Abstract: A confidence estimation tool uses a calibrated confidence mapping model to estimate confidence for a model-generated candidate root cause. The tool uses a generative artificial intelligence (“AI”) model to determine, based on a description of a current event, a candidate root cause of the current event. The tool determines a description-based confidence score using the description of the current event and descriptions of a set of relevant historical events in a target domain. The tool also determines a cause-based confidence score using the candidate root cause of the current event and root causes of the set of relevant historical events. Finally, the tool determines a final confidence score using the description-based and cause-based confidence scores. Even if the generative AI model is configured for general-domain applications, by referencing relevant historical events, the tool can accurately estimate confidence for a model-generated candidate root cause within the target domain.

Abstract: A technique uses a local language model, implemented by a local computing device, to customize original information based on a profile of a local entity (e.g., a user). A profile-generating system produces the profile by first using plural machine-trained local encoders to convert a set of content items to instances of local encoded information. The profile-generating system then uses a global encoder to convert the plural instances of local encoded information into profile information that expresses the profile. The technique passes a combination of the original information and the profile information to the local language model, optionally with level information that specifies an extent of customization to be applied to the original information. In some implementations, the original information originates from another language model that is larger than the local language model.

Abstract: Approaches to robust encoding and decoding of escape-coded pixels in a palette mode are described. For example, sample values of escape-coded pixels in palette mode are encoded/decoded using a binarization process that depends on a constant value of quantization parameter (“QP”) for the sample values. Or, as another example, sample values of escape-coded pixels in palette mode are encoded/decoded using a binarization process that depends on sample depth for the sample values. Or, as still another example, sample values of escape-coded pixels in palette mode are encoded/decoding using a binarization process that depends on some other fixed rule. In example implementations, these approaches avoid dependencies on unit-level QP values when parsing the sample values of escape-coded pixels, which can make encoding/decoding more robust to data loss.

Abstract: A camera system includes a camera, a privacy indicator configured to indicate an activation status of the camera, and an optical camera blocker disposed between the camera and an external environment. The optical camera blocker is configured to, when activated, obscure images captured by the camera. A privacy control circuit is configured to compare a state of a first circuit trace indicative of an intended operational status of the privacy indicator with a state of a second circuit trace indicative of an actual operational status of the privacy indicator, and based at least in part on detecting a mismatch between the state of the first circuit trace and the state of the second circuit trace, activate the optical camera blocker.

Abstract: The disclosed concepts relate to pretraining of machine learning models. One example method involves performing separate optimization of a first machine learning model and a second machine learning model. The first machine learning model can be optimized based at least on first predictions and the second machine learning model can be optimized based at least on second predictions. The first predictions can represent predictions of masked values in first sequences of values values, and the second predictions can represent whether or not the first values were replaced with different values predicted by the first machine learning model.

Abstract: Methods, systems, and computer programs are presented for labeling datasets. An example method can include generating rules for labeling data records within a first dataset. The rules can indicate an extent to which a data record matches query criteria. The method can further include generating an aggregated label for the corresponding data record based on the rules and training a machine learning model using the first dataset and the aggregated label. The method can include receiving an indication of user engagement and combining the indication of user engagement with the aggregated label to generate a score.

Abstract: Embodiments extract a key phrase from an online profile of a user of a professional social network (PSN) and identify a set of skills associated with the key phrase. A set of skill embeddings corresponding to the identified set of skills may be retrieved from a store of skill embeddings. The skill embeddings may be pre-created by a large language model (LLM). The retrieved skill embeddings may be aggregated to create a skill-centric digital representation of the user. Based on the skill-centric digital representation of the user, a subset of digital documents may be identified to present to the user via the PSN.

Abstract: Systems and methods for generating new content from a machine-learning model. A content generator extracts string content from existing slides of a slide presentation document and generates a text query using the existing slide content as context. The query is directed to a large language model. Additionally, prompt input from a user is received and combined with the context in the query. A response from the large language model is parsed and text output is separated into prospective slides to add to the slide presentation document. Upon selection of one or more prospective slides, the prospective slides are generated and added as new slides to the slide presentation document.

