Systems and methods for integrating a content item into one or more video scenes

Abstract

Systems and methods for integrating a content item into one or more video scenes are provided. The method includes receiving, by a computing device, input data comprising user data, video data, and entity data, wherein the entity data comprises entity representation data; generating, by a large language model (LLM) operating on the computing device, a textual prompt based on the user data and the entity data; identifying, by the LLM, one or more video scenes to present a content item based on the video data; processing, by the LLM, the textual prompt and the one or more video scenes to generate the content item; and integrating, by the LLM, the content item into the one or more video scenes.

Description

SYSTEMS AND METHODS FOR INTEGRATING A CONTENT ITEM INTO ONE OR MORE VIDEO SCENESFIELD

[0001] The present disclosure relates generally to systems and methods for integrating a content item into one or more video scenes.BACKGROUND

[0002] Traditional techniques for generating content items have become increasingly ineffective, primarily due to their inherently static nature. These methods necessitate that content items be pre-generated and stored prior to their deployment within a video game or other digital media, which demands considerable resources in terms of both storage capacity and computing power. Furthermore, the transmission of these pre-generated content items, particularly when they are not integrated into a video scene, often disrupts the user experience. These interruptions can deter users from interacting with both the content item and the game itself.

[0003] Accordingly, improved techniques for the generation and integration of content items are desired in the art. In particular, techniques for integrating content items into one or more video scenes that can be performed with reduced storage capacity and computing power would be advantageous.BRIEF DESCRIPTION

[0004] Aspects and advantages of the invention in accordance with the present disclosure will be set forth in part in the following description, or may be obvious from the description, or may be learned through the practice of the technology.

[0005] In accordance with one embodiment, a method for integrating a content item into one or more video scenes is provided. The method includes receiving, by a computing device, input data comprising user data, video data, and entity data, wherein the entity data comprises entity7representation data; generating, by a large language model (LLM) operating on the computing device, a textual prompt based on the user data and the entity data; identifying, by the LLM, one or more video scenes to present a content item based on the video data; processing, by the LLM, the textual prompt and the one or more video scenes to generate the content item; and integrating, by the LLM, the content item into the one or more video scenes.

[0006] In accordance with another embodiment, systems for integrating a content item into one or more video scenes are provided. The system includes one or more processors; and one or more transitory or non-transitory computer-readable media storing instructions that are executable to cause the one or more processors to perform operations, the operations comprising: receiving input data comprising user data, video data, and entity7data, wherein the entity data comprises entity7representation data; generating, by a large language model (LLM), a textual prompt based on the user data and the entity data; processing, by the LLM, the video data to identify a gameplay trigger; identifying, by the LLM, one or more video scenes to present a content item based on the video data and the gameplay trigger; processing, by the LLM, the textual prompt and the one or more video scenes to generate the content item; identify7, using the LLM, one or more content item locations based on the one or more video scenes and the content item; and integrating, by the LLM, the content item into the one or more video scenes based on the one or more content item locations.

[0007] In accordance with one embodiment, a method for integrating a content item into one or more video scenes is provided. The method includes receiving, by a computing device, input data comprising user data, video data, and entity7data, wherein the entity data comprises entity7representation data; generating, by a large language model (LLM) operating on the computing device, a textual prompt based on the user data and the entity7data; identifying, by the LLM, one or more video scenes to present a content item based on the video data; processing, by the LLM. the textual prompt and the one or more video scenes to generate the content item; integrating, by the LLM, the content item into the one or more video scenes; receiving, by the LLM, interaction data based on the integrated content item; generating, by the LLM, an interaction metric based on the interaction data; and determining, by the LLM, a temporal characteristic of the integrated content item based on the interaction metric, wherein the temporal characteristic includes a time period associated with presenting the content item.

[0008] These and other features, aspects, and advantages of the present invention will become better understood with reference to the following description and appended claims. The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the technology and, together with the description, serve to explain the principles of the technology.BRIEF DESCRIPTION OF THE DRAWINGS

[0009] A full and enabling disclosure of the present invention, including the best mode of making and using the present systems and methods, directed to one of ordinary skill in the art, is set forth in the specification, which makes reference to the appended figures, in which:

[0010] Figure 1 depicts an exemplary' block diagram of a system for the dynamic generation of content items using a generative model in accordance with embodiments of the present disclosure;

[0011] Figure 2 depicts an exemplary block diagram of a system for fine-tuning a generative model based on an interaction metric in accordance with embodiments of the present disclosure;

[0012] Figure 3 depicts an exemplary flow diagram of a method for the dynamic generation of content items using a generative model in accordance with embodiments of the present disclosure;

[0013] Figure 4 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;

[0014] Figure 5 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;

[0015] Figure 6 is a block diagram of an example sequence processing model according to example implementations of aspects of the present disclosure;

[0016] Figure 7 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;

[0017] Figure 8 is a block diagram of an example model development platform according to example implementations of aspects of the present disclosure;

[0018] Figure 9 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;

[0019] Figure 10 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;

[0020] Figure 11 is a block diagram of an example networked computing system according to example implementations of aspects of the present disclosure;

[0021] Figure 12 is a block diagram of an example computing device according to example implementations of aspects of the present disclosure; and

[0022] Figure 13 is a block diagram of an example computing device according to example implementations of aspects of the present disclosure.DETAILED DESCRIPTION

[0023] The present disclosure is directed to the dynamic generation of content items utilizing a generative model. Traditional techniques for generating content items have become increasingly ineffective, primarily due to their inherently static nature. These methods necessitate that content items be pre-generated and stored prior to their deployment within a video game or other digital media, which demands considerable resources in terms of both storage capacity and computing power. Furthermore, the transmission of these pre-generated content items, particularly when they are not integrated into a video scene, often disrupts the user experience. These interruptions can deter users from interacting with both the content item and the game itself.

[0024] To counteract these issues, generative models (i.e. a large language model) are used to enable the dynamic creation of content items. This approach effectively circumvents the requirement for pre-generating and storing the content items, allowing for on-the-fly generation of content items. As a result of this on-the-fly generation of content items, the resources required for both generation and storage of the content items are significantly reduced. This represents an improvement over the traditional techniques for the generation of content items due to the reduced burden on computing power and storage capacity.

[0025] Additionally, the generative model facilitates the customization of content items based on user data and video data. This allows the generative model to create content items that are specifically targeted at the user while being integrated into the video scene. The integration of the content item into the surrounding context of the gameplay environment allows the content item to become less disruptive of the gameplay.

[0026] In an embodiment, the system may be configured to receive information regarding the user’s interactions with the embedded content item. This allows for the generation of interaction metrics that quantify user engagement and behavior as they relate to the content item. These interactions may be aggregated to generate training datasets that reflect real user experiences and interactions. Through the use of these training datasets, the generative model can be fine-tuned to facilitate the creation of improved content items that optimize content creation based on actual user engagement patterns.

[0027] Benefits, other advantages, and solutions to problems are described below with regard to specific embodiments. However, the benefits, advantages, solutions to problems, and any feature(s) that may cause any benefit, advantage, or solution to occur or become more pronounced are not to be construed as a critical, required, or essential feature of any or all the claims.

[0028] Referring now to the drawings, Figure 1 illustrates an exemplary block diagram of a system for the dynamic generation of content items using a generative model. Figure 1 includes user data 102, entity data 104, entity representation data 106, video data 108, large language model (LLM) 110, textual prompt 112a-b, prompt expansion model 114, video scene 116, content item 118, gameplay trigger 120, and the like.

[0029] The operations include receiving input data that includes user data 102. As used in the cunent disclosure, user data 102 refers to information related to a user. User data 102 may include explicitly provided information, such as when a user fills out a profile and implicitly gathered information, such as when a user interacts with a system or platform over time. User data 102 may include contact information associated with the user. Contact information can include a user’s name, email address, phone number, physical address, gamer identifier, and the like. User data 102 may include demographic information about the user, such as age, gender, location, race, socioeconomic status, user preferences, purchasing habits, and the like.

[0030] User data 102 may include information about the items or services a user has purchased through a platform. This could include physical products, digital goods, in-game currency, subscription services, game titles, links clicked, and the like. In some cases, the user data 102 may include information about a user’s interaction history. This can include a record of how users have engaged with the platform or service, including clicks, searches, comments, or social media interactions. User data 102 may include information about how often and where a user engages with the platform, what times they are most active, and how long they spend in specific activities or levels.

[0031] In an embodiment, the user data 102 may include a user profile. As used in the current disclosure, the user profile refers to a collection of data that provides insights into a user's characteristics, history, and interactions with the system. The user profile may include a variety of information, such as demographic details (e.g., age, location, race, socioeconomic status), preferences (e.g., content types or genres the user enjoys), and past behaviors (e.g., games played, items interacted with, or in-game achievements).

[0032] The operations include receiving input data comprising entity data 104. As used in the current disclosure, entity data 104 refers to data associated with an entity’. Entity data 104may include information about the entity’s products and sen-ices, such as information about the entity's activities, products, and / or services. Exemplary embodiments of entity data 104 may include descriptions of the items or services the entity provides, along with specifications, features, pricing, availability, and distribution channels. For example, in the case of a retail company, entity data 104 could include a comprehensive list of products available for sale, along with their specifications (such as size, color, material, etc.), prices, and promotional offers.

[0033] Additionally, entity data 104 may include information about the target demographic for the products and services of the entity. This can include information about the specific groups of consumers or users that the entity aims to serve or attract. For example, the data might include demographic details such as age. gender, income level, socioeconomic factors, geographic location, psychographic factors, and the like.

