Privately sharing known information with and between agents

Abstract

Systems and methods for privately sharing known information with and between agents. The method can include obtaining a user query. The method can include processing the user query to perform an initial inference based on documents accessible by the client device. The method can include generating an initial response to the user query. The method can include transmitting data comprising the user query and hashed document data to a second agent. The method can include receiving a second response to the user query. The method can include processing the initial response and the second response to generate a composite response. The method can include providing the composite response to a user.

Description

PRIVATELY SHARING KNOWN INFORMATION WITH AND BETWEEN AGENTSFIELD

[0001] The present disclosure relates generally to machine learning processes and machine-learned devices and systems. More particularly, the present disclosure relates to privately sharing known information with and between agents.BACKGROUND

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

[0003] 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.

[0004] Example aspects of the present disclosure provide an example method. In some implementations, the example method can include obtaining, by a first agent on a client device, a user query; processing, by the first agent, the user query to perform an initial inference based on one or more documents accessible by the client device; generating, by the first agent, an initial response to the user query; transmitting, by the first agent, data comprising the user query and hashed document data to a second agent, wherein the hashed document data comprises at least one identifier associated with a first document of the one or more documents used to generate the initial response; receiving, by the first agent from the second agent, a second response to the user query, wherein the second response to the user query was generated by the second agent excluding the one or more documents used to generate the initial response; processing, by the first agent, the initial response and the secondresponse to generate a composite response; and providing, by the first agent, the composite response to a user.

[0005] Example aspects of the present disclosure provide one or more example non- transitory computer-readable media storing instructions that are executable by one or more processors to cause a computing system to perform example operations. In some implementations, the example operations can include accessing, by a second agent, a dataset comprising a plurality of hashed identifiers, wherein each respective hashed identifier of the plurality of hashed identifiers is associated with a document that the second agent has determined is relevant to a user query; determining, by the second agent, that a first hashed identifier associated with a first document is present in the dataset of hashed identifiers; removing, by the second agent, the first document from a context window for processing the user query based on determining that the first document was considered by a first agent computing system to generate an updated dataset; generating, by the second agent, a second response to the user query based on documents associated with the updated dataset; and transmitting, by the second agent, the second response to the first agent.

[0006] Example aspects of the present disclosure provide an example computing system that includes one or more processors and one or more example non-transitory computer-readable media storing instructions that are executable by one or more processors to cause a computing system to perform example operations. In some implementations, the example operations can include generating, by a first agent, an initial response to a user query based on one or more documents accessible by the first agent; transmitting, by the first agent, the user query and hashed document data to one or more second agents; accessing, by the one or more second agents, a document index, wherein the document index comprises a plurality of hashed identifiers of documents accessible by the second agent; excluding, by the one or more second agents, a first document based on the hashed identifier associated with the first document being in the hashed document data and the document index; and generating, by the one or more second agents, a second response based on one or more remaining documents accessible by the respective second agent.

[0007] Other example aspects of the present disclosure are directed to other systems, methods, apparatuses, tangible non-transitory computer-readable media, and devices for performing functions described herein. These and other features, aspects, and advantages of various implementations will become better understood with reference to the following description and appended claims. The accompanying drawings, which are incorporated in andconstitute a part of this specification, illustrate implementations of the present disclosure and, together with the description, help explain the related principles.BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 depicts a block diagram of example system to perform privately sharing known information between agents according to example embodiments of the present disclosure;

[0009] Figure 2 depicts a block diagram of an example data flow to perform privately sharing known information between agents according to example embodiments of the present disclosure;

[0010] Figure 3 depicts a block diagram of an example data flow to perform privately sharing known information between agents in cascade according to example embodiments of the present disclosure;

[0011] Figure 4 depicts a block diagram of an example data flow to perform privately sharing known information between agents in parallel according to example embodiments of the present disclosure;

[0012] Figure 5 depicts an example method for performing privately sharing known information between agents according to example embodiments of the present disclosure;

[0013] Figure 6 depicts an example method for performing privately sharing known information between agents according to example embodiments of the present disclosure;

[0014] Figure 7 depicts an example method for performing privately sharing known information between agents according to example embodiments of the present disclosure;

[0015] Figure 8 is a flow chart diagram illustrating an example method for training a machine-learned model according to example implementations of aspects of the present disclosure;

[0016] Figure 9 is a block diagram of an example processing flow for using machine- learned model(s) to process input(s) to generate output(s) according to example implementations of aspects of the present disclosure;

[0017] Figure 10 is a block diagram of an example sequence processing model according to example implementations of aspects of the present disclosure;

[0018] Figure 11 is a block diagram of an example technique for populating an example input sequence for processing by a sequence processing model according to example implementations of aspects of the present disclosure;

[0019] Figure 12 is a block diagram of an example model development platform according to example implementations of aspects of the present disclosure;

[0020] Figure 13 is a block diagram of an example training workflow for training a machine-learned model according to example implementations of aspects of the present disclosure;

[0021] Figure 14 is a block diagram of an inference system for operating one or more machine-learned model(s) to perform inference according to example implementations of aspects of the present disclosure;

[0022] Figure 15 is a block diagram of an example networked computing system according to example implementations of aspects of the present disclosure;

[0023] Figure 16 is a block diagram of an example computing device according to example implementations of aspects of the present disclosure; and

[0024] Figure 17 is a block diagram of an example computing device according to example implementations of aspects of the present disclosure.DETAILED DESCRIPTION

[0025] Generally, the present disclosure is directed to systems and methods that enable an agent to systematically search for new information without revealing what it knows to any outside party. In particular, the invention relates to the use of cryptographic hashes to represent documents or other information that is used by an agent to generate an initial response to a user query. The agent can pass the user query alongside a list of hashed identifiers for consideration by a remote agent in generating a response. As such, the remote agent can exclude documents that have already been considered by the agent (e.g., by performing a lookup for the hashed identifiers) and excluding those documents from being re-considered or reprocessed in generating a response.

[0026] In some implementations, the agent and remote agent can perform an initial communication to determine which hash functions are known by both systems. In some instances, the agent located on the client device can transmit a request for a list of hash functions which are understood by the remote agent. The agent located on the client device can apply a hash function to the identifiers of the documents that have been considered using a hash function in the list of hash functions which are understood by the remote agent. The hashed identifiers can then be transmitted to the remote agent along with the user query.

[0027] The remote agent can maintain a database of documents indexed by the hashes it understands. Additionally, or alternatively, the remote agent can determine a number of relevant documents and apply the known hash function to the identifier data associated with those selected documents. The remote agent can then retrieve documents relevant to the user query while excluding those that have already been considered by the agent. This can be performed by excluding any documents that are in both the list of hashed identifiers obtained from the agent and the list of hashed identifiers accessed by the remote agent. This can allow for the remote agent to only process new information to generate a response to the user’s query.

[0028] The present disclosure provides for a number of potential applications or use cases. Some example use cases can include mobile applications or assistants powered by large language models (LLMs). For instance, a mobile application or assistant could be tailored to a particular field or subject matter such as research, analytics, medical, or some other field. A user interacting with the agent can provide a request for information. For instance, the user can provide a query or natural language question. The agent can use the system and methods described herein to generate an initial response utilizing resources (e.g., documents) available to the agent and then make calls to additional agents to bolster the response using new resources in a privacy-preserving manner. The response can be bolstered by providing for the additional agents to exclude documents already considered by the initial agent while utilizing additionally accessible documents to provide a new response which can be transmitted to the initial agent, collated, and combined with the initial response to generate a composite response. The composite response can be transformed into a natural language form which can be provided for display to a user via a user interface of a client device.

[0029] The present disclosure provides for numerous technical effects and benefits. These advantages include the ability to allow the agent to systematically search for new information without revealing to any outside party what it knows. This can provide for security of a user and privacy of information accessible to the user via various agents. Additionally, the present disclosure provides for more efficient utilization of computing resources. For instance, the present disclosure allows for the agent to only process new information to generate a response to the user’s query. This can save the agent from performing unnecessary processing of previously processed information. Additionally, for agents (e.g., large language models (LLMs)) with finite context windows, the present disclosure can provide for a larger number of documents to be considered by the total of theagents compare to traditional systems which may consider the same documents more than once while excluding documents that may be cut off by traditional context window limits.

[0030] Additionally, the present disclosure allows the agent to continue reaching out to more agents until it finds the information it wants, e.g. until it has found an answer, or until it has a sufficiently detailed response. It then pieces together all the generated responses and returns a composite answer back to the user. Various example implementations are described herein with respect to the accompanying Figures. The improvements associated with the systems and methods discussed herein can be further understood with reference to the figures.

[0031] Reference now is made to the figures, which provide example arrangements of computing systems, model structures, and data flows for illustration purposes only. FIG. 1 illustrates an example data flow 100 to perform privately sharing known information between agents according to example embodiments of the present disclosure.

[0032] Figure 1 depicts a block diagram of example system 100 to perform privately sharing known information between agents according to example embodiments of the present disclosure. Example system 100 can include client device 101. Client device 101 can include an agent 102. Agent 102 can include or otherwise have access to one or more machine learned model(s) 104, indexed hashed identifiers 106, or documents 108. Client device 101 can communicate with one or more remote agents via network 140.

[0033] By way of example, client device 101 can communicate with first remote agent 110 via network 140. First remote agent 110 can include or otherwise have access to one or more machine learned model(s) 112, indexed hashed identifiers 114, or documents 116.

[0034] Client device 101 can communicate with second remote agent 120 via network 140. Second remote agent 120 can include or otherwise have access to one or more machine learned model(s) 122, indexed hashed identifiers 124, or documents 126.

