Techniques and systems are disclosed to perform language model prompt optimization for complex task resolution, such as for natural language prompts used with large language models (LLMs). An example method includes: forming different combinations of subtasks that accomplish some overall task; generating corresponding combinations of subprompts; producing a performance prediction of combinations of subprompts to accomplish the overall task using the language model; evaluating the performance predictions against a threshold performance condition; selecting a combination for performing the overall task that meets the threshold performance condition; executing the subprompts corresponding to the selected combination of subtasks using the language model to accomplish the overall task; and outputting a response based on a result of the executing of the subprompts.
An AR/VR headset that is configured to be worn on a user’s head. The headset includes a first sprocket, a second sprocket and a ball chain engaging the first and second sprockets. A rotating actuator engages the first sprocket such that the first sprocket rotates and the ball chain translates along the first and second sprockets when torque is applied to the rotating actuator. The ball chain applies torque to the second sprocket as the ball chain translates along the first and second sprockets. A transmission gear engages the second sprocket and rotates with the second sprocket. A first arm includes a first set of internal gear teeth, and a second arm includes a second set of internal gear teeth. The transmission gear engages the first and second sets of gear teeth such that rotation of the transmission gear moves the second arm relative to the first arm.
G06F 1/16 - Constructional details or arrangements
F16M 13/04 - Other supports for positioning apparatus or articlesMeans for steadying hand-held apparatus or articles for supporting on, or holding steady relative to, a person, e.g. by chains
3.
CONDITIONAL SIGNALLING OF REFERENCE PICTURE LIST MODIFICATION INFORMATION
Innovations in signaling of reference picture list (“RPL”) modification information. For example, a video encoder evaluates a condition that depends at least in part on a variable indicating a number of total reference pictures. Depending on the results of the evaluation, the encoder signals in a bitstream a flag that indicates whether an RPL is modified according to syntax elements explicitly signaled in the bitstream. A video decoder evaluates the condition and, depending on results of the evaluation, parses from a bitstream a flag that indicates whether an RPL is modified according to syntax elements explicitly signaled in the bitstream. The condition can be evaluated as part of processing for an RPL modification structure that includes the flag, or as part of processing for a slice header. The encoder and decoder can also evaluate other conditions that affect syntax elements for list entries of the RPL modification information.
H04N 19/58 - Motion compensation with long-term prediction, i.e. the reference frame for a current frame not being the temporally closest one
H04N 19/105 - Selection of the reference unit for prediction within a chosen coding or prediction mode, e.g. adaptive choice of position and number of pixels used for prediction
H04N 19/109 - Selection of coding mode or of prediction mode among a plurality of temporal predictive coding modes
H04N 19/134 - Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
H04N 19/46 - Embedding additional information in the video signal during the compression process
H04N 19/503 - Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
H04N 19/573 - Motion compensation with multiple frame prediction using two or more reference frames in a given prediction direction
H04N 19/70 - Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by syntax aspects related to video coding, e.g. related to compression standards
H04N 19/85 - Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using pre-processing or post-processing specially adapted for video compression
4.
CODE REVIEW COMMENT GENERATION VIA INSTRUCTION PROMPTING WITH INTENT
A code review is automatically generated by a large language model given a prompt that includes code changes made to a source code program, an associated intent, and an extended context. The intent represents an issue with the code changes from a code reviewer's perspective and is predicted from a neural classifier given the code changes in a code diff format. The neural classifier is a neural encoder transformer model pre-trained on various code review datasets and fine-tuned on code diff hunks of code changes labeled with an intent.
A hardware memory allocator of a compute system is described. The allocator stores batches of free identifiers (IDs) as elements of a linked list in a memory external to the system, each free ID identifying a respective segment of the memory, and transfers batch(es) of free IDs from the head of the linked list to an internal buffer. The allocator then responds to memory requests by allocating free IDs from among the batch(es) of free IDs in the buffer and to memory deallocation requests by deallocating previously-allocated free IDs. The allocator transfers batches of deallocated free IDs from the buffer to the head of the linked list, wherein transferring a batch of deallocated free IDs includes selecting one of the deallocated free IDs in the buffer exclusive of the batch and storing the batch of at the memory segment identified by the selected deallocated free ID.
There is provided a method that includes (a) obtaining a first voice vector that was derived from a signal of a voice that was sampled at a first sampling frequency, (b) obtaining a second voice vector that was derived from a signal of a voice that was sampled at a second sampling frequency, (c) mapping the second voice vector into a mapped voice vector in accordance with a machine learning model, and (d) comparing the first voice vector to the mapped voice vector to yield a score that indicates a probability that the first voice vector and the second voice vector originated from a same person.
A data processing system with a processor and memory is configured to: implement a feature pipeline to receive telemetry data generated in response to activity in a set of applications; analyze the telemetry data using a feature extraction unit of the feature pipeline to extract feature data from the telemetry data; operate a scoring pipeline to classify the user as working onsite or remotely based on the feature data extracted from the telemetry data, the scoring pipeline utilizing a machine-learning model trained to receive feature data as an input and to output a classification indicating that the user is working onsite at a location associated with the enterprise or remotely at a location not associated with the enterprise; and allocate computing and network resources using a resource allocation unit based on the classification.
H04L 47/76 - Admission controlResource allocation using dynamic resource allocation, e.g. in-call renegotiation requested by the user or requested by the network in response to changing network conditions
H04L 43/10 - Active monitoring, e.g. heartbeat, ping or trace-route
H04L 47/2441 - Traffic characterised by specific attributes, e.g. priority or QoS relying on flow classification, e.g. using integrated services [IntServ]
8.
ROUTING NETWORK TRAFFIC USING TIME ZONE INFORMATION
A DNS request is received from a requesting computing system at a DNS service. The DNS service serves the DNS request by identifying a service endpoint that is closest to the requesting computing system in terms of latency, based on a time zone in which the requesting computing system is located.
H04L 67/63 - Routing a service request depending on the request content or context
H04L 45/121 - Shortest path evaluation by minimising delays
H04L 61/4511 - Network directoriesName-to-address mapping using standardised directoriesNetwork directoriesName-to-address mapping using standardised directory access protocols using domain name system [DNS]
H04L 67/52 - Network services specially adapted for the location of the user terminal
9.
MACHINE LEARNING SYSTEM WITH MODALITY SELECTION MODEL
A computing system including one or more input devices, output devices, and processing devices. The one or more processing devices receive a user input via an input device and receive interaction context metadata associated with the user input. The interaction context metadata includes an input modality of the user input and output modalities of the one or more output devices. At a multimodal machine learning (ML) system, the one or more processing devices select one or more of the output modalities at a modality selection ML model based at least on the user input and the interaction context metadata. At one or more generative ML models, based at least on the user input, the one or more processing devices compute one or more respective model outputs that respectively have the one or more selected output modalities. The one or more processing devices output the one or more model outputs.
Implementations for capturing user activity as memory for question-answering are provided. One example includes a computing system comprising: processing circuitry and memory storing instructions that, during execution, cause the processing circuitry to: receive a query; perform semantic searching on an L2 index comprising a database of event summaries based on the query to retrieve an L2 answer comprising one or more relevant event summaries; generate one or more L1 answers based on the L2 answer using an L1 index comprising a database of screen summaries associated with a plurality of L0 screenshot images, wherein each of the one or more L1 answers comprises relevant screen summaries corresponding to L0 screenshot images associated with a respective relevant event summary; pass the query and the one or more L1 answers to a large language model to generate an answer to the query; and output the answer to the query.
The present disclosure relates to enabling a private connection to a Content Delivery Network (CDN). A data packet encapsulated with a tenant specific VNI is received by a driver, wherein the data packet comprising a user IP address in a source IP field and a subnet IP address in a destination IP field. The driver decapsulates the data packet to retrieve the tenant specific VNI and replaces the destination IP field with an internal IP address. The driver transmits the data packet to an edge server, wherein the data packet is processed by the edge server. The driver receives a returning data packet from the edge server comprising the internal IP address in the source IP field and the user IP address in the destination IP field. The driver replaces the source IP field with the subnet IP address and transmits the returning data packet to the edge router.
H04L 61/4511 - Network directoriesName-to-address mapping using standardised directoriesNetwork directoriesName-to-address mapping using standardised directory access protocols using domain name system [DNS]
12.
