Abstract: A text interaction record including interaction text from one or more messages between a client machine and a service provider is received at a database system. A search vector including a text embedding representing the interaction text in a multi-dimensional vector space may be determined based on the interaction text via a processor at the database system. A reference interaction record including reference interaction text may be retrieved from the database system based on the search vector. The reference interaction record may include a reference vector representing the reference interaction text in the multi-dimensional vector space. Recommended reply text is determined based on the interaction text and the reference interaction text by a large language model configured to generate the recommended reply text in response to a prompt that includes one or more natural language instructions.

Abstract: Embodiments described herein a CROss-Modal Distribution Alignment (CROMDA) model for vision-language pretraining, which can be used for retrieval downstream tasks. In the CROMDA mode, global cross-modal representations are aligned on each unimodality. Specifically, a uni-modal global similarity between an image/text and the image/text feature queue are computed. A softmax-normalized distribution is then generated based on the computed similarity. The distribution thus takes advantage of property of the global structure of the queue. CROMDA then aligns the two distributions and learns a modal invariant global representation. In this way, CROMDA is able to obtain invariant property in each modality, where images with similar text representations should be similar and vice versa.

Abstract: Embodiments provide a software framework for evaluating and troubleshooting real-world task-oriented bot systems. Specifically, the evaluation framework includes a generator that infers dialog acts and entities from bot definitions and generates test cases for the system via model-based paraphrasing. The framework may also include a simulator for task-oriented dialog user simulation that supports both regression testing and end-to-end evaluation. The framework may also include a remediator to analyze and visualize the simulation results, remedy some of the identified issues, and provide actionable suggestions for improving the task-oriented dialog system.

Abstract: Systems, methods, and computer-readable media are provided for reducing a number of potential code paths such that it is feasible to examine all possible code paths within source code. Source code may be received. The source code may be traversed such that the path is recorded. Predicates may cause the path to split such that both paths can be traversed with the result of the predicate stored such that the path does not need to split again when encountering a new predicate for which the stored predicate is determinative. The determined paths can then be used by applications.

Abstract: Techniques are disclosed relating to monitoring behavior of a computing system shared by multiple tenants. In some embodiments, a computer cluster is maintained that hosts containers accessible to a plurality of tenants of the computer cluster. First telemetry data collected about a particular one of the plurality of tenants is received from a container hosted at a first of a plurality of servers of the computer cluster. The first telemetry data identifies the particular tenant's consumption of a resource provided by the container. In response to the computer cluster migrating the container from the first server to a second of the plurality of servers, second telemetry data collected about the particular tenant's consumption of the resource is received from the migrated container hosted at the second server. An analysis is performed of the first and second telemetry data to identify whether the particular tenant's consumption of the resource has changed.

Abstract: An online system performs predictions for real-time tasks and near real-time tasks based on available network bandwidth. A client device receives a regression based machine learning model. Responsive to receiving a task, the client device determines an available network bandwidth for the client device. If the available network bandwidth is below a threshold, the client device uses the regression based machine learning model to perform the task. If the client device determines that the network bandwidth is above the threshold, the client device extracts features of the task, serializes the extracted features, and transmits the serialized features to an online system, causing the online system to use a different machine learning model to perform the task based on the serialized features.

Abstract: Embodiments described herein provide a semantic parsing framework which may be referred to as Uni-Parser. The Uni-Parser framework may be applied to question answering on both knowledge bases and databases. The three main stages of the Uni-Parser framework are enumeration, ranking, and generation. At the enumeration stage, primitives are enumerated based on matching the question to the data structure. After enumerating primitives, the Uni-Parser framework may rank the primitives used a trained ranker model. The top ranked primitives may then be used as inputs to a generator which is a learned sequence to sequence model which produces a logical form.

Abstract: Methods, computer readable media, and devices for efficient application programming interface (API) processing with privacy protection include one method of receiving a user request for content from a client, parsing the user request for content to identify one or more request portions having a type of public information, customized information, and personal information, transmitting the one or more request portions to a corresponding microservice based on the type of the portion of content being requested, receiving one or more response portions, determining a type of the one or more portions of content, and in response to determining the type of a portion of content is not personal information, caching the portion of content based on the type of the portion of content, combining the one or more response portions into a user response, and sending the user response to the client.

Abstract: Methods, systems, apparatuses, devices, and computer program products are described. A system may identify, from an event log including log entries for a tenant of a multi-tenant database system, a pattern of log entries corresponding to main actions and satisfying a frequency threshold. The system may identify log entries associated with the pattern and corresponding to the main actions, detailed actions, or both. The system may retrieve data corresponding to a history field of a data object associated with the pattern and may determine at least a portion of a process flow for the data object according to the pattern and based on the log entries and the historical data. The process flow may include operations to perform using the data object. In some cases, the system may transmit, to a user device, an indication of the portion of the process flow for user review and implementation.

Abstract: Embodiments described herein provide an intelligent method to select instances, by utilizing unsupervised tracking for videos. Using this freely available form of supervision, a temporal constraint is adopted for selecting instances that ensures that different instances contain the same object while sampling the temporal augmentation from the video. In addition, using the information on the spatial extent of the tracked object, spatial constraints are applied to ensure that sampled instances overlap meaningfully with the tracked object. Taken together, these spatiotemporal constraints result in better supervisory signal for contrastive learning from videos.