[0035] Client device 101 can communicate with Nth remote agent 130 via network 140. Nth remote agent 130 can include or otherwise have access to one or more machine learned model(s) 132, indexed hashed identifiers 134, or documents 136.

[0036] As will be described in further detail with regard to Figure 2 to Figure 7, client device 101 can obtain a user query and perform an initial inference based on one or more documents 108 accessible by the client device 101. Client device 101 can apply a hash function to identifier data associated with the one or more documents 108 used to generate the initial response to generate hashed document data. Client device 101 can transmit the userquery and the hashed document data to one or more remote agents (e.g., first remote agent 110, second remote agent 120, Nth remote agent 130, etc.).

[0037] Figure 2 depicts a swim lane diagram of an example data flow 200 to perform privately sharing known information between agents according to example embodiments of the present disclosure. Example data flow 200 can include a first agent 205 and a remote agent 210. While Figure 2 depicts a single remote agent 210, it should be understood that any number of remote agents can be utilized in accordance with the present disclosure (e.g., as depicted in Figure 3 and Figure 4).

[0038] At operation 215, first agent 205 can process a user query. For instance, the user query can include a natural language query. As described herein, a natural language query can include a question or request received from a user to be processed by an agent utilizing an LLM associated with the agent. For instance, a natural language query could include a question about a medical condition, a question relating to analytics for a content item or number of content items, a research question, or a general question.

[0039] At operation 220, first agent 205 can perform an initial inference based on the user query and one or more documents accessible by the client device. Performing the initial inference can include generating an initial response to the user query. In some implementations, the initial response can be a natural language response to the user query.

[0040] At operation 225, the first agent 205 can apply a hash function to identifier data associated with the one or more documents used to generate the initial response to generate hashed document data. Hashed document data can include one or more hashed identifiers. In some implementations, the first agent 205 can request a list of hash functions understood by the remote agent 210. Responsive to receiving the request, the remote agent 210 can transmit a list of hash functions understood by the remote agent 210 to the first agent 205. The first agent 205 can apply a hash function to the identifier data associated with the one or more documents used to generate the initial response to generate hashed document data. The hashed document data can include one or more hashed identifiers.

[0041] At operation 230, the first agent 205 can transmit the user query and the hashed identifiers to remote agent 210. Responsive to receiving the user query and the hashed identifiers, remote agent 210 can access a dataset comprising a plurality of hashed identifiers. As described herein, the hashed identifiers can be identifiers of documents that have had a hash function applied to the identifier such that the true identifier of the document cannot be determined by the remote agent unless the remote agent has access to the same document.

[0042] At operation 235, the remote agent 210 can access an index of hashed identifiers of documents that are accessible by remote agent 210. In some instances, remote agent 210 can have access to a number of hash functions which are also accessible by the first agent 205. The agents can communicate and establish a shared hash function to apply to the identifiers. Additionally, the system can include a standardized naming protocol for documents such that if the same hash function is applied to the same document identifier, the output is the same. In some instances, a known hash function can be selected or established ahead of time so that the call to request known hash functions from the remote agent can be avoided.

[0043] At operation 240, the remote agent 210 can determine that a first document was utilized by the first agent 205 to generate the initial response. At operation 245, remote agent 210 can exclude, responsive to determining that the first document was utilized by the first agent 205 to generate the initial response, the first document from being included in the context window while processing the user query. As such, the system can conserve computing resources by considering a smaller amount of documents as context for processing a query. In some instances, the system can conserve computing resources because the system can determine that there are no additional documents to be considered, and the process can be truncated early. Additionally, the system can provide for a more efficient utilization of computing resources by preventing redundant processing of the same document multiple times and providing an improved response to a query taking into account a larger number of documents (e.g., sources) in formulating an answer. Each of these provides improvements over existing systems which are unable to exclude documents from utilization by an agent without the first agent needing to disclose the documents that were used and are accessible by the first agent.

[0044] At operation 250, remote agent 210 can perform an initial inference to generate a second response to the user query based on one or more additional accessible documents. In some implementations, the second response can be a natural language response to the user query.

[0045] At operation 255, remote agent 210 can transmit the second response to the first agent 205. In some implementations, in addition to transmitting the second response to the first agent 205, the remote agent 210 can also transmit second hashed document data. For instance, second hashed document data can include one or more hashed identifiers for the one or more additional accessible documents.

[0046] At operation 260, the first agent 205 can process the initial response and the second response to generate a combined response to the user query. An additional example implementation can include a user providing a query relating to a specific topic. The agent which is located on the client device can utilize browsing history or other relevant information that is stored locally but should not be shared with third parties. The agent can hash uniform resource locators (URLs) associated with the websites viewed in the browsing history. Then a second agent, such as a search engine agent, can exclude websites previously viewed by the user in determining a response to the query. An additional or alternative use case can include a user requesting information about a specific product without want to divulge prior purchase history to a shopping chatbot. The present disclosure can provide for transmitting hashed identifiers associated with certain previously purchased or viewed items so that they can be excluded from consideration by the shopping chatbot. As such, the user’s purchase history can be taken into account for formulating the answer without the chatbot having the ability to explicitly call out the user’s purchase history.

[0047] At operation 265, the first agent 205 can provide the combined response to the user for display via a user interface of the user device. For instance, the response can include a natural language response, an image, a reference to a document, a natural language response which can be provided via audio, or any other accessible form.

[0048] Figure 3 depicts an example block diagram of an example data flow 300 to perform privately sharing known information between agents in cascade according to example embodiments of the present disclosure. Example data flow 300 can include client device 305. Client device 305 can include an agent 307. Agent 307 can be utilized to generate an initial response 309 to a user query utilizing one or more documents which are accessible to agent 307. Client device 305 can apply a hash function to identifier data associated with the one or more documents used to generate the initial response to generate hashed document data (e.g., first hashed identifiers 314). Client device 305 can transmit data 310 to a remote system 315. Data 310 can include user query 312 or first hashed identifiers 314.

[0049] Remote system 315 can include a remote agent 317. Remote agent 317 can be utilized to generate a second response 319 to the user query utilizing one or more documents which are accessible to remote agent 317. Remote system 315 can apply a hash function to identifier data associated with the one or more documents used to generate the second response to generate hashed document data (e.g., second hashed identifiers 324). Remote system 315 can transmit data 320 to second remote system 325. Data 320 can include user query 312, second hashed identifiers 324, or second response 326.

[0050] User query 312 can be the same initial user query obtained at client device 305. In some implementations, second hashed identifiers 324 can include both the first hashed identifiers 314 and the second hashed identifiers 324. In some implementations, second hashed identifiers 324 can include just the second hashed identifiers for the additional one or more documents utilized to generate the second response 326. The second response 326 can include a combination of the first response and the second response. Additionally, or alternatively, the second response can be a response which is independent of the first response and will be passed along from one remote system or remote agent to the next until the responses and hashed identifiers are eventually passed back to the client device and initial agent.

[0051] Second remote system 325 can include a remote agent 327. Remote agent 327 can be utilized to generate a third response 329 to the user query utilizing one or more documents which are accessible to remote agent 327. Remote system 325 can apply a hash function to identifier data associated with the one or more documents used to generate the third response to generate hashed document data (e.g., third hashed identifiers 334). Remote system 325 can transmit data 330 to third remote system 335. Data 330 can include user query 312, third hashed identifiers 334, or third response 336.

[0052] User query 312 can be the same initial user query obtained at client device 305. In some implementations, third hashed identifiers 334 can include both the first hashed identifiers 314, the second hashed identifiers 324, and the third hashed identifiers 334. In some implementations, third hashed identifiers 334 can include just the third hashed identifiers for the additional one or more documents utilized to generate the third response 336. The third response 336 can include a combination of the first response, the second response, and the third response. Additionally, or alternatively, the third response can be a response which is independent of the first response and the second additional one or more documents utilized to generate the third response 336. The third response 336 can include a combination of the first response, the second response, and the third response. Additionally, or alternatively, the third response can be a response which is independent of the first response and the second additional one or more documents utilized to generate the third response 336. The third response 336 can include a combination of the first response, the second response, and the third response. Additionally, or alternatively, the third response can be a response which is independent of the first response and the second response and will be passed along from one remote system or remote agent to the next until the responses and hashed identifiers are eventually passed back to the client device and initial agent.

[0053] The process described herein can be performed for any number of remote systems or remote agents. While the agents are depicted as being associated with discrete systems, in some implementations, the agents can be located on a same device. However, the agents can be associated with different authentication credentials.

[0054] Nth remote system 335 can include a remote agent 337. Remote agent 337 can be utilized to generate an Nth response 339 to the user query utilizing one or more documents which are accessible to remote agent 337. Remote system 335 can apply a hash function to identifier data associated with the one or more documents used to generate the Nth response to generate hashed document data (e.g., Nth hashed identifiers 344). Remote system 335 can transmit data 340 to client device 305. Data 340 can include user query 312, Nth hashed identifiers 344, or Nth response 346.

[0055] User query 312 can be the same initial user query obtained at client device 305. In some implementations, Nth hashed identifiers 344 can include both the first hashed identifiers 314, the second hashed identifiers 324, the third hashed identifiers 334, and the Nth hashed identifiers 344. In some implementations, Nth hashed identifiers 344 can include just the Nth hashed identifiers for the additional one or more documents utilized to generate the Nth response 346. The Nth response 346 can include a combination of the first response, the second response, the third response, and the Nth response. Additionally, or alternatively, the Nth response can be a response which is independent of the first response, the second response, and the third response and will be passed along from the remote agent to the client device 305 and initial agent 307.

