Generative Model Program Generation for Experiment Devices

Solution Overview

Problem

Existing generative models struggle to handle information not present in their training data, making it difficult to generate programs for experiment devices without sufficient teaching data, which is costly and time-consuming.

Innovation Solution

The system generates input information for a generative model that includes reference information related to the experiment device, such as definition information for interfaces and sample programs, allowing the model to generate programs that can be executed by the device.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If extensive training data is used to teach the generative model, then the model can generate programs for experiment devices, but the cost and time required increase significantly

Engineering Contradiction:

Improveprogram generation capabilityVSAvoidtraining time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent extracts only the essential interface definition information and sample programs from the experiment device documentation, rather than using the entire training dataset. This selective extraction allows the model to learn the necessary program generation patterns without processing all available data, thereby reducing training time and computational resources while maintaining program generation capability.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies partial action by training the generative model on a subset of the available training data that contains the necessary interface definitions and sample programs. This partial training approach is sufficient to achieve the desired program generation capability without requiring exhaustive training on all experiment device documentation, thus reducing the time and cost investment.

Inventive Principle:

Principle #16Partial or excessive action

2Reliability

If extensive training data is used to teach the generative model, then the model can generate programs for experiment devices, but the cost increases significantly

Engineering Contradiction:

Improveprogram generation capabilityVSAvoidtraining data quantity

Core Design Contradiction:

ReliabilityVSQuantity of substance

Solution Approach 1:

The patent extracts only the essential interface definition information and sample programs from the experiment device documentation, rather than using the entire training dataset. This selective extraction allows the model to learn the necessary program generation patterns without processing all available data, thereby reducing training time and computational resources while maintaining program generation capability.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies partial action by training the generative model on a subset of the available training data that contains the necessary interface definitions and sample programs. This partial training approach is sufficient to achieve the desired program generation capability without requiring exhaustive training on all experiment device documentation, thus reducing the time and cost investment.

Inventive Principle:

Principle #16Partial or excessive action

3Adaptability or versatility

If prompt optimization techniques are applied to add information not in training data, then the model can handle new interfaces, but the complexity of the system increases

Engineering Contradiction:

Improvehandling new interfacesVSAvoidsystem complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent incorporates interface definition information and sample programs into the training data before the model needs to generate programs for new interfaces. This preliminary inclusion of interface information in the training data allows the model to naturally handle new interfaces without requiring complex prompt optimization techniques or additional system components, thereby reducing overall system complexity while maintaining adaptability.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20250123882A1Information processing device, information processing method, and recording medium

Publication Date: 2025.04.17 PREFERRED NETWORKS INC

AI summary

An information processing device includes at least one memory and at least one processor. The at least one processor is configured to generate input information for generating a program to be executed by an experiment device; and acquire the program generated by inputting the input information into a generative model. The input information includes reference information related to the experiment device, and instruction information. The reference information includes definition information related to an interface for executing a function of the experiment device and a sample program using the interface. The instruction information includes information instructing the generative model to generate the program.