Alerta

COACHING OF ARTIFICIAL INTELLIGENCE AGENTS

Absstract of: US20260220543A1

A computer-implemented method for improving an operation of a first adaptive machine-learning model of an artificial intelligence agent by applying a second adaptive machine-learning model with at least one dataset comprising at least one sample is provided, the method comprises: inputting at least one hint as a part of at least one sample in the at least one dataset to the second adaptive machine-learning model; generating a target to an output of the first adaptive machine-learning model of the artificial intelligence agent by applying the at least one hint in an execution of the task with the second adaptive machine-learning model; and training the first adaptive machine-learning model by applying the target generated by the second adaptive machine-learning model. A computing system and a computer program are further provided.

SPECTRAL CLUSTERING FOR SPECTRAL TREE GUIDANCE IN HETEROGENEOUS MODEL TREE GENERATION OR OTHER FUNCTIONS FOR MACHINE LEARNING WITH DISCONTINUOUS DATASETS

Absstract of: US20260220490A1

0000 A method includes obtaining at least one dataset containing one or more discontinuities, where the one or more discontinuities split data of the at least one dataset into multiple partitions. The method also includes performing spectral clustering of the at least one dataset to identify multiple initial clusters of data in the at least one dataset. The method further includes trimming the initial clusters of data in order to identify an estimated number of partitions in the at least one dataset. The method also includes performing spectral clustering of the at least one dataset based on the estimated number of partitions to identify multiple updated clusters of data in the at least one dataset, where each updated cluster of data corresponds to one of the multiple partitions. In addition, the method includes providing the updated clusters of data as input to a machine learning algorithm.

MACHINE LEARNING MODEL EXPLANATION GENERATION

Absstract of: US20260220423A1

In some examples, a system obtains an approximation function for a machine learning (ML) model, the approximation function providing an approximate representation of a latent representation of latent features in a latent space produced by the ML model. The system presents a visualization of the latent features using the approximation function, and the system generates a refined latent space based on user input in the visualization. The system generates an explanation regarding which latent features of the latent space are primary contributors to a decision of the ML model.

SYSTEMS AND METHODS FOR SMART SETTING CONFIGURATION

Absstract of: US20260219896A1

Systems and methods are disclosed for smart setting configurations. Example methods may include ingesting a plurality of items of data, wherein each item of data is retrieved from a data source within a plurality of data sources associated with a user account on a social media platform, for each item of data, providing the item of data as input to a distinct machine learning algorithm in a plurality of contextual machine learning algorithms, and providing output of the plurality of contextual machine learning algorithms to an additional machine learning algorithm as input. Some methods may include formulating a setting configuration for the user account on the social media platform based at least in part on output of the additional machine learning algorithm, and generating a recommendation comprising the setting configuration to a user that operates the user account.

METHODS AND APPARATUS FOR PERSONALIZED GENERATIVE ARTIFICIAL INTELLIGENCE (GENAI) ASSISTANT

Absstract of: WO2026157015A1

An apparatus includes interface circuitry, machine-readable instructions, and at least one processor circuit to be programmed by the machine-readable instructions to obtain a first output from a natural language processing machine learning model engine, the first output generated with retrieval augmented generation (RAG) to respond to a user query, obtain a first feedback answer input associated with the user query, generate an embedding of the first feedback answer input and the user query, cause the embedding to be stored in a vector database to provide feedback for the RAG, and obtain a second output from the natural language processing machine learning model engine, the second output generated in response to the user query, the second output to include the first feedback answer input.

ESTIMATING LOCAL FEATURE IMPORTANCE

Absstract of: US20260220540A1

Here is generation of a local explanation of a machine learning (ML) inference, and this generation is accelerated by accurate estimation of local feature value importances from a multioutput regression by an attribution metamodel. Into a feature vector, a computer stores values for respective features and stores an inference, from the feature values, by an ML model. A multioutput regression by an attribution metamodel includes inferentially generating, from the feature vector, local value importances respectively for the feature values. Based on the local value importances, a local explanation of the inference is generated and displayed. For accelerated training with early stopping, an adaptively selected training corpus involves incrementally adding additional datapoints to the training corpus until the attribution metamodel accurately learns.