[0034] In some cases, the entity data 104 may include entity representation data 106. As used in the current disclosure, entity representation data 106 is data associated with the identifying phrases and imaging of the entity. Entity representation data 106 may include information about the entity's identity to the public. This may include information about the entity’s logos, branding, marketing, historical content items, and the like. For example, entity representation data 106 may include the visual identifiers that the entity uses to distinguish itself from competitors and to make its products or services instantly recognizable to consumers. Entity representation data 106 may include identifying phrases or taglines that are associated with the entity. Alternatively, the visual identifiers may include things like logos, color schemes, brand typography, mascots, and the like. Images associated with the entity, whether they are graphics, product imagery, or other visual representations, also fall under entity representation data 106.

[0035] The operations include receiving input data that includes video data 108. As used in the current disclosure, video data 108 is data associated with digital media. Digital media may include both interactive and non-interactive forms of media. Digital media may include video games, smartphone games, console games, computer games, videos, images, social media posts, movies, television shows, pre-recorded video, live events (z.e. sporting events and concerts), and the like. Video data 108 may include information about the parts and form of the digital media. The parts and form of the digital media may include information about upcoming video scenes and imagery, the temporal window associated with those video scenes and imagery, and the like.

[0036] In an embodiment, video data 108 may include gameplay data. As used in the current disclosure, gameplay data refers to data that is associated with video games. Gameplay data may include information about how players interact with the digital media. This may include information associated with the various game elements that define the gaming experience. These elements may include the levels, characters, environments, interactive objects, and the like within the game or digital media. The gameplay data may include information about the levels within the game. Levels refer to the stages or areas that players navigate through as they progress in the game. Gameplay data may include information related to how the players move through these levels, how long they spend on each stage, and whether they encounter specific challenges or obstacles.

[0037] Similarly, video data 108 may include information about game elements such as the characters within the game or digital media. This may include information about the player’s chosen character or avatar, how they evolve over time, and the various abilities or attributes that they gain as they progress through the game.

[0038] Video data 108 and / or gameplay data may include information related to game elements such as the game environment. This can include information about how players interact with different locations, areas, items, characters, and / or settings. An exemplary embodiment of this may include data on how players explore or revisit specific environments, and how often they engage with environmental objects. For example, video data 108 may track how players navigate through the in-game world. This may include insights into which areas are more appealing or engaging than others.

[0039] In some cases, video data 108 may include information about game elements such as interactive objects. Interactive objects may be the items or elements that players can interact with within the in-game world, such as weapons, power-ups, items, collectibles, and NPCs (non-playable characters). The video data 108 may capture how frequently players use these objects, which objects they prefer, and how the use of these items impacts gameplay.

[0040] Beyond individual game elements, the video data 108 may include information about the structure of gameplay, such as information about the sequence of events within the gameplay. The sequence of events may be described in terms of the obstacles that players encounter, how they overcome challenges, and the way in which the game’s narrative or objectives unfold.

[0041] Furthermore, video data 108 can include a temporal window associated with each game element. The temporal window may refer to the time frame in which particular events, scenes, or actions occur within the game. For instance, the temporal window may includeinformation about tracking when / how specific scenes or levels are expected to be introduced or how long certain gameplay sequences last. Video data 108 might include data on how long a player spends in a particular scene, when transitions between scenes occur, or how the pacing of the game might affect the overall experience.

[0042] With continued reference to Figure 1, system 100 includes a large language model (LLM) 110. As used in the current disclosure, a LLM 110 is a deep learning data structure that can recognize, summarize, translate, predict, and / or generate text and other content based on knowledge gained from datasets. Large language models 110 may be trained on large sets of data. Training sets may be draw n from diverse sets of data such as books, websites, advertising, content items, structured data, unstructured data, electronic records, video game data, digital media, and the like.

[0043] In an embodiment, the LLM 110 may include a text-to-image (T2I) model. As used in the current disclosure, the text-to-image model is designed to generate images based on textual inputs. The T2I model is configured to interpret textual data to generate image data. The generated image data can include but is not limited to all or a portion of the content item 118, video scenes, and any other image data discussed herein. The T2I model can use several natural language processing techniques to convert textual prompts into image data. The T2I model can include any of the machine learning models, natural language processing models, image processing models, large language models, and the like that are discussed herein below in Figures 4-13.

[0044] With continued reference to Figure 1 , the operations include generating a textual prompt 112a-b based on the user data 102 and entity data 104. As used in the current disclosure, the textual prompt 112a is a textual description of the content item 118. The textual prompt 112a is used to create tailored content items 118 that are personalized to individual users while also aligning with the identifying imagery or text of the entity. In a non-limiting example, if user data 102 reveals that a particular user frequently engages with a specific genre of digital media, the LLM 110 may create a textual prompt 112a that includes elements of that genre. The LLM 110 may evaluate user behavior, purchase history, interaction history, and the like to generate the textual prompt 112a that aligns with the user's demonstrated actions and choices.

[0045] The textual prompt 112a may incorporate information from the entity data 104 and the user data 102 simultaneously. The entity data 104 may be used to ensure that the textual prompt 112a is aligned with the entity's identifying phrases and imaging. This may be done by tailoring the textual prompt 112a to the entity's target demographic, products, services, imaging, and / or language. In a non-limiting example, if an entity is targeting a particulardemographic, the LLM 110 may tailor the description within the textual prompt 112a to appeal specifically to that demographic. This may be done by using appropriate tone, terminology, and reference points. In some cases, the textual prompt 112a could also include instructions to incorporate signature phrases, logos, or specific products into content item 118.

[0046] With continued reference to Figure 1, generating the textual prompt 112a-b may include processing, using a prompt expansion model 114. the textual prompt 112a to generate an expanded textual prompt 112b. As used in the current disclosure, the prompt expansion model 114 is a model that is configured to generate an expanded textual prompt 112b. PEM 114 can be consistent with the machine-learning models described herein below in Figures 4- 13. The PEM 114 operates by predicting and generating expanded versions of the textual prompt 112a based on the video data 108. The PEM 114 evaluates the video data 108 to predict and generate expanded versions of the textual prompt 112a-b that are aligned within the context of the digital media. For instance, if the video data 108 indicates that the game is set within medieval times, the PEM 114 can expand the textual prompt 112a-b to include descriptors that are aligned with the medieval theme of the game.

[0047] In an embodiment, the PEM 114 can include a neural network architecture. The PEM 114 can include multiple layers of interconnected nodes, or neurons, which are configured to process data in a hierarchical manner. Each layer of the neural network can be responsible for different aspects of the input, enabling the PEM 114 to leam complex patterns and relationships within the data. The textual prompt 112a can be processed using these layers where the neural network analyzes the text and identifies key components that can be expanded upon.

[0048] The nodes in the PEM 114 can be organized in a structured network, such as a convolutional neural network, which includes an input layer of nodes, one or more intermediate layers, and an output layer of nodes. During the training of the PEM 1 14, connections between these nodes can be established by applying elements from the training dataset to the nodes.

[0049] As used in the current disclosure, the expanded textual prompt 112b is a textual prompt 112a that has been augmented to include more details and contextual information. Augmenting the textual prompt 112a can include adding additional descriptors to the prompt, such as colors, details about the setting, emotional context, gameplay details, gameplay triggers, gameplay elements, and / or contextual details related to the user or the entity. In anon- limiting example, if a textual prompt 112a states "generate a content item associated with product A based on the user’s favorite character," the expanded textual prompt 112b caninclude an augmented prompt that specifies "generate a content item that has character A presenting product A within the medieval castle within the video scene."

[0050] In an embodiment, the operations may further comprise processing the textual prompt 112a-b and the video data 108 to generate a smart prompt. As used in the current disclosure, the smart prompt is a specialized prompt designed to interact with the LLM 110. The smart prompt is generated to maximize the performance of the LLM 110 by using specific wording, structure, contextual information, and / or video data 108. The smart prompt augments the textual prompt 112a-b to include specific details and context that can be used to guide the LLM 110 to generate a more refined content item 118. The smart prompt might provide additional information to the LLM 110 regarding which portion of the prompt to focus on, the length and format of the response, the key attributes, and the like.

[0051] In a non-limiting example, the textual prompt 112a-b describes "generating a content item 118 that includes product B and the user’s digital avatar." System 100 may leverage the video data 108 to generate a smart prompt that provides additional context and information to the LLM 110. Based on this additional context of the video data 108 may generate a smart prompt that states "Place logo B within a first location within the video scene and incorporate the user’s digital avatar interacting with product B within the content item." Through the use of this smart prompt, the LLM 110 is fed additional information about the desired content item.

[0052] With continued reference to Figure 1, the operations include identifying, by the LLM 1 10, one or more video scenes 1 16 to present a content item 1 18 based on the video data 108. As used in the current disclosure, video scenes 116 refer to a segment or moment within the digital media or gameplay. The video scene 116 may be defined by the environment, objectives, interactions, narrative progressions, and the like. The video scene 116 can include a variety of events, such as action sequences, puzzle-solving, exploration, dialogue, cutscenes, and the like. In some cases, video scenes 116 may include one or more gameplay scenes. These gameplay scenes can be seen as portions of the game where the content item 118 may be displayed.

[0053] In an embodiment, the video scene 116 might be defined by a change in the environment or scene within the digital media or game. In anon-limiting example, video scene 116 might begin when a character enters a new area of the in-game environment or by triggering specific objectives or interactions. Similarly, video scenes 116 can be delineated by moments in the digital media’s storyline, such as a boss fight, an in-game event, or a character interaction. Video scenes 116 may also be defined as one or more interactions between theplayer and the media environment. These could include interactions like engaging with another player, in-game items, NPC. dialogue choices, environments, levels, actions, and the like.