MACHINE LEARNING SYSTEM WITH ASYNCHRONOUS COMMUNICATION
A computing system (10) including one or more processing devices (12) configured to receive a first user input (300) at a user interface (20) of a machine learning (ML) system (30). In response to receiving the first user input, the one or more processing devices initiate a first model call (304) to a first ML model (312). During execution of the first model call, the one or more processing devices receive one or more second user inputs (302) and perform one or more second model calls (306) to a second ML model (314) to compute one or more second model outputs (310). The one or more processing devices output the one or more second model outputs to the user interface. Subsequently to outputting the one or more second model outputs, the one or more processing devices receive and output a first model output (308) computed at the first ML model in response to receiving the first user input.
Techniques and systems are disclosed to perform language model prompt optimization for complex task resolution, such as for natural language prompts used with large language models (LLMs). An example method includes: forming different combinations of subtasks that accomplish some overall task; generating corresponding combinations of subprompts; producing a performance prediction of combinations of subprompts to accomplish the overall task using the language model; evaluating the performance predictions against a threshold performance condition; selecting a combination for performing the overall task that meets the threshold performance condition; executing the subprompts corresponding to the selected combination of subtasks using the language model to accomplish the overall task; and outputting a response based on a result of the executing of the subprompts.
A system and method for providing automated event of interest detection and reporting for software testing. The technology described herein uses natural language processing to identify events of interest based on natural language included in log files messages and/or telemetry event items of software under test. Identified events of interest are reported to one or more downstream services. In some examples, events of interest are mapped to action sequences of interest that can be replayed to reproduce events of interest without requiring event of interest criteria to be defined. In some examples, events of interest are reported to a health/status dashboard and/or a listener that responds to reported events of interest.
Techniques are described herein that are capable of generating a content signature of a textual communication using OCR and text processing. The textual communication is rendered. Text is extracted from the rendered textual communication using OCR. Customization is removed from the text to provide a templatized version of the rendered textual communication that includes de-customized text. The de-customized text is parsed into tokens. Each token includes a respective subset of a plurality of characters. The tokens are converted into respective numbers. Each number is processed using fuzzy hash functions to provide respective hash values associated with the respective token. Representative hash values are selected for the respective fuzzy hash functions by selecting each representative hash value from the hash values that are processed using the respective fuzzy hash function. A content signature of the textual communication is generated by bitwise concatenating at least portions of the respective representative hash values.
Systems and methods for providing adaptive and conversational chat cards are provided herein. For example, a method may include receiving, from a client device, an indication to start a chat card exercise, determining, by a chat card engine, a source document based on the chat card exercise, and generating chat cards based on the source document. Each of the chat cards may correspond to a subtopic present in the source document. As such, the chat card engine may generate a first prompt based on a first subtopic and provide the first prompt within a first chat card to the client device. The chat card engine may receive a first query from the client device responsive to the first prompt and determine an understanding score. The chat card engine may then generate a second prompt based on the understanding score and provide the second prompt in the first chat card.
G09B 3/02 - Manually- or mechanically-operated teaching appliances working with questions and answers of the type wherein the student is expected to construct an answer to the question which is presented or wherein the machine gives an answer to the question presented by the student
H04L 51/02 - User-to-user messaging in packet-switching networks, transmitted according to store-and-forward or real-time protocols, e.g. e-mail using automatic reactions or user delegation, e.g. automatic replies or chatbot-generated messages
Some embodiments construct a set of build dependencies for a program without access to a full set of build instructions. When multiple clashing name resolutions are identified for a particular dependency, a union of the alternative versions is formed. Intermediate representations of the union of program versions, such as symbol tables, abstract syntax trees, and other internal compiler data structures, are emitted to persistent non-volatile storage, instead of using a single resolution to create temporary intermediate data to build an executable program. Security analysis and licensing analysis utilize the persisted program representations to analyze the union of multiple overlapping but different versions of the program.
Methods of secure resource authorization for external identities using remote principal objects are performed by systems and devices. An external entity creates a user group and defines entitlements to an owning entity’s secure resource as a set of permissions for the group. An immutable access template with the permissions and an access policy for the secure resource are provided to the owning entity for approval. On approval, a remote principal object is created in the owner directory according to the permissions and access policy. A remote principal that is a group member requests access via an interface to the owner domain using external domain credentials. The identity of the remote principal is verified against the remote principal object by a token service. Verification causes generation and issuance of a token, with the enumerated entitlements, to the remote principal interface affecting a redirect for access to the secure resource.
Examples of the present disclosure describe systems and methods for selecting test combinations of hardware and software features for product validation. In examples, a test scheduler of a system receives metrics associated with a first batch of test combinations for multiple computing entities. The metrics may include data associated with a fleet prevalence and a fleet risk determined for the multiple computing entities. The scheduler provides one or more of the metrics to a machine learning algorithm. The test scheduler determines a second batch of test combinations associated with a subset of the multiple computing entities based on an output of the machine learning algorithm. The second batch of combinations may include a subset of the first batch of test combinations. The system executes the second batch of test combinations for the subset of the multiple computing entities.
A cooling system for a set of solid state drive (SSD) modules includes a thermal mounting frame. The thermal mounting frame includes a thermal transfer plate, and a plurality of parallel fins that project outward from a first side of the thermal transfer plate along a length dimension. The first side of the thermal transfer plate in combination with each neighboring pair of fins of the plurality of parallel fins defines a slot that accommodates a corresponding SSD module of the set of SSD modules. The thermal transfer plate includes a plurality of parallel spines in which each spine projects into a core region of a corresponding fin of the plurality of parallel fins along a base portion of the length dimension of the fin. For each slot, each fin of the neighboring pair of fins that defines the slot includes an exterior thermal transfer layer.
A cached random access memory (RAM) counter system includes a plurality of linear feedback shift register (LFSR) counters associated with count values that maintain a unique sequence of bits associated with least-significant bits of a respective count value and a respective carry bit, a RAM circuit that stores a plurality of most-significant bits of each respective count value, a multi-bit multiplexer that receives the unique sequence of bits and the respective carry bit from each of the plurality of LFSR counters and selectively provides the unique sequence of bits and the respective carry bit for the respective count value; and an adder circuit that adds the carry bit output for the respective count value from the multi-bit multiplexer to the plurality of most-significant bits of the respective count value and provides updated most-significant bits of the respective count value for storage in the RAM circuit.
A method for recommending a mixed-precision data type configuration to accelerate execution of a language model includes identifying a plurality of mixed-precision test configurations that assign mixed-precision data types among components of the language model; measuring saliency metrics for the language model while the language model is executed in a test environment configured in accord with multiple of the plurality of mixed-precision test configurations; generating a saliency ranking that ranks the components based on the saliency metrics; and generating the recommended mixed-precision data type configuration. The recommended mixed-precision data type configuration assigns a first, reduced-precision data type to a least salient subset of the components identified in the saliency ranking while assigning a second, higher-precision data type to a most salient subset of the components identified in the saliency ranking.
Aspects herein pertain to a form of detection rule enrichment, in which a detection rule implemented in a security system is enriched with relevant context to assist actions such as reporting or remediation when the rule is triggered based on pattern matching to telemetry. A two-stage rule classification process is disclosed, where a first classifier assigns a first threat classification label (e.g., tactic) and a second classifier assigns a second threat classification label (e.g., technique). The classification is hierarchical in that second classifier uses the output of the first classifier (the first threat classification label) to determine the second threat classification label.
Methods, apparatuses, and products for database query acceleration. Methods include receiving, by a query execution platform, database query plan data from a database host, the database query plan data comprising a query execution plan and one or more database tables, translating the database query plan data into format associated with a coprocessor of the query execution platform, generating a coprocessor query execution plan for the coprocessor based on the query execution plan of the database query plan data, identifying one or more operators for executing the coprocessor query execution plan, selecting an implementation for each of the one or more operators based on one or more characteristics of data on which the one or more operators are to be applied based on the coprocessor execution plan, and executing the coprocessor query execution plan using the coprocessor.
Information is received indicating a plurality of files that are determined to be associated with a user's computing activity in the computing system and meet a criterion for being variants of one another. A polymorphic placeholder is generated that corresponds to the plurality of files. The plurality of files are selected based on a threshold relevance to the criterion, and the plurality of files are abstracted from underlying storage locations in the computing system.
