A log owner of a database system may receive a request from a log tailer of the database system to log a database operation in a transaction log. The log owner maintains the transaction log and the log tailer replays the transaction log. The request includes a record resulting from the database operation, and the record includes lineage information identifying a first previous version of the record observed by the log tailer when generating the record. The log owner then detects whether a record violation has occurred where a second previous version of the record has been committed that is a newer version than the first previous version. In response to detecting that the record violation occurred, the log owner prevents the database operation from being logged in the transaction log and may return an error to the log tailer that pertains to the record violation.
Systems, devices, and techniques are disclosed for management automation using database signals. An API call requesting actionable signals may be sent to a service of a database. An API response including the actionable signals may be received. The actionable signals may include indications of a subject and are generated from performance metrics generated from data stored in the database. Subjects may be selected from the actionable signals. An API call including the selected subjects may be sent to a generative system. Generated items based on the selected subjects may be received from the generative system. The subjects may include products or product categories sold by an online store. The generated items may include promotions for the products of the selected subjects.
A large language model (LLM) service for controlling LLM agents via an LLM may include obtaining, via a user interface, instructions for a first LLM, that is configured for agent evaluation, to evaluate inputs and outputs of LLM agents. The LLM service may obtain an input to an LLM agent. The LLM service may monitor the input to the LLM agent to evaluate whether the input to the LLM agent is in accordance with the instructions. The LLM service may output, to the LLM agent, the input based on evaluation of the input. The LLM service may obtain, from the LLM agent, an output in response to the input and monitor the output of the LLM agent to evaluate whether the output from the LLM agent is in accordance with the instructions. The LLM service may output the output from the LLM agent based on evaluation of the output.
Methods and corresponding systems and apparatuses for building customized applications from a base application are described. For example, source code for a base application and one or more assets that are required by the base application may be obtained. A build cache generated by a build automation tool upon building the base application may be obtained. A containerized software image file for the base application may be generated based at least in part on the source code, the one or more assets, and the build cache. One or more customizations to be applied to the base application may be determined. A customized version of the base application may be generated based at least in part on the containerized software image file for the base application and the one or more customizations.
A system may generate synthetic data via an iterative procedure that improves diversity and accuracy of the synthetic data by evaluating performance metrics of respective portions of the generated synthetic data prior to generating subsequent portions of the synthetic data. The system may receive an input indicating seed data and a task description and, based on the seed data and the task description, retrieve data from a data source. The system may generate first examples from the retrieved data, evaluate performance of the seed data based on the generated first examples, and weight the seed data accordingly. Using the weighted seed data and the first examples, the system may retrieve new data from the data source and perform another iteration of example generation and evaluation. The system may continue to perform iterations of example generation and evaluation to generate an amount of data requested as indicated by the input.
Embodiments described herein provide a method of configuring an artificial intelligence (AI) agent to respond to a user query. The method includes: receiving a user query; generating, by a first neural network based language model, a response to the user query; generating, by the first neural network based language model, a summary of an interaction history; and training, a second neural network based language model, using a dataset including the summary and the interaction history to generate a rating of the summary, an explanation of the rating, and a citation in the interaction history supporting the explanation conditioned on the summary and the interaction history in response to a training query. The method also includes: building, at a server, the AI agent through a first application programming interface (API) to the first neural network based language model and a second API to the trained second neural network based language model.
Embodiments described herein provide a training framework that trains an LLM to generate long reasoning by using synthesized reasoning data from a LLM with shorter reasoning, all without knowledge distillation from a long-reasoning ready LLM. Specifically, an LLM that is not pretrained with long reasoning data may generate multiple sample responses to an input prompt combining an input query and a set of long-reasoning examples as context to the input query. The generated sample responses each includes an associated long reasoning sequence and a respective final answer. The training framework filters the generated sample responses using an outcome reward model depending on accuracy of the final answers. The training framework then trains LLM using the filtered generated sample responses as training data for long reasoning. The training framework may refine the LLM by online training using new unseen queries.
Systems and methods are provided for receiving a first request of a plurality of requests at a first database server node for a first tenant of a plurality of database server nodes. The first database server node may reject the first request when it is determined to have a probability above a predetermined threshold. The rejected first request may be transmitted to a second database server node of the plurality of database server nodes.
H04L 67/1008 - Sélection du serveur pour la répartition de charge basée sur les paramètres des serveurs, p. ex. la mémoire disponible ou la charge de travail
G06F 16/27 - Réplication, distribution ou synchronisation de données entre bases de données ou dans un système de bases de données distribuéesArchitectures de systèmes de bases de données distribuées à cet effet
9.
SYSTEMS AND METHODS FOR VERIFYING TEXT-GENERATING NEURAL NETWORK MODELS
Embodiments described herein provide a method of configuring an artificial intelligence (AI) agent to respond to a user query. The method includes: receiving a user query; generating, by a first neural network based language model, a response to the user query; generating, by the first neural network based language model, a summary of an interaction history; and training, a second neural network based language model, using a dataset including the summary and the interaction history to generate a rating of the summary, an explanation of the rating, and a citation in the interaction history supporting the explanation conditioned on the summary and the interaction history in response to a training query. The method also includes: building, at a server, the AI agent through a first application programming interface (API) to the first neural network based language model and a second API to the trained second neural network based language model.
Systems, devices, and techniques are disclosed for response caching for machine learning systems. A vector including data and time values may be received. A point of interest, including data over a range of time values, in the vector may be determined. A request including the vector, the range of time values, and a request question may be sent to a machine learning system. A text response may be received from the machine learning system. The text response, a vector identifier that identifies the vector, and an index indicating the range of time values may be stored in a cache as a cached response. A user request including an indication of data of interest and an indication of a time period of interest may be received from a user. The text response stored in the cache may be determined to be responsive to the user request.
Techniques are disclosed for managing contention and improving transaction efficiency in distributed database systems through snapshot advancement. A system receives a query defining a first database transaction associated with a first snapshot of a database, where the transaction includes reading and updating a database value. The system evaluates the query against predefined criteria, such as accessing a single database row and deterministically locating the database value, to determine eligibility for snapshot advancement. Upon identifying contention over the database value between the first transaction and a second transaction, the system dynamically advances the transaction to a second snapshot of the database, avoiding a restart. This process involves granting snapshot advancement permission, waiting for a lock held by the second transaction to be released, and using the subsequent snapshot to continue execution. These techniques reduce transaction latency and resource overhead while maintaining consistency and efficiency in high-throughput database environments.
G06F 11/14 - Détection ou correction d'erreur dans les données par redondance dans les opérations, p. ex. en utilisant différentes séquences d'opérations aboutissant au même résultat
12.
SYSTEMS AND METHODS FOR A TESTING PIPELINE OF CODE GENERATION ARTIFICIAL INTELLIGENCE AGENTS
Embodiments described herein provide a testing pipelines for LLM-based software engineering agents. For example, the testing pipeline comprise a variety of tasks to evaluate a software engineering agent's abilities in crafting test cases under specific requirements, ranging from relatively simple ones, such as completing assertion statements, to more complex challenges, like writing test cases to cover specific code blocks across up to ten files. The testing pipeline comprises a data collection module that collects test programs, and a testing environment construction module that run the agent-generated programs on different repositories.
