Abstract: Techniques are described herein for a method of machine learning model compression. The method includes receiving a machine learning model comprising a plurality of blocks. The method further includes removing one or more blocks of the plurality of blocks to obtain an intermediate machine learning model comprising a subset of the plurality of blocks. The method further includes adding a block to the intermediate machine learning model to obtain a compressed machine learning model. The block generates an output corresponding to an output of the removed one or more blocks of the plurality of blocks. The method further includes executing the compressed machine learning model on a low resource device.

Abstract: A contextual processing engine architecture. The architecture utilizes data objects retrieved from a database to form a new transactional item data structure as input into a contextual processing engine. The transactional data structure includes a prior context pointer to point to historical context. The historical context can be null for new transactions or one or more basis transaction item data structures for contextual transactions. The processing engine processes the input using process functions lists and aggregates the results for output.

Abstract: Embodiments described herein provide a two-stage model-agnostic approach for generating counterfactual explanation via counterfactual feature selection and counterfactual feature optimization. Given a query instance, counterfactual feature selection picks a subset of feature columns and values that can potentially change the prediction and then counterfactual feature optimization determines the best feature value for the selected feature as a counterfactual example.

Abstract: Embodiments described herein provide a method of sanitizing a user input. A system receives the user input, and may retrieve one or more documents from a database based on the user input. The system then generates, via a first neural network based language model, a sanitized version of the user input in response to a determination to sanitize based on at least one of the user input or the one or more documents. The system then generates, via a second neural network based language model, an output text based on a prompt, the one or more documents, and the sanitized version of the user input.

Abstract: 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.

Abstract: Disclosed are systems, apparatus, methods, and computer readable media for suppressing network feed activities using an information feed in an on-demand database service environment. In one embodiment, a message is received, including data indicative of a user action. An entity associated with the user action is identified, where the entity is a type of record stored in a database. A type of the entity is identified. It is determined whether the entity type is a prohibited entity type. When the entity type is not a prohibited entity type, the message data is saved to one or more tables in the database. The tables are configured to store feed items of an information feed capable of being displayed on a device. When the entity type is a prohibited entity type, the saving of the message data, to the one or more tables in the database configured to store the feed items, is prohibited.

Abstract: 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.

Abstract: An exemplary system having a processor and a memory therein includes means for creating an isolation group, in which creating the isolation groups includes: defining isolation requirements, identifying a group of features utilizing call-out functions, and selecting from among the group of features utilizing call-out functions a group of features having the defined isolation requirements; deploying platform software integrating the isolation requirements, in which the platform software contains instructions to map the isolation requirements to a customer organization; creating the customer organization; creating a unique variant of the customer organization, in which creating the unique variant of the customer organization includes declaratively applying an isolation layer containing isolation requirements on top of a base layer for the customer organization; and deploying the unique variant of the customer organization onto the customer organization's computing infrastructure, in which the unique variant vali

Abstract: Embodiments herein relate to prediction, based on previous usage of a cloud-based computing resource by a user of one or more users of the cloud-based computing resource, future usage of the cloud-based computing resource. Based on the predicted future usage, embodiments relate to identifying that throttling of access to the cloud-based computing resource is to occur, and notifying the user of the throttling. Other embodiments may be described and/or claimed.

Abstract: Systems and techniques are provided that implement application programming interfaces (APIs) that can be executed synchronously or asynchronously depending on the expected response time of the API, the status of the API and/or systems that implement the API, the identity and/or type of user making a request via the API, historical requirements or operation of the API, and/or other factors.

Abstract: A server system is communicatively connected to a plurality of computing devices and one or more databases. The server system receives one or more queries from a computing device. The one or more queries specify a data source. The server system determines a level of security applicable to a user of the computing device. The server system translates the one or more queries into one or more logical queries according to semantics of the data source, and transmits the logical queries to a query pipeline of the server system. The server system executes the one or more queries against a first database of the one or more databases to retrieve query results from the data source. The server system applies the determined level of security to the query results to obtain one or more data sets, and returns the one or more data sets to the computing device.

Abstract: Example embodiments are disclosed where a narrative generation platform is integrated with a conversational interface such as a Chatbot to support interactive and dynamic narrative generation in response to speech inputs from users. Such a system can be further integrated with a visualization platform to allow the speech input through the conversational interface to interactively and dynamically focus and control visualizations and/or narratives that accompany the visualizations.