[0054] In some cases, video scenes 116 may be defined according to a temporal window. This temporal window may be defined based on its relationship to one or more events internal or external to the digital media. In a non-limiting example, the temporal window may be defined as a 15 second window that occurs prior to a character interaction.

[0055] Additionally, the LLM 110 may use contextual understanding of the digital media's world and objectives to make inferences about gameplay progression to identify the video scenes 116. This may be done by analyzing text logs within the video data 108 to identify the video scene 116 based on events that occur within the digital media. For example, if the game logs show that the player has completed a specific quest, moved into a new area, or encountered a significant narrative event, the LLM 110 might recognize that these actions constitute a scene transition.

[0056] In some cases, the LLM 110 can be used to identify the video scene 116 based on user traffic within the digital world. This may be done by analyzing patterns within character behavior and engagement across different segments of the game or digital world. In-game user traffic refers to the frequency or distribution of player actions and interactions within the game world. For example, in a multiplayer environment, user traffic could involve the number of players in a specific area, the rate of interactions w ith the environment, item, or the occurrence of player-driven events like quests or objectives.

[0057] During gameplay, certain scenes or moments may naturally generate spikes in user traffic. This elevated user traffic can be detected through system logs or data feeds that track user inputs, interactions, and movements. The LLM 110 can process this data to recognize patterns of increased activity, such as multiple players clustering in a particular area, a high number of in-game actions, or the activation of specific events that drive user engagement. By understanding these patterns, the LLM can identify an optimal video scene 116 to place the content item 118.

[0058] The LLM 110 could analyze the timing and distribution of character actions across various areas of the digital media’s world to identify in-game user traffic. This can be done by leveraging the patterns of user traffic to differentiate between regular gameplay traffic and moments of heightened gameplay traffic. For instance, a steady but lower volume of user traffic might be associated with routine gameplay, while elevated traffic patterns may indicate a transition into a more intense or narrative-driven video scene. By monitoring these trafficpatterns over time, the LLM 110 can leam to select a video scene 116 that features the optimal amount of in-game user traffic.

[0059] With continued reference to Figure 1 , the operations may include selecting a content item location based on the video scene 116. As used in the current disclosure, the content item location refers to the placement of the content item 118 within the video scene 116. The content item location may be configured to place the content item 118 in a location within the digital media with minimal disruption to gameplay or media consumption. The goal of selecting the content item location is to ensure that the content item 118 is both noticeable and accessible to the user while posing a minimal disruption to the flow or pacing of gameplay or media consumption.

[0060] The content item location may be selected based on a variety of dynamic or fixed elements within the video scene 116. A well-placed content item 1 18 may align with the cunent narrative, environmental cues, or gameplay mechanics that are active in that particular video scene 116. For example, if the user is in a combat-heavy scene, the content item 118 may be located within a health item where it is visible but not so easy to reach that it undermines the challenge of the encounter. In an additional example, if the user is watching a live sporting event, the content item 118 may be located within the field of play where it is visible but not distracting from the sporting event.

[0061] One of the key aspects of selecting the content item location is ensuring that it causes minimal disruption to gameplay or media consumption. Disruption could occur in various forms, such as breaking the flow of the game, making the game feel too easy or too difficult, interrupting the player’s immersion in the world, detracting attention from the digital media, or a content item 118 that feels out of place within the video scene 116. These interruptions can deter users from interacting with both the content item 118 and the digital media itself. In an effort to avoid these disruptions, content item locations are often chosen to complement the pacing and tone of the video scene. To accomplish this content item locations may be selected in such a way that they provide strategic vi sibility / accessibility without feeling out of place within the game.

[0062] Avoiding disruptions may be done by incorporating the content item 118 into a content item location that is integrated into gameplay or the digital media. For instance, an ingame collectible item could be placed in an area that the player is naturally drawn to, such as at the center of a room, on a pedestal, or within a well-lit area. In an additional or alternate embodiment, the content item location might be slightly less visible areas such as the background of the video scene. This might include replacing items (i.e. labels, products,branding, and the like) within the digital media with the content items 118. In a third embodiment, the content item location may be facilitated by contact with an NPC, where the NPC presents the user with the content item 118.

[0063] In some cases, the content item locations may be dynamic. Meaning that the content item locations may be procedurally generated based on the user’s presence in in-game environments or in response to the player's progress through a scene. For example, the content item locations may be iteratively generated or adjusted depending on real-time events. This allows the game to respond to player behavior and dynamically alter the content item placement to better suit the immediate context of gameplay.

[0064] With continued reference to Figure 1, wherein the operations may include processing, by the LLM 110, the video data 108 to identify a gameplay trigger 120. As used in the current disclosure, the gameplay trigger 120 refers to an event or condition within the game that serves as a signal or cue to System 100 that a particular video scene 116 is approaching. The gameplay triggers 120 may additionally signify to System 100 that certain actions should be taken, such as the generation of the content item 118. The gameplay triggers 120 may be a predefined or dynamic event that marks a transition or milestone in the game’s progression.

[0065] The gameplay triggers 120 can operate as an indicator for System 100 to anticipate the upcoming scene transition. For example, if a player is nearing a video scene 116, the gameplay triggers 120 might include the user entering into an area within the game, achieving a certain level of progress, or reaching a predetermined time or event threshold. At this point. System 100 can preemptively prepare the game environment by selecting the video scene 1 16, the content item locations, generating the content item 118, and the like.

[0066] In a non-limiting example, if System 100 detects that a player is nearing an obj ective or narrative milestone, the gameplay trigger 120 might activate when certain conditions are met, such as interacting with an NPC or completing a series of preliminary tasks. The system could then generate the content item 118 based on the gameplay trigger 120.

[0067] In an embodiment, the gameplay trigger 120 may be determined based on the video data 108. Data points such as player actions, system states, environmental variables, and event logs may be considered when identifying the gameplay trigger 120. The determination of the gameplay trigger 120 may come from both the player's direct input (like movement, combat, or interactions) and the game system's internal tracking of the game state (such as time elapsed, objectives completed, or enemy behavior). In some cases, System 100 may continuously monitor the player's position to determine and enact the gameplay trigger 120. Similarly,gameplay data related to player actions (e.g., defeating a group of enemies, interacting with an object, or completing a mini-quest) can be used to identify the gameplay trigger 120.

[0068] With continued reference to Figure 1, the operations include processing, by the LLM 110, the textual prompt 112a-b, and the one or more video scenes 116 to generate the content item 118. As used in the current disclosure, a content item refers to a piece of media that represents a portion of the entity data 104. The content item 118 may be dynamically tailored to the user data 102. the video scene 116, and / or the entity data 104. By tailoring the content item 118, System 100 can create a content item 1 18 that is not only relevant to the current video scene 116 but also selectively customizable to the player, the digital media, and the entity . Depending on the context, the content item 118 may take many forms, including images, audio, video, cut scenes, in-game items, text-based rewards, interactive objects, dialogue between characters, in-game locations, in-game characters, NPCs, and the like. In a non-limiting example, the content item 118 may take the form of a cut scene that displays story events or character interactions that incorporate a portion of the entity data 104. This cut scene could be triggered when the player reaches a certain video scene 116 or gameplay trigger 120.

[0069] The generation of the content item 118 from the textual prompt 112a-b is a process that is performed by LLM 110. The LLM 110 is used to analyze the textual data to produce the desired image, audio, cut scenes, in-game items, text-based rewards, interactive objects, and the like. The textual prompt 112a-b may be used to provide System 100 with a set of instructions, narrative elements, or parameters that guide the creation of the content item 118. Based on these instructions, the LLM 1 10 may employ its language-processing capabilities to interpret the prompt and generate the content item 118 based on that understanding.

[0070] Once the textual prompt is processed, the LLM 110 generates the content item 118 that aligns with each of the video scene 116, entity data 104, and user data 102. The LLM 110 may be configured to adjust the imagery, audio, or text of the content item 118 based on the player’s progress, choices, or current game state. For instance, the LLM 110 might pull from a set of available assets (like character models, animations, environmental objects, or sound files) and modify them based on the textual prompt to create something unique for the current gameplay moment. The output could be an image, a sound file, a segment of dialogue, or even a new in-game object, all of which would seamlessly integrate into the game world.

[0071] With continued reference to Figure 1, the operations further include integrating, by the LLM 110, the content item 118 into the one or more video scenes 116. Integrating the content item 118 into one or more video scenes 116 refers to the process of embedding new or external content into the digital media. The integration includes placing the content item 118in a particular location within the scene but also ensuring that it functions coherently with the surrounding environment, mechanics, and narrative flow of the game. The LLM 110 is used to integrate the attributes and characteristics of the content item 118 into the in-game universe. For example, if the content item 1 18 is an NPC, integrating the content item 118 may include placing the NPC within the one or more video scenes 116 and generating images, movements, dialogue, and / or text that would fit within the context of the video scene 116. This may include a determination of where the content item 118 would logically exist based on the narrative context, the game’s geographical layout, or its logical structure.

[0072] Inserting the content item 118 into the game scene requires the LLM 110 to consider several factors beyond mere placement. For instance, the item might need to align with the visual aesthetics of the scene. In an embodiment where the content item 118 is a new building or structure, integration of the content item 118 may include integrating the structure’s design with the game’s art style and spatial constraints. Similarly, if the content item is a gameplay mechanic, integration of the content item 118 may include incorporating the content item 118 within the existing game mechanics. In some cases, this may involve dynamically adjusting other aspects of the video scene 116 to accommodate the new item and create a cohesive experience for the player.

[0073] Integrating the content item 118 into the one or more video scenes 116 may include inserting the content item 118 into the content item location. This process includes placing the content item 1 18 at a defined spot within the game’s environment. The integration of the content item 1 18 may also include placing the content item in a location that is visible and accessible to the player but also maintains the integrity of the game’s pacing and narrative.