A computing system including one or more processing devices configured to receive a first user input at a user interface of a machine learning (ML) system. In response to receiving the first user input, the one or more processing devices initiate a first model call to a first ML model. During execution of the first model call, the one or more processing devices receive one or more second user inputs and perform one or more second model calls to a second ML model to compute one or more second model outputs. The one or more processing devices output the one or more second model outputs to the user interface. Subsequently to outputting the one or more second model outputs, the one or more processing devices receive and output a first model output computed at the first ML model in response to receiving the first user input.
Devices, systems, and methods for post-quantum (PQ) secure (PQS) communications are provided. A method includes authenticating, by a device and with a first server, using a legacy device identity, receiving and installing, by the device, a firmware upgrade that provides the device an ability to communicate in a PQS manner, establishing a PQS communication channel between the device and a second server, mutually authenticating, by the device and with the second server, using a PQ identity (PQI) of the device, the PQI different from the legacy identity, and responsive to successful mutual authentication, communicating between the device and the second server, using the PQS communication channel.
H04L 9/32 - Arrangements for secret or secure communicationsNetwork security protocols including means for verifying the identity or authority of a user of the system
The disclosed techniques provide a method for reducing power consumption in a logic circuit. A design of the logic circuit is provided. The logic circuit includes storage circuits that store state information and receive clock signals. Toggle activity data indicating whether the stored state information changes in response to a respective clock signal is determined. A graph that represents the toggle activity data is constructed. The graph includes nodes and edges. The nodes correspond to the storage circuits. The edges represent statistical relationships among the determined changes in the stored state information. A graph neural network to the graph is applied. The application includes generating an embedding that characterizes patterns in the toggle activity data. Via a reinforcement learning agent, control logic that selectively gates the clock signals provided to storage circuits for which the stored state information is predicted, based on the embedding, to remain unchanged is determined.
G06F 30/27 - Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
Aspects herein pertain to a form of detection rule enrichment, in which a detection rule implemented in a security system is enriched with relevant context to assist actions such as reporting or remediation when the rule is triggered based on pattern matching to telemetry. A two-stage rule classification process is disclosed, where a first classifier assigns a first threat classification label (e.g., tactic) and a second classifier assigns a second threat classification label (e.g., technique). The classification is hierarchical in that second classifier uses the output of the first classifier (the first threat classification label) to determine the second threat classification label.
A computing system (10) including one or more input devices (16), output devices (18), and processing devices (12). The one or more processing devices receive a user input (22) via an input device and receive interaction context metadata (50) associated with the user input. The interaction context metadata includes an input modality (52) of the user input and output modalities (54) of the one or more output devices. At a multimodal machine learning (ML) system (30), the one or more processing devices select one or more of the output modalities (72) at a modality selection ML model (70) based at least on the user input and the interaction context metadata. At one or more generative ML models (74), based at least on the user input, the one or more processing devices compute one or more respective model outputs (24) that respectively have the one or more selected output modalities. The one or more processing devices output the one or more model outputs.
A display device includes a plurality of image-forming elements arranged in a grid, including a plurality of rows and columns. A gate integrated circuit (IC) outputs a clock signal. A gate output buffer controls whether the clock signal is received by image-forming elements in a corresponding one or more rows. A source IC outputs data signals to control the plurality of image-forming elements. A block selection component controls, via a block selection signal, whether image-forming elements in a corresponding one or more columns are refreshed based at least in part on the data signals. A timing controller controls the output enable signal and the block selection signal to thereby control refresh rates for the plurality of image-forming elements, such that a first display region is refreshed at a first refresh rate, and a second display region is refreshed at a second refresh rate.
G09G 3/3266 - Details of drivers for scan electrodes
G09G 3/3275 - Details of drivers for data electrodes
G09G 3/36 - Control arrangements or circuits, of interest only in connection with visual indicators other than cathode-ray tubes for presentation of an assembly of a number of characters, e.g. a page, by composing the assembly by combination of individual elements arranged in a matrix by control of light from an independent source using liquid crystals
09 - Scientific and electric apparatus and instruments
42 - Scientific, technological and industrial services, research and design
Goods & Services
Computer hardware; computer software; downloadable software; computer firmware; computer chips; semiconductor chips; integrated circuits; cryptographic apparatus and instruments; data security apparatus and instruments; computer hardware, software and firmware for cryptography, data encryption and decryption, authentication, digital signatures, cryptographic key generation, storage, management and lifecycle management, identity and access management, computer security and data security; computer hardware and software for implementing and managing cryptographic and post-quantum cryptographic security protocols; computer hardware and software for managing, configuring, monitoring and operating cryptographic hardware and computer security hardware; computer hardware and software for managing user access, access permissions, access controls, authentication credentials and cryptographic credentials; computer hardware and software for the development, implementation and testing of cryptographic security systems; hardware security modules [HSMs]; integrated hardware security modules [HSMs]; cryptographic appliances; computer hardware, semiconductor chips and integrated circuits incorporating cryptographic security functionality. Software as a service [SaaS]; platform as a service [PaaS]; cloud computing; providing temporary use of online non-downloadable software; software as a service [SaaS] featuring software for cryptography, data encryption and decryption, authentication, digital signatures, cryptographic key generation, storage, management and lifecycle management, identity and access management, computer security and data security; software as a service [SaaS] featuring software for managing, configuring, monitoring and operating hardware security modules [HSMs], integrated hardware security modules [HSMs], cryptographic hardware and computer security hardware; software as a service [SaaS] featuring software for implementing and managing cryptographic and post-quantum cryptographic security protocols; software as a service [SaaS] featuring software for managing user access, access permissions, access controls, authentication credentials and cryptographic credentials; software as a service [SaaS] featuring software for the development, implementation and testing of hardware security modules [HSMs] and cryptographic security systems; providing temporary use of online non-downloadable software for the aforesaid purposes.
09 - Scientific and electric apparatus and instruments
42 - Scientific, technological and industrial services, research and design
Goods & Services
Downloadable software for managing, configuring, monitoring and operating hardware security modules (HSMs) and integrated HSMs; downloadable software for cryptographic key generation, storage, management and lifecycle management, data encryption and decryption, authentication, and cryptographic security; downloadable software for managing and monitoring cryptographic hardware, computer security hardware, semiconductor chips and integrated circuits; downloadable software and firmware for implementing and managing cryptographic and post-quantum cryptography security protocols; downloadable software for identity and access management, namely, managing user access, access permissions, access controls, authentication credentials and cryptographic credentials for hardware security modules (HSMs) and cryptographic systems; computer hardware and downloadable software for the development, implementation and testing of hardware security modules and cryptographic security systems; computer hardware; computer chips; semiconductor chips; integrated circuits; computer hardware, semiconductor chips and integrated circuits incorporating cryptographic security functionality; hardware security modules (HSMs); integrated hardware security modules (HSMs); secure cryptographic devices in the nature of hardware security modules (HSMs), cryptographic appliances for encryption, digital signature, authentication and cryptographic key management Software as a service (SaaS) featuring software for managing, configuring, monitoring and operating hardware security modules (HSMs) and integrated HSMs; software as a service (SaaS) featuring software for cryptographic key generation, storage, management and lifecycle management, data encryption and decryption, authentication, and cryptographic security; software as a service (SaaS) featuring software for managing and monitoring cryptographic hardware, computer security hardware, semiconductor chips and integrated circuits; software as a service (SaaS) featuring software for implementing and managing cryptographic and post-quantum cryptography security protocols; software as a service (SaaS) featuring software for identity and access management, namely, managing user access, access permissions, access controls, authentication credentials and cryptographic credentials for hardware security modules (HSMs) and cryptographic systems; software as a service (SaaS) featuring software for the development, implementation and testing of hardware security modules and cryptographic security systems.
34.
PIXEL-BASED AUTOMATED TESTING OF A NAVIGABLE SIMULATED ENVIRONMENT
A computing system for pixel-based automated testing of a navigable simulated environment includes one or more processors configured to execute, in a run-time inference phase, an application testing program. The application testing program includes a machine learning model trained to detect errors in the application under test. A screen image of the simulated environment is processed by an object detection module to determine if a key object is present in the screen image. If a key object is present in the screen image, the application testing program executes an object investigation module to generate investigation inputs to investigate the key object. If a key object is not present in the screen image, the application testing program executes an environment exploration module to generate an environment exploration action to be provided to the application under test as simulated user input.
