Client Model Task Prompts for Low-Resource Query Processing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Large language models require significant computational resources and network access, making them impractical for devices with limited memory and processing capabilities, and fine-tuning them for specific tasks is resource-intensive and time-consuming.

Innovation Solution

A client device uses a smaller client model trained on task prompts generated by a larger main-system model, allowing it to produce high-quality responses across various tasks without fine-tuning, while interacting with a network-accessible main system through an interface program.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a large language model is used to handle diverse queries, then the model's capability to understand and respond to various tasks is improved, but the computational resources and memory requirements increase significantly

Engineering Contradiction:

Improvemodel capabilityVSAvoidcomputational resources

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The system divides the language model functionality into two segments: a large main-system model hosted on a server that handles complex reasoning and diverse tasks, and a smaller client model that handles local inference. This segmentation allows the client device to operate with limited resources while still accessing the capabilities of the larger model through the API interface.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

An API interface acts as an intermediary between the client device and the main-system model. The client sends queries through this intermediary to the server, which processes them using the large language model and returns responses. This intermediary layer enables the client to access sophisticated model capabilities without needing to host the large model itself.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Use of energy by moving object

If a network-accessible language model is used, then the client device does not need large local resources, but latency and network dependency increase

Engineering Contradiction:

Improvelocal computational resourcesVSAvoidresponse latency

Core Design Contradiction:

Use of energy by moving objectVSSpeed

Solution Approach 1:

The system performs preliminary actions by pre-hosting the large language model on accessible servers and pre-establishing API interfaces. This allows the model to be ready for immediate query processing without requiring the client device to download or host the large model weights, reducing local resource requirements while maintaining accessibility.

Inventive Principle:

Principle #10Preliminary action

3Use of energy by moving object

If a smaller language model is used on the client device, then resource requirements are reduced, but the model's ability to understand broad ranges of queries deteriorates

Engineering Contradiction:

Improvecomputational resourcesVSAvoidquery understanding capability

Core Design Contradiction:

Use of energy by moving objectVSAdaptability or versatility

Solution Approach 1:

The main-system model on the server is designed with universal capabilities to handle diverse queries across multiple tasks and domains. The API interface is designed to be universally applicable, allowing the same client device to interact with the model for various purposes including text generation, analysis, and reasoning, thereby providing multi-functionality without requiring the client to host the large model.

Inventive Principle:

Principle #6Universality (Multi-functionality)

4Reliability

If a client model is fine-tuned for specific tasks, then task-specific performance is improved, but the time and resources required for training increase

Engineering Contradiction:

Improvetask-specific performanceVSAvoidtraining time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

Instead of fine-tuning the client model for each specific task, the system creates a copy of the universal main-system model on the server that can handle all tasks. The client simply copies the API interface and sends queries to the server, eliminating the need for time-consuming fine-tuning processes while maintaining task-specific performance through the server's sophisticated model.

Inventive Principle:

Principle #26Copying

Data Source

PatentUS12566765B2Executing a client model using a task prompt produced by a main system

Publication Date: 2026.03.03 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

A technique executes a client machine-trained model (“client model”) on a client device. In operation, the client device submits a description of a task to be performed by the client device to a network-accessible main system. The main system uses a main-system machine-trained model (“main-system model”) to produce a task prompt based on the task description. The client device subsequently uses the task prompt to process queries pertaining to the task. The main-system is trained to increase the accuracy of responses produced by the client model, while reducing the sizes of task prompts produced by the main system. The training process is performed by holding weights of the client model constant.