[0056] Client device 305 can process the first response 309 and the Nth response 346 to generate a combined response to the user query. Client device 305 can provide the combined response to the user for display via a user interface of the user device. In some implementations, client device 305 can store an aggregate list of hashed identifiers. Additionally, or alternatively, the client device 305 can be adjusted to alter the amount of time that the response and any associated data can be stored.

[0057] While the process can be performed via multiple computing devices or agents working in cascade (e.g., processing the user query one-by-one), an additional or alternatively embodiment can include a parallel approach as described in Figure 4.

[0058] Figure 4 depicts an example block diagram of an example data flow 400 to perform privately sharing known information between agents in parallel according to example embodiments of the present disclosure. Example data flow 400 can include client device 405. Client device 405 can include an agent 410. Agent 410 can generate an initialresponse 415 to a user query utilizing one or more documents which are accessible to agent 410. Client device 405 can transmit data 420 including user query and first hashed identifiers to remote system 425. Client device 405 can transmit data 445 including user query and first hashed identifiers to second remote system 450. Client device 405 can transmit data 470 including user query and first hashed identifiers to Nth remote system 475.

[0059] Remote system 425 can include remote agent 430 which can be used to generate second response 435. Second response 435 can be generated by excluding any documents considered by agent 410 in generating the initial response 415. As described herein, remote system 425 can perform a lookup and compare the first hashed identifiers with an index of hashed identifiers of documents accessible to the remote system 425. If any hashed identifiers are identified as belonging to both groups, the remote agent 430 can exclude the documents from being used in the context window for the remote agent 430 to generate second response 435.

[0060] Remote system 425 can transmit data 440 which can include the second response or hashed identifier data to client device 405. Client device 405 can aggregate the second response 435 with the initial response 415 as well as other obtained responses to generate a combined response. Client device 405 can provide the combined response to the user for display via a user interface of the user device.

[0061] Second remote system 450 can include remote agent 455 which can be used to generate third response 460. Third response 460 can be generated by excluding any documents considered by agent 410 in generating the initial response 415. As described herein, remote system 450 can perform a lookup and compare the first hashed identifiers with an index of hashed identifiers of documents accessible to the remote system 450. If any hashed identifiers are identified as belonging to both groups, the remote agent 455 can exclude the documents from being used in the context window for the remote agent 455 to generate third response 460.

[0062] Remote system 450 can transmit data 465 which can include the third response or hashed identifier data to client device 405. Client device 405 can aggregate the third response 460 with the initial response 415 as well as other obtained responses to generate a combined response. Client device 405 can provide the combined response to the user for display via a user interface of the user device.

[0063] Nth remote system 475 can include remote agent 480 which can be used to generate Nth response 485. Nth response 485 can be generated by excluding any documents considered by agent 410 in generating the initial response 415. As described herein, remotesystem 475 can perform a lookup and compare the first hashed identifiers with an index of hashed identifiers of documents accessible to the remote system 475. If any hashed identifiers are identified as belonging to both groups, the remote agent 480 can exclude the documents from being used in the context window for the can perform a lookup and compare the first hashed identifiers with an index of hashed identifiers of documents accessible to the remote system 475. If any hashed identifiers are identified as belonging to both groups, the remote agent 480 can exclude the documents from being used in the context window for the remote agent 480 to generate Nth response 485.

[0064] In some implementations the transmission of data to the respective remote systems (e.g., remote system 425, second remote system 450, or nth remote system 475) can be performed at the same time, or nearly the same time. In such an embodiment, it is possible that remote agent 43 and remote agent 455 can consider the same document. However, any document considered by agent 410 can be excluded from consideration. Additionally, or alternatively, the data can be transmitted to the respective remote systems sequentially such that the second remote system can exclude documents considered by both agent 410 and remote agent 430.

[0065] Figure 5 depicts a flow diagram of an example method 500 to perform privately sharing known information between agents according to example embodiments of the present disclosure. The method 500 can be performed by processing logic that can include hardware (e.g., processing device, circuitry, dedicated logic, programmable logic, microcode, hardware of a device, integrated circuit, etc.), software (e.g., instructions run or executed on a processing device), or a combination thereof. In some embodiments, method 500 is performed by a server computing system (e.g., server computing system 60) or client computing system (e.g., client computing device 50). Although shown in a particular sequence or order, unless otherwise specified, the order of the processes can be modified. Thus, the illustrated embodiments should be understood only as examples, and the illustrated processes can be performed in a different order, and some processes can be performed in parallel. Additionally, one or more processors can be omitted in various embodiments. Thus, not all processes are required in every embodiment. Other process flows are possible.

[0066] At operation 502, processing logic can obtain a user query. For instance, the user query can include a natural language query. A natural language query can include a question or request for information.

[0067] At operation 504, processing logic can process the user query to perform an initial inference based on one or more documents accessible by the client device. Forinstance, the initial inference can include generating an initial response to the user query. In some implementations, the initial response can be a natural language response to the user query. The initial response can be generated based on one or more documents which are accessible by the client device. The agent can keep track of which documents are utilized by the agent in performing the initial inference.

[0068] At operation 506, processing logic can generate an initial response to the user query. In some implementations, the initial response can be a natural language response to the user query. In some instances, processing the user query to generate an initial response to the user query can include processing logic performing retrieval-augment generation (RAG) based on the one or more documents accessible by the client device.

[0069] In some implementations, processing logic can transmit, by the first agent to a second agent, a request for a cryptographic hash understood by a second agent. For instance, the first agent can understand one or more cryptographic hashes such as hash functions. The first agent can confirm a common cryptographic hash to utilize. For instance, processing logic can receive, by the first agent from the second agent, a list of hash functions understood by the second agent. Processing logic can select, by the first agent, a first hash function of the list of hash functions. Processing logic can apply the first hash function to data associated with the first document. The data associated with the first document can include a document identifier such as a title or uniform resource locator (URL).

[0070] In some implementations, the cryptographic hash can include a user-specific salt. One way to achieve this is to use a user-specific salt, which can be a random value that is added to the input of the hash function before it is computed. This salt is unique to each user, and it is kept secret. When the agent hashes a document, it also includes the user's salt in the hash. This ensures that the hash is unique to the user, and that it cannot be easily reversed to determine the original document.

[0071] At operation 508, processing logic can transmit data comprising the user query and hashed document data to a second agent. The hashed document data can include at least one identifier associated with a first document of the one or more documents used to generate the initial response. In some instances, the second agent can be an agent located on the client device. Additionally, or alternatively, the second agent can include a remote agent. For instance, a remote agent can be an agent that is located on a device external to the client device or must be accessed by the first agent via a network or some form of tel ecommuni cati on .

[0072] The present disclosure can provide for a number of potential example use cases. For instance, applications can include personal use requesting an answer based on a corpus or private documents or can relate to larger scale use cases.

[0073] By way of example, a user may wish to obtain shopping recommendations. As such, a user can provide questions to various agents asking for product recommendations. A first agent can be associated with a first company. The first agent can identify products relevant to the query. These products can include identifiers such as barcodes, SKUs, or other universal or store-specific identifiers. The first agent can apply a hash to the identifiers and pass the hashed identifiers to one or more additional agents which can provide additional products while excluding those which have already been considered..

[0074] In addition, the present technology can be utilized in the context of performing research on an academic topic using a corpus of documents which require user authentication to access. For instance, a first agent can be located on a user device and can have information relating to a variety of profiles or credentials for a user for access to private databases (such as a digital library of publications, a subscription to a service such as a magazine or journal, etc.). As such, a user can provide a natural language query relating to a topic of interest. The first agent can generate a first response to the user query based on the user’s credentials and access to the private databases. The first agent can determine a number of second agents to communicate with to expand upon the first response. The first agent can transmit the user query to each of the second agents. In some implementations, the first agent can transmit the user query to each of the second agents at the same time. In some implementations, the first agent can transmit the user query to each of the second agents sequentially.

[0075] At operation 510, processing logic can receive a second response to the user query. The second response to the user query can be generated by the second agent excluding the one or more documents used to generate the initial response. For instance, the second agent can perform a lookup and compare the first hashed identifiers with an index of hashed identifiers of documents accessible to the second agent. If any hashed identifiers are identified as belonging to both groups, the second agent can exclude the documents from being used in the context window for the second agent to generate the second response. For instance, the agent can utilize a large language model to generate a response to the user query. The input to the large language model can include a prompt including a context window. The context window can be utilized to help generate better responses by honing in on details for the large language model to weigh more heavily in generating the response.

[0076] In some implementations, processing logic can receive, by the first agent from the second agent, second hashed document data. The second hashed document data can include one or more hashed document identifiers of documents used by the second agent in generating the second response to the user query.

[0077] At operation 512, processing logic can process the initial response and the second response to generate a composite response. The composite response can be a collated response generated by combining the initial response and the second response. By way of example, the composite response can include an aggregated response or a summary of the initial and second response.

[0078] In some instances, operations of the method described herein can be repeated for a number of second agents. The composite response can be generated based on responses obtained from the number of second agents. In some implementations, the operations described herein can be repeated for multiple second agents until the system determines that no new documents can be found that are relevant to the user query. As such, the lack of new documents can serve as an indicator that the processing logic should collate the responses to generate the composite response and refrain from calling additional agents for supplemental responses.