A system may monitor behavior of first web crawling bots traversing a plurality of web pages to generate behavior information associated with the plurality of first web crawling bots. The system may store at least a portion of the behavior information in a behavior matrix. The system may receive plain language configuration instructions that describe one or more rules for allocating processing resources associated with serving the plurality of web pages. The system may detect a presence of a second web crawling bot accessing one or more first web pages of the plurality of web pages. The system may allocate the processing resources for the plurality of web pages in accordance with the one or more rules and based at least in part on detecting the presence of the second web crawling bot.
Disclosed are some implementations of systems, apparatus, methods and computer program products for facilitating the authentication of computing system requests across tenants of at least one multi-tenant database system. Authentication is facilitated using a central registry that is accessible by and independent from the tenants of the multi-tenant database system.
H04L 9/30 - Clé publique, c.-à-d. l'algorithme de chiffrement étant impossible à inverser par ordinateur et les clés de chiffrement des utilisateurs n'exigeant pas le secret
H04L 9/32 - Dispositions pour les communications secrètes ou protégéesProtocoles réseaux de sécurité comprenant des moyens pour vérifier l'identité ou l'autorisation d'un utilisateur du système
15.
MECHANISMS FOR HANDLING WRITES IN DISTRIBUTED DATABASE SYSTEM
Techniques are disclosed for processing record writes in a database system having a log owner and one or more log tailers. The log owner maintains a transaction log, and the log tailers replay the transaction log to recreate a state of the log owner. A log tailer may receive a write request to perform a record write operation to write a base record to a database as part of a database transaction. The log tailer may determine additional records to write as a result of having to write the base record. The log tailer buffers the base record and additional records in a memory buffer and issues them as a single batch to the log owner to update the transaction log. The log owner may update the transaction log and return a response to the log tailer that indicates whether the log owner successfully updated the transaction log.
G06F 16/27 - Réplication, distribution ou synchronisation de données entre bases de données ou dans un système de bases de données distribuéesArchitectures de systèmes de bases de données distribuées à cet effet
16.
APPROACHES FOR DISASTER DETECTION AND RECOVERY OF CLOUD INFRASTRUCTURE AND SERVICES
Methods and corresponding systems and apparatuses for disaster detection and recovery of cloud infrastructure are described. Internal telemetry data corresponding to a plurality of levels of a cloud infrastructure may be obtained. At least one outage to at least one of the plurality of levels of the cloud infrastructure may be predicted. A threshold likelihood of at least one outage to the cloud infrastructure may be determined based at least in part on the at least one predicted outage. In response to determining the threshold likelihood of the at least one outage, an automated disaster recovery process to failover the at least one of the plurality of levels of the cloud infrastructure may be initiated .
H04L 41/0663 - Gestion des fautes, des événements, des alarmes ou des notifications en utilisant la reprise sur incident de réseau en réalisant des actions prédéfinies par la planification du basculement, p. ex. en passant à des éléments de réseau de secours
H04L 41/0604 - Gestion des fautes, des événements, des alarmes ou des notifications en utilisant du filtrage, p. ex. la réduction de l’information en utilisant la priorité, les types d’éléments, la position ou le temps
H04L 41/16 - Dispositions pour la maintenance, l’administration ou la gestion des réseaux de commutation de données, p. ex. des réseaux de commutation de paquets en utilisant l'apprentissage automatique ou l'intelligence artificielle
Techniques are disclosed for maintaining a database storing application data in a schema using schema layering. In some embodiments, a database system receives, from an instance of an application during the maintaining, a request to access application data stored in the database. The system may access schema overlays prior to accessing a base schema layer of the schema, where recent versions of objects stored in the schema overlays take precedence over corresponding prior versions of the objects stored in the base schema layer. After accessing the schema overlays, the system may access the base schema layer to retrieve objects that are not included in the schema overlays. The system may transmit, to the instance of the application, a notification indicating results of accessing the objects stored in the one or more schema overlays and the base schema layer.
Disclosed herein are system, method, and computer program product aspects for automatically generating recommended actions for a user. During operation, many different data points are examined, including user actions, calendar data, company guidelines, etc. A machine-learning model is used to analyze these data points and determine any number of recommended actions for the users. Such actions may include any known actions to be carried out in the course of the user’s business, such as following up with a customer, drafting an introductory email, submitting expense reports, etc. Once the recommended actions have been identified, they are provided to the user in a chat interface or list interface, where the user can view, select, edit, or authorize such actions. The model can then execute authorized actions as needed.
A computing services environment may include application servers providing computing services to a plurality of entities, a unified metadata framework including autonomous agent definitions referencing action definitions defining actions capable of being performed within the computing services environment, an agent studio configured to create a personal autonomous agent definition specific to a user account, an agent service configured to instantiate an autonomous agent instance based on a corresponding autonomous agent definition of the plurality of autonomous agent definitions, and an orchestration layer configured to determine an orchestration plan based on novel planning text generated by a generative language model. The orchestration plan may include a subset of the plurality of actions identified in the novel planning text. The computing services environment may be configured to execute the subset of the plurality of actions.
A large language model (LLM) may receive a query via a user interface of a client device. The LLM may generate one or more data queries from the query to query one or more data sets. The LLM may then transmit the one or more data queries to the one or more data sets. The LLM may then receive information associated with the query and a source for the information from the one or more data sets. The source may be indicative of a location within the one or more data sets from where the information was obtained. Following, the LLM may generate a response to the query that includes the information associated with the query and the source for the information and transmit the response to the user interface of the client device for display.
A user may input text or select a suggested prompt via a user interface. Based on the text or the suggested prompt, one or more models may identify a responsive topic and different aspects or categories of the topic. The aspects/categories may be represented as tiles that include content associated with the tile. The tiles and the associated content may be dynamically rendered in a canvas virtual space that is accessible to the user. The canvas virtual space may be dynamically updated based on requests from the user and/or the identification of additional data that is deemed to be relevant to the topic. As a result, the user can view relevant an up to date content that is responsive to their inputted text or suggested prompt via the canvas virtual space, without having to navigate to different resources (e.g., websites, internal documents, external documents, etc.) to obtain such information.
G06F 3/04845 - Techniques d’interaction fondées sur les interfaces utilisateur graphiques [GUI] pour la commande de fonctions ou d’opérations spécifiques, p. ex. sélection ou transformation d’un objet, d’une image ou d’un élément de texte affiché, détermination d’une valeur de paramètre ou sélection d’une plage de valeurs pour la transformation d’images, p. ex. glissement, rotation, agrandissement ou changement de couleur
G06F 3/0482 - Interaction avec des listes d’éléments sélectionnables, p. ex. des menus
24.
