USPTO Patent Applications - AI & Computing (G06N)

Automated High Throughput Protorheology

USPTO published patent application US20260092850A1 disclosing methods and systems for automatic, high-throughput estimation of rheological properties of fluidic materials using videography and neural-network processing. The invention enables parallel testing to achieve economical property predictions without expensive rheometric equipment.

Machine Learning System for Modifying ADAS Behavior for Optimum Vehicle Trajectory

USPTO published patent application US20260091794A1 by inventors Nakamura and Kim for a machine learning system that modifies Advanced Driver Assistance Systems (ADAS) behavior to achieve optimum vehicle trajectory. The system uses V2X communication to transmit vehicle-specific customized data and receives individualized optimum behavior data, enabling connected vehicles to implement region-wide optimum behaviors while satisfying individual vehicle needs.

VIASAT patent for ML-based RF apparatus tuning

VIASAT patent for ML-based RF apparatus tuning

Dynamic Microarchitecture Adaptation Using Machine Learning

The USPTO published patent application US20260093501A1 by inventors Zhu Zhou, Bin Li, Gilles Pokam, and Wessam Elhefnawy for dynamic microarchitecture adaptation using machine learning. The invention discloses a reinforcement learning model trained on microarchitectural performance data to configure processor settings based on telemetry and workload characteristics. Application No. 18900524 was filed September 27, 2024.

Configurable Artificial Intelligence Powered Software Assistant

The USPTO published patent application US20260093507A1 filed by UiPath, Inc. for a configurable AI-powered software assistant providing co-pilot guided automation experiences using contextual knowledge repositories. The application covers methods for AI model-based recommendation responses within a decision-making structure. Application No. 18941083 was filed on November 8, 2024.

Processing Parallelism for Machine Learning Model Training

The USPTO published patent application US20260093523A1 for a processing system that schedules parallel training of machine learning model instances based on microbatch counts. The system identifies expected idle cycles of processing units during training and schedules additional model instances during those periods. Inventors include Sumanth Gudaparthi, Yao Cui Fehlis, Karthik Ramu Sangaiah, and Sonali Singh. The application was filed on September 30, 2024.

AI system predicts task completion using resource history

USPTO published patent application US20260093528A1 by inventor John Graham Cvinar for an AI system that schedules tasks by predicting task completion using resource history, task completion history, experience, and availability data.

AI classifier training using anomaly detection methods

The USPTO published patent application US20260093782A1 by inventors Krishnaprasad and Somashekhar describing systems and methods for training AI classifiers using anomaly detection. The application covers a two-phase approach where anomaly detection rules initially label data instances, then a trained classification model takes over once performance thresholds are met. This applies to machine learning model development for anomaly detection in various data sets.

Method for Training ML Model to Generate Text

USPTO published patent application US20260093959A1 for a computer-implemented method training machine learning models for text generation. The method involves preprocessing input text into character vectors, encoding to word vectors, generating predictive word vectors via a backbone model, and decoding to character probabilities for model updates. Inventors: Lukas Balles and Pit Neitemeier. Application filed September 29, 2025.

LLM Speculative Decoding for AI Inference Acceleration

The USPTO published patent application US20260093960A1 by inventors Yao Cui Fehlis and Jalal Uddin Mahmud, disclosing a method for accelerating large language model inference using speculative decoding. The invention uses a neural network to output speculative decoding parameters iteratively, minimizing cumulative runtime. The patent covers CPC classifications G06N 3/047, G06F 40/284, and G06N 3/092.

Generalized Entanglement Forging with Slater Determinants

USPTO published patent application US20260087388A1 titled 'Generalized Entanglement Forging with Slater Determinants' covering quantum computing systems for electronic structure calculations. The application (No. 18892848), filed September 23, 2024, was submitted by inventors led by Mario Motta and describes systems for generalized entanglement forging using non-orthogonal Slater determinants and Jastrow ansatz on quantum systems.

METHOD FOR GENERATING TRAINING DATA, AND ELECTRONIC DEVICE

USPTO published patent application US20260087385A1 filed by Beijing Baidu Netcom Science Technology Co., Ltd. on March 26, 2026. The patent covers methods for generating AI training data using state transition images derived from user interaction data, trajectory data generation, and multimodal model-based reasoning reference extraction for training interaction agent models. The inventors are Le Zhang, Yu Shi, and Jingbo Zhou.