Abstract: Techniques for caching for an application executable on a mobile device, where the application has a user interface that includes a set of templates specifying data fields that reference record data from a database. The application is executable when the mobile device is online to communicate with an application server to access sufficient record data to fully render templates for those portions of the user interface selected by a user. The set of caching operations can include a record caching operation that caches records according to a default and/or custom set of rules for all versions of the application; a template caching operation that causes selected template data to be cached; and a template-record caching operation that causes a set of record data to be cached such that templates for the cached selected template data will fully render offline for a specified set of records.

Abstract: Disclosed herein are system, method, and computer program product embodiments for implementing variable a declarative authentication engine. The system generates a schema that includes a field and has a format defined by an authentication protocol associated with a service. The system then validates a connection request based on comparing the field of the generated schema to a field of the connection request for the service, wherein the connection request is formatted according to the schema and received from a client device. The system then provides the client device access to the service according to the connection request based on a result of the validating.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for backing up environments. One of the methods includes maintaining, for a cloud computing environment, first data that indicates one or more previously active sandbox environments; determining second data that indicates one or more most recently active sandbox environments; determining, using the second data, a newly added sandbox environment; determining, using a first identifier for the newly added sandbox environment and a second identifier for a prior sandbox environment from the one or more previously active sandbox environments, whether the newly added sandbox environment is likely a refresh of the prior sandbox environment; and performing one or more actions for the newly added sandbox environment using a result of the determination whether the newly added sandbox environment is likely a refresh of the prior sandbox environment.

Abstract: Media, methods, and systems are provided for screen sharing. A user may initiate a video recording. A recording preview interface may be displayed for the user to preview the video recording before beginning the recording. The recording preview interface may comprise a live video preview of the user and of a window or application being shared. When the user selects a window or application for previewing that may cause an infinity mirror effect to occur, a screenshot of the window or application may be generated. The screenshot may be rendered over the live video preview of the window or application to obscure the infinity mirror effect. When the user initiates recording, the screenshot may be removed to expose the live video. The screenshot may be cached and retrieved if the user navigates away and then returns to the preview of the window or application causing the infinity mirror effect.

Abstract: Methods, systems, apparatuses, devices, and computer program products are described. A group-based communication system may use machine learning techniques to classify teams of the system, determine discount messaging for teams of the system, or both. The group-based communication system may receive concurrency data for a team of users and may input the concurrency data (e.g., with one or more other features associated with the team) into a machine learning model to generate a classifier for the team. The classifier may indicate whether the team is a work team, an educational team, or a social team. Based on the classifier for the team, the system may send a message to at least one user of the team (e.g., an administrative user). In some examples, the system may use another machine learning model to generate a discount message for sending to the at least one user based on the team classifier.

Abstract: The system performs optimized upgrades for datacenters configured on a cloud platform. The optimized upgrade process combines upgrades of operating system and services on a node of a cluster of machines of the cloud platform. The system detects a service version change and operating system version change. If the system detects the version change for both a service and the operating system for a node, the system orchestrates the upgrade process for both service and operating system at the same time. The system restarts the node and service once the upgrade completes. This reduces the multiple restarts of the node/services thereby improving the availability of the system for clients.

Abstract: A computing services environment may include a database system storing database records for client organizations accessing computing services including a conversational chat interface, an application server providing access to the conversational chat interface, and an orchestration service configured to execute an orchestration process based on a natural language request message received via the conversational chat interface. An information enrichment and disambiguation process may be executed to determine candidate values corresponding to a text portion of the natural language request message. Novel clarification text requesting clarification of the candidate information may be determined and transmitted via the conversational chat interface. Updated information may be determined based on the candidate information and clarification input received via the conversational chat interface. Novel response text responsive to the natural language request message may be determined based on the updated information.

Abstract: Embodiments described herein provide systems and methods for federated learning. A central system may store a neural network model which has a body of a number of layers, and a classification layer comprising class prototypes which classifies the latent representations output by the body of the model. The central system may initialize the class prototypes so that they are uniformly distributed in the representation space. The model and class prototypes may be broadcast to a number of client systems, which update the body of the model locally while keeping the class prototypes fixed. The clients may return information to the central system including updated local model parameters, and a local representation of the classes based on the latent representation of items in the local training data. Based on the information from the clients, the neural network model may be updated. This process may be repeated iteratively.