[0079] For instance, processing logic can transmit, by the first agent, data including the user query and the hashed document data to a third agent. Processing logic can receive, by the first agent from the third agent, a third response to the user query. The third response to the user query was generated by the third agent excluding the one or more documents used to generate the initial response. For instance, the third agent can be called in parallel to, or independently of the call to the second agent. As such, the third agent may utilize overlapping documents utilized by the second agent. Processing logic can process, by the first agent, the initial response, the second response, and the third response to update the composite response.

[0080] Additionally, or alternatively, processing logic can receive, by the first agent from the third agent, a third response to the user query, wherein the third response to the user query was generated by the third agent excluding the one or more documents used to generate the initial response and excluding the one or more documents used to generate the second response. For instance, the third agent can be called sequentially such that it can take into account the documents considered by both the initial agent and the second agent. As described herein, processing logic can process, by the first agent, the initial response, the second response, and the third response to update the composite response.

[0081] In some instances, processing logic can receive data comprising second hashed document data. The second hashed document data can include at least one identifier associated with a first document of one or more documents used to generate the second response.

[0082] In some implementations, processing logic can transmit, by the first agent, data including the user query and the hashed document data to a third agent. Processing logic can receive, by the first agent from the third agent, a third response to the user query. The third response to the user query can include a message that no additional relevant documents are accessible by the third agent. For instance, the third agent can either be unable to access any additional relevant documents or make a determination that all of the relevant documents have already been considered.

[0083] As described herein, any or all of the agents can include machine-learned models. In some implementations the machine-learned models can include large language models (LLM). LLMs will be discussed with further detail with regard to Figure 8 to 17.

[0084] At operation 514, processing logic can provide the composite response to the user. For instance, processing logic can provide the composite response for display via a user interface of the user device. For instance, the response can include a natural language response, an image, a reference to a document, a natural language response which can be provided via audio, or any other accessible form.

[0085] Figure 6 depicts a flow diagram of an example method 600 to perform privately sharing known information between agents according to example embodiments of the present disclosure. The method 600 can be performed by processing logic that can include hardware (e.g., processing device, circuitry, dedicated logic, programmable logic, microcode, hardware of a device, integrated circuit, etc.), software (e.g., instructions run or executed on a processing device), or a combination thereof. In some embodiments, method 600 is performed by a server computing system (e.g., server computing system 60) or client computing system (e.g., client computing device 50). Although shown in a particular sequence or order, unless otherwise specified, the order of the processes can be modified. Thus, the illustrated embodiments should be understood only as examples, and the illustrated processes can be performed in a different order, and some processes can be performed in parallel. Additionally, one or more processors can be omitted in various embodiments. Thus, not all processes are required in every embodiment. Other process flows are possible.

[0086] At operation 602, processing logic can access a dataset comprising a plurality of hashed identifiers. Each respective hashed identifier of the plurality of hashed identifierscan be associated with a document that the second agent has determined is relevant to a user query.

[0087] At operation 604, processing logic can determine that a first hashed identifier associated with a first document is present in the dataset of hashed identifiers. As described herein, the first hashed identifier can be an identifier that has been obfuscated utilizing a oneway hash. A one-way hash

[0088] A one-way hash function is a mathematical function that takes an input of any size and produces a fixed-size output, known as a hash. The hash is a unique fingerprint of the input, and it is computationally infeasible to reverse the process and recover the original input from the hash. This means that given a hash, it is practically impossible to determine the original input that produced it. One-way hash functions are commonly used in cryptography to ensure the integrity and authenticity of data. For example, a message can be hashed to create a digital signature, which can be used to verify that the message has not been tampered with. This fingerprint can be used to verify the authenticity and integrity of the document or message.

[0089] One-way hash functions are essential for ensuring the security and integrity of data in various applications. Their one-way property makes it impossible to reverse the hashing process, thereby protecting sensitive information from unauthorized access or modification. As described herein, a hash function can be applied to data associated with the first document to generate hashed document data. For example, a hash function can be applied to an identifier associated with the first document to generate a hash of the identifier. The hash of the identifier can be stored in a database as hashed document data. The hash function can be such that small changes to the input of the hash function do not necessarily equate to small changes in the output. For instance, the distribution of the hash function should be nicely spread to ensure there are no clusters or collisions generated when performing the hash function on the input. In addition, the hash function should be computationally infeasible to invert while also providing for generation of the same output given the same input. The hash function should have the property that small changes in the input to the hash function do not equate to small changes in the output. For instance, the distribution of the hash function should be spread such that clusters or collisions are avoided. Additionally, the hash function can have a characteristic that given a particular input, the same output will be generated.

[0090] At operation 606, processing logic can remove the first document from a context window for processing the user query based on determining that the first documentwas considered by a first agent computing system to generate an updated dataset. For instance, processing logic can perform a lookup and compare the first hashed identifiers with the dataset of hashed identifiers. If any hashed identifiers are identified as belonging to both groups, the second agent can generate an updated dataset that does not include the first document. The second agent can exclude the documents from being used in the context window for the second agent to generate the second response. For instance, the agent can utilize a large language model to generate a response to the user query.

[0091] At operation 608, processing logic can generate a second response to the user query based on the documents associated with the updated dataset. For instance, the agent can utilize a large language model to generate a response to the user query. The input to the large language model can include a prompt including a context window. The context window can include relevant documents associated with the updated dataset. Byway of example, processing logic can determine a context window including a number of documents that can be used to generate the second response. Processing logic can select, by the second agent, responsive to removing the first document from the consideration for processing the user query, a replacement document for utilization in the context window for generating the second response.

[0092] At operation 610, processing logic can transmit the second response to the first agent. In some instances, processing logic can additionally transmit one or more hashed document identifiers such that the system can store which documents were considered by the agents when generating the composite response.

[0093] In some implementations, processing logic can generate updated hashed document data comprising a combined set of the plurality of hashed identifiers obtained from the first agent combined with one or more hashed identifiers of documents utilized by the second agent in generating the second response. Processing logic can transmit, by the second agent, data including the user query and updated hashed document data to a third agent. Processing logic can receive, by the second agent from the third agent, a third response to the user query. The third response to the user query was generated by the third agent excluding the one or more documents associated with the updated hashed document data. Processing logic can process, by the second agent, the second response and the third response to generate a composite response. Processing logic can transmit, by the second agent, the composite response to the first agent.

[0094] In some embodiments, processing logic can receive, by the second agent from the third agent, a third set of hashed identifiers used to generate the third response. Processinglogic can generate, by the second agent, updated hashed document data comprising the third set of hashed identifiers and the second set of hashed identifiers. Processing logic can transmit, from the second agent to the first agent, updated hashed document data.

[0095] In some implementations, processing logic can transmit a request for a cryptographic hash understood by a third agent. Processing logic can receive a list of hash functions understood by the third agent. Processing logic can select a first hash function of the list of hash functions. Processing logic can apply the first hash function to data associated with the first document.

[0096] Figure 7 depicts a flow diagram of an example method 700 to perform privately sharing known information between agents according to example embodiments of the present disclosure. The method 700 can be performed by processing logic that can include hardware (e.g., processing device, circuitry, dedicated logic, programmable logic, microcode, hardware of a device, integrated circuit, etc.), software (e.g., instructions run or executed on a processing device), or a combination thereof. In some embodiments, method 700 is performed by a server computing system (e.g., server computing system 60) or client computing system (e.g., client computing device 50). Although shown in a particular sequence or order, unless otherwise specified, the order of the processes can be modified. Thus, the illustrated embodiments should be understood only as examples, and the illustrated processes can be performed in a different order, and some processes can be performed in parallel. Additionally, one or more processors can be omitted in various embodiments. Thus, not all processes are required in every embodiment. Other process flows are possible.

[0097] At operation 702, processing logic can generate an initial response to a user query based on one or more documents accessible by a first agent.

[0098] At operation 704, processing logic can transmit the user query and hashed document data to one or more second agents.

[0099] At operation 706, processing logic can access a document index. The document index can include a plurality of hashed identifiers of documents accessible by the second agent.

[0100] At operation 708, processing logic can exclude a first document based on the hashed identifier associated with the first document being in the hashed document data and the document index.

[0101] At operation 710, processing logic can generate a second response based on one or more remaining documents accessible by the respective second agent.

[0102] In some implementations, processing logic can transmit the user query and hashed document data to a plurality of remote agents until the first agent has received a threshold number of unique hashed identifiers indicating a number of unique documents considered by the plurality of remote agents exceeds the threshold. Processing logic can collate responses generated by the remote agents to generate a composite response.

[0103] In some implementations, processing logic can transmit the user query and hashed document data to the plurality of agents in parallel. As such, the secondary agents can consider the hashed document data obtained from the initial agent, however they are unable to disregard documents that are considered by the other secondary agents.

[0104] In some implementations, processing logic can transmit the user query and the hashed document data to the plurality of agents sequentially. As such, the secondary agents can be run one-at-a-time such that each secondary agent obtains the initial hashed identifiers as well as hashed identifiers of any documents considered by any preceding secondary agent. As such, processing logic can disregard such documents as it generates various responses which can then be collated by the initial agent to be relayed to a user interface of a client device.

[0105] Figure 8 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 large language model such as a transformer or diffusion model associated with an agent or mobile application.

[0106] 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 8 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 8 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.

[0107] 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 trained using 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.

[0108] 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.

[0109] 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).

[0110] 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 backpropagated from the output (or another source of the evaluation signal) through the machine-learned model(s) to update one or more parameters of the model(s) (e.g., based on a gradient of the evaluation signal with respect to the parameter value(s)). For example, system(s) containing one or more machine-learned models can be trained in an end-to-end manner. Gradient descent techniques can be used to iteratively update the parameters over a number of training iterations. In some implementations, performing backwards propagation of errors can include performing truncated backpropagation through time. Example method800 can include implementing a number of generalization techniques (e.g., weight decays, dropouts, etc.) to improve the generalization capability of the models being trained.

