Absstract of: AU2025317582A1
System and methods for computer modeling in medicine. A sort of period table of medical models is described for personalized diagnostics, prognostics and therapeutics, including at least 80 major categories of medical models. Generative artificial intelligence and geometric deep learning techniques, and algorithms including 2D and 3D graph machine learning and GenAI algorithms, are described, tailored and applied to diagnostic disease description, prognostic prediction and therapeutic development and management, including generation of novel synthetic drugs. The AI and machine learning techniques and algorithms are applied to understand each individual's genetic, RNA and protein anomalies that represent the source of many unique patient diseases. AI-enabled software agents assist physicians and researchers in building patient medical models. Several personalized medicine applications of individualized medical modeling include cardiovascular
Absstract of: US20260350432A1
0000 Clinical event detection and recording using active learning are disclosed. The claimed techniques include maintaining sequences of sentences. Each sequence of sequences is associated with an entity, and each sentence in each sequence is associated with a label and a timestamp. The techniques include iteratively training a machine-learning model using each sentence of the sequences as input and the label as a ground-truth value. Each iteration of training includes generating a prediction indicating whether the timestamp of each sentence occurred before or after an event of the entity. Each iteration of training includes validating the machine-learning model based on the prediction and the label of each sentence and the ground-truth value.
Absstract of: US20260351387A1
A computer-implemented method includes: accessing electronic healthcare records of a group of patients, wherein each patient has received an elective surgical procedure; extracting a data structure encoding a plurality of features of each patient from the group of patients, wherein a subset of the group of patients undergo at least one hospital-based intervention after receiving the elective surgical procedure; determining, using a machine learning algorithm that operates on the data structure, a Shapley value for each of the features, wherein the Shapley value indicates a likelihood for each patient with a corresponding feature to receive at least one hospital-based intervention; identifying a subset of the plurality of features; and based on the identified subset of features, establishing a predictive tool to predict a combined likelihood for a patient to receive a hospital-based intervention.
Absstract of: WO2026229279A1
The invention relates to a smart matching system for use in the building sector, the system comprising a dynamic sectoral database (20), a machine learning multi-criteria matching engine (30), a self-adaptive dynamic scoring module (40) comprising a mechanism for automatically adjusting the prioritisation of the providers and that is configured to adapt continuously according to the evolution of the descriptive records, feedback from the users and other data relating to said providers, as well as a digital user interaction interface (10). The system is capable of receiving a service request in a free-form or structured language, and of returning a prioritised, contextualised and self-adaptive list of providers.
Absstract of: WO2026231819A1
Introduced here is an approach for addressing patient barriers to behavioral health change and measuring habit formation in relation to behavioral health changes. Methods and systems discussed herein apply specialized agents to a user's data to determine a barrier hindering the user's progress toward behavioral change, a strategy for addressing the barrier, and a tactic for implementing the strategy. An LLM then carries out the tactic using tailored content or conversations with the user. Methods and systems discussed herein further involve assessment of habit formation relating to behavioral change. Based on a dynamic assessment of a user's habit formation, measures to promote behavioral change can be taken.
Absstract of: WO2026231823A1
A material handling system includes a storage array. An automated package transport system stores and retrieves packages to and from the storage array. An automated palletizer, connected to the automated package transport for providing packages to the palletizer, places packages onto a pallet to form a pallet load. A controller has a pallet load generator having a pallet load planner with a plan model having plan parameters that inform generating a pallet load plan of the pallet load, and the pallet load generator has a machine learning resolver configured so as to heuristically resolve, via machine learning application, parameter variance from the plan parameters resulting in an optimizing solution for the pallet load plan.
Absstract of: WO2026231067A1
A system and a computer-implemented method for resolving hardware issues is disclosed. The method processes multi-modal inputs such as textual, auditory, visual, or sensor data, to identify hardware issues. Inputs are structured into contextual triads comprising an object, a connector, and meta information. The plurality of interrelationships among are detected to contextualize the hardware issue. A pretrained large language model analyzes the triads and the plurality of interrelationships to generate resolution triads which include step-by-step actions, contextual metadata, and prioritization based on criticality or urgency. The resolution triads are displayed via a user interface, enabling efficient issue resolution. The method uses a dynamically updated digital folksonomy dataset and machine learning to improve troubleshooting speed and accuracy in the industrial plant (106).
Absstract of: WO2026229177A1
A method for generating lighting plans based on a generative machine learning model, the lighting plans defining one or more parameters for operation of one or more lighting devices of a lighting system in an environment. The method comprising the steps of receiving an input from a user for generating a lighting plan, the input indicative of a lighting configuration and/or commissioning task, determining, by the generative machine learning model, a plurality of lighting plans based on the user input. The method further comprises determining for each lighting plan from the plurality of lighting plans a light effect associated with each lighting plan in the environment and determining a level of disorder in the plurality of lighting plans, wherein the level of disorder is indicative of a variability in the determined light effects associated with the plurality of lighting plans in the environment. If the level of disorder is above a disorder threshold, the method comprises providing a notification to the user.