GENERATING CITATIONS FOR LARGE LANGUAGE MODEL (LLM) RESPONSES
A computer-implemented method for generating citations for responses generated by a large language model (LLM). The method includes receiving a prompt from a user, sending the prompt to the LLM, receiving a response from the LLM, and receiving a context source description of the LLM response from a data service. The context source description includes an URL and an identifier corresponding to the LLM response. The method also includes segmenting the LLM response into semantic segments, and generating citation descriptor text strings corresponding to the segments. The text strings identify relevant portions of the context source description as source references for the segments. Accuracy of the source references are validated. The method further includes arranging the validated text strings in a citation reporting statement having a predetermined syntax and format based on the semantic segments, and returning the citation reporting statement to the user.
Users of a cloud platform may utilize a large language model (LLM) service. The service may configure a first and second agent with a first and second prompt and a third agent with the first prompt at the first agent. The service may obtain, from the first agent, a message with content generated via the first agent. The service may obtain, from the third agent, a control message that includes an indication associated with the content of the message. Further, in response to obtaining the control message and the control message indication, the service may switch from applying one set of parameter weights to an LLM associated the first agent to applying another set of parameter weights. Moreover, the service may output, to the second agent the message from the first agent and then the service may switch back to applying the initial set of parameter weights to the LLM.
Processor-implemented methods and systems are disclosed for interacting with multiple artificial intelligence (AI) agents. Logging data is generated that captures actions the AI agents take. The logging data includes data inputs, decision processes, and final decisions of the AI agents. Primary data features that guided the final decisions of the AI agents are determined. Control of the AI agents is in response to human-understandable explanations of the AI agents'final decisions.
Users of a cloud platform may utilize a large language model (LLM) service. The service may configure a first and second agent with a first and second prompt and a third agent with the first prompt at the first agent. The service may obtain, from the first agent, a message with content generated via the first agent. The service may obtain, from the third agent, a control message that includes an indication associated with the content of the message. Further, in response to obtaining the control message and the control message indication, the service may switch from applying one set of parameter weights to an LLM associated the first agent to applying another set of parameter weights. Moreover, the service may output, to the second agent the message from the first agent and then the service may switch back to applying the initial set of parameter weights to the LLM.
A computer-implemented method of providing targeted content to a user includes generating a query index from a data corpus, the query index including a plurality of market segment-based queries, wherein each market segment-based query of the plurality of queries is configured to provide targeted content on a browser user interface of a user determined to be within a corresponding market segment. The method further includes constructing the browser-executable library including the query index, where the browser-executable library is configured to execute within a local machine browser of the user, and transmitting the browser-executable library to the local machine browser of the user, wherein the browser-executable library is configured to determine that a query of the plurality of market segment-based queries matches user-specific data only stored in the local machine browser of the user, where the query matching the user-specific data stored in the local machine browser of the user is configured to cause the local machine browser to request the targeted content corresponding to the user-specific data.
A system may include a generative language model platform facilitating generation of novel text via a generative language model. The system may include a communication gateway communicating with client systems located outside of the computing services environment and within a first geographic region. The system may include a feedback service configured to write generative language model logging input data to a storage location within the geographic region. The system may include a data service instance configured to store database records determined based on generative language model logging input data retrieved from the storage location.
Systems and methods are provided for generating, at a server, a product digital twin that is a virtual representation of a product, where the product digital twin is dynamically updatable. A content digital twin may be generated that is a virtual representation of content for the product that is dynamically updatable. The server may receive data from at least one source, and generating updates for the product digital twin and/or the content digital twin based on the received data. The generated updates may be transmitted to a plurality of different platforms.
Systems, devices, and techniques are disclosed for minimal cache generation for cloud substrates. Records may be generated with security endpoint tools of a substrate of a cloud computing server system. Some of the records may include data identifying security violations found in applications of the substrate. A minimal cache including records may be generated based on the records including data identifying security violations. The records of the minimal cache may be smaller than the records generated with the endpoint security tools. The minimal cache may be sent to a second substrate. A compliance tracker of the second substrate may determine a security violation in an application of the second substrate based on the data identifying security violations from records of the minimal cache. The compliance tracker of the second substrate may perform an action to remediate the security violation in the application of the second substrate.
G06F 12/0802 - Adressage d’un niveau de mémoire dans lequel l’accès aux données ou aux blocs de données désirés nécessite des moyens d’adressage associatif, p. ex. mémoires cache
G06F 12/14 - Protection contre l'utilisation non autorisée de mémoire
Systems, devices, and techniques are disclosed for content delivery network configuration. An API request may include a description in an API language may be received. The description may include HTTP protocols, methods, header policies, certificate data, origins, and routing rules. Asynchronous processes may configure a first content delivery network based on the description. One of the asynchronous processes may interact with a certificate service. Additional asynchronous processes may configure a second content delivery network based on the description. The description may not be changed based on configuration parameters of the second content delivery network. The configuration parameters of the second content delivery network may be different from configuration parameters of the first content delivery network.
An interactive agent (e.g., a chatbot, an artificially intelligent assistant, a digital assistant, etc.) may utilize a dynamic button of a user interface to output a request suggestion the user can request while conversing with the interactive agent. The interactive agent may utilize personal account and interaction history in order to provide relevant request suggestions that the interactive agent can provide a response to when the dynamic button is selected. The dynamic buttons may be dynamically updated as the conversation progresses between the interactive agent and the user in order to provide relevant request suggestions to the user based on the conversation.
G06F 3/04842 - Sélection des objets affichés ou des éléments de texte affichés
G06F 9/451 - Dispositions d’exécution pour interfaces utilisateur
H04L 51/02 - Messagerie d'utilisateur à utilisateur dans des réseaux à commutation de paquets, transmise selon des protocoles de stockage et de retransmission ou en temps réel, p. ex. courriel en utilisant des réactions automatiques ou la délégation par l’utilisateur, p. ex. des réponses automatiques ou des messages générés par un agent conversationnel
34.
Systems and Methods for Generative Creation of an Autonomous Agent In a Database System
A computing services environment may include an agent creator configured to determine a plurality of metadata entries by analyzing flowchart description information via a generative language model. The computing services environment may also include a database system storing autonomous agent definition information defining an autonomous agent and referencing the plurality of metadata entries. The computing services environment may also include an orchestration engine configured to instantiate an instance of the autonomous agent based on the autonomous agent definition information. The computing services environment may also include an agent platform configured to execute the plurality of actions including retrieving data via the data retrieval action and updating information stored in the database system.
A computing services environment comprising may include a database system storing a plurality of database records for a plurality of client organizations accessing computing services, and may include a graphics rendering system configured to generate a graphics rendering environment comprising a simulation of robots included in a manufacturing environment. The computing services environment may also include an agent platform including metadata entries corresponding to a plurality of agent instances, and may also include a generative language model interface providing access to one or more generative language models. The computing services environment may additionally include an orchestration and planning service configured to instantiate and execute one or more agent instances associated with the graphics rendering environment, where executing one or more agent instances includes determining one or more topics based on an output of the graphics rendering system or the manufacturing environment.