DNN Inference Optimization Using Practical Early Exit Networks

USPTO published patent application US20260086912A1 disclosing methods and systems for optimizing DNN inference using early exit networks. The invention enables dynamic splitting of machine learning models based on processing load forecasts and adaptive batch sizing to improve computational efficiency. Application No. 19400394 was filed November 25, 2025.

Generative AI Content Retrieval Standardization Framework

The USPTO published patent application US20260087080A1 by inventor Jian JIAO for a generative AI framework that standardizes user queries and content items into a common object format with normalized values. The system improves content retrieval accuracy by combining selective online and offline calls to the generative AI model with a distilled encoder neural network, enabling real-time results.

AI System for Active Shooter Detection via Sensor Signals

USPTO published patent application US20260087415A1 disclosing an AI system for detecting active shooters using sensor signals from multiple devices. The system correlates sensor data exceeding threshold levels, applies machine learning to classify emergency types and severity, and generates response information for dispatch.

AI-Driven Structural Engineering Design System and Method

The USPTO published patent application US20260087189A1 filed by Alexander Davis on September 20, 2024, covering an AI-driven system for structural engineering design automation. The system uses machine learning trained on engineered structure datasets including structural failure instances to generate optimized structural designs, 3D CAD models, and code-compliant engineering documents.

US patent for cached MoE experts, Mar 26

US patent for cached MoE experts, Mar 26

Transactional Neural Reasoning AI (TNRAI) Patent Application

USPTO published patent application US20260087387A1 for Transactional Neural Reasoning AI (TNRAI), a novel AI system designed for complex decision-making. The application describes a five-pillar architecture including delta-path modeling, skew-based adversarial recognition, vector memory recall, ambient reasoning overlays, and a multi-logic arbitration engine. Filed by inventor Joshua B. Williamson under application number 19171498.

Fault Tolerant Quantum Computation via Logical Operators Measurement

USPTO published patent application US20260087389A1 by inventors Theodore James Yoder and Dominic Williamson disclosing systems and methods for low-overhead fault-tolerant quantum computation via measurement of logical operators. The invention utilizes a graph selection component to select an auxiliary graph and a measurement component to execute a deformed quantum stabilizer code on a quantum system. This patent has applications in quantum computing hardware and software development.

Quantum-Capacitance Simulation Using Gaussian Subspace Aggregation

The USPTO published Microsoft Technology Licensing's patent application US20260087390A1 for a quantum-capacitance simulation method using Gaussian-subspace aggregation. The method constructs and projects non-interacting Hamiltonians for material configurations, then uses sums-of-Gaussians procedures to approximate low-energy eigenstates under interacting Hamiltonians. Inventors are Samuel Boutin and Roman Bela Bauer.

Controlling Agents Using Ambiguity-Sensitive Neural Networks and Risk-Sensitive Neural Networks

USPTO published patent application US20260087311A1 for methods controlling AI agents using ambiguity-sensitive and risk-sensitive neural networks. The application covers action selection systems for agent control with four named inventors. Patent application was filed September 8, 2023, and published March 26, 2026.

Combined Deep Learning Inference and Compression Using Sensed Data

USPTO published patent application US20260087312A1 for a device and method combining deep learning inference with data compression using sensed data. The system encodes sensed data locally, transmits it in batches to a remote computing system, and receives optimized encoder and prediction models in return. Inventors: Damian Kelly, Megan O'Brien, Gregory Buckley, Colleen B. Caveney.

Deepfake Music Detection Apparatus Using AI Classification

The USPTO published patent application US20260087313A1 for BRAINDECK INC. on March 26, 2026, covering an AI-based apparatus and method for detecting deepfake music. The system analyzes audio features, voice separation probability, and neural vocoder usage to determine whether audio content is synthetically generated. Application No. 18929273 was filed on October 28, 2024.

Data Reconstruction Using Machine-Learning Predictive Coding

USPTO published patent application US20260087314A1 for a machine-learning method that reconstructs data samples in a time series using predictive coding. The method generates reconstructed versions of first and second data samples, then uses a neural network to predict intermediate data samples positioned between them. The application (No. 19107781) was filed July 27, 2023 and published March 26, 2026.