Absstract of: WO2026231813A1
A method can include: receive a historical data set, extract feature vectors from the historical data set corresponding to the input features for the historical recidivism model, split the feature vectors into a training set and a test set, generate candidate machine-learning models by executing machine-learning algorithms using as input the training set, calculate a relative accuracy of the recidivism model and the candidate machine-learning models using an error rate for the recidivism model and error rates of the candidate machine-learning models, determine a drift based on a difference between a feature importance of the input features for the recidivism model and a feature importance of the input features for the candidate machine¬ learning models, and in response to the drift exceeding a predetermined drift threshold and/or the relative accuracy being below a predetermined accuracy threshold, generate an alert including the drift and the relative accuracy.
Absstract of: WO2026229937A1
A machine-learning system processes an ordered stream of channelized token representations, each token position including one or more structured feature channels, such as token-content, selector, hierarchy or conditional-vocabulary, reconstruction, state/control, formatting, entity, and audit/visibility channels. Selector information activates, masks, or conditions channel-specific prediction paths, applicability masks, vocabulary or codebook slices, and decoder or reconstruction behavior. Predictable literals, formatting, grammatical envelopes, entity references, and object syntax may be carried in non-content channels attached to carrier tokens, reducing separate substantive token predictions while preserving surface reconstruction. The architecture supports conditional vocabularies, object and encoding-object templates, persistent or successor-predicted state, reasoning-control state, hidden audit channels, entity-register references, future-defined or training-deferred channels, and shared processing of text, image, audio, structured, and tool tokens in a unified multimodal stream.
Absstract of: US20260349818A1
0000 A computer implemented method for identifying an idle server is disclosed. The method includes receiving metric data from a server; determining, by a first machine learning system, a multidimensional embedding that defines a signature of the server, the multidimensional embedding being based on the metric data; applying, by a second machine learning system, a clustering algorithm to the multidimensional embedding to determine a first cluster from a set of clusters; determining, based on the first cluster, that the server is an idle server; and deactivating the idle server.
Absstract of: US20260350714A1
Methods, systems, and computer storage media for providing financial data analytics recommendations using a data analytics engine in a customer relationship management system. The recommendations can be a lead that is information associated with a model-generated suggested consumer solution, an alert of increased risk of attrition, or an alert of increased risk of default. The data analytics engine is configured to generate target variables associated with financial products or the customer relationship and utilize modeling techniques and apply rules to generate recommendations. Operationally, the recommendations are generated based on a data analytics model. Generating the recommendations is based on feature variables that are generated based on aggregation and transformation of customer data and utilizing machine learning models to detect patterns in the customer data using the feature variables. The recommendations can be presented via a financial data analytics interface along with insights that provide plain text explanations of the recommendations.
Absstract of: US20260346756A1
According to one embodiment of the present invention, there is provided a battery life evaluation system including: a data collection unit configured to obtain information about route information, charge/discharge patterns, driving patterns, and cell states from an electric vehicle; and a life evaluation unit configured to predict the life of a battery installed in the electric vehicle by performing machine learning based on the obtained information.
Absstract of: US20260350428A1
0000 Various embodiments of the present disclosure describe machine learning monitoring and retraining techniques for automatically triggering model retraining based on evaluation scores. The techniques include receiving a request to process an input data object with a target machine learning model that is previously trained using an at least partially synthetic training dataset. The techniques include identifying a synthetic data object from the training dataset that corresponds to the input data object and, in response, modifying a holistic evaluation score for the model, initiating the performance of a labeling process for assigning a ground truth label to the input data object, and augmenting a supplemental training dataset with the input data object and the ground truth label. In the event that the holistic evaluation score decreased beyond a threshold, the model may be retrained with the supplemental training dataset.
Absstract of: WO2026231781A1
The present disclosure provides a comprehensive framework for distinguishing germline and somatic variants in cell-free nucleic acid samples by integrating fragmentomics-based features with statistical modeling and machine learning classification. The disclosed approach leverages orthogonal biological signals derived from cfDNA fragmentation patterns to complement conventional allele-based metrics.
Absstract of: US20260350430A1
Systems and methods are disclosed for constructing a data model using an analytic platform. One method may include receiving a data project, the data project including a solution request, determining, based on the solution request, a first data set from one or more data repositories, accessing the one or more data repositories to locate data of the first data set, storing the first data set in one or more data repositories, processing the first data set to provide for processed data, constructing, by a trained machine learning model, a data model based on the processed data and the solution request, determining a solution based on the constructed data model and the solution request, and automatically providing the solution and the constructed data model to one or more applications.
