USPTO Patent Applications - AI & Computing (G06N)

Probabilistic Programming Approach to Intention Estimation in Human-Robot Teleoperated Assembly Tasks

USPTO published Honda Motor Co., Ltd.'s patent application US20260099738A1 for probabilistic modeling methods to estimate operator intentions in human-robot teleoperated assembly tasks. The method predicts next actions in assembly sequences. The application was filed on 2025-02-14.

Computer-Readable Recording Medium Storing Quantum Circuit Information Generation Program

The USPTO published patent application US20260099745A1 filed by Fujitsu Limited on October 7, 2024. The application covers a quantum circuit information generation program that specifies partial circuit types in target quantum circuits and stores circuit information with reference capabilities. The invention is classified under CPC G06N 10/20.

Quantum State Initialization Method and Device

USPTO published patent application US20260099746A1 for Korea Advanced Institute of Science and Technology (KAIST), covering a quantum state initialization device and method. The invention includes a control register, auxiliary registers, collective CNOT gate unit, detectors, and quantum state output unit implementing noise removal protocols. CPC classifications G06N 10/20 and G06N 10/40.

Quantum Processor Calibration Techniques by ColdQuanta

The USPTO published patent application US20260099747A1 for ColdQuanta's quantum processor calibration techniques. The application covers methods for efficiently calibrating control parameters of quantum information processors by executing quantum circuits multiple times while varying control parameters to determine calibrated values. Inventors include Daniel C. Cole, David Robert Mason, and Mark Saffman. The application was filed on October 8, 2024.

Intel Corporation Multimodal LLM Audio Video Tokenization Patent Application

USPTO published patent application US20260099522A1 by Intel Corporation inventors Kuba Lopatka, Adam Kupryjanow, and Tomasz Szmelczynski for power-efficient tokenization and long-context storage of audio and video data for multimodal large language models. The application covers systems with specialized subsystems for receiving input signals, generating discrete tokens, and buffering tokens for durations ranging from seconds to hours.

Expert Selection System for MoE Large Language Models

USPTO published patent application US20260099697A1 for an expert selection system in Mixture of Experts (MoE) large language models. The system includes a global selector that manages selection between global and local modes, and a pre-fetcher that pre-loads selected global expert sets from main memory into faster cache memory. Inventors include Usman Sajid, Marie Mai Nguyen, Shuyi Pei, Younghoon Kim, and Rekha Pitchumani. The filing date was October 1, 2025, under application number 19347703.

Quantum Circuit Simulation Using Tensor Networks Patent Application

The USPTO published patent application US20260099752A1 on April 9, 2026, covering systems and methods for quantum circuit simulation using tensor networks, invented by Mekena Metcalf. The application addresses quantum kernel methods for classification and demonstrates tensor network effectiveness at scaling this application. The application was filed on October 3, 2024.

Distributed Training of Compressed Machine Learning Models

USPTO published patent application US20260099761A1 for an apparatus enabling distributed training of compressed machine learning models using quantization-based parameter compression. The invention manages compressed ML model parameters across distributed systems by cycling between compression and decompression states, allowing efficient parameter synchronization over networks. The application was filed on October 7, 2024.

Syntiant Neuromorphic IC Sensor Processing Patent Application

USPTO published patent application US20260100186A1 assigned to Syntiant, covering sensor-processing systems with neuromorphic integrated circuits for keyword spotting. The application was filed December 11, 2025, with inventors Kurt F. Busch, Jeremiah H. Holleman III, Pieter Vorenkamp, Stephen W. Bailey, and David Christopher Garrett. The patent describes systems using neuromorphic ICs with neural networks to extract features from sensor data and arrive at actionable decisions for keyword spotting applications.

AI-Based Exam Question Generating Method and System

The USPTO published patent application US20260100140A1 for ViewSonic International Corporation, covering an AI-based exam question generating method and system. The system receives reference materials and exam parameters input by a user, then uses a pre-trained AI model to induce exam question rules and generate compliant exam questions from the reference material. The AI model is pre-trained on educational material datasets to learn question generation patterns.

Multi-Language Model-Based Filtering and Labeling of Datasets

The USPTO published patent application US20260100887A1 assigned to Cisco Technology, Inc. The patent covers methods for labeling telemetry data from computer network entities using a combination of N-gram models and large language models. The labeled data is then used to train predictive maintenance models for network infrastructure. The application was filed on July 28, 2025.

Systems and Methods for Privacy-Enabled Biometric Processing

The USPTO published patent application US20260100842A1 filed by Private Identity LLC on May 20, 2025, covering systems and methods for privacy-enabled biometric processing. The invention derives encrypted feature vectors from biometric data and uses deep neural networks to authenticate users while preserving privacy. Homomorphic encryption enables authentication comparisons without decrypting the underlying biometric data.