[0111] In some implementations, example method 800 can be implemented for training a machine-learned model from an initialized state to a fully trained state (e.g., when the model exhibits a desired performance profile, such as based on accuracy, precision, recall, etc.).

[0112] 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.

[0113] In some implementations, example method 800 can be implemented for finetuning 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)). In some implementations, example method 800 uses adapter modules. Adapters can be small trainable layers that are inserted between pre-existing layers of a pre-trained model. During the finetuning process, the original parameters of the pre-trained model are typically frozen, and only the parameters of the adapters are updated.

[0114] In some implementations, example method 800 can be implemented to execute parameter-efficient fine-tuning methods, such as Layerwise Optimization of Residuals (LoRA). LoRA can refine pre-trained models with minimal adjustments to the original parameters. This can be achieved by introducing trainable low-rank matrices that modify the behavior of the pre-trained weights without directly altering them. In some implementations, during fine-tuning, only these auxiliary matrices are updated, which significantly reduces the number of parameters that are trained.

[0115] An example fine-tuning approach includes reinforcement learning. Reinforcement learning can be based on user feedback on model performance during use.

[0116] Figure 9 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.

[0117] 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 nonlinear 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.

[0118] Machine-learned model(s) 1 can be or include, or otherwise be representative of any one or more of the machine-learned models described above with respect to the preceding figures. For example, machine-learned model(s) 1 can be or include, or otherwise be representative of any one or more of can include a large language model such as a transformer or diffusion model associated with an agent or mobile application, etc. Although various features, variations, and implementations described below are described with respect to machine-learned model(s) 1, it is to be understood that such features, variations, and implementations are to be understood as described with respect to each of can include a large language model such as a transformer or diffusion model associated with an agent or mobile application, etc., any other machine-learned component described herein.

[0119] 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.

[0120] 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 multiple different models or multiple different model portions configured to operate on data from input(s) 2.

[0121] Machine-learned model(s) 1 can include an ensemble of different models that can cooperatively interact to process data from input(s) 2. For example, a model ensemble can include multiple models that have different attributes (e.g., different architectures, trained with different recipes, etc.). The ensemble can output an overall output based on the individual outputs of the constituent models. In this manner, for instance, the diverse constituent models can work together to provide system-level robustness by effectivelyaggregating over individual strengths and weaknesses of any given model. The respective individual outputs can be combined in a weighted combination, using a voting or routing mechanism, or a learned output layer (e.g., one or more feedforward or fully-connected layers).

[0122] Machine-learned model(s) 1 can employ a mixture-of-experts structure. See, e.g., Zhou et al., Mixture -of -Experts with Expert Choice Routing, ARXIV:2202.09368V2 (Oct. 14, 2022). For example, different portions of a model can learn (explicitly or implicitly) different expertise areas, with pathways through the model being selected by a learned routing mechanism that engages the appropriate expert for a given input (e.g., a given portion of an input, such as on a per-token basis). For example, a feedforward network can be sparsely activated for a given portion of an input based on an output of a routing mechanism that processes the portion of the input. In this manner, for instance, the group of activated weights can form an “expert” that is selected by the router. On each forward pass, only a subset of the total model weights may be engaged, thereby decreasing a quantity of operations performed for processing a given input compared to a densely activated model. In this manner, for instance, the expressive and interpretive power of a high-parameter-count model can be achieved with more compute-efficient forward passes.

[0123] 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.

[0124] 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.

[0125] 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.

[0126] 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.

[0127] Figure 10 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-A , etc. obtained from input(s) 2. Sequence processing model 4 can process input sequence 5 using prediction layer(s) 6 to generate an output sequence 7. Output sequence 7 can include one or more output elements 7-1, 7-2, . . . , 7 -A, etc. generated based on input sequence 5. The system can generate output(s) 3 based on output sequence 7.

[0128] Sequence processing model(s) 4 can include one or multiple machine-learned model components configured to ingest, generate, or otherwise reason over sequences of information. For example, some example sequence processing models in the text domain are referred to as “Large Language Models,” or LLMs. See, e.g., PaLM 2 Technical Report, GOOGLE, https: / / ai.google / static / documents / palm2techreport.pdf (n.d.). Other example sequence processing models can operate in other domains, such as image domains, see, e.g., Dosovitskiy et al., An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale, ARXIV:2010.11929v2 (Jun. 3, 2021), audio domains, see, e.g., Agostinelli et al., MusicLM: Generating Music From Text, ARXIV:2301.1 1325V1 (Jan. 26, 2023), biochemical domains, see, e.g., Jumper et al., Highly accurate protein structure prediction with AlphaFold, 596 Nature 583 (Aug. 26, 2021), by way of example. Sequence processing model(s) 4 can process one or multiple types of data simultaneously. Sequence processing model(s) 4 caninclude relatively large models (e.g., more parameters, computationally expensive, etc.), relatively small models (e.g., fewer parameters, computationally lightweight, etc.), or both.

[0129] 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”).

[0130] 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.

[0131] 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.

[0132] For example, elements 5-1, 5-2, . . . , 5-M can represent tokens obtained using a tokenizer. For instance, a tokenizer can process a given portion of an input source and output a series of tokens (e.g., corresponding to input elements 5-1, 5-2, . . . , 5-A ) that represent the portion of the input source. Various approaches to tokenization can be used. For instance, textual input source(s) can be tokenized using a byte-pair encoding (BPE) technique. See, e.g., Kudo et al., SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processing, PROCEEDINGS OF THE 2018 CONFERENCE ON EMPIRICAL METHODS IN NATURAL LANGUAGE PROCESSING (System Demonstrations), pages 66-71 (October 31-November 4, 2018), https: / / aclanthology.org / D18-2012.pdf. Image-based input source(s) can be tokenized by extracting and serializing patches from an image.

[0133] 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 10 can be the tokens or can be the embedded representations thereof.

[0134] 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-learnedmodel 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.

[0135] 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.”

[0136] A transformer is an example architecture that can be used in prediction layer(s) 4. See, e.g., Vaswani et al., Atention Is All You Need, ARXIV: 1706.03762v7 (Aug. 2, 2023). A transformer is an example of a machine-learned model architecture that uses an attention mechanism to compute associations between items within a context window. The context window can include a sequence that contains input sequence 5 and potentially one or more output element(s) 7-1, 7-2, . . . , 7-N. A transformer block can include one or more attention layer(s) and one or more post-attention layer(s) (e.g., feedforward layer(s), such as a multi-layer perceptron).

[0137] 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.

[0138] 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.

[0139] 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.

[0140] Output sequence 7 can be generated autoregressively. For instance, for some applications, an output of one or more prediction layer(s) 6 can be passed through one or more output layers (e.g., softmax layer) to obtain a probability distribution over an output vocabulary (e.g., a textual or symbolic vocabulary) conditioned on a set of input elements in a context window. In this manner, for instance, output sequence 7 can be autoregressively generated by sampling a likely next output element, adding that element to the context window, and re-generating the probability distribution based on the updated context window, and sampling a likely next output element, and so forth.

[0141] Output sequence 7 can also be generated non-autoregressively. For instance, multiple output elements of output sequence 7 can be predicted together without explicit sequential conditioning on each other. See, e.g., Saharia et al., Non-Autoregressive Machine Translation with Latent Alignments, ARXIV:2004.07437V3 (NOV. 16, 2020).

[0142] 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.

[0143] Figure 11 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 canprocess 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.

[0144] 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 / Jdimensions. 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.

[0145] 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.

[0146] 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.

[0147] Task indicator 9 can include a model or model component configured to identify a task being performed and inject, into input sequence 8, an input value represented by element 8-0 that signals which task is being performed. For instance, the input value can be provided as a data type associated with an input modality and projected along with that input modality (e.g., the input value can be a textual task label that is embedded along with other textual data in the input; the input value can be a pixel-based representation of a task that is embedded along with other image data in the input; etc.). The input value can be provided as a data type that differs from or is at least independent from other input(s). For instance, the input value represented by element 8-0 can be a learned within a continuous embedding space.

[0148] 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).

[0149] Data-to-sequence models 11-1, 11-2, and 11-3 can be the same or different from each other. Data-to-sequence models 11-1, 11-2, and 11-3 can be adapted to each respective input modality 10-1, 10-2, and 10-3. For example, a textual data-to-sequence model can subdivide a portion of input text and project the subdivisions into element(s) in input sequence 8 (e.g., elements 8-1, 8-2, 8-3, etc.). An image data-to-sequence model can subdivide an input image and project the subdivisions into element(s) in input sequence 8 (e.g., elements 8-4, 8-5, 8-6, etc.). An arbitrary datatype data-to-sequence model can subdivide an input of that arbitrary datatype and project the subdivisions into element(s) in input sequence 8 (e.g., elements 8-7, 8-8, 8-9, etc.).

[0150] 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.

[0151] Figure 12 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 developmentplatform 12 can provide a number of different toolkits that developer systems can employ in the development of new or adapted machine-learned models.

[0152] 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. Model primitives 13-3 can include a library of pre-trained adapters or LoRA modules that can adapt a baseline foundational model to align its outputs with a desired performance profile, augment model capabilities (e.g., to adapt to a different input modality, etc.), and the like.

[0153] 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.

[0154] 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.