Abstract: Timestamps are used to prevent data tampering. A trusted data generator generates a concatenated data object including a data object and metadata, which includes a location indication, a timestamp, and, depending on the type of encryption, an integrity value. The trusted timestamp generator encrypts the concatenated data object, stores the encrypted concatenated data object (e.g., in untrusted storage) in accordance with a location indicated by the location indication, and protects the timestamp in trusted storage (e.g., by protecting at least the root timestamp). A trusted data validator extracts the data object and metadata from the decrypted concatenated data object and validates the data object by comparing the storage location to the extracted location indication, the protected timestamp to the extracted timestamp, and a calculated integrity value to the extracted integrity value.

Abstract: A cache service provides applications in a containerized, multi-tenant cloud-computing system low-latency access to secrets. The cache service may operate as a cluster-level service or a sidecar service. The cache service may store copies of secrets (which are located in one or more absolute stores) in a cache storage. The cache service and the cache storage may be closer to the applications than the one or more absolute stores are to the applications. The cache service may aggregate secrets associated with multiple entities in a single cache storage. The cache service may support isolation between secrets such that secrets of a first entity are isolated from secrets of a second entity. The cache service may enforce granulated access controls such that it can apply different access controls to secrets of a first entity than to secrets of a second entity.

Abstract: A computing system retrieves a value of a device identifier of itself and generates a device claim asserting the value of the device identifier. The device claim is then associated with an identifier of a user of the computing system. The computing system then generates and attach proof code to the device claim to turn the device claim into a verifiable device credential (VDC). The proof code proves that the VDC is issued by the user of the computing system. The VDC is later presented to a relying entity as part of an identity protection system to further protect the user's identity.

Abstract: Innovations in intra block copy (“BC”) prediction mode facilitate intra BC prediction that is more effective in terms of rate-distortion performance and/or computational efficiency of encoding and decoding. For example, some of the innovations provide ways to select block vector (“BV”) values more effectively. Other innovations provide ways to encode/decode BV values more efficiently. Still other innovations address how to perform in-loop deblock filtering when a block has intra BC prediction mode, or address how to perform intra BC prediction when constrained intra prediction is enabled.

Abstract: A method and system for providing a command-based communication are disclosed, comprising identifying a target user from a user database based on parsed entries subsequent to an action command in a compose field of a group thread, receiving a user selection of one of two operation modes in which the system operates, including a fork mode and a notification mode, and responsive to the user selection selecting the fork mode, creating and transmitting a fork communication thread between the user and a second user, the fork communication thread diverging from the group thread.

Abstract: Recommendations of access control assignments are disclosed. Paired activity data is determined based on historical activity. The paired activity data including task and resource pairs that indicate a task performed by an individual identity, and a resource on which the task was performed. Over-privileging metrics are determined for candidate access control assignments based on the tasks and resource scopes granted by the candidate access control assignments and the paired activity data. Recommended access control assignments are determined as a subset of the candidate access control assignments with a lowest aggregate over-privileging metric whose combined permissions and resource scopes cover at least a predetermined percentage of the paired activity data. A responsive action is performed based on the recommended access control assignments.

Abstract: A method for operating a computer system comprises transmitting first and second signals between a device interface and a first device over an elongate first conductor, the second signal including content of higher frequency than the first signal; transmitting a timing signal over an elongate second conductor, between a device interface and the first device, for synchronizing the second signal; and exposing at least the second signal and the timing signal to a second device distinct from the first device, the first device being substantially non-responsive to the second signal and to the timing signal, wherein at least the first or second elongate conductor conducts bidirectionally between the device interface and the first device.