[0074] With continued reference to Figure 1, the operations may include generating, by the LLM 110, the content item 118 during a predefined temporal window, wherein the predefined temporal window occurs at least a threshold duration prior to a gameplay trigger. Generating the content item 118 during a predefined temporal window refers to the process of creating or selecting the content item 118 at a specific point in time that occurs before an event or action, such as the gameplay triggers, to trigger the generation of the content item 118. As used in the current disclosure, the predefined temporal window is a time interval that is set by the system to allow for the content item 118 to be prepared in advance of user interaction. The purpose of this window is to ensure that the content item 118 is ready and available to be delivered at the right moment, while also optimizing system performance byreducing delays caused by real-time generation or retrieval of content.

[0075] The predefined temporal window may occur at least a threshold duration prior to the actual gameplay trigger 120, meaning that the system anticipates the user's potential interaction with the content item 1 18 in advance. This allows the LLM 110 to prepare the content item 118 in advance and ensure it is available when needed, reducing latency, and ensuring a smooth gameplay experience. The threshold duration could vary based on game mechanics, the complexity of the content, or system resources, but it generally allows enough time to generate and store the content item in memory before it is presented to the user.

[0076] Once the content item 118 has been generated during the predefined temporal window, the system may store the content item 118 in a cache memory associated with the computing device. Cache memory is a type of high-speed memory that is used to store data that is frequently accessed or expected to be needed soon. In this case, the content item 118 is stored in the cache so that it can be quickly retrieved and delivered to the user as soon as it is needed, without the system having to regenerate or retrieve it from slower storage sources like a hard disk or cloud server.

[0077] The cache is designed to store content in a way that optimizes performance, allowing for fast access to the content item when it is required during gameplay. This reduces the time spent loading or processing content. The LLM 110 can dynamically manage the cache, prioritizing which content items 118 are stored based on factors like predicted player behavior, recent interactions, or gameplay events. If the cache reaches capacity, older or less frequently used items may be replaced, ensuring that the most relevant content is always ready for immediate use.

[0078] Referring now to Figure 2, an exemplary' block diagram of a system for fine-tuning a generative model based on an interaction metric. Figure 2 includes interaction data 202, interaction metric 204, interaction score 206. interaction report 208, temporal characteristic 210, content preference metric 212, training data 214, and the like.

[0079] The operations may include receiving, by the LLM 110, interaction data 202 based on the integrated content item 118. As used in the current disclosure, the interaction data 202 refers to the data generated as a result of a user's interactions with content items 118 within a game or interactive environment. Interaction data 202 may be used to describe how users engage with various elements of the game, such as objects, characters, mechanics, and UI elements. The primary' purpose of interaction data 202 is to track and record the user’s choices, actions, and feedback related to the content item 118. This may include information associated with a wide range of behaviors that occur when the user engages with the content item 118. Interaction data 202 may be generated based on the user's interactions with the content item118. This can include a variety of actions such as collecting an object, initiating a dialogue with an NPC, using an item, triggering a game mechanic, clicking a link, using a promotional code, temporal-based interactions, physical interactions, and the like.

[0080] The operations may include generating, by the LLM 110, an interaction metric 204 based on the interaction data 202. As used in the current disclosure, the interaction metric 204 refers to a metric that is used to reflect the user’s engagement with the content item 118. The interaction metric 204 may be used to quantify the details around the user’s engagement with the content item 118. The interaction metric 204 aggregates and analyzes one or more users’ interactions to provide a quantification of how frequently users engage with content items 118.

[0081] The interaction metric 204 may include a breakdown of the types and styles of content items 118 which generate the most favorable interactions with users. By analyzing interaction data 202, the system can identify patterns in player behavior and determine the specific content items, gameplay features, or assets that drive the most engagement. The interaction metric 204 may include an evaluation of the factors that can drive engagement like the type of content item 118, the entity, the product, user demographics, the game demographics, and the like. For example, if a content item 118 is tied to a specific product or entity, the level of engagement might vary depending on how well-received that product or entity is by the player base.

[0082] The interaction metric 204 may be used to identify and quantify how various factors affect user engagement. This evaluation is useful because it is inevitable that some content items 1 18 may be more effective than others in generating positive user interactions based on their function or role within the digital media. Understanding the metrics and the underlying reasoning behind the phenomena can be used to improve the content items. Specifically, these insights can be used to train, re-train, or fine-tune the LLM 110. This may be done by using the previous inputs and outputs of the LLM 110 alongside the interaction metrics 204 as training data for the LLM 110. Systems and methods disclosed herein may be used to iteratively finetune the biases and weights associated with the LLM 110 using this training data until convergence is achieved. In some cases, this may include using reinforcement learning to finetune the biases and weights associated with the LLM 110.

[0083] The LLM 110 or another machine learning model may be used to generate the interaction metric 204. This can be done by tracking how often the user engages with the content item 118. The frequency of these interactions contributes to the overall interaction metric 204. allowing the system to calculate how engaging the content item 118 is to the player. Additionally, the LLM 110 may measure the duration or time spent interacting with the contentitem. Longer or more sustained interactions often indicate a higher level of interest or engagement, which may be reflected positively in the interaction metric 204. In some embodiments, the interaction metric 204 may include an evaluation of how long the content item 118 is on the user’s screen before the engagement rate, click-through rate, conversion rate, and the like start reacting negatively.

[0084] In some embodiments, the interaction metric 204 can be used to tailor the integration of the content item 118 to the users. For example, if the system detects that a particular player has shown a high level of engagement with certain types of content items 118, the LLM 110 may tailor future content to emphasize those types of interactions. This might involve presenting the player with content items 118 that are associated with products and entities that the user is known to have ties to. Additionally, this may involve selecting the content item location or the form of the content item 118 based on the historical level of engagement by the user or the user demographic groups. Exemplary embodiments of this include positioning items in areas where the player is most likely to notice them or presenting them in formats that align with the player’s interaction hi s lory.

[0085] Additionally, the interaction metric 204 may be used to evaluate the user’s in-game behavior. The LLM 110 may aggregate the interaction metrics 204 across the player base to identify trends in engagement. For instance, the interaction metric 204 might reveal that a particular gameplay mechanic is the optimal form of the content item 118. Conversely, a low interaction metric across the board for certain items or mechanics may highlight areas of the game that require adjustments or redesign.

[0086] In some embodiments, the interaction metric 204 may be represented as an interaction score 206. As used in the current disclosure, the interaction scores 206 is numerical representation of the engagement of the user with a particular content item 118. These scores are used to capture the degree to which players interact with and show interest in the content item 118. In an embodiment, the interaction scores 206 may be generated based on a variety of user behaviors and actions within the game. These actions can include clicks, item usage, time spent with the content item 118, completion of in-game objectives, and the like. System 200 analyzes these behaviors and aggregates them into the interaction scores 206. Interaction scores can be designed to capture multiple aspects of engagement, including frequency, intensity, and context.

[0087] In some embodiments, interaction scores 206 may also take into account the time spent interacting with a content item. For example, if a player spends a long period examining an item, reading dialogue, or participating in a related activity, this suggests a higher level ofengagement compared to a quick interaction. The interaction score 206 might reflect the extended engagement, assigning higher value to actions that demonstrate deeper involvement. Additionally, interaction scores 206 may also take into account the frequency of interactions. The content item 118 that is used regularly by a player, especially if those interactions occur over an extended period, may have a higher interaction score 206.

[0088] In some cases, the interaction scores 206 may reflect the engagement rate of the content items 118. As used in the current disclosure, the engagement rate refers how often a player interacts with a particular content item 118 relative to the total number of opportunities for interaction. This means that the engagement rate is essentially a measure of the frequency of interaction, providing insight into how likely players are to engage with a particular piece of content when they encounter it. If a content item is used frequently by a player, this will result in a higher engagement rate, and therefore, a higher interaction score. The engagement rate may be calculated by dividing the number of interactions with the content item by the number of potential interactions or opportunities for engagement. For example, if a content item 118 appears in the game world 100 times, but players interact with it 50 times, the engagement rate would be 50%. A higher engagement rate indicates that players are more consistently and actively engaging with the content item whenever it is presented, which often signifies that the content is particularly valuable, relevant, or appealing to the players.

[0089] By analyzing engagement rates and their corresponding interaction scores 206, the LLM 110 may gain valuable insights into player behavior and preferences. For instance, content items 118 with high engagement rates could be further highlighted or integrated into the digital media. In some cases, the LLM 110 may create additional content around these content items 118, make them more accessible, or design new mechanics that revolve around their frequent use. Conversely, if a content item has a low engagement rate, it could be refined or redesigned to improve its relevance, functionality, or appeal.

[0090] In some cases, the LLM 110 can also use interaction scores 206 to personalize content placement and availability. If the system detects that certain types of content items lead to higher engagement for specific players or demographics, it can recommend or present those items more frequently.

[0091] Additionally, the interaction scores 206 may be used to reflect the click through rate associated with the content items 118. As used in the current disclosure, the click-through rate (CTR) measures how often a user clicks on a content item or its associated prompt in comparison to how often it is displayed or made available. The CTR may be used to measure the effectiveness of content items 118 within a game, The CTR can be the ratio of how oftenplayers click on a specific content item or its associated prompt relative to the number of times the item is displayed or made available to the player. This metric provides insight into how compelling or attractive a content item is to players within the context of the game. A higher CTR indicates that the content item is attracting more user attention, leading to a higher interaction score.