SYSTEMS, METHODS, AND APPARATUS FOR AUTOTUNING OF RETRIEVAL AUGMENTED GENERATION PARAMETERS

Absstract of: WO2026157014A1

Systems, apparatus, and methods for autotuning of retrieval augmented generation parameters are disclosed. Example instructions to perform the same include instructions to vectorize a dataset using retrieval augmented generation (RAG) parameters, the RAG parameters identified based on a domain of the dataset, wherein the RAG parameters control at least one of ingestion of data or retrieval of data, execute a machine learning model using the RAG parameters, evaluate a quality of an output of the machine learning model, after determination that the quality does not meet a quality threshold, update the RAG parameters and cause re-execution of the machine learning model using the updated RAG parameters, and after determination that the quality meets the quality threshold, store the RAG parameters in association with the domain of the dataset.

FOUNDATION MODEL DOMAIN AND TASK-SPECIFIC FINE-TUNING

Absstract of: WO2026161237A1

A method includes loading a machine learning model having a set of parameters and trained using initial training data. A new set of training data may be used to determine a set of importance values for the set of petameters indicating a measure of importance for a given parameter to affect an accuracy of an output of the machine learning model. A first plurality of task-specific importance values can be loaded for a particular task using a respective task-specific training set. A second plurality of domain-specific importance values can be loaded for a particular domain using a respective domain-specific training set. Combined importance values can be determined for the set of parameters. The machine learning model can then be further trained using the combined importance values to determine fine-tuned values for the set of parameters.

Agentic approach to multilingual text-to-SQL conversion with intent detection and self-correction

Absstract of: US20260220129A1

0000 An intent associated with a user input is determined. An augmented prompt for generating a Standard Query Language (SQL) code snippet is generated. An SQL code snippet is generated by using the augmented prompt on a machine learning (ML) model. Results are generated by executing the SQL code snippet on a data lake.

REFINED PROMPT - RESPONSE PAIR GENERATION FOR RESPONSE QUALITY MANAGEMENT

Absstract of: US20260220550A1

0000 Methods and systems for managing operation of a request processing pipeline are provided. To service requests, the request processing pipeline may accept unstructured text based requests. The unstructured text may be used to obtain prompts for evaluation by a trained generative machine learning model. The pairs may be evaluated to rank order them with respect to one another. A best ranked one of the pair may be used to service the request. The request may be serviced by providing the response of the best ranked one of the pairs, by initiating provisioning of computer implemented services using the response, and/or via other processes.

GENERATING INTERFACE FOR COMMUNICATION ELEMENTS

Absstract of: US20260219899A1

Example implementations relate to communication element selection in a network environment. In an example, a plurality of user features is received and input data derived from the plurality of user features is processed using a window logic and provided to an attention-based machine learning model. A plurality of communication elements is provided to the attention-based machine learning model, which assigns one or more weights to each of the plurality of communication elements for a subsequent time period based on the plurality of user features obtained from a preceding time period. A score for each of the plurality of communication elements based on the one or more weights for each of the plurality of communication elements is calculated and an interface including a communication element of the plurality of communication elements having a highest calculated score is generated.

SYSTEM AND METHOD OF INVERSE IDENTIFICATION OF CAUSAL ASSETS AND PROCESSES

Absstract of: US20260220718A1

0000 Various embodiments described herein relates to inverse identification of one or more causal assets and/or processes corresponding to a plurality of products in a flexible manufacturing environment. In this regard, routing data and quality notification data associated with the plurality of products is received. Further, the routing data and the quality notification data is correlated using the supervised machine learning (ML) model. As a result, at least one causal asset and/or at least one causal process is identified based on the correlation of the routing data and the quality notification data. Further, real-time recommendations including one or more corrective actions associated with the at least one causal asset and/or the at least one causal process are generated using the supervised ML model.

EFFICIENT AUTOML DURING DATA DRIFTS VIA CONFIGURATION ESTIMATION

Absstract of: US20260220527A1

0000 For a new batch of datapoints during a detected data drift, without decreasing accuracy nor increasing volatility of accuracy, noniterative selection of a new hyperparameters configuration accelerates generating a retrained machine learning model. For each batch in a sequence of recent batches of datapoints that includes a last batch of datapoints, a computer stores: the batch, a hyperparameter value for an unsupervised model, and an accuracy measurement, for the unsupervised model, that is based on both of the hyperparameter value and the batch. With a retraining corpus that contains the new batch and the batches in the sequence of recent batches, a new hyperparameter value is predicted. Based on the retraining corpus and the new hyperparameter value, a new accuracy measurement is estimated. In response to detecting that the new accuracy measurement exceeds an accuracy threshold, retraining the unsupervised model with the retraining corpus is unconventionally accelerated.