A customer relationship management (CRM) system can leverage a generative machine learned model to extract customer insights from conversation data. In some examples, the generative machine learned model may be a large language model (LLM) that is trained to extract customer contact reasons from data associated with different communication channel types. The system may generate input data to input into the LLM. When generating the input data, the system can identify non-essential conversation data associated with the different communication channel types. Responsive to a prompt, the LLM may disregard the non-essential conversation data and extract a primary contact reason consistently across the different communication channels.
A customer relationship management (CRM) system can leverage a generative machine learned model to extract customer insights from conversation data. In some examples, the generative machine learned model may be a large language model (LLM) that is trained to extract customer contact reasons from data associated with different communication channel types. The system may generate input data to input into the LLM. When generating the input data, the system can identify non-essential conversation data associated with the different communication channel types. Responsive to a prompt, the LLM may disregard the non-essential conversation data and extract a primary contact reason consistently across the different communication channels.
Disclosed herein are system, method, and computer program product aspects for generating a connector using a connectivity language, which would generate a basic connectivity model from an artifact and allow for customization by developers, enabling reusability and support for new artifacts across different platforms. The methods include generating connectors for an external platform using the connectivity language and translating artifacts from the external platform using said connectors. The translated artifacts are used to connect an integration platform with an external platform.
G06F 40/58 - Utilisation de traduction automatisée, p. ex. pour recherches multilingues, pour fournir aux dispositifs clients une traduction effectuée par le serveur ou pour la traduction en temps réel
Embodiments described herein provide methods and systems for generating structured workflows for agentic implementation of user task requests by artificial intelligence (AI) agents. The approach includes transforming historic user-agent conversations into a predefined format using a language model, and generating embedded representations of the transformed conversations. A subset of conversations is selected based on similarity of the embedded representations for a specific user intent, and a workflow description is generated for that intent using another language model. The resulting workflow description is then utilized by the AI agent to perform agentic operations in response to user inputs, enabling more consistent and accurate task execution in a computing environment.
Techniques for transcribing and/or summarizing multimedia collaboration sessions are discussed herein. For example, users can communicate within a teleconferencing meeting associated with a channel. In some examples, a first machine learning model may be configured to receive audio-visual data and user interaction data (e.g., selected emojis, detected gestures, messages or text input by a user, a thread of messages, etc.) and output a teleconferencing meeting summary. In some examples, the teleconferencing meeting summary can include one or more highlights, action items, and/or artificial intelligence (AI) notes generated using a machine learning model(s).
Medium, method and system for ad hoc, ambient, synchronous multimedia collaboration in a group-based communication system. Embodiments of the invention provide a way for users to quickly establish a real-time collaboration among a group of peers without the burden and overhead of a conventional call or video meeting. Users can quickly and easily join and leave these synchronous multimedia collaboration sessions at any time, without disrupting the session for other users. This enables the user to experience the collaboration as a convenient forum rather than a burdensome scheduled event.
A computing device receives a user input directed to a dataset of time series data. The device converts the user input into a set of search terms, and executes a query against a search index for the dataset using the set of search terms to retrieve labeled trend events. Each labeled trend event corresponds to respective portion of a line chart representing the time series data and has a respective chart identifier. The device determines a composite score for each labeled trend event and assigns it to a group. The device ranks the one or more groups and retrieves, from the search index, data for a first subset of line charts having the respective chart identifiers of the ranked one or more groups. The device generates the first subset of line charts and displays one or more line charts of the first subset of line charts.
Disclosed herein are system, method, and computer program product embodiments for managing cloud-based document permissions. In an embodiment, a document generation system may generate and/or store a cloud-based document. The document generation system may also generate links to the cloud-based document. The generated links may be associated with different permissions. For example, a first link may allow a first client device accessing the first link to edit the document while a second link may allow a second client device accessing the second link to only view the document. Other permissions may also include generating distinct graphical user interfaces (GUIs) associated with different links. The GUIs may accept input messages and/or comments in a chat panel. The document generation system may facilitate the delivery of these input messages to other client device with document access.
G06F 21/62 - Protection de l’accès à des données via une plate-forme, p. ex. par clés ou règles de contrôle de l’accès
H04L 9/32 - Dispositions pour les communications secrètes ou protégéesProtocoles réseaux de sécurité comprenant des moyens pour vérifier l'identité ou l'autorisation d'un utilisateur du système
51.
SYSTEMS AND METHODS FOR ARTIFICIAL INTELLIGENCE AGENTS FOR FACTUAL CONSISTENT GENERATION
Embodiments described herein provide a method for enhancing factual consistency of an artificial intelligence (AI) agent. The method includes obtaining a first dataset of documents and corresponding seed summaries; generating, by a first neural network based language model, inconsistent summaries by replacing texts in one or more seed summaries; forming a second dataset of documents and corresponding inconsistent summaries for evaluating a second neural network based language model; and generating, by the second neural network based language model, a detection of a factual inconsistency and/or an explanation of the factual inconsistency. The method further includes generating, by a third neural network based language model, an evaluation score indicating an accuracy level of the detected factual inconsistency based at least in part on the explanation; and building, at a server the AI agent employing the second neural network based language model when the evaluation score is greater than a threshold.
Embodiments described herein provide a training framework with a data pipeline to synthesize datasets of unanswerable requests for any external knowledge database and automatically evaluate RAG systems. The framework assesses not only the ability of RAG systems to correctly respond to answerable requests, but also whether they can appropriately reject unanswerable requests. For example, the AI agent may be trained to reject user queries as unanswerable: Underspecified, False-presuppositions, Nonsensical, Modality-limited, Safety Concerns, and Out-of-Database.
Some implementations of the disclosed systems, apparatus, methods and computer program products are configured automatically generating service plans using large language models (LLMs).
Some implementations of the disclosed systems, apparatus, methods and computer program products are configured automatically generating service plans using large language models (LLMs).
Embodiments described herein construct a multi-agent system integrated with a messaging platform. The system includes a collaboration module that facilitates inter-agent communication, standardized collaboration protocols, and session management for context persistence. Developers can configure AI agents with defined personas, tools, and behaviors using a builder submodule. Agents are connected via message handlers and organized into workflow graphs that represent task execution sequences. A workflow abstraction encapsulates these graphs into scalable processes triggered within messaging channels. Agent tools are defined via functions, Pydantic models, or OpenAPI specifications, enabling seamless interaction with external systems and automation of complex, multi-step tasks.
A system may receive user input including a request for generation of the integration flow. The system may generate a query based on the request, integration flow grounding information, and conversation history. The system may transmit the query to the generative artificial intelligence (AI) model and receive a response including the integration flow. The system may perform a validation process on the integration flow based on one or more integration flow validation rules to generate an error message summary errors categorized into error patterns. The system may transmit, based on the validation process indicating an error, an error correction query to the generative AI model, the error correction query including an indication of the error message summary and error correction grounding information that is associated with the one or more error patterns. The system may receive a corrected response from the generative AI model.