[0092] Interaction scores 206 that are calculated based on the CTR may vary based on the context in which the content item 118 was presented. This context may factor in the visibility / accessibility of the content item 118 within the digital media. If a content item 118 is displayed in an obscure or hard-to-find location, even if it’s highly desirable, its CTR will likely be low simply because players are not aware of it or don't have easy access to it. Conversely, if the content item is placed in a central location where players are likely to encounter it the CTR will likely be higher. The system may adjust biases and weights associated with the interaction score 206 to account for the visibility / accessibility of the content item 118.

[0093] By analyzing the data associated with the CTR, the LLM 110 can determine which content items 118 are generating the most interest and engagement, and which ones may need further refinement. Items with a high CTR are clearly resonating with players, suggesting that they fulfill a need, offer a compelling reward, or align with the player’s current objectives. In such cases, the content item 118 may be featured more prominently or integrated into other aspects of the game to increase its impact.

[0094] In some embodiments, the conversion rate may be factored into the interaction score 206. As used in the current disclosure, the conversion rate measures how often a user takes a desired action after interacting with the content item. These actions might include making a purchase, clicking on a link, signing up for a service, completing an in-game objective, or even visiting an external website. The higher the conversion rate associated with the content item 1 18, the more effective that content item 118 is at prompting the user to take meaningful actions that align with the goals of the game or the business. By incorporating the conversion rate into the interaction score 206, the LLM 110 can evaluate how well content items 118 are fulfilling their intended purpose beyond simple engagement.

[0095] With continued reference to Figure 2, the operations may include generating, by the LLM 1 10, an interaction report 208 based on the interaction metric 204. As used in the current disclosure, the interaction reports 208 refers to a report that details the analysis of the interactions between users and the content item 118. The interaction reports 208 may be generated based on the data derived from the interaction metric 204. The interaction reports208 may be configured to provide insights into how players are interacting with one or more content items 118.

[0096] The interaction reports 208 may include a quantitative breakdown of interactions by different demographics. This can include user demographics such as age, location, gender, game preference, purchase history, or other user-specific attributes. By sorting interaction data by demographics, the system can uncover patterns of engagement that might not be immediately obvious from aggregate data alone. For example, a content item might perform particularly well with a certain age group or player type, or perhaps certain in-game events resonate more with players from specific regions.

[0097] Additionally, the interaction reports 208 can be used to assess the effectiveness of content item location, timing, and the like. The interaction report 208 may provide insights into how well a content item 118 fits within the larger context of the game. For example, if a particular in-game content item 118 is generating high interaction scores but few- conversions, this discrepancy could indicate that the ad is attracting attention but not offering a compelling enough incentive to drive purchases.

[0098] With continued reference to Figure 2. the operations include determining, by the LLM 110, a temporal characteristic 210 of the integrated content item 118 based on the interaction metric 204, wherein the temporal characteristic includes a time period associated with presenting the content item. As used in the current disclosure, the temporal characteristic 210 refers to the properties that define how long the content item 118 is presented. The temporal characteristic 210 may include various time-based parameters, such as the time period during which the content item 118 is visible or interactable. The temporal characteristic 210 may be defined by the time window during w hich the content item 118 is available for interaction. This could be defined by certain time-based conditions, such as a specific in-game event or narrative progression. The time period may be fixed or dynamic. In embodiments where the temporal characteristic 210 is dynamic the content item’s availability may be iteratively changed based on player actions or in-game factors.

[0099] The temporal characteristic 210 may also include time-limited interactions tied to the content item 118. For example, the content item 118 might be available for only a brief period before it expires or becomes inaccessible. Time-based incentives like this may be used in games to promote specific behaviors, drive user engagement, or align with game narrative arcs.

[0100] The LLM 110 may determine the temporal characteristic 210 of a content item 118 by analyzing both the interaction metric 204 and the video data 108. The LLM 110 mayconsider both predefined game parameters and dynamic, real-time factors. The LLM 110 uses these factors to decide when content should be made available to the player. The LLM 110 may also account for player-specific interaction metrics 204 to adjust the timing and availability of content. The LLM 110 may fine-tune the temporal characteristic 210 based on the player’s behaviors. This may be done by tracking patterns such as how often and when players engage with certain types of content items 118.

[0101] With continued reference to Figure 2. the operations include determining, by the LLM 110, a content preference metric 212 based on the interaction metric 204, wherein the content preference metric 212 comprises a quantification of a user’s preferences towards the content item 118. As used in the current disclosure, the content preference metric 212 is a measure that quantifies the preferences of users towards the content item 118. The content preference metric 212 provides a detailed analysis of what kind of content resonates with specific demographics of users. The content preference metric 212 may be generated from the aggregated data of the interaction metric 204 across a plurality of users. System 200 uses this to identify patterns and trends in how different user groups engage with the content item 118.

[0102] By analyzing the interaction metrics 204, the LLM 110 may gain insights on what demographic of users interact with the content items most frequently. These demographics can include characteristics like age, gender, ethnicity, skill level, geographic location, play frequency, and the like. The content preference metric 212 uses aggregated interaction metrics 204 to identify which types of content are most appealing to different groups. By quantifying the preferences of different user demographics, the LLM 1 10 can fine-tune content items 118 to the users based on their demonstrated interests. For example, if a certain demographic consistently interacts with event-based content, the system might prioritize similar time-limited events or in-game promotions for that group.

[0103] With continued reference to Figure 2, the operations may include generating, by the LLM 110, training data 214 based on the interaction metric 204. As used in the current disclosure, training data 214 is a collection of labeled examples that are used to train a machine learning model. Training data 214 may be referred to as a dataset that includes information on how users interact with various content items 118, capturing the correlations between content types and user preferences. By aggregating interaction metrics 204, the LLM 110 compiles data entries (i.e., training examples) that include multiple data elements, such as the frequency of interaction, engagement duration, and types of content the user interacts with.

[0104] The system 100 may create the training data 214 by collecting and aggregating interaction data 202 and interaction metrics 204 from a diverse set of users. This may enableSystem 200 to detect trends and correlations between different content items 118 and user demographics. These patterns may include identifying which content items 118 generate the most engagement within specific demographic groups, or how different types of content items 118 lead to higher conversion rates, or other desired actions. The training data 214 can also include non-categorized elements, allowing the system to dynamically generate new training examples based on emerging patterns or behaviors identified through data analysis.

[0105] Once training data 214 is generated, it may be formatted and organized to facilitate analysis and machine learning processes. The data elements within the training set may be associated with descriptors that categorize the data, such as user demographics, interaction frequency, or types or forms of content items 118. The system 200 may use this structured data to train machine learning algorithms, language processing models, T2I models, classifiers, and predictive models, which can classify or predict user preferences for specific content types. In some cases, the system may also sanitize the training data by removing outliers or poor-quality data to ensure that the machine learning process converges on meaningful insights.

[0106] With continued reference to Figure 2, the operations may include fine-tuning, by the computing device, the LLM 110 using the training data 214. Fine-tuning refers to the process of refining and adjusting the parameters of the LLM 110 based on the insights derived from the training data 214. This process is aimed at improving the accuracy and performance of the model by making incremental adjustments to its parameters, allowing it to better predict and respond to user interactions with content items 118. The training data 214 may be used as input to update the model's underlying algorithms, ensuring that the LLM 1 10 is more aligned with real-world user behaviors and preferences.

[0107] During fine-tuning, the LLM’s 110 parameters, such as weights, biases, and other hyperparameters, may be iteratively adjusted through machine learning techniques, such as backpropagation or gradient descent. The goal is to enhance the LLM’s 110 ability to make predictions about which types of content items 118 will generate the most engagement or conversions, based on the patterns observed in the interaction metrics 204. For example, the model might learn to prioritize content that is more likely to appeal to a specific user demographic or adjust the timing of content delivery to optimize engagement.

[0108] In some embodiments, the LLM 110 may be fine-tuned using a reinforcement learning (RL) process. Reinforcement learning is a type of machine learning where an agent (in this case, the LLM 110) learns to make decisions by interacting with an environment and receiving feedback in the form of rewards or penalties. The goal of reinforcement learning isto optimize the agent's decision-making strategy, known as a policy, by maximizing cumulative rewards over time.

[0109] The fine-tuning process through reinforcement learning may include defining a reward function that reflects desirable outcomes from the user’s interactions with the content items 118. For example, the reward could be based on user engagement metrics such as the click-through rate (CTR), conversion rate, or time spent interacting with content. Each time the LLM 110 integrates the content item 118 into a video scene, the LLM 110 may receive feedback (a reward or penalty) based on how well the content performs in achieving these goals. The LLM 110 then uses this feedback to adjust its decision-making process. Over time, as it interacts with more users and receives more feedback, the LLM 110 becomes better at predicting which types of content items 118 are likely to generate higher user engagement or other desired behaviors.

[0110] In some cases, the LLM 110 may employ policy optimization to iteratively improve its decision-making strategy by adjusting its parameters in response to the rewards or penalties received. This allows the model to leam from past experiences and refine its understanding of which actions (e.g., recommending specific types of content items or content item locations) lead to better user engagement. The learning process is dynamic, as the model continually adjusts based on new' data, ensuring that it adapts to changing user preferences and behaviors.

[0111] In some reinforcement learning frameworks, the fine-tuning can involve techniques such as Q-leaming or policy gradient methods, where the model leams to optimize the expected future rewards associated with each action it takes. These methods focus on improving the model’s decision-making by exploring different strategies for placing and generating the content item 118 and learning from both successes and failures. For example, if a certain ty pe of content item 118 consistently leads to higher conversion rates, the LLM 110 would adjust its policy to prioritize that content in future interactions.

[0112] By leveraging training data 214, the LLM 110 can be continuously improved to make better decisions regarding content item 118 placement, presentation, and recommendation. As new interaction data becomes available, the LLM 110 can be further finetuned, allowing the system to evolve and adapt to changing user dynamics.

