to determine a first subtask; processing, using a second machine-learned model from the plurality of machine- learned models, the first subtask to generate a first output; processing, using the first machine-learned model, the first output to generate a first response associated with the task; receiving, from a user device, user feedback associated with the first response; and processing, using the first machine-learned model, the user feedback to generate a first feedback for the second machine-learned model.

Images