Generative Model GUI With Dynamic Prompt Context Retention
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional chat-style interfaces for interacting with generative models like LLMs are sub-optimal for tasks with multiple steps or dependencies, leading to the displacement of previous responses off-screen, requiring users to scroll back and re-prompt the model, which can cause computational inefficiencies and loss of context.
Innovation Solution
A graphical user interface (GUI) that dynamically adjusts prompts based on user interactions, allowing items generated by the model to be displayed with associated prompts, and updates these prompts in real-time based on user selections, providing graphical shortcuts and reducing the need for re-prompting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a chat-style interface is used for interacting with generative models, then the interface is simple and easy to implement, but previous responses disappear off-screen and users must scroll back to view them
Solution Approach 1:
The patent transitions from a linear one-dimensional chat interface to a two-dimensional graphical interface with multiple panes and sections. The main chat area maintains the conversational flow while side panels display previous responses, generated items, and options. This dimensional expansion allows users to access historical information without scrolling through the entire chat history.
Solution Approach 2:
The interface is divided into multiple independent sections: main chat area, side panel with previous responses, item display area, and option buttons. Each segment serves a specific function and can be independently manipulated. This segmentation allows users to view multiple pieces of information simultaneously without them competing for the same screen space.
2Loss of information
If users scroll back through dialog sequence to view previous responses, then they can access historical information, but this increases interaction time and computational overhead
Solution Approach 1:
The system proactively displays previous responses and generated items in side panels before users need to reference them. This preliminary presentation of information eliminates the need for users to scroll through chat history to find previous responses. The graphical interface pre-organizes information in accessible locations, reducing retrieval time.
Solution Approach 2:
The graphical user interface acts as an intermediary between the user and the generative model's output. Instead of directly displaying only text in a linear chat sequence, the GUI intermediates by extracting key information and presenting it in structured visual formats with clickable items and option buttons, making previous responses easily accessible without scrolling.
3Loss of information
If re-prompting is used to obtain previously generated information, then users can retrieve context, but this causes the LLM to forget earlier information and incurs unnecessary computational costs
Solution Approach 1:
The graphical interface serves itself by maintaining and displaying context information locally in the UI elements. When users interact with the interface, the system updates the displayed items and options based on the current conversation state without requiring the LLM to regenerate or retrieve information. The interface keeps track of the conversation context and dynamically updates its display, eliminating the need for re-prompting.
Solution Approach 2:
The system implements feedback by monitoring user interactions with the graphical interface elements and automatically updating the displayed information. When users click on items or option buttons, the system processes these interactions and updates the context displayed in the interface, creating a continuous feedback loop that maintains context without requiring additional LLM calls.
4Productivity
If a graphical user interface with dynamic prompt adjustment is implemented, then user interaction efficiency is enhanced and context is maintained, but the device complexity increases
Solution Approach 1:
The graphical interface uses universal UI components that can serve multiple functions. For example, clickable items in the side panel can both display previous generated content and serve as triggers for regenerating or modifying that content. Option buttons can both present choices to users and automatically update the conversation context. This multi-functionality reduces the need for separate specialized components.
Solution Approach 2:
The patent merges multiple functions into integrated UI elements. The side panel combines display of previous responses with quick-access controls. Option buttons merge choice presentation with context updating functionality. This merging reduces the overall number of separate components needed while maintaining comprehensive functionality, thereby managing complexity.
Data Source
AI summary
Processor(s) of a system can: receive user input; process, using a generative model (GM), a GM input based upon the user input to generate a first GM output that includes a first set of items associated with a corresponding prompt for subsequent processing by the GM; cause the first set of items to be visually rendered using a first set of GUI elements; in response to receiving a user selection of a GUI element corresponding to an item of the first set of items, process, using the GM, the prompt associated with the selected item to generate second GM output that includes a second set of items; cause the second set of items to be visually rendered using a second set of GUI elements; and determine updated prompt(s) associated with the first set of items based upon a user interaction with the second set of GUI elements.