[0113] Referring now to Figure 3, an exemplary flow diagram of a method for the dynamic generation of content items using a generative model.

[0114] At step 302, the method includes receiving, by a computing device, input data comprising user data, video data, and entity data, wherein the entity data comprises entityrepresentation data in accordance with embodiments of the present disclosure. In an embodiment, the user data may include a user profile.

[0115] At step 304, the method includes generating, by a large language model (LLM) operating on the computing device, a textual prompt based on the user data and the entity data in accordance with embodiments of the present disclosure. In some cases, the LLM may include a text-to-image generation model.

[0116] At step 306, the method includes identifying, by the LLM, one or more video scenes to present a content item based on the video data in accordance with embodiments of the present disclosure. In an embodiment, the method may include: processing, by the LLM, the video data to identity' a gameplay trigger; and identifying, by the LLM, the one or more video scenes based on the video data and the gameplay trigger. In a second embodiment, the method may include identify ing, using the LLM, one or more content item locations based on the one or more video scenes and the content item.

[0117] At step 308, the method includes processing, by the LLM, the textual prompt and the one or more video scenes to generate the content item in accordance with embodiments of the present disclosure. In an embodiment, generating the textual prompt may include: processing, using a prompt expansion model, the textual prompt to generate an expanded textual prompt; and generating the content item based on the expanded textual prompt. In a second embodiment, the method may include generating, by the LLM, the content item during a predefined temporal window, wherein the predefined temporal window occurs at least a threshold duration prior to a gameplay trigger; and storing, by the LLM, the content item in a cache memory' associated with the computing device.

[0118] At step 310, the method includes integrating, by the LLM, the content item into the one or more video scenes in accordance with embodiments of the present disclosure.

[0119] In some cases, the method may include receiving, by the LLM, interaction data based on the integrated content item; and generating, by the LLM, an interaction metric based on the interaction data. The interaction metric may include an interaction score, wherein the interaction score is associated with an engagement rate of the content items. In some cases, the method may include generating, by the LLM, an interaction report based on the interaction metric. In some embodiments, the method may include determining, by the LLM, a temporal characteristic of the integrated content item based on the interaction metric, wherein the temporal characteristic includes a time period associated with presenting the content item. In an additional embodiment, the method may include: determining, by the LLM, a content preference metric based on the interaction metric, wherein the content preference metriccomprises a quantification of a user’s preferences towards the content item; and generating, by the LLM, the content item based on the content preference metric.

[0120] The method may include generating, by the LLM, training data based on the interaction metric; and fine-tuning, by the computing device, the LLM using the training data, wherein the fine-tuning includes updating one or more parameters of the LLM. Fine-tuning the LLM may include fine-tuning the LLM using one or more reinforcement learning techniques. In some cases, the method may include generating a second content item based on the textual prompt.

[0121] Figure 4 depicts a flowchart of a method 400 for training one or more machine- learned models according to aspects of the present disclosure. For instance, an example machine-learned model can include LLM 110.

[0122] One or more portion(s) of example method 400 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 400 can be performed by any (or any combination) of one or more computing devices. Moreover, one or more portion(s) of example method 400 can be implemented on the hardware components of the device(s) described herein, for example, to train one or more systems or models. Figure 4 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 4 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 400 can be performed additionally, or alternatively, by other systems.

[0123] At 402, example method 400 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 400 as a “training” instance, it is to be understood that runtime inferences can form training instances when a model is trained using an evaluation of the model’s performance on that runtime instance (e.g., online training / leaming). Example data types for the training instance and various tasks associated therewith are described throughout the present disclosure.

[0124] At 404, example method 400 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.

[0125] At 406, example method 400 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).

[0126] At 408, example method 400 can include updating the machine-learned model using the evaluation signal. For example, values for parameters of the machine-learned model(s) can be learned, in some embodiments, using various training or learning techniques, such as, for example, backwards propagation. For example, the evaluation signal can be backpropagated from the output (or another source of the evaluation signal) through the machine-learned model(s) to update one or more parameters of the model(s) (e.g., based on a gradient of the evaluation signal with respect to the parameter value(s)). For example, system(s) containing one or more machine-learned models can be trained in an end-to-end manner. Gradient descent techniques can be used to iteratively update the parameters over a number of training iterations. In some implementations, performing backwards propagation of errors can include performing truncated backpropagation through time. Example method 400 can include implementing a number of generalization techniques (e.g., weight decays, dropouts, etc.) to improve the generalization capability of the models being trained.

[0127] In some implementations, example method 400 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.).

[0128] In some implementations, example method 400 can be implemented for particular stages of a training procedure. For instance, in some implementations, example method 400 can be implemented for pre-training a machine-learned model. Pre-training can include, forinstance, large-scale training over potentially noisy data to achieve a broad base of performance levels across a variety of tasks / data types.

[0129] In some implementations, example method 400 can be implemented for fine-tuning a machine-learned model. Fine-tuning can include, for instance, smaller-scale training on higher-quality (e.g., labeled, curated, etc.) data. Fine-tuning can affect all or a portion of the parameters of a machine-learned model. For example, various portions of the machine-learned model can be “frozen” for certain training stages. For example, parameters associated with an embedding space can be “frozen” during fine-tuning (e.g., to retain information learned from a broader domain(s) than present in the fine-tuning dataset(s)). In some implementations, example method 400 uses adapter modules. Adapters can be small trainable layers that are inserted between pre-existing layers of a pre-trained model. During the fine-tuning process, the original parameters of the pre-trained model are typically frozen, and only the parameters of the adapters are updated.

[0130] In some implementations, example method 400 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.

[0131] An example fine-tuning approach includes reinforcement learning. Reinforcement learning can be based on user feedback on model performance during use.

[0132] Figure 5 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.

[0133] Machine-learned model(s) 1 can be or include one or multiple machine-learned models or model components. Example machine-learned models can include neural networks (e.g., deep neural networks). Example machine-learned models can include non-linear models or linear models. Example machine-learned models can use other architectures in lieu of or in addition to neural networks. Example machine-learned models can include decision tree-based models, support vector machines, hidden Markov models, Bayesian networks, linear regression models, k-means clustering models, etc.

[0134] 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 berepresentative of LLM 110 or any other model discussed herein. 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 LLM 110 or any other machine-learned component described herein.

[0135] 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.

[0136] 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.

[0137] 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 effectively aggregating 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).

[0138] 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, ARXlV:2202.09368v2 (Oct. 14, 2022). For example, different portions of a model can leam (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 thetotal 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.

[0139] 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.

[0140] 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.

[0141] 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.

[0142] 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 ty pes 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.

[0143] Figure 6 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- etc. obtained from input(s) 2. Sequence processing model 4 can process input sequence 5 using prediction layer(s) 6 to generate an output sequence 7. Output sequence 7 can include one or more output elements 7-1, 7-2. . . . , 7-N, etc. generated based on input sequence 5. The system can generate output(s) 3 based on output sequence 7.

[0144] 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:2301.11325V1 (Jan. 26, 2023), biochemical domains, see, e.g., Jumper et al., Highly accurate protein structure prediction with AlphaFold, 596 Nature 583 (Aug. 26, 2021), by way of example. Sequence processing model(s) 4 can process one or multiple types of data simultaneously. Sequence processing model(s) 4 can include relatively large models (e.g., more parameters, computationally expensive, etc.), relatively small models (e.g., fewer parameters, computationally lightweight, etc ), or both.

[0145] 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”).

[0146] 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.

[0147] 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, theelements 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.

[0148] For example, elements 5-1, 5-2, . . . , 5-M can represent tokens obtained using a tokenizer. For instance, a tokenizer can process a given portion of an input source and output a series of tokens (e.g., corresponding to input elements 5-1, 5-2, . . . , 5-M) that represent the portion of the input source. Various approaches to tokenization can be used. For instance, textual input source(s) can be tokenized using a byte-pair encoding (BPE) technique. See, e.g., Kudo et al., SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processing, PROCEEDINGS OF THE 2018 CONFERENCE ON EMPIRICAL METHODS IN 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.

[0149] 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 6 can be the tokens or can be the embedded representations thereof.

[0150] Prediction layer(s) 6 can predict one or more output elements 7-1, 7-2, . . . , 7-N based on the input elements. Prediction layer(s) 6 can include one or more machine-learned model architectures, such as one or more layers of learned parameters that manipulate and transform the input(s) to extract higher-order meaning from, and relationships between, input element(s) 5-1 , 5-2, . . . , 5-M. In this manner, for instance, example prediction layer(s) 6 can predict new output element(s) in view of the context provided by input sequence 5.

[0151] Prediction layer(s) 6 can evaluate associations between portions of input sequence5 and a particular output element. These associations can inform a prediction of the likelihood that a particular output follows the input context. For example, consider the textual snippet, “The carpenter’s toolbox was small and heavy. It was full of .” Example prediction layer(s)6 can identify that “It” refers back to “toolbox” by determining a relationship between 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.”

[0152] A transformer is an example architecture that can be used in prediction layer(s) 4. See, e.g, Vaswani et al., Attention Is All You Need, ARXlV: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 contextwindow can include a sequence that contains input sequence 5 and potentially one or more output element(s) 7-1, 7-2, . . . . T-N. A transformer block can include one or more attention layer(s) and one or more post-attention layer(s) (e.g., feedforward layer(s), such as a multilayer perceptron).

[0153] 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.

[0154] 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.

[0155] 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.

[0156] Output sequence 7 can be generated autoregressively. For instance, for some applications, an output of one or more prediction layer(s) 6 can be passed through one or more output layers (e.g., softmax layer) to obtain a probability distribution over an output vocabulary (e.g., a textual or symbolic vocabulary) conditioned on a set of input elements in a context window. In this manner, for instance, output sequence 7 can be autoregressively generated by sampling a likely next output element, adding that element to the context window, and regenerating the probability distribution based on the updated context window, and sampling a likely next output element, and so forth.