CO-OPTIMIZING EVALUATIONS AND SELECTION OF MACHINE LEARNING MODELS

Absstract of: US20260220421A1

Systems and methods are provided for co-optimized recommendation and evaluation of models, such as foundation models. Users may specify, in a natural language query, a request for (or requirements) regarding one or more models for application to a given use case. The query can be characterized along with models of a model repository to determine the suitability of models to satisfy the query. Benchmarks of a benchmark repository may also be analyzed to determine their suitability to be used to properly test any recommended models. The recommended models can then be tested on the recommended benchmarks. The characterization and analysis of models and benchmarks can reduce the compute and time costs associated with manual/conventional model and benchmark selection for use in a downstream application.

INFORMATION PROCESSING DEVICE, INFORMATION PROCESSING METHOD, AND PROGRAM

Absstract of: EP4783079A1

An information processing device includes: a machine learning unit (12) to learn a relationship between an evaluation value and a parameter on a basis of a search point of the parameter and an evaluation value of the search point, and predict the evaluation value for a search candidate point of the parameter; and a search progress acquiring unit (13) to acquire progress information indicating progress of a search on a basis of the search point, the evaluation value of the search point, the search candidate point, and the evaluation value of the search candidate point predicted by the machine learning unit (12).

CONTROL DEVICE FOR RADIO ACCESS NETWORK AND COMPUTER-READABLE STORAGE MEDIUM

Absstract of: EP4783645A1

0001 A control apparatus for a radio access network (RAN), includes: collection means configured to control the RAN based on a learning model and to collect first learning data from the RAN; determination means configured to determine usefulness of the first learning data in machine learning for the learning model; and processing means configured to perform processing to select second learning data from the first learning data for transmission to another control apparatus that performs the machine learning, based on the usefulness of the first learning data.

EVALUATING ON-DEVICE MACHINE LEARNING MODEL(S) BASED ON PERFORMANCE MEASURES OF CLIENT DEVICE(S) AND/OR THE ON-DEVICE MACHINE LEARNING MODEL(S)

Absstract of: EP4783078A2

Implementations disclosed herein are directed to systems and methods for evaluating on-device machine learning (ML) model(s) based on performance measure(s) of client device(s) and/or the on-device ML model(s). The client device(s) can include on-device memory that stores the on-device ML model(s) and a plurality of testing instances for the on-device ML model(s). When certain condition(s) are satisfied, the client device(s) can process, using the on-device ML model(s), the plurality of testing instances to generate the performance measure(s). The performance measure(s) can include, for example, latency measure(s), memory consumption measure(s), CPU usage measure(s), ML model measure(s) (e.g., precision and/or recall), and/or other measures. In some implementations, the on-device ML model(s) can be activated (or kept active) for use locally at the client device(s) based on the performance measure(s). In other implementations, the on-device ML model(s) can be sparsified based on the performance measure(s).

METHOD, APPARATUS, DEVICE, AND MEDIUM FOR MANAGING MACHINE LEARNING MODEL

Absstract of: US20260212281A1

A method, a device, and a medium for managing a machine learning model are provided. A prediction of a submission event between an object and a media item provided in the application is determined using a first machine learning model, the prediction of the submission event representing a probability that the object submits a response for a question associated with the media item. Based on the prediction of the submission event, a correction weight associated with the object is determined. A reference media item and a reference question associated with the reference media item are provided to the object in the application. In response to receiving a reference response submitted by the object for the reference question, the second machine learning model is updated based on the reference response and the correction weight, the second machine learning model describing an association relationship between the object and the reference media item.

FORECASTING MASTER DATA IN A SKEWED DATA SET USING HYBRID MACHINE LEARNING MODELS

Absstract of: US20260212254A1

Systems and methods described herein relate to hybrid machine learning techniques for forecasting master data, such as scrap data. A hybrid machine learning model includes a first machine learning model (e.g., balanced random forest classifier) that generates a prediction with respect to whether or not a manufacturing process will result in scrap generation. If the first machine learning model predicts that no scrap will be generated in the manufacturing process, then an output prediction is that no scrap will be generated in the manufacturing process. If the first machine learning model predicts that scrap will be generated in the manufacturing process, then a second machine learning model (e.g., gradient boost regressor) predicts a value of scrap percentage. The value of scrap percentage is provided as an output prediction of a percentage of scrap that will be generated in the manufacturing process.