G06F 3/04815 - Interaction with a metaphor-based environment or interaction object displayed as three-dimensional, e.g. changing the user viewpoint with respect to the environment or object
G06T 7/70 - Determining position or orientation of objects or cameras
G06T 19/00 - Manipulating 3D models or images for computer graphics
35.
CONTEXTUAL LONG-TERM SURVIVAL OPTIMIZATION FOR CONTENT MANAGEMENT SYSTEM CONTENT SELECTION
A method involves first receiving a set of data on rewards associated with previously chosen content variant choices, selected based on an initial content variant choice model. This initial model is informed by a prior set of data. A second, updated content variant choice model is then determined based on this first set of reward data. When a request for selecting a content variant choice is received, it comes with contextual features. The method involves estimating the expected rewards for a range of content variant choices, considering these contextual features. Subsequently, a specific content variant choice is chosen based on both the updated model and the anticipated rewards. Finally, the chosen content variant is displayed on a device, responding to the initial request.
Embodiments of the present disclosure include a multipurpose multiply-accumulator (MAC) array circuit comprising one or more input memories for receiving operands and a plurality of multiply-accumulator circuits each selectively coupled to the one or more input memories to receive at least a pair of operands and generate a result. Each of the plurality of multiply-accumulator circuits receives operands from the one or more input memories independently. Additionally, selection of operands from the one or more input memories is controlled based on at least an operation and/or data types, where different operation and/or data types configure the plurality of multiply-accumulator circuits to receive different pairs of operands from the one or more input memories to execute particular operation types.
G06F 7/544 - Methods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation using non-contact-making devices, e.g. tube, solid state deviceMethods or arrangements for performing computations using exclusively denominational number representation, e.g. using binary, ternary, decimal representation using unspecified devices for evaluating functions by calculation
G06F 5/06 - Methods or arrangements for data conversion without changing the order or content of the data handled for changing the speed of data flow, i.e. speed regularising
37.
SYSTEM AND METHOD FOR MULTI-CHANNEL SPEECH PRIVACY PROCESSING
A method, computer program product, and computing system for receiving a speech signal from a single microphone. A sensitive speech component is identified from the speech signal. In response to identifying the sensitive speech component, a filtered speech signal is generated by removing the sensitive speech component from the speech signal. A voice style transfer of the speech signal is generated. Speech processing is performed on the filtered speech signal and the voice style transfer of the speech signal.
Methods, computer systems, and computer-storage media are provided for facilitating implementation of high dynamic range display technology in a remote desktop environment. In embodiments, graphical data is obtained, at a video driver of a remote desktop server, for presenting in association with a remote desktop client device. The graphical data is provided to a high dynamic range (HDR) video encoder, logically separated from the video driver, based on an indication to use the HDR video encoder to encode. Thereafter, the graphical data is encoded via the HDR video encoder, and such encoded graphical data is provided to a remote access protocol stack in a terminal service of the remote desktop server for transmitting the encoded graphical data to the remote desktop client device associated with the HDR monitor.
G06F 9/451 - Execution arrangements for user interfaces
H04N 19/186 - Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a colour or a chrominance component
Systems and methods for dynamically modifying a cache associated with a neural network model of a natural language generator are described. In examples, a neural network model employs a beam search algorithm at a decoder when decoding output and generating predicted output candidates. The decoder utilizes caching techniques to improve a speed at which the neural network operations. When an amount of memory utilized by one or more caches of the neural network model is determined to exceed a threshold memory size, a layer-specific portion of a cache associated with a layer of the neural network model is identified. The identified layer-specific portion of the cache can be deleted when the amount of memory utilized by the cache of the neural network model exceeds the threshold memory size. In examples, data in the cache is deduplicated and/or deleted.
Methods and systems for processing digital audio are disclosed. A digital audio file is received and linear predictive coding (LPC) analysis is performed to decompose the digital audio file into a spectral envelope signal and an excitation signal. A spread spectrum watermark is generated using a first watermark key and the spectral envelope signal, and a self-correlated watermark is generated using a second watermark key and the excitation signal. The spread spectrum watermark and the self-correlated watermark are embedded into the digital audio file. LPC synthesis is performed to combine the spectral envelope signal with the embedded spread spectrum watermark and the excitation signal with the embedded self-correlated watermark to produce a watermarked digital audio file.
G10L 19/018 - Audio watermarking, i.e. embedding inaudible data in the audio signal
G10L 25/51 - Speech or voice analysis techniques not restricted to a single one of groups specially adapted for particular use for comparison or discrimination
41.
LATENT-CONTEXT ALERT CORRELATION ENGINE IN A SECURITY MANAGEMENT SYSTEM
Methods, systems, and computer storage media for providing security incident management using a latent-context alert correlation engine in a security management system. Security incident management is provided using the latent-context alert correlation engine that is operationally integrated into the security management system. In operation, first security data of a first alert and second security data of a second alert are accessed. The first alert and the second alert do not share a common entity identifiable in a security graph. Using the first security data and the second security data, a determination is made that the first alert is connected to the second alert based on a latent-context connection. The latent-context connection is a known attack path connection that indirectly connects alerts. Based on determining that the first alert is connected to the second alert, a security incident is generated for the alert. A notification comprising the security incident is communicated.
G06F 21/57 - Certifying or maintaining trusted computer platforms, e.g. secure boots or power-downs, version controls, system software checks, secure updates or assessing vulnerabilities
A multi-user virtual machine (VM) system is disclosed herein. Upon receiving a user login, first persistent storage and a first network interface card (NIC) are attached to a VM, where the VM operates as a personal computing device of a first user. Upon receiving a logout request for the user, the first persistent storage and the first NIC are detached from the VM. Thereafter, a second login is received, and second persistent storage and a second NIC are attached to the VM, where the VM operates as a personal computing device of a second user.
Aspects of the present disclosure relate to computing device headset input. In examples, sensor data from one or more sensors of a headset device are processed to identify implicit and/or explicit user input. A context may be determined for the user input, which may be used to process the identified input and generate an action that affects the behavior of a computing device accordingly. As a result, the headset device is usable to control one or more computing devices. As compared to other wearable devices, headset devices may be more prevalent and may therefore enable more convenient and more intuitive user input beyond merely providing audio output.
G06F 3/01 - Input arrangements or combined input and output arrangements for interaction between user and computer
H04Q 9/00 - Arrangements in telecontrol or telemetry systems for selectively calling a substation from a main station, in which substation desired apparatus is selected for applying a control signal thereto or for obtaining measured values therefrom
In a UDP firewall, a flow of packets is received from a public communications network, the flow of packets being sent into a private communications network. The firewall forwards a threshold amount of the flow of packets into the private communications network and validates the flow of packets in response to receiving a packet from the private communications network. In response to the validation failing the firewall blocks the flow of packets. In response to the validation succeeding the firewall allows the flow of packets to continue to be forwarded into the private communications network.
Systems and methods herein provide a disruptor validation engine and its related functions. In an aspect, a disruptor validation engine may determine disruption alerts issued by one or more disruptors, where a disruption alert indicates potential malicious activity within a tenant environment. Responsive to receiving the disruption alerts, the disruptor validation engine may generate a probability grade for a respective disruption alert. The probability grade may indicate a likelihood that the potential malicious activity triggering the disruption alert is actually malicious. The disruptor validation engine may then determine a disruption classification for the disruption alert based on a respective probability grade. Based on the disruption classification, the disruptor validation engine may generate a quality grade for the one or more disruptors indicating the validity and quality of the respective disruptor.
H04L 41/16 - Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
A generative artificial intelligence (AI) model application programming interface (API) receives a generative AI request and routes the generative AI request to a generative AI model. The generative AI model API identifies whether the generative AI request is an asynchronous or a synchronous request and identifies a likely length of generation requested by the generative AI request. The generative AI model API evaluates the available capacity of generative AI models in a shared pool of computing system resources based upon the generative AI model type requested by the generative AI request and based upon the length of the requested generation. The generative AI model API routes the generative AI request to a generative AI model based upon the available capacity, whether the generative AI request is synchronous or asynchronous, the generative AI model type, and the length of the requested generation.
The disclosed relates to a method and system for training language models using a reference-free multi-preference alignment approach. It addresses the challenge of aligning model outputs with human preferences by optimizing a contrastive loss function over selected responses. The method involves identifying candidate responses, calculating reward scores, and selecting a subset of responses for optimization. Deviations from mean reward scores are used to assign weights, and a weighted contrastive loss function is generated to update model parameters. This approach enhances model alignment by considering multiple preferences and mitigating biases toward brevity, resulting in richer and more detailed outputs.