[0157] Output sequence 7 can also be generated non-autoregressively. For instance, multiple output elements of output sequence 7 can be predicted together without explicitsequential conditioning on each other. See, e.g., Saharia et al., Non-Autoregressive Machine Translation with Latent Alignments. ARXIV:2004.07437V3 (Nov. 16, 2020).

[0158] 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.

[0159] Figure 7 is a block diagram of an example technique for populating an example input sequence 8. Input sequence 8 can include various functional elements that form part of the model infrastructure, such as an element 8-0 obtained from a task indicator 9 that signals to any model(s) that process input sequence 8 that a particular task is being performed (e g., to help adapt a performance of the model(s) to that particular task). Input sequence 8 can include various data elements from different data modalities. For instance, an input modality 10-1 can include one modality of data. A data-to-sequence model 11-1 can process data from input modality 10-1 to project the data into a format compatible with input sequence 8 (e.g., one or more vectors dimensioned according to the dimensions of input sequence 8) to obtain elements 8-1 , 8-2, 8-3. Another input modality 10-2 can include a different modality of data. A data-to- sequence model 11-2 can project data from input modality 10-2 into a format compatible with input sequence 8 to obtain elements 8-4, 8-5, 8-6. Another input modality 10-3 can include yet another different modality of data. A data-to-sequence model 11-3 can project data from input modality 10-3 into a format compatible with input sequence 8 to obtain elements 8-7, 8-8, 8-9.

[0160] Input sequence 8 can be the same as or different from input sequence 5. Input sequence 8 can be a multimodal input sequence that contains elements that represent data from different modalities using a common dimensional representation. For instance, an embedding space can have P dimensions. Input sequence 8 can be configured to contain a plurality’ of elements that have P dimensions. In this manner, for instance, example implementations can facilitate information extraction and reasoning across diverse data modalities by projecting data into elements in the same embedding space for comparison, combination, or other computations therebetween.

[0161] 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 datatypes can be broken down into continuously defined portions (e.g., image patches) that can be described using continuously distributed locations within the embedding space.

[0162] 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” wdiile 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.

[0163] 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 ty pe 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.

[0164] 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).

[0165] 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.).

[0166] 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.

[0167] Figure 8 is a block diagram of an example model development platform 12 that can facilitate creation, adaptation, and refinement of example machine-learned models (e.g.. machine-learned model(s) 1, sequence processing model(s) 4, etc ). Model development platform 12 can provide a number of different toolkits that developer systems can employ in the development of new or adapted machine-learned models.

[0168] 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.

[0169] 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.

[0170] 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.

[0171] 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).

[0172] 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.

[0173] 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., de-noising, 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.

[0174] Fine-tuning pipelines 17-3 can include a machine-learned model training workflow configured to refine the model parameters of development model 16 with higher-quality data. Fine-tuning pipelines 17-3 can update development model 16 by conducting supervised training with labeled dataset(s) in dataset(s) 17-1. Fine-tuning pipelines 17-3 can update development model 16 by conducting reinforcement learning using reward signals from user feedback signals. Workbench 15 can implement a fine-tuning pipeline 17-3 to fine-tune development model 16.

[0175] 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.

[0176] 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.

[0177] 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).

[0178] 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.

[0179] Prompt libraries 17-4 can include pipelines for prompt generation. For example, inputs can be generated using development model 16 itself or other machine-learned models. In this manner, for instance, a first model can process information about a task and output an input for a second model to process in order to perform a step of the task. The second model can be the same as or different from the first model. Workbench 15 can implement prompt generation pipelines in development model 16.

[0180] 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.

[0181] Although various training examples described herein with respect to model development platform 12 refer to ”pre-training“ and ‘Tine-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 400 described above.

[0182] 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 ofa 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.

[0183] 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”).

[0184] 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.

[0185] 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.

[0186] 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.

[0187] Model development platform 12 can include a computational optimization toolkit 19 for optimizing a computational performance of development model 16. For instance, toolsfor 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.

[0188] Workbench 15 can implement one. multiple, or none of the toolkits implemented in model development platform 12. Workbench 15 can output an output model 20 based on development model 16. Output model 20 can be a deployment version of development model 16. Output model 20 can be a development or training checkpoint of development model 16. Output model 20 can be a distilled, compressed, or otherwise optimized version of development model 16.

[0189] Figure 9 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. Figure 9 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 9 is described with reference to elements / terms described with respect to other systems and figures for exemplary illustratedpurposes 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.

[0190] 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 pretraining for the same or for a different model.

[0191] Initialized model 21 can undergo pre-training in a pre-training stage 22. Pre-training 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 pretrained (e.g., development model 16 contains, is, or is based on a pre-trained foundational model or an expert model).

[0192] 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.

[0193] Fine-tuned model 29 can then be a new version of development model 16, which can persist as development model 16 or as anew development model. Fine-tuned model 29 can be the initial state if development model 16 was already fine-tuned. Fine-tuned model 29 can undergo refinement with user feedback 26. For instance, refinement with user feedback 26 can include reinforcement learning, optionally based on human feedback from human users of finetuned model 25. As reinforcement learning can be a form of fine-tuning, it is to be understood that fine-tuning stage 24 can subsume the stage for refining with user feedback 26. Refinement with user feedback 26 can produce a refined model 27. Refined model 27 can be output to downstream system(s) 28 for deployment or further development.

[0194] In some implementations, computational optimization operations can be applied before, during, or after each stage. For instance, initialized model 21 can undergo computational optimization 29-1 (e.g., using computational optimization toolkit 19) before pretraining stage 22. Pre-trained model 23 can undergo computational optimization 29-2 (e.g., using computational optimization toolkit 19) before fine-tuning stage 24. Fine-tuned model 25 can undergo computational optimization 29-3 (e.g.. using computational optimization toolkit 19) before refinement with user feedback 26. Refined model 27 can undergo computationaloptimization 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.

[0195] Figure 10 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.

[0196] 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.

[0197] 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.

[0198] 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.

[0199] For example, model host 31 can operate on a server system that provides a machinelearning 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.

[0200] In some implementations, model host 31 can operate on a same device or system as chent(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.

[0201] Model instance(s) 31-1 can include one or more machine-learned models that are available for performing inference. Model instance(s) 31-1 can include weights or other model components that are stored on in persistent storage, temporarily cached, or loaded into highspeed memory. Model instance(s) 31-1 can include multiple instance(s) of the same model (e.g., for parallel execution of more requests on the same model). Model instance(s) 31-1 can include instance(s) of different model(s). Model instance(s) 31-1 can include cached intermediate states of active or inactive model(s) used to accelerate inference of those models. For instance, an inference session with a particular model may generate significant amounts of computational results that can be re-used for future inference runs (e.g., using a KV cache for transformer-based models). These computational results can be saved in association with that inference session so that session can be executed more efficiently when resumed.

[0202] 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.

[0203] 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.

[0204] 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.

[0205] 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.

[0206] 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.

[0207] 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.

[0208] Model host 31 can execute machine-learned model (s) 1 to perform inference for various tasks using various types of data. For example, various different input(s) 2 and output(s)3 can be used for various different tasks. In some implementations, input(s) 2 can be orotherwise represent image data. Machine-learned model(s) 1 can process the image data to generate an output. As an example, machine-learned model(s) 1 can process the image data to generate an image recognition output (e.g., a recognition of the image data, a latent embedding of the image data, an encoded representation of the image data, a hash of the image data, etc.). As another example, machine-learned model(s) 1 can process the image data to generate an image segmentation output. As another example, machine-learned model(s) 1 can process the image data to generate an image classification output. As another example, machine-learned model(s) 1 can process the image data to generate an image data modification output (e.g., an alteration of the image data, etc.). As another example, machine-learned model(s) 1 can process the image data to generate an encoded image data output (e.g., an encoded and / or compressed representation of the image data, etc.). As another example, machine-learned model(s) 1 can process the image data to generate an upscaled image data output. As another example, machine-learned model(s) 1 can process the image data to generate a prediction output.

[0209] 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.

[0210] 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 thenatural language data to generate alatent 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).

[0211] 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.

[0212] 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.

[0213] In some implementations, input(s) 2 can be or otherwise represent statistical data. Statistical data can be, represent, or otherwise include data computed and / or calculated from some other data source. Machine-learned model(s) 1 can process the statistical data to generate an output. As an example, machine-learned model(s) 1 can process the statistical data to generate a recognition output. As another example, machine-learned model(s) 1 can process the statistical data to generate a prediction output. As another example, machine-learned model(s) 1 can process the statistical data to generate a classification output. As another example, machine-learned model(s) 1 can process the statistical data to generate a segmentation output. As another example, machine-learned model(s) 1 can process the statistical data to generate a visualization output. As another example, machine-learned model(s) 1 can process the statistical data to generate a diagnostic output.

[0214] 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.

[0215] 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.

[0216] 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.

[0217] In some implementations, the task can be a text completion task. Machine-learned model(s) 1 can be configured to process input(s) 2 that represent textual data and to generate output(s) 3 that represent additional textual data that completes a textual sequence that includes input(s) 2. For instance, machine-learned model(s) 1 can be configured to generate output(s) 3 to complete a sentence, paragraph, or portion of text that follows from a portion of text represented by input(s) 2.

[0218] 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 atask 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.