Abstract: Techniques for generating structured data containers via a communication platform and messaging threads associated with the structured data containers are discussed herein. For example, the communication platform may, in response to receiving a request from a user of the communication platform, generate a structured data container representing a system of record associated with the communication platform, the structured data container containing a plurality of fields and information associated with an interaction via the communication platform. The communication platform may further generate and present, via a messaging thread, the structured data container including the plurality of fields.

Abstract: Techniques are disclosed in which a server computer system manages a database cache for multiple different tenants. The system may compare a key having an unidentified tenant with key segments having a number of occurrences within a database cache satisfying a threshold count, where the key is included in a key-value entry of the cache storing data for multiple tenants. Key segments for keys of the database cache and corresponding occurrence counts are stored in a central database. Based on the comparing, the system determines whether the unidentified key matches one of the key segments satisfying the threshold count. In response to the system determining a match, a tenant corresponding to the unidentified key is identified based on the matching key segment. The disclosed techniques may advantageously allow for analysis of cache metrics for tenants and more efficient use of the cache (e.g., by altering cache parameters for individual tenants).

Abstract: Implementation of a guided rules editor comprises displaying a view of a user interface (UI) of the guided rules editor to a user on the client device. Metadata of an organization to which the user belongs is fetched and used to identify which data of the organization is eligible for rule creation and attributes of the metadata are used to determine a layout of the one or more UI selection elements. One or more UI selection elements are rendered in one or more positions in the view on a display screen according to the layout and at least a portion of the one or more UI selection elements are populated with the data of the organization. Responsive to receiving user input of the one or more UI selection elements, an organization-specific rule is generated and saved to a data repository for execution by a rules engine after rule compilation.

Abstract: A method and system for defining feature variable configurations in a system defined by a codebase has been developed. The codebase is executed over a set of test cases to test the software feature behaviors of a software feature. Feature variable configurations including feature variables generated by the test cases are logged. Associations between the feature variables and individual software feature behaviors are identified based on the logged feature variable configurations. Impact scores for the feature variables associated with individual software feature behaviors are calculated based on an occurrence of the associations between the feature variables and individual software feature behaviors. Deterministic feature variables are identified for individual software feature behaviors based on the calculated impact scores.

Abstract: A method by one or more electronic devices to detect flapping in a computer system. The method includes obtaining a plurality of metric values indicating a state of the computer system at various times within a time window, determining a combined length of lines connecting time-wise consecutive points representing the plurality of metric values in a cartesian coordinate system, wherein a first axis of the cartesian coordinate system represents time and a second axis of the cartesian coordinate system represents a metric associated with the plurality of metric values, determining a baseline length based on a maximum metric value within the time window, a minimum metric value within the time window, and a length of the time window, and determining whether the computer system flapped during the time window based on a result of comparing the combined length to the baseline length.

Abstract: Techniques for displaying tasks and/or virtual spaces to a user within a group-based communication platform are discussed herein. In some examples, a user may use a first device to start a first session on a communication platform. As the user may transition from the first user device to a second device, the communication platform may determine that the user has changed devices based on determining that the first session has ended and/or the user having requested to start a second session using a second device. Upon receiving the request and/or based on the first session ending, the communication platform may identify or otherwise predict a task and/or virtual space that the user may want to access upon starting the second session on the second device. The communication platform may cause the task and/or virtual space to be displayed via a user interface of the second device.

Abstract: A request may be received to access a digital asset uniquely identified by an identifier within an on-demand database system configured to provide computing services to a plurality of entities via the internet. Ownership of the digital asset may be determined based on ownership of a designated smart contract recorded in a public trust ledger. A plurality of related smart contracts recorded in the public trust ledger may be identified. Each of the related smart contracts may be linked with the designated smart contract. Two or more of the smart contracts may each include a respective one or more non-fungible tokens related to the digital asset. A respective state for two or more of a plurality of modifiable values may be determined. Each of the modifiable values may be being stored within a respective one of the plurality of related smart contracts within the public trust ledger.

Abstract: In some embodiments, a method determines a first functional domain that includes a group of security policies that have been copied from a second functional domain. Network flow data is queried to determine network traffic that is associated with a security policy in the group of security policies in the first functional domain. The method analyzes utilization of the security policy based on the network traffic. Based on the analyzing, a recommendation is generated to change the security policy in the first functional domain.

Abstract: A method in a computing environment for application start up includes: generating a directed acyclic graph (DAG) for tasks that are performed during application start-up; performing topological sorting of the tasks based on the DAG to define an order for the tasks; allocating a tenant-specific priority queue; loading the tasks in the tenant-specific priority queue in the order determined by the topological sorting; allocating a plurality of computing threads to executing the tasks; executing the tasks using the plurality of computing threads that were allocated, wherein a plurality of the tasks are executed in parallel, and the tasks are assigned to the computing threads in an order based on the order in which the task are loaded in the tenant-specific priority queue; preparing a start page for use in accessing the application; and causing the start page to be displayed on a user access device for accessing the application.

Abstract: Apparatus and method for identifying code path contributors to resource utilization spikes. For example, in some implementations, stack trace data and profiling data are sampled from a plurality of code paths of an application to generate a plurality of code path samples, where the sampling is performed across a series of temporal windows. Processor utilization data is also sampled across a corresponding series of temporal windows. Correlations are detected between the code path samples and occurrences of processor utilization spikes to generate correlational scores for the plurality of code paths. A ranking of code paths is then generated in accordance with the correlational scores in which code paths relatively higher in the ranking are more likely to contribute to the occurrences of the processor utilization spikes than code paths relatively lower in the ranking.