Vehicle Interactive Agent Call Sequencing for Faster Accurate Responses
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for providing interactive agents in vehicle infotainment systems require inefficient multiple input/output processes of language models for complex user inputs, leading to increased execution time and cost, and rely on local optimization rather than global optimization, resulting in lower response accuracy.
Innovation Solution
A method and apparatus that utilize a pre-trained first language model to generate a call sequence, followed by a pre-trained second language model to generate an output text, where the call sequence includes a plurality of calls, optimizing information collection through a single input/output process to improve efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple input/output processes of language models are performed to process complex user input, then information collection is more thorough, but execution time increases and computing resources are wasted
Solution Approach 1:
The patent segments the language model processing into two distinct components: a first language model that generates call sequences and a second language model that generates output text. This segmentation allows each model to specialize in specific tasks, reducing the need for multiple iterative processes while maintaining response accuracy. The call sequence generation phase and output text generation phase are separated, enabling parallel execution paths that reduce overall execution time.
Solution Approach 2:
The patent performs preliminary action by using the first language model to generate complete call sequences before executing any API calls. This preliminary generation of the entire call sequence allows the system to plan all necessary information collection steps in advance, avoiding iterative back-and-forth processes. The call sequence is generated in a single pass, and then all calls are executed systematically, reducing execution time while ensuring thorough information collection.
2Measurement precision
If multiple input/output processes of language models are performed to process complex user input, then information collection is more thorough, but computing resource consumption increases
Solution Approach 1:
The patent segments the language model processing into two distinct components: a first language model that generates call sequences and a second language model that generates output text. This segmentation allows each model to specialize in specific tasks, reducing the need for multiple iterative processes while maintaining response accuracy. The call sequence generation phase and output text generation phase are separated, enabling parallel execution paths that reduce overall execution time.
Solution Approach 2:
The patent performs preliminary action by using the first language model to generate complete call sequences before executing any API calls. This preliminary generation of the entire call sequence allows the system to plan all necessary information collection steps in advance, avoiding iterative back-and-forth processes. The call sequence is generated in a single pass, and then all calls are executed systematically, reducing execution time while ensuring thorough information collection.
3Adaptability or versatility
If calls are generated sequentially for each input/output process, then the language model can adapt to each new piece of information, but the overall response accuracy decreases due to lack of global optimization
Solution Approach 1:
The patent extracts the call sequence generation function from the output text generation function by using a separate first language model. This extraction allows the call sequence to be generated independently with a focus on global optimization, while the second language model focuses on generating accurate output text. The call sequence is taken out as a distinct intermediate representation that guides all subsequent API calls, enabling global optimization without compromising adaptability.
Solution Approach 2:
The patent performs preliminary action by using the first language model to generate complete call sequences before executing any API calls. This preliminary generation of the entire call sequence allows the system to plan all necessary information collection steps in advance, avoiding iterative back-and-forth processes. The call sequence is generated in a single pass, and then all calls are executed systematically, reducing execution time while ensuring thorough information collection.
Data Source
AI summary
Provided are a method and apparatus for providing an interactive agent by using a call sequence. The method of providing an interactive agent by using a call sequence includes generating a call sequence based on an input text of a vehicle passenger, by using a first language model that is pre-trained, obtaining information of interest by executing the call sequence, and generating an output text corresponding to the input text, based on the information of interest, by using a second language model that is pre-trained, wherein the call sequence includes a plurality of calls.