Verizon Patent, Deterministic AI/ML Models, Apr 9

Verizon Patent, Deterministic AI/ML Models, Apr 9

AI Agent Feedback Adaptation for Wireless Networks Patent Application

The USPTO published patent application US20260099767A1 titled 'Feedback-Based AI/ML Models Adaptation in Wireless Networks.' The application covers an AI agent configured to report trained AI/ML models to an AI manager, receive performance feedback, and determine improvements based on that feedback. Inventors are Ping-Heng Kuo and Naveen Kumar R Palle Venkata. The application was filed on September 15, 2023.

Canva Files Patent on Multi-Label Classification Model Training

Canva Pty Ltd filed US Patent Application US20260099771A1 for methods and systems to train multi-label classification models to incorporate new entities added to a model dictionary. The method identifies candidate training instances based on the new entity, generates training records with the new label, and refines label sets to remove unrelated entries before model training. The application was published April 9, 2026, with a filing date of September 2, 2025.

Energy Cost Optimization Framework for EV Logistics Enterprise

USPTO has published patent application US20260099787A1, filed by Tata Consultancy Services Limited on June 24, 2025, for an AI-driven optimization framework that minimizes energy costs for electric vehicle (EV) logistics operations. The framework addresses cyclic dependency between electricity procurement costs and vehicle routing by iteratively optimizing routing costs and energy procurement from different energy sources.

Building Management System With Generative AI-Based Root Cause Prediction

USPTO published patent application US20260099773A1 titled 'Building Management System With Generative AI-Based Root Cause Prediction' on April 9, 2026. The application discloses a method for training a generative AI model using historical service requests to predict root causes of building equipment problems. Filing date was December 10, 2025.

Deep Learning Method for Lithium-Ion Battery Health Monitoring in Electric Propulsion

USPTO published Purdue Research Foundation's patent application US20260098906A1 on April 9, 2026, disclosing a deep learning method for predicting lithium-ion battery end-of-life in electric propulsion systems. The invention uses sensor data from real-time battery monitoring to generate operational recommendations for extending battery life and operational time during discharge cycles.

GAN Method and System for Processing Satellite Orbital Data and Predicting Maneuvers

The USPTO published patent application US20260098980A1 by Godwin et al. covering methods and systems for processing satellite orbital information using Generative Adversarial Networks (GANs). The technology enables evaluation of satellite orbital positions, prediction of future movements, and detection of orbital maneuvers potentially indicating nefarious intent. The patent has been assigned CPC classifications G01V 3/38, G06F 18/214, G06N 3/045, and G06N 3/08.

System and Method for Flowchart-Guided Dialogue Leveraging Large Language Models

Openstream Inc. has filed US Patent Application US20260099715A1 for a system and method that integrates flowcharts into large language model dialogue generation, enabling LLMs to handle both structured and unstructured conversations more effectively. The system is designed to manage user digressions from predefined dialogue paths, where users may skip steps or switch topics during conversations. The application was published on April 9, 2026, with a filing date of September 22, 2025, under Application No. 19335493.

Model Based Ranging Localization, Patent Filed

Model Based Ranging Localization, Patent Filed

ML-Directed Evolution Using Protein Language Models and CNN

Solugen, Inc. published patent application US20260099710A1 for a machine learning method used in protein engineering. The method generates fitness libraries using protein language model-guided site selection, trains convolutional neural networks to predict protein fitness, and optimizes sequences through phase transition-based algorithms with heating and cooling cycles.

AI Machine Learning Patent for Detecting Trade Spoofing Patterns

The USPTO has published patent application US20260099711A1 filed by Trading Technologies International, Inc. covering artificial intelligence and machine learning techniques for processing trading data to detect trade spoofing patterns. The application describes using semi-supervised machine learning with positively labeled and unlabeled training data to develop classification models distinguishing spoofing behavior from legitimate trading. Clustering techniques segment trading activity into assessable bursts for potential spoofing status. The application was filed on October 10, 2025, under application number 19355773.

Machine Learning Model Compression Patent Application

USPTO published patent application US20260099712A1 for Salesforce, Inc. on April 9, 2026. The application covers techniques for compressing machine learning models by removing and replacing blocks while preserving outputs, enabling execution on low-resource devices. The inventors are Romain Cosentino, Sarath Shekkizhar, Damjan Kalajdzievski, and Adam Earle.

Prediction Device Optimizes Language Model Learning Epochs

NEC Corporation filed patent application US20260099713A1 for a prediction device that optimizes the number of learning epochs for language models. The device includes an acquisition unit to obtain calculation resource constraints and target language resource amounts, along with a prediction unit that forecasts epoch ranges to minimize learning loss.

