Evolutionary Prompt Generation for Scored LLM Refinement

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Manual prompt engineering for large language models (LLMs) is inefficient, difficult to scale, prone to inaccuracies, and subjective, failing to capture domain complexities and nuances, leading to biased and nonsensical outputs.

Innovation Solution

Automated prompt engineering using evolutionary operators and machine learning to generate and score candidate prompts, applying crossover and mutation to iteratively refine prompts based on objective metrics, enhancing performance and reducing reliance on human intuition.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If manual prompt engineering is used, then prompt creation is simple and controllable, but it is time-consuming, subjective, and fails to capture domain complexities

Engineering Contradiction:

Improveprompt creation simplicityVSAvoidprompt generation efficiency

Core Design Contradiction:

Ease of operationVSProductivity

Solution Approach 1:

The system employs self-service by enabling the LLM to generate and evaluate prompts autonomously through iterative evolution. The model creates candidate prompts, scores them based on predefined criteria, and refines them without requiring manual intervention, thereby transforming a manual process into an automated self-improving system that maintains simplicity while dramatically increasing productivity

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces the mechanical manual process of prompt engineering with an automated computational system. Instead of humans manually crafting prompts, the system uses algorithmic generation, scoring, and evolution mechanisms to automatically create and optimize prompts, substituting human cognitive labor with automated AI-based processes

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Ease of operation

If manual prompt engineering is used, then human intuition and control are maintained, but inaccuracies and biases are introduced

Engineering Contradiction:

Improvehuman controlVSAvoidprompt accuracy

Core Design Contradiction:

Ease of operationVSReliability

Solution Approach 1:

The system implements feedback mechanisms where the LLM evaluates generated prompts against predefined criteria and uses this feedback to iteratively refine them. The scoring mechanism provides objective feedback on prompt quality, allowing the system to automatically correct inaccuracies and reduce biases through multiple generations of evolution until termination criteria are met

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent employs parameter changes by systematically varying prompt characteristics through evolutionary operators. The system modifies prompt parameters such as wording, structure, and complexity through controlled transformations, allowing objective optimization of prompt accuracy without relying on subjective human judgment

Inventive Principle:

Principle #35Parameter changes

3Use of energy by stationary object

If manual prompt engineering is used, then resource consumption is low, but scalability is limited

Engineering Contradiction:

Improveresource consumptionVSAvoidscaling capability

Core Design Contradiction:

Use of energy by stationary objectVSAdaptability or versatility

Solution Approach 1:

The system achieves universality by creating a multi-functional automated prompt generation framework that can handle diverse domains and tasks. The same evolutionary algorithm and scoring mechanism work across different application areas, enabling the system to scale from single-domain to multi-domain prompt generation without requiring domain-specific manual engineering for each case

Inventive Principle:

Principle #6Universality (Multi-functionality)

4Productivity

If automated prompt generation with evolutionary operators is used, then prompt generation efficiency and accuracy are improved, but system complexity increases

Engineering Contradiction:

Improveprompt generation efficiencyVSAvoidsystem complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent applies segmentation by dividing the complex prompt generation system into distinct functional modules: prompt generation, scoring/evaluation, selection, and evolution. Each module performs a specific function independently, making the overall complex system manageable through modular architecture while maintaining high productivity through automated coordination of these segments

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS20250278630A1Prompt generation with evolutionary operators

Publication Date: 2025.09.04 INTUIT INC

AI summary

Certain aspects of the disclosure pertain to prompt creation using language models in an evolutionary algorithm framework. A language model can be employed to generate an initial set of candidate prompts that return responsive replies to legitimate questions and disapproval replies to illegitimate questions. Candidate prompts can be scored. Subsequently, two or more candidates can be selected based on their scores. Additionally, candidate prompts can be generated with a language model by applying evolutionary operations to the two or more candidate prompts. Scores can be generated for the additional candidate prompts, and a termination criterion is evaluated to determine whether another iteration should be performed. After the termination criterion is satisfied, one or more candidate prompts can be output based on their score.