Abstract: A method, computer program product, and computing system for processing a plurality of compatibility constraints for a cloud computing environment. A plurality of prevalence metrics for the cloud computing environment are processed. A graph structure mapping compatibility relationships between cloud computing resources is generated based upon, at least in part, the plurality of compatibility constraints. A constrained optimization process is defined for the graph structure. A testing schedule for the cloud computing resources is generated using the graph structure and the constrained optimization process.

Abstract: Computing resources deployed on the cloud can be susceptible to different types of malicious attacks based on vulnerabilities introduced in computer program instructions that define the computing resources. To address these types of attacks, methods, systems, apparatuses, and computer-readable storage mediums are described for identifying a resource attack path. A vulnerability identifier scans a set of computer program instructions to identify a vulnerability therein. A resource mapper generates a resource map that identifies a relationship between a portion of the set of computer program instructions and a resource executing in a cloud. An attack path identifier obtains a log that identifies telemetry events in the cloud. The attack path identifier further identifies an attack path based at least on the identified vulnerability, the resource map, and the log. A security event remediator performs a remediation action in response to the identifying the attack path.

Abstract: An extended two tower network is used to make both a recommendation of a content item that a given user may be interested in, and a recommendation of a user that may be interested in a given content item. The extended two tower network includes a content item sub-tower, a user sub-tower, and a fusion sub-model that are jointly trained to predict a probability that a given user is interested in a given content item. The content item sub-tower and the user sub-tower are used to make an initial prediction that the given user will be interested in the given content item. The initial prediction is then input to the fusion sub-model to make a final prediction. In the case of a candidate invitee recommendation, the initial prediction may be combined with one or more interaction features and the combination input to the fusion sub-model to make the final prediction.

Abstract: Systems generate and utilize ledger records of verifiable evidence that attest to the authenticity and integrity attributes of hardware and software components of a particular computing environment. The ledger record can be sealed to provide a level of assurance that the attestations and evidence in the ledger record have not been tampered with. The ledger record can be shared with different systems and referenced to assess a level of trust and compliance of a computing environment relative to a particular policy context. Subsequent endorsements and controls can also be based on the assessed trust and compliance derived from the ledger records.

Abstract: An enhanced router is described that improves network performance for AI workloads by providing in-network primitives that improve the performance of operations such as Broadcast and Reduce operations. The enhanced router leverages the observation that operations such as the Broadcast and Reduce primitives are more performant (e.g., reduced latency and bandwidth) when performed in the network rather than in graphics processing unites (GPUs) or central processing units (CPUs) which are often deployed as leaf nodes in a typical network topology in a datacenter.

Abstract: Systems and methods for managing multi-tier key-value (KV) caching for a generative artificial intelligence (AI) model are described. A system includes a lower tier memory, a higher tier memory, and a KV cache manager to: (1) determine whether a key-value (KV) cache corresponding to a user associated with a session exists in the lower tier memory, and (2) if the KV cache, or a portion thereof, does not exist in the lower tier memory, upon determining that at least a portion of the KV cache exists in the higher tier memory, load at least the portion of the KV cache into the lower tier memory. The system includes a KV cache engine to execute: (1) an extended prefill phase to generate keys and values for extending the KV cache in the lower tier memory, and (2) a decode phase to generate output tokens until a stopping criteria is met.

Abstract: A method, computer program product, and computing system for: accessing a first siloed dataset concerning one or more medical notes; processing the first siloed dataset to identify clinical items within the one or more medical notes, thus defining a first tagged clinical item set; selecting a first set of templates for the first siloed dataset based, at least in part, upon the first tagged clinical item set within the first siloed dataset; defining a first instruct dataset for the first siloed dataset based, at least in part upon the first set of templates and first tagged clinical item set; and training a generative model using the first instruct dataset, thus defining a generative model trained for the first siloed dataset.

Abstract: Innovations in intra block copy (“BC”) prediction mode facilitate intra BC prediction that is more effective in terms of rate-distortion performance and/or computational efficiency of encoding and decoding. For example, some of the innovations provide ways to select block vector (“BV”) values more effectively. Other innovations provide ways to encode/decode BV values more efficiently. Still other innovations address how to perform in-loop deblock filtering when a block has intra BC prediction mode, or address how to perform intra BC prediction when constrained intra prediction is enabled.