An early stopping checking and verification process includes deriving a workload lower bound and a workload upper bound on what-if cost for an index tuning curve for a workload as what-if calls are being made. The index tuning curve represents a function that maps a number of what-if calls allocated to a percentage improvement of a best index configuration found so far. Early stopping checks are performed to identify a predetermined characteristic of the index tuning curve to determine if early stopping verification should be performed. A difference between the workload lower bound and the workload upper bound is compared to a threshold when the predetermined characteristic is detected. What-if calls for the workload are terminated based at least in part on the comparison between the difference between the workload lower bound and the workload upper bound and the predefined threshold.
Provided herein are systems, methods, machine-readable media, and the like, for bug mitigation in a software application. A method includes receiving a bug report associated with a bug, the bug report including a title, reproduction steps, and a description, generating, by a transformer model, respective embeddings of the title and the description resulting in a title embedding and a description embedding, generating a functional intent signature (FIS) of the bug based on the reproduction steps, clustering the bug based on the FIS, the title embedding and the description embedding, determining the bug belongs to a highest priority cluster, and mitigating the bug responsive to determining the bug belongs to a highest priority cluster resulting in an updated software application.
Techniques and related implementations for analyzing semantic sessions are provided. One implementation includes a computing system comprising: processing circuitry and memory storing instructions that, during execution, cause the processing circuitry to perform the steps of: receiving a plurality of screenshot images; for each of the plurality of screenshot images: identifying one or more image regions and one or more text regions on the screenshot image; and generating features for the screenshot image based on the one or more image regions and the one or more text regions using an artificial intelligence module; generating a user activity graph comprising nodes and edges, wherein the nodes correspond to the plurality of screenshot images, and wherein the edges correspond to commonalities between node pairs based on the generated features of the plurality of screenshot images; and identifying one or more node clusters from the user activity graph based on semantic relationships.
G06V 10/26 - Segmentation of patterns in the image fieldCutting or merging of image elements to establish the pattern region, e.g. clustering-based techniquesDetection of occlusion
G06V 10/40 - Extraction of image or video features
G06V 10/82 - Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
A computing system generates a trained super-resolution model using correlated high-resolution and low-resolution image pairs. A first image generator produces high-resolution images, which are processed by a downscaler network to create corresponding low-resolution images. The downscaler network is trained using an initial set of image pairs. Additional high-resolution images are generated and processed by the downscaler network to produce additional low-resolution images, forming a second set of image pairs. The super-resolution model is trained using the second set of image pairs. The trained super-resolution model processes low-resolution images to generate enhanced images with improved detail and reduced artifacts. The model can be used to increase resolution of input images or refine details without changing resolution. The trained model may be deployed on different devices and applied across various rendering environments.
A63F 13/52 - Controlling the output signals based on the game progress involving aspects of the displayed game scene
G06T 3/4046 - Scaling of whole images or parts thereof, e.g. expanding or contracting using neural networks
G06T 3/4053 - Scaling of whole images or parts thereof, e.g. expanding or contracting based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
G06T 5/60 - Image enhancement or restoration using machine learning, e.g. neural networks
An apparatus and method for mitigating gas line absorption of a signal transmitted via a hollow core fiber by a transmitter are described. The method comprises computing an adjustment to be made to a property of the signal, the adjustment being an adjustment to change one or more of a frequency, an amplitude, and a modulation format of the signal to reduce an impact of gas line absorption on the signal. The method further comprises configuring one or more of: the transmitter to cause the adjustment to be applied to the signal during transmission of the signal, via the hollow core fiber, by the transmitter; and a receiver configured to receive the signal transmitted via the hollow core fiber, to cause the adjustment to be applied to the signal after the receiver receives the signal.
According to implementations of the disclosure, a solution for performing network simulation with an FPGA device is provided. In this solution, at least one FPGA device is configured to configure a simulation network corresponding to a computing network based on configuration information of the computing network, the simulation network including a plurality of simulation endpoints and a plurality of simulation switches. The at least one FPGA device is configured to include a plurality of endpoint modules implementing the plurality of simulation endpoints in a multiplexing mode and a switch module implementing the plurality of simulation switches in a multiplexing mode. An endpoint module of the plurality of endpoint modules is configured to: execute an application configured in the endpoint module to process a first data packet. The switch module is configured to: transmit at least a second data packet to a further endpoint module of the plurality of endpoint modules.
H04L 41/16 - Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
G06F 30/27 - Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
54.
COORDINATING INTERACTIONS WITH HETEROGENOUS MACHINE LEARNING AGENTS
This document relates to facilitating interactions between a user and multiple agents, such as agents that provide different generative machine learning capabilities. In certain disclosed implementations, user input is received and two or more agents are selected to perform a particular task for the user. Then, interactions with the agents are coordinated on behalf of the user based on agent definitions of the selected agents. A result of the particular task is received and provided to the user. For instance, the user input can be received via a chat interface that allows the user to conduct a single chat session that involves the particular task being performed by multiple agents, thus giving the user the appearance of interacting with a single agent.
The present invention relates to a method for differential quantization of neural networks, such as large language models (LLMs), to improve operational efficiency without significant performance degradation. The method involves decomposing a neural network layer's matrix into multiple matrices using decomposition to identify significant components and insignificant components. The significant components undergo minimal or no quantization, while the less significant components undergo higher degrees of quantization. This differential quantization allows large neural networks to operate efficiently on edge devices with limited computing resources. The differential quantization technique enhances the deployment of neural networks in various hardware environments by reducing computational and memory bandwidth demands.
This disclosure describes a location-based response system that utilizes multimodal embedding and generative artificial intelligence (AI) models to generate location-based responses for user queries. For example, the location-based response system receives a real-time image, real-time location data, and a user query from a client device and determines a multimodal embedding in a multimodal vector embedding space based on the real-time image and the real-time location data. The multimodal embedding is then provided to a generative AI model, together with the data received from the client device, with instructions to generate a user query response. Upon receiving the user query response from the generative AI model, the system may provide it to the client device. Indeed, the location-based response system generates location-based responses for user queries by utilizing identified multimodal embedding and by instructing a generative AI model to respond to the user query based on the identified multimodal embedding.
The technology described herein automatically tunes large language model (LLM) prompts for improved performance. The system receives an LLM prompt as input, optimizes it, and outputs a more effective prompt. The optimization process involves converting the prompt into a hierarchical prompt tree, which organizes the prompt into distinct sections and subsections. This structured format allows for systematic refinements targeting specific components without altering the entire prompt. The system uses an actor/critic learning model, where the critic evaluates the prompts generated by the actor and provides feedback to guide future prompt generation decisions. The actor executes actions based on the feedback, continuously improving its performance. The iterative process generates replacement prompts, assigns quality scores, and selects high-quality prompts for further optimization.
Techniques and related implementations for analyzing semantic sessions are provided. One implementation includes a computing system (100) comprising: processing circuitry (102) and memory (104) storing instructions that, during execution, cause the processing circuitry (102) to perform the steps of: receiving a plurality of screenshot images (106); for each of the plurality of screenshot images (106): identifying one or more image regions (302) and one or more text regions (304) on the screenshot image (106); and generating features (110) for the screenshot image (106) based on the one or more image regions (302) and the one or more text regions (304) using an artificial intelligence module (112); generating a user activity graph (116) comprising nodes and edges, wherein the nodes correspond to the plurality of screenshot images (106), and wherein the edges correspond to commonalities between node pairs based on the generated features (110) of the plurality of screenshot images (106); and identifying one or more node clusters from the user activity graph (116) based on semantic relationships.
Disclosed are techniques for disabling security exploits. Some live code patches redirect execution from a vulnerable version of a function to a secure version of the function. However, the vulnerable version of the function may remain in memory, creating an opportunity for exploitation. Disclosed are techniques for determining when the vulnerable version of the function is no longer in use, and when this determination is made, rendering at least a vulnerable portion of the vulnerable version of the function inoperable. Reference counting, call stack inspection, and process lifecycle tracking may be used to determine when the vulnerable version of the function is no longer in use. Some implementations may employ compiler-assisted instrumentation, run-time generated instrumentation, or hardware support. The vulnerable version of the function may be rendered inoperable by overwriting it with illegal instructions, marking a page of memory that contains the vulnerable version of the function non-executing, etc.