[0155] Model alignment toolkit 17 can provide a number of tools for causing development model 16 to generate outputs aligned with desired behavioral characteristics. Alignment can include increasing an accuracy, precision, recall, etc. of model outputs. Alignment can include enforcing output styles, schema, or other preferential characteristics of model outputs. Alignment can be general or domain-specific. For instance, a pre-trained foundational model 13-1 can begin with an initial level of performance across multiple domains. Alignment of the pre-trained foundational model 13-1 can include improving a performance in a particular domain of information or tasks (e.g., even at the expense of performance in another domain of information or tasks).

[0156] 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.

[0157] 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.

[0158] 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 finetune development model 16.

[0159] 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.

[0160] 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.

[0161] 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).

[0162] Prompt libraries 17-4 can include one or more prompt engineering tools. Prompt engineering tools can provide workflows for retrieving or learning optimized prompt values. Prompt engineering tools can facilitate directly learning prompt values (e.g., input element values) based one or more training iterations. Workbench 15 can implement prompt engineering tools in development model 16.

[0163] 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 atask and output a 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.

[0164] 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.

[0165] 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.

[0166] 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.

[0167] 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”).

[0168] 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.

[0169] Model plugin toolkit 18 can include interfaces for calling external application programming interfaces (APIs) 18-3. For instance, in addition to or in lieu of implementing tool calls or tool code directly with development model 16, development model 16 can be aligned to output instruction that initiate API calls to send or obtain data via external systems.

[0170] 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.

[0171] 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 of development model 16 can be efficiently transferred to a smaller model for more efficient inference.

[0172] 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 ofdevelopment 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.

[0173] Figure 13 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.

[0174] 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.

[0175] 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).

[0176] Pre-trained model 23 can then be a new version of development model 16, which can persist as development model 16 or as a new development model. Pre-trained model 23 can be the initial state if development model 16 was already pre-trained. Pre-trained model 23 can undergo fine-tuning in a fine-tuning stage 24. Fine-tuning stage 24 can be implemented using one or more fine-tuning pipelines 17-3 over data from dataset(s) 17-1.Fine-tuning can be omitted, for example, if a pre-trained model as satisfactory performance, if the model was already fine-tuned, or if other tuning approaches are preferred.

[0177] Fine-tuned model 29 can then be a new version of development model 16, which can persist as development model 16 or as a new development model. Fine-tuned model 29 can be the initial state if development model 16 was already fine-tuned. Fine-tuned model 29 can undergo refinement with user feedback 26. For instance, refinement with user feedback 26 can include reinforcement learning, optionally based on human feedback from human users of fine-tuned 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.

[0178] In some implementations, computational optimization operations can be applied before, during, or after each stage. For instance, initialized model 21 can undergo computational optimization 29-1 (e.g., using computational optimization toolkit 19) before pre-training stage 22. Pre-trained model 23 can undergo computational optimization 29-2 (e.g., using computational optimization toolkit 19) before fine-tuning stage 24. Fine-tuned model 25 can undergo computational optimization 29-3 (e.g., using computational optimization toolkit 19) before refinement with user feedback 26. Refined model 27 can undergo computational optimization 29-4 (e.g., using computational optimization toolkit 19) before output 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.

[0179] Figure 14 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.

[0180] Model host 31 can perform inference on behalf of one or more client(s) 32. Client(s) 32 can transmit an input request 33 to model host 31. Using input request 33, model host 31 can obtain input(s) 2 for input to machine-learned model(s) 1. Machine-learned model(s) 1 can process input(s) 2 to generate output(s) 3. Using output(s) 3, model host 31 can return an output payload 34 for responding to input request 33 from client(s) 32. Output payload 34 can include or be based on output(s) 3.

[0181] Model host 31 can leverage various other resources and tools to augment the inference task. For instance, model host 31 can communicate with tool interfaces 35 to facilitate tool use by model instance(s) 31-1. Tool interfaces 35 can include local or remote APIs. Tool interfaces 35 can include integrated scripts or other software functionality. Model host 31 can engage online learning interface(s) 36 to facilitate ongoing improvements to machine-learned model(s) 1. For instance, online learning interface(s) 36 can be used within reinforcement learning loops to retrieve user feedback on inferences served by model host 31. Model host 31 can access runtime data source(s) 37 for augmenting input(s) 2 with additional contextual information. For instance, runtime data source(s) 37 can include a knowledge graph 37-1 that facilitates structured information retrieval for information associated with input request(s) 33 (e.g., a search engine service). Runtime data source(s) 37 can include public or private, external or local database(s) 37-2 that can store information associated with input request(s) 33 for augmenting input(s) 2. Runtime data source(s) 37 can include account data 37-3 which can be retrieved in association with a user account corresponding to a client 32 for customizing the behavior of model host 31 accordingly.

[0182] 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.

[0183] 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.

[0184] In some implementations, model host 31 can operate on a same device or system as client(s) 32. Model host 31 can be a machine-learning service that runs on-device to provide machine-learning functionality to one or multiple applications operating on a client device, which can include an application implementing client(s) 32. Model host 31 can be a part of a same application as client(s) 32. For instance, model host 31 can be a subroutine or method implemented by one part of an application, and client(s) 32 can be another subroutine or method that engages model host 31 to perform inference functions within the application. It is to be understood that model host 31 and client(s) 32 can have various different configurations.

[0185] 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 othermodel components that are stored on in persistent storage, temporarily cached, or loaded into high-speed 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.

[0186] 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.

[0187] 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.

[0188] Model host 31 can perform inference over batches of input requests 33 in parallel. For instance, a model instance 31-1 can be configured with an input structure that has a batch dimension. Separate input(s) 2 can be distributed across the batch dimension (e.g., rows of an array). The separate input(s) 2 can include completely different contexts. The separate input(s) 2 can be multiple inference steps of the same task. The separate input(s) 2 can be staggered in an input structure, such that any given inference cycle can be operating on different portions of the respective input(s) 2. In this manner, for instance, model host 31 can perform inference on the batch in parallel, such that output(s) 3 can also contain the batch dimension and return the inference results for the batched input(s) 2 in parallel. In this manner, for instance, batches of input request(s) 33 can be processed in parallel for higher throughput of output payload(s) 34.

[0189] 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.

[0190] 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.

[0191] Model host 31 can access a library of pre-trained adapters or LoRA modules that can adapt a baseline model to align its outputs with a desired performance profile, augment model capabilities (e.g., to adapt to a different input modality, etc.), and the like. For instance, model host 31 can receive an input request to load a customized model, and model host 31 can retrieve one or more components to adapt a baseline model to the custom profile. Model host 31 can determine that a particular functionality is needed for a particular task (e.g., based on an output of a model that preprocesses an input) and retrieve a pre-trained component accordingly.

[0192] 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.

[0193] In some implementations, the task is a computer vision task. In some cases, input(s) 2 includes pixel data for one or more images and the task is an image processing task. For example, the image processing task can be image classification, where the output is a set of scores, each score corresponding to a different object class and representing the likelihood that the one or more images depict an object belonging to the object class. The image processing task may be object detection, where the image processing output identifies one or more regions in the one or more images and, for each region, a likelihood that region depicts an object of interest. As another example, the image processing task can be image segmentation, where the image processing output defines, for each pixel in the one or more images, a respective likelihood for each category in a predetermined set of categories. For example, the set of categories can be foreground and background. As another example, the set of categories can be object classes. As another example, the image processing task can be depth estimation, where the image processing output defines, for each pixel in the one or more images, a respective depth value. As another example, the image processing task can be motion estimation, where the network input includes multiple images, and the image processing output defines, for each pixel of one of the input images, a motion of the scene depicted at the pixel between the images in the network input.

[0194] 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).

[0195] In some implementations, input(s) 2 can be or otherwise represent speech data (e.g., data describing spoken natural language, such as audio data, textual data, etc.). Machine-learned model(s) 1 can process the speech data to generate an output. As an example, machine-learned model(s) 1 can process the speech data to generate a speech recognition output. As another example, machine-learned model(s) 1 can process the speech data to generate a speech translation output. As another example, machine-learned model(s) 1 can process the speech data to generate a latent embedding output. As another example, machine-learned model(s) 1 can process the speech data to generate an encoded speech output (e.g., an encoded and / or compressed representation of the speech data, etc.). As another example, machine-learned model(s) 1 can process the speech data to generate an upscaled speech output (e.g., speech data that is higher quality than the input speech data, etc.). As another example, machine-learned model(s) 1 can process the speech data to generate a textual representation output (e.g., a textual representation of the input speech data, etc.). As another example, machine-learned model(s) 1 can process the speech data to generate a prediction output.

[0196] 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.

[0197] 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 thestatistical data to generate a visualization output. As another example, machine-learned model(s) 1 can process the statistical data to generate a diagnostic output.

[0198] In some implementations, input(s) 2 can be or otherwise represent sensor data. Machine-learned model(s) 1 can process the sensor data to generate an output. As an example, machine-learned model(s) 1 can process the sensor data to generate a recognition output. As another example, machine-learned model(s) 1 can process the sensor data to generate a prediction output. As another example, machine-learned model(s) 1 can process the sensor data to generate a classification output. As another example, machine-learned model(s) 1 can process the sensor data to generate a segmentation output. As another example, machine-learned model(s) 1 can process the sensor data to generate a visualization output. As another example, machine-learned model(s) 1 can process the sensor data to generate a diagnostic output. As another example, machine-learned model(s) 1 can process the sensor data to generate a detection output.

[0199] 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.

[0200] 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.

[0201] 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 togenerate output(s) 3 to complete a sentence, paragraph, or portion of text that follows from a portion of text represented by input(s) 2.