Abstract: In network security systems, graph-based techniques can be used to analyze data collected for a particular security incident, e.g., a command-and-control incident. In example embodiments, data extracted from data records of network activity and/or security alerts is used to generate a multipartite graph in which different entities (e.g., machines, processes, and domains or IP addresses) are represented as different types of nodes and relationships between the entities are represented as edges. The multipartite graph may be clustered, and the clusters be ranked according to some indicator of maliciousness (e.g., the number of associated security alerts or indicators of compromise (IoCs)). An output generated from the highest-ranking cluster(s) may serve, e.g., to identify new IoCs, or flow into mitigating actions taken in response to the incident.

Abstract: Systems, methods, and software are disclosed herein for discovering inconsistent latencies in data center environments. In an implementation, a method for detecting inconsistent latency in a network comprises generating test traffic comprising encapsulated packets to transmit along a route from a source device to a destination device and supplying the test traffic to a bundle of interfaces of the source device to transmit to the destination device over multiple physical paths. The method continues with identifying latencies of the encapsulated packets along the route, and identifying an occurrence of inconsistent latency with respect to the bundle of interfaces based at least on the latencies of the encapsulated packets.

Abstract: Implementations are directed to facilitating a data collaboration by debugging a data pipeline in production mode without exposing diagnostic logs generated by executing the data pipeline over production data contributed by collaborators. In an example implementation, a data collaboration tool treats the diagnostic logs as virtual data assets owned by the collaborators, permitting the collaborators to enable an option to save the diagnostic logs in the data trustee environment, define entitlements to reference the diagnostic logs in a pipeline, and build a debugging pipeline to evaluate the diagnostic logs. As such, a collaborator may trigger the data pipeline to generate a diagnostic log, and trigger a debugging pipeline to derive insights from the diagnostic log in the data trustee environment, without exposing the diagnostic log or the production data to the collaborators. As such, the insights may be exposed for debugging purposes without exposing collaborator data.

Abstract: Cloud computing techniques utilizing distributed application execution are disclosed herein. One example technique includes receiving a command to launch an application, and in response, determining an execution location corresponding to a type of data consumed by individual components of the application. Upon determining that one of the components is to be executed in a local computing facility, the example technique includes transmitting, from a public computing facility to the local computing facility, a request to execute the one of the components in the local computing facility instead of the public computing facility. Upon being authorized by the local computing facility, data is requested and received from the one of the components executed at the local computing facility without having direct access from the public computing facility to a data source at the local computing facility.

Abstract: A method of tracking a peripheral device is disclosed. A synchronization request signal is emitted via an infrared (IR) emitter of a head-mounted device (HMD). A peripheral device is configured to activate an IR constellation of IR emitters as acknowledgement of receiving the synchronization request signal. An image frame is captured via a camera of the HMD. It is determined that the peripheral device is synchronized with the HMD based at least on recognizing the activated IR constellation in the image. A peripheral device tracking image frame capture sequence is initiated. The sequence includes a plurality of odd image frames and even image frames. The peripheral device is configured to activate the IR constellation in odd image frames and deactivate the IR constellation in even image frames. A sequence of composite image frames is generated by subtracting content in corresponding even image frames from content in corresponding odd image frames.

Abstract: A method of registering a coordinate frame of an HMD with a second coordinate frame comprises receiving sensor data depicting the wearer of the HMD pointing at, gazing at, or touching, a real world feature, where the 3D position of the real world feature in the second coordinate frame is known. The method computes a 3D position of the real world feature, in the coordinate frame of the HMD, from the sensor data. A correspondence is stored comprising: the 3D position of the real world feature in the coordinate frame of the HMD, and a 3D position of the real world feature in the second coordinate frame. The method repeats so that a second correspondence is stored. The method registers the coordinate frame of the HMD and the second coordinate frame by computing registration from the correspondences.