Absstract of: US20260350166A1
In various examples, a security graph is generated based on computing resources within a computing environment, and a machine learning model is used to determine placement information for threat scanners. For example, control plane and/or data plane logs are used to generate the security graph which includes source and sink nodes. Continuing this example, the machine learning model generates predictions based on the security graph, and the predictions are used to determine the optimal placement of threat scanners within the computing environment.
Absstract of: EP4835697A1
A computer-implemented method for watermarking a supervised machine learning model, said method comprising: obtaining a training dataset of said supervised machine learning model ; modifying a subset of said training dataset by applying one or more functions to the input features of each training sample that belongs to said subset ; training said supervised machine learning model using said modified training data set.
Absstract of: US20250217676A1
In an embodiment, a computer-implemented method for decoding neural activity is provided. In the method, at least one machine learning model is trained using a training data set of EEG data and concurrently collected environmental data collected from data collection participants. Once the at least one machine learning model is trained, EEG data measured from sensors attached to or near a user's head is received. Environmental data describing stimulus the user is exposed to concurrently with the measurement of the EEG data is also received. The EEG data and the environmental data is input into the at least one machine learning model to determine an inference related to the neural activity. Based on the inference, an operation of a computer program is altered.
Absstract of: WO2025145040A1
This specification discloses systems, methods, and techniques for classifying a grade of cancer represented in a tumor tissue sample. The techniques include obtaining a whole-slide image (WSI) of the tumor tissue sample and partitioning the WSI into a collection of patches. For each patch, (i) a contextualized feature representation of the patch is generated based on intrinsic features of the patch, local dependencies between the patch and a subset of local patches of the WSI, and non-local dependencies between the patch and a subset of non-local patches of the WSI; and (ii) an attention weight is determined for the patch. A WSI-level cancer grade for the tumor tissue sample is predicted based on the contextualized feature representations and the attention weights for the plurality of patches.
Absstract of: EP4835743A1
0001 A system (100), method (900), and apparatus (102) for generating specification data of one or more charging units (114) are provided. The system (100) retrieves training data associated with each charging unit (114) of the one or more charging units (114). The training data of each charging unit (114) of the one or more charging units (114) comprises charging event data pertaining to one or more charging events associated with the corresponding charging unit (114) of the one or more charging units (114). The system (100) trains a machine learning model to predict specification data of a charging unit (114) based on the training data. The system (100) receives input data associated with the charging unit (114) and causes the machine learning model to predict the specification data of the charging unit (114) based on the input data. The system (100) generates a vehicle (800) navigation function for a vehicle (802) based on the specification data.
Absstract of: US20260340357A1
Automated artificial intelligence (AI) agent evaluation is described. In one or more implementations, a first machine-learning model generates test prompts for a target AI agent based on testing guidance stored in a database, the testing guidance initially provided by a developer of the target AI agent. A second machine-learning model analyzes responses from the target AI agent to the test prompts to generate an analysis report of the target AI agent in view of the testing guidance. The second machine-learning model updates the guidance in the database for subsequent test prompts based on analysis of the responses. The second machine-learning model outputs the analysis report.
Absstract of: US20260340432A1
A taxonomical data structure is generated based on multi-modal unstructured enterprise data and unstructured external data using a machine learning component. A taxonomized knowledge graph is constructed based on the taxonomical data structure using the machine learning component. New data is obtained from at least one of an external data source or an internal data source. An action associated with the new data is inferred using the machine learning component and based on the taxonomized knowledge graph and the new data. An indication of the action is output.
Absstract of: US20260340494A1
0000 A system may receive incident data that documents a reported service problem and convert the incident data to an input set of vector embeddings. The system may execute a support group identifying machine learning (ML) model with the input set of vector embeddings to generate a support group identifier. The system may compare the input set of vector embeddings to solution vector embeddings generated from historical service problem solutions to determine whether a probable solution to the reported service problem is present in the historical service problem solutions. The system may in response to determining that the probable solution is present, transmit data documenting how to implement the probable solution. The system may in response to determining that that the probable solution is not present, connect a first user device to a support system device for a support group indicated by the support group identifier.
Nº publicación: US20260340559A1 29/10/2026
Applicant:
IBM [US]
International Business Machines Corporation
Absstract of: US20260340559A1
Systems and techniques that facilitate comparisons of machine learning model generated summaries are provided. For example, one or more embodiments described herein can comprise a system, which can comprise a memory that can store computer executable components. The system can also comprise a processor, operably coupled to the memory that can execute the computer executable components stored in memory. The computer executable components can comprise an integrated regulatory compliance component, comprising a machine learning model, wherein the machine learning model: identifies new regulations and enterprise standards for an enterprise; determines an impact to the enterprise based on the new regulations and enterprise standards; and updates baseline enterprise standards with one or more compliance actions based on the impact, wherein the baseline enterprise standards comprise an enterprise protected information baseline and an artificial intelligence (AI) baseline.