NEC Corporation Predicts Language Model Performance Across Multiple Languages

NEC Corporation has obtained USPTO Patent Application US20260099714A1 for a prediction device that forecasts language model performance when trained across multiple languages. The device acquires model size, training data amount, and target language ratio to predict model loss using a function combining model size, training data, and a power of the target language ratio. The patent covers 5 CPC classifications including G06N 3/084 and G06F 18/24147.

Authority-Based LLM Training Patent Application US20260099716A1

USPTO published patent application US20260099716A1 by inventors Anna Luti and Paolo Antinori, covering an authority-based training process for large language models. The system generates data quality metrics and authority scores for training samples to dynamically adjust weights during LLM training. The application was filed on October 3, 2024, under CPC classification G06N 3/0895.

Human-in-the-Loop AI Training for Agentic Automation Patent Application

USPTO published patent application US20260099135A1 by UiPath, Inc. covering human-in-the-loop automation training using AI for agentic automation systems. The invention enables a listener to monitor user or AI agent interactions with computing systems and improve or personalize automation based on those interactions.

LLM Unlearning via Loss Adjustments - Accenture Global Solutions

USPTO published patent application US20260099772A1 by Accenture Global Solutions Limited disclosing a system and method for large language model unlearning via a forget data only loss adjustment (FLAT) function. The invention involves accessing forget data samples, associating template responses via LLMs, and training a target LLM using loss adjustments to maximize divergence between template and forget answers.

Neural Network Quantum Error Correction Decoding Method and Apparatus

USPTO published patent application US20260099754A1 by Tencent Technology (Shenzhen) on April 9, 2026. The application covers neural network-based methods for quantum error correction decoding, including error syndrome acquisition, feature extraction via neural network decoder, and error result determination for quantum circuits.

AI Models for Edge Case Driving Scenarios

The USPTO published patent application US20260099762A1 from AUTOBRAINS TECHNOLOGIES LTD describing methods for generating AI models for autonomous driving using clustered driving scenario data to enhance decision-making in edge case scenarios.

AI Model Explainer for Non-Numerical Data Types

The USPTO published patent application US20260099763A1 by inventors Wan et al. covering mechanisms for AI model explanation of non-numerical data. The system converts non-numerical feature data into numerical representations, processes these through an AI model explainer to generate explanations, and converts outputs back to non-numerical form using two trained computer models.

Hierarchical Speech Analysis Method for Age, Gender, and Emotion Detection

USPTO published patent application US20260100196A1 for Tencent America LLC, covering a hierarchical speech analysis method using two-stage neural networks to detect speaker age, gender, and emotion from voice signals. The first learning stage performs initial detection while the second stage refines these attributes. This patent application relates to AI-driven speech processing technology.

Machine Learning Predicts Gene Sequence Effects on Endophenotypes

USPTO published patent application US20260100241A1 by Inari Agriculture Technology, Inc. describing a machine-learning method for predicting how gene regulatory sequences affect endophenotypes. The method involves inputting gene regulatory sequences into a trained model to generate effect predictions and selecting sequences based on desired phenotypic profiles.

Bank of America AI Parameter Adjustment in Distributed Network Patent Application

USPTO published Bank of America Corporation's patent application for AI parameter adjustment systems in a distributed network. The application describes methods for analyzing user data with AI engines, generating reports based on defined parameters, and regenerating outputs based on user feedback. Filing date was October 9, 2024.

AI System Optimizes Heterogeneous Compute Memory Operations

USPTO published patent application US20260099366A1 for an AI system that optimizes operations across heterogeneous compute and memory resources, including systems with multiple base dies and attached memory dies. The application discloses methods for identifying and routing operation portions across distributed processing and memory resources.

Bank of America AI Cloud Resource Allocation Patent Application

The USPTO published Bank of America Corporation's patent application for an AI-based system that monitors cloud computing infrastructure utilization across deployed applications. The system uses artificial intelligence to analyze usage patterns, predict future resource needs, and generate optimization recommendations for cloud infrastructure. This patent application covers methods for maximizing efficient utilization of cloud computing resources across enterprise environments.

Conversational AI System for Real-Time Cooking Guidance

USPTO published patent application US20260099498A1, filed July 18, 2025 by inventor Paul Paturi, covering a computer-implemented conversational AI system that provides real-time cooking guidance. The system accepts voice or text input, uses natural language processing to convert speech to text, and delivers step-by-step recipe instructions with ingredient lists, time constraints, and reminders based on user profile and kitchen resources.