Abstract: Methods, systems, apparatuses, devices, and computer program products are described. A group-based communication system may support automatic layout updates for a document space. If a user adds an object into the document space, the system may generate a card representing the object. The card may include a link to a resource external to the document space and a preview image for the object, where the preview image may be rendered based on unfurling the link. The system may update a layout of the document space to include the generated card and may send, for display in a user interface (UI) of a user device, the updated layout. In some examples, a user may select to reposition the card within the UI (e.g., by dragging-and-dropping the card within the document space). The system may further update the layout of the document space based on the repositioning of the card.

Abstract: Embodiments provide a fast multi-agent reinforcement learning (RL) pipeline that runs the full RL workflow end-to-end on a single GPU, using a single store of data for simulation roll-outs, inference, and training. Specifically, simulations and agents in each simulation are run in tandem, taking advantage of the parallel capabilities of the GPU. This way, the costly GPU-CPU communication and copying is significantly reduced, and simulation sampling and learning rates are in turn improved. In this way, a large number of simulations may be concurrently run on the GPU, thus largely improving efficiency of the RL training.

Abstract: The embodiments are directed to training a momentum contrastive autoencoder using a contrastive learning framework. The contrastive learning framework learns a latent space distribution by matching latent representations of the momentum contrastive autoencoder to a pre-specified distribution, such as a distribution over a unit hyper-sphere. Once the latent space distribution is learned, samples for a new data set may be obtained from the latent space distribution. This results in a simple and scalable algorithm that avoids many of the optimization challenges of existing generative models, while retaining the advantage of efficient sampling.

Abstract: 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).

Abstract: 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.

Abstract: 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.

Abstract: Disclosed are methods, apparatus, systems, and computer readable storage media for providing access to a private resource in an enterprise social networking system. One or more servers may receive a request for access to a private resource to be granted to a user from a publisher. The publisher may be configured to publish a message as a feed item to one or more feeds, where the message includes a user identification identifying the user. The user does not have access to the private resource. The feed item may be provided to display in the one or more feeds. Access may be granted to the user via the one or more feeds. In some implementations, access may be granted in response to a user input from the feed item associated with a moderator or owner, the moderator or owner having a privilege to control user access to the private resource.

Abstract: Methods, apparatuses, and computer-program products are disclosed. The method may include storing, on a local machine, a first vector database and a second vector database that include first and second vector embeddings of first and second context information. The method may include receiving a query requesting generation of code. The method may include comparing a third vector embedding that is based on the query with the plurality of first vector embeddings and the plurality of second vector embeddings to identify one or more first vector embeddings and one or more second vector embeddings. The method may include generating a prompt that may include the query, first information and second information corresponding with the one or more first and second vector embeddings. The method may include generating, with a generative artificial intelligence (AI) model, a response to the query based on the query, the first information, and the second information.

Abstract: A computing services environment may provide computing services to a plurality of recipients via the Internet. The computing services environment may include application gateways receiving application-layer request messages from various sources. The computing services environment may also include an orchestration engine determining mitigation policies corresponding with the application gateways based on a classification of a subset of the application-layer request messages as being sent from sources associated with a distributed denial of service attack. The computing services environment may also include application-layer web application firewalls corresponding to the application gateways and transitioning from a deactivated state to an activated state upon receipt of an instruction from the orchestration engine.

Abstract: Methods, systems, apparatuses, devices, and computer program products are described. A data service may receive a first user input indicating a first set of entities for training an artificial intelligence (AI) model for propensity score-prediction. The data service may receive a second user input indicating a set of outcome conditions which define what a user would like to predict about a customer (e.g., propensity to purchase). The data service may generate the AI model accordingly, and based on executing the AI model, generate a set of prediction metrics (propensity scores) for a second set of entities. The data service may store an indication of the AI model for review by a user. When the user approves the AI model and publishes the AI model to the data service, the generated propensity scores may be used to generate a segment of entities of the second set of entities.

Abstract: Embodiments described herein provide systems and methods for providing zero-shot visual question answering. A first image and a first question relating to a visual content of the first image are received. One or more image captions relevant to the first question are determined using a visual-language neural model by determining portions of the first image relevant to the first question. Answer candidates are generated using the one or more image captions, answer candidates. Synthetic question-answer pairs are generated using synthetic questions generated using the answer candidates and the answer candidates. A prompt is generated by concatenating the synthetic question-answer pairs. A first answer to the first question is generated using a language network model using an input of the first question prepended with the prompt.

Abstract: Some implementations of the disclosed systems, apparatus, methods and computer program products are configured automatically generating service plans using large language models (LLMs).

Abstract: Some implementations of the disclosed systems, apparatus, methods and computer program products are configured automatically generating service plans using large language models (LLMs).