Embodiments described herein provide methods for retrieval augmented generation that leverage synthetic social graphs and persona-based data to improve AI personalization. In one approach, a large language model generates a social graph comprising multiple personas and their relationships, and creates profiles for these personas with randomly assigned characteristics. Synthetic documents, such as conversations between personas, are generated to reflect these characteristics. A training dataset is assembled from these documents, labeled according to the relevant persona attributes. A retrieval model is then trained to identify and retrieve conversations or documents in response to user queries about specific persona characteristics. An AI agent utilizes the trained retrieval model to generate responses to new user queries, drawing on the information retrieved from the synthetic documents to provide personalized answers.
Some implementations of the disclosed systems, apparatus, methods and computer program products are configured automatically generating service plans using large language models (LLMs).
G06Q 10/0637 - Gestion ou analyse stratégiques, p. ex. définition d’un objectif ou d’une cible pour une organisationPlanification des actions en fonction des objectifsAnalyse ou évaluation de l’efficacité des objectifs
Embodiments described herein provide a multi-agent collaborative environment for different types of agent within a multi-agent library. For example, multiple AI agents may be integrated with visual components for a variety of applications. Example AI agents may include a collaborative specialist team for enhancing business workflows and foster decision making from diverse perspectives, a proactive support specialist with in-depth knowledge of specific conversation channels (e.g., a scrum team channel) to assist every team member proactively, and a personal assistant agent that is deeply familiar with individual users, offering tailored support and solving private tasks efficiently.
G06Q 10/0631 - Planification, affectation, distribution ou ordonnancement de ressources d’entreprises ou d’organisations
H04L 51/02 - Messagerie d'utilisateur à utilisateur dans des réseaux à commutation de paquets, transmise selon des protocoles de stockage et de retransmission ou en temps réel, p. ex. courriel en utilisant des réactions automatiques ou la délégation par l’utilisateur, p. ex. des réponses automatiques ou des messages générés par un agent conversationnel
60.
INTEGRATION FLOW GENERATION USING LARGE LANGUAGE MODELS WITH ERROR CORRECTION AND QUALITY ENHANCEMENT
An application server may receive user input comprising a request for generation of an integration flow. The application server may generate a first version of the integration flow with a first LLM. The application server may perform, with a second LLM, a correctness evaluation of the first version of the integration flow, where the correctness evaluation comprises a plurality of evaluations that determine whether the first version of the integration flow complies with a plurality of correctness criteria and identify one or more logical errors included in the integration flow. The application server may perform, with a third LLM and based at least in part on a result of the correctness evaluation, a correctness enhancement procedure to generate a second version of the integration flow, where the correctness enhancement procedure corrects the one or more logical errors based at least in part on the plurality of correctness criteria.
G06F 18/21 - Conception ou mise en place de systèmes ou de techniquesExtraction de caractéristiques dans l'espace des caractéristiquesSéparation aveugle de sources
Techniques for determining which features to utilize in a distributed system with computer services that include a first computer service and a set of dependent computer services. A computer system determines a set of allowed features for a current version of the first computer service, which depends upon the set of dependent computer services. The computer system accesses, from a metadata service, a set of used features indicated as being used by the first computer service. The computer system then creates a maximal feature set for the first computer service from the set of allowed features and the set of used features. After identifying any features in the maximal feature set that are not indicated at the metadata service as having been used by all of the set of dependent computer services, the computer system prevents use of any identified features in the current version of the first computer service.
Embodiments described herein provide systems and methods for document recommendation. A system receives a set of training data including a plurality of documents. The system determines whether the set of training data includes annotated contextual information corresponding to the plurality of documents. The system trains supervised and/or unsupervised models based on the availability of data. The models are used to generate vectors representing the documents. During a live text conversation, text from the conversation may be vectorized using the models and the vectors compared to those representing the documents in order to find the most relevant documents. The system may generate an indication of a recommended document.
Techniques are disclosed for dynamically managing data structures during the execution of program instructions in a database system using Just-In-Time (JIT) compilation. A system receives program instructions that include function calls passing data structures as parameters. The system analyzes the call stack associated with the function calls to determine whether to pass the data structure in a first format (e.g., flattened format) or a second format (e.g., expanded format). Based on the analysis, the system inserts bytecode instructions that convert the data structure between formats, depending on whether operations such as write or read are detected. In some embodiments, the system executes the program instructions, including the inserted bytecode, and dynamically converts the data structure from a contiguous block of memory to a non-contiguous block upon detecting modification operations.
Techniques are disclosed that pertain to a database system having a log owner and log tailers. The log owner maintains a transaction log and the log tailers replay the transaction log. A log tailer may receive a set of requests to perform a database transaction that involves a write operation to write a record and a subsequent read operation to read that record. As a part of performing the transaction, the log tailer may issue a request to the log owner to log the write operation in the transaction log and the log tailer may insert the record into a local memory structure of the log tailer. After receiving a response from the log owner that the write operation has been logged, the log tailer may permit the subsequent read operation to access the record from the local memory structure without requesting the record from the log owner.
A computer system receives a user query for a large language model (LLM). The system encodes the user query into a query vector and queries a semantic cache to determine semantic vectors stored therein that are similar to the query vector. The querying includes determining, for each semantic vector in the semantic cache, a respective semantic similarity score. When the respective semantic similarity score satisfies a threshold score, the computer retrieves a cached response from the semantic cache and returns it as a response to the user query without querying the LLM. When the respective semantic similarity score does not satisfy the threshold score, the system retrieves the cached response, generates a prompt that includes the cached response as one of multiple context examples, inputs the prompt into the LLM, obtains an output from the LLM, and returns the output as the response to the user query.
Embodiments described herein provide systems and methods of subject-driven image generation. In at least one embodiment, a system receives, via a data interface, an image containing a subject, a text description of the subject in the image, and a text prompt relating to a different rendition of the subject. The system encodes, via an image encoder, the image into an image feature vector. The system encodes, via a text encoder, the text description int a text feature vector. The system generates, by a multimodal encoder, a vector representation of the subject based on the image feature vector and the text feature vector. The system generates, by a neural network based image generation model, an output image based on an input combining the text prompt and the vector representation.
G06V 10/74 - Appariement de motifs d’image ou de vidéoMesures de proximité dans les espaces de caractéristiques
G06V 10/82 - Dispositions pour la reconnaissance ou la compréhension d’images ou de vidéos utilisant la reconnaissance de formes ou l’apprentissage automatique utilisant les réseaux neuronaux
Techniques are disclosed relating to determining whether to utilize an alternative query plan for a subsequent execution of a database query. A computer system accesses a trace record detailing an execution of the database query in accordance with a first query plan selected by a query optimizer. The computer system may determine, based on the trace record, whether the execution of the database query satisfies a set of performance criteria. In response to determining that the execution of the database query does not satisfy the set of performance criteria, the computer system may determine whether one or more alternative query plans for the query include a second query plan that is predicted to be more performant than the first query plan. The computer system may provide an instruction to utilize the second query plan in place of the first query plan for the subsequent execution of the database query.
