Abstract: Techniques are disclosed for evaluating and improving color consistency across images generated by a text-to-image model in response to a common prompt. A plurality of images are generated using a fixed configuration, and color distribution features are extracted from each image. Similarity measures are computed between images, and an alignment score is calculated to reflect consistency in color composition across the set. If the alignment score falls below a threshold, one or more components of the configuration—such as the prompt, preprocessing logic, model selection, or post-processing—may be iteratively adjusted to improve consistency. In some implementations, spatial color consistency is also assessed using region-based analysis. Once an optimized configuration is identified, it can be used in a subsequent inference phase to generate final outputs with improved visual coherence. The approach is reference-free and applicable even when no explicit color specification or reference image is provided.

Abstract: Systems and methods include determination, for each of a plurality of different types of requests, of a sequence in which services are executed in response thereto, generation of a directed graph based on the sequences s, where each vertex of the directed graph represents a service, generation of a flow network graph by splitting each vertex of the directed graph into two vertices with a directed edge between, and associated with a capacity of the service represented by the vertex, determination of a maximum flow through the service system based on the flow network graph, determination of a residual capacity of each service based on the maximum flow and its capacity of each service, determination of services associated with a zero residual capacity, and increasing of computing resources available to the determined services.

Abstract: Techniques for implementing a differentially private variational autoencoder for data obfuscation are disclosed. In some embodiments, a computer system performs operations comprising: encoding input data into a latent space representation of the input data, the encoding of the input data comprising: inferring latent space parameters of a latent space distribution based on the input data, the latent space parameters comprising a mean and a standard deviation, the inferring of the latent space parameters comprising bounding the mean within a finite space and using a global value for the standard deviation, the global value being independent of the input data; and sampling data from the latent space distribution; and decoding the sampled data of the latent space representation into output data.

Abstract: A system and a process for provisioning a job through a trained machine-learning dynamic provisioning agent is provided herein. An input vector representing the job having one or more job components may be received. One or more additional data vectors representing additional job data may be obtained. For the one or more job components respectively, one or more action values corresponding to one or more provisioning options may be calculated based on the one or more additional data vectors. For the one or more job components respectively, one or more provisioning options for the respective one or more job components may be selected based on the corresponding one or more action values. The one or more selected provisioning options corresponding to the respective one or more job components may be aggregated. The aggregated selected provisioning options may be provided.

Abstract: Edge devices can provide large amounts of data, only some of which may be relevant to a downstream system. In addition, edge device data may need to be processed, including being enriched with additional data, before being used by the downstream system. Disclosed techniques provide an edge agent that receives edge device data, including sensor data. The edge agent performs processing on the edge device data, which can include operations such as filtering and aggregation, as well as enriching the edge device data with values for attributes in addition to attributes and associated values that are present in edge device data sent from the edge device to the edge agent. The use of the edge agent and its associated actions can reduce a data volume sent to the downstream system, as well as making the edge device data more immediately useable by the downstream system.

Abstract: According to some embodiments, systems and methods are provided including generating a first visualization based on a model; receiving a note on a first visualization data element; storing the received note with a path; generating a second visualization based on the model, the second visualization including a second visualization data element corresponding to the first visualization data element; generating, via execution of a member list instruction, a member list of note candidate members; determining a volume of note candidate members in the member list; selecting a path component based on the determined volume; executing the selected path component and outputting the path for each note candidate member in the generated member list; and rendering the note on the second visualization in a case the path for a first note candidate member in the generated member list matches the path stored with the received note. Numerous other aspects are provided.

Abstract: Systems and processes for providing a service to check for duplicates in data records and perform merges of duplicate data records in a database are provided. In a method for performing duplicate checks and merges of data, user input may be received corresponding to a new data record. The user input may be converted into one or more vectors using word embeddings, and the vectors may be compared to stored vectors for existing data records in a database using one or more functions configured to indicate a likelihood of a match between the newly-entered data and one or more of the existing data records. Data for the new data record may be merged with an identified duplicate data record using a language model.

Abstract: Arrangements for pre-deployment of configuration settings using shared container technology are provided. A customized table may be generated with the tenant identifier field being removed. A view of the database table which defines joins from the customized table including customer key ranges and a centralized table including vendor key ranges may be generated. Instead-of triggers used to modify the view upon an insert, upsert, or update operation on the view may be defined. An instead-of trigger may be executed in response to a modification action to the view. An instead-of trigger may determine whether the modification action applies to a customer key range or a vendor key range. The modification action may be redirected to the customized table in response to a determination that the modification action applies to the customer key range, or blocked in response to a determination that the modification action applies to the vendor key range.

Abstract: A metaverse environment reader performs semantic segmentation and object detection steps to identify a plurality of objects in a metaverse scene. Next, the reader determines an order of importance of the plurality of objects in the scene based at least on a location and a size of each object. Then, the reader sorts the plurality of objects of the scene based on the determined order of importance. Next, the reader indexes the objects based on the segmenting of the scene and based on the determined order of importance. Then, the reader creates a description of the scene based on the indexing, where the description is an audio, haptic, or braille representation of the scene. Next, the reader generates and conveys one or more electrical signals which include an encoding of the description of the scene to a user device to be presented on a user interface.

