Large Language Model Rule Sequencing for Text Compliance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large Language Models (LLMs) often generate content that lacks adherence to specific rules required in certain scenarios, limiting their application in fields that need strict compliance with standards or specifications.
Innovation Solution
A method that involves obtaining a question from a user to instruct the generation of a target type of text, selecting or determining a set of target rules corresponding to the text type, and sequentially inputting these rules into an LLM to constrain and improve the accuracy of the generated content, using nodes such as rule, parallel, cyclic, and judgment nodes to manage and apply the rules effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If LLM generates content freely without rule constraints, then generation speed and flexibility are improved, but compliance with specific standards and specifications deteriorates
Solution Approach 1:
The patent applies preliminary action by pre-defining multiple sets of rules corresponding to different text types (official documents, contracts, etc.) before the generation process. These rules are stored in a rule library and automatically selected based on the input question, constraining the LLM's generation process in advance to ensure compliance while maintaining generation efficiency
Solution Approach 2:
The patent segments the rule system into multiple independent rule sets, each corresponding to a specific text type. Each rule set contains specific constraints for that type (e.g., formal tone for official documents, specific clauses for contracts). This segmentation allows the system to apply only relevant rules for each generation task, maintaining both compliance and efficiency
2Manufacturing precision
If multiple rule sets are introduced to constrain LLM generation, then compliance accuracy is improved, but system complexity increases
Solution Approach 1:
The patent creates a universal rule management system that handles multiple text types through a common framework. The system uses a unified interface for rule storage, selection, and application across different text types (official documents, contracts, technical specifications, etc.), reducing the perceived complexity while maintaining comprehensive rule coverage
Solution Approach 2:
The patent introduces an intermediary rule selection mechanism that automatically matches input questions with appropriate rule sets. This intermediary layer (the rule selection module) simplifies the interaction between the user and the complex rule library by handling rule matching automatically based on keywords and text type identification
3Manufacturing precision
If sequential rule input is used to improve generation accuracy, then content quality is improved, but generation time increases
Solution Approach 1:
The patent applies partial action by selectively applying only the necessary rules from each rule set based on the specific requirements of the input question. The system identifies key constraints relevant to the current task and applies those rules, rather than sequentially applying all possible rules, thus reducing generation time while maintaining accuracy
Data Source
AI summary
The disclosure provides a method for optimizing content generated by a large model, an apparatus for optimizing content generated by a large model, an electronic device and a storage medium, and relates to the technical field of artificial intelligence, especially to the technical fields of text processing, large language model and the like. It can be applied to official document processing, automatic contract generation, legal document writing, enterprise internal system management and so on. The method includes: obtaining a question entered by a user, wherein the question is used to instruct a generation of a text of a target type; obtaining a set of target rules corresponding to the target type from a plurality of preset sets of rules, in which the set of target rules includes a plurality of target rules, and the target rules are rules followed by the target type of text; and according to a sequence of the target rules, inputting the plurality of target rules into a large language model sequentially to obtain a target text of the target type generated by the large language model. In this way, the accuracy of generating text following certain rules by the large language model is improved.