G06F 21/57 - Certifying or maintaining trusted computer platforms, e.g. secure boots or power-downs, version controls, system software checks, secure updates or assessing vulnerabilities
Implementations for operating system state (OS) context management are provided. One example includes a computing system (100) comprising processing circuitry (102) and memory (104) storing instructions that, during execution, cause the processing circuitry (102) to perform the steps of: receiving a query (106) containing a request for information relating to a current OS state; passing the query (106) to an artificial intelligence (AI) module (108); and generating an answer (114) to the query (106) by: calling an OS state retrieval tool (110) to perform a plurality of application programming interface (API) calls to retrieve current OS state information; formatting the current OS state information into a data structure describing graphical window context that is readable by the AI module (108); and passing, to the AI module (108), the formatted current OS state information and a prompt containing instructions for answering queries using the formatted current OS state information.
Methods, apparatuses, and products for context-based connection policies for securing remote virtual device connections, including: receiving, from a client device, a request to establish a session with a virtual computing device implemented in a cloud computing environment, wherein the request comprises an authentication context received by the client device from an authentication service, wherein the authentication context describes a user account associated with the client device and an operational state of the client device; establishing the session between the client device and the virtual computing device; and controlling one or more redirection features of the session based on the authentication context.
Disclosed are techniques for disabling security exploits. Some live code patches redirect execution from a vulnerable version of a function to a secure version of the function. However, the vulnerable version of the function may remain in memory, creating an opportunity for exploitation. Disclosed are techniques for determining when the vulnerable version of the function is no longer in use, and when this determination is made, rendering at least a vulnerable portion of the vulnerable version of the function inoperable. Reference counting, call stack inspection, and process lifecycle tracking may be used to determine when the vulnerable version of the function is no longer in use. Some implementations may employ compiler-assisted instrumentation, run-time generated instrumentation, or hardware support. The vulnerable version of the function may be rendered inoperable by overwriting it with illegal instructions, marking a page of memory that contains the vulnerable version of the function non-executing, etc.
The disclosed techniques provide a method for reducing power consumption in a logic circuit. A design of the logic circuit is provided. The logic circuit includes storage circuits that store state information and receive clock signals. Toggle activity data indicating whether the stored state information changes in response to a respective clock signal is determined. A graph that represents the toggle activity data is constructed. The graph includes nodes and edges. The nodes correspond to the storage circuits. The edges represent statistical relationships among the determined changes in the stored state information. A graph neural network to the graph is applied. The application includes generating an embedding that characterizes patterns in the toggle activity data. Via a reinforcement learning agent, control logic that selectively gates the clock signals provided to storage circuits for which the stored state information is predicted, based on the embedding, to remain unchanged is determined.
G06F 30/398 - Design verification or optimisation, e.g. using design rule check [DRC], layout versus schematics [LVS] or finite element methods [FEM]
G06F 30/27 - Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
A computer-implemented method of generating a mixed reality workflow is described. The method comprises identifying a series of tasks and generating an input task-to-object-mapping by analyzing data that defines a process performed by a first user interacting with objects in a first location. The input task-to-object-mapping that maps each task from the series of tasks to an object used in the respective task. A task-specific non-spatial characteristic of each object in the input task-to-object-mapping is determined and used to map each object in the input task-to-object-mapping to a candidate object identified at a second location to generate an output task-to-object-mapping. The series of tasks, location data defining a position of each candidate object in the second location and output task-to-object-mapping are used to generate a mapped workflow which is then output to a device in the second location.
A computing system for monitoring language model compliance with a rubric of one or more output characteristics. The computing system includes processing circuitry configured to interface with a trained generative language model that receives input of a prompt including natural language text input and, in response, generates an output that includes natural language text output. The processing circuitry is further configured to monitor compliance of the generative language model with the rubric, by feeding the output of the generative language model to a rubric classifier configured to generate a predicted classification for an output characteristic in the rubric, and output the predicted classification.
The present application relates to a network, apparatus, and method for resource allocation in a distributed wide area network including a plurality of computing resources configured to instantiate virtual machines. A workload manager for the network predicts, using one or more machine-learning algorithms, based on network utilization and eviction information, a set of stable interruptible virtual machine (IVM) resources having an IVM capacity. The network manager receives a plurality of requests for interruptible workloads. The network manager allocates the interruptible workloads to the stable IVM resources up to the IVM capacity. The network manager may predict a rate of requests for priority uninterruptible workloads and a capacity during a future time period, determine to protect at least one of the interruptible workloads based on the rate of requests and the capacity, and guarantee resources to the at least one interruptible workload as a protected workload for the future time period.
A microfluidic printed circuit board (PCB) including: a plurality of layers; a heat-generating component attached to the microfluidic PCB; and a network of microfluidic passages applied to one or more of the plurality of layers and in thermally conductive contact with the heat-generating component, the network of microfluidic passages electrically isolated from traces of one or more of the plurality of layer
A method performed by a processor of a computing system is described herein, where the method includes obtaining an image that includes an object having a shape, where a boundary of the shape of the object in the digital image is labeled in the digital image. The method also includes computing an encoding for the shape, where computing the encoding for the shape includes partitioning the shape into multiple partitions. Computing the encoding for the shape further includes, for the multiple partitions, computing angle-based contour descriptors that represent boundaries of the partitions, where the encoding for the shape of the object is based upon the angle-based contour descriptors.
The disclosed concepts relate to tracking and representing help sessions for video games where video game players are assisted by a helper, e.g., another video game player and/or a trained machine learning model. For instance, the disclosed implementations can graphically modify a controllable entity, such as a character or vehicle, to convey that the current video game player is being assisted by a helper. As another example, the disclosed implementations can graphically modify game achievements to indicate when a given achievement was earned with assistance from a helper.
A63F 13/5375 - Controlling the output signals based on the game progress involving additional visual information provided to the game scene, e.g. by overlay to simulate a head-up display [HUD] or displaying a laser sight in a shooting game using indicators, e.g. showing the condition of a game character on screen for graphically or textually suggesting an action, e.g. by displaying an arrow indicating a turn in a driving game
A63F 13/493 - Resuming a game, e.g. after pausing, malfunction or power failure
A63F 13/69 - Generating or modifying game content before or while executing the game program, e.g. authoring tools specially adapted for game development or game-integrated level editor by enabling or updating specific game elements, e.g. unlocking hidden features, items, levels or versions
A63F 13/79 - Game security or game management aspects involving player-related data, e.g. identities, accounts, preferences or play histories
A63F 13/86 - Watching games played by other players
A63F 13/87 - Communicating with other players during game play, e.g. by e-mail or chat
70.
COMPOSITE IDENTITY REPRESENTATION FOR PROVIDING USER INTERFACES OF NETWORK-BASED SERVICES
A trigger to generate a user interface element representing a presence status of a first user is identified, where the first user has a first account and a second account associated with the service. Based upon an identifier of the first user, a composite identity data structure is identified that identifies both the first account and the second account. In response to the data structure identifying the first user associating with both accounts, selection rules are applied to determine a unified presence status based on an activity associated with the first account or the second account. The user interface element is provided to a device of a viewing user, presenting the unified presence status as a single representation across the first account and the second account.
Methods, systems, and computer storage media for providing a tailored meeting-video segment associated with a meeting-video management engine of a meeting-video management system. The tailored meeting-video segment corresponds to a portion of meeting-video content that is programmatically generated based on features associated with video data, meeting data, and user data. A tailored meeting-video segment—or a plurality of tailored meeting-video segments—can be generated by employing a meeting-video tailoring machine learning model of the meeting-video management engine. In particular, the features—associated with video data comprising the plurality of clips, meeting data of the meeting, and user data of the user—are meeting-video tailoring features used by the meeting-video tailoring machine learning model to generate the tailored meeting-video segment. The tailored meeting-video segment is communicated to a user to enable uniquely tailored presentation and playback of meeting-video content computed to be relevant to the user via the meeting-video management engine.
G06V 20/40 - ScenesScene-specific elements in video content
G10L 25/57 - Speech or voice analysis techniques not restricted to a single one of groups specially adapted for particular use for comparison or discrimination for processing of video signals
H04L 65/403 - Arrangements for multi-party communication, e.g. for conferences
72.