Abstract: Methods, systems, and computer-readable storage media for receiving, by a management tool of a software vendor, a payload including system data that is representative of a software system experiencing an issue, generating, by the management tool, a service using the system data provided in the payload, executing the service to trigger an automated procedure that collects one or more log files from one or more source systems, uploads the one or more log files to a repository, and provides a uniform resource locator (URL) that provides a link to the one or more log files within the repository, and transmitting a response to a ticketing system, the response including the URL.

Abstract: Disclosed herein are a system, method, and computer program product embodiments for enabling document collection creation in accordance with a semi-structured data type, retrieving and filtering data from both relational databases and a document collection, and determining a data type in which the data is retrieved. For example, a statement configured to generate a collection of semi-structured documents in a document store based on a schema is processed. The statement specifies a semi-structured data type in which a plurality of entities from the collection are to be returned from the document store. A determination is made that an entity of the plurality of entities is defined by the schema as being a particular data type different from the semi-structured data type. A query for the entity is provided to the document store. The entity is received, based on the query, in accordance with the particular data type.

Abstract: Embodiments of the present disclosure include techniques for moving data. In one embodiment, computer systems import and/or export data. The computer systems may be application servers coupled to a content management system accessible over a plurality of tenant frontends. The data movement may use a plurality of jobs executing in parallel to move packages of content between the systems. In one embodiment, before a job is executed, a check is made of other jobs to ensure any overlap does not cause data conflicts.

Abstract: A in-memory certificate conversion system configured to determine that a configuration file contains a JKS certificate that is not compatible with a live data connector (LDC) and automatically convert the JKS certificate to a Public-Key Cryptography Standards (PKCS12) format certificate via a Java Virtual Machine of the LDC by generating a PKCS12 container containing the public key, private key, and the chain of trusted authority certificates for the JKS certificate and generating a PKCS12 format certificate by exporting the PKCS12 container as a bytes array. The in-memory conversion system provides, via the LDC, the PKCS12 format certificate to a web server for SSL configuration, where the web server is configured to establish secure communications using the PKCS12 format certificate in real time from a request to access a service via the web server.

Abstract: A transaction processing protocol for serverless database management systems can use a transaction scheduler to guarantee consistent serializable execution though analysis of the access pattern of transaction types and appropriate ordering of the transaction's events at runtime. A transaction topology is determined for each type of transaction and these are combined and used to generate a serialization graph. Cycles in the serialization graph are identified and breaking transaction types which may break the cycles are determined. When transaction requests are received, a breaking type of transaction is scheduled as a last transaction in the current epoch and later transactions not having the breaking transaction type are scheduled to execute in the next epoch.

Abstract: Embodiments facilitate deployment of customized code at a local site, for reference by a service that is being called by a remote system. At a design time, a visual code editor (e.g., Blockly) is utilized to create and store customized code at the local site. During a subsequent runtime, in response to dispatched service call initiated by the remote system, the customized code is retrieved and executed at the local site. By maintaining the customized code locally, embodiments confer security and avoid congestion associated with having the customized code stored remotely (with the remote system). This selective dispatch of a service call for handling by the local customized code, can be implemented based upon an extension scheme.

Abstract: An in-memory computing system for conducting on-line transaction processing and on-line analytical processing includes system tables in main memory to store runtime information. A statistics service can access the runtime information using script procedures stored in the main memory to collect monitoring data, generate historical data, and other system performance metrics while maintaining the runtime data and generated data in the main memory.

Abstract: In an example embodiment, a confidence score is computed for a predicted label (from a first model) for information extracted from a document. The confidence score is computed using a machine learned model different than the first model which is based on a Sliding-Window method. The Sliding-Window method may be based on convolutional neural networks classification, using sliding windows. It receives as input (1) the string of extracted information from an independent previous information extracted step (the “input text”), (2) the string's predicted class label, (3) the string's coordinate location in the document, and (4) the text of the document (for additional context information). The Sliding-Window method's task is to predict the confidence score to determine the correctness of the predicted label for the information.

Abstract: Some embodiments are directed to scheduling batch jobs for execution on a computer system. Multiple candidate job schedules may be generated that satisfy the defined limits. A job schedule quality may be predicted for the candidate job schedules using a predictive model previously trained on job schedule performance data. The candidate job schedule with suitable job schedule quality is used for autonomously scheduling the batch job execution.

Abstract: Systems and methods include reception receive of an instruction from an operator to check compatibility of an application executing on an application server with a machine learning scenario, the machine learning scenario comprising a machine learning model and a training pipeline, determination of a first plurality of object instances associated with the machine learning scenario, each of the first plurality of object instances comprising logic executable by a first object executor executing in the application server to perform a prerequisite check associated with the machine learning scenario and the application, instruction of the first object executor to execute the logic of the first plurality of object instances, and reception of first results corresponding to the executed logic from the first object executor.

Abstract: Techniques and solutions are provided for providing improved replay of a database workload. In particular, the present disclosure provides techniques for generating dependency graphs that have a reduced number of redundant edges. In particular, the dependency graphs can be generated using a technique that removes subpaths of a larger path of requests for database operations that access a common database object. Another technique removes subpaths of a larger path the incudes additional requests in multiple sessions, even if some requests access different database objects. Particular implementations process a set of requests to generate state information, where the state information can be used in generating a dependency graph that excludes one more types of redundant edges. Additional efficiency is provided in that graphs can be generated using a single scan of a given request.