Embodiments described herein provide a training framework with a data pipeline to generate synthesized multi-modal queries with chain-of-thoughts-and-action data for training a multi-modal language model. Specifically, given an image and question-answer (QA) pair as input, a multi-modal language model is prompted to generate a chain-of-thoughts-and-action (CoTA) output corresponding to the answer. The generated CoTA is then used as an input to guide the multi-modal language model to generate an answer to the question, so as to verify whether the CoTA is accurate. Additionally, the CoTA may be parsed into a sequence of actions in order to verify whether the parsed sequence actions leads to the correct answer. After the verification of the CoTA, the CoTA may be added to the image and the QA pair as training data to further train a multi-modal language model to generate an answer according to an input image.
G06V 10/82 - Dispositions pour la reconnaissance ou la compréhension d’images ou de vidéos utilisant la reconnaissance de formes ou l’apprentissage automatique utilisant les réseaux neuronaux
G06V 10/774 - Génération d'ensembles de motifs de formationTraitement des caractéristiques d’images ou de vidéos dans les espaces de caractéristiquesDispositions pour la reconnaissance ou la compréhension d’images ou de vidéos utilisant la reconnaissance de formes ou l’apprentissage automatique utilisant l’intégration et la réduction de données, p. ex. analyse en composantes principales [PCA] ou analyse en composantes indépendantes [ ICA] ou cartes auto-organisatrices [SOM]Séparation aveugle de source méthodes de Bootstrap, p. ex. "bagging” ou “boosting”
G06V 20/70 - Étiquetage du contenu de scène, p. ex. en tirant des représentations syntaxiques ou sémantiques
Techniques for filtering out undesirable generative machine learned model (or LLM) output are discussed herein. A system may receive a subset of an LLM output. That is, the system may stream the LLM output to a user device by receiving one or more tokens from the LLM and outputting such token(s) to a user device. However, prior to outputting the token(s) to the user device, the system may determine whether the token(s) include undesirable content that is to be blocked. The system may use synchronous blocking components (e.g., blocks the undesirable token(s) before such token(s) get output to the user device) and/or asynchronous blocking components (e.g., blocks the undesirable token(s) after the token(s) have been output to the user device) to filter out undesirable content. The synchronous and/or asynchronous blocking components may be designed to block one or more undesirable topics such as hateful speech, profanity, bias, toxicity, factualness, etc.
Techniques for synchronizing data across multiple virtual spaces are described herein. A communication platform may receive a request from a user profile to connect (e.g., link, associate, etc.) a first (already existing) virtual space to a second (to be generated) virtual space. When connecting the first virtual space to the second virtual space, the communication platform may use a previously generated template virtual space to generate the second virtual space. The communication platform may update the content in the first virtual space and the second virtual space when the other virtual space is modified. This continuous updating of the first and second virtual spaces may increase the ability of the communication platform to synchronize the virtual spaces.
Techniques for generating synthesized data within a virtual space are discussed herein. A communication platform may receive a request from a user profile of a communication platform. The request may include one or more instructions for the communication platform to send a survey to one or more user profiles to provide data (e.g., feedback) to the virtual space. Based on sending the request for feedback to the user profiles, the communication platform may receive feedback from the user profiles. Upon receiving such feedback, the communication platform may receive a request from a user profile synthesize the data. Upon receiving the request to synthesize the feedback, the communication platform may input the user feedback into a machine-learning model trained to output synthesized data. The communication platform may cause the synthesized data to be displayed via the virtual space.
G06T 19/00 - Transformation de modèles ou d'images tridimensionnels [3D] pour infographie
H04L 51/02 - Messagerie d'utilisateur à utilisateur dans des réseaux à commutation de paquets, transmise selon des protocoles de stockage et de retransmission ou en temps réel, p. ex. courriel en utilisant des réactions automatiques ou la délégation par l’utilisateur, p. ex. des réponses automatiques ou des messages générés par un agent conversationnel
H04L 51/48 - Adressage des messages, p. ex. format des adresses ou messages anonymes, alias
H04L 51/52 - Messagerie d'utilisateur à utilisateur dans des réseaux à commutation de paquets, transmise selon des protocoles de stockage et de retransmission ou en temps réel, p. ex. courriel pour la prise en charge des services des réseaux sociaux
Processor-implemented methods and systems are disclosed for optimizing pre-existing large language models for model inference. Different types of optimization techniques are provided to an offline optimization program. Within a generic model framework, different combinations of large language model serving configurations are generated. An automated online program optimizes the pre-existing large language models using large language model optimization configurations.
Embodiments described herein provide an AI guardrail system that integrates an agentic compliance evaluation framework with an agentic governance ontology framework to automate AI risk assessment and management. For the compliance evaluation framework, embodiments provide a Digital Risk and Compliance Officer (DRCO) agent built upon an LLM to verify AI compliance in an AI system based on a knowledge base managed by an AI governance ontology. For the governance ontology framework, embodiments provide an AI Governance Ontology (GO) agent built upon two LLMs that generate and update a knowledge base representing concepts, relationships, and metadata related to the governance of AI systems based on AI system generated conversation transcripts.
A computer system receives, via a user interface, a first sketch input corresponding to a first measure data field of a dataset. The computer system converts the first sketch input into a first set of line segments and determines respective values for a first set of parameters corresponding to the first set of line segments. The computer system executes a first query against a database of linearized data using the first set of parameters to identify one or more sets of linearized data. Each set of linearized data corresponds to a respective dimensional dataset for the first measure data field. The computer system retrieves, from the database, one or more first dimensional datasets corresponding to the one or more sets of linearized data, generates one or more first data visualizations from the one or more retrieved first dimensional datasets, and displays the first data visualizations via the user interface.
A system may utilize a large language model (LLM). The system may train and test agents via an LLM by using the LLM to generate a set of scenarios and a set of training data and then training the agents using the training data and executing the agents accordingly. The system may also perform evaluations of data within a data table via LLMs and evaluations of prompt templates using a set of data from the data table. In one example, the system may receive a set of instructions for data evaluation, evaluate the set of data via the LLM, and obtain a set of output data from the LLM. In another example, the system may receive a prompt template for the LLM, execute the prompt template via the LLM, obtain a set of output data from the LLM, and evaluate the performance of the prompt template via the LLM.
Embodiments described herein provide an LLM training framework that utilizes imperfect human labeled training data. In the context of text auto-complete for code, users are provided auto-complete text that they may explicitly accept or reject, or implicitly reject by continuing to type. This results in binary (yes/no) data that is “partially observed” by the fact that implicit rejections may not be true rejections. The probability of a specific output being accepted may be predicted based on the probability of the current model prediction according the model itself, and a “threshold” value generated by a trainable model. For a given predicted auto-complete output, a loss may be computed based on how the predicted probability aligns with the actual user selection (rejection/acceptance). The loss may be modified by a loss shaping function as illustrated here to compensate for the “imperfect” human labeling.
Embodiments described herein provide an AI guardrail system that integrates an agentic compliance evaluation framework with an agentic governance ontology framework to automate AI risk assessment and management. For the compliance evaluation framework, embodiments provide a Digital Risk and Compliance Officer (DRCO) agent built upon an LLM to verify AI compliance in an AI system based on a knowledge base managed by an AI governance ontology. For the governance ontology framework, embodiments provide an AI Governance Ontology (GO) agent built upon two LLMs that generate and update a knowledge base representing concepts, relationships, and metadata related to the governance of AI systems based on AI system generated conversation transcripts.