HYBRID LINEAR MACHINE LEARNING MODEL

Absstract of: US20260212264A1

0000 A hybrid machine learning model architecture that enhances the retrieval capabilities of recurrent models without compromising efficiency. A linear recurrent architecture is supplemented by a parallel retrieval branch that integrates retrieved knowledge into the model with relatively minimal computational overhead. The parallel retrieval branch operates alongside the primary recurrent processing pipeline. The parallel retrieval branch chunks the input sequence and embeds the chunks into one or more pool. Chunk-by-chunk causal retrieval produces results that are then injected into the primary linear pipeline using within-chunk cross-attention.

COMPUTER IMPLEMENTED METHOD OF TRAINING A MACHINE LEARNING MODEL CONFIGURED FOR PREDICTING A HYPOGLYCEMIC EVENT FOR A SUBJECT

Absstract of: WO2026154038A1

A computer implemented method of training a machine learning model configured for predicting a hypoglycemic event for a subject is proposed, comprising: i. (126) receiving at least two time series of glucose sensor data, wherein each time series of glucose sensor data comprises a plurality of glucose measurements measured by a glucose monitoring device (112) configured for detecting glucose in a bodily fluid of the subject, wherein a first time series of the two time series of glucose sensor data is measured at least during and/or after being exposed to a first predefined program of glycemic stimuli, and wherein a second time series of the two time series of glucose sensor data is measured at least during and/or after being exposed to a second predefined program of glycemic stimuli, wherein the second predefined program of glycemic stimuli is different from the first predefined program of glycemic stimuli; ii. (128) forming a first training dataset using the first time series and forming a second training dataset using the second time series; iii. (130) training the machine learning model with the first training dataset and the second training dataset.

METHOD FOR PERFORMING DIRECTORS AND OFFICERS RISK ASSESSMENT USING A DATA-SCIENCE AND RISK PREDICTION MODEL

Absstract of: US20260212298A1

The present invention is directed to a method, performed by a computer system, for the prediction of the directors and officers (D&O) risk of a query company, wherein said method is based on the use of machine learning to construct a quantitative risk-prediction model based on features produced from data that has been acquired from a plurality of data sources, and to train and validate said model. One key aspect of the invention is that the prediction model uses both features related to the query company and features related to the directors and officers of said company.

First Node, Second Node and Methods Performed Thereby, for Handling One or More Machine Learning Models

Absstract of: US20260212223A1

0000 A computer-implemented method performed by a first node (111). The methods is for handling one or more machine learning models. The first node (111) operates in a communications system (100). The first node (111) determines (404), using machine learning, ML, one or more ML models of an indicator of operation of the communications system (100). The determining (404) is based on a respective operation mode of Radio Access Technology (RAT), used by a respective set of nodes (121,122,123) wherefrom respective data has been collected to train, or infer, a respective ML model of the one or more ML models. The first node (111) also provides (405) a respective indication of the determined one or more ML models to a second node (112) operating in the communications system (100).

GENERATION OF SYNTHETIC DATASETS USING MACHINE LEARNING MODELS FOR ZERO-SHOT CLASSIFICATION

Absstract of: US20260212256A1

Generation of synthetic dataset using machine learning (ML) models for zero-shot classification is provided and includes receiving a first input that includes a plurality of labels. A first prompt associated with a first label of the plurality of labels is generated. A first ML model is applied to the first prompt and a first set of labelled datapoints is generated. At least a first sub-prompt associated with at least one labelled datapoint of the first set of labelled datapoints is generated. The first ML model is applied to the at least first sub-prompt and a first subset of labelled datapoints is generated. Based on the first set of labelled datapoints and the first subset of labelled datapoints a training dataset is generated and outputted.

PERTURBATION ANALYTICS OF MACHINE LEARNING MODELS

Nº publicación: US20260212231A1 23/07/2026

Applicant:

HEWLETT PACKARD ENTPR DEV LP [US]
HEWLETT PACKARD ENTERPRISE DEVELOPMENT LP

Absstract of: US20260212231A1

0000 Systems and methods are provided for implement a perturbation technique of a subset of input time series data (e.g., a “motif”) provided to a machine learning model to help determine the effect of the motif on the inference of the model, absent any knowledge of the weights/biases corresponding with that particular motif in the input data. The replacement of the motif may be implemented to help describe the significance of a repeating pattern in the time series data with respect to its effect on the inference output.