Large-Model Question Answering With Knowledge Conflict Resolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing retrieval augmented generation (RAG) models face conflicts between internal parameter knowledge and external knowledge, leading to inaccurate answers, particularly in complex long-tailed recognition scenarios, due to parameter knowledge conflicts and external knowledge conflicts.
Innovation Solution
A question answering method that determines a first result using a large model, retrieves additional information from a database, identifies the type of knowledge conflict between the two, and processes them to generate a target response, thereby improving accuracy by addressing these conflicts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a RAG model combines parameter knowledge from large language models with external knowledge from databases, then the accuracy of text generation and question answering is improved, but knowledge conflicts between internal parameter knowledge and external knowledge arise leading to inaccurate answers in complex long-tailed recognition scenarios
Solution Approach 1:
The patent introduces a knowledge conflict detection and resolution mechanism as an intermediary between the large language model and external database. This intermediary identifies conflicts between internal parameter knowledge and external knowledge, classifies them into different types (fact conflicts, logic conflicts, value conflicts), and applies resolution strategies to ensure consistent and accurate answers in complex long-tailed recognition scenarios.
2Adaptability or versatility
If external knowledge is retrieved from a database to supplement model knowledge, then the versatility and coverage of the system is improved, but conflicts between internal and external knowledge reduce the reliability of generated answers
Solution Approach 1:
The patent implements a feedback mechanism where the system continuously monitors for knowledge conflicts between internal model knowledge and external retrieved knowledge. When conflicts are detected, the system applies resolution strategies and feeds the resolved knowledge back into the generation process, ensuring that the expanded knowledge coverage does not compromise answer accuracy and reliability.
Data Source
AI summary
A question answering method based on a large model is performed by an electronic device. The method includes: determining a first result corresponding to a query statement based on a first large model, in which the first result includes a first response corresponding to the query statement, a first response logic corresponding to the first response, and first context knowledge corresponding to the first response; determining a retrieval result corresponding to the query statement by retrieving in a database based on the query statement; determining a type of a knowledge conflict between the first result and the retrieval result; and determining a target response corresponding to the query statement based on the first result, the retrieval result, and the type of the knowledge conflict.