A training framework includes a transparent and unbiased dataset generation pipeline to generate unbiased multi-modal training data for training a multi-modal LLM. Specifically, one or more images may be annotated using image recognition models, e.g., for object detection, attributes, relations, segmentation, etc. Then, the generated annotations are used to generate scene graph. A scene graph is a data structure that represents objects with attributes in an image as nodes and relationships between objects in an image as edges. Several computer programs are then used to systematically generate question-answer pairs. Because the computer programs include a set of known rules, how the data is generated is known explicitly and thus transparent. The question-answer pairs may be of several types, e.g., a type associated with each of the types of image recognition models. The question-answer pairs may be included in a training dataset and used to train a multimodal LLM.
G06V 10/774 - Génération d'ensembles de motifs de formationTraitement des caractéristiques d’images ou de vidéos dans les espaces de caractéristiquesDispositions pour la reconnaissance ou la compréhension d’images ou de vidéos utilisant la reconnaissance de formes ou l’apprentissage automatique utilisant l’intégration et la réduction de données, p. ex. analyse en composantes principales [PCA] ou analyse en composantes indépendantes [ ICA] ou cartes auto-organisatrices [SOM]Séparation aveugle de source méthodes de Bootstrap, p. ex. "bagging” ou “boosting”
G06V 10/25 - Détermination d’une région d’intérêt [ROI] ou d’un volume d’intérêt [VOI]
G06V 10/26 - Segmentation de formes dans le champ d’imageDécoupage ou fusion d’éléments d’image visant à établir la région de motif, p. ex. techniques de regroupementDétection d’occlusion
G06V 10/82 - Dispositions pour la reconnaissance ou la compréhension d’images ou de vidéos utilisant la reconnaissance de formes ou l’apprentissage automatique utilisant les réseaux neuronaux
G06V 10/94 - Architectures logicielles ou matérielles spécialement adaptées à la compréhension d’images ou de vidéos
82.
Artificial Intelligence Agent Procurement Management in a Database System
Systems disclosed herein include a computing services environment that includes a database system storing a plurality supplier relations management database records representing interactions between client organizations and supplier organizations providing production inputs to client organizations. Systems also include a unified metadata framework including a plurality of autonomous agent definitions referencing a plurality of action definitions defining a plurality of actions capable of being performed within the computing services environment. Systems further include an orchestration layer configured to determine an orchestration plan responsive to a request, the orchestration plan including a subset of the plurality of actions. Systems additionally include an agent service configured to execute a plurality of autonomous agent instances based on a corresponding subset of the autonomous agent definition within a shared virtualized environment, and configured to perform the subset of the plurality of actions via the plurality of autonomous agent instances.
G06Q 10/0631 - Planification, affectation, distribution ou ordonnancement de ressources d’entreprises ou d’organisations
G06F 9/455 - ÉmulationInterprétationSimulation de logiciel, p. ex. virtualisation ou émulation des moteurs d’exécution d’applications ou de systèmes d’exploitation
Techniques are disclosed that pertain to facilitating the execution of machine learning (ML) models. A computer system may implement an ML model layer that permits ML models built using any of a plurality of different ML model frameworks to be submitted without a submitting entity having to define execution logic for a submitted ML model. The computer system may receive, via the ML model layer, configuration metadata for a particular ML model. The computer system may then receive a prediction request from a user to produce a prediction based on the particular ML model. The computer system may produce a prediction based on the particular ML model. As a part of producing that prediction, the computer system may select, in accordance with the received configuration metadata, one of a plurality of types of hardware resources on which to load the particular ML model.
A computing services environment may include application servers providing computing services including access to a database system, a unified metadata framework including autonomous agent definitions referencing action definitions defining a plurality of actions capable of being performed within the computing services environment, an agent service configured to instantiate an autonomous agent instance based on an autonomous agent definition, and an orchestration layer configured to determine an orchestration plan based on novel planning text generated by a generative language model. The computing services environment may be configured to autonomously create, instantiate, and execute an autonomous agent based on natural language configuration input.
G06F 40/284 - Analyse lexicale, p. ex. segmentation en unités ou cooccurrence
H04L 51/02 - Messagerie d'utilisateur à utilisateur dans des réseaux à commutation de paquets, transmise selon des protocoles de stockage et de retransmission ou en temps réel, p. ex. courriel en utilisant des réactions automatiques ou la délégation par l’utilisateur, p. ex. des réponses automatiques ou des messages générés par un agent conversationnel
Techniques for transforming the structural formatting of content are described herein. A communication platform may receive a request from a user profile to transform the structural formatting of content included in an initial virtual space. The communication platform may receive user input data that represents one or more instructions regarding a type of new virtual space to generate. The communication platform may input the instruction(s) and/or content of the initial virtual space into one or more machine learned models that are trained to generate specific types of virtual spaces that include different types of formatting structures. The communication platform may receive the new virtual space from the machine learned model(s) and cause the new virtual space to be displayed via a user interface of a user device.
G06F 40/103 - Mise en forme, c.-à-d. modification de l’apparence des documents
G06F 3/0482 - Interaction avec des listes d’éléments sélectionnables, p. ex. des menus
G06F 3/0484 - Techniques d’interaction fondées sur les interfaces utilisateur graphiques [GUI] pour la commande de fonctions ou d’opérations spécifiques, p. ex. sélection ou transformation d’un objet, d’une image ou d’un élément de texte affiché, détermination d’une valeur de paramètre ou sélection d’une plage de valeurs
86.
CONTEXTUAL RETRIEVAL FOR MULTI-TENANT RETRIEVAL-AUGMENTED GENERATION (RAG) WITH ADAPTIVE LEARNING
Methods, systems, apparatuses, devices, and computer program products are described. A system may support retrieval-augmented generation (RAG) for a large language model (LLM). The system may use adaptive learning to improve the RAG process. For example, the system may implement a context-based embedding function to contextualize the RAG for the specific LLM or a specific tenant or user using the LLM. The context-based embedding function may project document vectors from a generic vector space into a context-based vector space for document retrieval. The system may retrieve a document using the context-based vector space to provide additional contextual information to the LLM to improve the LLM's output. The system may adaptively train the context-based embedding function based on the LLM, user feedback, or both. For example, the system may train the context-based embedding function to improve alignment of document retrieval likelihoods with confidence metrics for the outputs of the LLM.
A system and method for customized scheduling of sources, including breaking down a source of content into at least two categories, including posts and engagements, and gathering content related to a specific source. A scheduler handles scheduling of posts and engagement for a single source and entities that are due to be crawled are sent to a scheduling queue, in which each content type for a source can have its own queue. A process points to the correct scheduler queue in order to request content to be crawled, attaches to the proper queue, processes requests, queries the social network for content, parses the response and sends any new data to be saved to the system.
