Query Change Feedback for More Accurate Search Result Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Generative large models in Retrieval-Augmented Generation (RAG) struggle to effectively utilize behavioral feedback signals, particularly explicit human feedback, due to challenges such as quality and consistency, scale and cost, timeliness, and data bias, which affect the relevance and accuracy of generated results.
Innovation Solution
A method that includes acquiring a change query corresponding to an input query, obtaining a reference result through search, and generating an output result based on the input and change queries and reference result, utilizing a generation model trained with prompts containing original queries, change queries, search results, and target instructions to enhance relevance and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional search systems provide multiple results through recall, rough ranking, precise ranking, and fine-tuning, then the system can meet basic search needs, but the relevance and accuracy of results are insufficient compared to RAG systems
Solution Approach 1:
The patent applies feedback by acquiring change queries corresponding to input queries and using them to adjust search strategies. The system obtains reference results through search and generates output results by integrating the input query, change query, and reference result. This feedback loop from user query modifications to result generation improves result accuracy while managing system complexity through structured information flow.
Solution Approach 2:
The patent introduces change queries as an intermediary element between the input query and the final output result. The change query serves as a mediator that captures user intent modifications and guides the search process. This intermediary mechanism enables the system to achieve RAG-level accuracy by systematically incorporating user feedback without requiring the full complexity of a complete RAG system.
2Reliability
If generative large models in RAG are used to generate accurate and content-rich responses, then the relevance and accuracy of results improve, but the system faces challenges with quality and consistency of feedback signals, scale and cost, timeliness, and data bias
Solution Approach 1:
The patent extracts only the essential feedback signal - the change query - from complex user interactions. Instead of processing all possible feedback types (clicks, dwell time, rewrites, etc.), the system focuses on extracting change queries that directly indicate user intent modifications. This extraction approach improves result relevance while reducing feedback processing complexity by concentrating on the most informative signal.
Solution Approach 2:
The patent applies local quality by treating different types of feedback signals differently based on their informativeness. Rather than uniformly processing all feedback, the system specifically targets change queries as the most valuable local feedback signal. This selective approach allows the system to achieve high reliability in result relevance while avoiding the complexity of comprehensive feedback processing.
3Measurement precision
If the system acquires and processes change queries to generate more relevant output results, then the accuracy of generated results increases, but the processing time and computational resources increase
Solution Approach 1:
The patent applies preliminary action by acquiring change queries corresponding to input queries before generating the final output result. The system prepares the change query information in advance and uses it to guide the search and generation process. This preliminary preparation of feedback information enables more accurate result generation while optimizing processing time by avoiding last-minute computations.
Solution Approach 2:
The patent applies partial action by focusing on processing only the essential change query information rather than all possible feedback signals. The system selectively uses change queries to generate output results, achieving high accuracy without the computational overhead of processing excessive feedback data. This partial processing approach balances accuracy improvement with time efficiency.
Data Source
AI summary
Provided is a result generation method, a generation model training method, an electronic device and a storage medium, relating to the field of computer technologies, and in particular, to the field of search and generative model technologies. The result generation method includes: acquiring a change query corresponding to an input query; obtaining a reference result by searching according to the input query and the change query corresponding to the input query; and generating an output result corresponding to the input query according to the input query, the change query corresponding to the input query and the reference result.