[0202] In some implementations, the task can be an instruction following task. Machine-learned model(s) 1 can be configured to process input(s) 2 that represent instructions to perform a function and to generate output(s) 3 that advance a goal of satisfying the instruction function (e.g., at least a step of a multi-step procedure to perform the function). Output(s) 3 can represent data of the same or of a different modality as input(s) 2. For instance, input(s) 2 can represent textual data (e.g., natural language instructions for a task to be performed) and machine-learned model(s) 1 can process input(s) 2 to generate output(s) 3 that represent textual data responsive to the instructions (e.g., natural language responses, programming language responses, machine language responses, etc.). Input(s) 2 can represent image data (e.g., image-based instructions for a task to be performed, optionally accompanied by textual instructions) and machine-learned model(s) 1 can process input(s) 2 to generate output(s) 3 that represent textual data responsive to the instructions (e.g., natural language responses, programming language responses, machine language responses, etc.). One or more output(s) 3 can be iteratively or recursively generated 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.

[0203] 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 stepstoward 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.

[0204] In some implementations, the task can be an image generation task. Machine- learned model(s) 1 can be configured to process input(s) 2 that represent context regarding a desired portion of image content. The context can include text data, image data, audio data, etc. Machine-learned model(s) 1 can be configured to generate output(s) 3 that represent image data that depicts imagery related to the context. For instance, machine-learned model(s) 1 can be configured to generate pixel data of an image. Values for channel(s) associated with the pixels in the pixel data can be selected based on the context (e.g., based on a probability determined based on the context).

[0205] 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).

[0206] 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).

[0207] Figure 15 is a block diagram of an example networked computing system that can perform aspects of example implementations of the present disclosure. The system caninclude a number of computing devices and systems that are communicatively coupled over a network 49. An example computing device 50 is described to provide an example of a computing device that can perform any aspect of the present disclosure (e.g., implementing model host 31, client(s) 32, or both). An example server computing system 60 is described as an example of a server computing system that can perform any aspect of the present disclosure (e.g., implementing model host 31, client(s) 32, or both). Computing device 50 and server computing system(s) 60 can cooperatively interact (e.g., over network 49) to perform any aspect of the present disclosure (e.g., implementing model host 31, client(s) 32, or both). Model development platform system 70 is an example system that can host or serve model development platform(s) 12 for development of machine-learned models. Third-party system(s) 80 are example system(s) with which any of computing device 50, server computing system(s) 60, or model development platform system(s) 70 can interact in the performance of various aspects of the present disclosure (e.g., engaging third-party tools, accessing third-party databases or other resources, etc.).

[0208] Network 49 can be any type of communications network, such as a local area network (e.g., intranet), wide area network (e.g., Internet), or some combination thereof and can include any number of wired or wireless links. In general, communication over network 49 can be carried via any type of wired or wireless connection, using a wide variety of communication protocols (e.g., TCP / IP, HTTP, SMTP, FTP), encodings or formats (e.g., HTML, XML), or protection schemes (e.g., VPN, secure HTTP, SSL). Network 49 can also be implemented via a system bus. For instance, one or more devices or systems of Figure 15 can be co-located with, contained by, or otherwise integrated into one or more other devices or systems.

[0209] Computing device 50 can be any type of computing device, such as, for example, a personal computing device (e.g., laptop or desktop), a mobile computing device (e.g., smartphone or tablet), a gaming console or controller, a wearable computing device, an embedded computing device, a server computing device, a virtual machine operating on a host device, or any other type of computing device. Computing device 50 can be a client computing device. Computing device 50 can be an end-user computing device. Computing device 50 can be a computing device of a service provided that provides a service to an end user (who may use another computing device to interact with computing device 50).

[0210] Computing device 50 can include one or more processors 51 and a memory 52. Processor(s) 51 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and can be oneprocessor or a plurality of processors that are operatively connected. Memory 52 can include one or more non-transitory computer-readable storage media, such as HBM, RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. Memory 52 can store data 53 and instructions 54 which can be executed by processor(s) 51 to cause computing device 50 to perform operations. The operations can implement any one or multiple features described herein. The operations can implement example methods and techniques described herein.

[0211] Computing device 50 can also include one or more input components that receive user input. For example, a user input component can be a touch-sensitive component (e.g., 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, camera, LIDAR, a physical keyboard or other buttons, or other means by which a user can provide user input.

[0212] Computing device 50 can store or include one or more machine-learned models 55. Machine-learned models 55 can include one or more machine-learned model(s) 1, such as a sequence processing model 4. Machine-learned models 55 can include one or multiple model instance(s) 31-1. Machine-learned model(s) 55 can be received from server computing system(s) 60, model development platform system 70, third party system(s) 80 (e.g., an application distribution platform), or developed locally on computing device 50. Machine-learned model(s) 55 can be loaded into memory 52 and used or otherwise implemented by processor(s) 51. Computing device 50 can implement multiple parallel instances of machine-learned model(s) 55.

[0213] Server computing system(s) 60 can include one or more processors 61 and a memory 62. Processor(s) 61 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. Memory 62 can include one or more non-transitory computer-readable storage media, such as HBM, RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. Memory 62 can store data 63 and instructions 64 which can be executed by processor(s) 61 to cause server computing system(s) 60 to perform operations. The operations can implement any one or multiple features described herein. The operations can implement example methods and techniques described herein.

[0214] In some implementations, server computing system 60 includes or is otherwise implemented by one or multiple server computing devices. In instances in which server computing system 60 includes multiple server computing devices, such server computing devices can operate according to sequential computing architectures, parallel computing architectures, or some combination thereof.

[0215] Server computing system 60 can store or otherwise include one or more machine-learned models 65. Machine-learned model(s) 65 can be the same as or different from machine-learned model(s) 55. Machine-learned models 65 can include one or more machine-learned model(s) 1, such as a sequence processing model 4. Machine-learned models 65 can include one or multiple model instance(s) 31-1. Machine-learned model(s) 65 can be received from computing device 50, model development platform system 70, third party system(s) 80, or developed locally on server computing system(s) 60. Machine-learned model(s) 65 can be loaded into memory 62 and used or otherwise implemented by processor(s) 61. Server computing system(s) 60 can implement multiple parallel instances of machine-learned model(s) 65.

[0216] In an example configuration, machine-learned models 65 can be included in or otherwise stored and implemented by server computing system 60 to establish a client-server relationship with computing device 50 for serving model inferences. For instance, server computing system(s) 60 can implement model host 31 on behalf of client(s) 32 on computing device 50. For instance, machine-learned models 65 can be implemented by server computing system 60 as a portion of a web service (e.g., remote machine-learned model hosting service, such as an online interface for performing machine-learned model operations over a network on server computing system(s) 60). For instance, server computing system(s) 60 can communicate with computing device 50 over a local intranet or internet connection. For instance, computing device 50 can be a workstation or endpoint in communication with server computing system(s) 60, with implementation of machine-learned models 65 being managed by server computing system(s) 60 to remotely perform inference (e.g., for runtime or training operations), with output(s) returned (e.g., cast, streamed, etc.) to computing device 50. Machine-learned models 65 can work cooperatively or interoperatively with machine- learned models 55 on computing device 50 to perform various tasks.

[0217] Model development platform system(s) 70 can include one or more processors 71 and a memory 72. Processor(s) 71 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. Memory 72 caninclude one or more non-transitory computer-readable storage media, such as HBM, RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. Memory 72 can store data 73 and instructions 74 which can be executed by processor(s) 71 to cause model development platform system(s) 70 to perform operations. The operations can implement any one or multiple features described herein. The operations can implement example methods and techniques described herein. Example operations include the functionality described herein with respect to model development platform 12. This and other functionality can be implemented by developer tool(s) 75.

[0218] Third-party system(s) 80 can include one or more processors 81 and a memory 82. Processor(s) 81 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. Memory 82 can include one or more non-transitory computer-readable storage media, such as HBM, RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. Memory 82 can store data 83 and instructions 84 which can be executed by processor(s) 81 to cause third-party system(s) 80 to perform operations. The operations can implement any one or multiple features described herein. The operations can implement example methods and techniques described herein. Example operations include the functionality described herein with respect to tools and other external resources called when training or performing inference with machine-learned model(s) 1, 4, 16, 20, 55, 65, etc. (e.g., third-party resource(s) 85).

[0219] Figure 15illustrates one example arrangement of computing systems that can be used to implement the present disclosure. Other computing system configurations can be used as well. For example, in some implementations, one or both of computing system 50 or server computing system(s) 60 can implement all or a portion of the operations of model development platform system 70. For example, computing system 50 or server computing system(s) 60 can implement developer tool(s) 75 (or extensions thereof) to develop, update / train, or refine machine-learned models 1, 4, 16, 20, 55, 65, etc. using one or more techniques described herein with respect to model alignment toolkit 17. In this manner, for instance, computing system 50 or server computing system(s) 60 can develop, update / train, or refine machine-learned models based on local datasets (e.g., for model personalization / customization, as permitted by user data preference selections).

[0220] Figure 16 is a block diagram of an example computing device 98 that performs according to example embodiments of the present disclosure. Computing device 98can be a user computing device or a server computing device (e.g., computing device 50, server computing system(s) 60, etc.). Computing device 98 can implement model host 31. For instance, computing device 98 can include a number of applications (e.g., applications 1 through N). Each application can contain its own machine learning library and machine- learned model(s). 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, etc. As illustrated in Figure 16, each application can communicate with a number of other components of the computing device, such as, for example, one or more sensors, a context manager, a device state component, or additional components. In some implementations, each application can communicate with each device component using an API (e.g., a public API). In some implementations, the API used by each application is specific to that application.