H04L 51/52 - Messagerie d'utilisateur à utilisateur dans des réseaux à commutation de paquets, transmise selon des protocoles de stockage et de retransmission ou en temps réel, p. ex. courriel pour la prise en charge des services des réseaux sociaux
88.
Mechanisms for testing updates to database query optimizer system
Techniques are disclosed that relate to capturing and replaying database queries to assess the impacts of updates to a database system. A system may receive a plurality of queries from a set of users to execute against a database that stores data. The system identifies one or more of the queries that are deemed relevant to updates being made to the database system. The system executes the received queries and captures query execution information for the one or more identified queries. The system replays, based on the execution information, the one or more queries using the database system with the one or more updates enabled to determine a first performance of the database system. The system may generate a report indicating whether the first performance represents a reduction in performance relative to a second performance of the database system with the one or more updates disabled.
Methods, systems, apparatuses, and computer program products are described. A system may receive, via a cloud-based platform, user input comprising a request for generation of the integration flow. The system may generate a query based on the request and a query template including one or more example integration flows and a request to generate a natural language description of the integration flow. The system may transmit the query to the LLM and may receive, from the LLM, a response including the integration flow and the natural language description. The system may extract the integration flow and the natural language description from the response. The system may perform a validation process on the integration flow based at least in part on one or more integration flow validation rules.
Database systems and methods are provided for dynamic contextualization of an application provided at a client device. In response to a request for a graphical user interface (GUI) display associated with a virtual application, a service identifies a destination context associated with the client application based at least in part on the request, generates a constituent component corresponding to a GUI element associated with the GUI display based on the destination context, wherein the constituent component comprises behavioral code for retrieving data from a respective data source in accordance with the destination context, generates an aggregate component corresponding to the GUI display based on the destination context for incorporating the constituent component within the GUI display based at least in part on a configuration of the GUI display at the database system, and provides the components to the client application for execution responsive to the request.
A system may receive first reference information. The system may generate synthetic conversation records using a first generative artificial intelligence (AI) model. The system may generate response pairs based on the synthetic conversation records, where each response pair includes a respective positive response in accordance with the first reference information and a respective negative response in disaccord with the first reference information. The system may train a first set of generative AI parameters based on the plurality of response pairs and may merge the first set of generative AI parameters with a second set of generative AI parameters to generate a merged set of parameters. The system may receive a query and may provide a response generated by generative AI based on the merged set of parameters.
In an embodiment, a computer system performs a hash join. During a build phase, the computer system constructs a hash table in memory based on rows of a first table. The constructing may result in build batches of rows, including one build batch stored in memory and multiple build batches stored in a storage. The computer system determines whether any of the multiple build batches is skewed according to a data skew condition. In response to determining that there is at least one build batch that is skewed, the computer system loads one or more of the multiple build batches into memory such that there are at least two build batches stored in memory. During a probe phase, the computer system identifies, based on the at least two build batches stored in memory, rows of a second table to join with the rows of the first table.
Application servers may provide computing services to entities. A network ingress may receive application-level request messages and forward some or all of the request messages to an application server. A data aggregator may determine data buckets based on the application-level request messages. A data bucket may include information characterizing one or more features. The information may be determined based on a subset of the application-level request messages received during a respective period of time. A request analyzer may determine one or more of the data buckets and one or more of the features for analyzing an application-level request message and may determine a synthetic indicator for the request based on the one or more data buckets and the one or more features. A web application firewall may block the application-level request message upon determining that the synthetic indicator indicates that the request is illegitimate.
Systems, devices, and techniques are disclosed for style-based dynamic content generation. A seed image, entity design data, entity style data, and text items may be received. Bounding boxes that identify areas of the seed image for the placement of the text items may be generated for the seed image. Variant images may be generated from the seed image, the entity design data, and the entity style data. The variant images may be generated by placing text items in the bounding boxes based on the entity design data and rendering text of the text items using the entity style data.
G06V 10/22 - Prétraitement de l’image par la sélection d’une région spécifique contenant ou référençant une formeLocalisation ou traitement de régions spécifiques visant à guider la détection ou la reconnaissance
G06V 20/62 - Texte, p. ex. plaques d’immatriculation, textes superposés ou légendes des images de télévision
Systems and methods for generating adaptable summaries at varying levels of detail are disclosed. In one aspect, a computer system, in response to receiving a user input specifying a dataset and a chat history of an analytical conversation associated with the dataset, extracts conversational components from the chat history. The computer system generates and displays a user interface and renders the conversational components as interactive affordances in one or more panels of the user interface. The computer system, in response to receiving a user interaction with a first interactive affordance, displays an editing panel in the user interface. The computer system receives a selection of content from the one or more panels and placement of the content in the editing panel. The computer system generates a summary of the analytical conversation according to the selected content and displays the summary in the user interface.
Embodiments described herein provide a code generation framework that explores a code search space of code generation tasks through a tree-based structure. Specifically, the code generation framework comprises a Thinker model, a Solver model, and a Debugger model to implement strategy-planning, solution implementation, and solution improving correspondingly. posing comprehensive roles needed for code generation.
A method for performing a dialog session including receiving a user input from a chatbot application wherein the user input is indicative of a user intent associated with one of a plurality of skills, generating an input vector in response to the user input, generating a list of skills in response to a semantic search comparing the input vector to a plurality of skills vectors where each of the plurality of skills vectors is associated with one of the plurality of skills with the vector to determine a list of skills matching the vector, identifying an intended skill from the list of skills in response to the list of skills and the input vector using a large language model, and performing the intended skill in response to the identification of the intended skill.
A method may include finetuning a generative artificial intelligence (AI) model based on training data including finite state machine (FSM) conversation flows and corresponding intents. The method may include finetuning the generative AI model by tuning a set of base weights with a single set of weights that are based on the FSM conversation flows and the intents. The method may include receiving a query corresponding to an FSM conversation flow. The method may include determining that the first query contains a first intent associated with a first state of the FSM conversation flow. The method may include generating a first response to the first query that corresponds with the first state of the first FSM conversation flow. The method may include communicating commands to a service associated with the first FSM conversation flow, the commands corresponding with actions associated with the first intent and the first state.
Methods, apparatuses, and computer-program products are disclosed. The method may include monitoring user interactions with web services to identify web service elements and application programming interface (API) calls associated with the web service elements. The method may include generating, based on a machine learning analysis of the user interactions, a graph database including mappings between the web service elements and the API calls, where the mappings are organized into user-based clusters. The method may include receiving natural language input that includes instructions for scaling resources and processing the natural language input to identify resource scaling rules from the instructions. The method may include adjusting resource allocation parameters for the web services based on the mappings and the identified one or more resource scaling rules.
Embodiments described herein provide a unified framework for GUI agents to generate automatic task executions on GUIs based on vision observation and action spaces across diverse environments of GUIs. Specifically, vision-based grounding may be generated to improve generalization and reduce inference costs while employing a standardized action space with a plugin system to facilitate consistent learning and interaction across various platforms. In one embodiment, explicit visual planning and reasoning may be integrated into GUI agent, enabling autonomous navigation and interaction within complex digital environments.