[0219] 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 dataof the same or of a different modality as input(s) 2. For instance, input(s) 2 can represent textual data (e.g., natural language instructions for a task to be performed) and machine-learned model(s) 1 can process input(s) 2 to generate output(s) 3 that represent textual data responsive to the question (e.g., natural language responses, programming language responses, machine language responses, etc.). Input(s) 2 can represent image data (e g., image-based instructions for a task to be performed, optionally accompanied by textual instructions) and machine- learned model(s) 1 can process input(s) 2 to generate output(s) 3 that represent textual data responsive to the question (e.g., natural language responses, programming language responses, machine language responses, etc.). One or more output(s) 3 can be iteratively or recursively generated to sequentially process and accomplish steps toward answering the question. For instance, an initial output can be executed by an external system or be processed by machine- learned model(s) 1 to complete an initial step of obtaining an answer to the question (e.g., querying a database, performing a computation, executing a script, etc.). Multiple steps can be performed, with a final output being obtained that is responsive to the question.

[0220] 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).

[0221] 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).

[0222] 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 desiredportion 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. V alues for the data obj ect(s) can be selected based on the context (e.g., based on a probability determined based on the context).

[0223] Figure 11 is a block diagram of an example networked computing system that can perform aspects of example implementations of the present disclosure. The system can include a number of computing devices and systems that are communicatively coupled over a network 49. An example computing device 50 is described to provide an example of a computing device that can perform any aspect of the 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 ).

[0224] 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 11 can be co-located with, contained by, or otherwise integrated into one or more other devices or systems.

[0225] Computing device 50 can be any t pe 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 mayuse another computing device to interact with computing device 50).

[0226] 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., aprocessor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. Memory 52 can include one or more non- transitory computer-readable storage media, such as HBM, RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. Memory 52 can store data 53 and instructions 54 which can be executed by processor(s) 51 to cause computing device 50 to perform operations. The operations can implement any one or multiple features described herein. The operations can implement example methods and techniques described herein.

[0227] 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.

[0228] 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 part}- 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.

[0229] 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 oneprocessor 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 anyone or multiple features described herein. The operations can implement example methods and techniques described herein.

[0230] 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.

[0231] 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.

[0232] 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 serv ice (e.g., remote machine-learned model hosting service, such as an online interface for performing machine-1 earned 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 trainingoperations), 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.

[0233] Model development platform system(s) 70 can include one or more processors 71 and a memory 72. Processor(s) 71 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. Memory 72 can include one or more non-transitory computer-readable storage media, such as HBM, RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. Memory 72 can store data 73 and instructions 74 which can be executed by processor(s) 71 to cause model development platform system(s) 70 to perform operations. The operations can implement any one or multiple features described herein. The operations can implement example methods and techniques described herein. Example operations include the functionality described herein with respect to model development platform 12. This and other functionality can be implemented by developer tool(s) 75.

[0234] 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 plurality7of 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-party7system(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).

[0235] Figure 11 illustrates 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 thereol) to develop, update / train, or refine machine-learned models 1, 4, 16, 20, 55, 65, etc. using one or more techniques described hereinwith 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).

[0236] Figure 12 is a block diagram of an example computing device 98 that performs according to example embodiments of the present disclosure. Computing device 98 can be a user computing device or a server computing device (e.g., computing device 50. server computing system(s) 60, etc.). Computing device 98 can implement model host 31. For instance, computing device 98 can include a number of applications (e.g., applications 1 through N). Each application can contain its own machine learning library and machine-learned model(s). 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 12, 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.

[0237] Figure 13 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).

[0238] The central intelligence layer can include anumber of machine-learned models. For example, as illustrated in Figure 13, 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 theapplications. In some implementations, the central intelligence layer is included within or otherwise implemented by an operating system of computing device 99.

[0239] 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 13, the central device data layer can communicate with a number of other components of the computing device, such as, for example, one or more sensors, a context manager, a device state component, 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).

[0240] 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.

[0241] 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.

[0242] 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 aredescribed herein using lists of example elements joined by conjunctions such as "and." “or,” “but,” etc. It should be understood that such conjunctions are provided for explanatory purposes only. Clauses and other sequences of items joined by a particular conjunction such as “or,” for example, can refer to “and / or,” “at least one of’, “any combination of’ example elements listed therein, etc. Terms such as “based on” should be understood as “based at least in part on.”

[0243] 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.

[0244] 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.

[0245] This written description uses examples to disclose the invention, including the best mode, and also to enable any person skilled in the art to practice the invention, including making and using any devices or systems and performing any incorporated methods. The patentable scope of the invention is defined by the claims, and may include other examples that occur to those skilled in the art. Such other examples are intended to be within the scope of the claims if they include structural elements that do not differ from the literal language of the claims, or if they include equivalent structural elements with insubstantial differences from the literal language of the claims.

Claims

1. WHAT IS CLAIMED IS:

1. A computer-implemented method, comprising: receiving, by a computing device, input data comprising user data, video data, and entity data, wherein the entity data comprises entity representation data; generating, by a large language model (LLM) operating on the computing device, a textual prompt based on the user data and the entity data; identifying, by the LLM, one or more video scenes to present a content item based on the video data; processing, by the LLM, the textual prompt and the one or more video scenes to generate the content item; and integrating, by the LLM, the content item into the one or more video scenes.

2. The computer-implemented method of claim 1, wherein the method further comprises: processing, by the LLM, the video data to identify a gameplay trigger; and identifying, by the LLM, the one or more video scenes based on the video data and the gameplay trigger.

3. The computer-implemented method of claim 1, wherein the method further comprises: receiving, by the LLM, interaction data based on the integrated content item; and generating, by the LLM, an interaction metric based on the interaction data.

4. The computer-implemented method of claim 3, wherein the interaction metric comprises an interaction score, wherein the interaction score is associated with an engagement rate of the content items.

5. The computer-implemented method of claim 3, wherein the method further comprises generating, by the LLM, an interaction report based on the interaction metric.

6. The computer-implemented method of claim 3, wherein the method further comprises determining, by the LLM, a temporal characteristic of the integrated content itembased on the interaction metric, wherein the temporal characteristic includes a time period associated with presenting the content item.

7. The computer-implemented method of claim 3, wherein the method further comprises: determining, by the LLM, a content preference metric based on the interaction metric, wherein the content preference metric comprises a quantification of a user’s preferences towards the content item; and generating, by the LLM, the content item based on the content preference metric.

8. The computer-implemented method of claim 3, wherein the method further comprises: generating, by the LLM, training data based on the interaction metric; and fine-tuning, by the computing device, the LLM using the training data, wherein the fine-tuning includes updating one or more parameters of the LLM.

9. The computer-implemented method of claim 8, wherein the method further comprises generating, by the fine-tuned LLM, a second content item based on the textual prompt.

10. The computer-implemented method of claim 8, wherein fine-tuning the LLM comprises fine-tuning the LLM using one or more reinforcement learning techniques.

11. The computer-implemented method of claim 1, wherein the LLM comprises a text-to-image generation model.

12. The computer-implemented method of claim 1, wherein generating the textual prompt further comprises: processing, using a prompt expansion model, the textual prompt to generate an expanded textual prompt; and generating the content item based on the expanded textual prompt.

13. The computer-implemented method of claim 1, wherein the user data comprises a user profile.

14. The computer-implemented method of claim 1, wherein the method further comprises: generating, by the LLM, the content item during a predefined temporal window, wherein the predefined temporal window occurs at least a threshold duration prior to a gameplay trigger; and storing, by the LLM, the content item in a cache memory associated with the computing device.

15. The computer-implemented method of claim 1, wherein the method further comprises: identifying, by the LLM. the one or more video scenes based on the video data and a gameplay trigger; generating, by the LLM, the content item during a predefined temporal window based on the one or more video scenes and the gameplay trigger, wherein the predefined temporal window occurs at least a threshold duration prior to the gameplay trigger; and storing, by the LLM. the content item in a cache memory associated with the computing device.

16. The computer-implemented method of claim 1, wherein the method further comprises identifying, using the LLM, one or more content item locations based on the one or more video scenes and the content item.

17. The computer-implemented method of claim 1, wherein the video data comprises gameplay data.

18. A computing system, comprising: one or more processors; and one or more transitory7or non-transitory computer-readable media storing instructions that are executable to cause the one or more processors to perform operations, the operations comprising: receiving input data comprising user data, video data, and entity data, wherein the entity data comprises entity representation data; generating, by a large language model (LLM), a textual prompt based on the user data and the entity data; processing, by the LLM, the video data to identify a gameplay trigger;identifying, by the LLM, one or more video scenes to present a content item based on the video data and the gameplay trigger; processing, by the LLM, the textual prompt and the one or more video scenes to generate the content item; identify, using the LLM, one or more content item locations based on the one or more video scenes and the content item; and integrating, by the LLM, the content item into the one or more video scenes based on the one or more content item locations.

19. The computing system of claim 18, wherein the operations further comprise: generating, by the LLM, the content item during a predefined temporal window based on the one or more video scenes and the gameplay trigger, wherein the predefined temporal window occurs at least a threshold duration prior to the gameplay trigger; and storing, by the LLM, the content item in a cache memory associated with the computing device.

20. A computer-implemented method, comprising: receiving, by a computing device, input data comprising user data, video data, and entity data, wherein the entity data comprises entity representation data; generating, by a large language model (LLM) operating on the computing device, a textual prompt based on the user data and the entity data; identify ing, by the LLM, one or more video scenes to present a content item based on the video data; processing, by the LLM, the textual prompt and the one or more video scenes to generate the content item; integrating, by the LLM, the content item into the one or more video scenes; receiving, by the LLM, interaction data based on the integrated content item; generating, by the LLM, an interaction metric based on the interaction data; and determining, by the LLM, a temporal characteristic of the integrated content item based on the interaction metric, wherein the temporal characteristic includes a time period associated with presenting the content item.

Images