LLM Answer Feedback Interface with Candidate Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current feedback mechanisms for large language models are limited, as users can only passively receive answers or provide simple positive/negative feedback, making it difficult to obtain user-satisfactory answers and limiting the model's training effectiveness.
Innovation Solution
An answer feedback method and apparatus that allows users to interactively select from a candidate answer set displayed on a feedback page, enabling the selection of a new target answer and facilitating more accurate user feedback, which is then used for model training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If simple positive/negative feedback (like/dislike) is provided, then the feedback mechanism is easy to operate, but the training effectiveness is limited
Solution Approach 1:
The feedback mechanism is segmented into multiple levels: simple like/dislike buttons for quick feedback, and an expanded feedback form with multiple choice options and text input for detailed feedback. This segmentation allows users to choose the appropriate feedback depth based on their needs, maintaining ease of operation while improving training effectiveness when detailed feedback is provided.
Solution Approach 2:
The feedback interface dynamically expands from a simple like/dislike button to a comprehensive feedback form with multiple choice options and text input fields. This dynamic transformation allows the system to adapt to different feedback scenarios, providing both ease of operation for simple cases and improved training effectiveness for complex cases requiring detailed feedback.
2Loss of information
If users fill in specific reasons for dislike feedback, then more detailed feedback is obtained, but it is difficult to control the quality of the content
Solution Approach 1:
The system performs preliminary action by providing pre-defined multiple choice feedback options (e.g., 'The answer is inaccurate', 'The answer is incomplete', 'The answer is irrelevant') before users submit their feedback. This preliminary structuring of feedback options guides users to select from quality-controlled categories, ensuring consistent and useful feedback while reducing the complexity of content quality control compared to completely open text input.
3Device complexity
If only passive answer reception is implemented, then the system is simple, but the user experience is poor
Solution Approach 1:
The system implements a comprehensive feedback mechanism that goes beyond passive answer reception. Users can actively provide feedback through like/dislike buttons, select from multiple choice reasons for their feedback, and optionally add detailed text comments. This feedback loop transforms the system from a simple passive information delivery mechanism into an interactive system that adapts to user needs, significantly improving user experience while maintaining reasonable system complexity.
Data Source
AI summary
A method and an apparatus for answer feedback, which are applied to a large language model, are provided. The method includes receiving a question input by a user; generating a candidate answer set of the question by using a pre-trained large language model, and selecting an answer from the candidate answer set as a target answer, and displaying the target answer to the user; in response to receiving a feedback request for the target answer sent by the user: generating a feedback page and displaying the feedback page to the user, where content of the feedback page includes the candidate answer set; determining, in response to receiving an update request sent by the user based on the feedback page, an answer indicated by the update request from the candidate answer set as a new target answer, and displaying the new target answer to the user.