INTERACTIVE CONCEPT EDITING IN COMPUTER-HUMAN INTERACTIVE LEARNING
A collection of data that is extremely large can be difficult to search and/or analyze. Relevance may be dramatically improved by automatically classifying queries and web pages in useful categories, and using these classification scores as relevance features. A thorough approach may require building a large number of classifiers, corresponding to the various types of information, activities, and products. Creation of classifiers and schematizers is provided on large data sets. Exercising the classifiers and schematizers on hundreds of millions of items may expose value that is inherent to the data by adding usable meta-data. Some aspects include active labeling exploration, automatic regularization and cold start, scaling with the number of items and the number of classifiers, active featuring, and segmentation and schematization.
The technology described herein automatically tunes large language model (LLM) prompts for improved performance. The system receives an LLM prompt as input, optimizes it, and outputs a more effective prompt. The optimization process involves converting the prompt into a hierarchical prompt tree, which organizes the prompt into distinct sections and subsections. This structured format allows for systematic refinements targeting specific components without altering the entire prompt. The system uses an actor/critic learning model, where the critic evaluates the prompts generated by the actor and provides feedback to guide future prompt generation decisions. The actor executes actions based on the feedback, continuously improving its performance. The iterative process generates replacement prompts, assigns quality scores, and selects high-quality prompts for further optimization.
Provided herein are systems, methods, machine-readable media, and the like, for bug mitigation in a software application. A method includes accessing bug data of a bug of the software application that is not resolved, performing temporal decay analysis (TDS) on the bug that assesses developer activity on and lifecycle of the bug over time to identify that the bug has become stale, responsive to identifying the bug has become stale, flagging the bug as stale, performing (i) evolutionary relevance mapping on the bug to determine the bug is still relevant in a current instance of the software application or (ii) cross-version similarity modeling to identify the bug has recurred in the current instance of the software application, and mitigating the bug responsive to determining the bug is still relevant or has recurred in the current instance of the software application resulting in an updated software application.
Some embodiments get accessibility information of a web page, and then filter it to include specified context data and to exclude specified non-context data. A version of the context data is provided to a language model in a prompt, or along with a prompt, to influence computation of a response of the language model. In some cases, the provided version includes a markdown language translation of accessibility data extracted from the web page. Some embodiments run a browser script before extracting the accessibility data. In some, a portion of the markdown is chosen, with the choice being based on vector similarity or on association with a link. In some embodiments, links are consolidated.
This disclosure describes utilizing a suggestion item system to quickly, accurately, and efficiently generate suggestion items using generative artificial intelligence (AI) models and multiple embedding groups. For example, in response to detecting user input, which is often associated with search queries, the suggestion item system provides the user input to multiple generative AI models or model instances that utilize parallel decoding and distinct embedding groups to generate suggestion items. By using parallel decoding from separate embedding groups, the suggestion item system can quickly provide high-quality, accurate, and unique suggestion items in response to user inputs.
Implementations for operating system state (OS) context management are provided. One example includes a computing system comprising processing circuitry and memory storing instructions that, during execution, cause the processing circuitry to perform the steps of: receiving a query containing a request for information relating to a current OS state; passing the query to an artificial intelligence (AI) module; and generating an answer to the query by: calling an OS state retrieval tool to perform a plurality of application programming interface (API) calls to retrieve current OS state information; formatting the current OS state information into a data structure describing graphical window context that is readable by the AI module; and passing, to the AI module, the formatted current OS state information and a prompt containing instructions for answering queries using the formatted current OS state information.
G06F 16/583 - Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
78.
REDUCING POWER CONSUMPTION IN A LOGIC CIRCUIT AND RELATED METHODS
An exemplary method of inserting clock gate circuits in a logic circuit design includes grouping storage circuits in the logic circuit design into a first plurality of groups and also grouping the storage circuits into a second plurality of groups, where the second plurality of groups has fewer groups than the first plurality of groups. The method includes inserting clock gate circuits into the logic circuit design based on the first plurality of groups to generate a first updated circuit design and based on the second plurality of groups to generate a second updated circuit design. The method also includes selecting one of the first updated circuit design and the second updated circuit design based on a comparison of test results of the first updated circuit design and the second updated circuit design. The storage circuits may be grouped based on storage circuits that toggle in a sequence.
Systems and methods for generating synthetic data. One such method includes receiving, at a security unit, adaptive controls comprising both privacy budget constraints and compliance requirements for determining risk scores for sensitive information. Th method further includes receiving, at the security unit, an input dataset comprising a plurality of data entries including sensitive information and determining a risk score in accordance with the compliance requirements for each of the plurality of data entries. The method further includes dynamically allocating, for each of the plurality of data entries, a privacy budget based on the risk score for the data entry and the privacy budget constraints. The method further includes performing adaptive remediation on the input dataset to generate a synthetic dataset, wherein adaptive remediation comprises adaptive aggregation and/or adaptive differential noise injection.
An apparatus and method for mitigating gas line absorption of a signal transmitted via a hollow core fiber by a transmitter are described. The method comprises initiating generating a signal using a laser and initiating modulating the signal to produce a carved signal, wherein the carved signal comprises a plurality of channels. A spectrum of the carved signal comprises substantially zero signal intensity at a frequency within a gas line absorption profile, the frequency being between a first channel and a second channel of the carved signal. The method further comprises initiating transmission of the carved signal via the hollow core fiber by the transmitter.
Methods, systems, and computer storage media for programmatically retrieving data from a distributed data source by receiving a query to obtain data from the distributed data source that comprises one or more data sources and determining a completeness criteria corresponding to each data source of the distributed data source based on the query. The data is retrieved by generating a query plan comprising a plurality of sub-plans to satisfy the query based on the completeness criteria, generating a plurality of sub-queries corresponding to the sub-plans, submitting the plurality of sub-queries to the distributed data source, and providing a response to the query based on responses to the plurality of sub-queries.
Some embodiments get accessibility information of a web page, and then filter it to include specified context data and to exclude specified non-context data. A version of the context data is provided to a language model in a prompt, or along with a prompt, to influence computation of a response of the language model. In some cases, the provided version includes a markdown language translation of accessibility data extracted from the web page. Some embodiments run a browser script before extracting the accessibility data. In some, a portion of the markdown is chosen, with the choice being based on vector similarity or on association with a link. In some embodiments, links are consolidated.
Methods, apparatuses, and products for indexing semi-structured data, including: generating, for a base table comprising a semi-structured data field, an inverted index for the semi-structured data field, wherein the inverted index comprises a term and a posting, and wherein generating the inverted index comprises: storing, for each combination of a plurality of unique path and value combinations of the semi-structured data field, an entry in the inverted index comprising: a path for a corresponding combination in a first field of the term, a value for the corresponding combination in a second field of the term, and an indication of each record in the base table including the corresponding combination in the posting; and storing, for one or more paths of the plurality of unique path and value combinations comprising an array index, the array index in a third field the term.
The technology described herein relates to a method and system for generating a consolidated event explanation by combining content from multiple sources describing a single event. The system utilizes a natural language processing (NLP) architecture to analyze a diverse collection of materials, including online news articles, reference content, and social media content. The consolidated event explanation is structured into three layers: an event explanation summary, a main event explanation, and access to full articles. The event explanation summary includes a headline, key points, and an image, designed for display on news sites or applications. The main event explanation is divided into interest sections, each focusing on different aspects of the event, and may include quotations or summaries generated by a large language model (LLM). The system ensures comprehensive coverage by identifying and expanding on various reader interests, thereby providing a deep and multi-perspective understanding of the event.
Technology is disclosed herein for the integration of large language model (LLM) services for content analysis in productivity applications. In an implementation, an application identifies content associated with the application executing on the computing apparatus. The application identifies a context relating to the content, wherein the context comprises contextual information by which to evaluate the content. The application generates a prompt for an LLM service which includes the content and the context. The prompt requests an evaluation of the content to recommend supplemental content and submits the prompt to the LLM service. The application receives a response to the prompt from the LLM service which includes a suggestion for supplemental content based on the evaluation and displays, in a user interface of the application, a recommendation based on the suggestion from the LLM service.
The present disclosure provides systems and methods for detecting confidential information in input data. A subject entity and a first feature associated with the subject entity are extracted from input data. An indication of the subject entity is submitted to a machine learning feature extractor trained on a public dataset and entity-feature associations expressed in the public dataset. a second feature of the subject entity extracted from the public dataset is received from the ML feature extractor and compared with the first feature. It is determined, based on the comparison, that the first feature is not associated with the subject entity in the public dataset and in response, a data security mitigation action is performed.
