Computer system and information processing method

The system addresses the challenge of verifying LLM responses by identifying and presenting the document basis for answers, enhancing response validation efficiency in customer support operations.

JP2026028408APending Publication Date: 2026-02-20HITACHI LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
JP2024130789
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2024-08-07
Publication Date
2026-02-20

AI Technical Summary

Technical Problem

When using large-scale language models (LLMs) in customer support, it is time-consuming to locate the specific document sections that serve as the basis for generated responses, especially when documents are extensive, making it difficult for users to verify the correctness of the responses.

Method used

A computer system that includes a processor, storage, and a network interface, utilizing a large-scale language model to generate answers while identifying reference locations in external documents, tracking actions, and extracting basis information for answers, enabling the presentation of the document basis for the answers.

Benefits of technology

Enables users to verify the correctness of LLM-generated responses by presenting the document basis, thereby improving response validation efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2026028408000001_ABST
    Figure 2026028408000001_ABST
Patent Text Reader

Abstract

To present information referred to by a large-scale language model (LLM) together with an answer when generating the answer corresponding to a question by using the LLM.SOLUTION: A computer system executes answer generation processing for generating an answer to a question by using an LLM and ground extraction processing for specifying a reference place of external information referred to by the LLM for generating the answer. The answer generation processing includes processing of generating an answer instruction sentence for causing the LLM to execute a task of thinking about an action to be executed to obtain an answer and generating an answer sentence including a result of the thinking and at least one of information on the action required to be executed and the answer, and processing of executing the action and generating a history including an execution result of the action, meta information on the external information retrieved by the action, and the result of the thinking. The ground extraction processing includes processing of generating ground information for specifying a reference location of the external information referred to by the LLM based on the plurality of histories.SELECTED DRAWING: Figure 1
Need to check novelty before this filing date? Find Prior Art