Microsoft patent guards multimodal AI from malicious prompt attacks

Microsoft patent guards multimodal AI from malicious prompt attacks

Single Trajectory Policy Optimization for Generative Machine Learning Models

The USPTO published patent application US20260087409A1 for 'Single Trajectory Policy Optimization for Generative Machine Learning Models' filed by 14 inventors. The application covers methods for training generative ML models by optimizing an objective function based on likelihoods and quality scores of generated data items. Patent applications are informational publications and do not impose regulatory obligations.

Meta-learning with Diverse Tasks for Few-Shot Learning

USPTO published patent application US20260087412A1 by inventors Cresswell et al. covering methods for improving meta-learning models for few-shot learning of unseen tasks through improved task diversity scoring and generation of diverse training tasks using unsupervised analysis of disentangled latent features. Patent applications represent informational publications without creating immediate compliance obligations for third parties.

Federated Learning with Backbone Decoder Models

USPTO published patent application US20260087416A1 by Sony Group Corporation covering apparatus and methods for federated learning using a backbone-decoder model architecture. The invention enables servers to distribute decoder components to edge devices for localized training while maintaining backbone model parameters centrally, with aggregated decoder updates returned to the server for model refinement.

Secure Federated Learning System for Healthcare Data Management with Privacy Preservation

USPTO published patent application US20260087167A1 for a secure federated learning system enabling healthcare institutions to train machine learning models locally on sensitive data without transferring raw information. The system incorporates AES and RSA encryption, secure aggregation, differential privacy protocols, and automated HIPAA and GDPR compliance monitoring. Filing date was February 14, 2025.

Personal AI agent creation for computing device actions

USPTO published patent application US20260087384A1 for personal AI agent technology enabling automated computing device actions through LLM-based prompts and asynchronous data retrieval. Inventors: Jiachen Yang, Chih-Lun Lee, Hao Liu, Ang Li. Filing date: September 24, 2025.

Error robust quantum compiler, 6 inventors, March 2026

Error robust quantum compiler, 6 inventors, March 2026

Training Machine-Learned Models with Temporal Conditioning for Time-Aware Inference

USPTO published patent application US20260087404A1 for a method of training machine-learned models with temporal conditioning to enable time-aware inference. The application, filed by Florian Nils Hartmann and Matthew Sharifi, covers extracting temporal features from source data to construct training inputs and generate content predictions using computed loss functions. CPC classification is G06N 20/00 (Machine Learning).

AI evaluates threaded connections using torque, rotation measurements

USPTO published patent application US20260087405A1 for a method and apparatus using artificial intelligence to evaluate threaded connections based on torque and rotation measurements. The invention involves two AI systems where model parameters can be transmitted between them, enabling quality prediction even when the systems are not in communication. Inventors: Rainer RUEHMANN, Benjamin SACHTLEBEN, David GEISSLER.

USPTO Patent Application for Dynamic Spectrum Management

The USPTO has published a patent application (US20260082236A1) filed by Digital Global Systems, Inc. for a system, method, and apparatus for dynamic, prioritized spectrum management and utilization. The application details a system incorporating monitoring sensors, data analysis, and a semantic engine to create actionable data for spectrum management.

USPTO Patent Application: Adaptive Loop Filter Methods for Video

The USPTO has published a new patent application (US20260082042A1) filed by BEIJING DAJIA INTERNET INFORMATION TECHNOLOGY CO., LTD. The application details methods and apparatus for adaptive loop filter and cross-component adaptive loop filter for video decoding and encoding, utilizing AI-based classifiers.

USPTO Machine Learning for Channel Estimate Patent Application

The USPTO has published a patent application from Lenovo (United States) Inc. related to machine learning for channel estimation. The application details methods for generating channel estimates using machine learning models configured by reference signals.

USPTO Patent Application: Machine Learning Measurement Reporting in Wireless Communication

The USPTO has published a patent application from LG Electronics Inc. detailing a method for machine learning-based measurement reporting in wireless communication systems. The application describes a user equipment's process for configuring, obtaining, and transmitting measurement results based on machine learning models.

Samsung Patent Application for Semiconductor Device

The USPTO has published a patent application from Samsung Electronics Co., Ltd. for a novel semiconductor device. The application details a specific structure involving a channel layer, a ferroelectric layer with a unique interface region, and a gate electrode, aiming to improve semiconductor performance.