G06F 21/62 - Protecting access to data via a platform, e.g. using keys or access control rules
G06F 21/57 - Certifying or maintaining trusted computer platforms, e.g. secure boots or power-downs, version controls, system software checks, secure updates or assessing vulnerabilities
Systems and methods are provided for implementing improved memory defect management. Bad pages, referring to portions of memory affected by memory errors, are identified by an operating system (“OS”) of a computing system. A list of bad pages may be periodically uploaded to a cloud data storage. The OS may also send the list of bad pages to a basic input/output system (“BIOS”) for repair of some of the bad pages. A list of repaired bad pages or corresponding status information may be sent to the OS. In some examples, the list of bad pages is compared with the list of repaired bad pages and may be updated accordingly. The list of bad pages may be compared with a previously uploaded list of bad pages that has been received from the cloud data storage. One list may be updated with, or replaced by, the other list based on the comparison.
Processor-based systems employing configurable local frequency throttling management to manage power demand and consumption, and related methods. For example, such processor-based systems may include a processor and other power circuitry to control power to the processor. The processor includes a clock control circuit that is configured generate a clock signal(s) at a designated frequency to clock a processor core(s) in the processor at a desired operating frequency(ies). The clock control circuit is configured to dynamically throttle (i.e., limit and/or reduce) the frequency(ies) of a clock signal(s) clocking the processor in response to a frequency throttle event that may be an unexpected event. Reducing power demand may be important to ensure that the processor can continue to operate under interrupted or reduced power supply conditions. It may be faster to throttle the operating frequency of a processor than to throttle the operating voltage of power supplied to the processor.
Dynamically adjusting memory allocations based on Quality-of-Service (QoS) pool threshold values in processor-based devices is disclosed herein. In one exemplary embodiment, a processor-based device provides a storage controller circuit communicatively coupled to a memory device. The storage controller circuit comprises a dedicated storage device and a plurality of pool threshold values corresponding to a plurality of pools, wherein each pool corresponds to one or more QoS levels, and each pool threshold value defines an allocation of the dedicated storage device for the corresponding pool. For each pool of the plurality of pools, the storage controller circuit collects a usage statistic during a sampling period, and determines, based on the usage statistic, whether to modify the pool threshold value corresponding to the pool. The storage controller circuit then modifies the pool threshold value corresponding to the pool based on the usage statistic, responsive to determining to modify the pool threshold value.
Methods, apparatuses, and products for context-based connection policies for securing remote virtual device connections, including: receiving, from a client device, a request to establish a session with a virtual computing device implemented in a cloud computing environment, wherein the request comprises an authentication context received by the client device from an authentication service, wherein the authentication context describes a user account associated with the client device and an operational state of the client device; establishing the session between the client device and the virtual computing device; and controlling one or more redirection features of the session based on the authentication context.
Innovations in adaptive encoding for units of a video sequence can improve coding efficiency. For example, some of the innovations relate to encoding that includes adaptive switching of color spaces between units within a video sequence. Other innovations relate encoding that includes adaptive switching of color sampling rates between units within a video sequence. Still other innovations relate encoding that includes adaptive switching of bit depths between units within a video sequence.
H04N 19/59 - Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial sub-sampling or interpolation, e.g. alteration of picture size or resolution
H04N 19/117 - Filters, e.g. for pre-processing or post-processing
H04N 19/132 - Sampling, masking or truncation of coding units, e.g. adaptive resampling, frame skipping, frame interpolation or high-frequency transform coefficient masking
H04N 19/147 - Data rate or code amount at the encoder output according to rate distortion criteria
H04N 19/176 - Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a block, e.g. a macroblock
H04N 19/182 - Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a pixel
H04N 19/186 - Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a colour or a chrominance component
H04N 19/51 - Motion estimation or motion compensation
H04N 19/587 - Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal sub-sampling or interpolation, e.g. decimation or subsequent interpolation of pictures in a video sequence
H04N 19/593 - Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial prediction techniques
Described herein is a technique for generating data for a computer system. An apparatus comprises a high-speed communication bus, a memory unit communicatively coupled to the high-speed communication bus, an integrated circuit communicatively coupled to the high-speed communication bus, the integrated circuit to execute a machine learned model trained to receive input data and generate a set of implicit keywords from the input data to support a network service, and an implicit facet mapper to map the input data to an implicit facet of a knowledge graph, a database communicatively coupled to the high-speed communication bus, the database to store the set of implicit keywords and implicit facet in a data structure, and a network interface communicatively coupled to the high-speed communication bus to access a wireless network. Other embodiments are described and claimed.
A computer-implemented method, computer program product and computing system for: enabling a generative AI system to effectuate an analysis protocol; enabling a user to utilize the generative AI system during an interactive session concerning the analysis protocol; receiving content during the interactive session, thus defining received content; and determining whether the received content is relevant with respect to the analysis protocol.
In various examples, an image search is performed based on a query image where the result includes images with semantic information that matches the query image. For example, image encoders are used to extract information from images that are stored in an index. Continuing this example, when a query is obtained the index is searched to determine potential matches, and a comparison is performed between the query image and the potential matches to determine images with semantic information matching the query image.
The techniques disclosed herein enable systems to efficiently allocate computing resources for various computing workloads through a shared peak resource usage prediction. To achieve this, a predictive model analyzes a historical dataset defining resource usage of a computing environment for a past timeframe, and calculates a peak environment resource usage for a future or current timeframe. In addition, the predictive model estimates a peak number of computing workloads for the computing environment. Using the peak resource usage and/or the peak number of computing workloads, the system derives resource requests for allocating computing resources to a plurality of computing workloads. The computing workloads are subsequently assigned to computing nodes within the computing environment for execution. Furthermore, computing workloads within a computing node are configured to share computing resources to accommodate sudden surges in demand. In this way, the system can reduce computing resource wastage for computing environments.
A data processing system implements receiving a far-end signal associated with a first computing device participating in an online communication session and receiving a near-end signal associated with a second computing device participating in the online communication session. The near-end signal includes speech of a target speaker, a first interfering speaker, and an echo signal. The system further implements providing the far-end signal, the near-end signal, and an indication of the target speaker as an input to a machine learning model. The machine learning model trained to analyze the far-end signal and the near-end signal to perform personalized noise suppression (PNS) to remove speech from one or more interfering speakers and acoustic echo cancellation (AEC) to remove echoes. The model is trained to output an audio signal comprising speech of the target speaker. The system obtains the audio signal comprising the speech of the target speaker from the model.
G06N 3/0442 - Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
G10L 17/02 - Preprocessing operations, e.g. segment selectionPattern representation or modelling, e.g. based on linear discriminant analysis [LDA] or principal componentsFeature selection or extraction
G10L 17/04 - Training, enrolment or model building
G10L 17/06 - Decision making techniquesPattern matching strategies
A taxonomy-agnostic data protection framework includes a plugged-in data classification taxonomy definition having a taxonomy identifier and a set of data classification indicators. The taxonomy-agnostic data protection framework also includes a plugged-in set of data classification processing routines, and a plugged-in mapping mechanism which maps between data classification processing routines and data classification indicators. The framework facilitates efficient, accurate, and thorough implementation of data classification propagation per the plugged-in taxonomy, both within a given program and between programs that connect over a network. The framework also facilitates flexible implementation of per-taxonomy data protection actions such as deletion, redaction, encryption, anonymization, pseudonymization, hashing, or enrichment, in response to individual or combined data classification indicators. Static analysis of annotated source code determines whether data classifications are accurately and comprehensively propagated with a program.
Techniques are described herein that are capable of performing a security action based on anomaly detection using AI model profiles and user profiles. AI model profiles (e.g., a model-session profile, a model-response profile, and/or a model-response profile) associated with AI model(s) are generated. User profiles (e.g., user-session profiles, user-prompt profiles, and/or user-response profiles) associated with users of the AI model(s) are generated. A security action is performed with regard to an incoming AI prompt as a result of a difference between the incoming AI prompt and one or more of the AI model profiles and/or one or more of the user profiles being greater than or equal to a difference threshold.
G06F 21/54 - Monitoring users, programs or devices to maintain the integrity of platforms, e.g. of processors, firmware or operating systems during program execution, e.g. stack integrity, buffer overflow or preventing unwanted data erasure by adding security routines or objects to programs
