BG-RAG-SFL Knowledge Graph Retrieval for Multi-LLM Accuracy

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing retrieval-augmented generation (RAG) systems face challenges in accurately capturing complex interconnections between information pieces and managing multiple large language models (LLMs) due to differences in query modalities and syntaxes, leading to inefficiencies and inaccuracies in response generation.

Innovation Solution

The Bayesian Graph-Based Retrieval-Augmented Generation with Synthetic Feedback Loop (BaG-RAG-SFL) system integrates Bayesian evaluation, graph-based retrieval, and a synthetic feedback loop to optimize LLMs for accuracy, efficiency, and adaptability by using a knowledge graph, secondary ground-truth verification, multi-agent verification, and continuous improvement through synthetic data generation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional RAG systems are used, then information retrieval is performed, but complex interconnections between information pieces cannot be accurately captured

Engineering Contradiction:

Improveaccuracy of capturing interconnectionsVSAvoidcomplexity of information representation

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments information into knowledge graph entities and relationships, allowing complex interconnections to be represented as structured nodes and edges. This segmentation enables precise capture of relationships while maintaining manageable complexity through graph data structures.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent transitions from traditional flat text-based information representation to a multi-dimensional knowledge graph structure that captures entities, relationships, and metadata across multiple dimensions. This dimensional transformation enables accurate representation of complex interconnections that are difficult to capture in traditional RAG systems.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Adaptability or versatility

If multiple LLMs are used to handle different query modalities, then query handling versatility is improved, but system complexity increases due to differences in query modalities and syntaxes

Engineering Contradiction:

Improvecapability to handle different query modalitiesVSAvoidcomplexity of managing multiple LLMs

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements a unified query processing framework that can handle multiple query modalities (text, code, structured data) through a single system architecture. The knowledge graph serves as a universal representation that accommodates different query types, eliminating the need for separate specialized systems for each modality.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent introduces a query translator and knowledge graph intermediary layer that converts diverse query modalities into a unified representation. This intermediary component simplifies management of multiple LLMs by providing a standardized interface and common knowledge base that all models can access and process consistently.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Reliability

If RAG systems are used for response generation, then relevant information can be retrieved, but inaccuracies occur due to differences in query modalities and syntaxes across multiple LLMs

Engineering Contradiction:

Improveaccuracy of response generationVSAvoidcompatibility across different LLMs

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent standardizes query parameters and knowledge graph representations to ensure consistent processing across different LLMs. By normalizing query modalities and syntaxes through the knowledge graph framework, the system maintains reliable response generation while preserving adaptability to different model architectures.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent implements feedback mechanisms that allow the system to learn from and adapt to different LLM responses. The knowledge graph is continuously updated with verified information from multiple models, enabling the system to improve accuracy and compatibility over time through feedback loops that refine query processing and knowledge representation.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12437213B2Bayesian graph-based retrieval-augmented generation with synthetic feedback loop (BG-RAG-SFL)

Publication Date: 2025.10.07 ZON GLOBAL IP INC

AI summary

An advanced AI system, known as Bayesian Graph-Based Retrieval-Augmented Generation with Synthetic Feedback Loop (BG-RAG-SFL), combines Bayesian evaluation, graph-based retrieval, and synthetic data feedback to create a continuously improving AI platform. The present invention integrates multiple LLMs, optimizing their performance while managing complexities across different models. Key features include a knowledge graph-based RAG system, a Bayesian evaluation network, a secondary ground-truth graph for verification, synthetic data generation for ongoing improvement, and a multi-agent verification system. The system also functions as an AI operating system capable of acting as a virtual user with screen I/O control and managing multiple computers as an intelligent process automation system.