Abstract: A transparent substrate that encodes data therein having optically readable identification indicia corresponding to identifying information regarding the substrate. The optically readable identification indicia may be readable from the transparent substrate by altering the reflectivity of the transparent substrate in indicia portions that may be read by a scanner or reader (e.g., a barcode reader). The optically readable identification indicia may be provided on a common surface with a data zone in which data is encoded in the transparent substrate. Alternatively or additionally, the optically readable identification indicia may be provided on another, different surface than the surface on which the data is encoded in a data zone.

Abstract: An electronic device initiates a scan of a display frame in the display, the scan displaying a sequence of display rows from a starting point of the frame to an ending point of the display frame. A horizontal blanking period is positioned in time between each display row during which display emissions are turned off. The electronic device progressively scans a first sub-sequence of the display rows in the sequence during a first time period and then scans a display row having a quiet period inserted into a horizontal blanking period adjacent to the display row. Display pixel operations are delayed until expiration of the quiet period during the horizontal blanking period. The electronic device then progressively scans a second sub-sequence of the display rows in the sequence during a second time period.

Abstract: Aspects of the disclosure include methods and systems for proactive content service delivery. An exemplary method can include determining a delivery interval for a content update by defining delivery interval dimensions, each of a unique combination of time and user cohorts, determining a time cohort and a user cohort for the user, and selecting a predetermined delivery interval of the delivery interval dimension matching the time cohort and the user cohort of the user as the delivery interval. The method further includes determining a delivery content by training a classification model to predict whether content will result in an impression, inputting, into the classification model, the content update, receiving an impression score for the content update, and selecting one of a full update and a partial update as the delivery content from the impression score. The content update can be pushed to a client device according to the delivery interval.

Abstract: Innovations in control and use of chroma quantization parameter (“QP”) values that depend on luma QP values. More generally, the innovations relate to control and use of QP values for a secondary color component that depend on QP values for a primary color component. For example, during encoding, an encoder determines a QP index from a primary component QP and secondary component QP offset. The encoder maps the QP index to a secondary component QP, which has an extended range. The encoder outputs at least part of a bitstream including the encoded content. A corresponding decoder receives at least part of a bitstream including encoded content. During decoding, the decoder determines a QP index from a primary component QP and secondary component QP offset, then maps the QP index to a secondary component QP, which has an extended range.

Abstract: Innovations in control and use of chroma quantization parameter (“QP”) values that depend on luma QP values. More generally, the innovations relate to control and use of QP values for a secondary color component that depend on QP values for a primary color component. For example, during encoding, an encoder determines a QP index from a primary component QP and secondary component QP offset. The encoder maps the QP index to a secondary component QP, which has an extended range. The encoder outputs at least part of a bitstream including the encoded content. A corresponding decoder receives at least part of a bitstream including encoded content. During decoding, the decoder determines a QP index from a primary component QP and secondary component QP offset, then maps the QP index to a secondary component QP, which has an extended range.

Abstract: Innovations in intra block copy (“BC”) prediction mode facilitate intra BC prediction that is more effective in terms of rate-distortion performance and/or computational efficiency of encoding and decoding. For example, some of the innovations provide ways to select block vector (“BV”) values more effectively. Other innovations provide ways to encode/decode BV values more efficiently. Still other innovations address how to perform in-loop deblock filtering when a block has intra BC prediction mode, or address how to perform intra BC prediction when constrained intra prediction is enabled.

Abstract: Systems, methods, and devices are provided for dynamically mapping and rerouting a non-optimized binary to an optimized binary. Embodiments identify a position-independent portion within a non-optimized binary and receive metadata associated with an optimized binary. Further, embodiments identify a location identifier within the optimized binary from the metadata and generate a delta value between the position-independent portion and the location identifier. Embodiments also update the position-independent portion with the delta value.