Code Generation Method and Apparatus, Storage Medium and Electronic Device

USPTO published patent application US20260099303A1 for an AI code generation method and apparatus. The invention acquires target text (program code or natural language) and inputs it into a trained code generation model to produce target program code. The model is trained on both code understanding tasks (syntax and semantic features) and code generation tasks (producing new code from sample code).

Resilient Optimizer States for Fully Sharded Data Parallel Distributed ML Training

USPTO published patent application US20260099411A1 for systems and methods enabling failure resiliency in distributed machine learning model training. The invention allows compute nodes to store replicated optimizer shards and recover from node failures by reconstructing optimizer state from surviving replicas. The application names five inventors and claims priority to filing date December 11, 2025.

Energy Management AI Using Time Series Forecasting for Power Load Prediction

The USPTO published patent application US20260097684A1 disclosing an AI-based energy management system that creates synthetic training datasets to forecast power load using deep learning models. The system predicts energy storage device state of charge and controls charging operations based on projected load.

Apparatus and Method for Diagnosing Vehicle Exhaust Increase Using Autoencoder

USPTO published patent application US20260097776A1 by Hyundai Motor Company for an apparatus and method of diagnosing causes of increased vehicle exhaust gas using autoencoder neural networks. The system receives input data including sensed values, calculates reconstruction errors between input and restored data from a pre-learned autoencoder, and diagnoses exhaust increase causes based on these errors. This AI-based diagnostic technology is relevant to automotive manufacturers and emission control systems.

Vehicle Forward Blind Spot Object Detection System Using AI

The USPTO published patent application US20260097779A1 for an AI-based vehicle blind spot detection system that distinguishes between animate and inanimate objects. The system applies different closeness thresholds depending on object type and provides driver alerts when the vehicle's projected path intersects with detected objects within threshold distances. Inventor Andrew D. Johnson filed the application on October 9, 2024.

AutoBrains Patents AI Driving Scenario Activation Method

USPTO published patent application US20260097785A1 assigned to AutoBrains Technologies Ltd, covering methods for AI model activation based on driving scenarios. The system uses vehicle sensor data to generate signatures, matches them against a dictionary of concept signatures, and activates appropriate AI models for autonomous driving decisions. The application was filed on October 8, 2024, under Application No. 18908831.

Systems and Methods for Sampling-Based Krylov Quantum Diagonalization

The USPTO published patent application US20260094034A1 for sampling-based Krylov quantum diagonalization systems and methods, filed by 9 inventors including Kunal Sharma, Minh Tran, and Antonio Mezzacapo. The invention enables quantum devices to prepare Krylov basis states, sample them classically, and diagonalize Hamiltonians to approximate ground state energy. This A1 publication makes the application publicly available but does not grant patent rights.

Dynamic Explainable AI Pipeline Composability and Customization

USPTO published patent application US20260094033A1 for an explainable AI pipeline system enabling customizable explainability operations based on user persona roles. The system coordinates data analysis on AI model outputs and model performance analysis, outputting customized explanation data. Inventors: Ria Cheruvu, Harsha Bajpai, Arshad Mehmood, Saima Sharmin. Application No. 19111676 filed March 30, 2023.

Multi-Stage Federated Learning in Wireless Networks

USPTO published patent application US20260094032A1 for multi-stage federated learning in wireless networks, filed September 15, 2023. Inventors: Ping-Heng KUO and Alexander SIROTKIN. The application covers a method where AI agents train partial models, which are aggregated into a global model by an AI manager, with a trustworthiness determination gate before proceeding to the next stage.

Cause Estimation via Knowledge Graph

The USPTO published Fujitsu Limited's patent application US20260094031A1 for a cause estimation method using knowledge graphs. The invention selects nodes similar to phenomena requiring cause analysis, generates sub-knowledge graphs by tracing causal relationships, and determines confidence levels to identify cause candidates. The application (Filing Date: 2025-09-23) covers AI-based causal reasoning technology in the G06N classification.

Pulse-Regulated Temporal Architecture for Persistent Cognitive Machines

The USPTO published patent application US20260094030A1 filed by Brian Galvin covering a pulse-regulated temporal architecture for persistent cognitive machines. The system implements adaptive curvature-based feedback across fast, medium, and slow pulse layers to maintain coherent timing in artificial cognition processes. The application was filed on December 8, 2025, with CPC classifications in G06N 5/04 and G06F 1/324.

AI Agent Operation in SaaS Platforms Based on User Profiles and Interaction Patterns

The USPTO published patent application US20260094029A1 by monday.com Ltd. covering systems and methods for integrating generative AI agents within SaaS platforms as credentialed users with read/write privileges. The AI agents automate data operations, synchronize cross-platform workflows, and enable intent-based interactions. Filing date was September 29, 2025; application number 19344215.