[0221] Figure 17 is a block diagram of an example computing device 99 that performs according to example embodiments of the present disclosure. Computing device 99 can be the same as or different from computing device 98. Computing device 99 can be a user computing device or a server computing device (e.g., computing device 50, server computing system(s) 60, etc.). Computing device 98 can implement model host 31. For instance, computing device 99 can include a number of applications (e.g., applications 1 through N). Each application can be in communication with a central intelligence layer. Example applications include a text messaging application, an email application, a dictation application, a virtual keyboard application, a browser application, 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).

[0222] The central intelligence layer can include a number of machine-learned models. For example, as illustrated in Figure 17, a respective machine-learned model can be provided for each application and managed by the central intelligence layer. In other implementations, two or more applications can share a single machine-learned model. For example, in some implementations, the central intelligence layer can provide a single model for all of the applications. In some implementations, the central intelligence layer is included within or otherwise implemented by an operating system of computing device 99.

[0223] The central intelligence layer can communicate with a central device data layer. The central device data layer can be a centralized repository of data for computing device 99. As illustrated in Figure 17, the central device data layer can communicate with a number of other components of the computing device, such as, for example, one or moresensors, a context manager, a device state component, or additional components. In some implementations, the central device data layer can communicate with each device component using an API (e.g., a private API).

[0224] 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.

[0225] 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 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 cover such alterations, variations, and equivalents.

[0226] Aspects of the disclosure have been described in terms of illustrative embodiments thereof. Any and all features in the following claims can be combined or rearranged in any way possible, including combinations of claims not explicitly enumerated in combination together, as the example claim dependencies listed herein should not be read as limiting the scope of possible combinations of features disclosed herein. Accordingly, the scope of the present disclosure is by way of example rather than by way of limitation, and the subject disclosure does not preclude inclusion of such modifications, variations or additions to the present subject matter as would be readily apparent to one of ordinary skill in the art. Moreover, terms are described herein using lists of example elements joined by conjunctions such as “and,” “or,” “but,” etc. It should be understood that such conjunctions are provided for explanatory purposes only. Clauses and other sequences of items joined by a particular conjunction such as “or,” for example, can refer to “and / or,” “at least one of’, “anycombination of’ example elements listed therein, etc. Terms such as “based on” should be understood as “based at least in part on.”

[0227] The term “can” should be understood as referring to a possibility of a feature in various implementations and not as prescribing an ability that is necessarily present in every implementation. For example, the phrase “X can perform Y” should be understood as indicating that, in various implementations, X has the potential to be configured to perform Y, and not as indicating that in every instance X must always be able to perform Y. It should be understood that, in various implementations, X might be unable to perform Y and remain within the scope of the present disclosure.

[0228] The term “may” should be understood as referring to a possibility of a feature in various implementations and not as prescribing an ability that is necessarily present in every implementation. For example, the phrase “X may perform Y” should be understood as indicating that, in various implementations, X has the potential to be configured to perform Y, and not as indicating that in every instance X must always be able to perform Y. It should be understood that, in various implementations, X might be unable to perform Y and remain within the scope of the present disclosure.

Claims

WHAT IS CLAIMED IS:

1. A computer-implemented method for privately sharing known information between agents comprising: obtaining, by a first agent on a client device, a user query; processing, by the first agent, the user query to perform an initial inference based on one or more documents accessible by the client device; generating, by the first agent, an initial response to the user query; transmitting, by the first agent, data comprising the user query and hashed document data to a second agent, wherein the hashed document data comprises at least one identifier associated with a first document of the one or more documents used to generate the initial response; receiving, by the first agent from the second agent, a second response to the user query, wherein the second response to the user query was generated by the second agent excluding the one or more documents used to generate the initial response; processing, by the first agent, the initial response and the second response to generate a composite response; and providing, by the first agent, the composite response to a user.

2. The computer-implemented method of claim 1, comprising: transmitting, by the first agent, a request for a cryptographic hash understood by a second agent; receiving, by the first agent from the second agent, a list of hash functions understood by the second agent; selecting, by the first agent, a first hash function of the list of hash functions; and applying, by the first agent, the first hash function to data associated with the first document.

3. The computer-implemented method of any preceding claim, wherein generating, by the first agent, an initial response to the user query comprises performing retrieval -augmented generation based on the one or more documents accessible by the client device.

4. The computer-implemented method of any preceding claim, wherein the data associated with the first document comprises at least one of a title of the first document or a uniform resource locator (URL) associated with the first document.

5. The computer-implemented method of any preceding claim, comprising receiving, by the first agent from the second agent, second hashed document data, wherein the second hashed document data comprises one or more hashed document identifiers of documents used by the second agent in generating the second response to the user query.

6. The computer-implemented method of any preceding claim, comprising: transmitting, by the first agent, data comprising the user query and the hashed document data to a third agent; receiving, by the first agent from the third agent, a third response to the user query, wherein the third response to the user query was generated by the third agent excluding the one or more documents used to generate the initial response; and processing, by the first agent, the initial response, the second response, and the third response to update the composite response.

7. The computer-implemented method of any preceding claim, comprising: transmitting, by the first agent, data comprising the user query and the hashed document data to a third agent; receiving, by the first agent from the third agent, a third response to the user query, wherein the third response to the user query was generated by the third agent excluding the one or more documents used to generate the initial response and excluding the one or more documents used to generate the second response; and processing, by the first agent, the initial response, the second response, and the third response to update the composite response.

8. The computer-implemented method of any preceding claim, comprising: receiving, by the first agent, data comprising second hashed document data, wherein the second hashed document data comprises at least one identifier associated with a first document of one or more documents used to generate the second response.

9. The computer-implemented method of any preceding claim, comprising:transmitting, by the first agent, data comprising the user query and the hashed document data to a third agent; receiving, by the first agent from the third agent, a third response to the user query, wherein the third response to the user query comprises a message that no additional relevant documents are accessible by the third agent.

10. The computer-implemented method of any preceding claim, wherein the first agent comprises a large language model (LLM).

11. A computer-implemented method comprising: accessing, by a second agent, a dataset comprising a plurality of hashed identifiers, wherein each respective hashed identifier of the plurality of hashed identifiers is associated with a document that the second agent has determined is relevant to a user query; determining, by the second agent, that a first hashed identifier associated with a first document is present in the dataset of hashed identifiers; removing, by the second agent, the first document from a context window for processing the user query based on determining that the first document was considered by a first agent computing system to generate an updated dataset; generating, by the second agent, a second response to the user query based on documents associated with the updated dataset; and transmitting, by the second agent, the second response to the first agent.

12. The computer-implemented method of claim 10, comprising: determining, by the second agent, a context window comprising a number of documents that can be used to generate the second response; and selecting, by the second agent, responsive to removing the first document from consideration for processing the user query, a replacement document for utilization in the context window for generating the second response.

13. The computer-implemented method of claim 11, comprising: generating, by the second agent, updated hashed document data comprising a combined set of the plurality of hashed identifiers obtained from the first agent combined with one or more hashed identifiers of documents utilized by the second agent in generating the second response;transmitting, by the second agent, data comprising the user query and updated hashed document data to a third agent; receiving, by the second agent from the third agent, a third response to the user query, wherein the third response to the user query was generated by the third agent excluding the one or more documents associated with the updated hashed document data; processing, by the second agent, the second response and the third response to generate a composite response; and transmitting, by the second agent, the composite response to the first agent.

14. The computer-implemented method of claim 13, comprising: receiving, by the second agent from the third agent, a third set of hashed identifiers used to generate the third response; generating, by the second agent, updated hashed document data comprising the third set of hashed identifiers and the second set of hashed identifiers; and transmitting, from the second agent to the first agent, updated hashed document data.

15. The computer-implemented method of any of claim 11 to claim 14, comprising transmitting, by the second agent, a request for a cryptographic hash understood by a third agent; receiving, by the second agent from the third agent, a list of hash functions understood by the third agent; selecting, by the second agent, a first hash function of the list of hash functions; and applying, by the second agent, the first hash function to data associated with the first document.

16. The computer-implemented method of any of claim 11 to claim 15, wherein the second agent comprises a large language model (LLM).

17. A computer-implemented method comprising: generating, by a first agent, an initial response to a user query based on one or more documents accessible by the first agent; transmitting, by the first agent, the user query and hashed document data to one or more second agents;accessing, by the one or more second agents, a document index, wherein the document index comprises a plurality of hashed identifiers of documents accessible by the second agent; excluding, by the one or more second agents, a first document based on the hashed identifier associated with the first document being in the hashed document data and the document index; and generating, by the one or more second agents, a second response based on one or more remaining documents accessible by the respective second agent.

18. The computer-implemented method of claim 17 wherein the operations further comprise: transmitting, by the first agent, the user query and hashed document data to a plurality of remote agents until the first agent has received a threshold number of unique hashed identifiers indicating a number of unique documents considered by the plurality of remote agents exceeds the threshold; and collating responses generated by the remote agents to generate a composite response.

19. The computer-implemented method of claim 18, wherein transmitting the user query and hashed document data to the plurality of agents is performed in parallel.

20. The computer-implemented method of claim 18, wherein transmitting the user query and hashed document data to the plurality of agents is performed sequentially.

21. A computing system for privately sharing known information between agents comprising: one or more processors; and a transitory or non-transitory computer readable media storing a plurality of instructions that are executable to cause the one or more processors to perform operations, the operations comprising the method described in any of claims 1-19.

22. One or more transitory or non-transitory computer readable media storing instructions that are executable by one or more processors to perform operations comprising the method described in any of claims 1-20.

Images