LLM Query Decomposition for Comprehensive Low-Input Responses

Solution Overview

Problem

Current generative models require extensive user input and resource consumption to handle complex queries, leading to battery depletion and inefficiencies in generating comprehensive responses.

Innovation Solution

Implement a system that decomposes input queries into sub-queries, utilizes corresponding tools to generate sub-query responses, and iteratively refines comprehensive responses to ensure responsiveness, thereby reducing client device resource utilization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If a user provides extensive additional inputs to perform further computer actions, then the user can obtain comprehensive information and complete actions, but the time required and battery consumption increase significantly

Engineering Contradiction:

Improvecompleteness of informationVSAvoidtime required for additional inputs

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system performs preliminary actions by automatically executing multiple sub-queries and gathering comprehensive information before the user needs to take action. The AI agent proactively searches for product details, prices, delivery dates, and installation information, then presents processed results to the user, eliminating the need for the user to manually search and collate information.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The AI agent performs self-service by autonomously executing the complex task of gathering and processing information without requiring extensive user input. The system automatically interprets the initial query, breaks it down into sub-queries, executes searches, and synthesizes comprehensive responses, serving the user's information needs independently.

Inventive Principle:

Principle #25Self-service

2Reliability

If a user provides extensive additional inputs to perform further computer actions, then the user can obtain comprehensive information and complete actions, but battery resources of the client device are significantly depleted

Engineering Contradiction:

Improvecompleteness of informationVSAvoidbattery consumption

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The system performs preliminary actions by automatically executing multiple sub-queries and gathering comprehensive information before the user needs to take action. The AI agent proactively searches for product details, prices, delivery dates, and installation information, then presents processed results to the user, eliminating the need for the user to manually search and collate information.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The AI agent performs self-service by autonomously executing the complex task of gathering and processing information without requiring extensive user input. The system automatically interprets the initial query, breaks it down into sub-queries, executes searches, and synthesizes comprehensive responses, serving the user's information needs independently.

Inventive Principle:

Principle #25Self-service

3Productivity

If an LLM generates a response to a complex input query, then the user receives a response, but the response is often underspecified and requires additional dialog turns

Engineering Contradiction:

Improveresponse generation speedVSAvoidcompleteness of response

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system segments the complex input query into multiple smaller sub-queries that can be handled independently. Each sub-query is processed separately to gather specific information, and then the results are synthesized into a comprehensive response. This segmentation allows the system to maintain both speed and completeness by processing information in manageable chunks.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system incorporates feedback mechanisms where the AI agent evaluates the completeness of its generated response against the original query requirements. If information is missing or insufficient, the agent automatically identifies gaps and executes additional sub-queries to fill them, ensuring comprehensive coverage without requiring multiple dialog turns.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS20250348485A1Generative model based decomposition of input query into sub-queries and generation of comprehensive response based on responses to sub-queries

Publication Date: 2025.11.13 GOOGLE LLC

AI summary

Some implementations relate to utilization of generative model(s) (e.g., large language model(s)) in selectively generating a comprehensive response for an input query, where the comprehensive response is generated based on multiple sub-query responses, and where the multiple sub-query responses are generated based on multiple sub-queries decomposed from the input query and corresponding tools for the sub-queries. Generating the comprehensive response based on the multiple sub-query responses integrates, into the comprehensive response, detailed information and/or actionable content that are responsive to the multiple sub-queries decomposed from the input query.