Heuristic recursive training and artificial general intelligence twin system (agilts)
The AGILTS system addresses the limitations of current AI systems by using a White-Box Twin architecture with L-DAG to enable efficient deductive reasoning and novel problem-solving, enhancing the logical reasoning capabilities of AI systems.
WO2025117991A1PCT designated stage expired Publication Date: 2025-06-05XI WENDY WUSAN
Patent Information
- Application Number
- PCT/US2024/061837
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-12-25
- Filing Date
- 2024-12-24
- Publication Date
- 2025-06-05
AI Technical Summary
Technical Problem
Current AI systems, such as LLMs, lack deductive reasoning, interpretability, and efficiency in solving complex logical reasoning tasks, relying on probabilistic reasoning and inefficient brute-force methods.
Method used
The AGILTS system employs a White-Box Twin architecture with a Logical-Directed Acyclic Graph (L-DAG) process, combining modular subnetworks for knowledge, rules, methods, and tools, to enable efficient deductive reasoning and address novel problems.
Benefits of technology
AGILTS effectively solves complex deductive tasks, addresses novel problems through self-trading, and enhances AI systems' logical reasoning capabilities, providing transparent and interpretable solutions.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure US2024061837_05062025_PF_FP_ABST
Abstract
The Personalized Heuristic Iterative-Recursive QA 3D Self-Study Training Process (XI Training) trains learners to master XI Training principles via the Personalized Heuristic Iterative-Recursive Question-Answer Process (PHIR-QA), building logical reasoning and self-study abilities. It can be implemented through an interpretable AGRLTY (Artificial General Rational Intelligence Logic Twin System), comprising a White-Box Twin (WB-Twin), a Black-Box Twin (trained by the WB-Twin for Al intuition when needed), and a Rational Network Flowchart (referred to as L-DAG) process. The WB-Twin consists of three layers: White-Box Twin Central Network (WBT-CN), Integration Hubs (IH), and Clustered Modules (CM). The WBT-CN integrates Knowledge, Rule, Tool, and Method subnetworks, while each CM is a template-supported, adaptable unit with PHIR-QA clusters, and each IH links modules sharing the same WBT-CN elements. The L-DAG generates solutions for complex logical tasks and establishes logical connections within the system to identify conflicts, gain novel insights, and explore potential pathways.
Need to check novelty before this filing date? Find Prior Art
Citation Information
Patent Citations
Mechanism-Data Heterogeneous Information Fusion Method for Digital Twin Modeling of Rotating Machinery
CN112765748B
Method, device, vehicle and computer program for modeling and monitoring a warming-up behavior of a vehicle component
WO2022258229A1
Machine for device verification and anomaly checking
WO2023214910A1