Abstract: Some implementations of the disclosed systems, apparatus, methods and computer program products are configured automatically generating service plans using large language models (LLMs).

Abstract: Methods, apparatuses, and computer-program products are disclosed. The method may include deriving a first intermediate LLM based on a broad spectrum LLM. The first intermediate LLM may be associated with a first level of a hierarchy and deriving the first intermediate LLM may include training the first intermediate LLM on first training data associated with the first level of the hierarchy. The method may include deriving a user-level LLM based on the first intermediate LLM, where the user-level LLM is associated with a user associated with a second level of the hierarchy included in the first level of the hierarchy. Deriving the user-level LLM may include training the user-level LLM on second training data associated with the user, where deriving the user-level LLM may further include inheriting one or more first characteristics from the first intermediate LLM.

Abstract: Embodiments described herein provide a system and method for unsupervised anomaly detection. The system receives, via a communication interface, a dataset of instances that include anomalies. The system determines, via an inlier model, a set of noisy labels. The system trains a causality-based label-noise model based at least in part on the set of noisy labels and the set of high-confidence instances. The system determines an estimated proportion of anomalies in the dataset of instances. The system retrains the inlier model based on the estimated inlier samples. The system iteratively retrains the inlier model and the trained causality-based label-noise model based on the output from the corresponding retrained models not converging within the convergence threshold. The system extracts the anomaly detection model from the iteratively trained causality-based label-noise model.

Abstract: Systems and techniques are described for a unified data pipeline interface. The pipeline interface provides an interface illustrating key metrics, activities, insight, and/or other categories. The pipeline interface provides such data as to a specific opportunity. The systems and techniques described herein allow for the determination, presentation, and standardization of such data through a specifically configured graphical user interface.

Abstract: Techniques for streaming generative machine learned model (or LLM) output to a virtual space are described herein. A system may receive a request to perform an action. The system may leverage LLMs to assist in performing aspects of the requested action. The system can generate input data to input into the LLM. When generating the input data, the system can identify sensitive data associated with the request. The system can modify the input data to mask and/or anonymize the sensitive data. The system can input the input data into an LLM trained to output a subset (e.g., less than all) of the response at a time. The system can add the output subset to a buffer and upon identifying the masked data in the buffer, the system can demask the sensitive data and output the sensitive data to the user profile in a streaming manner.

Abstract: Disclosed are some implementations of systems, apparatus, methods and computer program products for implementing a database backup system.

Abstract: Embodiments described herein provide a unified debugging framework that adapts a pretrained programming language model for line-level debugging and repair. Specifically, the debugging framework follow the logic of programmers on how to debug their code. For example, the debugging framework first determines whether or not a function is buggy. If it is buggy, the debugging framework localizes the problematic line and provide a patch (repair).

Abstract: Techniques for modifying the display of task data are discussed herein. A user profile may request to modify a user interface object from a compressed state to an expanded state. Based on the request, the communication platform may convert the user interface object into an expanded state such that one or more columns of the task are presented in a single user interface. The communication platform may determine an updated organization of the columns and use such data to determine a modified size of the user interface object. The communication platform can generate a modified user interface object in an expanded state that includes an increased number of the columns. The communication platform may cause the modified user interface object to be displayed via a user interface of a mobile device.

Abstract: Embodiments described herein provide a universal item learning framework that generates universal item embeddings for zero-shot items. Specifically, the universal item learning framework performs generic features extraction of items and product knowledge characterization based on a product knowledge graph (PKG) to generate embeddings of input items. A pretrained language model (PLM) may be adopted to extract features from generic item side information, such as titles, descriptions, etc., of an item. A PKG may be constructed to represent recommendation-oriented knowledge, which comprise a plurality of nodes representing items and a plurality of edges connecting nodes represent different relations between items. As those relations in PKG are usually retrieved from user-item interactions, the PKG adapts the universal representation for recommendation with knowledge of user-item interactions.

Abstract: An automated data extraction pipeline for large language model (LLM) training may include extracting a set of code segments from a set of natural language question-answer (Q&A) combinations that each include a provided input, a provided output, and a provided code segment formatted to transform the provided input into the provided output. The data extraction pipeline may then generate a predicted output from a question portion of a first natural language Q&A combination using a first LLM. A first extracted code segment from the extracted set of code segments may then be executed to generate a first actual output of the first extracted code segment. One or more data samples may then be generated for training a second LLM based on a comparison of the first actual output to the predicted output. The second LLM may then be trained using the one or more data samples.