Abstract: An input tensor is received in an attention layer of machine learning (ML) network. A model query adapter generates a model query tensor based on the input tensor. An embedded knowledge base (KB) comprises a KB key tensor and a KB value tensor. A KB query adapter generated a KB query tensor based on the input tensor. An attention function combines attention over a model value tensor based on the model query tensor and a model key tensor with attention over the KB value tensor based on the KB query tensor and the KB key tensor, resulting in an output token.

Abstract: System, methods, apparatuses, and computer program products are disclosed for auto-scaling of a deployment based on resource utilization data for a workload executing on the deployment. A resource availability is determined based on the resource utilization data and a current resource allocation of the deployment. A severity of resource throttling of the workload may be determined based on the resource utilization data, and a scaling factor is determined based at least on the severity of resource throttling. In response to at least the resource availability satisfying a predetermined condition with a predetermined threshold, the deployment is scaled based on the scaling factor.

Abstract: Examples describe using a two dimensional floor plan, rather than a three dimensional (3D) reconstruction of a scene, to localize a camera. A floorplan of an environment is accessed and an image of the environment captured by a camera in the environment is received. The image is input to a trained neural network to predict an array of rays from the camera to surfaces in the environment indicated in the floorplan. Using the array of rays it is then possible to compute 3D position and orientation of the camera with respect to the floorplan.

Abstract: Technology is disclosed for programmatically determining, for a segment of a meeting recording, a user-specific adaptive playback speed, and generating a time-stretched segment playable at the adaptive playback speed. The adaptive playback speed is faster or slower than a default playback speed of the meeting recording. To determine the adaptive playback speed, this disclosure provides technologies to determine a playback data feature based on user-meeting data. The adaptive playback is generated based on the playback data feature. The segment is time-stretched to the adaptive playback speed to generate an updated meeting recording including the segment that is time-stretched and playable at the adaptive playback speed. In this manner, an updated meeting recording, specific to a user, and playable at an adaptive playback speed based on user-meeting data may reduce bandwidth associated with user's manually editing videos or rewinding playback, while improving user experience.

Abstract: Approaches to robust encoding and decoding of escape-coded pixels in a palette mode are described. For example, sample values of escape-coded pixels in palette mode are encoded/decoded using a binarization process that depends on a constant value of quantization parameter (“QP”) for the sample values. Or, as another example, sample values of escape-coded pixels in palette mode are encoded/decoded using a binarization process that depends on sample depth for the sample values. Or, as still another example, sample values of escape-coded pixels in palette mode are encoded/decoding using a binarization process that depends on some other fixed rule. In example implementations, these approaches avoid dependencies on unit-level QP values when parsing the sample values of escape-coded pixels, which can make encoding/decoding more robust to data loss.

Abstract: The present disclosure provides a system that controls an interim connection between two computing devices. The system first determines if a person if physically present at the host device by detecting an interaction with an input device attached to the host device. In response to the input, the system generates a connection identifier that is displayed at the host device. The display of a graphical element for communicating the connection identifier enables the guest device to receive the connection identifier. The graphical element can be a Quick Response (QR) code, bar code, text, etc. The guest device can send a connection request to the server using the connection identifier. In response to the request, the server verifies the connection identifier and establishes a data connection between the host device and the guest device. Data can be exchanged between the host device and the guest device via the server.

Abstract: A technique is described for compressing input information fed to a machine-trained generative model. The technique includes: receiving original input information having a plurality of sentences; performing word-level encoding of the original input information using a first part of a machine-trained transformer model, to provide word-level encoded information; performing sentence-level encoding of the word-level encoded information using a second part of the machine-trained transformer model, to provide scores associated with the first plurality of sentences; selecting a subset of the sentences in the original input information based on the scores, to provide modified input information; and providing the modified input information to the machine-trained generative model. The operation of word-level encoding performs parallel processing on portions of the original input information.

Abstract: This document relates to using machine learning models to automate the process of generating alert rules for computing systems. The disclosed techniques can generate templates that match log entries from system logs. Then, the templates can be employed to generate alert rules that specify certain conditions when alerts should be triggered. In some implementations, the machine learning model is invoked responsive to a determination that a given log entry does not already match an existing template. Thus, the number of invocations of the machine learning model is proportional to the number of templates, rather than the number of log entries. Since there are often many more log entries than templates, this approach can yield significant computational savings compared to approaches that might process each log entry using the machine learning model.

Abstract: A method for optical multiplication includes displaying a first encoded electromagnetic signal on a shared memory space, the first encoded electromagnetic signal being based on an input value and displaying a second encoded electromagnetic signal on the shared memory space, the second encoded electromagnetic signal being based on a learned parameter associated with the input value. The method includes detecting a cumulative electromagnetic signal on the shared memory space including a combination of at least some of the first encoded electromagnetic signal and at least some of the second encoded electromagnetic signal and determining a product of the input value and the learned parameter based on decoding the cumulative electromagnetic signal.

Abstract: Systems and methods are provided for implementing near-real-time artificial intelligence (“AI”) enhancement of video or augmented reality (“AR”) presentations of live events. In examples, a computing system identifies, from a plurality of spoken words in a live audio recording of individuals speaking, spoken words to modify, based on a set of parameters and/or a profile of a user, using an AI agent. The computing system causes a generative AI system to modify, in near-real-time, (a) the identified spoken words with corresponding replacement words in each of the live video recording and the corresponding live audio recording, based on the profile of the user; (b) portions of the live video recording and the corresponding live audio recording to generate video and audio smooth transitions before and after the replacement words are incorporated in the live video recording and the corresponding live audio recording, respectively.

Abstract: Innovations in encoder-side options for intra block copy (“BC”) prediction mode facilitate intra BC prediction that is more effective in terms of rate-distortion performance and/or computational efficiency of encoding. For example, some of the innovations relate to concurrently performing block vector (“BV”) estimation and making block splitting decisions for a block. Other innovations relate to selectively merging blocks into a larger block during BV estimation.

Abstract: Systems and methods are provided for determining a root cause of an incident that occurred in a 5G/6G multi-access edge computing and core network system. In particular, the disclosed technology is directed to using a pre-trained generative model to determine the root cause for an incident as recorded in event data of a system log. The present disclosure generates a prompt for the generative model to determine the root cause for an incident as recorded in the system log. The prompt comprises a combination including event data from a system log, function hierarchy graph data, and function information as grounding information in a prefix of the prompt. The prompt further includes a question that requests determining a root cause of an incident as recorded in the event data. An answer as generated by the pre-trained generative model indicates of the root cause and a function call associated with the incident.

Abstract: Systems and methods are provided for offloading a task from a central processor in a radio access network (RAN) server to one or more heterogeneous accelerators. For example, a task associated with one or more operational partitions (or a service application) associated with processing data traffic in the RAN is dynamically allocated for offloading from the central processor based on workload status information. One or more accelerators are dynamically allocated for executing the task, where the accelerators may be heterogeneous and my not comprise pre-programming for executing the task. The disclosed technology further enables generating specific application programs for execution on the respective heterogeneous accelerators based on a single set of program instructions.

Abstract: Systems and methods are provided for implementing post-production artificial intelligence (“AI”) enhancement of video presentations based on script modification. In examples, a computing system identifies, from a plurality of spoken words in an audio file of individuals speaking, spoken words to modify, based on a set of parameters and/or a profile of an intended audience, using an AI agent. If there is a script for the audio file, the computing system causes a generative AI system to modify the script to replace the identified spoken words with replacement words, and to modify corresponding portions of the audio file to replace the identified spoken words with the replacement words based on the modified script. The computing system further causes the generative AI system to generate audio transitions before and after the replacement words, and to modify the applicable portions of a corresponding video file based on the modified audio file.